142 P -Introductory-Food-Microbiology-F

Introduction 1

1INTRODUCTION

Food microbiology is the study of the microorganisms whichinhabit, create or contaminate food. Of major importance is thestudy of microorganisms causing food spoilage. However "good"bacteria such as probiotics are becoming increasingly importantin food science. In addition, microorganisms are vital for themanufacture of cheese, yoghurt, other fermented foods, bread, beerand wine.

FOOD SAFETY

Food safety is a major focus of food microbiology. Pathogenicbacteria, viruses and toxins produced by microorganisms are allpossible contaminants of food. However, microorganisms and theirproducts can also be used to combat these pathogenic microbes.Probiotic bacteria, including those which produce bacteriocinscan kill and inhibit pathogens. Alternatively, purified bacteriocinssuch as nisin can be added directly to food products. Finally,bacteriophage, viruses which only infect bacteria, can be used tokill bacterial pathogens. Thorough preparation of food, includingproper cooking will kill most bacteria and viruses. However, toxinsproduced by contaminants may not be heat-labile, and some willnot be eliminated by cooking.

FERMENTATION

Fermentation is one way microorganisms can change a food.Yeast, especially S. cerevisiae, is used to leaven bread, brew beerand make wine. Certain bacteria, including lactic acid bacteria,are used to make yogurt, cheese, hot sauce, pickles and dishes

Introductory Food Microbiology Introduction2 3

such as kimchi. A common effect of these fermentations is that thefood product is less hospitable to other microorganisms, includingpathogens and spoilage-causing microorganisms, thus extendingthe food's shelf-life.

Some cheese varieties also require mold microorganisms toripen and develop their characteristic flavors.

Foodborne Pathogens

Foodborne pathogens are the leading causes of illness anddeath in less developed countries killing approximately 1.8 millionpeople annually. In developed countries foodborne pathogens areresponsible for millions of cases of infectious gastrointestinaldiseases each year, costing billions of dollars in medical care andlost productivity. New foodborne pathogens and foodbornediseases are likely to emerge driven by factors such as pathogenevolution, changes in agricultural and food manufacturingpractices, and changes to the human host status. There are growingconcerns that terrorists could use pathogens to contaminate foodand water supplies in attempts to incapacitate thousands of peopleand disrupt economic growth.

Enteric Viruses

Food and waterborne viruses contribute to a substantialnumber of illnesses throughout the world. Among those mostcommonly known are hepatitis A virus, rotavirus, astrovirus, entericadenovirus, hepatitis E virus, and the human calicivirusesconsisting of the noroviruses and the Sapporo viruses. This diversegroup are transmitted by the fecal-oral route, often by ingestion ofcontaminated food and water.

Protozoan Parasites

Protozoan parasites associated with food and water can causeillness in humans. Although parasites are more commonly foundin developing countries, developed countries have also experiencedseveral foodborne outbreaks. Contaminants may be inadvertentlyintroduced to the foods by inadequate handling practices, eitheron the farm or during processing of foods. Protozoan parasitescan be found worldwide, either infecting wild animals or in waterand contaminating crops grown for human consumption. The

disease can be much more severe and prolonged inimmunocompromissed individuals.

Mycotoxins

Molds produce mycotoxins, which are secondary metabolitesthat can cause acute or chronic diseases in humans when ingestedfrom contaminated foods. Potential diseases include cancers andtumors in different organs (heart, liver, kidney, nerves),gastrointestinal disturbances, alteration of the immune system,and reproductive problems. Species of Aspergillus, Fusarium,Penicillium, and Claviceps grow in agricultural commodities orfoods and produce the mycotoxins such as aflatoxins,deoxynivalenol, ochratoxin A, fumonisins, ergot alkaloids, T-2toxin, and zearalenone and other minor mycotoxins such ascyclopiazonic acid and patulin. Mycotoxins occur mainly in cerealgrains (barley, maize, rye, wheat), coffee, dairy products, fruits,nuts and spices. Control of mycotoxins in foods has focused onminimizing mycotoxin production in the field, during storage ordestruction once produced. Monitoring foods for mycotoxins isimportant to manage strategies such as regulations and guidelines,which are used by 77 countries, and for developing exposureassessments essential for accurate risk characterization.

Yersinia Enterocolitica

Yersinia enterocolitica includes pathogens and environmentalstrains that are ubiquitous in terrestrial and fresh water ecosystems.Evidence from large outbreaks of yersiniosis and fromepidemiological studies of sporadic cases has shown that Y.enterocolitica is a foodborne pathogen. Pork is often implicated asthe source of infection. The pig is the only animal consumed byman that regularly harbours pathogenic Y. enterocolitica. Animportant property of the bacterium is its ability to multiply attemperatures near to 0°C, and therefore in many chilled foods. Thepathogenic serovars (mainly O:3, O:5,27, O:8 and O:9) showdifferent geographical distribution. However, the appearance ofstrains of serovars O:3 and O:9 in Europe, Japan in the 1970s, andin North America by the end of the 1980s, is an example of aglobal pandemic. There is a possible risk of reactive arthritisfollowing infection with Y. enterocolitica.


Vibrio

Vibrio species are prevalent in estuarine and marineenvironments and seven species can cause foodborne infectionsassociated with seafood. Vibrio cholerae O1 and O139 serovtypesproduce cholera toxin and are agents of cholera. However, fecal-oral route infections in the terrestrial environment are responsiblefor epidemic cholera. V. cholerae non-O1/O139 strains may causegastroenteritis through production of known toxins or unknownmechanism.

Vibrio parahaemolytitucs strains capable of producingthermostable direct hemolysin (TDH) and/or TDH-relatedhemolysin are most important cause of gastroenteritis associatedwith seafood consumption. Vibrio vulnificus is responsible forseafoodborne primary septicemia and its infectivity dependsprimarily on the risk factors of the host. V. vulnificus infection hasthe highest case fatality rate (50%) of any foodborne pathogen.Four other species (Vibrio mimicus, Vibrio hollisae, Vibrio fluvialis,and Vibrio furnissii) can cause gastroenteritis. Some strains ofthese species produce known toxins but the pathogenic mechanismis largely not understood. The ecology of and detection and controlmethods for all seafoodborne Vibrio pathogens are essentiallysimilar.

Staphylococcus Aureus

Staphylococcus aureus is a common cause of bacterialfoodborne disease worldwide. Symptoms include vomiting anddiarrhea that occur shortly after ingestion of S. aureus-contaminated food. The symptoms arise from ingestion of preformedenterotoxin, which accounts for the short incubation time.Staphylococcal enterotoxins are superantigens and, as such, haveadverse effects on the immune system.

The enterotoxin genes are accessory genetic elements in S.aureus, meaning that not all strains of this organism are enterotoxin-producing. The enterotoxin genes are found on prophage, plasmids,and pathogenicity islands in different strains of S. aureus.Expression of the enterotoxin genes is often under the control ofglobal virulence gene regulatory systems.

Campylobacter

Campylobacter spp., primarily C. jejuni subsp. jejuni is one ofthe major causes of bacterial gastroenteritis in the U.S. andworldwide. Campylobacter infection is primarily a foodborneillness, usually without complications; however, serious sequelaesuch as Guillain-Barre Syndrome occur in a small subset of infectedpatients. Detection of C. jejuni in clinical samples is readilyaccomplished by culture and non-culture methods.

Listeria Monocytogenes

Listeria monocytogenes is Gram-positive foodborne bacterialpathogen and the causative agent of human listeriosis. Listeriaeare acquired primarily through the consumption of contaminatedfoods including soft cheese, raw milk, deli salads, and ready-to-eat foods such as luncheon meats and frankfurters. Although L.monocytogenes infection is usually limited to individuals that areimmunocompromised, the high mortality rate associated withhuman listeriosis makes L. monocytogenes the leading cause ofdeath amongst foodborne bacterial pathogens. As a result,tremendous effort has been made at developing methods for theisolation, detection and control of L. monocytogenes in foods.

Salmonella

Salmonella serotypes continue to be a prominent threat tofood safety worldwide. Infections are commonly acquired by animalto human transmission though consumption of undercooked foodproducts derived from livestock or domestic fowl. The second halfof the 20th century saw the emergence of Salmonella serotypesthat became associated with new food sources (i.e. chicken eggs)and the emergence of Salmonella serotypes with resistance againstmultiple antibiotics.

Shigella

Shigella species are members of the family Enterobacteriacaeand are Gram negative, non-motile rods. Four subgroups existbased on O-antigen structure and biochemical properties;S. dysenteriae (subgroup A), S. flexneri (subgroup B), S. boydii(subgroup C) and S. sonnei (subgroup D). Symptoms include mildto severe diarrhea with or without blood, fever, tenesmus, and


abdominal pain. Further complications of the disease may beseizures, toxic megacolon, reactive arthritis and hemolytic uremicsyndrome. Transmission of the pathogen is by the fecal-oral route,commonly through food and water. The infectious dose rangesfrom 10-100 organisms. Shigella spp. have a sophisticatedpathogenic mechanism to invade colonic epithelial cells of thehost, man and higher primates, and the ability to multiplyintracellularly and spread from cell to adjacent cell via actinpolymerization. Shigellae are one of the leading causes of bacterialfoodborne illnesses and can spread quickly within a population.

Escherichia Coli

More information is available concerning Escherichia coli thanany other organism, thus making E. coli the most thoroughlystudied species in the microbial world. For many years, E. coli wasconsidered a commensal of human and animal intestinal tractswith low virulence potential. It is now known that many strainsof E. coli act as pathogens inducing serious gastrointestinal diseasesand even death in humans. There are six major categories of E. colistrains that cause enteric diseases in humans including the:

(1) enterohemorrhagic E. coli, which cause hemorrhagic colitisand hemolytic uremic syndrome,

(2) enterotoxigenic E. coli, which induce traveler's diarrhea,

(3) enteropathogenic E. coli, which cause a persistent diarrheain children living in developing countries,

(4) enteroaggregative E. coli, which provoke diarrhea inchildren,

(5) enteroinvasive E. coli that are biochemically and geneticallyrelated to Shigella species and can induce diarrhea, and

(6) diffusely adherent E. coli, which cause diarrhea and aredistinguished by a characteristic type of adherence tomammalian cells.

Clostridium Botulinum and Clostridium Perfringens

Clostridium botulinum produces extremely potent neurotoxinsthat result in the severe neuroparalytic disease, botulism. Theenterotoxin produced by C. perfringens during sporulation of

vegetative cells in the host intestine results in debilitating acutediarrhea and abdominal pain. Sales of refrigerated, processedfoods of extended durability including sous-vide foods, chilledready-to-eat meals, and cook-chill foods have increased over recentyears. Anaerobic spore-formers have been identified as the primarymicrobiological concerns in these foods. Heightened awarenessover intentional food source tampering with botulinum neurotoxinhas arisen with respect to genes encoding the toxins that arecapable of transfer to nontoxigenic clostridia.

Bacillus Cereus

The Bacillus cereus group comprises six members: B. anthracis,B. cereus, B. mycoides, B. pseudomycoides, B. thuringiensis and B.weihenstephanensis. These species are closely related and shouldbe placed within one species, except for B. anthracis that possessesspecific large virulence plasmids. B. cereus is a normal soilinhabitant and is frequently isolated from a variety of foods,including vegetables, dairy products and meat. It causes a vomitingor diarrhoea illness that is becoming increasingly important in theindustrialized world. Some patients may experience both types ofillness simultaneously. The diarrhoeal type of illness is mostprevalent in the western hemisphere, whereas the emetic type ismost prevalent in Japan. Desserts, meat dishes, and dairy productsare the foods most frequently associated with diarrhoeal illness,whereas rice and pasta are the most common vehicles of emeticillness.

The emetic toxin (cereulide) has been isolated andcharacterized; it is a small ring peptide synthesised non-ribosomally by a peptide synthetase. Three types of B. cereusenterotoxins involved in foodborne outbreaks have been identified.Two of these enterotoxins are three-component proteins and arerelated, while the last is a one-component protein (CytK). Deathshave been recorded both by strains that produce the emetic toxinand by a strain producing only CytK. Some strains of the B. cereusgroup are able to grow at refrigeration temperatures. These variantsraise concern about the safety of cooked, refrigerated foods withan extended shelf life. B. cereus spores adhere to many surfacesand survive normal washing and disinfection (except forhypochlorite and UVC) procedures. B. cereus foodborne illness is


likely underreported because of its relatively mild symptoms, whichare of short duration.

CARNOBACTERIUM: POSITIVE AND NEGATIVE EFFECTSIN THE ENVIRONMENT AND IN FOODS

The genus Carnobacterium contains nine species, but only C.divergens and C. maltaromaticum are frequently isolated fromnatural environments and foods. They are tolerant to freezing/thawing and high pressure and able to grow at low temperatures,anaerobically and with increased CO2 concentrations. Theymetabolize arginine and various carbohydrates, including chitin,and this may improve their survival in the environment.Carnobacterium divergens and C. maltaromaticum have beenextensively studied as protective cultures in order to inhibit growthof Listeria monocytogenes in fish and meat products. Severalcarnobacterial bacteriocins are known, and parameters that affecttheir production have been described. Currently, however, noisolates are commercially applied as protective cultures.Carnobacteria can spoil chilled foods, but spoilage activity showsintraspecies and interspecies variation. The responsible spoilagemetabolites are not well characterized, but branched alcohols andaldehydes play a partial role.

Their production of tyramine in foods is critical for susceptibleindividuals, but carnobacteria are not otherwise human pathogens.Carnobacterium maltaromaticum can be a fish pathogen, althoughcarnobacteria are also suggested as probiotic cultures for use inaquaculture. Representative genome sequences are not yet available,but would be valuable to answer questions associated withfundamental and applied aspects of this important genus.

Carnobacteria are ubiquitous lactic acid bacteria (LAB) isolatedfrom cold and temperate environments. More importantly from apractical viewpoint, they also frequently predominate in a rangeof foods, including fish, meat, and some dairy products. In thisregard, they have been extensively studied in the last two decadesas protective cultures, to inhibit pathogenic and spoilagemicroorganisms, and as potential spoilage bacteria in chilledseafood and meat products. In addition, owing to their presencein the aqueous environment, their importance as fish pathogens

or probiotic cultures in the aquaculture industry has beenexamined.

The genus Carnobacterium currently consists of nine species,but only two of these, Carnobacterium divergens and C.maltaromaticum (formerly C. piscicola) are frequently encounteredin the environment and in foods (Table 1). This review concernsthe importance of these two species in dairy, meat and fish productsand in the aquatic environment. The taxonomy and methods forisolation and identification of carnobacteria are not described. Itis, however, important to note that older studies frequently usedcarbohydrate fermentation patterns to distinguish between C.divergens and C. maltaromaticum. These methods are nowconsidered less reliable than molecular methods, and results relyingon phenotypic criteria must be interpreted with caution.

DISTRIBUTION IN THE NATURAL ENVIRONMENT ANDFOODS

Natural Environment

Among the nine reported species of Carnobacterium, only twospecies, C. divergens and C. maltaromaticum, are frequently isolatedfrom various sources. Four species, C. alterfunditum, C. funditum,C. gallinarum and C. mobile, have been isolated a few times fromat most three sources. Three species, C. inhibens, C. pleistoceniumand C. viridans, have only been isolated from one source, but itshould be noted that Carnobacterium spp. related to, for example,C. inhibens have been described. Carnobacterium spp. appear tohave both the temperate and polar aquatic environments as habitatsincluding live fish, marine sponges, Antarctic lakes, Arctic andAntarctic sea water as well as the deep sea, aquous alkaline tufacolumns from Greenland, and freshwater habitats from thetemperate clima zone, including a Sphagnum pond and rivers inthe northwest region of Spain.

Although C. maltaromaticum and/or C. divergens have beenisolated from tropical fish products, including smoked surubim, aBrazilian tropical freshwater fish, and from vacuum-packed tunacaught in the Indian Ocean and processed in Sri Lanka, it cannot,at least in the latter case, be excluded that this is due tocontamination associated with repackaging in Denmark.


The presence of carnobacteria has also been demonstrated inthe terrestrial environment, including a field treated with whey,Canadian winter soil, permafrost ice, a compost pile, a collapsedhorse manure pile, the larval midgut of a moth species, and othersources. It appears that the temperate/polar aquatic and terrestrialenvironments are both natural habitats.

Carnobacterium divergens and C. maltaromaticum possesstraits that may play a role in their survival in these surroundings.One study has reported that a Carnobacterium sp. soil isolaterelated to C. maltaromaticum survived 48 serial freeze-thaw cyclesbetter than Escherichia coli and an Enterococcus sp. soil isolate,but worse than a few other soil isolates, including an Acinetobactersp. Indeed, these organisms may survive freezing for considerableperiods of time, as witnessed by the isolation of a Carnobacteriumsp. preserved in a permafrost ice wedge for 25 000 years. Additionaldata on the abilities of carnobacteria to grow at low temperaturesand to survive frozen storage is available for food products. Also,a cold-active b-galactosidase from C. maltaromaticum and a cold-adapted alanine dehydrogenase from a Carnobacterium sp. relatedto C. alterfunditum have been reported. Some carnobacterial isolatesoriginate from natural high-pressure habitats, and additional dataon resistance to pressure have been reported for high-pressure-processed foods. It has not been described how other physical/chemical parameters such as salt content, atmosphere and pHaffect survival and growth of these organisms in the naturalenvironment.

With respect to energy-yielding substrates, one species, C.maltaromaticum, expresses chitinase activity. This may facilitateadherence to and survival on zooplankton as suggested forenterococci, and also explain the isolation of this species from themidgut of the larval stage of a species of moth. Interestingly, thearginine deiminase pathway is expressed by the most frequentlyencountered species, C. divergens, C. gallinarum, C.maltaromaticum and C. mobile but not by C. inhibens or C. viridans.Arginine is not a substrate that results in growth of C. alterfunditumand C. funditum, and information is not available for C.pleistocenium. This amino acid may represent an additional energysource for growth and survival when carbohydrates are scarce,

and it may offer protection against acid stress, as described forEnterococcus faecalis and related species. Finally, carnobacteriaare able to catabolize a range of carbohydrates, although there areconsiderable interspecies and intraspecies heterogeneities.

Examples of sources of such carbohydrates include animals(e.g. chitin and, to some extent, lactose, although the targets of thelactose hydrolytic activity shown by carnobacteria may instead bebyproducts of plant sugar polymers and saccharides adsorbed tohumic acid substances in the soil), plants (e.g. salicin and sucrose),fungi (e.g. chitin and trehalose), prokaryotes (N-acetylglucosamine),and living organisms in general (ribose). The importance of theseabilities for growth and survival in the environment deservesfurther study. The genome sizes of C. alterfunditum, C. divergensand C. pleistocenium strains and a Carnobacterium sp. AT7 deepsea isolate have been estimated to be 2.9, 3.2, 3.2 and 2.4Mb,respectively, and for the AT7 isolate, the data are reported in adatabase described by Liolios (2006). These sizes are relativelylarge in comparison to many other LAB, suggesting thatcarnobacterial genomes may encode a range of genes that makesthem well adapted to deal with environmental challenges.

Even if carnobacteria are well adapted to temperate or polaraqueous environments, this does not always provide improvedability to survive as compared to other bacteria. Thus, the actualabilities of strains of C. divergens and C. maltaromaticum isolatedfrom a Sphagnum pond to survive in water from this source werenot improved as compared to related gram-positive bacteriaoriginating from other sources. At present, therefore, the precisemechanisms by which Carnobacterium spp. persist in the naturalenvironment and their underlying genetics are not known. Finally,it should be noted that the environments from which carnobacteriacan be isolated are not always as extreme as they appear at firstsight. Thus, Carnobacterium spp. isolated from Lake Vanda,Antarctica were found at a depth of 61 m, where the temperaturewas c. 15-20 1C.

Dairy, Fish and Meat Products

High concentrations of bacteria (4106-107 CFUg_1) in foodare typically required before their activity is sufficient to influence


the sensory properties of a product. For this to happen, occurrenceand growth kinetics are the key parameters, and databases,including ComBase, are available to determine the effect of foodstorage conditions and product characteristics on growth of specificbacteria. However, information on carnobacteria has not yet beenincluded in such databases.

Carnobacterium maltaromaticum and another Carnobacteriumsp. have been isolated from milk, and they, in addition to C.divergens, have been detected in soft cheeses, including mold-ripened brie, mozzarella, Camembert and other types. The highconcentration of C. divergens in the curd of mozzarella, exposedto 10-37 1C during processing, is in agreement with the maximumgrowth temperature (40 1C) of this species.

The occurrence of Carnobacterium in dairy products (andother foods) is most likely underreported. This is due to the commonuse of acetate containing media, particularly MRS agar or Rogosaagar, for enumeration of LAB. Growth of Carnobacterium isinhibited by acetate, and both MRS and Rogosa agar mediasignificantly underestimate concentrations in food.

Carnobacterium divergens and C. maltaromaticum are presentin seafood and are able to grow to high concentrations in differentfresh and lightly preserved products.

Studies of naturally contaminated products suggest whichstorage conditions and product characteristics select forcarnobacteria as compared to the other bacteria present in seafood.For chilled fresh seafood, we have found no reports where C.divergens and C. maltaromaticum dominated the microbialcommunities in aerobically stored products, but this has beenreported for modified atmosphere-packed (MAP) coalfish, cod,pollack, rainbow trout, salmon, shrimp, swordfish and tuna.

After frozen storage (_20 to _30 1C for 5-8 weeks), C. divergensand C. maltaromaticum seem to be particularly prominent inchilled MAP fish, as has been reported for cod, garfish, salmon,and tuna. In addition to freezing, these carnobacteria are relativelyresistant to high-pressure processing and are found in highconcentrations in vacuum-packed and chilled squid mantle andcold-smoked salmon previously treated with 200-400MPa for 15-

20 min. In vacuum-packed coldsmoked or sugar-salted ('gravad')seafood with 3-7% NaCl in the water phase and a pH of 5.8-6.5,high concentrations of C. divergens and C. maltaromaticum areparticularly common, as reported for halibut, rainbow trout,salmon, surubim, and tuna.

High concentrations of C. maltaromaticum are also reportedin salted lumpfish roe, and it has been isolated from frozen, smokedmussels. Finally, for cooked seafood, high concentrations of C.maltaromaticum have been detected in MAP shrimp after storageat 2-8 1C. Both C. divergens and C. mobile were isolated fromcooked and brined MAP shrimps (with NaCl, benzoic acid, citricacid and sorbic acid) after storage at 2-8 1C. Clearly, carnobacteriaare common in chilled fresh and lightly preserved seafood, but athigher storage temperatures (15-25 1C), other species, includingEnterococcus spp., more frequently dominate the spoilage microbialcommunity of seafood.

Carnobacterium divergens and C. maltaromaticum are able togrow in meat products at temperatures as low as 2 to _1.5 1C, andthey are frequently predominant members (up to 50% C. divergensand up to 26% C. maltaromaticum of the gram-positive or LABisolates obtained) of the microbial community of raw meat (beef,pork, lamb, and poultry).

The two species are found irrespective of whether productshave been stored aerobically, vacuum packaged, or subjected tomodified atmospheres, including gas compositions of CO2/N2 (%)ranging from 10 : 90 to 80 : 20. One study demonstrated thepresence of C. divergens in beef stored in air but not in MAP beefwith increased concentrations of O2 (20-40%) in addition to CO2(40%). The growth of C. funditum is impaired by oxygen, and itwill be of interest to study further whether addition of O2 to thegas composition of modified atmospheres consistently inhibitsgrowth of carnobacteria, as has been observed for some otherbacteria.

Further studies are needed in order to determine whetherdifferences in the presence of carnobacteria in meat are due tovariations in storage conditions or variations in contaminationlevels at the processing plants. The source of carnobacteria inmeat products is most probably the processing plant, as these


organisms have not been isolated from the gastrointestinal systemor skin of chicken, cattle, pigs or sheep, except for one unpublishedstudy. Thus, the source of contamination of broiler carcasses by C.divergens and C. maltaromaticum was shown to be the air in theprocessing plant and not incoming broiler chickens. In fact, twostudies confirmed that C. divergens and C. maltaromaticum presentin the raw material were eliminated or reduced in number bycooking of a processed meat product, ham.

Carnobacterium divergens and C. maltaromaticum have,however, been detected in a variety of processed meat products,including the cured pork product bacon, a Danish processed porkproduct ('rullep_lse'), various Spanish processed meat products,cooked poultry meat, pressure-treated (408-888MPa) chicken, andirradiated pork and chicken. On a few occasions, C.maltaromaticum has even been isolated from fermented sausages;as noted by Hammes & Hertel (2003), this contrasts with the usualnonaciduric habitats of carnobacteria. Other Carnobacterium spp.isolated from processed meat products include C. gallinarium(irradiated chicken), C. mobile (cooked turkey) and C. viridans(vacuum-packed Bologna sausage). Although these organisms arefrequently isolated from processed meat products, it has beensuggested that they are rarely present in high numbers, and thatterminal spoilage of cooked, cured meats is primarily caused byaciduric LAB.

Finally, a Carnobacterium sp. has also been isolated from eggcontents, and it was shown that this isolate had a similar abilityto penetrate the eggshell and contaminate the whole egg as variousother gram-positive and gram-negative bacteria.

Transmission routes for carnobacteria from the naturalenvironment into food-manufacturing plants and further on todairy, fish and meat products are not known to any extent.Knowledge on this topic is essential to evaluate if species andintraspecific clusters that differ in spoilage ability (as discussedlater) also have different colonization abilities. Such knowledgewould also illuminate the feasibility of using the production ofmetabolites, especially tyramine, by C. divergens and C.maltaromaticum as an index of microbial spoilage of specific fishand meat products.

FUNCTIONAL PROPERTIES OF CARNOBACTERIA

Bacteriocins and Antimicrobial Properties

Carnobacteria with the ability to produce antimicrobialpeptides, bacteriocins, are commonly encountered in foods. Thecharacterized carnobacterial bacteriocins belong to class I andclass II [e.g. Drider et al. (2006) for bacteriocin nomenclature]. Sofar, only one class I bacteriocin (lantibiotic), UI149, has beenreported to be produced by Carnobacterium. In contrast, ten aminoacidsequenced bacteriocins belong to class II, and most of thesemore specifically to class IIa, which comprises small pediocin- likepeptides. The inhibition spectrum of carnobacterial class IIabacteriocins includes Listeria, and antimicrobial activity is exertedby pore formation, dissipation of membrane potential, and leakageof internal low molecular weight substances.

Class IIa bacteriocins are ribosomally synthesized as inactiveprepeptides that are modified by posttranslational cleavage of theN-terminal peptide leader at a doubleglycine site in order to releasemature and active cationic peptides. Carnobacterial class IIabacteriocins have similar amino acid sequences, with aYGNGV(X)C(X)4C motif (X denotes any amino acid) near the N-terminus of the mature peptide. Two cysteine residues form adisulfide bond in the N-terminal region, and there is anamphipathic a-helix near the C-terminus. The N-terminal regionis relatively hydrophilic and conserved, whereas the C-terminusis hydrophobic and diverse. To establish the structure-activityrelationships of carnobacterial bacteriocins, the structures ofcarnobacteriocin B2 and divercin V41 were modified. Thus, aminoacid substitutions closer to the N-terminus in carnobacteriocin B2drastically reduced or eliminated antimicrobial activity, whereasthis was not so for substitutions close to the C-terminal part.Divercin V41 contains a second disulfide bond located in the C-terminal region.

When the C-terminal region of divercin V41 was separatedfrom the N-terminus by endoproteinase Asp-N, only the C-terminalfragment was active. After trypsin cleavage next to lysine at position42 or disulfide reduction, the C-terminus lost its inhibitory activity.These results suggested that both hydrophobicity and foldingimposed by this second disulfide bond were essential for


antilisterial activity of the C-terminal hydrophobic peptide.Chemical oxidation of tryptophan residues by N-bromosuccinimideshowed that these residues were crucial for inhibitory activity, asmodification of any one of them rendered divercin V41 inactive.The three-dimensional structure of carnobacteriocin B2, itsprecursor and the corresponding immunity protein have beensolved by nuclear magnetic resonance. These are, at present, theonly reported carnobacterial protein structures.

Carnobacterial class II bacteriocins that contain adoubleglycine- type leader peptide are transported by a dedicatedATP-binding cassette (ABC) transport system. In contrast, divergicinA does not require a secretion protein, as the transport depend onthe general cellular secretion (sec) pathway. The genes involved incarnobacterial bacteriocin production are generally clustered inoperons. In the simplest case, corresponding to the sec-dependentdivergicin A, the bacteriocin operon is composed of only thestructural gene followed downstream by the gene encoding animmunity protein that protects the cell from is own bacteriocin.Production of class IIa bacteriocins requires four genes as aminimum, including genes encoding bacteriocin, immunity protein,ABC of transport protein and its membrane-bound accessoryprotein. Genes for bacteriocin production may be encoded on thechromosome or on plasmids. Carnobacterium maltaromaticumLV17 produces at least three class II bacteriocins.

Carnobacteriocins A and B2 are encoded on different butcompatible plasmids, pCP49 (72 kb) and pCP40 (61 kb), respectively.The carnobacteriocin BM1 structural gene and its immunity geneare localized on the chromosome. Activation and export ofcarnobacteriocin BM1 depend on genes located on plasmid pCP40encoding carnobacteriocin B2. For carnobacteriocins BM1 and B2,a peptide-pheromone dependent quorum-sensing mode caused bythe bacteriocin itself or an autoinducer peptide is involved in theregulation of bacteriocin production by two- and three-componentsignal transduction systems.

The components of this regulatory system consist of aninduction factor, a histidine protein kinase, and a responseregulator. Four carnobacterial bacteriocin operons including genesfor immunity proteins and regulatory proteins involved in

bacteriocin expression and secretion have been sequenced. Externalparameters also affect bacteriocin production. Thus, increasingconcentrations of NaCl (2-7%) reduced production of A9b/B2bacteriocin, increasing the temperature from below 19 to 25 1Cinhibited production of piscicolin 126, and reducing the pH from6.5 to 5.5 inhibited production of carnobacteriocins BM1 and B2.Acetate (A9b bacteriocin) or the presence of a bacteriocin-sensitivestrain belonging to the same genus induced bacteriocin production.It is clear that, in order to apply carnobacterial bacteriocins for thebiopreservation of foods, detailed knowledge of the factors thatinfluence production of the bacteriocin is necessary.

Finally, it should be noted that the bacteriocinogenic activitymay be lost, as shown for a C. divergens divercin V41- producingstrain that retained only partial activity after being subjected tospray-drying. Interest in applying carnobacterial bacteriocins infoods and feeds has been directed towards inhibiting Listeriamonocytogenes and spoilage microorganisms, to extend the shelf-life of lightly preserved seafood. These attempts have generally,but not always [e.g. regarding prevention of spoilage], met withsuccess, although diversity in sensitivity of the target organismhas been observed.

However, the occurrence of resistant L. monocytogenes targetorganisms has led to the suggestion that bacteriocin-negative LABmay be more suitable for practical use as bioprotective agentsagainst L. monocytogenes in ready-to-eat foods. Indeed, L.monocytogenes is inhibited by carnobacterial cultures that do notproduce bacteriocins, and this is partly due to glucose depletion.One study also reported a successful application of cultures withno demonstrated bacteriocinogenic activity for extension of theshelflife of vacuum-packed cold-smoked salmon. This effect wasnot, however, observed for a C. divergens strain that was unableto produce the class IIa divercin V41.

Resistance of L. monocytogenes to the class IIa divergicin M35was probably due to modification of the cell wall fatty acidcomposition. Listeria monocytogenes strains resistant to the classIIa divercin V41 also showed substantial differences in proteinexpressions as compared with the wild type strain. A s54-dependent PTS permease of the mannose family, which belongs to


the phosphotransferase system (PTS), is responsible for sensitivityof L. monocytogenes to class IIa bacteriocins. These resultssuggested that EIIt Man encoded by the mptACD operon might bea target molecule for class IIa bacteriocins. Recently, two genescoding for a glycerophosphoryl diester phosphodiesterase (GlpQ)and a protein with a putative phosphoesterase function (Pde)were identified as also being involved in the class IIa sensitivityof En. faecalis. Finally, it is important to note that the susceptibilityof the target strain is affected by various environmental conditionssuch as NaCl and pH.

Another cause of concern when using carnobacteria forbiopreservation is that C. divergens and C. maltaromaticum bothproduce tyramine, as discussed further below. Therefore, it hasbeen suggested that a C. divergens strain in which the tyrosinedecarboxylase gene is inactivated by mutagenesis could be usedas a protective culture to prevent growth of L. monocytogenes incold-smoked salmon. Currently, no carnobacterial culture,bacteriocin producing or not, is commercially applied for protectionagainst the growth of L. monocytogenes or other bacterial pathogensor spoilage organisms in food.

Effect of Catabolic Activities on Sensory Characteristics and Safetyof Foods

As shown in Table 4, the catabolic activities of carnobacteriamay result in sensory spoilage of inoculated fish and meat products.Whether this is so for cheese products is not clear. In naturallycontaminated products, it seems other members of the bacterialcommunity are typically more important with regard to sensoryeffects, including spoilage.

It is of interest that spoilage was enhanced if moderate-spoilagestrains of C. maltaromaticum were inoculated with nonspoilageVibrio sp. strains or the moderate-spoilage organism Brochothrixthermosphacta into cold-smoked salmon that was subsequentlyvacuum-packaged. However, this was not so for combinations ofC. maltaromaticum and Photobacterium phosphoreum in thisproduct. In addition, a spoilage synergy effect was observed forcombinations of B. thermosphacta and C. divergens, C.maltaromaticum or C. mobile in MAP shrimp.

We will first focus on catabolic reactions with carbohydratesand/or organic acids as substrates and with pyruvic acid as anintermediate metabolite. Respiration might occur in the presenceof hematin, as shown for C. maltaromaticum, and, indeed, thisspecies consumes substantial proportions of oxygen duringexponential growth under aerobic conditions. Carnobacterium spp.are, however, considered to be homofermentative organisms thatproduce lactic acid from glucose.

The presence of the glycolytic pathway in C. divergens hasbeen demonstrated. It can be debated whether carnobacteria shouldinstead be considered facultative or atypical heterofermentativeorganisms, as C. divergens and C. maltaromaticum are able toutilize ribose and gluconic acid as substrates for growth, and mayproduce acetic acid, formic acid, and CO2 as end-products of somesecondary decarboxylation/ dissimilation reactions of pyruvic acid.Indeed, carnobacteria were initially described as heterofermenters.

Acetic acid production by C. divergens and C. maltaromaticumcan be substantial during growth in laboratory media under aerobicconditions or in MAP shrimp, and quantitatively it can exceedlactic acid production. Production of acetic acid by C.maltaromaticum is also increased relative to lactic acid if glucoseis substituted by ribose. Furthermore, C. maltaromaticum canproduce large amounts of ethanol from glucose and ribose duringgrowth in a shrimp extract under anaerobic conditions.

Acetoin can be generated by C. maltaromaticum from pyruvicacid, and this reaction is also found for C. divergens and C.gallinarum but not for C. alterfunditum, C. funditum, C. inhibensand C. viridance, according to reactions in the Voges-Proskauertest. Carnobacterium mobile gives variable reactions, andinformation is not available for C. pleistocenium.

The factors affecting acetoin production are not well known,but production is increased by resting cells of C. maltaromaticumin the presence of hematin. In addition, two out of four strains ofC. divergens and C. maltaromaticum isolated from mozzarellacheese were reported to be able to metabolize citric acid, but thepotential sensory role of this reaction, e.g. by production of acetoinand diacetyl, was not examined.


Production of diacetyl and 2,3-pentanedione by C.maltaromaticum during growth in cold-smoked salmon resultedin a butter-like odor but not in spoilage. In conclusion, even, ifcarbohydrate catabolism by carnobacteria appears to result in adiverse number of metabolites, these have generally a limited effecton the sensory attributes of foods. H2O2 may be produced by C.divergens, and formation of this compound by C. viridans hasbeen associated with spoilage in the form of green discolorationof cured Bologna ham.

Whether carnobacteria possess proteolytic activities of potentialimportance for taste in products such as cheese is not known, andthis deserves further investigation. In contrast, metabolites resultingfrom the degradation of amino acids certainly cause sensory effectsin foods. Generation of branched alcohols and aldehydes (2-methyl-1-butanal, 2-methyl-1-butanol, 3-methyl-1-butanal, 3-methyl-1-butanol, 2-methylpropanal, and 2-methylpropanol) bytransamination, decarboxylation and reduction of the amino acidsvaline, leucine and isoleucine appears to be particularly important.Production of some of these compounds, such as 3-methyl-1-butanal, is strain or species dependent. Production by a C.maltaromaticum strain of 3-methylbutanal, 3-methylbutanol and3-methylbutanoic acid from leucine was, in general, increased atpH values of 6.5 or higher and in the presence of increasedconcentrations of a-ketoisocaproic acid and glucose. This strainaffected the odor and reduced the leucine content of a sausagemince, but a clear causal relationship was not demonstrated.Production by C. maltaromaticum of alcohols and aldehydes fromvaline, leucine and/or isoleucine resulted in a malty, green aromain skimmed milk and shrimp, and has also caused spoilage ofcured ham. The association between the presence of thesecompounds and spoilage is, however, not always straightforward,as production of 3-methyl-1- butanal and/or 3-methyl-1-butanolin shrimp by certain C. maltaromaticum strains did not result inmalty off-flavors.

Production of NH4 from arginine as a result of its catabolismcatalyzed by the arginine deaminase pathway may, in theory,cause spoilage of shrimp or squid that contain high levels of freearginine. Carnobacterium maltaromaticum and C. divergens but

not C. mobile were able to produce NH4 in MAP shrimp, but thiswas not the cause of spoilage of this product. Production of indolefrom the amino acid tryptophan is also a potential cause of spoilage,although C. divergens, C. maltaromaticum and C. mobile are allunable to carry out this reaction in standard laboratory media. Ithas, however, been observed that C. divergens and strainsbelonging to a major phenotypic cluster of C. maltaromaticumwere able to use tryptophan as a substrate during growth in MAPshrimp. Whether this resulted in generation of indole was notexamined.

Production of tyramine from tyrosine is the only knownmetabolic reaction by Carnobacterium spp. that constitutes a causeof concern regarding food safety. Not all carnobacterial speciespossess this ability, as C. mobile does not produce tyramine duringgrowth in shrimp, and variation exists among different strainsand phenotypic clusters of C. divergens and C. maltaromaticumfor amounts of tyramine produced. Tyramine production by oneC. divergens strain was at its maximum during the stationarygrowth phase and at a low initial pH in the presence of 0.6%glucose, whereas NaCl (10%) was inhibitory. Regardless of strainvariation and the effects of environmental parameters, tyramineproduction by C. divergens and/or C. maltaromaticum has beenreported in a range of foods, including meat (up to 28 mg kg_1),a meat-fat mixture (up to 121 mg kg_1), cold-smoked salmon (upto c. 370 mg kg_1), frozen and thawed salmon (up to 40-60 mgkg_1), and shrimp (up to 20-60 mg kg_1).

These levels have no adverse effects on most consumers, butfor sensitive individuals, e.g. with reduced monoamine oxidaseactivity due to medication or hereditary deficiency, very littletyramine can cause migraine headaches, and an intake of no morethan 5mg of tyramine per meal has been recommended. Typicalconsumption of fish and meat products is 50-150 g per meal. Asdescribed above, C. divergens and/or C. maltaromaticum are ableto form c. 20-370 mg tyramine kg_1 in these products,corresponding to c. 1-55 mg of tyramine per meal. Consequently,tyramine formation by carnobacteria in specific foods can representa hazard for sensitive individuals who might suffer migraineheadaches. This risk for sensitive individuals can be minimized


by restricting their consumption of fish and meat to products thatare newly processed, i.e. that are not close to the declared shelf-life expiry day.

Growth of Carnobacterium spp. in food products may resultin accumulation of a range of volatile alcohols, ketones,hydrocarbons, and other compounds. The metabolic reactionsleading to these products have not been studied in detail forCarnobacterium, in contrast to the situation for other LAB, butmay be a result of NAD1-generating reactions or nonenzymaticreactions. Several ketones have characteristic odors, but whetherthey contribute to spoilage off-flavors as a result of carnobacterialmetabolic activity is not well understood. With regard to lipids, C.maltaromaticum has been reported to hydrolyze tributyrin, atriglyceride with butyrate as fatty acid, but it is not known whetherthis lipase activity contributes to flavor changes of dairy, fish andmeat products. Carnobacterium divergens and C.maltaromaticumwere not able to limit oxidation of linoleic acid during growth.This indicates that these species may not prevent qualitydeterioration of meat products by lipid oxidation. Finally, C.maltaromaticum has been shown to cause a softer texture of salmonfillets when inoculated in high numbers.

It is worth noting that interspecies and intraspecies differencesexist regarding carnobacterial spoilage capacity. Laursen et al.(2006) showed that C. mobile and a minor phenotypic cluster ofC. maltaromaticum did not spoil MAP shrimp, in contrast to whatwas seen with C. divergens and the major phenotypic cluster ofC. maltaromaticum. It would be of interest to examine whethersimilar heterogeneities exist for Carnobacterium spp. growing inother dairy, fish and meat products. As environmental parametersreadily affect the metabolism of carbohydrates and amino acids,it is not surprising that product-specific and storage-specificconditions also play important roles. The ability of two C.maltaromaticum strains to relatively rapidly spoil meat duringgrowth under aerobic conditions at 7 1C but not or only after along storage period in vacuum packs at 2 1C serves as an example.

Carnobacteria as Pathogenic Organisms and/or Probiotic Cultures

Two human clinical cases caused by opportunistic infectionwith C. maltaromaticum and a Carnobacterium sp. have been

described. Two clinical isolates of C. divergens have also beenobtained. Carnobacteria are not known members of the humangastrointestinal microbial community, unlike several other LAB.The exceptions are an unpublished study reporting an unculturedCarnobacterium sp. clone from cow rumen and two studiesdemonstrating the presence of carnobacteria in pig and horseeffluent-impacted environments. Nevertheless, it is safe to concludethat the presence of Carnobacterium spp. in food does not presenta risk factor for human illness, except for the production of thebiogenic amine, tyramine, as discussed previously. Neither doCarnobacterium spp. appear to present a risk for nocosomialinfections in hospitals and similar institutions, although theyhave, in one instance, been isolated from contaminated bloodplasma.

The presence of virulence factors in carnobacteria is not welldocumented. Thus C. maltaromaticum strains isolated fromdiseased fish lacked hemolytic and phospholipolytic activities.Carnobacterium viridans shows, however, b-hemolytic activity onsheep blood agar. Although several isolates of carnobacteriaproduce bacteriocins, none of these compounds has been reportedto exert a cytolytic activity as shown for the En. faecalis class I-related bacteriocin cytolysin. Regarding chemotherapeutic agents,fish pathogenic strains of C. maltaromaticum were resistant toseveral of the agents widely used in aquaculture, such asoxytetracycline, quinolones, nitrofurans and potentiatedsulfonamides, but sensitive to erythromycin. This species is notresistant to lysozyme from salmonid eggs, which supports theidea that it is not vertically transmitted, at least in this fish species.Finally, it should be noted that two clinical isolates of C. divergenshave been reported to possess a gene encoding a new class ofpenicillinase. It will be of interest to further explore the distributionof this enzyme among carnobacterial species as well as itsfunctionality.

It has been documented that some strains of C. maltaromaticumare pathogenic for several fish species, including Australiansalmonids, carp, rainbow trout, striped bass and channel catfish,and salmon. Pathologic effects vary, and include, for example,septicemia, peritonitis, exophthalmia, accumulation of ascitic fluid,


and hemorrhages. In most (but not all) instances, the virulenceappears, however, to be low and only causes problems in fishundergoing severe stress, e.g. due to spawning. It is, therefore, notsurprising that C. maltaromaticum and C. maltaromaticum-likestrains are also found in the intestine or gills of healthy fish,including Arctic charr, Atlantic cod, Atlantic salmon, brownbullhead, trout, and various freshwater fish.

The presence of other carnobacterial species in healthy fishhas also been reported, including C. divergens, Atlantic salmon,Atlantic cod and wolffish, trout, and freshwater fish, C.alterfunditum-like [rainbow trout], C. funditum-like [Arctic charrand Atlantic cod], C. gallinarum [Atlantic cod], C. inhibens[Atlantic salmon], C. mobilelike [Arctic charr and Atlantic cod],and Carnobacterium spp. [Arctic charr and carp]. Clearly, furtherresearch is necessary to clarify under what circumstancesCarnobacterium spp. may be pathogenic for fish and whether theinfective capability is species and clone specific.

The reported pathogenicity of C. maltaromaticum has nothindered research into the possibility of using other carnobacteriaas probiotic cultures in aquaculture, including Carnobacteriumsp., C. alterfunditum, C. divergens, and C. inhibens. Isolates ofthese species, in addition to C. maltaromaticum, exhibit inhibitoryactivity towards bacterial fish pathogens. Carnobacteriumdivergens and C. maltaromaticum enhance the cellular andhumoral immune responses and cytokine expression ratios ofrainbow trout. Use of carnobacteria as probiotics leads to increasedsurvival of the fish in some instances [larvae of cod fry andAtlantic salmon fry, rainbow trout, and salmon], but not in others[larvae of cod fry, and rainbow trout]. One study showed that invitro exposure to C. divergens did not reverse the effects of bacterialpathogens, although these were alleviated. Additional research isnecessary to validate the usefulness of carnobacteria as probioticcultures.

Carnobacterium maltaromaticum has also been reported to bepathogenic for the fruit fly, Drosophila melanogaster, upon injectioninto the thorax. This specific case of pathogenesis is probably bestunderstood as a result of opportunistic infection, although C.maltaromaticum (but not C. divergens, C. gallinarum, C. inhibens,

C. mobile or C. viridans) exerts chitinolytic acitivity, a feature thatcan be perceived as targeting the chitin-containing exoskeleton ofinsects. Indeed, insects may serve as a host for this species.

Genomics

The complete chromosomes of many LAB species have nowbeen sequenced. Currently, 19 complete genome sequences ofstreptococci and 18 complete genome sequences of thenonpathogenic LAB representing 14 species from the orderLactobacillales are available. However, no genome sequence orphysical genetic map is available for Carnobacterium. The LABhave relatively small genomes for nonobligatory bacterial parasitesor symbionts, ranging from c. 1.6 to c. 3Mb. For C. divergens andC. pleistocenium, the genome size was estimated to 3.2Mb, and forC. alterfunditum pf4T, it was estimated to be 2.9Mb. At the timeof writing (March 2007), there is only one Carnobacterium genomesequencing project. Carnobacterium sp. AT7, a piezophilic strainisolated from the Aleutian trench at a depth of 2500m, is beingsequenced for the Moore Foundation Marine Microbial GenomeSequencing Project with a grant to the J. Craig Venter Institute. Thedata already available indicate that the Carnobacterium sp. AT7genome contains 2.4Mb and encodes 2388 proteins.Carnobacterium sp. AT7 is closely related to C. alterfunditum andC. pleistocenium.

To date, knowledge on the genes and DNA sequences ofCarnobacterium is comparatively sparse, concerning mostlybacteriocin-related genes in the species C. divergens and C.maltaromaticum and 16S rRNA and 16S-23S rRNA gene intergenicspacer sequences. Genes involved in some important carnobacterialmetabolic traits have been sequenced. The genetic informationobtained is, for most sequences, derived from just one strain, andmay therefore be strain specific, as shown for a number of L.monocytogenes genes.

Clearly, our knowledge of important phenotypes of strainsand species of Carnobacterium would benefit from the availabilityof a complete genome sequence for a strain representative of C.maltaromaticum, which is the species with most importance forthe food and aquaculture industries. The most extensive study on


phenotypic variation of C. maltaromaticum revealed that themajority of strains studied belonged to one phenotypic cluster(cluster H). Thus, in our opinion, a suitable candidate for a wholegenomic sequence could be either C. maltaromaticum LMG 22901(isolated from pork meat), LMG 22899 (isolated from cod) or LMG22898 (isolated from salmon), which all belong to this phenotypiccluster. LMG 22901 does not harbor plasmids, and thereforeappears to be a prime candidate for such an endeavor.

CONCLUDING REMARKS

We have come some way towards understanding importantaspects of the presence of carnobacteria in foods and theenvironment, particularly concerning the genetics of carnobacterialbacteriocins and their application. However, there are importantareas where knowledge on these bacteria is limited. These includethe following:

(1) Distribution and Quantitative Microbial Ecology:Carnobacterium spp. have not been reported from a numberof habitats, such as plants, fermented vegetable foods andthe gastrointestinal system of birds and mammals,including humans, that have otherwise been associatedwith several genera and species of LAB. This may, in someinstances, be due to faulty methodology for detectingcarnobacteria. Generally, however, our quantitativeunderstanding is limited regarding factors that influenceinactivation, survival or growth of carnobacteria in variousnatural environments. In foods, more detailed informationis available, but mathematical models for quantitativeprediction of processing, product and storage effects ongrowth and metabolic activity, including bacteriocin andtyramine formation, awaits further development togetherwith recording of responses in databases. In addition,further studies are needed to elucidate mechanismsunderlying the relationship between environmentalparameters and kinetic responses of carnobacteria.

(2) Routes of Food Contamination: To reduce the negative effectsof carnobacteria in foods (spoilage metabolites andtyramine), further information on routes of contamination

is desirable. In fact, very little research has been performedso far concerning this aspect of carnobacteria in food.

(3) Metabolic Activity: It is known that carnobacteria, or atleast C. maltaromaticum and C. divergens, have the capacityto produce a wide range of metabolites in chilled food.Further studies on the importance of these metaboliteswith respect to positive and negative sensory attributes ofvarious foods are, however, needed.

(4) Diversity: We are beginning to appreciate how interspeciesand intraspecies variation determine the positive andnegative effects of the presence of carnobacteria in foodand the environment, but we still do not understand howsuch variation affects, for instance, their survival in food-processing plants and contamination of foods. Thesignificance of such variation for their potential as fishpathogens also needs to be substantiated. This is also thecase for their role as spoilage organisms in meat and fishproducts, with the exception of a few products such asvacuum-packed smoked salmon and cooked MAP shrimps.

(5) Genomics: Representative genome sequences would offer avery valuable road map in order to answer many of theoutstanding questions associated with this importantgenus.

Staphylococcus Aureus Growth Boundaries: Moving towardsMechanistic Predictive Models Based on Solute - Specific Effects

The formulation of shelf-stable intermediate-moisture productsis a critical food safety issue. Therefore, knowing the preciseboundary for the growth-no-growth interface of Staphylococcusaureus is necessary for food safety risk assessment. This studywas designed to examine the effects of various humectants and toproduce growth boundary models as tools for risk assessment.The molecular mobility and the effects of various physicalproperties of humectants, such as their glass transitiontemperatures, their membrane permeability, and their ionic andnonionic properties, on S. aureus growth were investigated. Theeffects of relative humidity (RH; 84 to 95%, adjusted by sucroseplus fructose, glycerol, or NaCl), initial pH (4.5 to 7.0, adjusted by


HCl), and potassium sorbate concentration (0 or 1,000 ppm) onthe growth of S. aureus were determined. Growth was monitoredby turbidity over a 24-week period. Toxin production wasdetermined by enterotoxin assay. The 1,792 data points generatedwere analyzed by LIFEREG procedures, which showed that allparameters studied significantly affected the growth responses ofS. aureus. Differences were observed in the growth-no-growthboundary when different humectants were used to achieve thedesired RH values in both the absence and the presence ofpotassium sorbate. Sucrose plus fructose was most inhibitory atneutral pH values, while NaCl was most inhibitory at low pHvalues. The addition of potassium sorbate greatly increased theno-growth regions, particularly when pH was <6.0. Publishedkinetic growth and survival models were compared with boundarymodels developed in this work. The effects of solutes and differencesin modeling approaches are discussed.

Olsen et al. reported that Staphylococcus aureus was involvedin 42 documented outbreaks of food-borne poisoning in the UnitedStates, with 1,413 cases and 1 death occurring from 1993 to 1997.S. aureus has been estimated to cause approximately 185,000illnesses, 1,750 hospitalizations, and 2 deaths per year in theUnited States, all from consumption of contaminated foods. Food-borne staphylococcal poisoning, caused by the ingestion of one ormore preformed toxins in food contaminated with S. aureus, is oneof the most prevalent causes of gastroenteritis worldwide. Recently,powdered milk produced in Taiki, Hokkaido, Japan, was linkedto contaminated dairy products that made more than 14,700 peoplein Japan ill in June and July of 2000. Health authorities determinedthat dry skim milk powder was contaminated with staphylococcalenterotoxin A.

Scott related the relative vapor pressures (rvp's) of foods to thethermodynamic activity of water, using the definition aw = p/po,where aw is the water activity as derived from the laws ofequilibrium thermodynamics, p is the vapor pressure of the sample,and po is the vapor pressure of pure water. He showed a clearcorrelation between the rvp of the growth medium and the rate ofS. aureus growth. In the field of food science, the general acceptanceand application of this concept relating the rvp's of foods to a

minimum aw for microbial growth began with the review by Scott.The important point is that the equality of aw with p/po is basedon the assumption of the existence of thermodynamic equilibriumand that the value is true only for specified conditions oftemperature and pressure. Although dilute solutions obey thelaws of equilibrium thermodynamics, the same is not necessarilytrue for intermediate-moisture systems and is even less likely forlow-moisture systems, the properties of which are determinedmainly by slow reaction rates. The concept of aw then becomesmeaningless, because the measured rvp of water is no longer thesame as the equilibrium vapor pressure. The equation aw = p/potherefore applies only in the range 0.95 <aw <1.0, where equilibriumis established rapidly. With many foods, this assumption ofthermodynamic equilibrium is violated, so the appropriate term isrvp or relative humidity (RH) rather than aw.

For the past several decades, in the food science literature, thephenomenon denoted by the term "aw" has been widely used topredict microbial growth as well as the relationship between manycommon food deterioration reactions and aw. Although RH is abetter indicator of food stability and safety than the water contentof a system, it is not always a reliable predictor. For example,several authors have reported that the growth boundaries forvarious genera of microorganisms differ depending on the type ofhumectant used to depress RH rather than the absolute value ofRH. The literature on microbial osmoregulation via compatiblesolutes has also shown that the type of humectant, rather than theabsolute value of RH, is critical.

In recent years, increasing evidence based on glass transitiontheory has shown that molecular mobility (Mm) may be an attributethat deserves further attention, as it is related to many importantdiffusion-limiting properties of foods. This is because thetemperature at which a viscous liquid changes to a "glass" (glasstransition temperature [Tg]), which governs Mm, is a solute-specificproperty that is inversely linearly related to the measured RH ofa system. Although the use of RH has served the food industrywell for the past several decades since Scott introduced the conceptof aw in 1953, further study to understand these solute-specificeffects on microbial growth should continue.


Microbial growth models are typically developed when theobjective is to understand the responses of microorganisms whenpart of the range of conditions studied permits growth to occur.Such models can describe the increase in numbers with time(kinetic models), the conditions allowing growth or no growth(boundary models), or the chance of growth (probabilistic models).Kinetic growth models are typically based on growth curves overa range of conditions and can allow growth rates to be predicted.In the 1980s, the United Kingdom Ministry of Agriculture, Fisheries,and Food and the U.S. Department of Agriculture (USDA) fundedprograms to develop kinetic models for the growth, survival, anddeath of food-borne pathogens, including S. aureus. This workresulted in suites of models which are currently available. NaClwas the humectant chosen to control the RH in these models, andthe resulting kinetic models do not allow the growth-no-growthboundary to be estimated.

The statistical effects and interactions of RH, initial pH, andpotassium sorbate concentration on the boundary for S. aureusgrowth were modeled in this study by using the approachdescribed previously. Additionally, the concepts of Mm and theeffects of various physical properties of humectants, such as theTg, membrane permeability, and ionic and nonionic properties ofsolutes, on S. aureus growth were investigated by controlling theRH with sucrose-fructose (50:50, wt/wt), glycerol, or NaCl.

MATERIALS AND METHODS

Preparation of Bacteria and Media

Five S. aureus strains were used as a cocktail in this study:ATCC 13565, ATCC 14458, and ATCC 27664, from the AmericanType Culture Collection, Manassas, Va.; D-2, from ToxinTechnology, Inc., Sarasota, Fla.; and A-100, from the U.S. ArmyNatick Laboratories, Natick, Mass. A cocktail of strains was chosenfor use because it helps to overcome the variability between strainsand increase the chance of detecting the fastest growth at differentpoints in the experimental matrix. These benefits of using a cocktailhelp to create a fail-safe model without repeating the experimentmany times with different strains. All strains showed typical growthon Baird-Parker agar (Difco, Detroit, Mich.) plates and were

coagulase positive. Their identification as S. aureus was confirmedby the use of a Riboprinter. Stock cultures were grown overnightat 37°C in brain heart infusion (BHI) broth (Difco), suspended insolution containing a 2:1 BHI broth-glycerol solution, and storedat -80°C until needed.

For this study, BHI broth was reconstituted with theappropriate ratio of water to sucrose-fructose (50:50, wt/wt) or toNaCl to achieve the desired RHs for the study (84, 88, 92, or 95%).Appropriate amounts of 10% stock solutions of potassium sorbate(Hoechst, Frankfurt, Germany) were added to achieve a finalconcentration of 1,000 ppm. The pH was adjusted to 4.5, 5.0, 6.0,or 7.0 by using 0.1 or 1.0 N HCl. The final RH was determined byusing an Aqualab CX-2 aw meter. The RHs of the media measuredat ambient temperature differed by no more than ±0.003% relativeto those measured at 37°C. Each of the 64 broths was filter sterilizedand stored in screw-cap tubes at 4°C until needed.

Bioscreen Microtiter Plate Preparation, Incubation, andMeasurement

The five stock cultures were removed from -80°C storage, andone loopful of each was transferred separately into 9 ml of BHIbroth. The inoculated broths were incubated for 18 h at 37°C. Thefive cultures were individually adjusted to optical densities (ODs)at 530 nm of 0.750 to 0.780 (measured with a Perkin-Elmer model35 spectrophotometer) with 0.1% sterile peptone to achieve aconcentration of approximately 108 CFU/ml. Diluted cultures (10ml) were transferred into a single sterile test tube and vortexed tocreate the S. aureus cocktail used in our experiments. Three 1/10dilutions with sterile 0.1% peptone water were used to make afinal working cocktail with a concentration of approximately 105

CFU/ml. This working cocktail was kept on ice until it was usedto inoculate the various media used in the experiments. Samples(1 ml) of the working cocktail were taken, serially diluted, spreadplated onto Baird-Parker agar plates, and incubated at 37°C for 48h, and cells were counted to determine their initial concentrations.

Each of the 64 broths (10 ml) was aseptically transferred to asterile tube, and the S. aureus cocktail (100 µl) was added to thebroth to achieve an initial concentration of approximately 103


CFU/ml. The inoculated broths (400 µl) were asepticallytransferred into sterile 100-well microtiter plates, with eight replicatewells per medium. The outer wells of the microtiter plates werefilled with uninoculated medium to act both as a partial moisturebarrier and as uninoculated controls. Each plate contained mediaat only one RH level. A piece of sterile thermaseal film was placedon top of the open wells under the lid during incubation. Theplates were placed on a rack in a sealed plastic container thatcontained the appropriate saturated salt slurries needed tomaintain the environmental RH and thereby ensure the stabilityof the RHs of the media. The saturated salt solutions used wereas follows: ZnSO4 · H2O (83% RH at 37°C), KNO (89% RH at37°C), KPO4 (93% RH at 37°C), and K2Cr2O7 (96% RH at 37°C).The containers were closed and incubated at 37°C. Plates wereperiodically removed and shaken for 60 s, and the ODs of the1,024 individual wells were measured for up to 24 weeks by usingthe wideband filter on the Bioscreen C system (Labsystems Oy,Helsinki, Finland). The experiments were repeated on separatedates. A total of 1,024 wells were monitored.

Data Collection

The S. aureus population was considered to have grown whenthe Bioscreen OD measured with a wideband filter (ODwb)increased from an initial reading between 0.200 and 0.220 to anODwb of 0.350 or higher. The time to growth (TTG) was determinedby calculating the geometric mean of the time of the lastmeasurement that showed no growth (ODwb < 0.350) and the firsttime point that showed growth (ODwb > 0.350). In instances of nogrowth, TTG was censored at the final measurement time of 168days. A threshold ODwb value of 0.350 was used to score wellswith S. aureus growth (ODwb ³ 0.350) and wells with no growth(ODwb < 0.350). This threshold value was determined as discussedin detail by Stewart et al. (40). In brief, ODwb measurements werecompared with plate counts, and an ODwb of 0.350 was identifiedas the lowest ODwb value at which an increase in turbidityregularly correlated to an increase in plate counts.

OD measurements were transferred from the Bioscreen C ASCIIfile to a DOS text editor and then imported into Statistica software

for initial analysis. Data were transferred to a Microsoft Excelspreadsheet to determine TTG and then to SAS software formodeling by the LIFEREG procedure.

An additional data set was also included in the developmentof the model. This data set included information on TTG for thesame S. aureus cocktail obtained by an identical experimentalprocedure, with the exception that the humectant used to achievethe desired RH of the BHI broth was glycerol, a glass-formingmembrane-permeable solute. The experimental design of Stewartet al. included three levels of potassium sorbate (0, 500, and 1,000ppm). A total of 768 data points from the glycerol data set wereincluded in the data analysis.

Statistical Analysis

The TTG data along with a censoring indicator, the percentRH, the pH, the potassium sorbate level, and the humectant type(sucrose-fructose, glycerol, or NaCl) were input data for theconstruction of the growth boundary model. RH, pH, andpotassium sorbate values were normalized [(RH - 89.75)/4.15],[(pH - 5.625)/0.96], [(sorbate - 500)/500] prior to model development.The TTG results from replicate test wells were averaged beforemodeling to minimize the effect of measurement error. The SASLIFEREG procedure was used to develop predictive models of thenatural logarithm of TTG (ln TTG) as a function of the factorspercent RH, pH, preservative level, and humectant type. Theresulting model can easily be transformed into a regular timescale. It should be noted that the type of humectant is considereda categorical variable (i.e., nonnumeric) in the SAS procedure. Ingrowth modeling, there are often conditions under which nogrowth occurs; therefore, TTG may be missing or censored. Underthese conditions, ordinary least-squares regression is not applicableand special procedures are required.

The LIFEREG procedure can accommodate such censored dataand uses maximum-likelihood estimation methods to findregression coefficients. Further details of this method of analysisare given by Allison (1). The LIFEREG procedure works best withmany replicates in which the intervals between growth parametersare equally spaced, with approximately 50% of conditions limiting


growth. The ranges of the parameters studied (RH, pH, andpreservative levels) were chosen to satisfy these requirements.

Toxin Production

Toxin production for experimental conditions close to thegrowth-no-growth boundary was measured by the VIDAS staphenterotoxin assay. The VSEA is a qualitative enzyme-linkedfluorescent immunoassay performed in an automated mini-VIDASinstrument. The media from the microtiter plates were removed forassay after the final ODwb measurements were taken (24 weeks).The broth from each of the eight replicate wells was removed viapipette, combined into two microcentrifuge tubes, and centrifugedfor 10 min. The supernatants were combined, and the pH valueswere adjusted with 1 N NaOH to 6.0 to 8.0. Samples (0.5 ml) wereplaced into the appropriate wells in a test strip, and the assayswere run in the mini-Vidas. Positive and negative controls wererun with each set of test strips. The VSEA is sensitive to 1 ng oftoxin/ml of sample.

RESULTS

Model Development and Diagnostics

Models were developed from the 1,792 data points generated.The models included all three main effects: percent RH, pH, andpotassium sorbate concentration ("sorb" in the equations); theirquadratic effects (RH x RH = RH2; pH x pH = pH2; and sorb xsorb = sorb2); and the three two-factor interactions (RH x pH, RHx sorb, and pH x sorb). The impact of the humectant type on themain effects and interactions was assessed by including indicatorvariables in the model.

These main effects, quadratic effects, and two-factor interactionsare collectively referred to as the factors. SAS LIFEREG allows theuser to specify the error distribution to account for the variationin TTG not explained by the regression model. For modeldevelopment purposes, Weibull, lognormal, and loglogisticdistributions were considered. All three distributions gave verysimilar models (the regression coefficients were similar in signand magnitude), and we selected the loglogistic distribution model,which has been used in previous work.

Model development is an iterative process. The first analysisof the data created models which were all-inclusive; this allowedfor the determination of those factors which had a significanteffect on the growth of S. aureus. All of the main effects, quadraticeffects, and two-factor interactions were significant (P < 0.005),except for the potassium sorbate concentration quadratic term andthe interaction between RH and potassium sorbate concentration.The insignificant factors were dropped, and the data werereanalyzed to develop the final models. SAS LIFEREG outputs atable of regression coefficient estimates and approximate chi-squaredistribution P values for each factor in the model. The relativeimportance of each factor can be judged by its P value: factors withlow P values have the greatest influence on and are most predictiveof the TTG.

The resulting model equation for the use of sucrose-fructose asthe humectant was as follows (where TTG is measured in days,RH is a percentage, and the potassium sorbate concentration ismeasured in parts per million): ln TTG = 4.46 - (5.90 x RH) - (2.62x pH) + (1.42 x sorb) + (1.78 x RH2) + (1.06 x pH2) + (0.48 x RHx pH) - (0.94 x pH x sorb).

The resulting model equation for the use of glycerol as thehumectant was ln TTG = 2.24 - (3.86 x RH) - (2.29 x pH) + (1.00x sorb) + (1.78 x RH2) + (1.06 x pH2) + (0.48 x RH x pH) - (0.94x pH x sorb).

The resulting model equation for the use of NaCl as thehumectant was ln TTG = 3.15 - (3.59 x RH) - (3.81 x pH) + (1.64x sorb) + (1.78 x RH2) + (1.06 x pH2) + (0.48 x RH x pH) - (0.94x pH x sorb).

From the model equations, the TTG contour plots for sucrose-fructose, glycerol, and NaCl were created. The contour plots werecreated by generating a grid of evenly spaced points in theexperimental design space of pH, RH, and potassium sorbateconcentration. With a grid of 2,268 points, a predicted TTG wascalculated for each point and humectant by using the appropriatemodel shown above.

The grid of points and predicted TTG values was input intothe SAS procedure "proc gcontour". Diagnostic methods to check


for departures from model assumptions can be used in a mannerthat is similar to their use in ordinary regression analysis. However,when nonnormal distributions are being fitted, an appropriatestandardization of residuals must be used (25). The SAS reliabilityprocedure was used for residual analysis. A probability plot ofstandardized residuals was examined, and no unusual patternsor anomalies were detected.

Model Development and Diagnostics

Models were developed from the 1,792 data points generated.The models included all three main effects: percent RH, pH, andpotassium sorbate concentration (“sorb” in the equations); theirquadratic effects (RH x RH = RH2; pH x pH = pH2; and sorb x sorb= sorb2); and the three two-factor interactions (RH x pH, RH xsorb, and pH x sorb). The impact of the humectant type on themain effects and interactions was assessed by including indicatorvariables in the model. These main effects, quadratic effects, andtwo-factor interactions are collectively referred to as the factors.SAS LIFEREG allows the user to specify the error distribution toaccount for the variation in TTG not explained by the regressionmodel. For model development purposes, Weibull, lognormal, andloglogistic distributions were considered. All three distributionsgave very similar models (the regression coefficients were similarin sign and magnitude), and we selected the loglogistic distributionmodel, which has been used in previous work.

Model development is an iterative process. The first analysisof the data created models which were all-inclusive; this allowedfor the determination of those factors which had a significant effecton the growth of S. aureus. All of the main effects, quadratic effects,and two-factor interactions were significant (P < 0.005), except forthe potassium sorbate concentration quadratic term and theinteraction between RH and potassium sorbate concentration. Theinsignificant factors were dropped, and the data were reanalyzedto develop the final models. SAS LIFEREG outputs a table ofregression coefficient estimates and approximate chi-squaredistribution P values for each factor in the model. The relativeimportance of each factor can be judged by its P value: factors withlow P values have the greatest influence on and are most predictiveof the TTG.

The resulting model equation for the use of sucrose-fructose asthe humectant was as follows (where TTG is measured in days,RH is a percentage, and the potassium sorbate concentration ismeasured in parts per million): ln TTG = 4.46 - (5.90 x RH) - (2.62x pH) + (1.42 x sorb) + (1.78 x RH2) + (1.06 x pH2) + (0.48 x RHx pH) - (0.94 x pH x sorb).

The resulting model equation for the use of glycerol as thehumectant was ln TTG = 2.24 - (3.86 x RH) - (2.29 x pH) + (1.00x sorb) + (1.78 x RH2) + (1.06 x pH2) + (0.48 x RH x pH) - (0.94x pH x sorb).

The resulting model equation for the use of NaCl as thehumectant was ln TTG = 3.15 - (3.59 x RH) - (3.81 x pH) + (1.64x sorb) + (1.78 x RH2) + (1.06 x pH2) + (0.48 x RH x pH) - (0.94x pH x sorb).

From the model equations, the TTG contour plots for sucrose-fructose, glycerol, and NaCl were created. The contour plots werecreated by generating a grid of evenly spaced points in theexperimental design space of pH, RH, and potassium sorbateconcentration. With a grid of 2,268 points, a predicted TTG wascalculated for each point and humectant by using the appropriatemodel shown above. The grid of points and predicted TTG valueswas input into the SAS procedure “proc gcontour”. Diagnosticmethods to check for departures from model assumptions can beused in a manner that is similar to their use in ordinary regressionanalysis. However, when nonnormal distributions are being fitted,an appropriate standardization of residuals must be used. TheSAS reliability procedure (SAS Institute, Inc.) was used for residualanalysis. A probability plot of standardized residuals wasexamined, and no unusual patterns or anomalies were detected.

MODEL PREDICTIONS AND COMPARISON WITHPUBLISHED WORK

As conditions became increasingly unfavorable for growth,the contour lines drew closer together, indicating that conditionswere approaching those that do not allow growth. As the RH orpH of the system decreased, a corresponding increase in TTG wasseen and the no-growth area of the contour plot increased in size.The addition of potassium sorbate at low pH values (4.5 to 5.5)


dramatically changed the contour plots in the low-pH area of theplot. The effect of potassium sorbate decreased as the pH valueapproached 7.0.

Differences were observed in the growth-no-growth boundarywhen different humectants were used to achieve the desired RHvalues in both the absence and the presence of potassium sorbate.The use of sucrose-fructose as the humectant resulted in the greatestdelay of S. aureus growth under the highest pH and RH combination(i.e., pH 7.0 and 88% RH) either with or without potassium sorbate.Even with the addition of potassium sorbate at 1,000 ppm at pH7.0, glycerol and NaCl were less effective in delaying S. aureusgrowth than sucrose-fructose with no added preservative. At pHvalues above ca. 5.3 and in the absence of potassium sorbate, arank order was observed, with NaCl being the least inhibitory,followed by glycerol, and with sucrose-fructose being the mostinhibitory; by contrast, at pH values below ca. 5.3, NaCl showeddramatically increased inhibition.

With the addition of potassium sorbate at 1,000 ppm, the samerank order occurred at pH values above ca. 6.1. It has generallybeen accepted that the limits for the growth of S. aureus are 85%RH when NaCl is used as the humectant and 89% when glycerolis used.

Our models showed the limiting RH for growth to beapproximately 88% (pH 7.0) when sucrose-fructose was used, 86%when glycerol was used, and 85% when NaCl was used. Thelimiting RH with glycerol as the humectant was markedly differentfrom that reported by Marshall et al., who found that the growthlimits were 89 and 86% RH when glycerol and NaCl were usedas the humectants, respectively.

This disagreement may be due to strain variation or to the factthat only quarter-strength BHI broth was used in the work byMarshall et al. while full-strength BHI broth was used in ourmodel system. Whatever the reason, this result suggests that ifglycerol makes a large contribution to the humectant content of afood product, it would be wise to consider the target RH carefully.

Additionally, there is published literature that describes theability of S. aureus to grow in various food systems, mainly meats

and cheeses. Unfortunately, these data cannot be used for modelvalidation, as RH was not reported and/or other parameters notincluded in our models (such as oxygen concentration ortemperature) were varied. Additionally, the pH and/or ingredientlists of the products used in these studies were not reported.

NaCl was most inhibitory at pH values below ca. 5.3 in theabsence of potassium sorbate and at pH values below ca. 6.1 in thepresence of potassium sorbate at 1,000 ppm. For example, no growthwas seen even at conditions of 95% RH. This finding disagreeswith predictions made by the USDA’s pathogen modeling program(PMP) version 5.1, which indicates that at 95% RH and pH 4.5,S. aureus will grow from 103 to 107 CFU/ml in 1.65 days.Comparison of the sucrose-fructose and glycerol models developedin this study with other published work is difficult, as the majorityof previously published studies have used NaCl to control the RHof the system.

As discussed previously, differences in the physical propertiesof humectants can lead to various responses, even when RH isconstant. The differences in modeling approaches also makecomparisons difficult. Neither the Food Micro Model of the UnitedKingdom Ministry of Agriculture, Fisheries, and Food nor PMPwas created from data sets containing many replicated experimentswhere the bacteria were subjected to stressed conditions. There isalso the danger of extrapolation beyond the conditions under whichdata were actually collected. The PMP gives a prediction for growthunder the conditions of pH 4.5 and 95% RH (with NaCl as thehumectant), although data were not collected for this combinationof parameters. A full factorial design was utilized for the boundarymodels presented in this paper, and the models were constructedby using interpolation only.

Enterotoxin Assays

When the final ODwb was close to 0.350, the samples weretested for enterotoxin. Assays were also conducted for all sampleswith combinations of 84% RH and pH 7.0 or 84% RH and pH 4.5.A total of 94 assays were run, and in every case where the datawere scored as “no growth,” the toxin assay results were negativefor the presence of toxin, while in every case where the data were


scored as “growth,” the toxin assay results were positive for thepresence of toxin. This further supports the choice of an ODwb of0.350 as an appropriate cutoff value for TTG in the individualwells and greatly increases confidence in the practical value of themodels.

DISCUSSION

Using RH without considering the effect of the type ofhumectant has often led to difficulties in predicting the stabilityand microbial safety of intermediate-moisture foods. For example,S. aureus is highly salt tolerant and has been reported to grow atRHs as low as 85% in NaCl concentrations up to 25% (wt/wt);however, RH values limiting growth are typically higher whenhumectants other than NaCl are used to control RH.

Notermans and Heuvelman showed that the RH limiting S.aureus growth varied depending on whether sucrose (with whichit was reported that an aw of 0.90 limited growth) or NaCl (withwhich it was reported that an aw of 0.87 limited growth) was usedas the humectant.

Marshall et al. reported that the level of inhibition of the growthof S. aureus by glycerol at neutral pH was about 10% greater thanthat caused by NaCl in the rvp range (reported as aw) from 0.96to 0.90.

Clearly, the choice of humectant used to depress the RH in afood system is critical and its effects on microbial survival and/or growth depend not only on the RH but also on the chemical andphysical properties of the humectant, as well as the biologicaleffects of the humectants on the microorganisms.

As stated previously, although RH is a better indicator forfood stability and safety than water content, it is not always areliable predictor of stability. As described by Fennema, increasingevidence has shown that the Tg and Mm may be important meansto explain many diffusion-limiting properties of food.

Interest in Mm was first generated by Luyet and associates inthe United States as well as by Rey in France, who showedrelationships between Mm and biological materials, while basicconcepts relating nonequilibrium Mm in synthetic amorphous

polymer systems were formulated by Ferry. The importance of therole of glassy states in various sugar-containing foods wasdescribed by White and Cakebread, who suggested that these statesplayed an important role in the stability and processability ofmany food systems.

The basic principles that apply to synthetic amorphouspolymers, as described by Ferry and others, also apply to thebehavior of glass-forming foods. Foods of this type include starch-containing products like pasta and bread, boiled confections, protein-based foods, intermediate-moisture foods, and dried, frozen, orfreeze-dried foods.

The RH and Mm approaches to food stability arecomplementary, not contradictory. RH, referred to as aw, focuseson the availability of water in a food matrix, while Mm defines thisavailability in terms of microviscosity and diffusibility, with thelatter being dependent on the properties of water as a plasticizerto depress the Tg

of the food matrix.

Humectants with various characteristics were employed inthis study specifically to begin to explore the influence of variousphysical properties of humectants, such as their Tgs, on microbialgrowth under osmotically stressed conditions. When glass-formingsolutes are used to achieve desired RH values in systems, theunderlying Tg and the relaxation behavior of the glasses becomethe controlling parameters of the system. Within each glass-formingsystem, Tg is inversely linearly related to the measured RH of thesystem.

However, across different glass-forming solute systems, Tgincreases as the weight-average molecular weight of the solutesincreases. As Tg increases, the viscosity of the system increases,which in turn decreases the mobility of molecules in the system(including the mobility of water molecules). This relative decreasein mobility between glass-forming systems may partially explainthe varied responses of microorganisms in systems with matchingRH values but in which different humectants were utilized toachieve the targeted RH.

The humectants used in this study included sucrose-fructose(non-membrane-permeable glass-forming solutes with a reference


Tg of approximately -32°C), glycerol (a membrane-permeable glass-forming solute with a reference Tg of approximately -65°C), andNaCl (a non-glass-forming ionic solute which may affect variousion channels in the cell).

The limits of microbial growth cannot be predicted by the RHof the system alone, as seen by comparison of the growthboundaries developed with the three types of humectants used inthis study. Instead, the RH of the system should be considered inconjunction with the physical properties of the solutes and theireffects on the biological systems. Ionic solutes, such as NaCl, mayplace both ionic and osmotic stresses on the cell. Of the threehumectants studied, NaCl was the most inhibitory below pHvalues of ca. 5.3 in the absence of potassium sorbate.

The no-growth region of NaCl was expanded to pH valuesbelow ca. 6.1 with the addition of potassium sorbate at 1,000 ppm.This exaggerated interaction between NaCl, low pH, and potassiumsorbate may suggest that the Na+ proton antiport may be thecritical ion channel involved.

These data suggest that growth inhibition may be due tooverwhelming ionic stress. The membrane permeability of solutessuch as glycerol may influence the degree of osmotic stressexperienced by the cell.

Additionally, the influence of the Tg of the solute on Mmdirectly affects the degree of osmotic stress a specific solute (e.g.,sucrose-fructose or glycerol) places on the cell. Sucrose-fructosewith a higher Tg than that of glycerol (and, therefore, a largerconstraint on Mm than that of glycerol) was most inhibitory at pHvalues above ca. 5.3 or 6.1 in the absence or presence of potassiumsorbate, respectively.

Glycerol is also membrane permeable, and therefore, the cellmay experience less osmotic stress than when sucrose-fructose(membrane impermeable) is utilized. Physiological effects on S.aureus cells have been shown to differ when various solutes areused to cause osmotic stress. Cellular transport systems responsiblefor the uptake of compatible solutes by S. aureus during osmoticstress have been shown to respond differently to the stresses causedby NaCl, sucrose, and glycerol.

Additionally, the effects of osmotic stress on S. aureus cell sizealso differ depending on whether NaCl, sucrose, or glycerol isused as the humectant. Although these models are empirical innature, they demonstrate that the limits of growth based on RHwould be different when various humectants were used to controlRH, so comparing studies based on RH alone may lead to falseconclusions.

These results also allow us to begin to explore the possibilityof moving towards a mechanistic approach to creating predictivemodels. Further work in this laboratory to study solute-specificeffects on cellular energy (ATP levels) is ongoing in an effort tocontinue to work towards a better understanding of themechanism(s) of bacterial inactivation.

Three statistical models describing the growth boundaries forS. aureus with respect to RH controlled by three types of humectants,pH, and preservative were developed, and results were confirmedby toxin assays. Building on previous work that has establishedregions of growth and no growth, these mathematical modelspredict actual TTG as a function of the humectant type, RH, pH,and preservative level.

These predictions can be used to develop much more detailedand informative boundary surfaces. Good agreement between allthree of these growth boundary surfaces and PMP growth kineticmodels was found in the relatively unstressed regions. Growthboundary predictions were outside the confidence limits of thePMP growth kinetic model under more stressful conditions, evenwhen NaCl was used as the humectant.

Kinetic models and growth boundary models can be used ascomplementary tools for the development of microbially safeproducts with short and long shelf lives. These models will allowproduct developers to visualize the “safe space” for the formulationof shelf-stable intermediate-moisture foods by employing thesepreservation factors, will allow microbiologists to assess risksmore effectively over a wide range of products, and will ultimatelyallow the consumer to have greater assurance of food safety.

The use of sucrose-fructose as the humectant results in thegreatest delay of S. aureus growth when the highest pH and RH

Introductory Food Microbiology44 Role of Ctc from Listeria Monocytogenes in Osmotolerance 45

(i.e., pH 7.0 and 88% RH) are used either with or without potassiumsorbate. NaCl is the most inhibitory humectant at pH values belowca. 5.3 in the absence of potassium sorbate and at pH values belowca. 6.1 in the presence of potassium sorbate at 1,000 ppm. Solute-specific effects on bacterial growth were evident, as growthboundaries differed at the same RH values when differenthumectants were used.

Over the almost 50 years since Scott first published his workon the water relations of S. aureus, measurement techniques andthe fields of physical chemistry, microbial physiology, and microbialmodeling have progressed significantly. Scott was well aware thataw would not necessarily be adequate to describe all of theproperties of solutions that influence microbial growth, metabolism,and survival. He suggested that, with further work to fill some ofthe obvious gaps in our knowledge, we would discover whetheraw could be replaced by something more meaningful. With carefulexperimentation and understanding, we will continue to fill thegaps and allow the food industry to produce the highest-qualityproducts while ensuring food safety.

2ROLE OF CTC FROM LISTERIA

MONOCYTOGENES INOSMOTOLERANCE

Listeria monocytogenes is a food-borne pathogen with the abilityto grow under conditions of high osmolarity. In a previous study,we reported the identification of 12 proteins showing highinduction after salt stress. One of these proteins is highly similarto the general stress protein Ctc of Bacillus subtilis. In this study,induction of Ctc after salt stress was confirmed at thetranscriptional level by using RNA slot blot experiments. To explorethe role of the ctc gene product in resistance to stresses, weconstructed a ctc insertional mutant. No difference in growth wasobserved between the wild-type strain LO28 and the ctc mutanteither in rich medium after osmotic or heat stress or in minimalmedium after heat stress. However, in minimal medium afterosmotic stress, the growth rate of the mutant was increased by afactor of 2. Moreover, electron microscopy analysis showed impairedmorphology of the mutant grown under osmotic stress conditionsin minimal medium. Addition of the osmoprotectant glycine betaineto the medium completely abolished the osmotic sensitivityphenotype of the ctc mutant. Altogether, these results suggest thatthe Ctc protein of L. monocytogenes is involved in osmotic stresstolerance in the absence of any osmoprotectant in the medium.

INTRODUCTION

Listeria monocytogenes is a food-borne pathogen widelydistributed in the environment. This microorganism is of particular

Introductory Food Microbiology Role of Ctc from Listeria Monocytogenes in Osmotolerance46 47

concern in the food industry because of its ability to survive andfrequently to grow under a wide range of adverse conditions usedto preserve food such as low temperature, low pH, and highosmolarity. Growth of L. monocytogenes has been reported at NaClconcentrations as high as 10%.

Most bacteria cope with elevated osmolarity in the environmentby intracellular accumulation of compatible solutes, calledosmolytes. Among the compatible solutes efficient in L.monocytogenes, two quaternary amines, glycine betaine andcarnitine, are the most effective. The accumulation of theseosmoprotectants in L. monocytogenes occurs through osmoticactivation of their transport from the medium rather than throughde novo synthesis. Accumulation of glycine betaine and carnitineoccurs via at least two glycine betaine transporters encoded by thebetL gene and the gbu operon and one carnitine transporter encodedby the opuC operon. A betL knockout mutant and a mutant of gbuobtained by transposition were significantly affected in their abilitiesto accumulate glycine betaine and were unable to withstandconcentrations of salt as high as the isogenic parent strain canwithstand. Similarly, a mutant with an insertional inactivation ofopuCA was defective in the uptake of carnitine and had impairedgrowth at high osmolarity. Proline has been identified as anosmolyte for L. monocytogenes. The proline transport mechanismhas not been characterized yet. However, the proBA operon, codingfor the enzymes that catalyze the two first steps of prolinebiosynthesis, has recently been identified. Disruption of this operonsignificantly reduced the growth of the corresponding mutant athigh salt concentrations. However, little information is availableconcerning other mechanisms that take place in L. monocytogenesto enable the organism to cope with osmotic stress, especiallywhen osmolytes are not available in the environment.

In a previous study, by proteomic analysis and massspectrometry or N-terminal sequencing, 12 proteins showing highinduction after a salt stress were identified. One of these proteinsis similar to the Ctc protein of Bacillus subtilis, a general stressprotein which belongs to the L25 family of ribosomal proteins. InB. subtilis, the ctc gene is induced in response to osmotic, heat, andoxidative stress and glucose limitation (14, 41). Regulation of the

expression of the ctc gene of B. subtilis occurs via the sB RNApolymerase subunit. The ctc promoter was one of the first sB-dependent promoters identified and for this reason has beenextensively studied. It is the best-characterized sB-dependentpromoter and has become the promoter of choice in nearly allinvestigations of sB regulation. In contrast to the wealth ofinformation regarding the ctc promoter, the function of the ctcproduct itself in B. subtilis is less clear and seems to be dispensable.Only reduced sporulation efficiency at high temperatures has beenobserved in a ctc null mutant.

To investigate the function of the Ctc protein in L. monocytogenes,especially with regard to the stress resistance of the bacterium, weanalyzed the sequence of the corresponding gene and inactivatedit by insertional mutation. Physiological studies indicate that theCtc protein facilitates growth in minimal medium under conditionsof high osmolarity and in the absence of an osmoprotectant. Thisis the first time that a role has been assigned to Ctc, which belongsto a family of unknown proteins.


Bacterial Strains and Plasmids

L. monocytogenes LO28, a clinical isolate, was obtained from P.Cossart (Institut Pasteur, Paris, France). Bacterial plasmids werepropagated in Escherichia coli TG1. Plasmid pHT315 was used asa cloning vector for sequencing, and plasmid pAUL-A was usedfor gene disruption.

Culture Media and Stress Conditions

Cells were grown on complex culture media: brain heartinfusion (BHI) broth or agar (Difco Laboratories, Detroit, Mich.). Achemically defined minimal medium called Improved MinimalMedium, or IMM, was also used, but pyridoxal, which is notnecessary for growth, was not added. Different stress conditionsused for growth rate or microscopy experiments were inducedaccording to the following procedure. Overnight cultures of strainLO28 or the ctc mutant were used to inoculate fresh medium at aninitial optical density at 600 nm (OD600) of 0.05. For heat stress,BHI medium or IMM was inoculated with a preculture grown in


the same medium, and the cultures were incubated with shakingat 45°C. For osmotic stress, either BHI medium with or without5.5% NaCl (final concentration, 6%) was inoculated with apreculture grown in BHI medium, or IMM with or without 3.5%NaCl or with 0.6 M xylose was inoculated with a preculture grownin IMM. Betaine (1 mM) or carnitine (1 mM) was added to IMMcontaining NaCl when required. The cultures were incubated withshaking at 37°C. Growth rate experiments were performed with aMicrobiology Reader Bioscreen C (Labsystems, Helsinki, Finland)in 100-well sterile microplates, each well containing 300 µl ofculture medium. The OD600 was monitored. Experiments wereperformed at least in duplicate and were repeated independently,twice for heat stress experiments and three times for osmotic stressexperiments.

Antibiotics were used at the following concentrations:ampicillin at 100 µg ml-1 for E. coli, and erythromycin at 5 µg ml-

1 and rifampin at 200 µg ml-1 for L. monocytogenes.

Cloning and Sequencing

Plasmids were prepared using the Plasmid Midi kit (Qiagen,S.A., Courtabuf, France). Bacterial chromosomal DNA was isolatedas described previously. Restriction endonucleases, T4 DNA ligase,and Taq polymerases were used as recommended by themanufacturer (Roche Molecular Biochemicals, Mannheim,Germany). DNA restriction fragments were purified from agarosegels by using the QIAquick gel extraction kit (Qiagen).Oligonucleotides were synthesized by MGW-Biotech. Plasmidswere introduced into E. coli by standard methods, while for L.monocytogenes, electroporation was achieved as describedpreviously.

DNA sequencing was performed with the BigDye terminatorcycle sequencing ready reaction kit (Applied Biosystems, Courtabuf,France), and sequences were analyzed with an automatic DNAsequencer (ABI Prism 310 genetic sequencer; Applied Biosystems).Searches for sequence homology were performed with the FASTAprogram. Sequencing of the ctc gene of L. monocytogenes LO28 wasperformed as follows. A 1,185-bp chromosomal DNA fragmentcontaining the ctc gene was amplified by PCR using primers OD1

and OD2 with incorporation of two restriction sites, HindIII andBamHI. These primers were designed using the complete nucleotidesequence of L. monocytogenes strain EGDe. The PCR product wasdigested with the HindIII and BamHI restriction enzymes and wascloned into similarly digested plasmid pHT315. Inserts of twoplasmids were sequenced.

Construction of a Ctc Insertional Mutant

The ctc gene was insertionally inactivated by a simplerecombination event using the temperature-sensitive suicide vectorpAUL-A. A 435-bp internal ctc fragment was PCR amplified fromchromosomal DNA with primers OD3 and OD4 with theincorporation of two restriction sites, HindIII and BamHI. Thepurified PCR product was digested with HindIII and BamHI, ligatedinto similarly digested plasmid pAUL-A, and then transformedinto E. coli. The resultant plasmid pAUL-CTC was used to createa knockout in ctc by homologous recombination with the L.monocytogenes chromosome as described previously. In the resultingmutant, 135 bp in the 3' end of ctc was deleted. Southern blotanalysis and PCR (data not shown) confirmed the authenticity ofa single integration of pAUL-CTC in the chromosome of L.monocytogenes strain LO28. The stability of the pAUL-A insertionwas confirmed by PCR analysis of cultures grown in BHI mediumwith or without erythromycin selection at 37°C.

RNA Isolation and Analysis

RNA samples were prepared from 10 ml of mid-logarithmic-phase cells (OD600, 0.5) grown in BHI medium or IMM, 30 minafter the addition (or not) of NaCl (3.5% in IMM and 5.5% in BHImedium). After the shock, cells were harvested by centrifugation at7,000 x g for 4 min at 4°C. Cells were treated with rifampin andRNAprotect Bacteria reagent (Qiagen) and centrifuged at 7,000 xg for 10 min at 4°C. The pellet was resuspended in 400 µl of bufferS (10% glucose, 12.5 mM Tris [pH 7.6], 66 mM EDTA, 0.5% sodiumdodecyl sulfate) containing 10 mg of lysozyme ml-1, 200 µg ofproteinase K ml-1, and 200 µg of rifampin ml-1. After the additionof glass beads, cells were subjected to mechanical disruption. RNAwas purified by a Trizol extraction, two chloroform-isoamyl alcoholextractions, and an isopropanol precipitation. The final RNA pellet


was dissolved in water, treated with DNase, and quantifiedspectrophotometrically before storage at -70°C. The integrity andthe relative concentrations of RNA samples were checked byagarose gel electrophoresis and ethidium bromide staining.

The mRNA of ctc was semiquantitatively analyzed by a slotblot technique as follows. RNA samples were treated at 65°C in20% formaldehyde for 15 min. The samples were then vacuumblotted with a Bio-Rad slot blot apparatus onto positively chargednylon membranes. RNA was cross-linked to the membrane by UVradiation. Transcription of ctc was monitored using an intragenicdigoxigenin (DIG)-labeled probe generated by PCR with primersOD3 and OD4. Detection of the labeled probe was mediated byaddition of an anti-DIG alkaline phosphatase-conjugated enzymeand CDP-star substrate (Roche Molecular Biochemicals). Lightemission was captured by standard autoradiography (Hyperfilm;Amersham Biosciences). A 16S rRNA probe was used to controlthe loading uniformity of RNA extracted under different conditionsof stress.

Electron Microscopy

Stationary-phase-grown bacteria were processed forobservation by scanning electron microscopy (SEM). Bacteria werefixed for 1 h with 4% glutaraldehyde in a 0.2 M cacodylate buffer.Cells were dehydrated using a graded ethanol series (70, 95, and100% ethanol, three times for 10 min each time) and subjected toan acetone dehydration series of 30, 50, 70, and 100% acetone for10 min each. One drop was spread on a microcover, coated withgold in an Emscope SC500, and observed with the Philips SEM505.

RESULTS

Sequence Analysis of the Ctc Gene of L. Monocytogenes LO28

Primers designed by using the sequences surrounding the ctcgene (lmo00211) of L. monocytogenes strain EGDe, whose genomehas been entirely sequenced, were used to amplify the ctc gene ofstrain LO28 by PCR. The PCR yielded one band, and the nucleotidesequence of this DNA fragment revealed 100% identity betweenthe ctc sequences of strains LO28 and EGDe.

The ctc gene starts at an ATG codon, 9 nucleotides downstreamof a potential ribosome binding site (5' GGAG 3'), and ends witha TAA stop codon. The deduced polypeptide contains 207 residueswith a calculated molecular weight of 22,654 (pI 4.14). Only oneamino acid differs between the Ctc sequences of L. monocytogenesLO28 and Listeria innocua CLIP 11262 (Lys201® Asn substitutionin L. innocua Ctc protein). Homology searches revealed a significantdegree of similarity with the Ctc protein of B. subtilis (42% identity),the equivalent protein, TL5, from Thermus thermophilus (36%identity), and all members of the Ctc protein family identifiedduring different genome-sequencing projects. Like those of otherCtc proteins, the N-terminal part of the L. monocytogenes Ctc proteinpresents similarities with members of the 50S ribosomal proteinL25 family, for instance, 28% identity with the L25 protein of E.coli, RplY.

A putative sB-dependent promoter was found 45 bp upstreamfrom the ATG start codon (5' GTTT-N15-GGGTAG 3') based on acomparison with the sB-dependent consensus promoter (5'GTTT[15/16 nt]GGGTAA3'). A stem-and-loop structure(ATGGAAGATTCGCA TTTGTT TGCGATATCTTCCAT; DG = -77kJ mol-1) followed by a T track is located 16 bp downstream fromthe TAA stop codon of the ctc gene. This palindromic structure isa putative transcriptional terminator. Analysis of the regionssurrounding the ctc gene in strain EGDe revealed an upstreamgene (lmo00210) transcribed in the opposite direction and encodinga putative lactate dehydrogenase (67% identity with Lactobacilluscasei lactate dehydrogenase). A gene (lmo00212) encoding a proteinwith no significant similarities with any of the proteins recordedin the databases is located downstream of the ctc gene and in thesame direction of transcription. This gene is separated from the ctcgene by the putative terminator mentioned above.

Transcriptional Analysis of the Ctc Gene

In a previous study, the Ctc protein was identified as a proteinshowing high induction after salt stress. Moreover, as describedabove, we found a putative sB promoter in front of the ctc gene, andB transcription is strongly induced after an osmotic upshift.Therefore, semiquantitative analysis of the ctc gene transcription


was carried out using slot blots with RNA extracted from culturesof strain LO28 grown in rich or minimal medium before or afteran osmotic shock. Similar results were obtained in both media.Under growth conditions without added NaCl, constitutiveexpression of ctc was observed. Thirty minutes of exposure to NaCl(3.5% in IMM and 5.5% in BHI medium [final concentration, 6%])resulted in an increase in the level of transcription.

Role of Ctc in Osmotic and High-temperature Adaptation

To investigate the function of the ctc gene of L. monocytogenesLO28, it was mutated by insertional mutation by using the pAUL-A plasmid, a temperature-sensitive suicide vector, as described inMaterials and Methods. Phenotypic analysis did not show anydifference between the ctc mutant and the wild-type strain withrespect to the aspect of colonies, catalase, hemolysis on blood agar,and metabolic profiles, as determined by using API-CH50microplate assays.

The stress tolerance of the ctc mutant was compared with thatof strain LO28. Because the Ctc protein was primarily identifiedas a protein induced by an osmotic upshift, we first examinedwhether the absence of Ctc would have any effect on the abilityof L. monocytogenes to grow under conditions of high osmoticstrength in a rich (BHI) or minimal (IMM) medium. Growth rates(doubling times) were found identical for the wild-type strain andthe ctc mutant in BHI medium (55 ± 5 min). No significant differencewas found between the growth rates of the two strains after theaddition of NaCl to BHI medium (85 ± 10 min for the wild-typestrain and 110 ± 20 min for the ctc mutant) or in IMM withoutadded NaCl (100 ± 15 min for the wild-type strain and 125 ± 15min for the ctc mutant). However, in IMM supplemented with 3.5%NaCl, the growth of the ctc mutant was significantly impaired.

The growth rate reached 620 ± 140 min, whereas it was 270± 15 min for the wild-type strain. In order to test if the sensitivityof the ctc mutant was linked to salt stress or osmotic stress, growthwas performed in IMM supplemented with 0.6 M xylose. Anaverage increase of 95% for the growth rate of the ctc mutant wasobtained. The ability of L. monocytogenes to survive high saltconcentrations is attributed mainly to the accumulation of

compatible solutes such as glycine betaine or carnitine. In order totest if the ctc mutation still had an effect in the presence ofosmoprotectants, we added 1 mM glycine betaine or carnitine toIMM supplemented with 3.5% NaCl. Addition of glycine betainenearly allowed strain LO28 and the ctc mutant to recover thegrowth rate of strain LO28 cultivated in IMM without NaCl. Theeffect of the addition of carnitine was intermediate, but nosignificant difference between the growth rates of the two strainswas observed in this situation.

In order to test if the Ctc protein was involved in general stresstolerance, the growth of the ctc mutant versus LO28 was measuredunder heat stress conditions at 45°C in BHI medium and IMM.Under these conditions, the growth of the mutant was similar tothat of the wild-type strain. This suggests that the role of Ctc instress tolerance is restricted to osmotic tolerance.

The Morphology of the Ctc Mutant is Impaired during StationaryPhase in Minimal Medium Containing NaCl

The wild-type strain LO28 and the ctc mutant weresubsequently examined using photonic microscopy during theexponential and stationary phases of growth at 37°C in BHI mediumand IMM with or without NaCl. No difference in morphology wasobserved between the two strains during growth in BHI mediumwith or without NaCl or in IMM without NaCl (data not shown).This was confirmed by transmission electron microscopy (TEM)using negative staining and by SEM by examining 48-h-growncultures (data not shown). In these three different media, bacteriaappeared as small rods measuring 1 to 2 µm in length. In contrast,in IMM containing 3.5% NaCl, the ctc mutant displayed a differentmorphology than the wild-type strain as observed by photonicmicroscopy, TEM (data not shown), and SEM. After 48 h of growthin this medium, the morphology of strain LO28 was characterizedby a rod shape with a variable size ranging between 1 and 4 µm.Approximately 1% of the cells displayed a bent rod shape. The ctcmutant also displayed a rod shape with a variable size rangingbetween 1 and 6 µm, but 80% of the cells had a bent or twisted rodshape. The morphology of the ctc mutant did not differ significantlyfrom that of the LO28 strain when 1 mM glycine betaine wasadded to the medium (data not shown).


We have shown that the ctc gene is involved in the resistanceof L. monocytogenes to high osmolarity in the absence ofosmoprotectants such as glycine betaine and carnitine in themedium. Thus, a ctc insertional mutant grew twice as slowly as thewild-type strain LO28 under conditions of high osmolarity (0.6 MNaCl or xylose) in minimal medium.

Moreover, the morphology of the ctc mutant was impaired inthis growth condition. Whereas the morphology of the wild-typestrain LO28 was characterized by a rod shape, the ctc mutantmorphology was characterized by a bent or twisted rod shapeunder osmotic stress conditions. When glycine betaine or carnitine,known to be the most efficient osmoprotectants in L. monocytogenes,was added to this medium, the growth of the mutant becameidentical to the growth of its isogenic parent strain. This canexplain why the growth rates of the mutant and the wild-typestrain were identical in rich medium (BHI) supplemented with5.5% NaCl. The BHI medium contains carnitine, which is relativelyabundant in some mammalian tissues. The role of ctc in stresstolerance seems to be restricted to osmotic stress tolerance, since nodifference between the wild-type strain and the ctc mutant wasobserved under conditions of growth at high temperatures in richor minimal medium.

Few genes involved in salt stress tolerance have been identifiedin L. monocytogenes until now. Survival of L. monocytogenes at highsalt concentrations is attributed mainly to the accumulation ofthree osmoprotectants, glycine betaine, carnitine, and proline.Independently of genes involved in the transport or biosynthesisof osmoprotectants, two genes encoding proteins of the Clp familyhave been identified, clpC and clpP. Inactivation of these genesconfers a general stress sensitivity phenotype, including salt stresssensitivity, on the corresponding mutants. These genes are knownto encode general stress proteins, chaperones assisting the properfolding, refolding, or assembly of proteins and proteases processingthose that cannot be refolded. A recent study identified relA, a geneencoding a (p)ppGpp synthetase, as a gene involved inosmotolerance. The authors showed that (p)ppGpp is involved inthe growth of L. monocytogenes under high osmotic pressure andthat the intracellular accumulation of (p)ppGpp is probably

controlled by mechanisms distinct from accumulation of compatiblesolutes. The last gene which has clearly been associated withosmotolerance in L. monocytogenes is sB. The absence of sB impairedthe ability of L. monocytogenes to use glycine betaine or carnitine asan osmoprotectant and impaired the transport of glycine betaine.

The transport of carnitine has not been studied. A potential sB-dependent promoter has been identified upstream of the betL geneand the opuC operon. This suggests that sB plays a key role inosmotolerance of L. monocytogenes via regulation of the expressionof two major osmoprotectant transport systems. We have identifieda putative sB-dependent promoter upstream of the ctc gene. Incontrast to that of B. subtilis, the L. monocytogenes ctc gene does notseem to belong to an operon. Moreover, expression of the ctc geneis strongly induced by an osmotic upshift, like that of the sB gene.Taken together, these observations suggest that the ctc gene may beregulated, at least in part, at the transcriptional level by sB. Thisemphasizes the role of sB, which is probably a key regulator ofosmotolerance in L. monocytogenes in the presence or absence ofcompatible solutes in the environment.

Currently, the function of the Ctc proteins is unknown. Ourresults suggest that the Ctc protein of L. monocytogenes belongs toa novel system utilized by this bacterium to adapt to an osmoticupshift in the absence of an osmoprotectant. This is the first timethat a role has been assigned to the Ctc protein, whose gene iswidely distributed in bacterial genomes. The product of the ctcgene of L. monocytogenes, like other ctc gene products, presentssimilarities in its N-terminal part with the 50S ribosomal L25protein and consequently belongs to the L25 ribosomal proteinfamily. According to the COG database, which compares the proteinsequences encoded in 43 complete genomes, representing 30 majorphylogenetic lineages, no L25 homologue is present in the archaealgenomes, but L25 homologues are present in nearly all eubacterialgenomes.

L25 homologues are found in all gram-negative bacteria andin all gram-positive bacteria except Lactococcus lactis, Streptococcuspyogenes, Mycoplasma pneumoniae, and Mycoplasma genitalium. Thesequence homologies observed between the Ctc proteins and theL25 proteins include many conserved residues, which the 5S rRNA-

Introductory Food Microbiology56 Identification of Listeria Monocytogenes Genes 57

L25 structure confirmed to be involved in the rRNA-proteinbinding interaction, thereby confirming that these two groups ofproteins are strongly related. It is highly probable that the Ctcproteins are associated at least in their N-terminal parts with theribosome and bind the 5S rRNA. Ribosomes have been implicatedas sensors of heat and cold shock in E. coli. Recent results implicatedthe ribosome as a possible mediator of the activity of Obg, anessential GTP-binding protein, and the stress induction of sB,suggesting that ribosomes are also general sensors in B. subtilis. Wecan hypothesize that the ribosome is also a sensor of salt stress,at least in L. monocytogenes, through the activity of Ctc. Furtherinvestigations will be required to clarify the function of Ctc in L.monocytogenes and in other bacteria.

3IDENTIFICATION OF LISTERIA

MONOCYTOGENES GENES

The capacity of Listeria monocytogenes to tolerate salt andalkaline stresses is of particular importance, as this pathogen isoften exposed to such environments during food processing andfood preservation. We screened a library of Tn917-lacZ insertionalmutants in order to identify genes involved in salt and/or alkalinetolerance. We isolated six mutants sensitive to salt stress and 12mutants sensitive to salt and alkaline stresses. The position of theinsertion of the transposon was located in 15 of these mutants. Insix mutants the transposon was inserted in intergenic regions, andin nine mutants it was inserted in genes. Most of the genes haveunknown functions, but sequence comparisons indicated that theyencode putative transporters.

INTRODUCTION

Listeria monocytogenes is a food-borne pathogen that is widelydistributed in the environment. This microorganism is of particularconcern in the food industry because of its ability to survive, andfrequently to grow, under a wide range of adverse conditions usedto preserve food, such as low temperature, low pH, and highosmolarity, or used to clean and sanitize equipment, such as highpH. Growth of L. monocytogenes has been reported at NaClconcentrations as high as 10% and at pHs as high as 9.

There is little information on the mechanisms that allow thisbacterium to cope with alkaline environments. Knowledgeconcerning the mechanisms used by gram-positive bacteria foradaptation and growth at alkaline pHs comes mainly from studies

Introductory Food Microbiology Identification of Listeria Monocytogenes Genes58 59

of alkaliphilic strains of Bacillus species, such as Bacillus haloduransC-125 or Bacillus pseudofirmus OF4. There is strong evidence thatmonovalent cation-proton antiporters are essential for maintaininga neutral cytoplasmic pH and, therefore, for growth under alkalineconditions. In addition, the acidic cell wall polymers teichuronicacid and teichuronopeptides contribute to pH homeostasis. Thesewall macromolecules may provide a passive barrier to ion flux andelevation of the cytoplasmic buffering capacity at highly alkalinegrowth pHs. In Bacillus subtilis, monovalent cation/protonantiporters also seem to be important since the mrpA gene encodingan Na+/H+ antiporter and the tetA(L) gene encoding amultifunctional tetracycline-metal/H+ antiporter that also exhibitsmonovalent cation/H+ antiport activity are involved in Na+-dependent pH homeostasis.

Most bacteria cope with elevated osmolarity in the environmentby intracellular accumulation of particular osmolytes calledcompatible solutes. These compatible solutes act in the cytosol bycounterbalancing the external osmolarity, thus preventing waterloss from the cell and plasmolysis without adversely affectingmacromolecular structure and function. Compatible solutes can beeither transported into the cell or synthesized de novo. Survival ofL. monocytogenes at high salt concentrations is attributed mainly tothe accumulation of three compatible solutes: glycine betaine,carnitine, and proline. Accumulation of glycine betaine andcarnitine occurs via two glycine betaine transporters encoded bythe betL gene and the gbu operon and one carnitine transporterencoded by the opuC operon. Disruption of these genes reducedthe osmotolerance of L. monocytogenes. Both betL and opuC haveputative sB-dependent promoters. The absence of sB impaired theability of L. monocytogenes to use glycine betaine or carnitine as acompatible solute.

Proline transport has not been characterized yet. However,disruption of the proBA operon (proline biosynthesis-encodingoperon) reduced the growth of the corresponding mutant at highsalt concentrations. Little information is available concerning othermechanisms that L. monocytogenes uses to cope with salt stress,especially when compatible solutes are not available in theenvironment. Two genes, clpC and clpP, encoding a ClpC ATPase

and a ClpP serine protease, respectively, have been identified.Inactivation of these genes conferred a general stress sensitivityphenotype, including sensitivity to salt stress, to the correspondingmutant. In a recent study workers identified relA, a gene encodinga (p)ppGpp synthetase, as a gene involved in osmotolerance viaa mechanism different from the mechanism involving accumulationof compatible solutes. It has also been shown that the generalstress protein Ctc of L. monocytogenes is involved in osmotolerancein the absence of any compatible solutes in the environment.

In order to obtain a better understanding of the mechanismsinvolved in salt and alkaline tolerance, we used a library oftransposon insertional mutants of L. monocytogenes LO28 to isolatemutants with decreased NaCl and/or alkaline tolerance. Wesucceeded in identifying different mutants that exhibit lessresistance to salt and/or alkaline stress than the parental strainand characterized the genes interrupted.


Bacterial Strains

The L. monocytogenes strains used were LO28, a clinical isolateof serotype 1/2C, and a library of Tn917-lacZ mutants of strainLO28. Bacterial plasmids were propagated in Escherichia coli strainTG1.

Culture Media and Stress Conditions

Cells were grown on complex culture media, including brainheat infusion (BHI) broth or agar (Difco Laboratories, Detroit, Mich.).Screening of the library of Tn917-lacZ mutants for sensitivity tosalt and alkaline stresses was performed as follows. Wells ofmicroplates containing 100 µl of BHI medium with erythromycin(BHI-erm) were inoculated with the different mutants. Themicroplates were incubated at 37°C overnight and subsequentlyused to transfer the mutants, after 1:1 dilution with BHI-erm, ontoagar plates with a replicator. The plates used contained BHI-ermwith or without 5.5% NaCl (final concentration, 6%) and BHI-ermadjusted to pH 8.6 with NaOH. The growth of the mutants wasrecorded after 48 h. Mutants selected after this first step were usedto perform liquid growth experiments with a Microbiology Reader


Bioscreen C (Labsystems, Helsinki, Finland) in 100-well sterilemicroplates, and each well contained 300 µl of culture medium, asfollows.

Overnight cultures of Tn917-lacZ mutants in BHI-erm wereused to inoculate different media (BHI-erm with or without 5.5%NaCl and BHI-erm with the pH adjusted to 8.5) at an initial opticaldensity at 600 nm of ~0.1. The cultures were incubated with shakingat 37°C. The optical density was monitored at 600 nm. Experimentswere repeated independently at least twice. The LO28 strain wasused as a control and was inoculated into BHI medium lackingerythromycin. For more detailed physiological characterization ofmutants sal5 and sal11, liquid growth experiments were performedby using the same procedure except that the pH of BHI mediumwas adjusted to 8.5 with a glycine-NaOH-NaCl buffer. Growthexperiments were also performed in BHI medium supplementedwith 7% KCl or 15% xylose and were repeated independently atleast four times. Growth curves were fitted by using a modifiedGompertz equation, and the generation time was calculated byusing nonlinear regression with the Statistica statistical software(Statsoft, Tulsa, Okla.).

Antibiotics were used at the following concentrations: 100 µgof ampicillin ml-1 for E. coli and 5 µg of erythromycin ml-1 for L.monocytogenes.

Identification of Transposition Target

Inverse PCR was used to amplify the DNA fragment next tothe region downstream from the Tn917-lacZ chromosomalinsertions. Bacterial chromosomal DNA was isolated as describedpreviously. Chromosomal DNA was digested with HindIII formutants sal1, -2, -5, -6, -11, -17, -22, and -23 and mutants sl7, -10,-13, -14, and -25 and with NdeI for mutants sl12 and sal21 and wassubsequently circularized by self-ligation. The region downstreamfrom the Tn917-lacZ insertion was amplified by using primersRG7 (5'-ATTCCGTCTGAAGCAGTGGT-3') and RG9 (5'-GAACGCCGTCTACTTACAAG-3') for HindIII-digested DNA andprimers RG9 and RG11 (5'-GAATCACGTGTCCCTTTGCG-3') forNdeI-digested DNA. Amplification products were sequenced eitherdirectly or after they were cloned into the pGEM-T plasmid

(Promega France, Charbonnières, France), a 3' T-end vectorspecifically designed for cloning PCR fragments, by following themanufacturer’s specifications. DNA sequencing was done with aBigDye terminator cycle sequencing Ready Reaction kit (AppliedBiosystems, Courtaboeuf, France). The reactions were performedwith unlabeled primers and fluorescent dideoxynucleotides, andthen the reaction mixtures were analyzed with an automatic DNAsequencer (ABI Prism 310 genetic sequencer; Applied Biosystems).Blast sequence homology analyses were performed by using theNational Center for Biotechnology Information network service.The primers used for the sequence were RG1 (5'-CCCACTAAGCGCTCGGG-3'), RG7, and RG9. Oligonucleotideswere synthesized by MGW-Biotech (Courtaboeuf, France).

RESULTS

Selection of Salt and Alkaline Stress-sensitive Mutants

A library of approximately 2,500 Tn917-lacZ insertion mutantswas screened for salt and/or alkaline stress sensitivity. Each mutantwas grown on BHI-erm plates with or without 5.5% NaCl or withthe pH adjusted to 8.6. Twenty-three mutants showed a growthdelay on at least one of the two stress media. Six mutants seemedto be affected under salt stress conditions, nine mutants seemed tobe affected under alkaline stress conditions, and eight mutantsseemed to be affected under both conditions. The phenotypes ofthe mutants were further confirmed in liquid growth experimentsby comparing the growth curves to that of the LO28 wild-typestrain. Five mutants were removed after this second step. Thegrowth of two mutants sensitive to both stresses was impaired inBHI medium, and the alkaline sensitivity phenotype of threemutants was not confirmed. Finally, we isolated six mutantssensitive to salt stress and 12 mutants sensitive to both salt stressand alkaline stress. The designations of mutants that were sensitiveonly to salt stress begin with sl (for salt sensitivity locus), and thedesignations of mutants that were that were sensitive to salt stressand alkaline stress begin with sal (for salt and alkaline sensitivitylocus). Southern hybridization with HindIII-digested chromosomalDNA and, when required, EcoRI-digested chromosomal DNA witha digoxigenin-labeled DNA probe specific for the Tn917-lacZ


transposon revealed that the 18 mutants contained a single copyof the transposon and that the loci corresponded to 18 independentinsertion loci (data not shown). The 18 remaining mutants werekept for further characterization.

Identification of the Transposition Target

The inverse PCR method enabled us to clone and sequence thedownstream transposon-chromosome junctions of 15 mutants. Wedid not succeed in cloning the junctions of three mutants, sal3, -4, and -19. The sequences were compared with the completegenome sequence of L. monocytogenes strain EGDe. In nine mutants,the transposon was inserted into open reading frames, whereas itwas inserted into intergenic regions in six mutants.

Genes Encoding Proteins with an Identified Function or a PutativeFunction

In the sl12 mutant, the transposon was inserted into the mutSgene, which is involved in DNA mismatch repair, whereas inmutants sal1 and -11 and sl14 it was inserted into transporters orputative transporters. In sal1, the transposon was inserted into theATPase subunit of an ATP binding cassette transporter, and insal11 it was inserted into the permease subunit of another ATPbinding cassette transporter. In sl14, the transposon was insertedinto the mdrL gene, which encodes a multidrug efflux transporterof the major facilitator superfamily. If the orientation of the genesand the presence of putative terminators were taken into account,the phenotypes of mutants sal1 and -11 and sl14 could not belinked to a polar effect of the mutation on downstream genes.However, the phenotype of mutant sl12 could be linked to a polareffect of the mutation of a downstream gene, mutL coding for aDNA mismatch repair protein or lmo1405 coding for a putativeantiterminator regulatory protein.

Genes Encoding Proteins with Unknown Functions

In sl7 and sal22, the transposon was inserted into the lmo1432gene at positions separated by 29 bp. The deduced protein encodedby the lmo1432 gene is specific to Listeria. In sal2 and -17, thetransposon was inserted into two genes, lmo1443 and lmo2232,respectively, which encode proteins having unknown functions in

other organisms. lmo1443 has orthologues in different gram-positive bacteria, and lmo2232 has a paralogue (lmo2399) and fiveorthologues in B. subtilis yhdP, yrkA, yqhB, yugS, and yhdT and inother eubacteria. Considering the orientation of the genes and thepresence of putative terminators, the phenotypes of these fourmutants could not be linked to a polar effect of the mutation ondownstream genes. This was not the case for sal5, in which thetransposon was inserted into lmo0992, a gene of unknown functionlocated upstream from lmo0991 which encodes a protein withsimilarities to NtpJ of Enterococcus hirae, a K+/Na+ transporter.

In these five mutants, similarity searches could not assign aputative function to the genes interrupted, but a search fortransmembrane domains with the DAS program revealed that allcorresponding proteins have putative transmembrane domains.

Intergenic Regions

For six mutants, the position of the insertion of the transposonwas located in intergenic regions.

The physiology of two mutants, sal5 and sal11, which wereidentified as sensitive to salt and alkaline stresses during thescreening analysis, was characterized by quantifying the growthof these organisms in different media.

Physiology of the sal5 and sal11 Mutants in Response to Osmoticand Alkaline Stresses

Growth of the sal5 and sal11 mutants was examined by usingBHI medium, BHI medium supplemented with 5.5% NaCl, 7%KCl, or 15% xylose, and BHI medium with the pH adjusted to 8.5.For the growth experiments under osmotic stress conditions, theconcentration of each of the three solutes (NaCl, KCl, and xylose)was approximately 1 M. The results showed that the phenotypeidentified during the screening analysis was confirmed. Bothmutants were sensitive to NaCl stress and alkaline stress. However,the sensitivity was far more pronounced with the alkaline stressconditions; under these conditions the growth rate (doubling time)of both mutants was approximately 300 min and was fivefoldgreater than the growth rate of the wild-type strain. Under NaClosmotic stress conditions, the growth rates of mutants sal5 and


sal11 were 1.4- and 1.3-fold greater, respectively. Moreover, bothmutants were also sensitive to KCl stress (but to a lesser extent)and slightly sensitive to xylose stress.

DISCUSSION

We isolated six Tn917-lacZ insertional mutants of L.monocytogenes strain LO28 sensitive to salt stress (sl mutants) andnine mutants sensitive to both salt and alkaline stresses (salmutants) and located the positions of the insertions of thetransposons in the genomes of the different mutants. We used thecomplete genome sequence of L. monocytogenes strain EGDe torapidly identify the positions of the insertions of the transposons.We sequenced 15 fragments corresponding to a total of 6,249nucleotides and found only 4 nucleotides which differed in strainsLO28 and EGDe. These results suggest that the DNA sequences ofthese two strains are very similar and justify utilization of thecomplete genome sequence of L. monocytogenes strain EGDe tostudy the L. monocytogenes strain LO28 genome.

On the basis of the functions or putative functions of the genesdisturbed by insertion of the transposons, we classified the mutantsin four categories.

Genes with known Functions Directly Linked to Salt Stress

In mutants sal6 and sl10, the transposon was inserted in frontof the gbu operon. This operon, which is very similar to the opuAoperon of B. subtilis, encodes an ATP-dependent transporterbelonging to the ATP binding cassette transporter superfamily andis involved in the transport of glycine betaine. A mutant with thisoperon inactivated was found to be salt sensitive. Ko and Smithalso identified sequences with significant similarities to the sA-type -35 and -10 promoter recognition sequence TTGTGT-N15-TATTGC. In the sl10 mutant, the transposon was located betweenthe putative promoter and the coding DNA sequence of the gbuoperon. In this mutant, the promoter of the gbu operon is separatedfrom the gbu operon by the transposon.

Consequently, the gbu operon cannot be transcribed. The saltstress sensitivity of the sl10 mutant confirmed the results of Ko andSmith. This result provides good validation of our screening

protocol. In sal6, the transposon was inserted 109 bp upstreamfrom the putative promoter of the gbu operon. This result indicatesthat regions upstream from the putative promoter are probablyinvolved in the transcription of this operon. In B. subtilis, twopromoters have been identified in front of the opuA operon, and wehypothesize that this is also the case for L. monocytogenes.Surprisingly, the sal6 mutant was found to be sensitive to salt andalkaline stresses. However, the alkaline phenotype is weak. Usingtwo-dimensional electrophoresis, we identified the GbuA protein,the ATPase subunit of the Gbu transporter, as a protein inducedby salt stress. In L. monocytogenes, expression of the sB factor isstrongly induced by salt stress, and consequently, expression ofthe genes under the control of this sigma factor is also induced bysalt stress. However, no sB promoter recognition sequences arepresent in front of the gbu operon, indicating that another saltinduction pathway is present in L. monocytogenes.

Genes with known Functions not Directly Linked to Salt Stress

In mutants sl12 and sl14 the transposon was inserted into themutS and mdrL genes, respectively. The mutSL locus of L.monocytogenes is involved in both mismatch repair and homologousrecombination. The mdrL gene encodes a multidrug effluxtransporter and is involved in the efflux of ethidium bromide. Inboth cases, no salt stress sensitivity of the mdrL or mutSL mutanthas been mentioned previously; thus, this is the first time thatthese two loci have been associated with salt stress. MdrL extrudesdifferent toxic components, and we hypothesize that this transporteralso extrudes Na+, perhaps in a nonspecific manner.

GENES ENCODING PUTATIVE TRANSPORTERS

In mutants sal1 and sal11, even though the function is unknown,sequence comparisons indicated a putative function. In bothmutants, the transposons were inserted into genes encodingdomains of two distinct ATP binding cassette transporters. In sal1,the gene interrupted, ykpA, encodes the ATP binding proteindomain of the transporter, and in sal11 the gene interrupted, lmo668,encodes the permease protein domain. ykpA has orthologues ingram-positive bacteria. The corresponding protein has beenidentified in Staphylococcus aureus as an immunodominant antigen,


and antisense ablation of the ykpA gene led to a growth-inhibitingeffect.

The lmo668 gene is located downstream from its putativeassociated ATP binding protein. Lmo668 has very significantsimilarities with Yadh of gram-negative bacteria. In these cases itis highly probable that ykpA and lmo0668 encode transporters, butthe transported solutes remain unknown. The transportersprobably extrude Na+ and/or H+, perhaps in a nonspecific manner.Growth experiments with mutant sal11 indicated that the lmo0668gene is probably involved in pH homeostasis because the sal11mutant was highly sensitive to alkaline pH, moderately sensitiveto NaCl stress, and slightly sensitive to KCl stress and xylosestress. The differences observed among the effects of the threeosmolytes, NaCl, KCl, and xylose, which were added atconcentrations of approximately 1 M to the medium, were probablydue to the fact that NaCl tends to be more stressful than KCl andthe fact that xylose provides only one-half the osmotic stress ofNaCl and KCl.

The last mutant in which expression of genes encoding putativetransporters are disturbed is mutant sal5. lmo0992, the geneinterrupted by the transposon, belongs to an operon consisting offive genes. The first gene, lmo0989, encodes a putativetranscriptional regulator of the MarR family. The next three genes,lmo0990, lmo0991, and lmo0992, encode proteins whose functionsare unknown but which are putative integral membrane proteins.Lmo0991 and Lmo0992 are very similar to YkoY of Bacillusmegaterium, B. subtilis, and Bacillus anthracis. The last gene, lmo0993,encodes a protein with similarity to the NtpJ protein of E. hirae(33% identity). The ntpJ gene encodes a component of the KtrIIuptake system. NtpJ mediates K+ and Na+ cotransport.

The growth of an ntpJ-disrupted mutant is impaired at pH 10in K+-limited medium. The phenotypes of an ntpJ-disrupted mutantand the sal5 mutant are similar; the sal5 mutant is highly sensitiveto alkaline pH in BHI medium. Thus, the phenotype of the sal5mutant is probably due to a polar effect of the mutation on lmo0993,the ntpJ-like gene. In E. hirae, the ntpJ gene is the last gene of anoperon (ntp) encoding a V-type Na+-ATPase, which is importantfor Na+ extrusion at high pH. The NtpJ K+/Na+ uptake system

functions with the V-type Na+-ATPase at high pH and/or highNa+ concentrations in order to eliminate sodium ions and to drivepotassium ion uptake. We found no similarities between thesequences of lmo0990, lmo991, and lmo992 and the sequences ofthe ntp genes encoding the V-type Na+-ATPase, and further workis needed to check if there is an analogy of function among thegenes.

Genes with Unknown Functions

For mutants sal2, sl7, sal22, and sal17, the function of the geneinterrupted is completely unknown, but the three genes interruptedencode proteins with putative transmembrane domains. Thetransposon was inserted into the same gene, lmo1432, in mutantssl7 and sal22. This interrupted gene seems to be specific to thegenus Listeria. In sal2 the interrupted gene seems to be specific togram-positive bacteria, whereas in sal17 it seems to be specific toeubacteria. In mutants sal21 and sl25 the transposon was insertedin front of putative operons encoding proteins with unknownfunctions.

We were not able to identify sA-type or sB-type promoterrecognition sequences in front of these operons. However, wehypothesize that in these two cases transcription of the operonswas eliminated by insertion of the transposon in the promoterregion and that the phenotypes of the mutants are linked to at leastone of the genes of the operons.

Finally, the salt sensitivity phenotype of mutant sl13 and thesalt and alkaline sensitivity phenotype of mutant sal23 are difficultto explain because in sl13 the transposon was inserted betweentwo convergent genes and in sal23 it was inserted between the endand the terminator of the lmo1140 gene.

We did not identify the relA gene during our screeningprocedure. This gene encodes a (p)ppGpp synthase and was recentlyidentified during a screening procedure very similar to ourprocedure. However, this is not surprising because a library oftransposons cannot be exhaustive. Moreover, we did not succeedin identifying the positions of insertion of the transposons in threemutants, and we cannot exclude the possibility that one of thesethree mutants corresponds to a relA mutant.


We were interested in studying salt stress because it is one ofthe most commonly used methods for food conservation, and wewere interested in studying alkaline stress because of the alkalinenature of most of the detergents and some of the chemical sanitizersused to clean and sanitize equipment in food processing plants.Little information is available concerning the physiology of L.monocytogenes in response to alkaline stress. Growth of food factoryisolates was reported at pHs as high as 9, and this pathogen hasbeen shown to be resistant to storage at pHs up to 12. It has alsobeen shown that alkaline pH induces cross-protection of L.monocytogenes against heat. Moreover, there is no informationconcerning the mechanisms that take place in L. monocytogenes inorder to cope with alkaline stress. It was interesting to study thesetwo stresses in parallel because a combination of salt stress andalkaline stress is more effective in decreasing the survival of L.monocytogenes than an individual type of stress. It is also knownthat homeostasis of Na+ and H+ ions is tightly linked, and in mostcell membranes there are proteins that couple the fluxes of the twoions via the Na+/H+ antiporter or the Ntp complex, for example.During the screening procedure used in this study we identifiedfew genes involved in Na+ and/or alkaline stress tolerance. Furtherwork is needed to investigate the function of these genes, but thefact that they encode putative integral membrane proteins indicatesthat they probably encode ion transporters.

CELL-WALL PROTEINASES PRTS AND PRTB HAVE ADIFFERENT ROLE IN STREPTOCOCCUS THERMOPHILUS /LACTOBACILLUS BULGARICUS MIXED CULTURES IN MILK

The manufacture of yoghurt relies on the simultaneousutilization of two starters: Streptococcus thermophilus and Lactobacillusdelbrueckii subsp. bulgaricus (Lb. bulgaricus). A protocooperationusually takes place between the two species, which often resultsin enhanced milk acidification and aroma formation compared topure cultures. Cell-wall proteinases of Lactococcus lactis andlactobacilli have been shown to be essential to growth in milk inpure cultures. In this study, the role of proteinases PrtS from S.thermophilus and PrtB from Lb. bulgaricus in bacterial growth inmilk was evaluated; a negative mutant for the prtS gene of S.thermophilus CNRZ 385 was constructed for this purpose. Pure

cultures of S. thermophilus CNRZ 385 and its PrtS-negative mutantwere made in milk as well as mixed cultures of S. thermophilus andLb. bulgaricus: S. thermophilus CNRZ 385 or its PrtS-negative mutantwas associated with several strains of Lb. bulgaricus, including aPrtB-negative strain.

The pH and growth of bacterial populations of the resultingmixed cultures were followed, and the Lactobacillus strain wasfound to influence both the extent of the benefit of Lb. bulgaricus/S. thermophilus association on milk acidification and the magnitudeof S. thermophilus population dominance at the end of fermentation.In all mixed cultures, the sequential growth of S. thermophilus thenof Lb. bulgarius and finally of both bacteria was observed. Althoughproteinase PrtS was essential to S. thermophilus growth in milk inpure culture, it had no effect on bacterial growth and thus on thefinal pH of mixed cultures in the presence of PrtB. In contrast,proteinase PrtB was necessary for the growth of S. thermophilus,and its absence resulted in a higher final pH. From these results,a model of growth of both bacteria in mixed cultures in milk isproposed.

Streptococcus thermophilus is a thermophilic lactic acid bacterium(LAB), widely used as a starter to produce fermented dairyproducts. It is generally used in association with other micro-organisms, in particular with Lactobacillus delbrueckii subsp.bulgaricus (Lb. bulgaricus) for the manufacture of yoghurt. For thisapplication, the fast-growing capacity of these bacteria in milk iscrucial to enable intense and rapid acidification of milk. LAB arefastidious micro-organisms, which have in particular several aminoacid auxotrophies. Most S. thermophilus strains are stimulated bythe supply of two to five amino acids, whereas lactobacilli requirebetween three and 14 amino acids.

The optimal growth of LAB in milk thus depends on theirproteolytic system, which hydrolyses milk caseins into peptidesand amino acids. The cell-wall proteinases of LAB are of majorimportance in this process, as they are responsible for the first stepof casein breakdown. They belong to the same multi-domainproteinase family and show significant homologies, even thoughdifferences in specificity, bacterial anchor and domain organizationhave been described. The cell-wall proteinase of Lactococcus lactis


(PrtP), which is very frequent in this species, has been extensivelystudied. In milk, Lc. lactis PrtP-negative strains only reach 10% ofthe cell densities observed with PrtP-positive strains. In S.thermophilus, the presence of a cell-wall proteinase, PrtS, recentlycharacterized, is less common than in Lc. lactis. In this species,high cell-wall proteinase activities are associated with high milk-acidifying capacities. In Lb. bulgaricus, the cell-wall proteinase, PrtB,is also essential for optimal growth in milk; a proteinase-negativestrain reaches only 22% of the final biomass of a proteinase-positivestrain when grown in milk.

In yoghurt, S. thermophilus and Lb. bulgaricus are grown inassociation, which often results in a positive interaction. Thisrelationship, called protocooperation, has a beneficial effect ongrowth of both species and on acid and aroma production. S.thermophilus indeed produces pyruvic acid, formic acid and CO2,which stimulate the growth of Lb. bulgaricus. In turn, Lb. bulgaricusproduces peptides and amino acids that stimulate S. thermophilusgrowth.

In the present study, we wished to determine the role of thecell-wall proteinases PrtS from S. thermophilus and PrtB from Lb.bulgaricus in bacterial growth in milk. We therefore constructed anegative mutant for the cell-wall proteinase gene (prtS gene) of S.thermophilus CNRZ 385, which was recently sequenced andcharacterized in our laboratory. The latter mutant was used tostudy the role of PrtS on the growth of S. thermophilus in pureculture in milk. We also took advantage of the availability of aPrtB-negative mutant of Lb. bulgaricus to evaluate the role of cell-wall proteinases PrtS and PrtB on growth and thus on acidificationin S. thermophilus/Lb. bulgaricus mixed cultures.

METHODS

Strains of S. thermophilus and Lb. bulgaricus were grown inthree different media: reconstituted skim milk (Nilac Low HeatMilk powder, NIZO) heated for 10 min at 95 °C, supplementedwith yeast extract (3 g l-1; Difco) when required, M17 mediumsupplemented with 20 g lactose l-1, and MRS mediumsupplemented with 20 lactose g l-1 and acidified at pH 5·2,supplemented with streptomycin (Sigma) (2 mg ml-1) when required.

The Escherichia coli strain was grown at 37 °C on Luria-Bertani(Difco) medium with shaking, in the presence of erythromycin(Ery) (150 µg ml-1) when required.

Stock cultures of each strain of S. thermophilus and Lb. bulgaricuswere prepared after growth at 42 °C on skim milk, supplementedwith yeast extract for proteinase-negative strains, from overnightskim milk cultures, supplemented with yeast extract when required.The pH was then measured, and bacterial numbers were estimatedby plating, with an automatic spiral plater, appropriate dilutionsof the culture on agar medium: M17Lac was used for specificenumeration of S. thermophilus cells, and MRSLac pH 5·2,supplemented with streptomycin when required, for specificenumeration of Lb. bulgaricus cells.

For S. thermophilus strains and before dilution, chains of cellswere disrupted for 30 s in a mechanical blender. After 48 hincubation at 42 °C in anaerobic jars, cells were enumerated withthe EC1 colony counter software. At the end of culture, bacteriawere directly frozen in liquid nitrogen and kept at -80 °C.

Growth rates of S. thermophilus 385 and 385-PrtS strains weredetermined in M17 at 42 °C using a Microbiology Reader BioscreenC (Labsystems) in 100-well, sterile, covered microplates. Each well,containing 200 µl M17Lac, was inoculated at 1% with overnightM17Lac cultures of S. thermophilus and covered with one drop ofparaffin oil. The optical density was measured at 600 nm every20 min, after gentle shaking. The apparent growth rate (µmax) wasdefined as the maximum slope of semi-logarithmic representationof growth curves, assessed by optical density measurements.

Mixed cultures of S. thermophilus and Lb. bulgaricus strainswere performed at 42 °C by inoculating skim milk with 5x106

c.f.u. ml-1 of stock cultures of each strain. For proteinase-negativestrains, cells from stock culture were washed three times in 50 mMTris buffer (pH 7) before inoculation to avoid peptides and/oramino acids being supplied in the mixed culture. Every 20 min, thepH of the culture was measured, and bacteria were enumerated asdescribed above. Total bacterial populations were estimated byaddition of data from enumerations of each bacterial species onspecific medium to the others, as indicated above.


Proteinase Assay

The PrtS proteinase phenotype of S. thermophilus strains wasdetermined on bacterial colonies in two ways. First, bacteria weregrown on FSDA medium (Fast Slow Difference Agar) (Huggins &Sandine, 1984 ). This milk-based agar medium made it possible todifferentiate bacteria exhibiting slow or limited growth in milkfrom those exhibiting rapid growth; in particular, bacteriapossessing a cell-wall proteinase activity appeared as white,opaque, rounded colonies, whereas proteinase-negative colonieswere small, flat and translucent.

Second, bacteria from an overnight skim milk culture werediluted and plated on agar skim milk plates (cell culture dishes,35 mm in diameter). After 24-48 h incubation at 42 °C in anaerobicjars, colonies were covered by a solution containing Tris buffer(50 mM, pH 7), a chromogenic substrate of proteinase PrtS (Suc-Ala-Ala-Pro-Phe-ßNA, 10 mg ml-1; Novabiochem), 10 mg ml-1 Fast-Garnet (GBC, Sigma) and 10-50 mM CaCl2. PrtS-positive clonesappeared as red colonies, whereas PrtS-negative clones remainedwhite.

Proteinase activity was measured on cellular suspensionsusing [14C]casein as the substrate according to the method ofMonnet et al. (1987) , modified as follows. Cell suspensions wereprepared from 4 ml overnight M17 cultures; cells were recoveredby centrifugation (20 min, 8000 g, 4 °C) and washed three times inTris buffer (50 mM, pH 7). The last pellet was suspended in 150 µlBistris buffer (50 mM, pH 6·5) containing 10 mM CaCl2.

Fiftymicrolitres of cell suspension was incubated with 50 µl of 14Ccasein solution (0·1%) at 37 °C for 15, 60 and 120 min. Enzymereactions were stopped by the addition of 100 µl TCA (12%), leftfor 30 min at room temperature and centrifuged for 2 min at 10000 g,and the radioactivity was then measured in the supernatants.Protease activity corresponded to the percentage of caseinhydrolysis in 10 min.

DNA Manipulations and Sequencing

Total DNA Preparation

Total DNA of S. thermophilus CNRZ 385 was prepared asdescribed by Pospiech & Neumann (1995).

Preparation of Electrocompetent Cells of S. thermophilus and Lc.lactis

Electrocompetent cells of S. thermophilus CNRZ 385 and Lc.lactis MG1363 were prepared according to the method of Holo &Nes (1989) , modified as follows. From an overnight culture inM17Lac, a culture was performed at 37 °C (S. thermophilus) or at30 °C (Lc. lactis) by 1% inoculation of M17Lac containing DL-Thr(100 mM) for S. thermophilus or Gly (1·5%) for Lc. lactis until theOD600 reached 0·6-1. Cells were collected by centrifugation at 5000g for 5 min and washed four times in 0·5 M sucrose/10% glycerolsolution. They were then resuspended in 10% glycerol/30%PEG2000 solution for S. thermophilus or in 0·5 M sucrose/10%glycerol solution for Lc. lactis and immediately frozen in liquid N2and stored at -80 °C.

DNA Sequencing

The Sanger method of DNA sequencing was carried out ondouble-strand DNA plasmids and on PCR products with theBigDye Terminator cycle sequencing ready reaction kit (370A DNAsequencer, Applied Biosystems).

Construction of a Negative Mutant for PrtS

A 3776 bp PCR product containing part of the prtS gene wasamplified using oligonucleotides 1 (5' CAT CAC GGA AAG TCTAGG 3') and 2 (5' AAC GTA TTG ATA CTT ACC 3') from totalDNA of S. thermophilus CNRZ 385 strain. Streptococcal DNA(100 ng) was added to a PCR mixture containing 2·5 U of Taqpolymerase (Quantum Appligene) and 0·26 µM of eacholigonucleotide (Life Technology). After 5 min of denaturation at94 °C, 30 cycles of 30 s annealing at 50 °C and 3 min of elongationat 72 °C were carried out using a Perkin-Elmer DNA thermal cycler(model 480). The amplified fragment was purified from 0·7%agarose gel with the QIAquick gel-extraction kit (Qiagen). It wasthen ligated to pCR-XL-TOPO vector (Invitrogen) and cloned bytransformation of electrocompetent TOP10 E. coli cells (Invitrogen)according to the manufacturer’s protocol. The recombinant vector,pCR-XL-TOPO-DprtS-1, was purified with QIAprep Spin MiniprepKit (Qiagen) from the recombinant cells and digested with BsgI(New England Biolabs).


A 5·4 kb fragment containing the TopoXL vector and part ofthe prtS gene was then purified from 0·7% agarose gel withQIAquick gel extraction kit (Qiagen) and blunt-ended with T4polymerase 3'® 5' exonuclease (Life Technologies) according to thesupplier’s protocol. It was then circularized by self-ligation withFast-link DNA ligation kit (Epicentre Technology); the resultingplasmid, pCR®-XL-TOPO-DprtS-2, was produced bytransformation of electrocompetent TOP10 E. coli cells and purifiedas described above. It was then digested with NotI and SpeI(Eurogentec), and the resulting 2·078 kb fragment was purified asalready described above. The 2·078 kb fragment (~200 ng) wasligated to pGhost9 vector (1~00 ng) (Maguin et al., 1996 ), digestedwith NotI and SpeI. The ligation mix was used to electrotransform100 µl of electrocompetent cells of Lc. lactis MG1363, as describedby Holo & Nes (1989) . Recombinant clones were selected onM17Lac Ery plates after incubation at 28 °C. The recombinantvector, pG+h9::DprtS, was purified as described above, and 20 µgwas used to transform electrocompetent cells of S. thermophilusCNRZ 385, as previously described (Garault et al., 2000 ). Integrationof pG+h9::DprtS into the streptococcal chromosome was performedas described by Garault et al. (2000) with the following modification:to induce chromosomal integration of the plasmid, the culture wasdiluted and plated on M17Lac Ery plates.

Finally, the mutant for PrtS was obtained by successiveincubations of the culture containing the chromosomal integrationat 37 °C to favour the excision of the pGhost9 plasmid.

PrtS is Essential to S. thermophilus Growth in Milk

S. thermophilus PrtS-negative mutant construction. We havedescribed here for the first time the construction of a targetednegative mutant for S. thermophilus cell-wall proteinase PrtS. Thismutant of S. thermophilus CNRZ 385 was constructed by genereplacement using a truncated copy of prtS gene cloned in pGhost9plasmid. DNA sequencing confirmed that this copy was insertedat the prtS locus and that pGhost9 was subsequently excised,resulting in a truncated prtS gene. As expected, the truncated genewas deprived of part of the signal sequence, all the pro-region(removed after maturation of the protein in the parental strain),and almost all of the region encoding the catalytic domain of the

enzyme. Only the region encoding the six C-terminal amino acidsamong the 495 constituting the catalytic domain (PR domain) wasstill present in the mutant and did not include the sequenceencoding the residues involved in the catalytic activity of theproteinase. Furthermore, protein exportation signals were no longerpresent in the mutant; the signal sequence was truncated, and theexpected peptide cleavage site was excised.

Cell-wall proteinase activity of the wild-type and PrtS-negativemutant of S. thermophilus. Using two different methods, we checkedthat the S. thermophilus PrtS-negative mutant lacked cell-wallproteinase activity. First, using 14C-labelled casein as a substrate,we observed that cell suspensions of the PrtS-positive strain werecapable of hydrolysing casein (12·5% of total casein hydrolysedwithin 10 min), whereas PrtS-negative cells had no detectablecaseinolytic activity. Second, we set up a rapid test on coloniesusing a chromogenic substrate of PrtS. Three strains of S.thermophilus were used: the proteinase-negative strain CNRZ 302as negative control and the two proteinase-positive strains CNRZ385 and CNRZ 703, which have a high cell-wall proteinase activity.After growing on milk agar plates, colonies were covered with asolution containing the substrate Suc-A-A-P-F-ßNA, Fast-Garnetand different concentrations of CaCl2, the latter being an activatorof PrtS proteinase. Whatever the CaCl2

concentration (10, 20 or50 mM), colonies of strains 703 and 385 rapidly became red,whereas those of the negative strain 302 remained white. Usingthis test, we confirmed that the mutant strain was PrtS-negative,as colonies remained white even after several hours of contactwith the substrate solution. This test functioned on milk plates butnot on rich medium M17 plates for strain 703, which confirmed aprobable regulation of prtS expression by the growth medium asalready observed for this strain. This test will be useful to screenfor S. thermophilus PrtS-negative strains in milk and also for PrtS-deregulated strains in M17.

Growth characteristics of the wild-type and PrtS-negativemutant of S. thermophilus. By comparing the phenotypes of theparental strain 385 and its PrtS- mutant on FSDA, and their growthcurves in liquid M17 and milk, we showed that proteinase PrtSwas essential to the growth of S. thermophilus in milk.


The PrtS- mutant, plated on FSDA, appeared as flat andtranslucent colonies, as expected for PrtS- bacteria, whereas thePrtS+ parental strain appeared as white, opaque, rounded colonies.

In M17, both strains had similar growth curves with a µmaxof 0·89 and 0·85 h-1 for the parental strain and the mutant strain,respectively. In milk, streptococcal growth was determinedindirectly by pH measurement. Growth of the PrtS- mutant wasseverely impaired in milk, as indicated by the reduced acidificationof milk by this strain compared to the parental strain 385. For thePrtS- strain, milk acidification, and thus bacterial growth, wasrestored to the same extent as that for the wild-type strain, after theaddition of yeast extract to milk.

The Lactobacillus strain influences the extent of the positiveeffect of S. thermophilus/Lb. bulgaricus association.

Mixed cultures of S. thermophilus and Lb. bulgaricus were madeusing two different strains of Lb. bulgaricus. To choose the last twostrains, we first determined the effect of the co-culture of Lb.bulgaricus strains with the S. thermophilus CNRZ385 strain on milkacidification, compared to the pure culture of Lb. bulgaricus. Amongthe three strains of lactobacilli tested, the effect of adding strain385 on the acidification was greatest with strains Lb. bulgaricus 397and 1038; indeed, for these two Lactobacillus strains, the additionof the Streptococcus highly enhanced the acidification rate comparedto the Lb. bulgaricus strain alone. Furthermore, the positive effect ofthe bacterial association on milk acidification was more intense forstrain 1038, as, for this strain, the acidification rate and the finalpH were higher and lower, respectively, in the mixed culture thanin the pure culture. In contrast, addition of S. thermophilus strain385 had no significant effect on milk acidification by Lb. bulgaricusstrain 752. Thus, strains 397 and 1038 of Lactobacillus were kept forthe following study. In addition, a proteinase-negative mutant ofstrain Lb. bulgaricus CNRZ 397 was available and was used for thefollowing experiments.

In mixed cultures, proteinase PrtS has no effect on final pH andbacterial populations, but PrtB affects both.

The effect of proteinases PrtS and PrtB on acidification ofmixed cultures and bacterial populations was estimated by

measuring the final pHs, final total bacterial populations and finalindividual populations of cultures performed with a strain of S.thermophilus PrtS+ (strain 385) or PrtS- (strain 385-PrtS) and a strainof Lb. bulgaricus PrtB+ (strains 397 and 1038) or PrtB- (strain 397-PrtB).

The presence of proteinase PrtS had no effect either on thefinal bacterial populations or on the final pHs of mixed culturesinvolving PrtB+ Lactobacillus strains. The final total populationswere always similar in the presence or not of proteinase PrtS: 1·39and 1·36x109 c.f.u. ml-1, respectively, for cultures involving Lb.bulgaricus strain 1038, and 7·05 and 6·27x108 c.f.u. ml-1, respectively,with strain 397. The absence of any differences in final totalpopulations corresponded to similar final individual populationsof S. thermophilus and Lb. bulgaricus, regardless of the presence ofPrtS: 1·3x109 c.f.u. ml-1 for strains 385 and 385-PrtS and 8·9x107 c.f.u.ml-1 for strain 1038 for mixed cultures involving strain 1038,5·5x108 c.f.u. ml-1 for strains 385 and 385-PrtS and 1·1x108 c.f.u. ml-

1 for strain 397 in mixed cultures involving strain 397. Thiscorrelated well with the similar final pH obtained: 4·72 and 4·85 formixed cultures involving, respectively, strain 1038 and strain 397.

It is noteworthy that both the final total bacterial populationsand the acidification rates varied according to the Lactobacillusstrain associated with S. thermophilus strain 385. The final totalpopulation when using Lactobacillus strain 1038 (1·38x109 c.f.u. ml-

1) was twice as high as that of strain 397 (6·67x108 c.f.u. ml-1),because the Streptococcus populations were more than twice ashigh with strain 1038, Lactobacillus populations remaining constant.Final pHs were not significantly different, but the time required toreach these pHs was shorter for mixed cultures, including strain1038, than those including strain 397 (4·66 h with strain 1038versus 5·66 h with strain 397).

In contrast, the presence of proteinase PrtB affected both thefinal bacterial populations and the final pHs. Final total bacterialpopulations were threefold higher in the presence of PrtB than inits absence (7·05x108 versus 2·8x108 c.f.u. ml-1). This resulted fromhigher final populations of S. thermophilus in the presence of PrtB(6·1x108 versus 2·06x108 c.f.u. ml-1) and led to a significantly betteracidification in the presence of PrtB (final pH 4·86 versus 5·42).


In our conditions of inoculation (Streptococcus/Lactobacillus ratioof 1:1), S. thermophilus was systematically predominant in the totalfinal populations, regardless of the strain of Lactobacillus and thepresence of proteinases PrtS and PrtB. The magnitude of thispredominance depended on the Lb. bulgaricus strain used: withstrain 1038, S. thermophilus populations were 15-fold higher thanLactobacillus populations and fivefold higher with strain 397, whenPrtB was present. This predominance was less marked in theabsence of PrtB, as the S. thermophilus populations were threefoldlower (6·1x108 c.f.u. ml-1) than populations reached in the presenceof PrtB (2·06x108 c.f.u. ml-1).

Variation of individual populations of S. thermophilus and Lb.bulgaricus throughout mixed cultures in milk.

Proteinase PrtS had no significant effect on the variation of pHand of individual populations throughout the culture and,regardless of the mixed culture considered (except that involvingstrain PrtB-), the variation of these two parameters remained similar.Fig. 4a gives an example of this variation (a mixed culture madeof S. thermophilus 385 and Lb. bulgaricus 1038); for mixed culturesincluding Lb. bulgaricus strain 397, we observed the same behaviour.

During the first 60-90 min, which corresponded to the firstacidification phase, S. thermophilus grew exponentially, whereasLb. bulgaricus did not grow significantly. Then, as the pH remainedconstant, the streptococcal population stabilized for about 60-90 min, whereas Lb. bulgaricus started to grow regularly andcontinuously. Finally, during the last 2 or 3 h of fermentation,when the acidification rate was the highest, both the Lactobacillusand the Streptococcus grew regularly.

In contrast, proteinase PrtB was clearly involved in thevariation of bacterial populations and of pH, as demonstratedwith mixed cultures involving strain 397-PrtB. In fact, regardlessof the presence of PrtB, the first two phases of acidificationcorresponding to the sequential growth of S. thermophilus and Lb.bulgaricus were similar. However, during the third acidificationphase, the growth of S. thermophilus slowed down in the absenceof PrtB, the bacterial populations remaining almost constant duringthe last 2 h of fermentation. This reduced growth resulted in a

reduced acidification rate and an increased final pH (pH 5·42 inthe absence of PrtB and 4·86 in the presence of PrtB).

DISCUSSION

The present work aimed at evaluating the role of proteinasePrtS from S. thermophilus in the growth in milk of S. thermophilusin a pure culture. We also determined the effect of the presence ofboth PrtS and PrtB from Lb. bulgaricus on S. thermophilus/Lb.bulgaricus mixed cultures. For this purpose, we constructed atargeted negative mutant of proteinase PrtS from S. thermophilusCNRZ 385 and performed pure cultures of S. thermophilus andmixed cultures with Lb. bulgaricus in milk.

In milk, the extent of the beneficial effect of the S. thermophilus/Lb. bulgaricus association varies.

We observed that the effect of the co-culture of Lb. bulgaricusstrains with S. thermophilus strain 385 on the acidification of milk,and thus the benefit of the bacterial association, depends on thestrain of Lb. bulgaricus used. In fact, with Lb. bulgaricus strain 752,we did not obtain a marked beneficial effect of the association withS. thermophilus 385 as already observed by several authors withother strains. In contrast, mixed cultures of strains 1038 and 397resulted in a higher acidification than pure cultures. Acidificationwas higher with strain 1038 than with strain 397 due to higher S.thermophilus populations, Lb. bulgaricus populations being similar.These higher S. thermophilus populations probably resulted from abetter peptide and/or amino acid supply by one Lactobacillus straincompared to the other as these nitrogen compounds are growth-limiting for S. thermophilus in milk.

The two strains of Lb. bulgaricus thus probably differ in theirproteolytic potential, which is in agreement with the differencesobserved in the final quantities of free amino acids and free NH2groups in the supernatants of mixed cultures including these twostrains (data not shown). Some authors have also reported avariability in the Lb. bulgaricus proteolytic potential. This variabilitycould be related to the presence of one cell-wall proteinase in Lb.bulgaricus, which is the case of strain 397, or of two proteinases,as reported for other strains.


PrtS is essential to the growth of S. thermophilus in milk inpure culture but not in mixed culture.

Proteinase PrtS is essential to the growth of S. thermophilusgrowth in milk as its PrtS- mutant was unable to grow efficientlyin milk until a nutritional complement [yeast extract orbactotryptone (data not shown)] was added. This indicated thatproteinase PrtS was involved in nitrogen supply to the cell, viacasein hydrolysis, which is consistent with data previouslyobtained with cell-wall proteinases of other lactic acid bacteria.

However, we demonstrated here that proteinase PrtS had nosignificant effect on the growth of S. thermophilus in mixed culturesin milk with Lb. bulgaricus; the growth of the parental strain 385and of the PrtS- mutant in mixed culture was similar when Lb.bulgaricus proteinase PrtB was present. This indicates thatassimilable nitrogen compounds necessary for S. thermophilusgrowth are supplied by PrtB, as confirmed by the fact that theabsence of PrtB led to lower streptococcal populations. Furthermore,as the streptococcal population was higher in the presence of PrtBthan in the presence of PrtS, we can assume that PrtB is moreefficient in the supply of peptides to S. thermophilus than PrtS. Thiscan be explained by a more active proteinase PrtB compared toPrtS, as previous studies reported that the global proteolyticactivities of Lb. bulgaricus strains were 25-70 times higher than thatof S. thermophilus strains. We cannot rule out the possibility thatPrtS and PrtB have different substrate specificity, which leads tothe production of different peptides, some being more assimilablethan others. Indeed, PrtS is capable of hydrolysing MS-Arg-Pro-Tyr-pNA, a substrate also hydrolysed by lactococcal proteinasePrtP but not by PrtB. Furthermore, when comparing the substrate-binding region of proteinases PrtS and PrtB, in particular theresidues 138, 166, 748, which have been identified as being directlyinvolved in substrate specificity in lactococci, we noticed that theyare totally different in PrtS (Thr, Ala, Asp) and PrtB (Gly, Val, Thr).

Model of growth of S. thermophilus associated with Lb.bulgaricus and effect on acidification.

In all the mixed cultures performed in milk, we observed thesequential development of S. thermophilus and then of Lb. bulgaricus,which is in agreement with previous studies. Recently, the growth

of S. thermophilus in pure culture in milk has been characterized,in particular with regard to nitrogen nutrition; it consists of twoexponential growth phases, interrupted by a non-exponentialgrowth phase. From these latter results and those of the presentwork, we propose the following model of growth of S. thermophilusin mixed cultures with Lb. bulgaricus with three S. thermophilusgrowth phases corresponding to three acidification steps.

During the first acidification step, characterized by a smalldecrease in pH (<0·5 pH units), S. thermophilus grows exponentially,whereas Lb. bulgaricus does not grow; S. thermophilus is thusresponsible for this first acidification, as first observed by Pette &Lolkema (1950c ). The preferential growth of S. thermophilus can beexplained first by the fact that S. thermophilus has fewer nutritionalrequirements than lactobacilli in milk. In particular, S. thermophilusrequires few amino acids and is capable of synthesizing branched-chain amino acids; its growth can probably be supported by freeamino acids and peptides present in milk, as previouslydemonstrated in pure culture, regardless of the presence of PrtS. Incontrast, Lb. bulgaricus is much more demanding from a nutritionalpoint of view than S. thermophilus (Letort, 2001 ); its optimal growthrelies on the supply of essential factors (CO2, pyruvate, formate)produced by S. thermophilus. Second, in our study, mixed cultureswere performed at 42 °C, a temperature more favourable for S.thermophilus, whose optimal growth temperature ranges between40 and 45 °C, versus 45-50 °C for Lb. bulgaricus.

Then, the S. thermophilus exponential growth pauses and,concomitantly, the acidification, while Lb. bulgaricus begins to growslowly and regularly until the end of fermentation. This pauseprobably corresponds to depletion of amino acids and peptides inmilk, due to their consumption by S. thermophilus, as shown recentlyby Letort et al. (2002) in pure culture, and the absence ofcompensatory production by cell-wall proteinases. These authorsactually demonstrated that proteinase PrtS synthesis starts in themiddle of this phase and is maximal during the second exponentialgrowth phase in pure culture. Concerning the growth of Lb.bulgaricus, we assume that as S. thermophilus reaches a high cellulardensity during its first growth phase, it probably produces enoughgrowth-stimulating factors to favour the growth of Lb. bulgaricus.

Introductory Food Microbiology82 Experimental Validation of Low Virulence in Field Strains 83

Finally, during the following acidification phase, which leadsto a high pH decrease (about 1·5 pH units), Lb. bulgaricus continuesto grow; at the same time, S. thermophilus starts a second exponentialgrowth phase. We suggest that this acidification results not onlyfrom the growth of Lb. bulgaricus but also from that of S. thermophilus.This acidification phase is indeed greatly improved by the additionof S. thermophilus to a Lb. bulgaricus culture; furthermore, in theabsence of PrtB, acidification is reduced, while only S. thermophiluspopulations significantly decrease. The growth of S. thermophilusprobably occurs because of the utilization of peptides produced byPrtS (when PrtB is absent) but also mainly by PrtB. No differencesin the growth of S. thermophilus were observed in the presence orabsence of PrtS when PrtB was present, and S. thermophiluspopulations were significantly reduced in the absence of PrtB, i.e.when PrtS was the sole source of peptide production.

In conclusion, we have determined the role of cell-wallproteinases PrtS and PrtB in the growth of S. thermophilus and Lb.bulgaricus in mixed cultures. We have shown that PrtB is involvedin the optimal growth of S. thermophilus, whereas PrtS does notplay a significant role when PrtB is present. Studies of the effectof these proteinases on the free amino acid and peptide contentsas well as on the aroma profiles of mixed cultures are in progress.As precursors, amino acids are involved in the formation of aromain dairy products, and variations in their composition can affectaroma development. However, the different pH values observed inthe present study at the end of fermentation, when varying thepresence of proteinase PrtB, can modify the yoghurt flavour.

4EXPERIMENTAL VALIDATION OF

LOW VIRULENCE IN FIELD STRAINS

Several reports have described Listeria monocytogenes strainswhich were nonpathogenic or weakly pathogenic, but little isknown about these low-virulence strains. We found that 9 field L.monocytogenes strains were hypovirulent and 17 were avirulent,based on the number of mice contaminated and the colonizationof their spleens after subcutaneous inoculation. All these strainspossessed the known virulence genes. We have now assessed thelow virulence of these strains in other assays before determininghow they differ from virulent strains. We have shown that the low-virulence strains exhibited a phenotypic stability and were not amixture of virulent and avirulent bacteria. They did not recovervirulence after many passages in mice and colonized the spleensof mice more poorly than virulent strains after i.v. inoculation.Their lethal capacities, determined by 50% lethal dose (LD50), werelower than those of virulent strains. Like Listeria innocua, 14 of 17avirulent strains had no LD50

and were eliminated by the lymphnodes after subcutaneous inoculation. The virulent, hypovirulent,and avirulent strains were always significantly different, whateverthe tests of virulence used, confirming the importance of these low-virulence field strains in identifying the proteins involved invirulence.

INTRODUCTION

Listeria monocytogenes organisms are ubiquitous gram-positivebacteria. They are widespread in the environment and have beenisolated from many sources, including soil, sewage, decayingvegetation, and food. They are responsible for human diseases

Introductory Food Microbiology Experimental Validation of Low Virulence in Field Strains84 85

characterized by meningitis, meningoencephalitis, septicemia,abortion, and gastroenteritis. Through contaminated food, bacteriareach the gastrointestinal tract and can translocate the intestinalbarrier to infect lymph nodes. Then, through lymph and blood, afraction of the bacteria reach the spleen and liver. Apoptosis,neutrophils, and phagocytic cells contribute to the rapid clearingof the bacteria before complete abolition by the specific immuneresponse. In some cases, such as the immunocompromised host,bacteria multiply unrestrictedly in the hepatocytes from whichthey disseminate through blood to the brain and placenta. AlthoughL. monocytogenes is also present in the environment and is probablyfrequently ingested by humans, listeriosis is very rare. The incidenceis very low, around two to eight sporadic cases annually permillion people in Europe and the United States. If we exclude thesusceptibility of the host, another reason for this conflicting evidencemay lie in the variability of virulence in the L. monocytogenes strains.Serotypes of L. monocytogenes could also be linked to the level ofvirulence, as only three serotypes (1/2a, 1/2b, and 4b) have beenimplicated in human cases. However, no bacterial genes related tothe serotype have yet been found.

Studies using different assays have shown that virulence variesfrom one strain of L. monocytogenes to another. The mouse assaysare extremely sensitive assays for evaluating the pathogenicity ofL. monocytogenes by the systemic route. The immunocompromisedmouse model has shown a considerable difference in the 50%lethal doses (LD50s) of virulent and nonvirulent strains. In thesame way, subcutaneous (s.c.) inoculation of immunocompetentmice is very sensitive and specific, depending on the clinicalorigin of the strains. Tissue culture assays, i.e., cytopathogenictests, have also been developed to distinguish pathogenic andnonpathogenic L. monocytogenes strains. Certain genetic orphenotypic markers have been linked to the virulence of the strains.

In our previous paper, the virulence of L. monocytogenes strainswas evaluated with a plaque-forming (PF) assay on HT-29 cells,followed by s.c. injections of immunocompetent mice. We found 26low-virulence field L. monocytogenes strains identified ashypovirulent or avirulent. All these strains possessed the knownvirulence genes and exhibited the same growth in nonselectivemedia by a bioscreen study. However, the cause of the low virulence

is presently unknown. As a preliminary step toward understandingthe cause of this low virulence, it seemed important to knowwhether low virulence is a stable character over time which cannotbe enhanced after in vivo passages. It was also important to knowwhether these strains are also attenuated after intravenous (i.v.)inoculation and whether their lethality is modified.


Listeria Strains: The Listeria strains used and theircharacteristics are given in Table 1 . Virulence was estimated bythe method of Roche et al. Different studies allowed the detectionof 9 hypovirulent strains and 17 avirulent strains. To analyzethese strains, 13 virulent L. monocytogenes strains and 1 Listeriainnocua strain were added as control strains. The strains weremaintained in storage medium at 4°C. For analysis, they werecultured in brain heart infusion broth (3 ml) at 37°C for 8 h. BHIagar (BHIA; Difco) slopes were then seeded and incubated overnightat 37°C. The colonies were suspended in 2 ml of phosphate-bufferedsaline (PBS) (pH 7.3), standardized turbidimetrically, and dilutedappropriately for each test.

Cell Line and Culture Conditions: The human adenocarcinomacell line HT-29 (no. 85061109; European Collection of Animal CellCultures, Salisbury, United Kingdom) between passages 27 and67 was used. Cells were grown in 75-cm2 plastic tissue cultureflasks in Dulbecco’s modified Eagle’s medium with glucose (4.5 g/liter) (Invitrogen) supplemented with 10% (vol/vol) fetal calf serum(Invitrogen) and 2 mM L-glutamine (Invitrogen). Antibiotics (100IU of penicillin per ml and 100 µg of streptomycin per ml; Sigma,Saint-Quentin Fallavier, France) were routinely added to the culturemedium except for the virulence assays. Cells were maintained ina humidified incubator (at least 90% relative humidity) at 37°Cunder 5% (vol/vol) CO2.

Phenotypic Stability: An initial PF assay was done on fourstrains to demonstrate that hypovirulence was not the result oftwo populations of bacteria, one infecting the cells and other unableto infect the cells. The bacteria forming plaques were collected andused for a second PF assay. The first PF assay was performed withconfluent monolayers of HT-29 cells in six-well tissue culture


plates. Cells were infected with 5 log CFU (105 CFU) per wellsuspended in Dulbecco’s modified Eagle’s medium for 2 h at 37°C.Incubation was continued for a further 1.5 h with 100 µg ofgentamicin (Sigma) per ml in the culture medium. Each well wasthen overlaid with an agarose gel containing 0.48% indubiose inculture medium supplemented with 10 µg of gentamicin per ml.The same medium was then added to prevent cell starvation, andincubation was continued for 3 days. The cells including bacteriaaround plaques were recovered and lysed, and the bacteria wereused for a new PF assay (15). Bacteria maintained in the storagemedium were also tested. The results are expressed as the numberof plaques obtained for 7 log CFU deposited per well.

Virulence Recovery after in Vivo Passages: Spleen colonizationby five strains was monitored during 10 successive passages inmice, and a PF assay was performed at the end of the experiment.Groups of five 7-week-old conventional Swiss female mice (Iffa-Credo, Saint-Germain-sur-l’Arbresle, France) were injected s.c. intheir left hind footpad with 6 log CFU suspended in 50 µl of PBS.Each inoculum was checked by a viability count on tryptic soyagar (TSA) plates (Bio-Mérieux, Marcy l’Etoile, France). The plateswere incubated for 48 h at 37°C. Mice were killed 3 days afterinjection. Their spleens were removed aseptically, pooled, andhomogenized. Aliquots of each homogenate were used to assessthe bacteria in the spleens or to prepare the inoculum for thepassage in the next mouse.

The spleen colonization assessed on TSA plates is expressedas the number of log CFU per homogenate (homogenates from thespleens from the five mice in the group were pooled). The inoculumfor the next passage was prepared by incubating 100 µl ofhomogenate in BHI broth at 37°C for 32 h (step enrichment). Theincubated homogenate was then seeded on BHIA slopes andincubated for 17 h at 37°C. The strains isolated after passage 10were compared to bacteria maintained in the storage medium ina PF assay. The results are expressed as the number of plaques per7 log CFU deposited per well.

Determination of Lethal doses in Mice: DBA/2 breeder micewere purchased from Iffa-Credo. The mice were kept in the animalhouse of the laboratory in level 2 containment facilities, and themice reproduced. Groups of six 8- to 10-week-old female mice were

inoculated s.c. in their left hind footpad with 50 µl of bacteriasuspended in PBS. The inocula contained approximately 1 to 9 logCFU for the virulent strains, 2 to 9 log CFU for the hypovirulentstains, and 4 to 9 log CFU for the avirulent or nonpathogenicstrains. Each inoculum was checked by counting viable cells afterincubation on TSA plates for 48 h at 37°C. Mice were observedevery day for 15 days, and all deaths were recorded. All the miceremaining on day 15 were killed. The LD50s were calculated usinga probit dose-response model, considering a log transformation ofdose rates and the total number of mice that died. The percentageof mice that died in the three groups were also analyzed by logisticregression.

i.v. Injection: Female Swiss mice (6 to 9 weeks old) (Iffa-Credo)were injected with hypovirulent and avirulent strains. The micewere kept under controlled conditions (humidity, temperature, fooddelivery, and stress) during the experiments. Bacteria (4.5 log CFU)were suspended in 0.5 ml of PBS and injected i.v. The mice werekilled with carbon dioxide, 2 days after inoculation. Their spleenswere removed and homogenized in PBS using a glass homogenizerwith a loose-fitting pestle. Triton X-100 was added to a finalconcentration of 0.001%, and dilutions were made immediately.Four mice were used for each strain. The viable bacteria in theinoculates and spleens was counted on Columbia agar. Resultswere compared by analysis of variance and analyzed by the Tukey-Kramer multiple comparison method.

Kinetics of Colonization: Kinetics of colonization were analyzedfor four strains. Groups of five 6-week-old conventional Swissfemale mice (Iffa-Credo) were injected s.c. in their left hind footpadwith 4 log CFU suspended in 50 µl of PBS. The mice were killed1, 24, 48, and 72 h after injection.

The left and right popliteal lymph nodes, lumbar lymph nodes,spleen, liver, and lungs were removed from each mouse aseptically.Samples were homogenized, and the homogenates were diluted inBHI broth. Appropriate dilutions were plated onto TSA plates andincubated at 37°C for 48 h. Viable bacteria were counted. The meanlog CFU per organ was calculated only for samples with bacteria.Homogenates were kept overnight in BHI at 37°C for enrichment.Enriched homogenates in which Listeria strains were not detectedwere isolated on TSA plates and further incubated overnight at


37°C. The results are expressed as the number (log) of CFU perorgan.

RESULTS

Phenotypic Stability: All the low-virulence strains studiedwere cloned. We checked the possibility that low virulence couldbe the result of expression of two phenotypes, one of which couldinfect cells while the other could not. If this was the case, a smallsubpopulation of bacteria would be able to form plaques and if werecovered these bacteria, the number of plaques in a second testshould be higher. Among the low-virulence strains, we used thefour strains that produced few plaques (strains BO34, CR282, 449,and 442). Indeed, in order to observe a possible increase or decreasein plaque number, we could not choose strains forming a highnumber of plaques or no plaque at all. We found no difference inthe virulence of the bacteria recovered from the plaques and thecontrol bacteria under the conditions we used, confirming that allbacteria within the population have the same level of virulence.

Lack of Recovery of Virulence after in Vivo Passages: In orderto analyze a possible recovery of virulence in low-virulence strainsafter 10 passages in mice, we chose the three strains that infectedthree of five mice (436, BO34, and 464) and the two strains thatinfected one of five mice (SO205 and BO43). As these strains werehypovirulent, we pooled the five spleens in order to increase thechance of recovering bacteria. The numbers of bacteria recoveredfrom the spleens of mice infected with strains 436, 464, and BO34were fairly constant during the experiment. The numbers of bacteriafound in the spleens of mice infected with strains SO205 and BO43were equal to the threshold of detection, because they were notdirectly recovered from spleen homogenate but were recoveredafter 1 and 10 passages, respectively, in spleen homogenates onlyafter 32 h of growth in BHI medium. After four passages in vivo,the SO205 strain was not recovered from the spleens of the fivemice, despite the enrichment step. Thus, this value was below thethreshold. The number of plaques of bacteria recovered from thespleens after 10 passages was then compared to the number ofplaques of bacteria maintained in storage medium. The bacteriathat had been passaged in vivo formed 10 times fewer plaquesthan the control bacteria.

Lethality Study: Under our study conditions, L. monocytogenesstrains were virulent, hypovirulent, or avirulent, depending ontheir ability to colonize the spleens of mice. In order to knowwhether these strains exhibited the same lethality, the numbers ofmice dying were recorded during the 15 days after inoculationwith increasing doses of Listeria strains. DBA/2 mice were chosenfor these study because they are more sensitive to L. monocytogenesthan the Swiss mice. Some LD50s were calculated from very fewobservations due to the few deaths caused by some strains. Forsome strains, none of the mice died after 15 days or there were notenough deaths to calculate an LD50. In that case, maximal injecteddoses are used and indicate the difference in the virulence of thethree groups (virulent, hypovirulent, and avirulent strains). AnLD50 could be calculated for only 3 of the 17 avirulent L.monocytogenes strains; they were between 8.7 and 9.3 log CFU. The14 other strains were not lethal. The LD50s could be determinedfor only five of the nine hypovirulent L. monocytogenes strains (8.3to 9.0 log CFU). The LD50s for the 13 virulent strains were 4.1 to7.9 log CFU. We also compared the percentages of dead mice in thethree groups by logistic regression. The difference was highlysignificant (P < 0.0001), depending on the injected dose. Thedifference between the hypovirulent and avirulent strains wasalso highly significant (P < 0.0001), with the hypovirulent strainsbeing more lethal after injection of 9 log CFU.

i.v. injection: Differences in spleen colonization and lethalcapability were observed after s.c. injections. Bacteria were injectedi.v. to determine whether their virulence was modified when themode of inoculation changed. Only 4 of the 13 virulent strainswere tested, and their mean virulence ranged from 5.8 to 7.2 logCFU per spleen homogenate, with a mean of 6.58 log CFU for the4 strains. The mean virulence of the hypovirulent strains waslower than that of the virulent strains (4.0 to 6.3 log CFU perspleen homogenate; mean, 5.07 log CFU). The avirulent strainsincluded 11 strains with a virulence of 3.5 to 5.2 log CFU perspleen homogenate, with a mean of 2.93 log CFU. The numbers ofCFU recovered for six avirulent strains were below the thresholdfor the four mice (2 log CFU per spleen homogenate). The percentageof mice infected by the hypovirulent and virulent strains (100% of51 mice) was significantly different from those infected by the


avirulent strains (42% of 60 mice). Analysis of variance of themeans of log CFU indicates a highly significant difference (P <0.0001) between the three groups of strains. Analysis of the pairwisedifferences in mean numbers of bacteria per spleen homogenatewas obtained by the Tukey-Kramer multiple comparison method.All the different values compared, namely, the values forhypovirulent and virulent strains, hypovirulent and avirulentstrains, and virulent and avirulent strains, were statisticallysignificant. All the simultaneous 95% confidence intervals by theTukey method exclude zero.

Rate of Colonization: The rates of colonization by the virulentL. monocytogenes strain EGDe, the hypovirulent strain BO34, theavirulent strain 442, and L. innocua BUG499 were measured todetermine how fast the bacteria spread in mice after s.c. inoculationin their left hind footpad. Bacteria of the virulent strain were foundin all the organs studied 1 h postinoculation. The number of miceinfected and the degree of infection increased with time, so that allthe mice were infected on day three. The hypovirulent strain spreadmore slowly than the virulent strain did. Bacteria were recoveredin the spleens only on and after the second day and only in twoor three of the five mice. The liver was never infected. The avirulentstrain infected only the lymph nodes. The bacteria spread from theleft popliteal lymph nodes to the lumbar lymph nodes but not tothe spleen or liver, suggesting that the lymph nodes were sufficientto eliminate the bacteria. No bacteria were found in the blood, evenafter enrichment, but the lymph nodes were more severely infectedthan those of mice injected with the L. innocua strain.

DISCUSSION

It is difficult to detect and characterize low-virulence L.monocytogenes strains for several reasons. They grow at the samerate as virulent strains on nonselective media (i.e., TSA and BHIA),but detecting low-virulence strains on some selective media isproblematic. Indeed, some of these strains could be detected onPalcam medium only after 3 days of growth and were generallynot detected or poorly detected on Rapid L’mono medium.Moreover, the low virulence of these strains is often ascertained bya single test, and there is no standard, well accepted method foridentifying and defining low-virulence L. monocytogenes strains,

although several virulence assays have been described. We recentlydeveloped a virulence assay based on a PF assay and the s.c.infection of mice. It allowed the detection of 26 low-virulence fieldstrains of L. monocytogenes. However, it is important to bettercharacterize this low virulence before undertaking geneticcharacterization of the strains. In 1987, Pine et al. reported that theATCC 35152 strain contains nonvirulent, nonhemolytic coloniesthat originated as spontaneous variants from the hemolytic parentstrain. We therefore looked for such nonphenotypic stability, asany uncertainty could raise questions about the reliability of alldata obtained with such strains. The bacteria recovered from andaround a plaque formed the same number of plaques (of the samesize) as the parent bacterial culture, showing that low-virulencestrains did not consist of a mixture of virulent and avirulentbacteria.

Waseem et al. demonstrated that passaging the L. monocytogenesNCTC 7973 strain increases its virulence in rabbits, as evaluatedby the recovery of viable bacteria from the infected organs. In thesame way, Wirsing von Koenig et al. have shown that mice becamemore resistant after many passages as evaluated by LD50s. Thus,it is possible that the virulence of our strains also increases afterin vivo passages. Our data clearly show that bacteria conserved instorage medium do not increase their virulence during successivesubculturings. The number of Listeria per spleen after 10 passagesin mice was the same as those obtained at the first passage. Inaddition, the virulence of the strains after 10 passages was notincreased over that of their parent strains and was even diminishedfor some strains.

We also confirmed the low virulence of these strains by severalclassical tests. The lethality and the bacterial load of organs weretherefore used as criteria of pathogenicity after both s.c. and i.v.inoculation. We did not use the oral inoculation route in mice,because the results were less reproducible. Moreover, it is notrepresentative of human infection because mouse enterocytes donot express the same E-cadherin that human enterocytes do,according to the observations of Lecuit et al. Virulent strains haveLD50s from 1 x 104 to 8 x 107 CFU, and the spleens are heavilycolonized after both s.c. and i.v. inoculation. The hypovirulent andavirulent strains were much less virulent than the virulent strains

Introductory Food Microbiology92 The Effect of Inoculum Size on the Lag Phase 93

in all the assays used. The hypovirulent strains were close toavirulent strains in term of lethality, with an LD50 greater than 2x 108 CFU. Mice inoculated with 109 CFU of L. innocua or avirulentstrains were only lethargic and had ruffled for the first few daysafter infection and they recovered soon afterwards. However, thehypovirulent strains colonized the organs better than the avirulentstrains, particularly the spleen. Taking into account the capacityof the strains to colonize host organs, our results suggest that theavirulent strains spread better than the nonpathogenic L. innocuastrain BUG 499 because L. innocua did not colonize the lumbarlymph nodes after inoculation into the footpad.

The rate of infection after s.c. inoculation suggested that thevirulent strains spread quickly to heavily infect all the internalorgans and lymph nodes. The lumbar lymph nodes and liver seemto play a key role in the elimination of hypovirulent strains injectedinto the hind footpad, whereas the lymph nodes alone are sufficientto eliminate avirulent strains and L. innocua. These data agree withstudies showing that Listeria bacteria are cleared rapidly from thelymph nodes by CD8+ T cells and from the bloodstream by theneutrophils and Kupffer’s cells in the liver.

Thus, our data show that all the L. monocytogenes strainsscreened by our PF assay are low-virulence strains and that thislow virulence is not an artifact. Although the virulence of theavirulent L. monocytogenes strains is very similar to that of L. innocua,the rates at which they colonize organs are different. They alsohave the same in vitro and in vivo phenotypes as the strainsattenuated by deletion of actA/plcB and are no danger to humanhealth. However, the hypovirulent L. monocytogenes strains have alow but real virulence that is confirmed by the clinical origin oftwo strains. All the virulence assays used (LD50, spleen colonizationafter i.v. inoculation or s.c. inoculation in their left hind footpad)clearly showed significant differences between the three levels ofvirulence established by our virulence assays. All these strainshave the main known genes of virulence, but in-frame mutationscould decrease their virulence. Genetic and phenotypic analyses ofhypovirulent and avirulent L. monocytogenes strains are now inprogress in our laboratory.

5THE EFFECT OF INOCULUM SIZE

ON THE LAG PHASE

The effect of inoculum size on population lag times of Listeriamonocytogenes was investigated using the Bioscreen automatedmicrotitre plate incubator and reader. Under optimum conditions,lag times were little affected by inoculum size and there was littlevariation between replicate inocula even at very low cell numbers.However, in media containing inhibitory concentrations of NaCl,both the mean lag time and variation between replicate inoculaincreased as the inoculum size became smaller.

The variation in lag time of cells within a population wasinvestigated in more detail by measuring the distribution ofdetection times from 64 replicate inocula containing only one ortwo cells capable of initiating growth. The variance of the lag timedistribution increased with increasing salt concentration and wasgreater in exponential than in stationary phase inocula. The numberof cells required to initiate growth increased from one cell underoptimum conditions to 105 cells in medium with 1.8 M NaCl.

The addition of spent medium from a stationary phase culturereduced the variance and decreased lag times. The ability to initiategrowth under severe salt stress appears to depend on the presenceof a resistant sub-fraction of the population, although high celldensities assist adaptation of those resistant cells to theunfavourable growth conditions by some unspecified mediumconditioning effect. These results are relevant to the prediction oflag times and probability of growth from low numbers of stressedcells in food.

Introductory Food Microbiology The Effect of Inoculum Size on the Lag Phase94 95

INTRODUCTION

The lag phase of the microbial growth cycle represents thetime period needed for bacteria to adapt to a new environmentbefore cell multiplication commences. Recent work modelling thebehaviour of bacteria in foods has shown that the lag phase ismore difficult to predict than is the specific growth rate. Thisimplies that certain relevant variables affecting lag time are nottaken into account, particularly the physiological state of theinoculum and the inoculum size.

The physiological state of cells will be affected by their previousgrowth environment and by exposure to stress conditions. Cellularinjury caused by heating freezing or drying can extend the lagtime considerably and also increase its variability. Starved cellsmay also have very extended lag phases. The temperature historyof the inoculum culture and its growth conditions have a profoundeffect on lag. Some mathematical models used in predictive foodmicrobiology contain a parameter related to the physiological stateof the inoculum but it has not been possible so far to measurephysiological state directly in terms that are useful for predictingbehaviour. This may become possible eventually with developmentsin methods for measuring single cell metabolic activity.

Two classes of inoculum size effect on population lag may beenvisaged (a) cooperative or inhibitory effects of high cellconcentrations or (b) statistical effects at low cell concentrationsarising from the variability in individual lag times. There is littlespecific information about the possible effects of cell–cellinteractions on lag time although cell signalling has been shownto affect the emergence of cells from dormancy and the lag time ofpopulations in biofilms. The effect of substances produced inculture supernatants on microbial growth was reviewed byKaprelyants et al. (1999)

Mathematical treatments of the statistical effects of inoculumsize on population lag were provided by Baranyi (1998) andBaranyi and Pin (1999). These authors showed that as the cellnumber in the inoculum decreases, the population lag increasesby an amount that depends on the distribution of individual lagtimes and the maximum specific growth rate. The mathematicalanalyses predict that under optimum growth conditions, this

statistical effect would only be expected at inoculum levels ofbelow about 102–103 cells. McKellar and Knight (2000) have alsopresented a model of the lag phase of Listeria monocytogenes thattakes into account the variability of individual cells.

There have been relatively few experimental tests of the effectof inoculum size on lag, but available data suggests that the effectsare small in broth cultures under optimum growth conditions andalso in food. However more recently, it has been shown that thelag time of L. monocytogenes was extended in stressed cells whenthe inoculum size was small. Stephens et al. (1997) also showedthat the length of the lag phase is not only very variable in heatinjured cultures of low cell density, but also on average, longerthan those of high cell density.

The purpose of this study was to investigate the effect ofinoculum size on the population lag time of L. monocytogenesunder conditions of salt stress and to assess the heterogeneity inlag times of inocula containing very low cell numbers. During thecourse of this work, an effect of population size on the probabilityof initiating growth became evident.


Organism and Culture Conditions

L. monocytogenes NCTC 11994 was used throughout theseexperiments. It was kept on glass beads at “70 °C and maintainedon tryptone soya agar slopes at 5 °C. Stationary phase inoculawere prepared by inoculating 20 ml tryptone soya broth andincubating overnight at 37 °C. To produce an exponential phaseinoculum, 0.1 ml of a stationary phase culture was inoculated into20 ml TSB (pre-warmed to 37 °C) and incubated until the opticaldensity at 680 nm (OD680) reached 0.15 measured in aspectrophotometer with a light path of 1 cm. Viable counts weremeasured by using spread plates. Tenfold dilutions of culturewere made in Maximum Recovery Diluent (Oxoid) and duplicate50 µl samples spread on TSA plates.

Bioscreen

The effect of inoculum size on the variation in the populationlag was investigated by preparing tenfold serial dilutions of


exponential and stationary phase cultures in either TSB, TSB with1.2 M NaCl or TSB with 1.6 M NaCl.

Ten replicate 300µl inocula from each dilution were placed inthe wells of a microtitre plate and the plates were then incubatedat 37 °C in the Labsystems Bioscreen C. The increase in turbidityat 600 nm was monitored automatically every 10–30 min for 2–12days. The plates were shaken between measurements.

Detection times obtained from the outermost wells of themicrotitre plates were generally longer than those from the otherwells, particularly when the medium contained a high saltconcentration. On further investigation, we found that for longincubations (greater than 7 days), there was a decrease in thevolume of media in these outer wells of up to 10%. We thereforediscarded the results obtained from all the outer wells in allexperiments. So, for the initial experiments, there were only eightreplicates rather than 10 per dilution.

To investigate variability in the length of the lag phase at verylow inoculum levels, cultures were diluted to a level that gavegrowth in approximately 50% of 64 wells. According to the Poissondistribution, 70% of positive wells would then contain a singlecell and 24% two cells. Three salt concentrations, 0, 1.2 and 1.6 M,were investigated.

The possible effect of medium conditioning (alteration of somephysicochemical aspect of the growth medium by the previousgrowth of a culture) or cell signalling was checked by preparingdilutions of log phase cells using 50:50 spent broth/fresh TSB,with added NaCl as before. The spent broth was prepared byfiltering a 24-h culture through a 0.2-mµ pore size nylon membranefilter (Nalgene, Milton Keynes, UK). All experiments were repeatedat least twice.

Probability of Growth

The effect of inoculum size on the probability of initiatinggrowth in a culture was investigated by incubating 40 replicatesof a few critical inoculum levels, taken around the growth/no-growth inoculum size boundary at three salt concentrations, 0, 1.7and 1.8 M.

Mathematical Treatment of Data

Data from the Bioscreen were analysed using the approach ofBaranyi (1998). Detection time is converted to a corresponding‘physiological state’ parameter (a) that is an expression of the‘readiness of cells to grow’ using the following transformation

( )T

rmax detexp µ

α−

=

where µmax is the maximum specific growth rate, Tdet is the detectiontime, r is the ratio between the initial and the detected concentrationreached at Tdet.

RESULTS

The Effect of Inoculum Size on Lag Time

To test whether there was an inoculum size effect, tenfoldserial dilutions of cultures were made in TSB containing differentlevels of NaCl, and the detection times recorded. Assuming thatspecific growth rate is unaffected by inoculum size, the differencein detection time between consecutive dilutions should be constantand proportional to the maximum specific growth rate (µmax)provided there is no effect of inoculum size on population lagtimes. Hence, a plot of detection time versus the logarithm of theinoculum size should give a straight line. If, on the other handpopulation lag times were affected by inoculum size, a deviationin linearity would be expected. Under optimum growth conditions,i.e. in TSB with no added NaCl, detection times were indeedlinearly related to log inoculum size. The slopes of theexperimentally determined lines were close to those predictedfrom the growth rate estimated from viable count data from culturesgrowing in the wells. The specific growth rates estimated from thedetection times in Fig. 1a were 1.12 and 1.22 h”1 which is close toa growth rate of 1.27 h”1 estimated from viable count data fromcultures growing in the wells. There was little variability in thedetection times of replicate inocula at each inoculum level althoughslight scatter was observed at the lower inoculum levels (1–10cells). Under these conditions, there was no marked effect ofinoculum size on population lag and no difference in the responseof exponential and stationary phase cells.


As conditions became more stressful and growth ratedecreased, i.e. in TSB containing 1.2 M NaCl, there was greatervariation in lag between replicates particularly for the exponentialphase inoculum. The variability increased with decreasinginoculum size especially below 104 cells per well. The trend of thescatter was towards longer detection times rather than beingdistributed around the expected detection time. The net effect wasthat as inoculum size decreased, the mean detection time deviatedmore from that predicted assuming that lag time and growth rateremain constant. Assuming that the maximum growth rate isindependent of inoculum size, the observed deviation representsan increase in lag at low cell numbers. An additional effect wasthat, with exponential phase cultures, no growth was seen frominocula containing less than 10 cells. Cultures inoculated withstationary-phase cells showed a linear relationship betweeninoculum size and detection time, except at very low cell numbers,where detection times were longer than predicted and fewer wellsshowed growth.

The results from experiments with TSB containing 1.6 M NaClshow a continuation of this trend, such that the relationshipbetween inoculum size and mean detection time is no longerlinear. Less than 50% of the cultures grew when inoculated with104 cells, and none grew with inocula of less than 103 cells,irrespective of the growth phase. Exponential phase inocula gaveconsistently longer detection times than stationary phase inocula.

The detection time data were transformed according to themethod of Baranyi and Pin (1999) to yield values. If the lag timesof the individual cells are not affected by the inoculum size, thenthese physiological state values should be scattered around aconstant, independently of inoculum size, but with increasingvariance as the inoculum size decreases. When data from stationaryphase cells inoculated into 1.2 M NaCl were transformed in thisway, values for stationary phase cells did scatter around a constant,except at the lowest inoculum level, indicating that inoculum sizedid not influence readiness of cells to grow except at the lowestinoculum level.

However, with exponentially growing cells, the values deviatedprogressively from the constant showing that inoculum size affected

lag by a mechanism separate from the statistical effect due to theinherent lag time distribution.

The distribution of detection times was studied in more detailfor very small inocula. Sixty-four replicates were prepared atinoculum levels that gave growth in approximately half thereplicates. The average number of cells per inoculum required togive growth in half the replicates was determined independentlyby plate counting.

For TSB containing 0, 1.2 or 1.6 M NaCl, the mean inoculumsizes were 1.2, 5.0 and 2000 cells, respectively. However, sinceonly half the wells actually showed growth, the effective meaninoculum size according to the Poisson distribution would havebeen 0.7 cells per well in each case. The data were normalised bysetting histogram bin size equal to the doubling time for each saltconcentration, hence, the increase in variability in detection timesshown in the figures was not simply an effect of growth rate. InTSB alone, there was very little variation in detection times (andhence lag times) between inocula of approximately one cell.

In medium containing 1.2 M NaCl, there was a pronounceddifference between the responses of exponential and stationaryphase cells. Stationary phase cells showed relatively little variationin detection times a slight tailing towards longer detection times.The range between shortest and longest detection time wasequivalent to about five doubling times. Under the same conditions,exponential phase cells showed a much larger variation, with thelongest detection times being approximately twice that of theshortest (a range of about 22 doubling times). A similar effect wasobserved at the higher salt concentration of 1.6 M.

Effect of Inoculum Size on the Ability to Initiate Growth

When cells were inoculated into TSB with no added salt, thepercentage of wells showing growth was close to that predictedfrom the Poisson distribution, showing that growth was possiblefrom single cells. However, with increasing salt concentration, thenumber of cells required to initiate growth in 50% of the wellsincreased to more than 103 cells in media containing 1.7 M NaCland around 106 cells the presence of 1.8 M NaCl.


Effect of Spent Medium

To further investigate the cause of the inoculum size effects,log phase cells were inoculated into fresh TSB mixed with anequal volume of spent broth. The hypothesis was that inoculumsize effects (a shortened lag phase or increased probability ofgrowth at high cell density) might be caused either by carry-overof substances with the inoculum, or the presence of a signalchemical produced during the lag phase that may also be presentin medium from stationary phase cultures. Spent medium had noeffect at NaCl concentrations of 1.2 M or less.

However, with 1.6 M NaCl detection times were shortenedwith inocula of 3000 cells per inoculum or less, and the variabilitybetween samples reduced. A simple ANOVA was run on thedetection times measured at the two lowest inoculum levels, wherethe variabilty of the lag times of individual cells has the biggesteffect on the population lag. An F-test showed that the probabilitythat the shorter avarage lag times were observed only by chancewas less than 10"3. The standard deviation of the detection timesdecreased by more than 50% for both inoculum levels, confirmingthe significance of the spent medium effect.

DISCUSSION

Use of the Bioscreen

The Bioscreen proved to be a useful means of producing largequantities of growth curve data. It has been used for a number ofdifferent applications such as determining bacterial growth rates,studying the effect of different conditions on growth, growthinhibition studies, enumeration of bacteria from food samples,and comparison of pre-enrichment media for resuscitation ofinjured salmonellae. However, the data produced need to beanalysed with care. The spurious readings obtained from theoutermost wells were most obvious when the cells were grownwith high NaCl concentrations under conditions that requiredlong incubation times before growth became detectable. Thisprobably resulted in evaporation from the outermost wells, a featurethat limits the incubation times that can be employed in thesestudies.

THE EFFECT OF INOCULUM SIZE ON LAG TIME

Our basic assumption was that the maximum specific growthrate (µmax), within the physiological range of any given organism,is independent of cell history and is uniquely determined by itsenvironment. This is shown by the global constancy of reportedµmax values for particular organisms grown under the sameconditions. A recent report by Membré et al. (1999) describing aneffect of inoculum pre-treatment on growth rate of L. monocytogenesis an exception to this rule, though the effect was very small.Therefore, for any set of conditions, a plot of the logarithm ofinoculum size versus detection time should yield a straight linewhose slope is proportional to specific growth rate, provided lagtime is constant and unaffected by inoculum size. If the units onthe abscissa are Ln inoculum size, the slope will equal 1/µmax, ifthe units are log2 inoculum size, the slope will be equal to doublingtime. From the above, a deviation from linearity would imply thatpopulation lag was not independent of inoculum size.

The lag time of a population is less than the average lag timeof individual cells because cells with shortest lags begin to multiplysoonest and their descendants dominate the populationnumerically. This effect was analysed by Baranyi (1998) whoderived a mathematical relationship between population lag andthe distribution of individual cell lag times and growth rate. Thisanalysis revealed that as cell numbers in the inoculum increasedfrom 1 to 103, the scatter of lag times from replicate inoculadecreased. The greatest variation in lag should thus occur ininocula containing low numbers of cells from populations thathad a wide scatter of individual cell lag times. Any variation inlag time between replicate inocula will be reflected in acorresponding variation in detection times.

Because plots of inoculum size versus detection time wereessentially linear for growth in TSB, we conclude that, underoptimum conditions, there was no effect of inoculum size onpopulation lag, except that due to statistical variation which wasnoticeable only at low cell numbers (100 cells per well or less). Ourresults obtained under optimum conditions agree with the workof Jason (1983) who found that lag and growth rate of Escherichiacoli were independent of inoculum size. Duffy et al. (1994b) also


concluded from viable count data that the lag times of L.monocytogenes in Listeria Selective Broth, Palcam broth or a mixtureof beef mince and broth, were unaffected by inoculum size.Although an increase in lag time was observed with decreasinginoculum size this was statistically insignificant due to the widescatter of the data.

Under the more exacting growth conditions provided by TSBcontaining 1.2 or 1.6 M NaCl, there was a wider scatter of detectiontimes as inoculum size decreased, and the mean value divergedincreasingly from that predicted assuming a constant lag.This was particularly noticeable with exponential phaseinocula. Exponential phase cells are generally more fragile andsusceptible to injury than those in the stationary phase and thismay account for their longer and more variable lag times followingosmotic shock. Transformation of the data according to Baranyiand Pin (1999) showed that inoculum size had an effect on lagthat was separate from the statistical effect caused by thedistribution of lag times of single cells, i.e. there was a cooperativepopulation effect.

An effect of inoculum size on lag has been shown previouslywith L. monocytogenes inoculated into broth at pH 5.9 and anincubation temperature of 14 °C, designed to simulate ripening ofCamembert cheese. The biggest inoculum size effect was seenwhen cells were grown at 30 °C then held for 4 weeks at 4 °Cbefore inoculation into broth at 14 °C, when an inoculum of 103

cells gave a lag of less than 21 h compared with 7.7 days for aninoculum of 10 cells. An extension of lag time and an increase invariability between replicate inocula has also been reported forheat-injured cells of Salmonella typhimurium. This is similar to theeffect described here with salt-stressed cells of L monocytogenes.

Effect of Inoculum Size on the Probability of Initiating Growth

Under optimum conditions a single cell was able to initiategrowth in TSB but, in the presence of high salt concentrations,much larger inocula were needed. A similar observation was madeby Razavilar and Genigeorgis (1988) who found that the numberof cells of L. monocytogenes required to initiate growth at 30 °C wasunaffected by NaCl concentrations up to 1.37 M. However, at

concentrations of 1.71 to 2.05 M NaCl, at least 105 cells wererequired for growth to occur, comparable with an inoculum sizeof 103 with 1.7 M and 106 with 1.8 M NaCl that we observed at37 °C.

Statistical Explanations of Population Size Effects

Studies with inocula containing predominantly single cellssuggested that the increase in mean population lag time at lowcell number under stressful growth conditions could have beenexplained simply by the corresponding increase in scatter in thelag times of single cells. In unstressed populations the variationof lag times was only about ±2 doubling times. Part of that variationcan be accounted for by variation in the number of cells per wellas predicted from the Poisson distribution.

Since the number of wells showing growth was around 50%,about 70% of positive wells would contain one cell and 24%would contain two cells. The detection times of wells containingtwo cells would be shorter by one generation time than thosecontaining only a single cell. Under stressful conditions, the scatterincreased to about ±10 doubling times implying that the variabilityin detection time was largely due to greatly extended lags ratherthan different numbers of cells in the wells.

The question nevertheless arises whether inoculum size effectson lag time and the probability of initiating growth can beexplained solely by the distribution of some resistance attributewithin the population. In broth containing 1.2 M NaCl, a widescatter of lag times was seen even with inocula containing 3000cells, a number which, from the theoretical results of Baranyi(1998), should have ensured that the lag times of replicates wouldhave converged close to the population mean.

However, a comparison of the dilution at which wells showedno growth with the known number of cells inoculated, showedthat the number of cells able to initiate growth was much less thanthe number present in the wells, and this may therefore bring theeffective inoculum size within the range where statistical effectsbecome apparent. The same argument applies with cells in 1.6 MNaCl except the fraction of cells able to initiate growth was evensmaller.


Cooperative Effects on Lag and the Ability to Initiate Growth

An alternative explanation for the inoculum size effects is thatsome kind of conditioning of the culture medium or chemicalsignalling is required before growth can occur under stressfulconditions. There appears to be a critical threshold cell densitybelow which growth initiation is not possible and this thresholdis related to the severity of the culture conditions.

Cells may produce a chemical or physicochemical change insitu, or there may be carry-over of substances from the inoculum.The addition of spent medium to TSB containing 1.6 M NaClshortened detection times, and decreased lag time variability, butno change was recorded in the probability of initiating growth. Arequirement for specific signal molecules in recovery from stresshas been reported in some bacteria. Cells of Nitrosomonas europaearecovered from starvation and commenced growth sooner whencells were present in a biofilm.

Acylated homoserine lactones were shown to reduce the lengthof the lag phase in a concentration dependent manner, suggestingthat these signal molecules responsible for this cooperative effect.In the Gram positive Micrococcus luteus a peptide signal moleculethat is necessary for the revival of dormant cells has been isolatedfrom culture medium.

The growth stimulatory effects described here may also involvesignal molecules, but a non-specific conditioning effect could alsoexplain the phenomenon. For example, the protective effect of highcell densities on stressed bacteria may be due to non-specificeffects such as leakage of magnesium or other unspecified materialfrom dead and injured cells that protects the remaining cells.Oxygen tension and redox potential are other examples of non-specific factors that can markedly effect the recovery of stressedcells.

The results reported here support the view that the ability toinitiate growth under severe salt stress depends on the presenceof a resistant sub-fraction of the population, but that high celldensities appear to assist the adaptation of those cells to theunfavourable growth conditions by some unspecified mediumconditioning effect.

RELEVANCE TO GROWTH IN FOOD

The results presented here predict that, under inhibitoryconditions, the mean lag time of cells that were sparsely distributedin food would be appreciably longer than that indicated frombroth experiments with large inocula. Mackey and Kerridge (1988)found that lag times of salmonellae growing in minced beef attemperatures between 10 and 35 °C were much more variable thanthe corresponding growth rates. Considering all data, there wereno statistically significant differences between the lag times oflarge and small inocula (40 or 104 cells g”1, respectively) but at thelowest most stressful temperature measured (10 °C), the lag timewas 50% longer (60 h compared with 40.6 h) for the lower inoculumlevel.

The results presented here also suggest that, under inhibitoryconditions, the probability that a pathogen could initiate growthin any single food item containing very low numbers of cellswould be less than that suggested by growth studies in brothinoculated with thousands of cells. The exact probability valueper pack would depend on the distribution of resistance withinthe population and the actual number of cells present per pack.The overall probability that growth would occur at least in somepacks would be the same as that calculated from broth studieswith large inocula. Any reduction of lag time by cooperative orconditioning effects would depend on cell concentration andproximity; for example whether cells were present as microcoloniesor single cells. Predicting the behaviour of bacteria in food is oftenbased on experiments with initial inocula in excess of 103 cellsml”1. From our work, it would appear that beneath this level ofinoculation, the lag phase becomes longer and the probability ofgrowth is less. This may have significance in estimating risk andcalculating safe storage times for foods.

Introductory Food Microbiology Modelling the Growth of Listeria Monocytogenes...106 107

6MODELLING THE GROWTH OFLISTERIA MONOCYTOGENES IN

DYNAMIC CONDITIONS

A recurrent neural network for the prediction of Listeriamonocytogenes growth under pH and aw variable conditions wasdeveloped. The use of this model offered the possibility to take intoaccount the consequences of the variations of the factors on L.monocytogenes growth. The effects of solutions, such as NaCl, aceticacid and NaOH, and their interactions on the response of L.monocytogenes cells were studied. Furthermore, the results showedthe capacity of the recurrent neural network to predict growthscarried out in different experimental conditions without usingthose used for its elaboration.

INTRODUCTION

Predictive microbiology combined the knowledge of bacterialgrowth responses over a range of conditions with the power ofmathematical modelling to enable predictions of growth.Mathematical modelling techniques can help to predict how foodpreservation systems may affect growth kinetics. Most mathematicalmodels developed to simulate the growth of micro-organisms inrelation to environmental conditions, were elaborated from datacoming from growth carried out in constant conditions ofenvironmental factors, such as aw or pH. But, food-manufacturingprocesses can decrease the pH of food, produce organic acids, e.g.,pickling or fermentation, or reduce water activity, e.g., by additionof an agent such as sodium. These processes are extensively used

as mechanisms to prevent microbial growth and to ensure foodsafety. Therefore, it is important to understand and to be able topredict the responses of micro-organisms, more particularlypathogenic micro-organisms as Listeria monocytogenes, in thepresence of variable environmental factors. In recent years, severalinvestigations which described the growth of micro-organisms inthe presence of temperature changes, were carried out. Whenbuilding these dynamic models, these authors made the hypothesisthat the transposition of results obtained from constant conditionsto variable conditions was possible. Cheroutre-Vialette et al. (1998)demonstrated the importance of taking into account the variationsof environmental factors, which can occur during a food processingand induce a stress situation for the micro-organisms. The objectiveof this work was to produce a dynamic model that predicts thegrowth of L. monocytogenes as a function of fluctuating conditionsof acid pH, alkaline pH and concentration of NaCl. To this end,an alternative approach to conventional methods of microbialgrowth predictions, that is, a recurrent multilayer neural networkapproach, was used.

MATERIAL AND METHODS

Strain and Medium

L. monocytogenes 14 (serotype 4b), obtained from an industrialenvironment, was used throughout the study. All growthexperiments were conducted in a tryptic meat broth (TMB). TMBregulated at pH=7 and aw=1 was called standard medium.

Experimental Procedure

An automated turbidimeter was used to follow the growth ofthe strain in the micro-titer plates. The working volume in eachwell of the micro-titer plate was 400 l. Optical density (O.D.) wasread at a wavelength of 600 nm. According to the procedureexpoused by Cheroutre-Vialette et al. (1998), for all experimentsthree conditions were studied: standard, limiting and shockconditions. Under standard and limiting conditions, bacteria weregrown in TMB adjusted to the desired aw and pH values. Theosmotic variation was achieved by the addition of NaCl accordingto Chirife and Resnik (1984). Acetic acid and NaOH (Prolabo)


were added to adjust low and high pH respectively. The shockcondition was defined as follows: the bacteria were grown instandard medium until the beginning of the exponential phaseand were then shocked by the abrupt addition of shock solutions.These shock solutions were prepared in order to obtain a finalvalue similar to those indicated for the limiting conditions. Thetemperature was 20°C. For each combination, six growth repetitionswere carried out.

Experimental Design

The data of L. monocytogenes 14 growths at 20°C were takenfrom an experimental study using a combination of two centralcomposite designs and a factorial design. The ranges of the differentenvironmental parameters were as follows: NaCl: 0–8%; acid pH:5.6–7.0 or alkaline pH: 7.0–9.5. Each factor was studied at fivelevels and for each combination, the shock and limiting conditionswere studied. At least 50 experiments were performed accordingto 25 growths in shock condition and 25 growths in limitingcondition.

Additional Experiments

A two-litre fermentor SET2M was used. O.D. of the growthwas measured with a spectrophotometer at 600 nm. The protocolwas similar to that in Bioscreen C adjusting the different volumes.Growths in shock conditions carried out in the fermentor allowedthe study of the osmotic shock effects by continuous additionmode. Seven hundred and fifty milliliters of standard mediumwere inoculated. At the beginning of exponential phase, 250 ml ofosmotic solution giving a final concentration of 8% NaCl wereadded using a peristaltic pump. The addition mode of shocksolution in Bioscreen C was by steps of 2%.

Data Analysis

Averages of the O.D. were calculated for the six repetitions ofinoculated media and for the four repetitions of non-inoculatedmedia. The data were then analysed using the procedure describedby Bégot et al. (1996). Four quantities were calculated at time t:

• (O.D.i)t, the mean of the O.D. of the 6 repetitions ofinoculated media for each combination;

• (O.D.ni)t, the mean of the O.D. of the 4 repetitions of non-inoculated media for each combination;

• (DO.D.)t=(O.D.i)t–(O.D.ni)t;

• Yt=log10[(DO.D.)t/(DO.D.min)] where DO.D.min was thelowest DO.D. value above the detection threshold.

Recurrent Neural Network (RNN) Model

Neural network (NN) maps inputs to outputs. Supervisedneural networks are capable of learning from previous examplesthrough iteration without the requirement of a priori knowledgeof the relationships between the process variables. NNs are madeof a number of simple, highly connected processing elements (PE)called neurones. The architecture of a NN describes how the NNis constructed from layers of PEs. Each PE receives inputs fromother PEs or from the outside. The PEs in the input layer onlytransfer scaled inputs to the appropriate PEs in the hidden layerthrough weighted connections. Each PE of the hidden layer andoutput layer calculates the weighted sum of its inputs and passesthe result through a transfer function. Most often a non-linearsigmoid transfer function is employed. The NN is iteratively trainedby presenting to it representative exemplar input/output vectors.The weights of the neural connections are adjusted in order tominimise a cost function equal to the mean square of the outputerror. The weights are usually randomly initialised. It has beenshown that one hidden layer of neurones is sufficient toapproximate any continuous non-linear function, although morecomplex networks may be employed in special applications.

The distinction between feedforward and recurrent multilayerNNs is the following: for the feedforward multilayer NNs anelement of a given layer can only receive information from previouslayers; on the contrary, for recurrent multilayer NNs, the valuescalculated by the NN can be fed back to the NN input layer or anyprevious layers. The difference in structure induces a difference intraining: recurrent multilayer NNs have to be trained on a sequenceof predictions, whereas training feedforward multilayer NNsinvolves only instantaneous predictions. Training a recurrentneural network (RNN) is more time-consuming but it is able toprovide better results than a feedforward NN when simulating the


evolution of phenomena with time. RNNs trained to learn dynamicevolutions of a process are more constrained and are forced to finda greater stability as well as a representation of the dynamic linksbetween control and process variables. Therefore, RNNs providea better prediction confidence than feedforward multilayer NNsthat only predict the outputs at a fixed horizon. Furthermore, thistechnique is more interesting in an advanced control perspective.

In the present study, a recurrent multilayer structure whichcontains one input layer, one hidden layer and one output layer,was used. The architecture of the RNN is presented in Fig. 2. Itwas designed to contain:

1. in the input layer, five input parameters: Yt–Dt, Yt–2.Dt,Yt–3.Dt, pHt–Dt and NaClt–Dt (%)

2. in the output layer, one output parameter: Yt whichrepresented the predicted response.

A priori, there are no rules for the choice of the hiddenstructure. It is determined empirically: the structure that gives thebest results is chosen. The optimum neurone number of the hiddenlayer was iteratively determined by developing several RNNs thatvary with the size of the hidden layer (3 to 10 neurones weretested) and simultaneously observing the change in the meansquare of the output error.

This was carried out with the training and testing data. Sixneurones in the hidden layer was determined as the best structure.The sigmoid function f(x)=1/(1+exp(-x)) was chosen as anactivation function for each neurone. The RNN was trained byiteration using a repeated presentation of representative exemplarinput/output vector pairs.

The weights of the neural connections, initially chosenrandomly, are adjusted by a non-linear optimisation technique:the quasi-newtonian formula of Shanno (1970) in order to minimisea cost function equal to the mean square of the output error. Thelearning base was used to adjust the weights, the testing base toprovide over-learning during weights optimisation, and thevalidation base for validation of results. At least, 60% of experimentswere included in the learning and testing bases, the last 40% werein the validation base. In order to ensure the validity of the study,

the growth results carried out in fermentor or in Bioscreen C withvarious modes of shock exposure were included into the validationbase.

RESULTS

Growth Predictions of the Validation Base

The analysis of the growth predictions in the limitingconditions showed that the RNN represented satisfactorily theexperimental data whatever the conditions tested (alkaline–osmoticor acid–osmotic). The different characteristics of the L. monocytogenesresponse, i.e. induction of a lag time and growth recovery differentto those observed in the new environment, were taken into accountby the RNN whatever the combination, alkaline–osmotic or acid–osmotic. Furthermore, RNN was able to predict the effect of thetype of shocks and their combinations. There was a good agreementbetween the experimental growth and the prediction.

Growth Predictions of the Additional Experiments

The previous results of the validation base demonstrated theability of the RNN to predict growths under shock conditionswhen the pH and aw transitions were carried out abruptly by onestep. The objective of additional experiments was to investigatethe capacity of the RNN to represent the response of L.monocytogenes cells shocked in exponential phase with 8% NaCladded by repeated steps of 2% NaCl or continuously during aduration of one or four generation times. The effects of adding 8%NaCl in 7.7 h (four generation times). The extrapolation to newexperimental conditions (mode of shock exposure) was made witha good agreement by the RNN. It confirmed that the RNN has thecapacity to predict growths carried out in different experimentalconditions from those used for its elaboration.

DISCUSSION

These results obtained at variable conditions showed thatneural networks can effectively be used to study the complexeffects of fluctuating environmental conditions on micro-organismbehaviour. Such dynamic model could follow the microbial impactof different steps associated with production, distribution and


retailing of a food and so could be an important support to HACCPand food safety systems.

Growth of Listeria monocytogenes as a function of dynamicenvironment at 10°C and accuracy of growth predictions withavailable models.

A combination of a factorial design and two central compositedesigns was used to assess quantita-tively the effects andinteractions of water activity (1-0.95) and pH (5.6-9.5) variationson the growth of Listeria monocytogenes in a meat broth at 10°C. Atinoculation or at the beginning of the exponen-tial phase, the cellswere exposed to the addition of NaCI and acetic acid or NaCI andNaOH.

The effects of abrupt fluctuating conditions on the generationand lag times were analysed using turbidity mea-surements. Thedata indicated that the cells exposed to osmotic and acid or alkalinevariable conditions from the time of inoculation were less affectedthan cells exposed at the beginning of the exponential phase. Inthis last case, a lag phase could be induced and the growthrecovery was different from those expected in the new environment.Generation time values were estimated by three available predictivemodels which describe the effects of temperature, salt concentrationand pH on L. monocytogenes growth to highlight the potentialproblems of variable conditions.

The aim of predictive microbiology lies in predicting the growthkinetics of micro-organisms at a given moment during foodprocessing and so predicting the evolution of food during storage,manufacturing or preservation accidents. One of the great interestsof predictive microbiology is to optimize the shelf life of a highrisk food product by the use of appropriate predictions.Mathematical models have been developed to simulate the growthof micro-organisms in relation to given environ-mental conditionsand can be classified by the microbiological event studied, themodelling approach used or the variables considered.

Most of these models were elaborated from data acquiredun-der constant environmental conditions as poly-nomial models,e.g. Food micromodel, Pathogen modeling program, whichexpressed bacterial growth as a function of factors such as pH,

water activity (%NaCl), temperature. But envir-onmental factors,such as temperature, Aw, or pH which are often the most importantfactors governing microbial behaviour in food, may vary extensivelyduring food processing, throughout the complete production anddis-tribution chain of a food product. Indeed, a food product isexposed to environmental var-iations during certain stages in aprocess, e.g. food cooling, natural product acidification or productbrining. In this way, it is important to be able to quantifyundesirable micro-organism growths, like Listeria monocytogenes,in envir-onmentally variable conditions.

The purpose of this study was to determine the effects of thecombined alkaline-osmotic and acid-osmotic conditions, constantor vari-able during time, on the growth of a L. monocy-togenesstrain at low temperature. Considering the temperatures likely tobe used for refriger-ated storage (0-10°C) or for working meatpro-duct premises, the temperature of the study was establishedat 10°C.

NaCl, acetic acid and NaOH were used to regulate wateractivity, acid pH and alkaline pH respectively. As the capacity ofBioscreen C for studying growth kinetics under variable conditionswas ever established, this material was used in this study.Moreover, with the objective to illustrate the potential pro-blemsassociated with the modelisation of growth under fluctuating pH/Aw, conditions, we proposed to analyse the accuracy of thegen-eration time predictions with available models andcharacterize these predictions with our ex-perimental data.

Materials and Methods

Strain and Media

Listeria monocytogenes 14 (serotype 4b, ob-tained from industrialenvironment) was used throughout the study. Stock cultures weremaintained in TSA agar slopes and stored at 4°C.

For preparation of inocula, cultures on TSA were subculturedin a meat medium (MM) which contained meat peptone 10 gl-1,yeast extract (Dif-co) 5 gl-1 and glucose 5 gl-1. The culture mediumfor growth studies was a tryptic meat broth (TMB). It was bufferedwith a K2HPO4-KH2PO4 (Merck) 0.1 mol l-1 solution in proportion1:1 (v/v) to pH 7.0.


Monitoring of Bacteria/growth

An automated turbidimeter was used to follow the growth ofL. monocytogenes 14 in the micro-titer plates. Optical density (OD)was read at a wavelength of 600 nm.


The strain was incubated on TSA agar slopes for 7 h at 37°Cand then transfered to MM that was incubated in a rotary shakerwaterbath for 20 h at 20°C, by which time the strain had reachedthe stationary phase. A proportion of the culture was inoculatedin various media to give a con-centration of about 3.107 cfu ml-1,in order to be above the detection threshold of the Bio-screen C.The viable number of bacteria was confirmed by TSA plate counts.

Figure. Experimental plan showing the combination of osmotic and acidstresses using NaCI and acetic acid respectively ((D); and the combinationof osmotic and alkaline stresses using NaCI and NaOH respectively (Q).The points of the factorial design are represented by n. The experimentnumbers are indicated in italics.

For all experiments, three conditions were studied: standard,limiting and shock conditions. Under standard and limitingconditions, bacteria were grown in TMB or in TMB ad-justed tothe desired AR, and pH values. For the inoculated TMB medium,300 µl were dispensed in each of six wells and the same vo-lumeof non-inoculated medium was dispensed in each of four wells inorder to determine the OD of the growth medium and to detect

possi-ble contamination. The well numbers were con-sidered asreplicates. The shock exposures were arranged as follows: 300 µlof inoculated and 300 µl of non-inoculated TMB were dispensedin each of six and four wells, respectively. An aliquot (100 µl) ofconcentrated ( x 4) shock so-lutions or standard TMB was addedwhen the OD was near 0.2, corresponding to the begin-ning of theexponential phase. The osmotic shock was achieved by theaddition of NaCl (Prolabo) according to Chirife and Resnik (1984).Acetic acid was added to adjust acid pH. The effect of high pHstresses was studied with the presence of NaOH (Prolabo). Thetime which elapsed before adding the shock solutions to 90 wellswas about approximately 10 min. These plates were placed in theBioscreen C previously adjusted to 10°C.

Experimental Design

The growth of L. monocytogenes 14 at 10°C in the presence ofsalt (0-8%), corresponding to Aw value (1-0.95) and in differentvalues of acid pH (5.6-7.0) or alkaline pH (7.0-9.5) was stu-died byusing an experimental design. This design was a combination oftwo central com-posite designs and a factorial design. Each factorwas studied at five levels and for each combination, the shock andlimiting conditions were studied. At least 50 experi-ments wereperformed.

Data Analysis

Averages of the OD were calculated for the six repetitions ofinoculated media and for the four repetitions of non-inoculatedmedia. The data were then analysed using the procedure

1. (ODi)t, the mean OD of the six repetitions of inoculatedmedia;

2. (ODni)t, the mean of the OD of the four repetitions of non-inoculated media;

3. (DOD)t=(ODi)t-(ODni)t;

4. log10[(ÄOD)t/(ÄODmin)] where ÄODmin was the lowestDOD value above the detection threshold.

As indicated by Cheroutre-Vialette et al. (1998), a modifiedGompertz equation (Zwietering et al. 1990) was used to fit thegrowth curves log10[(DOD)t/(DODmin)]=f(t). Growth parameters


such as A (logarithmic increase of population), LT (lag time) andµ (maximal growth rate) were determined by non-linear regressionwith STAT-ITCF statistic software. GT (generation time) was derivedfrom µ using the relatio: GT=log10(2)/µ.

In standard and limiting conditions, the time of inoculationwas taken as zero time when considering the growth curve. Inshock conditions, the time of the addition of the shock solutionwas taken as zero time in the calculation of growth parameters.

Model Predictions

The ratio of predicted and observed generation times wascalculated for the studied combinations (predicted generation time/observed generation time). One complementary measure, proposedby Ross (1996) as simple index of the performance of models inpredictive microbiology, was also evaluated. This value, termedthe ‘bias factor, may be interpreted as the average ratio of thepredicted and observed values and is defined as follows:

where y predicted is the predicted generation time, y observedis the observed value, and n is the number of observations usedin the calcu-lation. Perfect agreement between predictions andobservations will lead to a bias factor of 1.

Results

Growth kinetics of L. Monocytogenes

The abrupt addition of 8% NaCl has a marked effect on thegrowth recovery of the micro-organism compared to the effectproduced by the presence of NaOH to alkaline pH 9.5. The growthcurve asso-ciated to the combined alkaline-osmotic shock solutionsdiffered from those of uncombined shocks suggesting an interactiveeffect of NaOH and NaCl at 10°C. The effects of acid shock (pH6.3), osmotic shock (8% NaCl) and combined acid (pH 6.3) - osmotic(8%) shocks at 10°C on the exponential phase of L. monocy-togenes14 growth. The acid-osmotic variable conditions have particu-larlyaffected the growth recovery of the strain. Indeed, the organic acidand the solute pre-sented an interactive effect on the growth ofmicro-organism.

In shock condition, L. monocytogenes 14 respondedinstantaneously to small changes of pH and Ate,. Small acid and

acid-osmotic variations involved an unusual growth curve asshown by Fig. 3. Indeed, consecutively to these shocks, the growthof micro-organism contin-ued and a rupture of the sigmoid curvewas seen before the growth recovery. The Gompertz equationfitting left this phenomenon out of account and considered thegrowth recovery cor-responding to the generation time values. Inexchange, large environmental conditions dur-ing exponentialphase (i.e. shock condition) of the micro-organism induced a lagphase before the growth recovery. This lag phase was morepronounced especially as the concentrations of shock solutionswere higher in case of alkaline-osmotic shocks. This induced lagperiod was really observed and cal-culated when a highconcentration of NaCl, e.g. 6.8 and 8%, was added with the organicacid in the culture.

Some growths were particularly affected by the exposure tothe combined shock solutions in exponential phase. Indeed, noincrease of op-tical density during the experimental period, i.e. 21days, underlying no increase of population, was noted for thecombination pH 5.6 and 4 or 8% NaCl.

An adaptation period to the new environ-ment (temperature,pH and water activity) was observed in limiting condition, i.e.when the shock solutions were present at the begin-ning of theculture. This period reached 12 days if the cells were placed inpresence of higher concentrations of NaCl and alkaline pH, 8%and 9.5, respectively (data not shown). An experimental condition(pH 5.6, 8% NaCl) caused no increase of optical density during theexperimental period. The generation times calculated with theGompertz equation revealed that growth re-covery obtained inlimiting conditions was dif-ferent to those observed in shockconditions. Indeed, the shock condition pre-sented a generationtime value higher than calculated in limiting condition whateverthe environmental variation.

In conclusion, two principal pheno-mena were observed whenthe bacteria sub-mitted to abrupt change of pH and Aw (i) largeenvironmental variations induced a lag phase following theexposure to shock solutions, and (ii) the growth continued witha generation time value different from that observed before thechange.


Polynomial Models Predictions

The aim of this part was not to com-pare polynomial models,but to show the position of their predictions with regard to variableenvironmental conditions. The range of predicted values was wide.Considering the limiting condition, the ratio calculated was ingeneral less than 1.0, indicating that the pre-dicted value wasinferior to those observed.

Table : Generation times, expressed in h, observed in shock or limiting conditions. In brackets: asymp- Totic 95% confidence interval

Experiment number

pH % NaCI Generation times

Limiting conditions

Shock condition

22

5.6 0 26.6 (±0.4) 64 (±2)

2 6.3 0 9.6 (±0.1) 14.0 (±0.2) 15

5.8 1.2 19.9 (±0.3) 68 (±6)

14

6.8 1.2 6.8 (±0.1) 9.2 (±0.3)

5 5.6 4 120 (±1) NI 8 6.3 4 19.9 (±0.3) 41(±1) 9 6.3 4 20.6 (±0.3) 39 (±1) 19

5.8 6.8 190 (±2) 215 (±3)

18

6.8 6.8 10.2 (±0.1) 18.0 (±0.3)

25

5.6 8 NI NI

11

6.3 8 56.8 (±0.6) 129 (±2)

20 (standard condition)

7.0 0 5.1(±0.1)

3 7.0 4 6.3 (±0.1) 9.5 (±0.1) 24

7.0 8 9.3 (±0.1) 17.9 (±0.2)

a: Growth parameter was calculated according to zero time considered as the time of inoculation. b: Growth parameter was calculated according to zero time considered as the moment of shock. NI: No increase in optical density during 21 days.

The predictions were more especially conser-vative as theculture conditions were severe, i.e. low pH and high saltconcentration. Few plots were above the identity line indicatingthat the predicted parameters were higher than the observed valuesin limiting conditions. This was observed for growth predictionsassociated to combined alkaline pH-osmotic conditions oruncombined osmotic conditions. In cases of growth predictionsunder low pH (pH < 6.3), the predictions values were in generalthree or four times be-low the observed values and might be 14times inferior in presence of low pH and higher salt concentration,i.e. pH 5.8 and 6.8% or pH 6.3 and 8%. Considering the shockcondition, the trend of models to have ‘fail safe’ predictions wereconfirmed and increased. All the points were below the identityline, representing predictions which were shorter than the observedgeneration time. These results were confirmed by the bias factorswhich maybe interpreted as the geometric mean ratio of predictedand observed generation times. The bias factors calculated inshock condition showed, on average, 1.5-fold greater bias thanthose observed in limiting condition.

Discussion

For several years, it was largely demonstrated that refrigerationin conjunction with other factors, e.g. salt or acid, may severelyretard or prevent the growth of micro-organisms. As L.monocytogenes is common in the food proces-sing environment andas a food product is often exposed to environmental variations, itis important to evaluate the adapta-tion of this bacteria to variableconditions of environmental factors as pH and Aw at lowtem-perature. A previous study, Cheroutre-Vialette et al. (1998)has reported that the generation times calculated for the growthsof L. monocy-togenes in shock and limiting conditions under variablepH or Aw conditions at 20°C were sig-nificantly different.Moreover, an additional lag time in shock condition could beobserved. The results obtained in the present study car-ried out atlow temperature and under com-bined pH-Aw conditions were inagreement with the conclusions established in this pre-vious study.

The aim of this paper was to introduce more advances into thefield of predictive microbiology, by specifying the microbial


behavior under dynamic environment of pH and water activityand collecting experimental data. In this way, the use of BioscreenC allowed to rapidly ex-plore the microbial impact of varying pHand Aw conditions. In recent years, the interest in developingdynamic mathematical models that describe the growth ofmicroorganisms in the presence of environmental factor variationshas increased. With the objective to take into account the effectsof temperature variations on microbial growth, Baranyi et al. (1993)observed that the abrupt transitions could lead to adjustmentperiods, i.e. microbial cultures, when shifted abruptly from onetemperature to another, may exhibit a transient growth rate beforeassuming the growth rate expected at the new temperature.

Table : Comparison of model predictions of the data PMP FMM Lm14

Limiting Shock Limiting Shock Limiting Shock na 15 14 15 14 9 8

Ratiob average 0.53 0.33 0.55 0.34 0.62 0.43 Ratio minima 0.07 0.06 0.05 0.05 0.08 0.12 Ratio maxima 1.15 0.60 1.19 0.63 1.25 0.82 Bias factor 0.40 0.25 0.39 0.25 0.50 0.33

a: n ˆ number of observations used in the calculation.

b: ratio ˆ predicted generation time/observed generation time.

Our data under higher pH and/or Aw, variations confirmedthe pre-sence of adjustment period which was repre-sented by aninduced lag phase. When these environmental changes weresmaller, an unusual growth curve was observed. Thischar-acteristic raised some questions about its origin. This couldbe assigned to the material used, the Bioscreen C. The lateralagitation of the plates in Bioscreen C could have been insufficientfor homogenous distribution of the compound in the medium. Atthe moment of shock (acid or acid-osmotic), a time is needed forthe repartition of the shock solutions in the well. A study ofJorgensen et al. (1995) has also reported that the cells of L.monocytogenes incubated in media containing NaCl became rapidlylonger than cells grown without salt. Such morphological changescould also lead to variations in the turbidity of the medium.

The experiments of the present study showed also that therecovered growth rate after the change of pH and/or Aw wasusually less than those associated to the new environment. Other

investigations of bacterial growth pre-dictions with changes intemperature by Zwie-tering et al. (1994) andVan Impe et al. (1995),or with changes in pH by Rosso (1995) admitted the possibletransposition of results obtained from constant conditions tovariable condi-tions. Considering our results, such postulate isnot suitable to take into account the physio-logical response ofmicro-organisms.

The results of this study highlighted the pro-blems associatedwith the variable conditions and the available models establishedin con-stant conditions. The results obtained with the predictionsestimated with the models (FMM, PMP, Lm14) are not surprisingtaking into con-sideration that models were developed with strainsof different origins, in media of various compositions and indifferent conditions of growth. For example, the pH was adjustedwith HCl in PMP or Lm14, lactic acid in FMM. Acetic acid wasused for pH regulation in our growth experiments and it was wellestab-lished that this organic acid presents a higher inhibitoryactivity on L. monocytogenes. As can be pointed out by this study,the predictions of these models based on data generated underconstant conditions can reliably predict growth under fluctuatingconditions of pH and Aw. But, a systematic over-prediction wasrevealed under dynamic environment. Under a dynamicenvironment, the bias factor showed that the recovered growthrates caused much larger generation times than those underconstant conditions. In consequence, the structure of these modelsseemed to be inappropriate to take into account the microbialresponse to variable en-vironmental conditions.

Moreover, the induced lag phase observed in dynamicenvironment of pH and Aw could not be integrated by these models.The necessity to determine a new way to model changing pHvalues, varying Aw or other food paramenters was underlined.Neuronal networks could represent a new approach to take intoaccount the varia-bility of cell response in variable conditions andproduce a dynamic model. A previous study provided a goodprediction of L. monocytogenes growth under variable conditions.Such dynamic models could follow the microbial impact of thedifferent steps associated with the production, distribu-tion andretailing of a food.


EFFECTS OF PH OR AW STRESS ON GROWTH OF LISTERIAMONOCYTOGENES

The growth of three strains of Listeria monocytogenes at 20°C ina meat broth of different pH or water activity was investigated. Atinoculation or at the beginning of the exponential phase, cellswere exposed to stress by the addition of NaOH or NH4

+, aceticacid, NaCl or KCl, in order to reach a pH of either 9.0 or 5.6, oran aw of 0.950 or 0.965, respectively.

The effects of the exposure to stress on the generation and lagtimes of each strain were analysed by turbidity measurements forcultures in micro-titer plates. Results were confirmed by conductingthe same experiments in a fermentor, except for the maximalpopulation reached. The three strains showed similar behaviour.Cells were able to overcome the alkaline stress rapidly whereasacid and osmotic shocks induced important changes of the growthparameters. Cells exposed to acid or osmotic conditions from thetime of inoculation were less affected than cells exposed at thebeginning of the mid-exponential phase.

Listeria monocytogenes, a psychrotrophic bacteria, is consideredas an important food-borne pathogen. It can persist and grow atlow pH, high pH, low water activity and at refrigerationtemperatures.

Work on predictive microbiology has been carried out in orderto improve the shelf life and safety of foods. Predictive models ofmicrobial growths were set out with respect to the main controllingfactors of the environment such as temperature, pH, and wateractivity.

These factors were often considered constant during growth.But, environmental factors, particularly the temperature, may varyextensively throughout food processing. To take the environmentalvariations during time into account, different authors haveimproved predictive models and proposed dynamic models whichdescribe growths at varying temperature profiles.

In this study, we report the effects of water activity and pHshifts on the growth of three strains of L. monocytogenes usingdifferent osmotic solutes (NaCl, KCl), organic acid (acetic acid)and bases (NaOH, NH4

+).

Monitoring of Bacterial Growth

An automated turbidimeter (Bioscreen C, Labsystem, LabsystemFrance SA, Les Ulis, France) was used to follow the growth ofListeria strains in the micro-titer plates. Optical density (O.D.) wasread at a wavelength of 600 nm.

A two-litre fermentor SET2M was used. The O.D. of the growthwas measured with a spectrophotometer (UV-160A, Shimadzu,Japan) at 600 nm.

Microorganisms

The strains used by Bégot et al. (1997) in the studies withBioscreen C were chosen for this work: L. monocytogenes CLIP19804 isolated from meat products, L. monocytogenes 14 obtainedfrom the food processing environment and L. monocytogenes CAWisconsin associated with a cheese-borne listeriosis outbreak. Allwere serotype 4b which was involved in the recent epidemics inFrance. L. monocytogenes 14 strain was also used in the fermentorstudies. Stock cultures were maintained on TSA agar slopes andstored at 4°C.

Media

For preparation of inocula, cultures on TSA were subculturedin meat medium (MM) which contained meat peptone 10 g/l,yeast extract (Difco) 5 g/l and glucose 5 g/l.

The culture medium for growth studies was a tryptic meatbroth (TMB) (Fournaud et al., 1973). It was buffered with aK2HPO4–KH2PO4 (Merck) 0.1 mol/l solution in proportion 1:1 (v/v) and adjusted to pH 7.0 with NaOH (40 g/l).

Types of Experiment

For all experiments, three conditions were studied: standard,limiting and shock conditions. Under standard and limitingconditions, bacteria were grown in TMB or in TMB adjusted to thedesired aw or pH values. The following additions were made toTMB: NaOH (solution of 200 g/l; Prolabo) 12.2 g/l or NH4

+ (Merck)7 g/l to obtain pH of 9.0, acetic acid (Carlo Erba, Nanterre, France)4.1 g/l to obtain a pH of 5.6, 80 g/l of NaCl (Prolabo) or 80 g/l of KCl (Prolabo) (Jakobsen et al., 1972) to obtain aw=0.95 and a


pH of 7.0 or aw=0.965 and pH of 7.0, respectively. The shockconditions were defined as follows: the bacteria were grown instandard TMB until the beginning of the exponential phase andwere then exposed to the pH and aw values described above.

Growth in Microplates

Each strain was incubated on TSA slopes for 7 h at 37°C andthen transferred to MM that was incubated in a rotary shakerwaterbath (Aquatron, Infors, Switzerland) for 18 h at 20°C, bywhich time growth of all strains had reached the stationary phase.

A proportion of the culture was inoculated in the variousmedia to give a concentration of about 5.107 CFU/ml in order tobe above the detection threshold of the Bioscreen C. For theinoculated TMB medium, 300 l were dispensed in 6 wells and thesame volume of non-inoculated medium was dispensed in 4 wellsin order to determine the O.D. of the growth medium and to detectpossible contamination. The well numbers were considered asreplicates. The same procedure was used for the pH and awadjusted TMB. The shock exposures were arranged as follows: 300l of inoculated and 300 l of non-inoculated TMB were dispensedin 6 and 4 wells, respectively. An aliquot (100 l) of concentrated(×4) stock solution or standard TMB was added when the O.D.was near 0.2. The delay in adding the solutions to 70 wells wasapproximately 7 min. These plates were incubated in the BioscreenC previously adjusted to 20°C.

Growth in Fermentor

The protocol was similar to that in Bioscreen C except for thefermentor volume with additions adjusted accordingly. Theadditions lasted about 10 min. The experimental conditions were20°C with a shaking speed of 100 rpm and an aeration regulatedat 0.2 litre of sterile air per litre of medium per min.

Data Analysis

Averages of the O.D. were calculated for the six repetitions ofinoculated media and for the four repetitions of non-inoculatedmedia. The data were then analysed using the procedure describedby Bégot et al. (1996). Four quantities were calculated at time t:

1. (O.D.i)t, the mean of the O.D. for the 6 repetitions ofinoculated media;

2. (O.D.ni)t, the mean of the O.D. for the 4 repetitions of non-inoculated media;

3. (DO.D.)t=(O.D.i)t–(O.D.ni)t;

4. log10 [(DO.D.)t/(DO.D.min)] where DO.D.min was the lowestDO.D. value above the detection threshold.

A modified Gompertz equation (Zwietering et al., 1990) wasused to fit the growth curves log10 [(DO.D.)t/(DO.D.min)]=f(t).Growth parameters such as A (logarithmic increase of population),L (lag time) and (maximal growth rate) were determined by non-linear regression with STAT-ITCF software. Tg (generation time)was derived from µusing the relation:

gT 10log (2)

µ=

In standard and limiting conditions, the time of inoculationwas taken as zero time when considering the growth curve. Inshock conditions, the time of the addition of the shock solutionwas taken as zero time in the calculation of growth parameters.

RESULTS

The growth curves of L. monocytogenes 14 in the presence ofacetic acid and NaCl can be seen in Fig. 1a and b, respectively.When the shock treatments (acid and osmotic) were applied afterthe beginning of growth of L. monocytogenes in standard media,lags were shorter, but generation times increased, as comparedwith cells grown in limiting media. This phenomenon wasparticularly pronounced for the acetic acid example.

In comparison with growth in standard media, growth of thestrain was not really affected by the presence of NaOH and NH4

+

whatever the treatment, limiting or shock conditions. The growthparameter values, i.e. generation and lag times, confirmed thatcells adapted rapidly to the alkaline upshift to pH 9.0 with NaOHor NH4

+. Indeed, these values calculated in alkaline media wereconsistent with those found in standard media. Under acid andosmotic conditions, cell growth was more affected. Media


containing acetic acid was the most severe culture condition forthe growth of the three strains. Under limiting conditions, althoughthe presence of acetic acid did not cause increasing lag times,generation times were at least 3.5 times greater compared to thevalues in standard media. The solute used to control aw caninfluence the growth pattern of the microorganism. At the testedconcentration (80 g/l), L. monocytogenes which is also consideredas a salt-tolerant microorganism, tolerated KCl better than NaCl.Indeed, the generation and lag times in limiting and shockconditions, were higher in the presence of NaCl 80 g/l. Thisinfluence of NaCl can be due to a larger aw effect.

Generally L. monocytogenes 14 and 19804 had similargeneration times whereas L. monocytogenes CA had longergeneration times under limiting conditions. Variations in lag timewere noted among the strains in all tested conditions.

Generation and lag times of cultures in the fermentor arepresented in Table 3. The values of generation times were similarto those obtained in Bioscreen C, but large variations in lag phaseswere noted. The results obtained from the fermentor cultureexperiments confirmed the observations made with the Bioscreen C.

DISCUSSION

In previous studies about growths in non-variable conditions,the interest in using the Bioscreen C for analysis of microbialgrowth parameters and its good reproductivity was established.Our study confirmed these observations. The results obtained fromthe experiments carried out in the automated turbidimetric systemwere in agreement with those obtained in the fermentor. The increasein the number of L. monocytogenes was underestimated in theBioscreen C as compared to the fermentor. The homogeneousaeration and agitation which were controlled in the fermentormight be a possible explanation for this observation. Nevertheless,the differences found in the generation times between the twomethods were not significant. This suggests, therefore, that theBioscreen C is suitable for studying the growth of bacteria undervariable conditions. A high concentration of cells is required inthe Bioscreen C for a detectable change in absorbance, butBuchanan and Phillips (1990) and Dalgaard et al. (1994)

demonstrated that the growth kinetics of L. monocytogenes wereunaffected by the size of the initial inoculum i.e., the generationand lag times derived from the Gompertz equation were not affected.

The L. monocytogenes strains showed similar responsesfollowing aw or pH stresses. Variations were generally larger inlag times than in generation times. There was very little differencein generation times between L. monocytogenes strains 14 and 19804for most of the experiments. Any major variations were observedwith the most severe culture conditions, i.e., in the presence ofacetic acid. L. monocytogenes strain CA was more affected by thedifferent stress conditions and in particularly under limitingconditions. These findings are consistent with those ofPapageorgiou and Marth (1989) who also reported that L.monocytogenes CA was less salt tolerant than strains in whey andskimmed milk containing 12% NaCl.

In recent years, the interest in developing dynamicmathematical models that describe the growth of microorganismsin the presence of temperature variations, has increased. Whenbuilding dynamic models to predict bacterial growth with changein temperature, two hypotheses have been formulated: (i) thebacteria show no additional lag phase due to a change intemperature during the exponential phase, and (ii) growthcontinues immediately at the specific growth rate associated withthe temperature post-shift. More recently, for the elaboration ofdynamic models, Van Impe et al. (1995) and Rosso (1995) acceptedas a postulate, that variable environmental conditions do notinduce stress conditions on a microbial population, i.e., theyconsidered that the microflora responds instantaneously to pH(Rosso, 1995) or temperature changes (Van Impe et al., 1995 andRosso, 1995). Our tests carried out with the three L. monocytogenesstrains showed that growth under modified osmotic or acidenvironmental conditions in exponential phase did not agree withthese hypotheses. Since 1992, Van Impe et al. (1992) haverecognized the need to take into account the previous history ofa product, and therefore the previous history of a strain whichcould have an important impact on the lag time prediction of amicroorganism. It is also necessary to quantify the ability of amicroorganism to grow under variable conditions. This could


allow the optimal combination of temperature, pH, aw and otherfactors with the purpose of increasing the shelf life of a givenproduct.

VALIDATING THE USE OF GREEN FLUORESCENT-MARKED ESCHERICHIA COLI O157

Monitoring bacterial kinetics in food is of great importance infood safety. The targeted micro-organism has to be identifiedaccurately among competitive flora. Using green fluorescent protein(GFP)-transformed strains is a possible answer to such issues.However, quantitative studies require that this transformation doesnot alter the micro-organism behaviour: parent and transformedorganisms were thus compared.

Three Escherichia coli O157:H7 strains were transformed usinga GFP-plasmid expressing. Parent and transformed strains werecompared according to their genetic characteristics and serotypes.Growth ability was also assessed in constant and fluctuatingtemperature profiles. Cardinal values of pH, water activity andtemperature were computed. No differences were observed betweenparent and transformed strains for all these experiments. Theplasmid was satisfactorily maintained within transformed strainsthroughout the studies. Growth was eventually monitored in beefmeat.

Although Escherichia coli is a common inhabitant of the humanand animal gastrointestinal tract, several pathogenic types of thespecies are able to cause a variety of human diseases. Firstrecognized as a human pathogen in 1982, E. coli O157:H7 [Shigatoxin-producing E. coli (STEC)] has since been associated withoutbreaks of food-borne illness in countries across the world, withmany foods identified as vehicles of infection. The main virulencefactor of STEC is the ability to form cytotoxic exotoxins. Two majorcategories of Shiga toxins have been distinguished, Shiga toxin 1(stx1) and Shiga toxin 2 (stx2). The other virulence factors are theproperty of producing attachment-effacement lesions and thepresence of an entero-haemolysin gene.

Given the severity of the illness associated with E. coli O157:H7,it is necessary to provide the food industry with procedures forpreventing its presence and controlling its growth. The aim of

predictive microbiology lies in predicting growth kinetics of micro-organisms at a given moment during food processing and thuspredicting the evolution of food during storage, manufacturing orpreservation accidents. Rosso et al. (1993, 1995) proposed, in theircardinal approach, a model in which the maximum microbialspecific growth rate ( max) is a function of temperature or pH. Thecardinal values of the considered strain are then determined. Asan example, for temperature values, the following parameters areconsidered: T min (the temperature below which growth is nolonger observed), T max (the temperature above which no growthoccurs) and T opt (the temperature at which the maximum specificgrowth rate max equals its optimal value opt). The same procedurecan be applied to pH and water activity a w. The minimal andmaximal values are important factors to food preservation, andthe optimal value is a crucial parameter in predictive microbiology.In this approach, for a given strain, opt is a food-dependentparameter.

The development of sensitive methods for monitoring bacteriain foods, especially in presence of natural contaminating flora, isof marked importance. Indeed, models have to be validated and itis usually more appropriate to collect new data in food to have agood validation. In this way, the objective of this work was firstlythe construction of E. coli O157:H7 strains marked with the greenfluorescent protein (GFP) gene carried by a plasmid. The GFP fromthe jellyfish Aequorea victoria, a versatile 27-kDa protein, has provedto be valuable as a tool for studying a variety of biological questions,e.g. to study gene expression, to determine protein distribution orto tag a cell lineage in vivo, in situ and in real time. The GFP emitsgreen light when excited with ultraviolet (u.v.) radiation and nosubstrate or cofactor is required for fluorescence.

Ajjarapu and Shelef (1999) used GFP-bearing E. coli O157:H7in ground beef and were able to monitor its evolution despite thepresence of a background microflora. It should yet be verified thatusing GFP-transformed strains instead of nontransformed strainsis a right way, i.e. transformed strains and parent strains shouldbehave in a similar manner. Fratamico et al. (1997) found nodifferences according to morphological or biochemical criteria orusing PCR. In a second part of the present work, the influence of


temperature on the growth kinetics of both GFP-E. coli strains andtheir parental strains was assessed in liquid medium. GFPexpression was evaluated during growth. Cardinal values of pH,a w and temperature of GFP or parental strains were determinedcomparatively. Furthermore, experiments at varying temperatureprofiles were carried out in order to simulate temperature-changingconditions (food processing, food storage, etc.). The results obtainedin liquid medium were complemented by experiments in food.

Strains and Media

Three E. coli O157:H7 strains were used in this study. One ofthe strains was isolated from bovine faeces and will be referred toas strain 1, the other two were of clinical origin and will bereferred to as strains 2 and 3. These strains were transformed witha plasmid vector pBAD-GFPuv (BD Biosciences - Clontech Inc.,Palo Alto, CA, USA), adapted for visualization under u.v. light.Stock cultures were maintained at 80°C in cryobank.

All strains were resuscitated before use by inoculation into10 ml of brain-heart infusion (BHI) broth (Oxoid Ltd, Basingstoke,Hampshire, UK), followed by incubation at 37°C for 8 h. Thesubculture and culture medium was BHI (Oxoid) supplementedwith yeast extract (Oxoid) 3 g l 1 and glucose 2 g l 1. The BHIsupplemented medium was sterilized by filtration (0·20 m poresize). This will be referred to as modified BHI.

The count medium was plate count agar for the E. coli strains.Plates were incubated at 37°C for 24 h.

Construction and Identification of GFP-expressing E. coli

Escherichia coli strains were transformed with pBAD-GFPuvby the calcium chloride method. The pBAD-GFPuv plasmid vectorcontains an arabinose-induced promoter and an ampicillinresistance gene. Transformants were selected by plating on PCAsupplemented with arabinose 2 mg ml–1 and ampicillin 50 µg ml–

1 (PCA + Ara + Amp). Colonies on PCA + Ara + Amp, that wereinduced by arabinose and ampicillin resistant, were fluorescentunder u.v. light. Stability of fluorescence was monitored. Indeed,the percentage of fluorescent colonies before and afterexperimentations was also examined throughout the present study.

Parent and transformed strains were compared according totheir genetic characteristics (stx genes), serotypes and biochemicalcharacteristics. Detection of stx genes was performed withdegenerate primers ES149 and ES151 as described by Read et al.(1992). These primers amplified a conserved sequence of stx 1 andstx 2 genes. For DNA preparation, 1 ml of an overnight samplebroth (BHI) was centrifuged at 12 000 g for 3 min. The pellet waswashed in PBS buffer (pH 7·4; Sigma). The DNA from pelletedcells was released by boiling and purified using Instagen DNApurification matrix. Ten microlitres of the DNA were used astemplate for PCR detection of stx genes. The conditions for thePCR amplification were the same as those described by Uyttendaeleet al. (1998).

Serotyping of the different strains was performed for O157and H7 determination with Difco antisera.

Parent or transformed strains were biochemically confirmedby using the API 20E test strips.

Growth Kinetics as Function of Constant or Dynamic Temperaturein CARY 100

Behaviour of parent and GFP-expressing strains in function oftemperature was compared in modified BHI. Parent andtransformed strains were grown simultaneously in aspectrophotometer CARY 100. Three replicates were carried outfor each condition.

After resuscitation (in modified BHI, 8 h at 37°C), the parentor GFP-expressing strains were subcultured in modified BHI andin modified BHI + ampicillin (50 g ml–1), respectively at the definedtemperature until the beginning of exponential phase. A 0·5% ofthis inoculum, corresponding to inoculum size of 106 CFU ml–1,was then transferred to the culture medium (modified BHI) forgrowth assessment. Growth was monitored in constanttemperature conditions at 37, 20 and 10°C. Moreover, growthevaluation was carried out in dynamic temperature conditions asfollows: the strains were first placed at a start temperature untilbeginning of exponential phase and then, a new temperature wasapplied. Studied conditions were: 37°C followed by 20°C, 37°Cfollowed by 10°C, 20°C followed by 10°C and 10°C followed by


20°C. The temperature gradient was usually achieved in <15 min.In order to check fluorescence, plate counts (PCA + Ara) wereperformed at the beginning and at the end (cells in stationaryphase) of every experiment.

Determination of Cardinal Values of pH, a w, Temperature inBioscreen C

The methodology was defined in the framework of the FrenchPredictive Microbiology Project, Sym’Previus as described byMembre et al. (2002). The detection time method was used tocalculate max for different conditions of temperature (tested range:8-45°C), pH (adjusted by HCl: pH 4-7) or a w (regulated by NaCl:0·5-8%) in a Bioscreen C. Briefly, this method is based on the factthat successive binary dilutions produce growth curves that areshifted from each other by one-generation time value.

The obtained values of µmax allowed determination of cardinalvalues (T min, T opt, T max, pHmin, pHopt, pHmax, a w min, a w opt, a w

max) and µopt with Rosso secondary model. Cardinal values a w maxwere set to 1. Cardinal values pHmax were chosen so that pHmodel would be symmetric, i.e.:

pHopt = (pHmin + pHmax)/2

pHmax = 2pHopt – pHmin.

The cardinal values of pH, a w and temperature were obtainedin modified BHI for the parent strain 2. For the transformed strain2, these cardinal values were determined in modified BHI + Ara.The determination of fluorescence percentage was studied asdescribed before.

Experiments in Food

In order to evaluate the potential utilization of GFP strain infood product, growth of GFP-transformed strain 2 was monitoredin raw ground beef at 10°C. The strain was first subcultured inmodified BHI + Amp at 37°C for 8 h; a second subculture wasconducted in the same medium at 10°C for 4 days.

The suspension was diluted to produce an inoculation levelin beef of ca 2·5 103 CFU g 1. Inoculated meat was divided into10 g aliquots and placed in sterile bags.

A growth curve was produced with 10-15 points and repeatedonce. On each measurement time, one bag was opened forenumeration. Meat was placed in 90 ml of tryptone salt broth, andthen homogenized for 1 min. Dilutions were made in tryptone saltbroth to obtain appropriate levels. Plating was made on PCA fortotal bacterial count, on PCA + arabinose (0·2%) (PCA + Ara) forvisualization of the GFP strain under u.v. light, and onPCA + arabinose (0·2%) + ampicillin (50 g ml–1)(PCA + Ara + Amp) as a selective medium allowing expression offluorescence.

Statistical Analysis of Data

Analysis of variance, regressions or confidence intervalscalculations were performed using software S-Plus 2000. Graphicaloutputs were also produced with Microsoft Excel 2000.

The primary model chosen in this study was the model ofRosso (1995). It was used for constant and dynamic temperaturestudies and for food experiments, in order to calculate primarymodel parameters (mainly lag time and growth rate max). A cardinalvalues model was chosen as secondary model.

RESULTS

Construction and Identification of GFP-expressing E. coli

pBAD-GFPuv was successfully introduced into the three E.coli O157:H7 strains. Colonies of the transformed and parent strainshad identical morphological characteristics on PCA + Ara + Amp(for GFP strains) and PCA (for parent strains), except that coloniesof transformants appeared green under u.v. visualization, whereasthose of the parent strains remained nonfluorescent.

The comparison of the parent and transformed strain geneticcharacteristics, i.e. detection of stx genes, showed that virulencegenes were present after the plasmidic transformation. Furthermore,no serotypic (i.e. O157:H7) or biochemical difference was recordedbetween parent and transformed strains.

Stability of fluorescence was also monitored. Thetransformants were cultured at 37°C in the nonselective liquidmedium, i.e. without ampicillin. The plasmid was maintained inmore than 60% of colonies after 13 days of overnight cultures.


Growth Kinetics as Function of Constant or Dynamic Temperature

GFP-expressing strains were compared with their respectiveparent strains according to growth rate and lag time values(calculated from Rosso primary model) under various temperatureconditions. Correct agreement between the strain results wasobserved.

Loss of GFP expression was also examined for transformedstrains. Colonies plated on PCA + Ara were counted, and anothercount was made under u.v. light in order to evaluate the proportionof fluorescent colonies. The percentage of nonfluorescent coloniescorresponded to plasmid loss within the bacterial population.Mean plasmid loss percentage was under 10% in all the cases.The maximal range of observed plasmid loss was 0-21%. Theseobservations indicated of a correct GFP expression, and thereforea conservation of the plasmid at whatever the environmentaltemperature conditions. A constant temperature of 10°C gave lessfavourable results, but experiments were longer (5 days). Strainsseemed to keep their fluorescence better in dynamic conditions(rapid temperature changes) of growth.

Cardinal values of a w, pH, Temperature

Cardinal values of a w, pH and temperature for strain 2 areindicated in Table 2 with their confidence intervals. As shown bythese results, the confidence intervals between the parent and thetransformed strains are overlapping. This trend confirmed theprevious observations, showing no differences between the parentstrain and the GFP-expressing strain behaviour. It must be notedthat the pHmax was estimated by a symmetric model and noexperiment was conducted to alkaline value. This could explainthe fluctuation of calculated pHmax values.

The loss of GFP-expression for transformed strain 2 wasevaluated for all tested conditions of a w, pH and temperature. Themeans of the plasmid loss percentage, as well as the minimal andmaximal percentages, were calculated and indicated in Table 3.Whatever the environmental conditions, strain 2 exhibited a goodconservation of the plasmid. In general, the plasmid loss wasbelow 10%. This value could be more pronounced for the mostsevere culture conditions, e.g. in presence of the highest salt

concentration, the lowest acid pH or the highest temperature. Butthe percentage never exceeded 33%.

Growth in Beef Meat

The potentiality of use of GFP-expressing strain in foodproduct was evaluated. A challenge-test with the transformedstrain 2 was carried out in raw ground beef at 10°C. High countsof a competitive flora, up to 105 CFU ml–1, were initially found inmeat, which represented particularly hard condition to follow thegrowth of a specific E. coli strain. The total flora was counted onPCA, and GFP strain on PCA + Ara and PCA + Ara + Amp (as aselective medium). The evolution of the GFP strain was easilydistinguished from competitive flora thanks to its fluorescence: itssurvival could be monitored throughout the experiment. The resultsobtained in the two plating media were similar.

DISCUSSION

Using GFP transformation allowed for successful applicationsin assessing various biological issues. In the present work, theconstruction and evaluation of E. coli strains carrying the GFP-expressing plasmid pBAD-GFPuv are reported.

The presence of this plasmid does not affect the intrinsiccharacteristics of the E. coli strains, i.e. the serotypic or biochemicaltraits and virulence genes. No significant behaviour differencebetween marked strains and parent strains could be found, asindicated by overlapping confidence intervals. A major limitationof the GFP marker is the small number of studies on the influenceof environmental conditions on the GFP expression, and moreparticularly, for a plasmid marker.

Our results have demonstrated the stability of the marker inunfavourable environment, i.e. in presence of acid, salt and incooling/heating environment. Indeed, the study of the transformantstrain cardinal values showed that the GFP persists in large rangesof pH values (4·0-7·0), NaCl concentrations (0·5-8%) andtemperature (8-45°C). Furthermore, the plasmid stability wasconfirmed in dynamic conditions of temperature, where the extentof variation could reach 27°C. It is well known that this factor mayvary extensively throughout food processing. The stress induced


by the abrupt change of temperature did not increase the plasmidloss.

Most foods are complex systems with heterogeneous microbialpopulations. The introduction of the GFP marker provides a simplemethod to differentiate between inoculated strains and naturalcompetitive flora of the product. Indeed, in the experiments onfood described above, we could rapidly distinguish individualspecies within the natural flora of the product. GFP labelling inmicro-organisms makes it a marker of choice for ecological studies.Furthermore, the stability of the plasmid in the absence of selectionobviates the need for the addition of antibiotics to these systemsin short-time experiments (14 days in the tested food experiment).In laboratory medium, GFP-labelled strain could be detected onthe basis of green fluorescence, even after 13 days of inoculation.This observation is in accordance with the previus study ofBloemberg et al. (1997), who examined the stability of a GFP-containing plasmid in Pseudomonas spp.

Marking cells with GFP expressed from a plasmid providesthe flexibility to transform a wide range of strains. Indeed, theconstruction of such strains is relatively easy.

Predictive food microbiology has been focusing on modellingthe microbial responses to food environment, in the interest offood safety and for avoiding spoilage. Predictive models havebeen regularly published to describe the growth of a strain or amixture of strains as a function of the environmental factors offood. A good way of validating a model is to compare its predictionto real data obtained from food products. The standard method ofdata collection is the total viable count, which is a very labour-intensive method. In this way, the use of GFP-marked strain ofproviding a useful system for monitoring the evolution of a definedmicro-organism and even developing among competitive ones,presents strong advantages.

GROWTHS KINETICS COMPARISON OF CLINICAL ANDSEAFOOD LISTERIA MONOCYTOGENES ISOLATES IN ACIDAND OSMOTIC ENVIRONMENT

Comparison of pathogenic bacterial strains of clinical originwith strains of the same species isolated from the environment

may be a valuable tool for microbial risk assessment, especially forfoodborne pathogens. Thus, a number of Listeria monocytogenesstrains responsible for human cases of listeriosis, in relation to theconsumption of contaminated seafood, have been compared with“natural” L. monocytogenes strains isolated from similar seafoodproducts. Complete factorial designs were used to assessquantitatively the growth abilities of four clinical and four seafoodisolates of L. monocytogenes placed in various environmentalconditions. The cells were submitted to acid and osmotic stress asthey were in stationary phase (constant condition) or inexponential phase (dynamic condition). The effects and interactionsof pH (5–7) and NaCl concentration (0.5–8% v/v) were studied attwo growth temperatures (10 and 20 °C). Growth parameters (lagand generation times calculated with Gompertz equation) wereused to compare the behavior of the strains with respect to theconditions of culture. The results indicated an overall weak effectof acid stress alone, whereas osmotic stress clearly affected bacterialgrowth and a synergic effect between these two factors wasobserved. Clinical strains displayed better adaptation than seafoodstrains in stationary phase, however, this difference was not verifiedin exponential phase. Low temperature (10 °C) usually confirmedthe observations at 20 °C, and the differences between clinical andfood strains were more pronounced. Finally, a classification of theeight strains, based on the collected data, showed three groups: (i)seafood strains, (ii) three clinical strains and (iii) the last clinicalstrain, alone due to its high resistance to adverse conditions.

Listeria monocytogenes assumed public health significance as aresult of its presence in foods linked to several outbreaks oflisteriosis in Europe and North America. Most cases of humanlisteriosis occur in immunocompromised individuals, pregnantwomen and the elderly.

L. monocytogenes is a widespread microorganism that has beenisolated from a variety offoods including fish. Although severalkinds of food products have been incriminated as source ofsporadic or epidemic listeriosis cases, the involvement of fish andfishery products is still very rare. In 1992, two perinatal listeriosiscases, which occurred in Auckland, New Zealand, were associatedto the consumption of smoked mussels. Mussels were also involved


in cases occurred in Sydney, Australia. An outbreak caused byrainbow trout was reported in Sweden.

The prevalence of L. monocytogenes in naturally contaminatedseafood was studied by Jorgensen and Huss (1998): the highestprevalence was found in cold-smoked fish (34–60%), while thelowest was found in heat-treated and cured seafood (4–12%). L.monocytogenes has been found in smoked salmon in several studies.Ben Embarek (1994) reported a contamination rate of between 0%and 75%, with an overall prevalence of 10%, in cold-smokedsalmon samples.

In fresh as well as in several lightly preserved seafoodsincluding cold-smoked salmon, none of the processing stepseliminate L. monocytogenes. Furthermore, cold-smoked fish products,which are typically consumed without cooking, are among theready-to-eat foods of particular concern due to the lack of a heatinactivation step during processing. The ability of a microorganismto survive in various foods depends on several combinedparameters in the foods such as temperature, pH, salt concentration(water activity). A range of seafoods, particularly the lightlypreserved products (<6% salt concentration, pH>5) such as smokedfish products (hot and cold smoked), lightly salted products (brinedcooked shrimp) or marinated products, are capable of supportingthe growth of L. monocytogenes (Huss et al., 2000). Moreover, L.monocytogenes is able to survive or grow at refrigerationtemperatures. The pathogen growth was observed in vacuum-packed smoked salmon stored at 10 or 2 °C. It is important to notethat cold smoking is conducted at temperatures below 30 °C, mostfrequently 19–22 °C, for 2 to 3 h and, therefore, provides a possibleenvironment for bacterial growth.

Several studies have shown that the virulence of individual L.monocytogenes strains may differ. Furthermore, Datta (1994)demonstrated that the pathogenicity of this microorganism can beaffected by various substrate factors, i.e. in foods. There are fewerpublished systematic studies in which growths of clinical andfood strains of L. monocytogenes under unfavourable conditions(found in industrial environment) were compared. The comparisonof pathogenic bacterial strains of clinical origin to strains of thesame species isolated from environment may be a valuable tool for

microbial risk assessment, especially for foodborne pathogens.Therefore, it was of interest to evaluate and compare the capacityof growth of clinical and food strains, in order to determine whetherthere was a relationship between strain origin and the growthpotentialities in function of environmental factors.

Consequently, the aim of this work was to investigate thegrowth behaviors of four L. monocytogenes strains responsible forhuman cases of listeriosis in relation to consumption ofcontaminated seafood and four L. monocytogenes strains isolatedfrom seafood products, in function of temperature, pH and saltconcentration. Furthermore, the effects of the pH and saltconcentration variations on the growth of the strains were takeninto account, with the objective to simulate the variations whichcan occur during food processing and induce a stress situation forthe microorganism.

Monocytogenes Strains and Media

Experiments were carried out with eight L. monocytogenesstrains. Four were clinical strains associated to fish or fish productsand four were isolated from seafoods.

Stock cultures were maintained at “80 °C in cryobank. Allstrains were resuscitated before use by inoculations into 10 ml ofBrain Heart Infusion (BHI) broth (Oxoid, Basingstoke, Hampshire,UK), followed by incubation at 37 °C for 8 h.

The subculture and culture medium was BHI (Oxoid)supplemented with yeast extract (Oxoid) 3 g l”1, glucose (Prolabo,Fontenay sous bois, France) 2 g l”1 and buffered with a K2HPO4–KH2PO4 0.1 mol l”1 solution in proportion 1:1 (v/v) to pH 7.0.

Monitoring of Bacterial Growth

An automated turbidimeter was used to follow the growth ofL. monocytogenes strains in the micro-titer plates. Optical density(O.D.) was read at a wavelength of 600 nm. The working volumein each well of the micro-titer plate was 400µl.


The ability of the eight L. monocytogenes strains to grow inosmotic and acid environment was determined as described by


Cheroutre-Vialette et al. (1998). After resuscitation, strains weresubcultured (1%) in the supplemented BHI 18 h at 20 °C, by whichtime the strains had reached stationary phase. 0.5% of thisinoculum was then transferred in the culture medium for thegrowth studies. The concentration of cells in the culture mediumwas about 107 cfu ml”1, in order to be above the detection thresholdof the Bioscreen C. The viable number of bacteria was confirmedby PCA plate counts (AES). Plates were incubated for 24 h at 37°C and enumerated.

For all experiments, three conditions were studied: control,constant and dynamic conditions. Under control condition, cellsin stationary phase (after subculture) were grown in culturemedium BHI, without addition of salt and acid. In constantconditions, bacteria in stationary phase were grown in culturemedium adjusted to the desired aw and pH values. The osmoticvariation was achieved by the addition of NaCl (Merck). A solution5 M of HCl (Merck, 32%) was added to regulate low pH. Thedynamic condition was defined as follows: the bacteria were grownin culture medium (control condition, i.e. without addition ofNaCl and HCl) until the beginning of the exponential phase andwere then shocked by the abrupt addition of shock solutions.These shock solutions (osmotic and acid) were prepared in orderto obtain final values similar to those indicated for the constantconditions. An aliquot (100 l) of concentrated (×4) shock solutionswas added when the O.D. was near 0.2, corresponding to thebeginning of exponential phase.

For each combination, five growth repetitions were carriedout.

Experimental Design

A factorial design was used to assess quantitatively the effectsand interactions of NaCl and pH (HCl) conditions on the growthof L. monocytogenes clinical or seafood strains at two temperatures.The combinations of the following conditions were used:

Temperature (°C): 10, 20 °C

pH: 5, 6, 7

NaCl (% v/v): 0.5, 4, 8

For each combination, the constant and dynamic conditionsof pH and NaCl concentration were studied. At least, for eachstrain, 36 experiments were performed according to 18 growths inconstant condition and 18 growths in dynamic condition.

Statistical Analysis of Data

Experimental data were statistically analysed by the use of S-PLUS 2000 software.

Averages of the O.D. were calculated for the five repetitions ofinoculated media and for the two repetitions of non-inoculatedmedia. The data were then analysed using the procedure describedby Bégot et al. (1996). Four quantities were calculated at time t:

• (O.D.i)t, the mean of the O.D. of the five repetitions ofinoculated media for each combination;

• (O.D.ni)t, the mean of the O.D. of the two repetitions of non-inoculated media for each combination;

• (DO.D.)t=(O.D.i)t–(O.D.ni)t;

• Yt=log10[(DO.D.)t/(DO.D.min)], where DO.D.min was thelowest DO.D. value above the detection threshold.

As indicated by Cheroutre-Vialette et al. (1998), a modifiedGompertz equation was used to fit the growth curves Yt=f(t).Growth parameters, A (logarithmic increase of population), l(lagtime) and µ(maximal growth rate), were determined by non-linearregression (S-PLUS software). GT (generation time) was derivedfrom musing the relation: GT=log10(2)/m.

In control and constant conditions, the time of inoculationwas taken as zero time when considering the growth curve. Indynamic condition, the time of addition of the shock solution wastaken as zero time in the calculation of the growth parameters.

To summarize the similarities in behavior of the strains, ahierarchical clustering method was used. A global data set,including all the conditions of the experimental design, was built.Lag and generation times for the 36 treatments (two temperatures,three pH, three NaCl concentrations, static and dynamicconditions) were used as variables measured on the eight strains.Data were standardized for each variable. The first stage of thismethod is the computation of a distance measure between the


individuals. Each individual is considered as an initial cluster.The two “closest” clusters are merged in a bigger one at each stepof the process. The distance between clusters is calculated usingthe average distance method: the distance between clusters A andB is the average of the distances between the points of A and thepoints of B. The algorithm stops when there is only one cluster left.The distances between merged clusters are retained: this will allowthe program to produce a graph in the form of a classification tree.Each node in the tree appears at a height equal to the distance atwhich the clusters were merged.

Growth of L. monocytogenes Strains in Constant Acid and OsmoticEnvironment

At 20 °C in control condition (i.e. in culture medium BHI), allthe strains, clinical or food isolates, presented similar growthparameters, i.e. lag and generation times.

The pH had negligible effect on the different strains. Indeed,whatever the strain origin, the calculated lag and generationstimes at pH 6.0 were similar to those obtained with the controlgrowths. The growth parameters values of a clinical strain, Lm-C4, and a food strain, Lm-E4 in control condition and in acidconstant condition (pH 6.0 and pH 5.0). The growths at pH 5.0have shown longer lag times. On the other hand, at 20 °C, thestrains were more sensitive to NaCl. The presence of salt (8%)induced longer lag phase and increased the generation time by afactor 2.5, on average, compared to the corresponding controlcondition. All the results obtained at 20 °C showed that nodifferences were observed between clinical and seafood strainswhen testing acid environment alone or osmotic environment alone.On the other hand, at this growth temperature, it was observeddifferences between strains when the two factors were consideredsimultaneously. On the whole, the calculated generation times ofthe clinical strains were lower than those of food strains forpHd”6.0 combined with 4% or 8% NaCl. Among the clinicalstrains, Lm-C1 was less affected by the combined acid and osmoticconditions, more particularly in the most severe culture condition.Variations in lag times were noted among the strains in the testedconditions.

At refrigerated temperature (10 °C), the strains were moresensitive to low pH. The interactive effect of the factors pH andconcentration of NaCl on the L. monocytogenes growth were morepronounced at 10 °C than at 20 °C. Furthermore, the differencesbetween clinical and seafood strains were more underlined,especially as the culture conditions were severe, i.e. low pH andpresence of NaCl. No food strains could grow in medium regulatedto pH 5.0 supplemented to 4% of NaCl, whereas the growthrecovery of the whole clinical strains could be observed. The abilityof Lm-C1 to grow in severe conditions was confirmed by theseresults. No strains were able to grow, i.e. no increase of opticaldensity was observed during the experiment duration (14 days),at 10 °C, pH 5.0, 8% NaCl.

Comparison with Growths in Dynamic Conditions

In dynamic condition, L. monocytogenes strains were subject toabrupt variations of pH and salt concentration during thebeginning of exponential phase at 20 or 10 °C. The stress inducedby these variations could induce a lag phase, i.e. an adaptationperiod to new environment. The duration of this period wasvariable and no relation between this period value (lag time) andthe strain origin or the environmental factors could be established.

The generation times associated to growth recoveryconsecutively to the environmental variation was different fromthose calculated in constant condition. For example, in Table 4,the values of ratios of GT calculated in dynamic condition and GTin constant condition for strains Lm-F4 and Lm-C4 are indicated.In most cases, the growth recovery of the L. monocytogenes strainsexposed to environmental variations in exponential phase, i.e.dynamic condition, showed a generation time longer than inconstant condition, when the strains were in stationary phase.The conditions of culture in correlation with the presence of NaClinduced the higher ratios, i.e. the generation time observed indynamic condition was longer than in constant condition.

The better adaptation of clinical strains, as compared to foodstrains observed in constant condition was not established indynamic condition. The parameters values of the strains werecomparable. It must be noted that, contrary to results in constantconditions, all the strains were capable to grow in the condition


pH 5.0, 4% NaCl, 10 °C. At this temperature, no growths wereobserved at the most severe condition, pH 5.0, 8% NaCl, as shownin constant condition.

Among the clinical strains used in this work, two strains, Lm-C3 and Lm-C4, were implicated in the same outbreak, but isolatedfrom different patients and had different clonal types. Therefore,it was interesting to analyze their characteristics of growths. At 20°C, Lm-C3 and Lm-C4 had a similar behavior in front of thestudied factors and their variations. The growth parameter values,i.e. lag and generation times, did not show significative difference.As shown in Fig. 4 for 10 °C, few differences could be observedwhatever the studied condition. In constant condition, at pH 6.0–8% NaCl or pH 5.0–4% NaCl, Lm-C4 presented longer lag andgeneration times, e.g. at the last combination the lag time of Lm-C4 was elongated of 17 h compared to Lm-C3.

Classification of the Different Strains

The data base collected in this present work allowed theconstitution of a classification of the different strains, with theobjective to verify the global behavior of the clinical and foodstrains. The result, confirmed the previous conclusions. Threegroups were created. One of the groups was constituted by all theseafood L. monocytogenes strains (Lm-F1, Lm-F2, Lm-F3, Lm-F4), asecond one contained three of the clinical L. monocytogenes strains(Lm-C2, Lm-C3, Lm-C4). This means that the food strains were ingeneral “closer” to each other than to the clinical strains. Theparticular behavior of the clinical strain Lm-C1 underlined in ourresults was confirmed by this classification, since it was isolatedfrom the other ones. Indeed, as indicated by the experimentalresults, this clinical strain presented a better adaptation tounfavorable environment of temperature, acid pH and saltconcentration.

DISCUSSION

It is well known that L. monocytogenes is a microorganismwhich is able to survive, and frequently grow, under a wide rangeof adverse conditions such as low temperature, low pH and highosmolarity. In this work, significant differences in growth kineticswere found between clinical and seafood strains in acid and/or

osmotic environment, at different temperatures. Indeed, the clinicalstrains, in stationary phase, were revealed to be more resistant tothese environmental conditions, in comparison to environmentalstrains. These results are in accordance with the study of Dykesand Moorhead (2000) who examined the acid stress response ofclinical or meat L. monocytogenes strains and demonstrated theability of clinical strains to survive an acid shock. Likewise, Averyand Buncic (1997) focused on another environmental factor, thetemperature, and showed that 15 clinical strains of L. monocytogeneshad higher resistance to the effects of unfavorable storageconditions, compared with 15 meat strains.

Our results confirmed that more investigations were necessaryto allow the evaluation of the growth characteristics of clinicalstrains, compared to food strains, in particular in conditions foundin the food environment, with the objective to improve the riskassessment. Eight strains were studied in the present work. Thebehavior of a larger number of strains (implicated in the mainlisteriosis cases in the world or associated with different foods) infunction of different environmental factors should now be exploredto consolidate our conclusions. Indeed, when foodborne listeriosisis considered, it could be hypothesized that a high resistance ofsome L. monocytogenes strains to environmental factors found infoods or industrial environment may contribute to the particularcapability of certain strains to cause illness and, consequently, tobecome clinical strains. The influence of environmental factors(such as temperature, acid, etc.) on the expression of virulencemarkers was also highlighted.

The experiments of the present study, comparing the strainsbehavior in constant and dynamic condition, showed differenceson the calculated generation times. This observation is in agreementwith the previous studies of Cheroutre-Vialette et al. (1998) andCheroutre-Vialette and Lebert (2000a). They observed that cells ofL. monocytogenes isolated from meat products or food processingenvironment submitted to environmental variations of pH/aw inexponential phase were usually more sensitive than cells instationary phase. It was largely demonstrated that, phenotypically,bacteria in stationary-phase growth are more thermotolerant, acid-resistant and are better equipped to survive osmotic stress (Hilland Rees). Nevertheless, in this work, for a studied combination


(pH 5.0, 4% NaCl, 10 °C), the food strains were able to overcomethe stress in exponential phase but not in stationary phase (thisexperiment was repeated three times with different inoculum).

The presence of L. monocytogenes in seafoods has beenexamined by several authors. The prevalence of this pathogen canbe associated to the light preservation process (salting, etc.), theextended shelf life at refrigerated temperatures, capable ofsupporting the growth of L. monocytogenes, the consumptionwithout further cooking. Therefore, it is important to evaluate theadaptation of this microorganism in the food processingenvironment which could include variations (temperature, salting,acidification, etc.) during time. Mathematical modeling techniquesare gradually becoming recognized and accepted as powerfultools for predicting the effects of food-preservation systems onbacterial growth kinetics. The applicability of available predictivemodels to fish products, considering the products characteristics,was approached. Recently, several studies have demonstrated thesignificative interest to take into account the dynamic condition inthe field of predictive microbiology, and to collect experimentaldata in this way.

With the objective to introduce more advances into the field ofpredictive microbiology and increase the safety of products, studiesallowing correlations between clinical strains characteristics (suchas growth kinetics, cell physiology, e.g. stress condition, expressionof pathogenicity in function of particular environment found infood) prove to be necessary. Another prospect is to study thegenetical background of environmental strains in order to selectthe ones that may be most representative of the whole naturalpopulation. The knowledge of typing data (serotype, ribotype,pulse field gel electrophoresis) for such strains and theircomparison to the corresponding epidemiological data shouldprovide a useful tool for the Listeria risk assessment.

VARIABILITY OF THE RESPONSE OF 66 LISTERIAMONOCYTOGENES AND LISTERIA INNOCUA STRAINS TODIFFERENT GROWTH CONDITIONS

The growth of 58 strains of Listeria monocytogenes and eightstrains of Listeria innocua iso-lated from meat products (68%) and

industrial sites (23%), were compared in four con-ditions oftemperature, water activity (aw) and pH. Temperatures rangedfrom 10-37 °C, pH from 5·6-7·0 and aw from 0·96-1. Growths wereperformed in a meat broth with an auto-mated turbidimeter(Bioscreen C, Labsystem). Growth curves were fitted using theGom-pertz function, and growth parameters were calculated. Thedifferences between strains in lag phase duration were muchgreater than in growth rate. The greatest differences occurred at 10C, pH 7 and aw 0·96 : lag time values ranged from 4 h to 4 days.the Listeria population was separated into five groups, accordingto the lag time and maximal growth rate values using clusteringanalysis. The majority of the strains isolated from industrial siteswere grouped together and showed faster growth than the othersin the four con-ditions studied. The serotype or the nature of themeat from which the strains were isolated did not influence growth.The variability observed among strains raises questions about theconsequences in quantitative risk assessment and about theconstruction of models in pre-dictive modeling.

Listeria bacteria are widespread in the environment. Amongthe different species, Listeria monocytogenes and Listeria innocua arethe two most commonly isolated from food processing. L.monocytogenes can cause the death of both the very young andimmunocompromised individuals. Most cases are traced to thecontamination of raw or processed foods with L. monocytogenes.Listeriosis is therefore a major threat to human health.

Listeria monocytogenes can grow at low temperatures, low pHand low water activity (aw); they are therefore able to survive andmultiply in a wide range of food products.

Work on predictive microbiology has been carried out in anattempt to improve the shelf life and safety of food. Predic-tivemodels of microbial growths are set out with respect to the maincontrolling factors of the environment such as temperature, pHand water activity. Despite variations in growth among strainsnumerous models have been constructed using only one or asmall pool of strains. In this work, we compared the growth of 66strains of L. monocytogenes and L. innocua, isolated from meat, meatprod-ucts and industrials sites. Strains were clus-tered accordingto their growth in a broth medium in four conditions of temperature,


pH and aw. Consequences on the construction of models arediscussed.


Strains

Fifty-eight strains of L. monocytogenes and eight strains of L.innocua were studied: 45 strains were isolated from meat and meatproducts, 15 from meat and dairy plants (materials, floors, walls)four were involved in outbreaks. Five of 13 serotypes of L.monocytogenes were represented: 1/2a, 1/2b, 1/2c, 4b, 4d. Listeriacultures were stored on tryptic soy agar slopes at 4 C.

Media

Culture inocula were prepared in a meat medium (MM): meatpeptone (Merck) 10 g l 1, yeast extract (Difco) 5 g l 1, glucose(Merck) 5 g l 1. The pH was adjusted to 7·0 with NaOH (Prolabo)1 mol l 1. The medium was autoclaved at 120 C for 20 min.

Growth experiments were carried out in a tryptic meat broth(TMB) which was steril-ized by filtration derived from tryptonemeat agar (GTV5p, Fournaud et al. 1973): meat extract (Merck) 10g l 1, proteose peptone (Merck) 10 g l 1, tryptone (Difco) 5 g l 1,

glu-cose 5 g l 1. The medium was buffered with a K2HOP4-KH2PO4(Merck), 0·1 mol l 1 solution in proportion 1:1 (v/v) and adjustedto pH 7·0 with NaOH 1 mol l 1. aw was set by adding NaCl(Merck) in accordance with Chirife and Resnik (1984).

Tryptone sodium chloride medium (TSC) (tryptone 1 g l 1,

NaCl 8·5 g l 1, pH 7·0) was used for dilutions, and TSA and APTagar (Difco) were used for plate counts and agar slope cultures.

Growth

Growths were measured by optical density in an automatedturbidimeter, Bioscreen C. All strains were inoculated in MM andincubated for the same period of time (17 h) at 30 C. All cultureswere therefore in stationary phase, thus avoiding disparities inthe subsequent growth kinetics. Nine millilitres of TMB wereinoculated with the subculture. The inoculum size was confirmedby plate counts. The inocu-lated TMB was dispensed asepticallyin 300 µl volumes into honeycomb microplates (10 10) of the

Bioscreen C. For each strain, eight successive wells of the samecolumn were filled. The last two wells received the same volumeof non-inoculated medium in order to determine the growth mediumoptical density (OD) and controlled any possible contamination.Ws methodology is simi-lar to that used by Be got et al. (1996).

Experimental Design

A Plackett Burman design (Plackett and Bur-man 1943) wasused to test the effects of three factors: temperature, pH and wateractivity. Each factor was studied twice at two levels: high and low(Table 3). While a fac-torial design would have required eightexperiments, four were sufficient with a Plackett Burman design.

Curve Fitting

OD data were transferred from the Bioscreen C to Excelsoftware (Microsoft Windows) and transformed for eachmeasurement. Four quantities were calculated at time t: (ODi)t, theaverage of the OD of eight replicates; (ODni)t, the average of theOD of the non-inoculated medium

(DOD)t=(ODi)t (ODni)t

Log10[(DOD)t/DODmin]

where DODmin was the lowest DOD value above the detectionthreshold.

In the linear range of the Bioscreen C ( DOD<1’2) (Begot et al.1996), growth curves were fitted using the modified Gompertzequation:

Table : Serotypes, food sources and laboratory references oftested strains of Listeria

Species Serotype Food source

1 Listeria monocytogenes P1a 1/2c Pig carcass

2 L. innocua 2138a 6a Minced meat

3 L. monocytogenes 24631a 1/2b Rib of lamb

4 L. monocytogenes 3670a 1/2a Minced meat

5 L. monocytogenes 2141a 1/2c Minced meat

6 L. monocytogenes 27795a 1/2a Minced meat

7 L. monocytogenes 4133a 1/2c Minced meat


8 L. monocytogenes 28423a 1/2c Minced meat 9 L. monocytogenes 2143a 1/2a Minced meat 10 L. monocytogenes 5602a 4b Pig shoulder 11 L. monocytogenes 880030b 1/2a Meat products 12 L. monocytogenes 880390b 4b Meat products 13 L. monocytogenes 880398b 1/2c Meat products 14 L. monocytogenes 890307b 4b Meat products 15 L. monocytogenes 890313b 4d Meat products 16 L. monocytogenes 890467b 4b Meat products 17 L. monocytogenes CLIP 19908c 1/2c Meat products 18 L. monocytogenes CLIP 19910c 1/2a Meat products 19 L. monocytogenes CLIP 19532c 1/2c Meat products 20 L. monocytogenes CLIP 19534c 1/2c Meat products 21 L. monocytogenes CLIP 19536c 1/2c Meat products 22 L. monocytogenes CLIP 19712c 1/2a Meat products 23 L. monocytogenes CLIP 19734c 1/2a Meat products 24 L. monocytogenes CLIP 19802c 4b Meat products 25 L. monocytogenes CLIP 19804c 4b Meat products 26 L. monocytogenes CLIP 19884c 1/2c Meat products 27 L. monocytogenes CLIP 19887c 1/2b Meat products 28 L. monocytogenes 925331d 1/2a Chicken 29 L. monocytogenes 925318d 1/2b Guinea-fowl 30 L. monocytogenes 925321d 1/2c Sausages 31 L. monocytogenes 925222d 1/2a Chicken 32 L. monocytogenes 925228d 4b Chicken 33 L. monocytogenes 925267d 1/2a Sausages 34 L. monocytogenes 925257d 1/2c Minced meat 35 L. monocytogenes 925261d 1/2c Minced meat 36 L. monocytogenes 925253d 1/2a Sausages 37 L. monocytogenes 925201d 1/2a Sausages 38 L. monocytogenes ATCC 19111 1/2a Poultry 39 L. monocytogenes ATCC 19115 4b Human

cerebrospinal fluid40 L. monocytogenes 1e 4b Industrial sites 41 L. monocytogenes 4e 1/2b Industrial sites 42 L. monocytogenes 10e 1/2a Industrial sites 43 L. monocytogenes 13e 1/2b Industrial sites 44 L. monocytogenes 14e 4b Industrial sites 45 L. monocytogenes 16e 4b Industrial sites

46 L. monocytogenes 17e 1/2b Industrial sites 47 L. monocytogenes 18e 4b Industrial sites 48 L. monocytogenes 29e 4b Industrial sites 49 L. monocytogenes 39e 4b Industrial sites 50 L. monocytogenes 41e 4b Industrial sites 51 L. monocytogenes 42e 1/2c Industrial sites 52 L. monocytogenes 44e 4b Industrial sites 53 L. monocytogenes 70e 4b Industrial sites 54 L. monocytogenes 99e 4b Industrial sites 55 L. monocytogenes Lo 28f 1/2a Human isolate 56 L. monocytogenes CDC Atlanta 9g 4b Milk 57 L. monocytogenes CA Wisconsing 4b Cheese58 L. monocytogenes OH Wisconsing 4b Cheese59 L. monocytogenes Scott A Wisconsing 4b Human isolate60 L. innocua CLIP 20719g Meat61 L. innocua CLIP 20728g Poultry62 L. innocua CLIP 20741g Meat63 L. innocua CLIP 20595g Meat64 L. innocua CLIP 20600g Poultry65 L. innocua CLIP 12511g Poultry66 L. innocua CLIP 12512g PoultryStrains of Listeria were donated by: aDr Nicolas (Regional Laboratory of Haute-Vienne, France).bDr Marly (INRA of Nouzilly, France).cDr Rocourt (Pasteur Institute, France).dDr Courtieu (National Center of Listeria references, France).eDr Jolivet (SOREDAB, France).fDr Cossart (Pasteur Institute, France).gDr Richard (National Institute of Agronomic Research of Jouy-en-Josas,France).

( )t

ODNLog Log

N OD10 100 min

(1)⎛ ⎞∆⎛ ⎞

= ⎜ ⎟⎜ ⎟ ⎜ ⎟∆⎝ ⎠ ⎝ ⎠

µ eA L t

A

.. exp exp ( ) 1

⎛ ⎞⎛ ⎞= − − +⎜ ⎟⎜ ⎟⎝ ⎠⎝ ⎠

where A is the logarithmic increase of bac-terial population, L, thelag time, µ, the maxi-mal growth rate and t, the time.


Growth parameters were determined by non-linear regressionwith Statitcf software (Technical Institute of Cereals and Fodders,France). Generation time (T) was derived from the maximal growthrate (2):

T = [Log10(2)]/m

Table: Conditions of culture incubation and growthmeasurements of Listeria strains in anautomated turbidimeter

(Bioscreen C)

Temperature 37 C 10 C

Number of wells 100 200

Wavelength 600 nm

Measurement time 30 h 240 h

4t 5 min 60 min

Shaking frequency 30 s per 1 min

Shaking intensity medium


Two techniques were used to compare the growth of strains.Lag times (L) and maxi-mum growth rates ( ) were both considered.Average and standard deviation were calcu-lated for both variables.Each of the 66 values were then normalized: the average wassub-tracted and the result divided by the stan-dard deviation.

Each strain was depicted in a space of 2x4=8 dimensions.Principal component analy-sis (PCA) was used to reduce thedimension of the sample space. The method looks for the uniquesubspace of a given dimension which explains the major part ofthe variance in the original data. This analysis was performedusing Statitcf software. A cluster analysis was used and calculatedwith Statitcf software. Classification is achieved through successiveaggre-gations of individuals and then groups of individuals. Theprinciple is to regroup indi-viduals according to their distancefrom the centre of gravity for the set of experimental points. Twoindividuals or two classes were aggregated when their fusioninvolved a minimal increase of the intra-class dispersion. Theaverage value of each parameter was then calculated for eachgroup and an overall average was calculated for the 66 strains.

Table: Plackett Burman design to study the effect oftemperature, aw, pH on the growth of 66Listeria strains

Conditions Name aw Temperature( C) pH

1 010 0·96 37 5·6

2 100 1·00 10 5·6

3 001 0·96 10 7·0

4 111 1·00 37 7·0

Results

The logarithmic increase of popu-lation, A, was maximal incondition 111. Par-ameter A was higher in both conditions of pH7·0 than those of pH 5·6.

The mean, below and above which there is an equal numberof strains, was below the average value for the parametersconsidered in all conditions. A proportion of 50-60% of the strainswas under the average for each parameter.

Large variations in lag time and gener-ation time were notedamong the strains in all conditions. The largest differences in lagtime among strains were recorded in con-dition 001. In thiscondition, the values were multiplied by a factor 25 from theminimum to the maximum value, taking the value from 4 h to 4days. Larger variations among lag times were found successivelyin condition 001, 100, 010 and 111.

The more severe the growth conditions, the larger the numberof disparities found in lag time. Fewer differences were found ingeneration time. In condition 100, generation times tripled fromthe minimal to the maximal value. In the other conditions,generation times doubled when comparing the slowest andfastest strain for each condition. Slight variations were found inparameter A which corresponds to the logarithm increase of thepopulation density. We observed that strains exhibiting the longestgeneration time of the entire population did not show the maximallag time.

For example L. monocytogenes ATCC 19115 which has thelongest generation time in condition 001, showed a lag time of50·8 h whereas the longest lag time was 97·9 h. Similarly, the


strain exhibiting the shortest generation time did not show theshortest lag time. Results were identical in all conditions tested.

Table: Effect of temperature, aw and pH on the growth of 66Listeria strains

Parameter aw Temp. pH Min. Max. Average Mean

A001 1·2 1·9 1·5 1·5

T001 0·96 10 7·0 8·9 20·1 13·3 13·0

L001 3·9 97·9 31·8 28·2

A010 1·1 1·6 1·4 1·4

T010 0·96 37 5·6 1·3 2·9 1·7 1·7

L010 1·2 10·2 4·0 3·4

A100 0·8 1·8 1·3 1·3

T100 1·00 10 5·6 8·4 24·2 12·9 11·8

L100 2·4 40·2 17·4 16·6

A111 1·1 1·9 1·7 1·8

T111 1·00 37 7·0 0·5 0·9 0·7 0·7

L111 0·2 2·6 0·9 0·6

The variability of the growth response was expressed by thefollowing parameters: logarithm increase of population density(A), generation time (T) and lag time (L). T and L were expressedin hours, A was an increase of Log10.

Strains were separated according to their ability to grow ineach condition by using the variables L and µ. We chose threeaxes which explained 68% of the initial information. Axis 1separated strains according to the values of parameters L and µ.Lag time and maximal growth rate were negatively and highlycorre-lated in all conditions.

Strains exhibiting fast growth (low lag time and high maximalgrowth rate) were isolated from strains which grew slowly (highlag time and low maximal growth rate). The second axis scaledthe conditions from worst to best: 001-100-010-111. Strainsseparated again according to growth conditions. Axis 3 segregatedcondition 001 from the others. Strains that grew well in thiscondition, where temperature and water activity were low, wereclustered.

Cluster Analysis

The cluster analysis divided the population into five groups.

Group 1 (10 strains) was clearly separated from other groups.It was comprised of strains which grew faster than the overallaverage in all the conditions studied. The majority (66%) of thestrains belonging to this group originated from industrials sites.

Group 2 (15 strains) was characterized by strains which grewmore slowly than the overall average, which is in opposition togroup 1. L. innocua was not present.

Group 3 (28 strains) was intermediate. Strains exhibited lagtimes and maximal growth rates around the overall average. It islocated at the centre of the PCA figures. All origins were represented.

Group 4 (eight strains) clustered which grew more slowlythan the overall average except for condition 100 where they grewfaster. No epidemiological strains of L. monocytogenes were presentin this group.

Group 5 (five strains) was composed of strains which grewmore slowly than the average of the strains, except for condition001 where they grew faster. 80% of group 5 strains were L. innocua.L. monocytogenes Scott A, which has been widely used to set upmodels, belonged to this group.

The nature of the meat they were isolated from had no effecton the cluster analysis:

Table: Comparison of the generation times (h) and lag times (h)averages calculated on the 66 Listeria strains and on each

group defined by the cluster analysis

10 C aw 0·96 37 C aw 0·96 10 C aw 1·0037 C aw 1·00 pH 7·0 pH 5·6 pH 5·6 pH 7·0 T001 L001 T010 L010 T100 L100 T111 L111

Overall average 13·3 31·8 1·7 4·0 12·9 17·4 0·7 0·9

Group 1 10·7 16·2 1·5 2·3 9·7 9·2 0·6 0·5

Group 2 13·9 44·8 1·9 6·6 13·5 23·2 0·6 0·6

Group 3 13·8 36·0 1·6 3·2 12·1 18·7 0·7 0·7

Group 4 13·3 28·2 2·1 4·2 10·8 10·1 0·8 1·8

Group 5 10·2 6·5 1·8 4·0 19·1 20·1 0·7 1·3


Table: Distribution of the serotypes of L. monocytogenes strainsin the five groups defined by the cluster analysis

Group 1/2a 1/2b 1/2c 4b 4d L. innocua Non- Totaldetermined

1 2 3 1 3 0 1 0 10

2 4 1 5 3 0 1 1 15

3 8 2 5 11 1 1 0 28

4 1 0 3 3 0 1 0 8

5 0 0 0 1 0 4 0 5

Total 15 6 14 21 1 8 1 66

each group was composed of strains of diff-erent serotypes.

Discussion

This study highlighted differences in growth among L.monocytogenes strains. Larger variations between the 66 strainswere shown in lag times rather than in generation times.Particularly wide variations were observed in condition 001 (10 C,aw 0·96, pH 7·0) where lag times extended from 4 h to 4 days.These results tally with those of Barbosa et al. (1994) who reportedthat among 39 strains of L. monocytogenes, lag times ranged from1·5-2·9 days at 10 C, pH 7·2. Lag time is known to be the timerequired for the organ-isms to adapt their cellular components tothe new environment. As a consequence, it is affected by thedifference between subculture and culture condition, thephysiological state of the inoculum and by the strain. Indeed, inour study, as we were careful to homogenize the conditions of thesub-cultures so as to avoid disparities in the sub-sequent growthkinetics, the differences we obtained in the growth parameterswere attributed to the nature of the strain.

Such variability in lag times added to the variability ingeneration times characterizes strains with slow or fast growth.As the Inter-national Commission on Microbial Specifi-cation forFoods set the toler-ance limit in food at <100 L. monocytogenes pergram at the point of consumption, it was possible to calculate thetime needed to reach this limit in the four conditions studied forboth ‘slow’ and ‘fast’ strains as the inoculum varied. The maximal

difference between both strains is seven days and is reached incondition 001, when the inoculum is 1 cfu ml 1. Such results canhave great consequences in quantitative risk assessment. Listeriosiscauses illness in certain categories of the population(immuno-compromised, old or pregnant people) but the conditionsof foodborne outbreaks are still not well known. The infectiousdose is supposed to be high but the minimal dose causing illnessdepends on the vulnerability of the person who ingested the food,the type and quality of the food, the level of the pathogen in thefood and the virulence of the strain. As a consequence, suchunknown data linked to the variability of the growth responseshow the difficulties than can be faced in estimating the probabilityof occurrence of listeriosis.

In our study, the strains were clustered according to theirgrowth ability in the four conditions. They were not groupedround their serotype nor the nature of the meat from which theywere isolated. However it is interesting to note that group 1 wasin majority composed of strains isolated from industrial sites.These strains grew faster than the others in all the conditionstested. This suggests that they are accustomed to growing in awide range of harsh conditions (e.g. high osmolarities, lowtemperatures). Studies have recently shown the ability of Listeriato adapt their cellular physiology to harsh environments. Most ofthe L. innocua studied were found in group 5, and grew moreslowly than the average population, except in condition 001 (10C, aw 0·96, pH 7·0). These results concord with those of Barbosaet al. (1994) but not with studies where modified or selectivemedia were used. Listeria monocytogenes Scott A was also clusteredin group 5; in condition 001, it has the fastest lag time (3·9 h) andthe fastest generation time (9·4 h) of all the 66 strains. Theseresults do not tally with those of Bar-bosa et al. (1994) who studiedthis strain among 39 other strains and showed that it had thelongest lag time at 4 and 10 C. Other strains that were the causesof epidemic out-breaks were found in different groups: L.monocytogenes OH Wisconsin was clustered in group 1(characterized by fast growth), L. monocytogenes CDC Atlanta ingroup 2 (slow growth) and L. monocytogenes CA Wisconsin, ingroup 3 (average growth).


Table: Prediction of time (in days) necessary to reach a limitlevel of 102 cfu ml 1 as the inoculum varied from 1-10 cfu ml1

Conditions aw Temp. pH Strains 1 cfu ml 1 10 cfu ml 1

001 0·96 10 7·0 Fast 2,6 1,2

Slow 9,6 6,5

010 0·96 37 5·6 Fast 0,4 0,2

Slow 1,2 0,8

100 1·00 10 5·6 Fast 2,4 1,1

Slow 8,2 4,5

111 1·00 37 7·0 Fast 0,2 0,1

Slow 0,4 0,2

The time was calculated in the four conditions of growth for a ‘slow’‘ strain and a ‘fast’ ‘ strain.

Such results point out the difficulties in choosing a strain tobuild a predictive model. Indeed numerous studies have involvedthe testing of one or a cocktail of three or five strains. L.monocytogenes Scott A was fre-quently used. The results of Barbosaet al. (1994) along with those presented here raise some questionsas to predictive models: is it wise to construct models with onlyone strain? If so, which strain? That with an average, fast or slowgrowth? If not, how can we integrate the growth variability inmod-els? If the fastest strain were considered, the model wouldalways predict a shorter lag time and lower growth rate thanthose found for the majority of strains, and would thus alwaysprovide safer growth estimates. But in this case, the industrycould not support the costs involved in reducing the shelf life ofa product. Furthermore, there is nothing to guarantee that a fasterstrain will not be found in the future. For public health safety, theslowest strains should be avoided. In order to take the variabilityof the strains into account, two ways are proposed: the firstapproach is the production of two predictive models using twostrains that are charac-terized by a fast and a slow growth; thiswould result in giving a growth response interval. The secondapproach would be a model integrating first the study of astrainhaving an average behavior in a large range of experimentalconditions, and second the results of this study.

We have shown the importance of studying a large number ofstrains of various origins before constructing models to assessstrain variability. However, other factors should also be consideredto produce effective mod-els, for example inoculum level,interactions between bacteria, interactive effects between strainand product as shown by Rosenow and Marth (1987) andcomparison of growth in liquid media to growth on solid media.

Growth of Pseudomonas fluorescens and Pseudomonas fragi in ameat medium as affected by pH (5.8 – 7.0), water activity (0.97– 1.00) and temperature (7 – 25°C)

A total of 59 strains of Pseudomonas, isolated from meat products,were grown in micro-titer plates in a meat medium over a rangeof pH (5.8–7.0), aw (0.97–1.00) and temperature (7–25°C). Growthswere performed in a meat broth with an automated turbidimeter(Bioscreen C, Labsystem, France). The growth curves obtained inthis study did not have sigmoidal shapes making it impossible tocalculate growth parameters using the Gompertz equation. Themedium was weakly oxygenated in the micro-titer plates andreached 0%-dissolved oxygen at the beginning of the exponentialphase. Strains were separated into two groups: P. fragi and P.fluorescens. Strains of P. fragi had shorter lag times than those ofP. fluorescens. The impact of such results is interesting in that thesecould assist to explain the succession of flora that is observedduring the processing of meat: P. fluorescens is the dominant bacteriaamong Pseudomonas spp. at the beginning of a slaughter line andP. fragi becomes dominant during the chilling process.

Many flora of spoilage-bacteria have an effect on the shelf-lifeof refrigerated food products. The main flora responsible for spoilagein fresh meat and milk products during aerobic storage howeverare the Pseudomonas species. These are dominant in poultry meat,pork, and beef and lamb. In milk and cream the major spoilagebacteria have been identified as P. fluorescens and P. fragi. Widderset al. (1995) reported that the average contamination on wholecarcasses at 4°C peaked at 3.90 log10 cfu/cm2 and 4.54 log10 cfu/cm2 on meat surfaces at the same temperature. P. fragi and P.fluorescens cause deterioration in quality of meat and milk productsdue to the production of extracellular proteases and lipases at lowtemperatures. Off-odours occur in milk when the population of


Pseudomonas spp. reaches 2.2×106 to 3.6×107 cfu/ml or on meatwhen the population reaches 107 to 108 cfu/cm2. P. fragi strainsproduce fruity and putrid odours on beef and have a deleteriouseffect on the colour of meat stored at 1°C, resulting in a green andslimy appearance.

Although many studies have been carried out on the occurrenceand levels of contaminations in meat products, there has beenlittle work on a comparison of relative growth of Pseudomonasspecies and on the growth variations among these strains. Theaim of this study therefore was to compare the growth rates ofstrains of P. fragi and P. fluorescens in a selected media over arange of pH, water activity (aw) and temperature, and to attemptto describe the resulting growth curves.

Strains

A total of 59 strains of Pseudomonas were selected from bacterialcounts made from meats of different origin. All were isolated fromor grown on Pseudomonas Agar Base supplemented withPseudomonas CFC Supplement (Oxoïd). The strains were identifiedas P. fluorescens or P. fragi using standard tests. This involved:growing on Plate Count Agar at 4°C, and a range of tests includingGram stain, oxidase and catalase reaction, mobility, production ofacid in the presence of maltose and cellobiose, degradation ofgelatine and production of pigment on King B medium. The strainswere stored at 4°C on PCA slants and transferred monthly.

Media

Two media were used for successive subcultures — thesewere PCA slants and meat medium (MM) which contained meatpeptone (Merck) 10 g/l, yeast extract (Difco, OSI, Maurepas, France)5 g/l, and glucose (Prolabo, Fontenay sous Bois, France) 5 g/l. ThepH was adjusted to 7.0 by NaOH (40 g/l). The culture mediumwas a tryptic meat broth (TMB) (Fournaud et al., 1973) and wasbuffered with a K2HPO4-KH2PO4 (Merck), 0.1 mol/l solution toadjust the pH. Water activity was controlled by adding NaCl(Merck) according to Chirife and Resnik (1984). An aliquot of 0.5ml/l of anti-foam ANBIO 15 (1% v/v) was added for growth ina fermenter.

Materials

An automated turbidimeter was used to follow the growth ofPseudomonas spp. in the micro-titer plates. Optical density (OD) ofthe growth was measured on a spectrophotometer. OD was readat a wavelength of 600 nm. The threshold of detection for theBioscreen C, and for the spectrophotometer, were determined at6.0×106 cfu/ml and 1.0×107 cfu/ml, respectively. The range whereOD was proportional to the bacterial population extended from0.010 to 0.8 OD for both apparatus.

A 2-l fermenter was used with pH monitored by a heat-sterilisable electrode. Dissolved oxygen was measured directly inthe fermenter with an oxygen probe and the zero calibrationdetermined with an oxygen-free medium provided by Setric GénieIndustriel. A full-scale reading was obtained by saturation of themedium with air supplied continuously to the culture.

Inoculum

Each strain was incubated on PCA slants at 25°C for 8 h andthen transferred to MM that was shaken in a rotary shakerwaterbath for 17 h at 25°C, by which time growth of all the strainshad reached the stationary phase. A proportion of the culture wasinoculated in TMB to give a concentration of about 107 cfu/ml inorder to be above the detection threshold of the spectrophotometerand the Bioscreen C. The viable number of bacteria was determinedon PCA plates using standard microbiological practice immediatelyafter inoculation of the growth medium. Viable numbers for growthin Bioscreen C ranged from 6.9–7.1 log10 cfu/ml and from 7.1–7.3log10 cfu/ml for growth in fermenter.

Growth Experiments

Growth of the strains was obtained on micro-titer plates usingthe protocol developed by Bégot et al. (1996): for each strain, 300l of inoculated TMB was dispensed in eight successive cuvettes inthe micro-titer plates. Some cuvettes were filled with non-inoculatedTMB and were used as controls. OD was measured at 600 nm.

The working volume of the fermenter was 1 l. The stirringspeed was regulated to 70 rev./min. Prior to inoculation themedium was saturated with oxygen, so that growth began at a


level between 90 and 95% of dissolved oxygen. Air could besupplied at the following volumetric rates: 0, 0.24, 0.40 and 0.80l/l of medium/min. A fraction of the fermenter volume wasperiodically removed to measure OD, however, no more than 10%of the working volume was pipetted during growth.

Growth Conditions

A fractional factorial design for each of the environmentalvariables, aw, temperature, pH at three levels, gave nine conditionsfor growth for each of the 59 Pseudomonas strains. These growthconditions are given in Table 2. Growth in the fermenter wascarried out with P. fragi 103 in condition E111.

Growth Curve Treatments

Growth data obtained as a function of OD versus time weretransferred from Bioscreen C to Excel software (Microsoft Windows)and transformed for each measurement. Four quantities werecalculated at time t:

1. (ODi)t, the OD average for the eight replicates

2. (ODni)t, the OD average for the non-inoculated TMB

3. (DOD)t=(ODi)t”(ODni)t

4. f(t)=Log10[(DOD)t/DODmin] where DODmin was thelowest DOD value above the detection threshold.

The function f(t) was used to calculate two parameters for the59 strains in the nine conditions: the maximal growth rate, Vmax,and the lag time, L. The maximum growth rate of all the growthrates Vmax, was calculated for each time tn (Eq. 2). L correspondedto the intersection of the x-axis and the tangent in Vmax:

( ) ( )n

n nt

n n

f t f tV

t t1 1

1 1

+ −

+ −

−=

− (1)

Vmax = maximum (Vtn) (2)


Average, mean, minimum and maximum were calculated forthe two variables L and GT. Strains were classified according totheir ability to grow in the different conditions. A cluster analysiswas carried out with STAT-ITCF software.

RESULTS

Strains of both species grew at 4°C on PCA, were Gram-negative, motile, gave positive oxidase and catalase reactions. P.fluorescens strains produced fluorescent pigments and gelatinase,P. fragi strains did not. P. fragi strains produced acid in thepresence of maltose and cellobiose, P. fluorescens strains did not.

As shown in numerous studies, a decrease in temperature, pHor aw increased the duration of both lag and generation times. Fig.1 shows for most growth conditions, growth was characterised byan unusually shaped growth curve, i.e. the lag phase was clearlyvisible with the growth phase generally reaching its maximalgrowth rate at the beginning of this phase. After the maximalgrowth rate, there was a break in the curve leading to a decreasingphase which did not reach a steady stationary phase, even afterlong periods of incubation (up to 45 h in condition E111). Thesegrowth curves are thus characterised by an inflection point situatednot long after the end of the lag phase and by a non-symmetricalshape. The two parameters GT (generation time) and L (lag time).It is seen that the interval between minima and maxima increasedas the conditions of pH, aw and temperature moved away from themost favourable condition E111. It is to be noted that the gapbetween minima and maxima increased more for L than for GT.

P. fragi 103 rapidly metabolised dissolved oxygen when no airwas supplied, i.e. there is a break in growth as the dissolvedoxygen level reached 10% 1.5 h after inoculation. The time neededto reach 10% of the saturated value for dissolved oxygen was 1.5,1.5, 2.5 and 3.5 h when the rates of air were respectively 0, 0.24,0.4 and 0.8 l per l of medium per min. Zero percent dissolvedoxygen was reached 1 h following these periods. Growth curvesin these four conditions are shown in Fig. 3. For 2 h, growthcurves are similar, then they separated and rose faster as the airsupply increased. The growth curve obtained in Bioscreen C wasclose to that obtained in the fermenter at a flow rate of 0.24 l perl of medium per min. It assumed confidently that conditions ofaeration in the Bioscreen C cuvettes are similar to this type ofaeration.

The classification carried out on all the strains revealed twomain groups. The first was comprised of all the P. fragi and four


P. fluorescens namely 51, 11.6A, M2L3 and P×10. The secondcomprised the remaining P. fluorescens. Statistical values for allstrains. L, averages, means, minima and maxima for each group.It can be seen that shorter lag times were obtained for the firstgroup. Between each group differences were greater in conditionsE133 (aw 1.00, 7°C, pH 5.8), E232 (aw 0.985, 7°C, pH 6.4), E331 (aw0.97, 7°C, pH 7.0) related to the lowest temperature.

DISCUSSION

The growth curves obtained in this study did not havesigmoidal shapes making it impossible to calculate growthparameters with the Gompertz equation which is usually used byauthors. A steady stationary phase was rarely obtained and forlong experiments a biofilm could be seen on the top of the mediumin the cuvettes. In addition to possible insufficient shaking of themicro-titer plates in Bioscreen C, it was supposed that the aerationwas also too weak. Growth in the fermenter confirmed this andhighlighted that the aeration in the cuvettes was equivalent to 0.24l of air per l of medium per min. These conditions of aerationplayed too important a role for aerobic bacteria and became alimiting factor for the growth of the Pseudomonas spp. A possibleexplanation for the unexpected shape of the growth curves is thatthe combined effect of the depletion of oxygen and the change inthe metabolism: indeed, in some cases, Pseudomonas spp. can usenitrate as an alternate electron acceptor and can growanaerobically.

For many studies carried out in milk and milk products withPseudomonas spp. the generation time was the main parameterfor comparison. Cox and MacRae (1988) reported that P. fragigrew faster than P. fluorescens in both U.H.T. and raw milk at 4°C.In meat products, few studies have compared the behaviour of thespecies. This work carried out on 59 strains of Pseudomonas hasshown small differences between calculated generation times thatare confirmed by results of Pooni and Mead (1984) who workedon strains isolated from poultry products and grown in heartinfusion broth. These authors showed that pigmented and non-pigmented Pseudomonas made little difference in average generationtimes. The differences in lag time observed between both speciesare important. For the 59 strains separated into the two groups,

only four strains of P. fluorescens were in the P. fragi group. Thesewere characterised by shorter lag times in condition E111 (25°C,aw 1, pH 7). This explains their presence in the first group. The P.fragi group had on average shorter lag times than P. fluorescens.These differences have been observed on a great number of strainsand in a wide range of temperature, pH and aw; e.g. average L wastwo-fold shorter in the P. fragi group at 7°C.

These results could help to explain the changes in the ecologyof Pseudomonas flora shown by different authors who have studiedlevels of contamination and areas of isolation during meatprocessing and retailing. Lahellec and Colin (1981) observed thespoilage strains in poultry: pigmented strains represented themajor population isolated in processing plants, but during storagenon-pigmented strains outgrew the former. On beef, lamb andpork, studies have shown the predominance of P. fluorescens fromthe slaughter line to the chilling process. P. fluorescens is knownto be largely present in the environment (floor, water), on animals(hide, skin) or also in water and surfaces in meat factories. Oncutting lines and during storage and retailing, P. fragi was thenfound as the dominant flora on meat. Drosinos and Board (1995)studied the evolution of the bacterial flora in minced lamb storedaerobically at 4°C and showed that P. fragi outgrew P. fluorescens.They tried to explain this succession of flora by differences in themetabolism of both species. In our study shorter lag times observedfor P. fragi strains could be linked to their ability to adapt betterto new environmental conditions compared to P. fluorescens.

Our findings make it possible to choose a limited number ofrepresentative strains from the 59 Pseudomonas — so as to diminishthe number of experiments needed to study these few strains withmore precision in broth or on surface tissue of meat. A ‘low’,‘quick’ or ‘average’ strain characterised by a slow, rapid or averagegrowth in all or few conditions of the experimental design, couldthen be chosen in both species. Further studies should beundertaken: to substantiate the differences in lag times betweenthe ‘average’ strains of both species and compare their generationtimes in better conditions of aeration; to study a ‘low’ strain anda ‘quick’ strain and observe the variability in their growthresponses; or to compare the growth parameters obtained in brothand on the surface tissue of meat.


DIVERGICIN 750, A NOVEL BACTERIOCIN PRODUCED BYCARNOBACTERIUM DIVERGENS 750

Divergicin 750, a bacteriocin produced by Carnobacteriumdivergens 750, preferentially inhibited the growth of strains ofCarnobacterium and Enterococcus. Selected strains of Listeriamonocytogenes and Clostridium perfringens were also inhibited.The bacteriocin was purified to homogeneity by ammonium sulfateprecipitation and sequential S-Sepharose, hydrophobic interactionand reversed phase chromatography. The complete amino acidsequence was determined by Edman degradation. The peptideconsisted of 34 amino acid residues. The calculated M(r) from thepeptide sequence, 3447.7, agreed well with that obtained by massspectrometry. Divergicin 750 did not show any sequencesimilarities to other known bacteriocins. The plasmid-locatedstructural gene encoding divergicin 750 (dvn750) was cloned andsequenced. The gene encoded a primary translation product of 63amino acids with a deduced M(r) = 6789.4 which is cleavedbetween amino acid residues 29 and 30 to yield the maturebacteriocin. Lactic acid bacteria produce a variety of com- poundswith antimicrobial activity. Some of these are proteins or peptidesand are termed bacteriocins. Bacteriocins from lactic acid bacteriaare currently being divided into four classes. Class II bacteriocinsare small and have little post-translational modifications. Theyare usually heat stable, hydrophobic and often act on the cellmembrane of susceptible target cells. Bacteriocins are commonlysecreted by a dedicated transport and maturation system. Thebacteriocins usually inhibit the growth of closely related species.Some bacteriocins also inhibit the growth of pathogens and spoilageorgan-isms and may thus be of interest for enhancing food safetyand food hygiene.

Carnobacteria grow at low temperature, produce relativelylittle acid and volatile compounds and they generally also havelow proteolytic and lipolytic activity. Little is known aboutbacteriocins from carnobacteria. Some bacteriocin-producingcarnobacteria have been identified. We have previously purifiedand characterised piscicolin 61 from Carnobacterium piscicola LV61.C. piscicola LV 17 produced three bacteriocins, termedcarnobac-teriocin A, BM1 and B2 of which carnobacteriocin Awas identical to piscicolin 61. Carnobacteriocin BM 1 and B2

show significant sequence similar-ity and contain the -YGNGVXC-motif typical of bacteriocins active against Listeria. Recently,diver-gicin A, produced by C. dicergens LV 13 was char-acterised.It constitutes an interesting exception in being a small class II-likebacteriocin which uses the cells’ sec system for translocation. Alanthion-ine-containing bacteriocin, carnocin UI49 has beenpurified as well. Here we describe the purification, characterisationand cloning of divergicin 750, a novel bacteriocin from C. diuergens750.

Growth of Bacteria

Thirty-seven strains of carnobacteria were screened forproduction of bacteriocin. Carnobac-terium dicergens 750 from thelaboratory stock of Bundesanstalt fur Fleischforschung, Kulmbach.Ger-many, produced divergicin 750. For production of bacteriocin,cells were grown at 25°C overnight in cMRS medium containingPeptone proteose no. 3 (10 g l-1 ), yeast extract (5 g l-1), sucrose (20g l-1 ), K HPO4 (2 g l-1 ), (NH4)2 citrate (2 g l -1 ), MgSO4 (0.02 g l-

1), MnSO4 (0.02 g l-1 ). pH was adjusted to 8.5 with NaOH and themedium auto-claved for 20 min. In growth kinetic experiments C.diuergens 750 was grown in D-MRS broth adjusted to initial pH6.6 and pH 8.2. The broths were inoculated at a 0.1% level ofovernight cultures and incubated for 48 h at 25°C. At appropriatetime intervals the number of colony-forming units and bacteriocinactivity in the culture supernatant was determined. Growth of C.dicergens 750 in D-MRS broth of different pH values was monitoredusing an automated turbidometer, BIOSCREEN C. 10 µl of a ten-fold diluted 24-h culture were added to honeycomb wellscon-taining 190 µl D-MRS broth of pH 4.5, 5.0, 6.0. 7.0, 8.0, 9.0,and 10.0. Al! inoculations were done in triplicate and incubationwas for 48 h at 30°C. After 14 h and 38 h, samples were withdrawnfrom a second honeycomb plate and tested for bacteriocin activity.To test the bacteriocin for pH stability, fractions of culturesupernatants were adjusted to pH values between 2 and 11 with5 M NaOH or 2 N HCI. After 1 h of incubation at 20°C, residualactiv-ity was examined by use of the agar spot assay.

Bacteriocin Activity Assays

Bacteriocin activity was quantified by using serial dilutions ofthe bacteriocin in the agar spot test as described previously. One


arbitrary unit (AU) was defined as the reciprocal of the highestdilution yielding a definite zone of inhibition on the indicatorlawn. Alternatively, bacteriocin activity was quanti-fied in amicroliter plate assay system by measuring the growth of theindicator organism in two-fold dilutions of bacteriocin in mediumas described previously.

When necessary, bacteriocin fractions were adjusted to pH 6.5and sterilised by filtration through Millipore filters (0.22 µm) priorto activity measurements. One bacteriocin unit (BU) was definedas the amount of bacteriocin which inhibited growth of theindicator organism by 50% as compared to a control culturewithout bacteriocin. Unless otherwise stated C. dicergens L66 wasused as an indicator strain.

Purification and Analysis of Divergicin 750

Bacteriocin was purified essentially as described previously.In short, a 2-1 culture of C. diuer-gens 750 was grown to thestationary phase and cells were removed by centrifugation. Thebacteriocin pre-sent in the supernatant fraction was concentratedby ammonium sulfate precipitation and subjected to ion exchangechromatography (S-Sepharose), hydropho-bic interactionchromatography (Octyl-Sepharose) and reversed phase FPLC(Pharmacia). The purified bacteriocin was stable in 40% (v/v) 2-propanol containing 0.1% trifluoroacetic acid at - 20°C.

Amino acid sequencing and mass analysis of puri-fieddivergicin 750 were done as described previously.

Recombinant DNA Techniques

Basic cloning techniques were used. The probes were end-labelled with 32 P using a terminal transferase end-labelling kit.Plas-mid DNA from C. divergens 750 was digested with variousrestriction endonucleases and the resulting fragments wereseparated on agarose gels, blotted onto nylon filters and hybridisedwith a divergicin 750-specific degenerate oligonucleotide probe.

Two partly overlapping fragments, a 2-kb EcoRI frag-mentand a 0.2-kb Mbol fragment, hybridising to the oligonucleotideprobe, were purified by excising them from a low-melting-pointagarose gel and subse-quently employing the MagicTM clean-up

system. Cloning was done in Escherichia coli DH5 a using thecloning vector pGEM-7Zf( + ) (Promega). Positive clones wereidentified by colony hybridisation. The cloned DNA fragmentswere se-quenced completely on both strands using the Seque-naseversion 2.0 DNA sequencing kit (Amersham) and a primer walkingstrategy. Computer analyses were carried out on an IBM personalcomputer employing the DNASIS sequence analysis program andon a UNIX computer employing the GCG programme package.

RESULTS

Production and Inhibition Spectrum of Divergicin 750

Growth of C. divergens 750 at different initial pH values of thegrowth media was measured. The bacteria grew well at high pHwhile growth was significantly retarded at pH 5. Bacteriocin waspro-duced over a wide range of pH values and could be detectedeven down to pH 5. It appeared that diver-gicin 750 was producedin the late exponential phase of growth as the cells entered theearly stationary phase and remained fairly stable in the growthsuper-natant after production.

When C. divergens 750 was grown at different temperatures,maximum production of bacteriocin occurred at 25°C . Bacteriocinproduction was de-tected down to growth at 4°C. At 37°C noactivity was observed. Consequently, cells were grown at initialpH 8.5 and 25°C to enhance the yield of bacteriocin.

Table: Inhibition spectrum of purified divergicin 750

Indicator bacteria Number of sensitive/number tested

Carnobacterium dicergens 1 / 1

Carnobacterium piscicola I / 1

Carnobacteriumgallinarum 0/1

Enterococcus faecalis 2/2

Enterococcus faecium 0/

Lactobacillus sake 0/1

Bacillus cereus 0/1

Brochothrix thermosphacta 0/1

Clostridium butvricum 0/1


Clostridium perfringens 1 / 1

Clostridium sporogenes 0/ 1

Listeria innocua 0/2

Listeria icanocii 0/1

Listeria monocytogenes 1 /5

Listeria seeligeri 0/1

Listeria welshimeri 0/ 1

Propionibacterium acnes 0/ t

Staphylococcus aureus 0/4

Streptococcus mutans 0/ 1

Inhibition was tested by using the purified bacteriocin (2048 AU ml-1)in the agar spot test assay described in Materials and methods.

Some lactic acid bacteria are known to produce more than onebacteriocin. To rule out interference from other inhibitorysubstances, purified divergicin 750 was used in determination ofthe inhibition spectrum. Divergicin 750 preferably inhibited strainsof Carnobacterium and Ente-rococcus. Selected strains of Clostridiumperfringens and Listeria monocytogenes were also inhibited.

Purification of Dirergicin 750

Divergicin 750 was purified by ammonium sulfateprecipitation, sequential cation exchange, hydropho-bic interactionand reversed-phase chromatography. The purified bacteriocinappeared homogeneous when subjected to FPLC-reversed phasechromatography. Overall, a more that 600-fold increase in specificactivity was observed (BU mg-1 protein). The purified bacteriocinhad an activ-ity of approx. 10 000 BU µg -1 protein.

The complete amino acid sequence of the purified bacteriocinwas determined by Edman degradation. The sequence wasidentical to that deduced from the structural gene after cloning.

Divergicin 750 consisted of one polypeptide chain of 34 aminoacid residues. No unusual amino acid residues such as lanthioninswere found. The calcu-lated Mr of 3447.7 was almost identical tothe value of 3448.5 Da obtained by PDMS. The extra 1 Da in themass analysis probably originated from protonation of thepolypeptide during desorption.

Purification of Divergicin 750

Purification stage Fraction Total Specific Yieldvolume protein activity

(ml) (mg) (BU mg-I X 10-4) (%)

Culture supernatant 2000 120 1.6 100

(NH4)>SO4 conc. (Fr I) 200 90 4.2 190

S-Seph. chrom. (Fr))) 50 21 0.2 2.5

Octyl-Seph. chrom. (FrIIl) 9 1.4 100 70

Rev. phase FPLC (FrIV) 2 0.07 1000 35

Cloning and DNA Sequencing of the Divergicin 750 Structural Gene

Hybridisation of a degenerate oligonucleotide probe to plasmidDNA from C. dirergens 750 gave one strong hybridisation signalto a 2-kb EcoRI fragment and a partly overlapping 0.2-kb Mbolfrag-ment. The EcoRl and Mbol fragments were cloned andsequenced. The gene encoded a primary translation prod-uct of 63amino acids with a calculated Mr = 6789.4. Cleavage of this peptidebetween amino acid residue 29 and 30 would give the maturedivergicin 750. Uncertain amino acid residue determinations inposi-tions 1, 6, 11 and 26 of divergicin 750 were corrobo-rated bydeduction from the sequence of the cloned gene. Downstream ofthe structural gene, a region of dyad symmetry was found.

DISCUSSION

Divergicin 750 is a low molecular mass, hy-drophobic andbasic peptide and thus shares charac-teristics with otherbacteriocins of lactic acid bacte-ria. Divergicin appeared to beproduced only in the late exponential phase of growth . This is incontrast to many other bacteriocins, for example sakacin A fromL. sake Lb706 and sakacin P from L. sake Lb674, which areconstitutively synthesized. Nothing is known about the mechanismof regulation of divergicin 750. The bacteriocin was purified by aprocedure simi-lar to that used for other bacteriocins. Duringpurification a drop in activity was observed after elution from theS-Sepharose column, whilst the activity was regained after thenext purification step. The reason for this apparent drop isunknown. The molecular mass of divergicin 750 as calcu-latedfrom the deduced amino acid sequence was very close to that

Introductory Food Microbiology172 Epidemiological Typing of Bacillus spp Isolated from Food 173

determined by PDMS and is consistent with the proposed aminoacid sequence. Taken together with the information that nolanthion-ins were detected upon amino acid composition anal-ysis,divergicin 750 appeared not to be subjected to extensive post-translational covalent modifications and thus belonged to theclass II bacteriocin group. No sequence similarities of divergicin750 with other proteins encoded by sequences in the EMBL andGenBank sequence databases were found. Divergicin 750 thusdiffers from other bacteriocins of the lactic acid bacteria.

A general mechanism of action for the class II bacteriocins hasbeen suggested. In this model the bacteriocin molecules are thoughtto form am-phiphilic a-helices that combine to make a pore in thecell membrane of susceptible cells. Nothing is known about themechanism of action of divergicin 750. The region encompassingamino acid residues 7-28 may be able to form an amphiphilic a-helix and may indicate a similar mode of action for divergicin 750.

Divergicin 750 is processed from a longer pri-mary translationproduct. The N-terminal 29 amino acid residues are cleaved offadjacent to a glycine doublet to yield the mature product. Thisleader sequence, although unusually long, shows typicalsignatures of bacteriocin leader peptides of the dou-ble-glycinetype, for example a hydrophilic re-gion at positions - 8 to - 11flanked by hydropho-bic amino acid residues. Other bacteriocinsof lactic acid bacteria with this type of leader invariably have adedicated secretion system for their export. Indeed, a putativeopen reading frame starting ap-prox. 1 kb downstream of thestructural gene showed a high degree of similarity to bacteriocinABC ex-porter genes such as sapT. This indicates that dvn750 maybe part of a larger gene cluster involved in production and secretionof divergicin 750. Bacteriocins and bacteriocinogenic strains maybe used by the food industry as additives to prevent outgrowth ofpathogens and spoilage organisms, giv-ing an enhanced controlover production processes. The study of the structure of bacteriocinsis an important step in establishing knowledge about the parts ofthe molecule that are responsible for the inhibitory activity.Subsequently, it may be possible to increase the inhibitory spectrumand enhance the activity by changing specific amino acid residuesin the bacteriocin molecules.

7EPIDEMIOLOGICAL TYPING OFBACILLUS SPP ISOLATED FROM

FOOD

Biotypes, fatty acid profiles, and restriction fragment lengthpolymorphisms of a PCR product (PCR-RFLP of the cereolysin ABgene) were compared for 62 isolates of the Bacillus cereus group.Eleven isolates originated from various foods, and 51 isolateswere obtained from pasteurized milk which had been processedby two different dairies. The isolates were clustered into 6 biotypes,10 fatty acid groups, or 7 PCR-RFLP clusters. Isolates withmesophilic or psychrotrophic characteristics were preferentiallydistributed into specific fatty acid or PCR-RFLP groups (P = 0.004).Unique fatty acid clusters were predominantly found in milksamples of each dairy (P < 0.0001), suggesting that certain dairyplants may harbor plant-specific B. cereus which might constantlycontribute to postpasteurization contamination.

Improving microbial safety and extending the shelf life ofpasteurized milk and related products have always been animportant concern to the dairy industry. A major factor limitingrealization of these goals is microorganisms surviving thepasteurization process and/or contributing to postpasteurizationcontamination. Bacillus cereus, Bacillus thuringiensis, and Bacillusmycoides are frequently found in pasteurized milk, causingspoilage because of the production of lipases and proteases. Theycan also exhibit a health risk to the consumer since they produceenterotoxins. It has been proposed to merge B. cereus, B.thuringiensis, and B. mycoides into one single species. DNA-DNA

Introductory Food Microbiology Epidemiological Typing of Bacillus spp Isolated from Food174 175

hybridization experiments and pulsed-field gel electrophoresis ofchromosomal DNA demonstrated a very close genetic relationshipamong the three species. Furthermore, the 16s rRNA sequences ofthese species are more than 99% similar. In the present article, B.cereus, B. thuringiensis, and B. mycoides are, therefore, notdistinguished and are referred to as B. cereus.

Many dairy scientists believe that contamination of pasteurizedmilk with Bacillus spp. results from spores that were present inthe raw milk and survived the pasteurization process. However,other data in the literature suggest that postpasteurizationcontamination might significantly contribute to Bacillus counts inpasteurized milk. This controversy can partly be related to the lackof adequate typing methods for B. cereus. Ternstro¨m et al. describedan attempt to use numerical phenotypic analysis forcharacterization of dairy Bacillus spp. However, the authors didnot show any relationship between Bacillus isolates from raw andpasteurized milk. Va¨isa¨nen et al. used fatty acid analysis andphages to type dairy B. cereus isolates. Although they found bothtechniques useful for typing of B. cereus, the authors did notsystematically compare B. cereus isolated throughout theprocessing lines of single dairies.

Thus, their data do not allow conclusions about the origin ofB. cereus found in pasteurized milk. The objective of the presentstudy was to evaluate the feasibility of three typing techniques forB. cereus. The techniques chosen were biotyping, analysis of cellularfatty acids (CFA), and restriction fragment length polymorphismsof PCR products (PCR-RFLP). To allow direct comparison, thethree techniques were applied to a defined collection of B. cereusstrains. Because a large proportion of the Bacillus strains studiedwere isolated from pasteurized milk, additional epidemiologicaldata for dairy Bacillus spp. were also acquired.

Bacterial Strains

A total of 62 isolates of B. cereus were used. Eleven of theseisolates, originating from food-borne B. cereus outbreaks and fromrandomly collected milk samples, were obtained from our culturecollection. The remaining 51 isolates were cultured from 23 retailcartons (250 ml and 500 ml) of pasteurized milk that had been

processed by two different dairies. Isolation procedures followedthe methods described in the Association of Official AnalyticalChemists’ Bacteriological Analytical Manual, except that B. cereusselective agar supplemented with polymyxin B and egg yolkemulsion was used as selective plating medium.

Growth of B. cereus at 8°C

Growth curves for the 62 B. cereus isolates were determinedwith the multiwell photometric plate reader Bioscreen C. Bacteriawere grown for 18 h in brain heart infusion broth at 30°C andwere diluted in fresh brain heart infusion broth to an opticaldensity at 600 nm (OD600) of 0.005, and the suspensions wereused to inoculate the multiwell plate of the Bioscreen C. Theturbidity development (OD600) of the suspended bacteria at 8°Cwas recorded for 144 h. Growth curves (i.e., turbidity developmentcurves) were subsequently generated from the OD readings byusing the computer software BIOLINK.

Preliminary experiments had shown that growth velocity at8°C can be used to divide mesophilic and psychrotrophic B. cereusstrains by recording the detection time for each strain. The detectiontime is defined by the BIOLINK software as the time at which anOD of 0.05 is reached. The detection time of psychrotrophic B.cereus isolates is less than 100 h, whereas it is considerablyhigher than 100 h for mesophilic isolates.

Biochemical Properties

Twenty-one biochemical variables of the 62 B. cereus isolateswere determined by using the VITEK Jr. system according to themanufacturer’s instructions. From the biochemical patternsobtained, a binary profile was established for each isolate. Thedistances between these profiles were calculated by Jaccard’s index[d = 1 - c/(p + q + c)], where c is the number of variables presentin both strains, and p and q are the numbers of variables presentin each strain.

A dendrogram was constructed from the resulting distancematrix by using the unweighted pair-group method with arithmeticaverages (UGPMA). Calculations were made with the computerprograms MacDendro and MacMul.


PCR-RFLP

It has been demonstrated that the lecithinase andsphingomyelinase genes (cerAB) of B. cereus display geneticvariations among different isolates. This diversity and its possibleapplication for typing of B. cereus were further explored in thepresent study. A PCR product encompassing a 1.5-kb fragment ofthe cerAB gene was used to assess the genetic diversity of the 62B. cereus isolates at that locus. PCR amplifications with primersspecific for cerAB were performed as described previously. Ten-microliter aliquots of each amplicon were digested for 2 h with 10U of the following restriction endonucleases: MnlI, HgaI, Sau96I,PvuII, and BstXI. Reaction conditions were set as recommended bythe manufacturers of the enzymes. RFLP profiles were analyzedby electrophoresis on 1.5% agarose gels stained with ethidiumbromide. A binary matrix of the profiles was used to calculate thedistances between strains by Jaccard’s index. A dendrogram wasconstructed from the distance matrix by UGPMA as describedabove.

Analysis of Whole-cell Fatty Acid Profiles with the MIDI System

Fatty acid methyl esters of B. cereus were analyzed by usingthe Microbial Identification System. The analysis was conductedaccording to the instructions given by MIDI (10). Briefly, isolateswere grown on Trypticase soy agar (BBL) for 24 ±1 h at 28±1°C.At that time, 45 ±1 mg (wet weight) of cells was harvested fromsingle colonies. Total fatty acids of the bacteria were converted tofatty acid methyl esters which were subsequently separated andquantified with a Hewlett-Packard 5890 series II gas liquidchromatography column equipped with a flame ionization detector.A dendrogram based on the fatty acid profiles was established for137 B. cereus isolates (62 isolates described in this study plus 75B. cereus isolates isolated from various foods) by using the MISlibrary generation software. This program calculates a similaritycoefficient based on the Euclidean distances (ED) between pairs ofisolates.

Clustering of the isolates is determined by UGPMA. Thephenotypic expression of relative amounts of fatty acids withinbacterial cells depends on various factors, which include medium

composition, growth temperature, and growth rate. To determinean ED value that is not affected by such variations in CFA andthat could therefore be used to obtain a reproducible grouping ofthe B. cereus isolates within the dendrogram, the variability of thefatty acid profile analysis was assessed by repeated testing of B.cereus ATCC 14579. Repeated testing was conducted during a 3-week period in the following manner. Two separate batches ofmedium plates were prepared independently and stored at 4°C.During each week, the test strain was subcultured daily onto freshTrypticase soy agar plates.

The second, fourth, and sixth subcultures were analyzed. Eachof the subcultures tested was streaked onto duplicate plates ofboth batches of media. Each preparation of fatty acid methyl esterswas split into two subsamples, and the fatty acid profiles wereanalyzed separately with the MIS.

ED were calculated for the following datum pairs: (i) duplicatereadings of 33 identical sample preparations, (ii) samplepreparations from 33 identical subcultures streaked onto duplicateplates, (iii) sample preparations from 33 identical subculturesgrown on two different batches of media, (iv) 43 pairs of samplepreparations obtained from common subcultures grown duringeach week (e.g., second, fourth, or sixth subculture obtained in the1st and 2nd, 2nd and 3rd, or 1st and 3rd weeks), and (v) forty-four pairs of sample preparations obtained from serialsubcultures grown during each week (e.g., second and fourth,fourth and sixth, or second and sixth subcultures obtained in eachweek).

In this study, identical subcultures of the B. cereus isolateswere used to generate the fatty acid profiles. The critical valueobserved with two batches of media (ED, 6.7) could, therefore, beused to group the 137 B. cereus isolates, resulting in 11 fatty acidclusters. Although the MIS was originally designed for speciesidentification, it has been suggested that it may also be valuablefor epidemiological typing of bacteria. Consequently, the datumlibrary generated with 137 B. cereus isolates was used to evaluatethe relationships of the 62 B. cereus isolates. The fatty acid profilesof these isolates were attributed to 10 (Mi 2 to Mi 11) of the 11 B.cereus fatty acid groups.


Correlations

Comparison of growth characteristics and fatty acid profilesof the 62 isolates showed that psychrotrophic isolates werepreferentially (P , 0.0001) grouped into fatty acid clusters Mi 2 toMi 6, whereas mesophilic isolates belonged to fatty acid groupsMi 7 to Mi 11. A similar tendency (P 5 0.004) was perceived whengrowth temperature and PCR-RFLP groups were compared. Noassociation could be observed among biotypes and growthcharacteristics, PCR-RFLP groups, or fatty acid groups. Evaluationof the fatty acid groups for the 51 B. cereus isolates from pasteurizedmilk revealed interesting tendencies. A majority of the isolatesfound in samples processed by dairy A belonged to the subgroupMi 5, whereas isolates found in milk from dairy B belonged togroups Mi 6 and Mi 8 (P , 0.0001). Va¨isa¨nen et al. exploitedphage typing, minimum growth temperature, and analysis of CFAfor differentiation of dairy B. cereus isolates.

They found a correlation of fatty acid composition andminimum growth temperature; however, they did not examine thefatty acid data with regard to further typing of the isolates. Ourresults confirm the observations regarding composition of CFAand ability to grow at low temperatures. Additionally, we foundindications that psychrotrophic B. cereus strains with a specificCFA profile were predominantly found in pasteurized milkprocessed by one dairy, while they were less frequently found inmilk from the other processing plant. This plantspecific distributionof B. cereus in pasteurized milk suggests that the strains mayoriginate from common reservoirs within dairy plants or withincertain farms supplying raw milk. This research was funded bythe Ontario Food Quality and Safety Research Fund, by the DairyFarmers of Ontario, and by the Natural Sciences and EngineeringResearch Council of Canada. H.S. was supported by a fellowshipfrom the Commission for the Advancement of Young Scholars,Zu¨rich, Switzerland.

RAPID ASSESSEMENT OF THE QUALITY OF POULTRYCARCASSES

An assay has been developed whereby it is possible to assessmicrobial concentrations in poultry carcass rinses within 10

minutes. The test is a modification of one previously developed fordetermining raw milk quality. In the latter test, milk samples weretreated with a detergent to lyse non-microbial cells present in themilk. Microbial cells were removed by filtration and the ATPpresent in the cells extracted and assayed using the luciferase-luciferin reaction. A number of problems were encountered whenthis protocol was applied to chicken carcass rinses. The primaryproblem was a quenching effect on the light emission by theluciferase reaction caused by the presence of particulate and lipidmaterial in the rinse water. The quenching effect was substantiallyreduced by incorporating a pre-filtration step and an enzymetreatment prior to filtration through the bacteria-retaining filter.Using the modified assay, an excellent correlation (r=0.9) wasobtained when the ATP count obtained on 150 chicken rinsesamples was compared with plate counts carried out on the samerinse washes.

The test can be used to make a rapid assessment of poultrycarcass quality, based on selective cut-offs predetermined by theprocessor. If this level of count is exceeded, the carcass can besubjected to a heat treatment before sale. The accuracy of theprediction was high. Work was undertaken to compare methodsused to enumerate bacteria in chicken carcass rinses. The methodsstudied included plate count (by the Spiral plater), ATPbioluminescence, impediometry (using the Bactometer), HGMF,and turbidiometry. This enabled comparison of the repeatabilityand accuracy of the various methods. Only the ATPbioluminescence assay and HGMF correlated well with plate countsfor chicken carcass rinses and the repeatability of these test methodswas excellent.

The ATP method was used to monitor critical control pointsin the poultry processing plant. The CCP’s studied included thescald, the prechill and the chill tank waters. It was shown that theATP test could be used to give a reliable indication of the microbialcontent of process waters during production and results could beobtained within 10 minutes. The microbial load in the scald tankwater was shown to increase throughout the working day andthis was shown to be due to accumulation of microorganismswashed from the carcass rather than growth. Levels of microbial


contamination of the pre-chill and chill tanks were low in theplant studied on several sampling days. The microbial load in thetanks was dependent on flow rate of water through the tanks andit was concluded that, based on microbial load in the chill tank,water consumption at the plant could be reduced substantially.This will enable plant personnel to more efficiently respond tohigh microbial loads that may develop during chilling and resultin more economical management of the water supply.

ANTIMICROBIAL COMPOUNDS AND EXTRACELLULARPOLYSACCHARIDES PRODUCED BY LACTIC ACIDBACTERIA STRUCTURES AND PROPERTIES,DISSERTATION

In a variety of ecological niches, microorganisms competewith each other for survival and through evolution form uniqueflora. In some food ecosystems, lactic acid bacteria (LAB) constitutethe dominant microflora.

These organisms are able to produce antimicrobial compoundsagainst competing flora, including food-borne spoilage andpathogenic bacteria. Under unfavorable environmental conditionsmany species of LAB also produce exopolysaccharides (EPSs),which protect themselves against desiccation, bacteriophage andprotozoan attack.

Lactic acid bacteria provide the major preservative effects infood fermentation which mankind has practiced for thousands ofyears. The primary antimicrobial effect exerted by LAB is theproduction of lactic acid and reduction of pH. In addition, LABproduce various antimicrobial compounds, which can be classifiedas low-molecular-mass (LMM) compounds such as hydrogenperoxide (H2O2), carbon dioxide (CO2), diacetyl (2,3-butanedione),uncharacterized compounds, and high-molecular-mass (HMM)compounds like bacteriocins.

Among bacteriocins so far characterized, nisin is best defined,and the only purified bacteriocin produced by LAB that has beenapproved for use in food products. A LMM antimicrobialcompound, reuterin, has also been chemically identified, and thereuterin-producing Lactobacillus reuteri strain has been applied asa probiotic in dairy products.

The EPSs produced by LAB are either present as a capsuleattached to the cell surface, or secreted into the environment.Based on their sugar compositions, the EPSs can be divided intohomopolysaccharides, composed of a single type ofmonosaccharide, and heteropolysaccharides, containing severaltypes of monosaccharide. Dextran is the most importanthomopolysaccharide, which is a glucan produced, e.g. byLeuconostoc mesenteroides. The heteropolysaccharides produced byLAB are generally composed of repeating units of up to eightmonosaccharide residues; the chain length and degree of branchingvary with the producing strains. The rheological properties of thepolysaccharides depend on the monomeric composition, thenumber of side chains, the chain length and the charge (neutralor anionic) of the polysaccharides, as well as the anomericconfiguration of the monosaccharides and the sequence in whichthey are arranged. The EPSs produced by LAB may act asviscosifying agents to improve the texture and consistency offermented foods. Since LAB are food-grade microorganisms withthe GRAS status (Generally Recognized As Safe), the use of thesecreted EPSs as natural alternatives to produce all-natural foodproducts without additives has received increased attention. Ithas also been claimed that EPSs isolated from LAB cultures haveantitumor activity.

Lactic acid bacteria are able to produce a large variety ofcompounds which contribute to the flavor, color, texture andconsistency of fermented foods. However, the present study focuseson two potentially important group of compounds, antimicrobialcompounds and EPSs, which differ largely in their chemistry andfunctionalities. Attention has been paid to developing methodssuitable for separation and purification of LMM antimicrobialcompounds aiming at inhibition of food-borne spoilage bacteria.In order to understand the relation between the structures andrheological properties of the EPSs, a knowledge of their primarymolecular structures is required. In this study, the primarymolecular structures of the EPSs produced by Lb. helveticus strainshave been studied by NMR spectroscopy. The EPSs produced byseveral slime-forming Lactococcus lactis ssp. cremoris strains havebeen characterized to understand the roles of the EPSs in therheology of fermented foods.


Antimicrobial Compounds Produced by Lactic Acid Bacteria

Lactic acid bacteria are a physiologically diverse group oforganisms, which can be generally described as Gram-positive,nonsporing cocci or rods with lactic acid as the major product ofcarbohydrate fermentation. Traditionally, LAB comprise four generaLactobacillus, Leuconostoc, Pediococcus, and Streptococcus. However,several new genera have been suggested for inclusion in the groupof LAB due to a recent taxonomic revision. The genus Streptococcushas been reorganized into Enterococcus, Lactococcus, Streptococcusand Vagococcus. LAB are involved in the fermentation of a rangeof milk, meat, cereal and vegetable foods. The antimicrobialcompounds produced by LAB can inhibit the growth of pathogenicbacteria of possible contaminants in the fermented products. Inthe past two decades, there have been many reports on thebacteriocins produced by LAB. These bacteriocins are of aproteinaceous nature and they have been grouped into class I,lantibiotics which are small peptides (e.g. nisin), class II, smallheat-stable peptides, class III, large heat-labile proteins, and classIV, complex bacteriocins which are not well defined. In thefollowing text, the LMM antimicrobial compounds produced byLAB will be discussed.

Organic Acids

Fermentation by LAB is characterized by the accumulation oforganic acids and the accompanying reduction in pH. The levelsand types of organic acids produced during the fermentationprocess depend on the species of organisms, culture compositionand growth conditions. The antimicrobial effect of organic acidslies in the reduction of pH, as well as the undissociated form ofthe molecules. It has been proposed that the low external pHcauses acidification of the cell cytoplasm, while the undissociatedacid, being lipophilic, can diffuse passively across the membrane.The undissociated acid acts by collapsing the electrochemicalproton gradient, or by altering the cell membrane permeabilitywhich results in disruption of substrate transport systems.

Lactic acid is the major metabolite of LAB fermentation whereit is in equilibrium with its undissociated and dissociated forms,and the extent of the dissociation depends on pH. At low pH, a

large amount of lactic acid is in the undissociated form, and it istoxic to many bacteria, fungi and yeasts. However, differentmicroorganisms vary considerably in their sensitivity to lacticacid. At pH 5.0 lactic acid was inhibitory toward spore-formingbacteria but was ineffective against yeasts and moulds. It waspossible to grow Aspergillus parasiticus NRRL 2999 in a mediumcontaining 0.5 or 0.75% lactic acid at pH 3.5 or 4.5. Lindgren andDobrogosz (1990) showed that at different pH ranges the minimuminhibitory concentration (MIC) of the undissociated lactic acidwas different against Clostridium tyrobutyricum, Enterobacter sp.and Propionibacterium freudenreichii ssp. shermanii. In addition, thestereoisomers of lactic acid also differ in antimicrobial activity, L-lactic acid being more inhibitory than the D-isomer.

Acetic and propionic acids produced by LAB strains throughheterofermentative pathways, may interact with cell membranes,and cause intracellular acidification and protein denaturation.They are more antimicrobially effective than lactic acid due totheir higher pKa values (lactic acid 3.08, acetic acid 4.75, andpropionic acid 4.87), and higher percent of undissociated acidsthan lactic acid at a given pH. Acetic acid was more inhibitorythan lactic and citric acids toward Listeria monocytogenes. Aceticacid also acted synergistically with lactic acid; lactic acid decreasesthe pH of the medium, thereby increasing the toxicity of aceticacid.

Hydrogen Peroxide and Carbon Dioxide

Hydrogen peroxide is produced by LAB in the presence ofoxygen as a result of the action of flavoprotein oxidases ornicotinamide adenine hydroxy dinucleotide (NADH) peroxidase.The antimicrobial effect of H2O2 may result from the oxidation ofsulfhydryl groups causing denaturing of a number of enzymes,and from the peroxidation of membrane lipids thus the increasedmembrane permeability. H2O2 may also be as a precursor forthe production of bactericidal free radicals such as superoxide(O2 -) and hydroxyl (OH.) radicals which can damage DNA.

It has been reported that the production of H2O2 by Lactobacillusand Lactococcus strains inhibited Staphylococcus aureus, Pseudomonassp. and various psychotrophic microorganisms in foods. In raw


milk, H2O2 activates the lactoperoxidase system, producinghypothiocyanate (OSCN-), higher oxyacids (O2SCN- andO3SCN-) and intermediate oxidation products that are inhibitoryto a wide spectrum of Gram-positive and Gram-negative bacteria.

Carbon dioxide is mainly produced by heterofermentative LAB.The precise mechanism of its antimicrobial action is still unknown.However, CO2 may play a role in creating an anaerobicenvironment which inhibits enzymatic decarboxylations, and theaccumulation of CO2 in the membrane lipid bilayer may cause adysfunction in permeability. CO2 can effectively inhibit the growthof many food spoilage microorganisms, especially Gram-negativepsychrotrophic bacteria. The degree of inhibition by CO2 variesconsiderably between the organisms. CO2 at 10% could lower thetotal bacterial counts by 50%, and at 20-50% it had a strongantifungal activity.

Aroma Components

Diacetyl is produced by strains within all genera of LAB bycitrate fermentation. The antimicrobial effect of diacetyl has beenknown since the 1930s. It inhibits the growth of Gram-negativebacteria by reacting with the arginine-binding protein, thusaffecting the arginine utilization.

Jay (1982) showed that Gram-negative bacteria were moresensitive to diacetyl than Grampositive bacteria; the former wasinhibited by diacetyl at 200 ìg/mL and the latter at 300 ìg/mL.Diacetyl at 344 ìg/mL inhibited strains of Listeria, Salmonella,Yersinia, Escherichia coli, and Aeromonas. Since the production ofdiacetyl during lactic fermentation is low, e.g. 4 ìg/mL producedby Lc. lactis ssp. diacetylactis (Cogan 1980), and the acceptablesensory levels of diacetyl are at 2-7 ìg/mL, its practical use as afood preservative is limited. However, diacetyl may actsynergistically with other antimicrobial factors and contribute tocombined preservation systems in fermented foods.

Acetaldehyde is produced by Lb. delbrueckii ssp. bulgaricus bythe action of a threonine aldolase, which cleaves threonine intoacetaldehyde and glycine. Since Lb. delbrueckii ssp. bulgaricus andS. thermophilus in yoghurt cannot metabolize acetaldehyde, itaccumulates in the product at a concentration of about 25 ppm.

Acetaldehyde at 10-100 ppm inhibits the growth of Staphylococcusaureus, Salmonella typhimurium and E. coli in dairy products.

Fatty Acids

Under certain conditions, some lactobacilli and lactococcipossessing lipolytic activities may produce significant amounts offatty acids, e.g. in dry fermented sausage and fermented milk. Theantimicrobial activity of fatty acids has been recognized for manyyears. The unsaturated fatty acids are active against Gram-positivebacteria, and the antifungal activity of fatty acids is dependent onchain length, concentration, and pH of the medium. Theantimicrobial action of fatty acids has been thought to be due tothe undissociated molecule, not the anion, since pH had profoundeffects on their activity, with a more rapid killing effect at lowerpH.

Reuterin and Other Low-molecular-mass Compounds

Reuterin is produced by Lb. reuteri, a heterofermentative speciesinhabiting the gastrointestinal tract of humans and animals. It isformed during the anaerobic growth of Lb. reuteri by the action ofglycerol dehydratase which catalyzes the conversion of glycerolinto reuterin. Reuterin has been chemically identified to be 3-hydroxypropanal (â- hydroxypropionaldehyde), a highly solublepH-neutral compound which is in equilibrium with its hydratedmonomeric and cyclic dimeric forms. The biosynthesis pathwayfrom glycerol to the three forms of reuterin is shown below.

Methods for Evaluation of Antimicrobial Activity

Among many methods available for evaluation of antimicrobialactivity, the methods described below have been used fordetermining the antimicrobial activity of compounds produced byLAB.

The Agar Diffusion Method

The agar diffusion method has long been used for testingantimicrobial activity, and it was first used by Fleming in 1924.The method has been widely used for evaluation of antimicrobialactivity, specially for biologically derived compounds. It includesagar well diffusion assay and disc assay. In this test, an


antimicrobial compound is applied to an agar plate on a paperdisc or in a well. The compound diffuses into agar resulting in aconcentration gradient that is inversely proportional to the distancefrom the disc or well. The size of the inhibition zone around thedisc or well is a measure of the degree of inhibition. The resultsof the test are generally qualitative. The method requires that theindicator organisms must grow rapidly, uniformly, and aerobically.Since highly hydrophobic antimicrobial compounds cannot diffusein agar, they are not suitable for tests by this method used an agarwell diffusion assay for testing the antimicrobial activity of Lb.rhamnosus GG by addition of a 10-fold concentrate of the GG strainor MRS broth to wells (diameter 4 mm) in agar against variousanaerobic and facultative bacteria. The activity of an antimicrobialsubstance produced by Lb. delbrueckii ssp. bugaricus 7994 wastested quantitatively with a disc assay procedure, using paperassay discs 12.7 or 6.35 mm in diameter wetted with 30 or 10 ìlof sample against Pseudomonas fragi and Staphylococcus aureus. Theassay methods used for determination of the antimicrobial activityof different species of LAB were slightly different with respect tothe sizes of the wells, discs and samples, and the incubationconditions were dependent on the indicator organisms used.Several modified procedures based on the agar diffusion methodhave also been used for testing antimicrobial activity of LAB.These procedures include the agar spot test, deferred antagonismassay, and spot-onlawn assay.

The Agar and Broth Dilution Methods

Agar and broth dilution methods are used as quantitativemethods, suitable for microorganisms with variable growth rateand for anaerobic, microaerophilic microorganisms. The resultsare expressed as MIC, which is the lowest concentration of anantimicrobial that prevents growth of a microorganism after aspecific incubation period. In this test, an antimicrobial is seriallydiluted and a single concentration added to a culture tube or plateadded with nonselective broth or melted agar medium, which isthen inoculated with test organisms and incubated. The MIC isdefined as the lowest concentration at which no growth occurs(absence of turbidity) in a medium following incubation. The brothdilution assay has been used for the determination of the

antimicrobial activity of reuterin produced by Lb. reuteri, and theactivity of reuterin was expressed as MIC values or as the maximumdilutions of the reuterin fraction.

The Automated Turbidometric Assay

A turbidometric assay based on automated systems determinesthe effect of a compound on the growth or death kinetics of amicroorganism. It provides information concerning the effect of anantimicrobial that may cause a delayed lag phase or reducedgrowth rate at concentrations below the MIC. Since the bacterialgrowth is monitored by measuring the turbidity of the brothmedium, the method demands that the instrument be highlysensitive. Growth at levels below log 5.0 CFU/ml may not bedetectable. Skyttä and Mattila-Sandholm (1991) described aquantitative method based on automated turbidometry for assayingantimicrobial activity, which was expressed as area reductionpercentage values measured under the growth curve. The methodhas been used to test the antimicrobial activity of antimicrobialcompounds produced by P. damnosus and P. pentosaceus and Lb.plantarum.

Exopolysaccharides Produced by LAB

Lactic acid bacteria produce polysaccharides as cell wallcomponents and storage polymers, and also in many species, asa capsule or slime. In the dairy industry, the slime-forming LABstrains have traditionally been used in the production of fermentedmilk products, e.g. yogurts, Finnish ‘viili’ and Scandinavian‘långfil’. It has been generally acknowledged that the secretedEPSs by LAB play an important role in the rheological behaviorand texture of the products.

Homopolysaccharides

Homopolysaccharides are a group of polysaccharidescomposed of one monosaccharide type. Several species of LAB areable to utilize sucrose as a specific substrate to produce dextrans,mutans, and levans. Dextrans are a large class of extracellularlyformed glucans produced by the genus Lactobacillus, Leuconostoc,and Streptococcus, of which Leuc. mesenteroides and Leuc. dextranicumare the well-known dextran producers. Although each bacterial


strain produces a unique glucan, a common structural feature ofall dextrans is a high percentage (up to 95%) of á- 1,6 linkageswith a smaller proportion of á-1,2, á-1,3, or á-1,4 linkages resultingin a highly branched molecule (Franz 1986). Dextrans aresynthesized outside the cell by dextransucrase, which catalyzessucrose to produce D-fructose and D-glucose, and transfers thelatter to an acceptor to form dextran. The reaction is as follows:sucrose + glucan acceptor dextran or mutans + D-fructose Mutansare synthesized in a similar way by S. mutans and S. sobrinus.However, mutans differ from dextrans in containing a highpercentage of á-1,3 linkages, which are attributed to the insolublenature of this type of polymers. Some S. salivarius strains are ableto produce fructans of the levan type with 2,6-linked â-fructofuranoside residues. An extracellular enzyme levansucraseis involved in hydrolyzing sucrose and transferring D-fructose togrowing fructan chains to form levans: sucrose + fructan acceptorlevan + D-glucose

Another type of homopolysaccharide is the galactan producedby Lc. lactis ssp. cremoris H414, which is composed of a branchedpentasaccharide repeating unit as shown below.

A Pediococcus strain produced a â-D-glucan with a trisacchariderepeating unit. Lactobacillus spp. G-77 has been shown to producea 2-substituted- (1’!3)-â-D-glucan, identical to the EPS producedby P. damnosus 2.6. Lactobacillus spp. G-77 also produced a á-D-glucan composed of a trisaccharide repeating unit. Recently, vanGeel-Schutten et al. (1998) reported for the first time the productionof a fructan by Lb. reuteri strain LB121 with raffinose as a sugarsubstrate; this strain also produced both a glucan and a fructanon sucrose.

Heteropolysaccharides

A wide range of LAB strains can produceheteropolysaccharides, which are composed of repeating units.The monosaccharide compositions of these EPSs are mostlygalactose and glucose, and also small amounts of rhamnose,fructose, mannose, and galactosamine. In comparison with thehomopolysaccharides, the production of heteropolysaccharidesby LAB is much lower (60 to 400 mg L-1). Generally, the

heteropolysaccharides are synthesized intracellularly at thecytoplasmic membrane utilizing sugar nucleotides as precursorsfor the assembly of polysaccharide chains.

Lactobacillus

The ability of lactobacilli to produce EPSs has been recognizedfor many years. In 1968, Kooiman first reported the structure of aheteropolysaccharide produced by a Lb. brevis strain isolated fromkefir grains. This polysaccharide consists of a hexasacchariderepeating unit with D-galactose and D-glucose in the molar ratio1:1. In the last decade, a number of heteropolysaccharides producedby the Lactobacillus species have been investigated.

Lb. helveticus strains produce several EPSs with varyingrepeating units, though all containing galactose and glucose. TheEPS produced by Lb. helveticus 776 has hexasaccharide repeatingunits containing D-galactose and D-glucose. The EPS produced byLb. helveticus TY1-2 consists of heptasaccharide repeating unitswith Dgalactopyranosyl and D-glucopyranosyl, and 2-acetamido-2-deoxy-D-glucopyranosyl residues. Recently, Stingele et al. (1997)showed that the EPS produced by Lb. helveticus Lh59 had anidentical primary molecular structure as the one produced by Lb.helveticus TN-4, a presumed spontaneous mutant of the strainTY1-2. This polymer is composed of a tetrasaccharide backbonewith a lactosyl side-chain, and the molar ratio of D-galactose andD-glucose is 1:1.

Robijn et al. (1995b) reported the primary molecular structureof a viscous EPS produced by Lb. sake 0-1 which was isolated fromfermented meat products. The EPS consists of a pentasacchariderepeating unit of glucose, rhamnose, and glycerol phosphate. Thethreedimensional structure of this polymer has been further studiedby molecular mechanics calculations. The helics generated by apolysaccharide builder program are highly extended, with either2-fold or 3- or 4-fold right-handed chiralities. Grobben et al. (1997)showed that Lb. delbrueckii ssp. bulgaricus NCFB 2772 produced anEPS made up of galactose, and small quantities of glucose andrhamnose, and another EPS that, according to Sikkema and Oba(1998), was similar to the structure of the EPS produced by Lb.delbrueckii ssp. bulgaricus rr. The enzymes involved in the production


of the sugar nucleotides of strain NCFB 2772 have been analyzed,and based on this analysis a biosynthetic pathway for the EPS hasbeen proposed. Growth of the strain in a fructose-based mediumled to the absence of the enzyme activities for the synthesis of therhamnose nucleotide, and accordingly no rhamnose was presentin the polysaccharide produced. The EPSs produced by Lb.acidophilus LMG9433, Lb. kefiranofaciens K1, and Lb. paracasei 34-1have also been structurally evaluated, the repeating units beingpentamers, hexamers and tetramers, respectively. Recently, Lb.rhamnosus strain C83 has been shown to produce an EPS composedof a pentasaccharide repeating unit with a linear structure. Thisstrain, as well as Lb. casei CG11 and Lb. sake 0-1, produced moreEPSs at lower temperatures, whereas several other Lactobacillusstrains produced more EPSs at higher temperatures (comparedwith the optimum temperatures of growth).

Lactococcus

Among lactococci, only the slime-forming Lc. lactis ssp. cremorisstrains have been investigated. These strains producing EPSs playa role in the proper consistency of the fermented milk. The sugarcomponents of the EPSs are most frequently galactose, glucose,and very often rhamnose. Nakajima et al. (1992a) reported aphosphate-containing heteropolysaccharide, named ‘viilian’,produced by Lc. lactis ssp. cremoris SBT 0495 which was isolatedfrom a Finnish ‘viili’ starter culture. The EPS consists of thefollowing repeating unit:

Lc. lactis ssp. cremoris B40 has also been found to produce aphosphopolysaccharide with an identical repeating unit as shownabove. Lc. lactis ssp. cremoris strain LC330 appeared to produceconcurrently two EPSs; an anionic EPS composed of galactose,glucose, rhamnose, glucosamine and phosphate, and a neutralEPS containing galactose, glucose and glucosamine with branchedterminal galactose moieties. The mechanism for the biosynthesisof the EPS by Lc. lactis ssp. cremoris has not been investigated indetail. Recently, Oba et al. (1996) proposed a biosynthetic pathwayfor the production of ‘viilian’ by strain SBT 0495 with the followingsteps: preparation of the membraneembedded lipid carrier;incorporation of the first monosaccharide with the phosphate onC-1; assembly of the intact repeating unit.

The biosynthesis of the EPSs produced by Lc. lactis strains isgenerally associated with a plasmid. Transferring mucoidityplasmids from Lc. lactis ssp. cremoris ARH87 and MS to nonmucoidLc. lactis strains proved the latter to be mucoid. van Kranenburget al. (1997) described a novel 12 kb EPS gene cluster located ona 40 kb plasmid, which was essential for the EPS synthesis of Lc.lactis ssp. cremoris NIZO B40. Introduction of the EPS gene clusterfrom S. thermophilus Sfi6 to a non-EPS-producing Lc. lactis MG1363produced an EPS with a different structure from the EPS of thenative host (Stingele et al. 1999). The absence of the GalNAc residuein the EPS of Lc. lactis MG1363 was probably caused by the lackof a UDP-N-acetylglucosamine C4- epimerase activity.

Streptococcus

S. thermophilus strains are used in combination with Lb.delbrueckii ssp. bulgaricus strains in yoghurt starters. The EPSsproduced by several S. thermophilus strains have been found tohave similar or identical primary molecular structures. Doco et al.(1990) first reported the structure of an EPS produced by ropy S.thermophilus strains CNCMI 733, CNCMI 734 and CNCMI 735,which consisted of a tetrasaccharide repeating unit of D-galactose,D-glucose, and N-acetyl-D-galactosamine in a molar ratio 2:1:1.An EPS with an identical repeating unit structure has been reportedto be produced by S. thermophilus Sfi6. Lemoine et al. (1997) showedthat the EPSs produced by S. thermophilus Sfi12 and Sfi39 hadmolecular masses greater than 2 x 106, and both yielded a slimytexture rather than a thickened one in yoghurt. However, they haddifferent sugar compositions and structures; the former consistingof a hexasaccharide repeating unit of galactose, glucose andrhamnose, and the latter a tetrasaccharide repeating unit ofgalactose and glucose.

Recently, Faber et al. (1998) showed that S. thermophilus Rsand Sts produced EPSs of identical repeating units, but they haddifferent molecular masses, resulting in a difference in viscosity intheir milk cultures. Bubb et al. (1997) also showed that the EPSproduced by S. thermophilus OR 901 had a similar repeating unitto the one of the strains Rs and Sts; all being branchedheptasaccharide repeating units of D-galactose and L-rhamnosein the same molar ratio: 5:2.


Isolation and Structural Elucidation of EPSs

Isolation of EPSs

EPSs produced by LAB can be isolated by precipitation withtrichloroacetic acid (TCA) to remove proteins from the culturemedia, and subsequently by ethanol and/or acetone to precipitatepolysaccharides. Gel filtration or ion exchange chromatogrphy isoften used for further purification of EPSs. Robijn et al. (1995a,1995b, 1996a, 1996b) purified EPSs produced by severalLactobacillus strains by gel filtration with different columns (Sephacel500, Sephacryl S-500, and Superrose-6).

Gel filtration techniques were also used for purification of theEPSs produced by Lc. lactis ssp. cremoris H414, S. thermophilusSfi16, and S. thermophilus CNCMI 733. Since many EPSs arenegatively charged, they can be bound to an anion exchanger.This technique has been used for purification of the anionic EPSsproduced by Lc. lactis ssp. cremoris B40, and Lb. helveticus TY1-2and TN-4. Marshall et al. (1995) showed that Lc. lactis ssp. cremorisstrain LC330 produced at the same time both neutral and anionicEPSs in the medium; these two EPSs were effectively separatedand purified by anion exchange chromatography.

Sugar and Methylation Analyses

In the analysis of monosaccharide compositions of EPSsproduced by LAB, the EPSs are hydrolyzed with trifluoroaceticacid (TFA), reduced and acetylated, and the acetate derivatives areanalyzed with GC. For the methylation analysis, monosaccharidesare usually derivatized into partially methylated alditol acetates,which are introduced into the EI source of MS from a GC, or GLCinterface.

The substitution pattern of the monosaccharides can bedetermined by comparing their fragmentation pattern withreference EI-MS spectra. Yamamoto et al. (1994, 1995) studied thesubstitution pattern of the monosaccharides in the EPSs producedLb. helveticus TY1-2 and TN-4 by GLC-MS of the methylated andacetylated sugar residues. The EPS produced by Lb. helveticus 766was analyzed by GLC-MS on DB-1 of the partially methylatedalditol acetates.

Nuclear Magnetic Resonance Spectroscopy

NMR spectroscopy relies on the interaction of radio-frequencyelectromagnetic radiation with magnetically active nuclei in astrong magnetic field. The radio frequencies used range from 200to 800 MHz, corresponding to magnetic fields from 4.7 to 18.8Tesla. 1H and 13C are the spin-active nuclei most frequentlyencountered in carbohydrates. 1H and 13C NMR spectroscopy, including one- (1D) and two-dimensional (2D), is a powerful toolfor structural studies of carbohydrates, which also includepolysaccharides produced by LAB. 1D 1H NMR spectroscopy canbe used for rapid identification or to check the purity of apolysaccharide sample. Signals in the anomeric region (about 4.3-5.5 ppm) of the spectrum and the coupling of the anomeric protons(JH1,H2) may provide useful information about the number ofresidues in a repeating unit, and the anomeric configuration,respectively.

Yamamoto et al. (1995) recorded the 500 MHz 1H NMRspectrum of the EPS produced by Lb. helveticus TN-4 in D2O at 70oC, and found six signals in the anomeric region with nearlyequal integrated intensities, suggesting there was a hexasacchariderepeating unit for this EPS. A study on the EPS produced by Lb.helveticus Lh59, by 400 MHz 1H NMR spectroscopy in Me2SO-d6at 80 oC produced four anomeric protons signals in a molar ratio1:2:2:1, which also indicated a hexasaccharide repeating unit. The1H NMR spectrum of the EPS produced by Lb. paracasei had fourdoublets in the anomeric region, and the coupling constants(JH1,H2) of these signals (8.3 Hz, 7.7 Hz, 7.3 Hz, 7.7 Hz) were ofthe pyranoid ring form with all the residues in the â configuration.

Since the natural abundance of 13C is very low (1.1% relativeto 12C), the peak intensity of 13C has to be enhanced in 1D 13CNMR spectroscopy by using a large number of pulses, by takingadvantage of the nuclear Overhauser effect (NOE), or by usingdistortionless enhancement by polarization transfer (DEPT)experiments. The values of 13C chemical shift and 13C-1Hcoupling (1JC,H) provide structural information of thepolysaccharides. Robijn et al. (1995a) found six signals in theanomeric region of the 13C NMR spectrum of the EPS producedby Lb. helveticus 766, confirming the suggested hexasaccharide


repeating unit by 1H NMR spectroscopy. Based on the C-1 chemicalshifts in the 13C NMR spectrum of the EPS produced by Lb.rhamnosus strain C83, Vanhaverbeke et al. (1998) assigned twodownfield signals (ä 110.12, 107.67) to two residues having âconfiguration, and three signals at ä 99.73, 99.99 and 100.73 tothree residues having á configuration.

The 1D NMR techniques are often used for the assignment ofsignals in the anomeric region. For detailed assignment for thespin system of sugar residues, 2D techniques are needed. Thesetechniques include 1H,1H-correlated spectroscopy (COSY), totalcorrelation spectroscopy (TOCSY), homonuclear Hartmann-Hahnspectroscopy (HOHAHA), gradient selected heteronuclear singlequantum coherence (gHSQC), gradient selected heteronuclearmultiple-bond correlation (gHMBC) experiment, and nuclearOverhauser effect spectroscopy (NOESY). Bubb et al. (1997) useda TOCSY experiment to further assign two signals of anomericprotons in the 1H NMR spectrum of the EPS produced by S.thermophilus OR 901. By means of 2D COSY, HOHAHA, NOESY,and 13C-1H HMQC (heteronuclear multiple quantum coherence),Robijn et al. (1996a) assigned almost all 1H and 13C resonancesin 1D 1H and 13C NMR spectra of the EPS produced by Lb.paracasei 34-1. The sequence of the monosaccharide residues in arepeating unit can be established by 2D NOESY and HMBCexperiments. The former experiment gives information about theinterresidue linkage from observation of the NOE between anomericprotons and the protons at the substituted positions ofneighbouring sugar residues. The latter experiment gives rise tocrosspeak between proton and carbon atoms that are long-rangescalar coupled. Faber et al. (1998) used 2D NOESY together withHSQC-NOE experiments to determine the sequence of the sugarresidues in the EPS produced by S. thermophilus Rs and Sts. Themonosaccharide sequence in the EPS produced by Lb. paracasei 34-1 was established by 2D NOESY experiments.

Rheological Characterization of EPSs

Solution Viscosity

Viscosity ç is defined as the ratio between shear stress andshear rate. The intrinsic viscosity [ç], which measures the

hydrodynamic volume of a molecule, is obtained by extrapolatingthe Huggins equation to zero concentration: çsp /c = [ç] + k’[ç]2c,where çsp is specific viscosity, c is polymer concentration, and k’is a constant for a series of polymers of different molecular massin a given solvent. çsp /c is also defined as the reduced viscosityçred. For ionic polysaccharides in aqueous solutions, the value ofçred increases with decreasing concentration, showing apolyelectrolyte effect. The behavior of the polyelectrolytes isinfluenced by intrachain Coulombic interaction, ionic strength,pH and specific counterions. Oba et al. (1999) suggested that in astrain sweep test at very high dilution of the EPS produced by Lc.lactic ssp. cremoris SBT 0495, the higher cross-over frequency of theEPS in 0.1 M NaCl compared to that in pure water was due to thepolyelectrolyte effect of this EPS. van den Berg et al. (1995) showedthat over a wide range of shear rates, the viscosity of a 1% solutionof the EPS produced by Lb. sake 0-1 decreased with increasingshear rates, indicating a shear-thinning behavior, and the viscositywas comparable to that of xanthan gum.

Dynamic Viscoelasticity

In response to an applied stress, polysacharides may show aviscoelastic behavior, i.e. a combination of truely viscous flow andperfectly elastic response. In a dynamic test, the polysaccharidesample is subject to sinusoidal shear oscillation with a wide rangeof frequencies (0.01-300 Hz). The relative magnitudes of G’ (storagemodulus) and G” (loss modulus) vary with the state of thepolysaccharide. For entangled solutions, where there is a greatercontribution from the viscous element, G’ is low. When frequencydecreases, there is a crossover in G’ and G”, and they flow as highviscosity liquids at very low frequencies. For gel systems, G’ andG” are parallel, with G’ > G” and largely frequency independent.Oba et al. (1999) showed that in a dynamic and steady shearmeasurement the aqueous solutions of the EPS produced by Lc.lactis ssp. cremoris SBT 0495 behaved as an entangled solution butnot as a weak gel. Nishinari (1997) reported the frequency (0.01- 10 ù/radÅ”s-1) dependence of G’ and G” for 1-3% solutions ofgellan gum, an EPS produced by Pseudomonas elodea.

The 1% (0-30 °C) and 2% (30 °C) solutions had a typical dilutesolution behavior with G” > G’. The 2% (15 °C, 25 °C) and 3% (30


°C) solutions, however, had a concentrated solution behavior witha crossover of G’ and G” and G’ > G” at higher frequencies.Gelation occurs at 3% at 0-25 °C with G’ and G” being slightlyfrequency dependent.

Gelling Properties

Gels are defined as loose three-dimensional networks withstructures ranging from homogeneous solutions (enthalphy-drivenaggregations) to heterogeneous rigid porous systems. Clark andFarrer (1995) described the mechanisms of gel formation in threemain classes, firstly by point crosslinking with covalent bonds,secondly by chain association driven by changes in temperature,pH and ionic strength, and the presence of small molecules andspecific counterions, and thirdly by particle aggregation. Manypolysaccharide gels are formed by thermoreversable physicalassociations, involving Coulombic, dipole-dipole, van der Waals,charge transfer, and hydrophobic and hydrogen bondinginteractions, as well as double-helix formation and aggregation.Gels can be characterized to be strong or weak based on theirresponse to deformation. At large deformations, strong gels willrupture and fail, while weak gels flow without fracture, and showrecovery of solid character. Xanthan gum, the EPS produced byXanthomonas campestris, forms a weak gel, with large deformationfluid properties, but it also forms strong gel under extremeconditions.

Rheological behaviors of EPSs in Fermented Milk

The use of EPS-producing LAB strains may improve therheological properties of fermented milk. The gel structure andviscosity of the products are affected by the gel formationconditions, as well as the amount and the type of the EPSsproduced. Hammelehle et al. (1998) showed that fast warmingrates (20-50 °C) during acidification increased the gel firmnessand storage modulus, and decreased the syneresis of a milk gel.Skim milk fermented by ropy EPS-producing strains exhibitedsimilar rheological properties, and had greater viscosity than skimmilk fermented by non-ropy strains. Ropy EPS-producing strainsalso increased the viscosity of yoghurt when compared to yoghurtmade with non-ropy cultures, and improved the texture of quarg.

As described ealier, S. thermophilus Rs and Sts produced EPSs ofthe same structure, but had different viscosities in the milk culturesdue to their different molecular masses.

The rheological behavior of the polysaccharides is also relatedto their three-dimensional structure. In addition to the viscosifyingeffect of the polysaccharides, the interactions between the EPSsand the milk proteins, e.g. caseins, also play a role. Studies of ayogurt gel with a scanning electron microscopy showed that thecells were attached to the protein coagulates by a network structureconsisting of polysaccharide filaments. The microorganisms and/or the EPSs that they produce may affect the protein aggregation,thereby affecting the physical properties of the milk gel. A recentstudy showed that the rheological properties of stirred yoghurtwere affected by the type of EPSproducing strains used, suggestingan effect due to the interaction between the polymer and milkproteins. Hess et al. (1997) proposed a model for shear-induceddegradation of the microstructure of EPS-producing yogurt. Sincethe associations of EPS with bacterial cells or casein micelles arestronger than the associations between the casein micelles, anincrease in shear will first disrupt the casein micelle network thatis not associated with EPS, subsequentely the associations betweenthe cells and EPS, and then the portion of the casein network thatis associated with EPS.

Applications in Foods and Health Aspects

Antimicrobial Compounds as Natural Food Preservatives

The quality of most foods deteriorates during storage. Inaddition to physical, chemical and enzymatic factors which mayalter the sensory characteristics, the microbiological changes infoods may bring about a wide range of spoilage reactions, includingfood poisoning. Therefore, it is of significance to inhibit the growthof spoilage microorganisms in foods. Due to a strong demand fornatural and minimally processed foods, there has been a growinginterest in the use of antimicrobial compounds produced by LABas a safe and natural way of food preservation.

In addition to nisin which has been widely used in foods,another antimicrobial compound that has been proposed for usein food preservation is reuterin produced by Lb. reuteri. Addition


of reuterin to ground beef was found to inhibit the growth of E.coli. Surface treatment of herring with a mixture of Lb. reuteri andglycerol significantly improved the shelf-life of the product. Lb.reuteri has been commercially used in combination withBifidobacterium infantis and Lb. acidophilus in sweet and fermentedmilk under the trade name BRA-mjölk.

Antimicrobial compounds can be applied to foods either aspurified chemical agents, or as viable cultures in the case offermented products. Novel purified antimicrobial compoundsrequire data to substantiate their lack of toxicity in order to obtainapproval for their use in foods. Traditional fermented productsthat naturally contain antimicrobial compounds have beenconsumed for centuries, and starter cultures with selectedantimicrobial properties may be used to replace those used intraditional fermented foods. However, problems may arise withrespect to retaining the flavor and texture of the products.

EPSs in Food Applications

Polysaccharides may function in foods as viscosifying agents,stabilizers, emulsifiers, gelling agents, or water-binding agents.The majority of the polysaccharides used in foods are of plantorigin. Most of them are chemically or enzymatically modified inorder to improve their rheological properties, e.g. cellulose, starch,pectin, alginate and carrageenan. Therefore, their use is stronglyrestricted. EPSs of microbial origin have unique rheologicalproperties because of their capability of forming very viscoussolutions at low concentrations and their pseudoplastic nature.The EPSs produced by foodgrade LAB have been considered as anew generation of food thickeners to improve the rheologicalproperties of foods. Dextran is the first industrial polysaccharideproduced by LAB. It was discovered in 1880 in sugar cane or beetsyrups where dextran was found to be responsible for thethickening and gelation of the syrups. Due to their structuraldifferences, some dextrans are water soluble and others areinsoluble. Dextran can be used in confectionary to improve moistureretention, viscosity and inhibit sugar crystallization. In gum andjelly candies it acts as a gelling agents. In ice cream it acts as acrystallization inhibitor, and in pudding mixes it provides thedesirable body and mouth feel. In addition, dextran has also been

used as blood plasma extenders and as the basic component ofmany chromatographic stationary phases.

Xanthan gum is the second microbial EPS which was approvedfor use in foods in 1969. Although it is produced by the plant-pathogen Xanthomonas campetris, Sutherland (1998) describedxanthan as the “benchmark” product with respect to its importancein both food and nonfood applications, which include dairyproducts, drinks, confectionary, dressing, bakery products, syrupsand pet foods, as well as the oil, pharmaceutical, cosmetic, paper,paint and textile industries. The production of xanthan is relativelyinexpensive because of the high conversion of substrate (glucose)to polymer (60-70%). According to Becker et al. (1998), xanthan insolutions exhibits a high viscosity at low concentrations and strongpseudoplasticity, and it is stable over a wide range of pHs,temperatures and ionic strengths.

Another probiotic Lb. reuteri also produced an antimicrobialcompound with a wide spectrum of activities. Studies on bacterialadhesion showed that capsular polysaccharide might promote theadherence of bacteria to biological surfaces, thereby facilitating thecolonization of various ecological niches. The EPSs were found tobe present in adherent biofilms; the EPSs might function as initialadhesion, and permanent adhesion compounds. As well as livebacteria (probiotics) which can improve intestinal balance topromote health, dietary carbohydrates may function as prebiotics,beneficially affecting the colonic microflora. These dietarycarbohydrates include polysaccharides of plant origin (resistantstarch, â-glucan, cellulose, inulin), oligosaccharides (fructo-, gluco-, malto-, xylo- and soybean oligosaccharides), and lactosederivatives. There have been no reports of the use of EPSs producedby LAB as prebiotics. Although milk fermented with an EPS-producing strain Lc. lactis ssp. cremoris SBT0495 had cholesterollowering activity, the mechanism is unknown.

Oda et al. (1983) reported an antitumor EPS produced by Lb.helveticus ssp. jugurti. The antitumor activity of the EPS was testedagainst ascites Sarcoma-180 by injecting the EPS preparationintraperitoneally. Mice given a 20 mg kg-1 dose for nine succesivedays had an increased life span value of 144%, and a value ofgreater than 233% corresponding to a 40 or 80 mg kg-1 dose. The


authors concluded that the antitumor activity of the EPS might bebased on its host-mediated actions. In order to understand theantitumor activity, the effect of the EPSs or the EPS-producing cellson the immune system has been investigated. Forsén et al. (1987)showed that cell surface materials, possibly lipoteichoic acids, ofLc. lactis ssp. cremoris T5 produced Tcell mitogenic activity inhuman lymphocytes. The slime produced by Bifidobacteriumadolescentis had immunomodifying effects on mouse splennocytes.Kitazawa et al. (1992) showed that the slime-forming Lc. lactis ssp.cremoris KVS20 had antitumor activity, and the slime containedstrong B-cell dependent mitogenic substances.

AIMS OF THE STUDY

One of the aims was to study the antimicrobial compoundsproduced by dairy lactic acid bacteria, particularly the low-molecular-mass compound inhibitory toward various spoilage andpathogenic bacteria in foods. Another aim was to study theextracellular polysaccharides produced by dairy lactic acid bacteriain view of the role of the exopolysaccharides in the improvementof the texture and consistency of fermented foods. The specificaims of the study were:

1. To separate, purify and identify low-molecular-massantimicrobial compounds produced by the lactic acidbacterial strains, and to study the antimicrobial propertiesof these compounds.

2. To isolate exopolysaccharides produced by the lactic acidbacterial strains, to evaluate the primary molecularstructures of the exopolysaccharides, and to study therheological properties of the viscous exopolysaccharides.


Antimicrobial Compounds Produced by LAB (I, II)

Bacterial Strains and Growth Conditions

All bacterial strains used in the study of antimicrobialcompounds were obtained from Valio Ltd, Research andDevelopment Service, Helsinki, Finland. The bacterial cultureswere maintained at -80 °C in glass beads and they were subcultured

twice before use. The LAB strains examined for producingantimicrobial compounds were grown in MRS, KCA, and wheypermeate or whey media, and incubated at 30 °C or 37 °C. Foodspoilage bacteria and also LAB strains were used as indicatororganisms for antimicrobial tests.

Separation and Purification of Antimicrobial Compounds

After the growth of the LAB strains under proper conditions,cells in the culture broth were filtered, and the cell-free broth wasconcentrated 10-fold by lyophilization. The concentrate was thenprecipitated stepwise by ethanol from 30 to 97.5% with intermediatecentrifugation (30 min, 22 000g, 4 °C). The precipitates obtainedfrom each addition of ethanol and/or the final supernatantsshowing antimicrobial activity were further purified bychromatographic methods.

Gel filtration was performed using a Bio-Rad Econo System.The sample (100 mg) was loaded onto a column (75 x 1.5 cm) onBio-Gel P-2 polyacrylamide gel (M=100-1800, -400 mesh, Bio-Rad)eluted with 0.05 M ammonium acetate (NH4OAc) at a flow rate of10 ml h-1, and the eluant was monitored at 280 nm. The activefractions were collected, lyophilized, and subjected to anionexchange chromatography using a Bio-Rad Econo system with acolumn (25 x 1.5 cm) on weakly basic Fractogel TSK DEAE-650(S)gel. Elution was carried out at a flow rate of 1.0 ml min-1 usinga stepwise elution program: fractions 1-30 with water; fractions31-65 with 0.04 M NH4OAc adjusted to pH 5.5 with acetic acid(AcOH); fractions 66-90 with 0.5 M NH4OAc adjusted to pH 5.5with AcOH.

A fraction was collected every four minutes with monitoringat 254 nm. The active fractions (except fractions containing lacticacid) from anion exchange chromatography were further purifiedby RP-HPLC using a model 600 E multisolvent delivery systemequipped with a Baseline 810 software. The mobile phase, 0.02 MNH4OAc containing 1% AcOH (pH 3.80), was used after filtrationthrough a membrane filter (pore size, 0.2 ìm). Elution was performedisocratically from a Spherisorb S5 C8 column (250 x 4.6 mm, PhaseSeparations Ltd, Chester, England), fitted with a C8 precolumn(Millipore) at a flow rate of 0.75 ml min-1, and at 40 °C for 30 min.


A fraction was collected each minute. The absorbance wasmonitored at a range of wavelengths from 190 to 300 nm at aninterval of 5 or 10 nm.

Identification of Antimicrobial Compounds

1H and 13C NMR measurements were carried out on a BrukerAM 400 WB spectrometer (Karlsruhe, Germany), operating at 400.1MHz for 1H. Spectra were recorded with sample solutions inH2O/D2O (90/10) at ambient temperature and referenced tosodium 3-trimethylsilyl- [2,2,3,3-2H4]propanoate.

The electron impact (EI) and fast atom bombardment (FAB)mass spectra were recorded on a Jeol SX-102 double-focusingspectrometer.

EI: The sample was injected into the direct probe and thesolvent (water) evaporated. The probe was inserted into the ionsource (250 °C). The filament was heated at a rate of 16 °C/minup to 300 °C/min, the ionization current being 300 mA. Theionization energy was 70 eV and the accelerating voltage 10 kV.The spectra were recorded over the range 10-500 m/z. Calibrationwas based on PFK (perfluorokerosin, positive ion mode).

FAB: The sample was introduced on the target plate directlyinto the ion source (40 °C) in a glycerol matrix. The target wasbombarded with xenon atoms having a maximum of 6 kV energy.The acceleration voltage of generated ions was 10 kV. The spectrawere recorded at a scan range of 0-800 m/z. Calibration was basedon solid CsI (cesium iodide, positive ion mode).

Antimicrobial assay

The agar diffusion method was performed using a disc testand a spot test. The disc test was performed according to aprocedure developed by Pulusani et al. (1979) with somemodifications: 10 ml of the melted agar medium was seeded with100 ìl of an 18 ± 2 h old broth culture of the test organism in asterile petri dish. When the soft agar had hardened, an antibiotictest disc (diameter 6 mm, Schleicher & Schuell) was placed on theagar surface, and 22 ìl of the sample was spotted onto the disc.After incubation for 20 ± 2 h at the appropriate temperature foreach organism tested, the diameter of the inhibition zone around

the disc was measured. The spot test was done by spotting theliquid sample (3 ìl) directly onto the surface of the solidified,seeded agar medium, and the diameter of the inhibition zone wasmeasured after incubation. Turbidometric assays were performedusing a Bioscreen C automated turbidometer equipped with aBiolink software. The growth of indicator organisms in broth(300 ìl) containing antimicrobial compounds was studied in plates(100 wells). Each well was inoculated with 100 ìl broth culture(grown overnight) of the test organism diluted to 106 to 107 CFUml-1. The optical density was measured automatically at 30 min-interval, using a wideband filter (405-600 nm), and the plates wereshaken at 3 min-interval for 20 s. The growth curves weredetermined from the turbidity data.

EPSs Produced by LAB (III, IV, V)

Bacterial Strains and Growth Conditions

The LAB strains examined for producing EPSs and theirgrowth conditions are shown in Table 4. The source and methodsof maintainance of these strains were the same as described abovefor the LAB strains examined for producing antimicrobialcompounds in this study.

Isolation of EPSs

For the isolation of the EPS produced by Lb. helveticus Äki4grown in MRS broth, bacterial cells were filtered from the medium,and the cell-free supernatant was concentrated 10- fold bylyophilization. The concentrate was fractionally precipitated withethanol from 40 to 95% with intermediate centrifugation. Thepolysaccharide precipitated at 40% ethanol was washed, anddissolved in water. After filtration through a syringe filter (0.8 ì/0.2 ìl), it was freeze-dried. The crude polysaccharide (20 mg) waspurified by anion-exchange chromatography with a column (25 x1.5 cm) on Fractogel TSK DEAE-650(S) gel using a Bio-Rad Econosystem. The column was eluted at about 60 ml h-1 first with waterfor 80 min, and subsequently with 0.06 M NH4OAc adjusted topH 5.5 with AcOH for 120 min. A fraction was collected everyeight minutes with monitoring at 254 nm, and the presence ofsugar was tested with a Molish reagent.


For the isolation of the EPSs produced by other LAB strainsgrown in milk or whey medium, proteins and cells were initiallyprecipitated by addition of 4% (w/v) TCA (Merck) to the culture,and the mixture was stirred for 2 h. After centrifugation (35 min,22 000 g, 4 °C), the supernatant was collected and filtered. Coldethanol was then gradually added to the cell-free supernatantfrom one to two, and three volumes of the supernatant withintermediate centrifugation. The EPS precipitated was washedand dissolved in water. The aqueous solutions of EPS were filtered,and then extensively dialyzed against water overnight at 4 °Cwith two changes of water, and finally lyophilized.

The purity of the EPS material was examined by gel filtrationusing a column (75 x 1.5 cm) of Bio-Gel P-30 polyacrylamide gel(exclusion limit 40 000 daltons, 100-200 mesh). The sample (1 mg)was loaded onto the column and eluted with 0.05 M NH4OAcwith UV monitoring at 280 nm. To check the ionic nature of theEPSs, anionexchange chromatography of the EPS solutions(~1 mg mL-1) was performed using a column (25 x 1.5 cm) ofweakly basic Fractogel TSK DEAE-650(S) gel. Elution was carriedout at 1.1 mL min-1; first with water for 2 h, and subsequentlywith NH4OAc from 0.1 to 0.5 M using a linear increasing gradient.

Structural Elucidation of EPSs

GC analysis of alditol acetates was performed on a HP-5fused silica column (0.20 mm x 25 m) using a temperature programof 180 °C for 1 min followed by 3 °C min-1 to 250 °C. Hydrogenwas used as the carrier gas. The column was fitted to a Hewlett-Packard model 5890 series II gas chromatograph equipped witha flame ionization detector. GLC-MS analysis was performed ona Hewlett-Packard model 5970 mass spectrometer equipped withan HP-5MS fused silica column (0.2 mm x 25 m). A temperatureprogram of 170 °C for 3 min followed by 3 °C min-1 to 250 °C wasused with helium as the carrier gas.

In sugar analysis, the EPS samples were hydrolyzed with 2 MTFA at 120 °C for 2 h. After reduction with sodium borohydride(NaBH4) and acetylation, the samples were analyzed by GC. Theabsolute configuration of the sugars present in the EPSs wasdetermined essentially as devised by Leontein et al. (1978) but

with (+)-2-butanol. Methylation analysis was performed accordingto Hakomori (1964) using sodium methylsulfinylmethanide andiodomethane in dimethyl sulfoxide. The methylated compoundswere recovered by use of Sep-Pak C18 cartridges using the methodof Waeghe et al. (1983). The purified methylated sample was thenhydrolyzed (2 M TFA, 120 °C, 2 h), reduced, and acetylated.

The partially methylated alditol acetates were analyzed byGLC-MS. NMR spectra of solutions in D2O were recorded at 65 °Cand pD 5.5, using a Jeol GSX- 270, Jeol Alpha-400, or Varian Inova600 or 800 MHz instrument. Chemical shifts are reported in ppmrelative to sodium 3-trimethyl-(2,2,3,3-2H4)propanoate (äH 0.00)or acetone (äc 31.00) as internal references, or dioxan as an externalreference. Data processing was performed using standard Jeolsoftware, VNMR software, or Felix 2.3 (Biosym/MSI, San Diego,CA, USA). 1H,1H-COSY, relayed COSY, double-relayed COSY,TOCSY, 13C,1H-COSY, gHSQC (Wilker et al. 1993) and HMBC(Bax and Summers 1986) experiments were used to assign signalsand performed according to standard pulse sequences. For inter-residue correlations, 2D NOESY experiments with mixing times of300 and 400 ms (III), or 75 and 150 ms (IV), and HMBC experimentswith 60 and 90 ms (III), or 45 and 90 ms (IV) delays for theevolution of long-range couplings were used.

Rheological Measurements of the EPSs Produced by Lc. lactis ssp.Cremoris Strains

The viscosities of the dilute solutions of EPS at concentrationsof 0.01 up to 0.1 g dL-1 were measured at 25 °C with an Ubbelohdecapillary viscometer (536 13/Ic, SCHOTT35 GERÄTE, Hofheim,Germany), which allows the determination of the flow times withan accuracy of 0.03 s. The aqueous solutions of EPS with orwithout the addition of salt were prepared by dissolving ameasured amount of EPS in 0.1 M sodium chloride (NaCl) solutionor in deionized water. Sample dilution to the various requiredconcentrations of EPS was done directly in the viscometer. Afterabout 5 min for temperature equilibration, flow times were taken,and each flow time was reproduced six times. The reduced viscosityand intrinsic viscosity of the EPS solutions were calculated fromthe collected data.


The effect of temperature, pH and salts on the rheologicalbehavior of the EPS solutions (1%, w /v) was studied with aBohlin VOR rheometer (Bohlin Instruments Ltd, England) usingconcentric cylinders (C14) with a gap of 0.5 mm between theupper and lower geometries. The viscometry measurements wereperformed at 5, 25, 40 or 60 °C, with increasing shear rates up to291 s-1 in 29 steps. The pH of the EPS solutions was adjusted withlactic acid to 4.0, 5.0 and 6.5. The EPS solutions containing saltswere prepared by addition of EPS to 0.1 M NaCl or 0.1 M calciumchloride (CaCl2) solutions. The oscillation measurements wereperformed for the EPS (1%, w/v) in aqueous and salt solutions,and in skim milk at a frequency sweep from 0.01 to 15 Hz in 19steps. For every temperature-dependent measurement a thermalequilibration time of about 60 min was used.

RESULTS AND DISCUSSION

Antimicrobial Compounds Produced by LAB (I, II, unpublishedresults)

Separation, Purification and Identification of AntimicrobialCompounds

After ethanol precipitation of the cell-free cultures of the LABstrains, the obtained precipitates and the final supernatants weretested for antimicrobial activity. The supernatants containingantimicrobial activity were subjected to chromatographic separationand purification. The active fractions appeared in a relativelynarrow part of the chromatogram on gel-filtration on Bio-Gel P-2of the supernatants. Anion exchange chromatography of theseactive fractions resulted in two different ranges of active fractions(52-54 and 58- 63), fractions 58-63 containing lactic acid. Furtherpurification by RP-HPLC of the fractions 52- 54 gave rise to onemajor peak at retention time 4.96 min and two small peaks at 3.65and 5.78 min. The absorption maximum of the antimicrobialcompound was at 215 nm. Fractions of these peaks were collectedand tested for antimicrobial activity. Only the fraction of the peakat 4.96 min was found to contain antimicrobial activivity.Identification of this active fraction by both NMR (1H and 13C)and MS (EI and FAB) spectra indicated that the antimicrobialcompound was 2-pyrrolidone-5-carboxylic acid (PCA), also known

as pyroglutamic acid. Among the twenty LAB strains examined,thirteen Lactobacillus and five Pediococcus strains were found toproduce PCA under the growth conditions used in this study.Four Lactobacillus strains produced, in addition to PCA, also HMMantimicrobial compounds, as indicated by the presence ofantimicrobial activity in the precipitates obtained from ethanolprecipitation. Since these HMM compounds exhibited a rathernarrow range of activity they were not subjected to further studies.On the basis of the results of this study, it seems that many LABstrains, particularly Lactobacillus strains, are able to produce PCA.

Since LAB produce relatively large amounts of lactic acid,being antimicrobially active, it is important to remove the lacticacid in order to find other antimicrobial compounds, especiallythose of low molecular mass. In the separation and purificationprocedures developed in this study, both PCA and lactic acidwere present in the culture supernatant obtained from ethanolprecipitation, and they also appeared in almost the same range offractions after gel filtration. However, complete separation of lacticacid was achieved by anion exchange chromatography based ona gel matrix (Fractogel TSK) suitable for separation of biomolecules.Previously, size exclusion HPLC was used to separate lactic acidfrom LMM antimicrobials, but all the fractions obtained werefound to contain antimicrobial activity because the mobile phase(sodium phosphate) used was antimicrobially active.

Niku-Paavola et al. (1999) reported the separation of lacticacid by gel filtration on Sephadex G-10 with water as an eluentand found several LMM antimicrobial compounds produced byLb. plantarum. Although there were some reports on the separationand purification of LMM antimicrobial compounds produced byLAB, the techniques for seperation of lactic acid appeared not tobe clearly demonstrated. In addition, there were reports on the useof a neutralization technique to eliminate the antimicrobial effectof lactic acid, but lactic acid could not be separated with thistechnique, and the activity of acidic antimicrobials might besuppressed due to neutralization. PCA is a natural constituent offoods of plant origin, including vegetables and fruits, andfermented soybean and cereal products. Among LAB strains, onlyS. bovis has previously been shown to produce PCA by conversion


of glutamine. PCA can also be synthesized by heating glutamicacid using a dehydration process. It has been reported that PCAis able to increase cerebral blood flow and decrease the resistanceof brain vessels, which result in enhanced brain metabolism, i.e.increased glucose uptake and utilization by cerebral tissues anddecreased brain lactate dehydrogenase activity. Other biologicalfunctions of PCA are related to its presence as an amino-terminalresidue in many biologically significant peptides and proteins,e.g., eisenine, LH-RH (luteinizing hormone releasing hormone),and TRH (thyrotropin releasing hormone).

The Antimicrobial Activity of 2-pyrrolidone-5-carboxylic AcidProduced by LAB

Although LAB are able to produce a variety of antimicrobialcompounds, the present study, for the first time, reports that theproduction of a certain cyclic amino acid (PCA) is involved in theantimicrobial action of LAB. PCA has been considered, followingthe identification of reuterin, to be another well identified LMMantimicrobial compound produced by LAB.

The antimicrobial assay of PCA by agar diffusion methodshowed that PCA at 2% inhibited Bacillus subtilis, E. coli, Enterobacter,Klebsiella, and Pseudomonas strains. The most sensitive strainsEnterobacter cloacae 1575, Pseudomonas fluorescens KJL G, andPseudomonas putida 1560-2 were inhibited by PCA at 0.5% and1.0%. Among all bacterial strains tested, the Gram-positive strainssuch as LAB strains, and several Listeria and Staphylococcus strainswere not inhibited by PCA. An Enterococcus faecalis strain wasmoderately inhibited by 0.5% PCA in BHI broth at 37 °C. It seemedthat Gramnegative bacteria were more sensitive to PCA than theGram-positive ones. This is in agreement with previous studieswhich showed that organic acids (lactic and acetic acids) weremore inhibitory toward Gram-negative bacteria than Gram-positiveones. PCA was heat stable and the antimicrobial activity did notchange after heat treatments (63 °C 30 min, 72 °C 15 s or 121 °C20 min). Although raising the pH of the PCA in water solutionsreduced the antimicrobial activity, and completely destroyed theactivity at pH 3.8-4.0, the antimicrobial activity of PCA seemed tobe not solely due to the pH effect, but other factors may be possiblyinvolved. The turbidometric assay of antimicrobial activity showed

that PCA in broths was inhibitory against the test organisms atpH 5.0-5.89. At pH values when 1% PCA showed no activity, 2%PCA was still active. Earlier studies showed that neutralizationcaused the loss of the antimicrobial activity of Lb. acidophilus, sincemany inhibitory substances produced by lactic cultures wererelative stable in an acidic pH. In addition, the antimicrobialactivity of PCA, like other organic acids, may also be dependenton the undissociated form of the acid.

PCA was found to be less inhibitory than lactic acid. At thesame concentrations, lactic acid was more active than PCA againstseveral indicator organisms tested. During the course of this study,we also observed that the production of PCA varied with strainsand it was generally small when compared with the amount oflactic acid produced. However, the significance of PCA is that, inaddition to its wide spectrum of antimicrobial activity, it is alsoinvolved in many important biological functions as discussedabove. Among the PCA-producing LAB strains of this study, Lb.rhmnosus GG was found to produce relative large amounts of PCA.Lb. rhamnosus GG has also been reported to produce a microcin-like LMM antimicrobial compound, and the strain has beencommercially applied in the production of probiotic milk products,e.g. Gefilus in Finland. Considering the potential applications ofPCA, e.g. as PCA-producing starters in food preservation, furtherstudies are needed on the optimal production of PCA by LABstrains, the antimicrobial effects of PCA in foods and the effectivePCA concentrations, as well as the sensory quality of foods whenPCA is used.

EPSs Produced by LAB

Of the thirteen LAB strains examined, except for Lb. fermentumG.1.2.1, Lb. rhamnosus LC705 and Lc. lactis ssp. cremoris SEPH 11which did not produce EPSs, ten strains produced neutral oranionic EPSs of varying amounts when they were grown underthe conditions of this study.

Previous studies have shown that the production of EPS isgrowth-associated, and it is influenced by medium compositions,culture temperature and pH. Gamar et al. (1997) showed thatvariations in carbon sources in the growth medium resulted in


different yields of the EPS produced by Lb. rhamnosus strain C83.Addition of glucose or sucrose to the milk medium stimulated theEPS production and modified the sugar composition of the EPSproduced by Lb. casei ssp. casei NCIB 4114.

Several authors have also reported that more EPSs could beproduced by different LAB strains at lower temperatures, or withlimited nitrogen sources. Although bacteria grow well, and morecells may be obtained under optimized growth conditions,unfavourable culture conditions may stimulate EPS production bythe cells as a form of bacterial self-protection. Therefore, thesefactors discussed above need to be considered for optimizing EPSproduction by the LAB strains in this study.

EPSs Produced by Lb. Helveticus Strains

Lb. helveticus Äki4 grown in MRS broth, and strains Lb161and K16 grown in skim milk were found to produce EPSs.Viscometric measurements by Ubbelohdetype capillary viscometershowed that at 0.5% (w/v) EPS in aqueous solutions, the EPSs ofstrains Lb161 and K16 were viscous, but the EPS of strain Äki wasnot viscous. The primary molecular structures of the EPSs producedby strains Äki4 and Lb161 have been studied by sugar andmethylation analyses, and 1D and 2D NMR spectroscopy.

Structural Elucidation of the EPSs Produced by Lb. Helveticus Aki4(III) and Lb161 (IV)

The EPS produced by strain Äki4 was shown to be composedof a hexasaccharide repeating unit containing one terminal â-D-galactose, one 4- substituted á-D-galactose, one 6-substituted â-D-galactose, one 4-substituted â-D-glucose, one 3,6-substituted â-D-galactose and one 6-substituted â-D-glucose residue. Theassignment of the spin systems for the six sugar residues wasperformed using 2D homo- and heteronuclear techniques. Eachspin system with a specific sugar residue and substitution patternwas identified from the JH,H values, indicating the anomericconfiguration, and from the 1H and 13C NMR chemical shifts.The sequence of the sugar residues was determined using 2DNOESY and HMBC experiments. The structure of the repeatingunit of the EPS is as follows:

The EPS produced by strain Lb161 was shown to be composedof a heptasaccharide repeating unit containing two terminal â-D-glucose, one 2,3-substituted á-D-glucose, one 3- substituted á-D-galactose, one 4-substituted á-D-glucose, one 3,4-substituted â-D-galactose and one 3-substituted â-D-glucose residue. Theassignments of the spin systems for the seven sugar residues wereperformed using 2D homo- and heteronuclear techniques. Eachspin system with a specific sugar residue and substitution patternwas identified on the basis of the JH,H values and the 1H and 13CNMR chemical shifts. The sequence of the repeating unit wasestablished using 2D NOESY and HMBC experiments with thefollowing structure:

Structural Variations Among the EPSs Produced by Lb. HelveticusStrains

As shown above, the repeating units of the EPSs of Lb helveticusÄki4 and Lb161 are a hexamer and a heptamer made up of a mainchain branched by one and two side chains, respectively. Structuralstudies on the EPS of strain K16 showed that the repeating unitof the EPS was composed of six sugars: one terminal D-galactose,one terminal D-glucose, one 4- substituted D-galactose, one 4-substituted D-glucose, one 2,4-substituted D-glucose and one 4,6-substituted D-glucose.

Comparing with the so far published structures of the EPSs ofother four Lb. helveticus strains, the common structural feature ofthe EPSs produced by all these Lb. helveticus strains is that theEPSs contain either a hexa- or heptasaccharide repeating unit ofD-galactose and D-glucose. The EPS produced by Lb. helveticusstrain TY1-2 contains N-acetyl-D-glucosamine in addition to D-galactose and D-glucose. Rhamnose and other monosaccharideswhich have been found in the EPSs produced by many LABstrains are not present in the EPSs of the Lb. helveticus strains.Since the rheological properties of polysaccharides are closelyrelated to their primary molecular structures and threedimensionalstructures, it would be of interest to study the relation betweenstructures and functional properties of the EPSs produced by LABstrains within the same species. The difference in the viscositiesof the EPSs of the Lb. helveticus strains of this study is probablydue to their different primary molecular structures (linkage and


degree of branching), molecular masses, three-dimensionalstructures and chain-chain interactions of the polysaccharides insolutions.

EPSs Produced by Lc. Lactis ssp. Cremoris Strains and theirRheological Characterization (V)

The growth of Lc. lactis ssp. cremoris strains (ARH53, ARH74,ARH84, ARH87, B40) in skim milk at 25 °C for 18-20 h resultedin very viscous cultures as compared to the Lb. helveticus strainsof this study. The EPSs (164-263 mg L-1) produced by these ‘viili’strains were separated and purified from the cultures by treatmentwith 4% TCA and ethanol precipitation. The EPSs were shown tobe anionic in nature by anion exchange chromatography, andthey contained the same monosaccharides rhamnose, glucose andgalactose in similar molar ratios. The combined evidence fromsugar analysis and NMR spectroscopy indicates that the primarymolecular structures of these EPSs are identical or closely relatedto that from Lc. lactic ssp. cremoris SBT 0495. The underestimationof the ratio of galactose resulted from incomplete release by acidhydrolysis. The galactose residues were involved in thephophodiester linkage in the polysaccharide and it was hard torelease them.

In dilute aqueous solutions (up to 0.1 g dL-1), the EPSsproduced by Lc. lactis ssp. cremoris strains exhibited a polyelectrolyteeffect. At very low concentrations ( 0.02 g dL-1), an increase inreduced viscosity with a decrease in concentration was observed.In the presence of 0.1 M NaCl, the reduced viscosities of all theEPS solutions were significantly lowered, and the polyelectrolyteeffect disappeared as indicated by the straight lines of the Hugginsplot. The intrinsic viscosities of the EPSs were therefore obtainedby a linear extrapolation of the plots to zero EPS concentration.Since the EPS produced by strain ARH53 gave the highest intrinsicviscosity (19.62 dL g-1), this EPS might possess a relatively largemolecular size than the other Lc. lactis ssp. cremoris strains studied.At a higher concentration (1%, w/v), the viscosities of the aqueoussolutions of EPS produced by strain ARH53 were temperature, pHand salt dependent. At the same pH (4.0, 5.0 or 6.5), increasingtemperature from 5 °C to 60 °C caused a decrease in viscosity. Athigher temperatures (40 °C and 60 °C), the viscosity was clearly

lowered in the order of pH 6.5 > pH 5.0 > pH 4.0. The higherviscosity of EPS in 0.1 M CaCl2 solutions than that in 0.1 M NaClsolutions is due to different interactions between cations andpolyelectrolyte chains, Ca2+ being more effective than Na+ in thisrespect. Changing Na+ into Ca2+ ion was observed also to increasethe storage modulus of the sample. Samain et al. (1997) comparedthe effect of salts on the viscosity of a gel-forming EPS producedby Alteromonas sp. strain 1644 and showed that the divalentmagnesium cation was more effective than the monovalent sodiumcation.

The aqueous solutions of the EPS (1%, w/v) of strain ARH53behave viscoelastically. As shown in the oscillation measurementsat 5, 25 and 40 °C, the EPS solutions behaved as a viscoelasticfluid at lower frequencies with G” > G’, and showed a predominantelastic character (G’ > G”) at higher frequencies. The drastic decreasein viscosity of the EPS aqueous solution with increasing shear ratealso clearly shows the non-Newtonian behavior (shear-thinning)of the solution.

In skim milk to which the EPS of strain ARH53 was added,the viscosity was much higher as compared with the EPS inaqueous solutions at 5, 25 and 40 °C. At shear rates lower than9.21 s-1, an upward shift in viscosity was observed at 40 °C. Theviscosity was higher at 40 °C than at 25 °C, and even higher thanat 5 °C at shear rates lower than 0.37 s-1. In the oscillationmeasurements, at 5 °C there is a moderate frequency dependenceof G’ with a crossover of G’ and G” at the lower end of thefrequency range, suggesting the formation of a weak gel. At 25 °Cthe system exhibited a viscoelastic behavior with a crossover of G’and G” at about 0.8 Hz. Increasing temperature to 40 °C led to theformation of a strong gel, as indicated by G’ » G”, and the parallelcurves of G’ and G” with only a slight frequency dependence, andthe value of G’ being 198 Pascals at 15 Hz. The interactionsbetween EPS and casein micelles of skim milk were possiblyinvolved in the gelation. Hess et al. (1997) proposed that in yoghurtthe interactions between EPS and casein micelles are strongerthan those between the casein micelles. Intercalation of the EPSsinto the casein matrix was also involved in the improvement of thetexture of quarg using ropy cultures. It seemed that the interactions


between the EPS and casein micelles increased at highertemperatures as observed in this study.

The production of EPSs by Lc. lactic ssp. cremoris strains hasbeen reported earlier. However, the rheological properties of theseEPSs were poorly understood. Recently, Oba et al. (1999) reportedthe viscoelastic properties of the EPS produced by strain SBT 0495.It is of interest that the rheological behavior of this EPS in aqueoussolutions, e.g. polyelectrolyte effects and viscoelasticity, is similarto that of the EPS of strain ARH53. This is propably due to theirsimilar or identical structures as shown in this study. The similarityin rheological behavior was also noticed among the EPSs producedby the five ‘viili’ strains of this study. The difference in viscosityof the EPSs in dilute, as well as in concentrated solutions waspossibly due to their different molecular masses.

It has previously been known that slime production is animportant factor contributing to the special slimy characteristic ofFinnish fermented milk ‘viili’. The increase in viscosity and gellingat 40 °C resulted from the addition of the EPS produced by Lc.lactis ssp. cremoris demonstrated in this study is of interest withrespect to the possible use of the EPS to improve the rheologicalproperties of milk products, for instance, yoghurts withfermentation at about 42 °C. Considering the potential applicationsof the EPSs, future work on the optimization of the EPS productionand further physical characterization is required.

SUMMARY AND CONCLUSIONS

Lactic acid bacteria are able to produce a large variety ofcompounds which give fermented foods their characteristic flavorand color, and also impart improved safety and rheology to thefoods. In this study, the production of antimicrobial compoundsand extracellular polysaccharides by LAB strains obtained fromthe dairy industry has been investigated in order to find the LABstrains with potential applications in foods.

Studies on the twenty dairy LAB strains from Lactobacillus,Lactococcus, Pediococcus and Streptococcus showed that eighteenstrains produced a LMM antimicrobial compound, 2-pyrrolidone-5-carboxylic acid (PCA). Separation and purification of PCA fromthe growth media were achieved by ethanol precipitation and

chromatographic methods. The technique of anion exchangechromatography developed in this study was essential for effectiveseparation of PCA from lactic acid, facilitating the identificationof PCA by NMR and mass spectrometry. The chromatographicprocedure based on this technique can be used for separation andpurification of other LMM antimicrobial compounds. To ourknowledge, this is the first report of a cyclic amino acid, PCA,which is involved in the antimicrobial action of LAB. PCA can beconsidered as a well identified LMM antimicrobial compoundfollowing the identification of reuterin produced by Lb.reuteri.

PCA was shown to be inhibitory toward many food-bornespoilage bacteria such as Bacillus subtilis, Enterobacter, E. coli,Klebsiella and Pseudomonas. In antimicrobial tests against differentorganisms, the Gram-negative spoilage bacteria were found to bemore sensitive to PCA than the Gram-positive ones. Heat treatmentof PCA did not change its antimicrobial activity. Although PCAwas active under acidic conditions, it appeared that the activitywas not solely due to the pH effect. Regarding the mechanism ofantimicrobial action of PCA, more study is needed on the mode ofaction of PCA on sensitive bacterial cells and the MIC value, aswell as the biosynthesis of PCA in different species of LAB. Inaddition, further investigation on a wide range of LAB strains isneeded in order to find out whether production of PCA is acommon characteristic of LAB.

During the course of the studies on the antimicrobialcompounds, we gradually precipitated HMM antimicrobialcompounds from culture supernatants by ethanol, while at thesame time following how polysaccharides were precipitated. Thefirst EPS, which was precipitated at 40-60% ethanol, was foundto be produced by Lb. helveticus Äki4 grown in MRS broth. NMRspectroscopic studies of the EPS showed that it was composed ofa hexasaccharide repeating unit of D-galactose and D-glucose ina molar ratio of 2:1, and the main chain was branched with a sidechain of D-galactose. Although there were relatively large amountsof the EPS (~500 mg L-1) produced by Lb. helveticus Äki4, this EPSwas found to be not viscous.

In the screening study of LAB strains producing viscous EPSs,another Lb. helveticus strain, strain Lb161 grown in skim milk was

Introductory Food Microbiology216 Modelling the Growth Boundary of Staphylococcus Aureus 217

found to produce a viscous EPS. The EPS was isolated by thetreatment of the growth medium with TCA to precipitate theproteins first, and subsequently by ethanol precipitation to obtainthe polysaccharide. The primary molecular structure of the EPShas been elucidated by NMR spectroscopy. The EPS was shownto consist of a heptasaccharide repeating unit of D-galactose andD-glucose in a molar ratio 2:5 with two side chains, each consistingof a D-glucose residue.

Further studies on slime-forming Lc. lactis ssp. cremoris strainsshowed that growth of the strains ARH53, ARH74, ARH84, ARH87or B30 in skim milk resulted in very slimy cultures. By usingbasically the same methods as for the EPS of Lb. helveticus Lb161,polysacharides of varying amounts were isolated from thesecultures. Sugar analysis and NMR spectroscopy showed that theEPSs had a rather similar or probably identical structure to theone reported earlier. Rheological studies showed that the EPSs indilute aqueous solutions behaved as polyelectrolytes, and the EPSof strain ARH53 gave the highest intrinsic viscosity. Furthercharacterization of the EPS of strain ARH53 showed that theviscosity of the EPS in concentrated solutions was dependent onthe temperature, pH and ionic strength of the solutions. In skimmilk, addition of the EPS of strain ARH53 resulted in a clearincrease in viscosity and a gel was formed at 40 °C.

In our future studies of EPSs, we have planned to continue thestudy of the EPSs produced by LAB, as well as bifidobacterialstrains of food origins, and strains from the human intestine,aiming at applications in food, functional food or clinical products.Although the present study provides data on the structures andrheological properties of EPSs, further studies are needed in orderto understand the structure-function relations of the EPSs, theinteraction of polysaccharides and proteins (e.g. casein), and theroles of EPSs in their adhesion interaction with EPS-producingstrains in the human intestine.

8MODELLING THE GROWTH

BOUNDARY OF STAPHYLOCOCCUSAUREUS

Knowing the precise boundary for growth of Staphylococcusaureus is critical for food safety risk assessment, especially in theformulation of safe, shelf-stable foods with intermediate relativehumidity (RH) values. To date, most studies and resulting modelshave led to the presumption that S. aureus is osmotolerant.However, most studies and resulting models have focused ongrowth kinetics using NaCl as the humectant. In this study, glycerolwas used to investigate the effects of a glassforming nonionichumectant to avoid speci? c metabolic aspects of membrane iontransport. The experiments were designed to produce a growthboundary model as a tool for risk assessment.

The statistical effects and interactions of RH (84 to 95%adjusted by glycerol), initial pH (4.5 to 7.0 adjusted by HCl), andpotassium sorbate (0, 500, or 1,000 ppm) or calcium propionate (0,500, or 1,000 ppm) on the aerobic growth of a ? ve-strain S. aureuscocktail in brain heart infusion broth were explored. Inoculatedbroths were distributed into microtiter plates and incubated at37°C over appropriate saturated salt slurries to maintain RH.Growth was monitored by turbidity during a 24-week period.Toxin production was explored by enterotoxin assay. The 1,280generated data points were analyzed by SAS LIFEREG procedures,which showed all studied parameters significantly affected thegrowth responses of S. aureus with interactions between RH andpH. The resulting growth/no growth boundary is presented.

Introductory Food Microbiology Modelling the Growth Boundary of Staphylococcus Aureus218 219

INTRODUCTION

Staphylococcus aureus is one of the leading causes of foodborneillness and is ranked as one of the most prevalent causes ofgastroenteritis worldwide, even though the occurrences of thisfoodborne disease are grossly underreported. S. aureus wasdetermined to be the etiological agent in 367 (19.6%) of 1,869documented bacterial foodborne outbreaks in the United States.Approximately 25 major outbreaks of Staphylococcus food poisoningoccur annually in the United States. S. aureus has been estimatedby the Centers for Disease Control and Prevention to cause 185,060illnesses, 1,753 hospitalizations, and 2 deaths per year in theUnited States, all of which are via consumption of contaminatedfoods.

Because of its prevalence as a food poisoning organism, S.aureus has been extensively studied to determine the physical andchemical parameters that affect its growth and toxin formation. S.aureus is ubiquitous to the mucous membranes and skin of warm-blooded animals. It is a poor competitor with other bacteria andis easily destroyed by cooking temperatures; however, its toxinscan survive heat processing equivalent to that given to low-acidcanned foods. Staphylococcal food poisoning frequently occurswhen food is contaminated after cooking by a person carrying theorganism, and subsequently the food is temperature abused forseveral hours, which allows for toxin production beforeconsumption. It is the ingestion of the toxin that causes thefoodborne disease.

S. aureus is highly salt tolerant and has been reported to growat relative humidities (RHs) as low as 85% in NaCl concentrationsup to 25% (wt/wt) (11). The RHs limiting growth are typicallyhigher when humectants other than NaCl are used to control RH(11). Notermans and Heuvelman (19) reported that at a wateractivity of 0.98 to 0.93, sucrose was more favorable for growththan NaCl, but at a water activity of 0.87, NaCl was more favorable.Marshall et al. reported that glycerol inhibition of the growth rateof S. aureus was about 10% greater than that caused by NaCl atwater activity levels between 0.96 and 0.90. Empiricalmicrobiological models can be designed to predict howmicroorganisms relate to the environment.

Predictive models are typically broadly applied for thescreening of products and/or processes before challenge or shelflifestudies. However, it must be remembered that models are onlytools to be used to help design products with food safety concernsaddressed at the beginning of the development process and thatthere must be validation of the system through challenge or shelf-life studies with the real product.

Both the U.S. Department of Agriculture Pathogen ModelingProgram (PMP) and the Food MicroModel (FMM) developed in theUnited Kingdom through the Ministry of Agriculture, Fisheriesand Food include S. aureus growth kinetic models. The PMP modelsthe effects of temperature, initial pH, NaCl concentration, andsodium nitrite concentration in a broth system on aerobic andanaerobic growth kinetics of the organism.

The FMM models the growth responses as affected by NaClconcentration, pH, and storage temperature in a model brothsystem. The PMP also includes a model for the survival of S. aureusin nongrowth conditions, which was developed from a modelbroth system that includes the effects and interactions of pHcontrolled by lactate buffer, lactic acid concentration, NaClconcentration, sodium nitrite concentration, and varyingtemperatures.

For all of the above-mentionedmodels, NaCl was the humectantchosen to control the RH in the system, and the resulting modelsare kinetic in nature and, therefore, do not allow the boundary forgrowth/no growth to be estimated. In contrast to kinetic modeling,probability modeling focuses on determining if the microorganismof concern will or will not grow, in other words, determining thegrowth/no growth interface.

This becomes increasingly important when pathogens are ofconcern, because the rate of growth may be less important than thefact that the organism is present and may be able to grow to aninfectious dose or produce toxins.

In this study, the statistical effects and interactions ofRH controlled by glycerol, initial pH, and potassium sorbate orcalcium propionate on the boundary for S. aureus growth weremodeled



Organisms and Media Preparation

The . ve S. aureus strains used to create the bacterial cocktailused in this study were S. aureus ATCC 13565 (American TypeCulture Collection, Rockville, Md.), which produces staphylococcalenterotoxin A (SEA), ATCC 14458 (SEB), and ATCC 27664 (SEE);D-2 (SED) obtained from Toxin Technology, Inc. (Sarasota, Fla.);and A-100 (SEA) obtained from U.S. Army Natick Labs. All strainsshowed typical growth on Baird-Parker agar plates and werecoagulase positive. Their identi. cation as S. aureus was con- . rmedby use of a Riboprinter. Stock cultures were grown overnight at37°C in brain heart infusion broth, suspended in a 2:1 broth-glycerol solution, and stored at -80°C until needed. The BHI brothwas reconstituted with the appropriate ratio of water to glycerolto achieve the desired RHs for the study (84, 88, 92, or 95% RH).Appropriate amounts of 10% stock solutions of potassium sorbateor calcium propionate were added to achieve . nal concentrationsof 500 or 1,000 ppm. The pH was adjusted using 0.1 or 1.0 N HCl(J. T. Baker, Inc., Phillipsburg, N.J.; pH 4.5, 5.0, 6.0, or 7.0). FinalRH was determined using the Aqualab CX-2 water activity meter.It was determined that the RH of the media did not change morethan 60.003% when measured at ambient temperature versus 37°C.Each of the 80 broths was . lter sterilized and stored in screw-capped tubes at 4°C until needed.

Bioscreen microtiter plate preparation, incubation, andmeasurements.

The . ve stock cultures were removed from -80°C storage, andone loopful was transferred into 9 ml of BHI broth, separately. Theinoculated broths were incubated for 18 h at 37°C. The . ve cultureswere individually adjusted to an optical density (OD) at 530 nmof 0.750 to 0.780 (Perkin-Elmer 35 spectrophotometer) with 0.1%sterile peptone to achieve a concentration of approximately 108CFU/ml. Two milliliters of each diluted culture was transferredinto one sterile test tube and vortexed to create the S. aureus cocktailused in the studies. Three 1:10 dilutions with sterile 0.1% peptonewater were used to make a . nal working cocktail with aconcentration of approximately 105 CFU/ ml. This working

cocktail was kept on ice until used to inoculate the various mediaused in each experiment. A 1-ml sample of the working cocktailwas taken, serially diluted, spread plated onto Baird-Parker agarplates, incubated at 37°C for 48 h, and counted to determine initialconcentrations of cells.

Ten milliliters of each of the 80 broths as described above wasaseptically transferred to sterile tubes, and 100 ml of the S. aureuscocktail was added to the broth to achieve an initial concentrationof approximately 103 CFU/ml. The inoculated broth (400 ml) wasaseptically transferred into sterile, 100-well microtiter plates ineight replicate wells per medium. The outer wells of the microtiterplates were . lled with uninoculated medium to act as a partialmoisture loss barrier and as uninoculated controls. Each platecontained media at one RH level only. A piece of sterile Thermaseal. lm was placed on top of the open wells, and the lid was also inplace during incubation. The plates were placed on a rack in aplastic sealable container, which had the bottom . lled withappropriate saturated salt slurries to maintain the environmentalRH to ensure RH of the media in the plates remained as stable aspossible. The saturated salt solutions used were as follows: ZnSO4· H2O (83% RH @ 37°C), KNO (89% RH @ 37°C), KPO4 (93% RH@ 37°C), and K2Cr2O7 (96% RH @ 37°C). The containers wereclosed and placed into a 37°C incubator. The plates wereperiodically removed from incubation, and the OD of the 640individual wells was measured for up to 6 months using thewideband . lter on the Bioscreen C system (Labsystems Oy, Helsinki,Finland). The experimentswere repeated on separate dates. A totalof 1,280 wells were monitored.

Data Collection

The S. aureus population was considered to have shown growthwhen the Bioscreen wideband OD (ODwb) measurement increasedfrom an initial reading of 0.200 to 0.220 to 0.350 or higher. Timeto growth (TTG) was determined by calculating the geometricmean of the time of the last measurement that showed no growth(ODwb <0.350) and the . rst time point that showed growth (ODwb>0.350). In instances of no growth, TTG was censored at the . nalmeasurement time of 168 days.


OD measurements were made and transferred from theBioscreen C ASCII . le to a DOS text editor and then imported intoStatistica software for initial analysis. Data were then transferredto a Microsoft Excel spreadsheet to determine TTG and then toSAS for modeling using LIFEREG.


The TTG data along with a censoring indicator and the RH,pH, and potassium sorbate or calcium propionate levels wereinput data for construction of two separate models: one based onthe potassiumsorbate data set and one based on the calciumpropionate data set. These two data sets used a common block ofdata generated from the experimental conditions, which had nopreservatives added. The SAS LIFEREG procedure was used todevelop predictive models of ln TTG as a function of the factorsRH, pH, and preservative level. By default, the procedure . ts amodel to the log of the dependent variable.

The resulting model can easily be transformed to a regulartime scale. In growth modeling, there are often conditions whereno growth occurs; therefore, TTG is sometimes censored. Underthese conditions, ordinary least-squares regression is not applicable,and special procedures are required. The LIFEREG procedure canaccommodate such censored data and uses maximum likelihoodestimation methods to . nd regression coef. cients. When creatingmodels with the LIFEREG procedure, there is a need to have manyreplicates of conditions equally spaced out about the matrix, withapproximately 50% of the conditions allowing and 50% notallowing growth for best-. tting purposes. Therefore, the ranges ofconditions studied (RH, pH, and preservative levels) were chosento satisfy these requirements.

Toxin Production

Toxin production for experimental conditions close to thegrowth or no growth border was explored by enterotoxin assayusing the VIDAS Staph Enterotoxin Assay. The VIDAS StaphEnterotoxin Assay is a qualitative enzyme-linked • uorescentimmunoassay performed in the automated mini-VIDAS instrument(bioMerieux, Inc., Hazelwood, Mo.). After the . nal ODwbmeasurements were completed (24 weeks), the media from the

microtiter plates were removed for assay. The broth from each ofthe eight replicate wells was removed via pipet and combined intotwo microcentrifuge tubes. The samples were spun down for 10min, and the supernatant from the two tubes was combined. ThepH of the supernatant was adjusted with 1 N NaOH to a pH of6.0 to 8.0. A half a milliliter of the sample was placed into theappropriate well in a test strip, the strips were placed into themini-VIDAS, and the assays were run. A positive and negativecontrol were run with each set of test strips. The test is sensitiveto 1 ng of toxin per ml of sample.

RESULTS

A threshold ODwb value of 0.350 was used to score wellswith S. aureus growth (ODwb $ 0.350) versus wells with no growth(ODwb, 0.350). This threshold value was determined in two ways.First, ODwb measurements were compared with plate counts, andit was determined that an ODwb reading of 0.350 wasapproximately the . rst ODwb value where an increase in turbidityregularly correlated to an increase in plate counts, indicating theapproximate ODwb where the instrument was sensitive enough toaccurately determine an increase in cell numbers by a change inturbidity. Second, the data were analyzed to determine TTG ineach well using threshold ODwb values for growth of 0.300, 0.350,and 0.400, and models were created and compared. The modelscreated with the threshold value set at 0.300 did not makebiological sense. We suspect that this was because this ODwbcorresponded to a level that is lower than the sensitivity of theinstrument to measure an increase in cell numbers. When thethreshold ODwb value was raised to 0.400, the models were similarto those created with the TTG data when the threshold ODwb of0.350 was used; however, the TTG data when the threshold ODwbof 0.350 was used gave a slightly more conservative model. Forthese reasons, the models were developed with TTG data obtainedby making the ODwb of 0.350 or more the lower limit fordetermining that growth had occurred in a particular well.

The models included all three main effects: RH (rh), pH (ph),and potassium sorbate (sorb) or calcium propionate (cal), theirquadratic effects (RH·RH 5 rh2, pH·pH 5 ph2, and potassium


sorbate·potassium sorbate 5 sorb2 or calcium propionate· calciumpropionate 5 cal2), and the three two-factor interactions (rh ·ph, rh·sorb, ph·sorb or rh ·ph, rh ·cal, ph· cal). These main effects, quadraticeffects, and interactions are collectively referred to as the factors.LIFEREG outputs a table of regression coef. cient estimates andapproximate chi-square distribution P values for each factor in themodel. The relative importance of each factor can be judged by theP value: factors with small P values are most in• uential andpredictive of the TTG. LIFEREG also allows the user to specify theerror distribution to account for the variation in TTG not explainedby the regression model. For model development purposes, Weibull,lognormal, and log-logistic distributions were considered. It wasfound that all three distributions gave very similar models; theregression coef. cients were similar in sign and magnitude. Sincethe three distributions are not all from the same class ofdistributions, it is not possible to formally test for goodness of . tusing likelihood ratio tests. However, since all models producedsimilar regression equations, we selected the log-logisticdistribution model, because it had the largest sample log likelihood.

The model development process is an iterative process. The .rst analysis of the data created models that were all inclusive; thisallowed for the determination of those factors that had a signi.cant effect on the growth of S. aureus. All of the main effects,quadratic effects, and two-factor interactions were signi. cant foreach model (P <0.005) except for the interaction between RH andpotassium sorbate in the potassium sorbate model and the calciumpropionate quadratic term in the calcium propionate model. Forthis reason, in each model the insigni. cant factor was droppedand the data were reanalyzed to develop the . nal models. Theresulting model equation from the calcium propionate data was2.250 to 3.703· rh - 1.488·ph + 0.418· cal + 1.572· rh2 + 0.343·ph2

+ 0.814· rh·ph - 0.130·rh · cal - 0.221·ph·cal, and the resultingmodel equation from the potassium sorbate data was 2.624 to3.938·rh - 2.421·ph + 0.907· sorb + 1.638· rh2 + 0.887·ph2 - 0.190·sorb2 + 0.903·rh ·ph - 0.756·ph·sorb. From the model equations, theTTG contour plots for calcium propionate and for potassiumsorbate were created. For diagnostic purposes, plots of ln predictedtime to growth versus ln observed time to growth were examined.

Plots of residuals versus main effects were also examined, and nounusual patterns or anomalies were detected.

There was a slight difference in the curvature of the contourlines in the plots for the media with no preservatives data for eachmodel, despite the fact that the same data for the no preservativecondition were used for developing each model. This differenceoccurs because two distinctly different mathematicalmodels wereproduced by the analysis, and when different models are used topredict TTG at a zero preservative level, they will produce slightlydifferent predictions. It should be noted that the differences aresmall and mainly in the low pH areas of the plots. In practice, thetarget product formulation would be well away from the growthboundary to allow for inherent process variation.

All of the models make good biological sense. As conditionsbecome more and more unfavorable for growth, the contour linesare closer together, indicating conditions are approaching thosethat do not allow growth of the organism. As the RH or pH of thesystem decreases, a corresponding increase in TTG is seen, andthe no growth area of the contour plot increases in size. Theaddition of potassium sorbate at low pH (pH 4.5 to 5.5)dramatically changes the contour plots with steep curvature in thelow pH area of the plots. The no growth boundary with respectto RH is increased from approximately 89% to approximately 93%when 1,000 ppm of potassium sorbate is present at pH 4.5. Thiseffect is not as evident when calcium propionate is present; the nogrowth boundary with respect to RH only increased approximately1.5% from 88 to 89.5% when 1,000 ppm of calcium propionate ispresent at pH 4.5. There is little change in the boundary withrespect to RH when the system is at pH 7.0 with the addition ofeither preservative studied.

Various regions on the contour plots generated by the modelswere explored with enterotoxin assays for all conditions wherethe . nal ODwb measurement was borderline. Also, assays wereconducted for all samples with treatments at 84% RH, pH 7.0, and84% RH, pH 4.5, although all ODwb measurements for treatmentsat 84% RH were less than 0.350. The results are indicated by aplus sign (toxin present) or minus sign (no toxin present) on thecontour plots. A total of 71 assays were run, and in every case


where the data were scored as ‘‘no growth,’’ the toxin assayresults were negative for toxin, and in every case where the datawere scored as ‘‘growth,’’ the toxin assay results were positive forthe presence of toxin. This further supports the choice of an ODwbof 0.350 as the correct cutoff value for time to growth in theindividual wells.

DISCUSSION

Comparison of the models developed in this study with otherpublished work is dif. cult, because most previously publishedstudies have used NaCl to control the RH of the system, whereasthis study has used glycerol. The limits of S. aureus growth, or thegrowth of any microorganism, cannot be predicted by the RH ofthe system alone; instead, both the RH of the system and thephysical properties of the humectant used should be considered.For example, NaCl is a non–glass-forming ionic humectant thathas specific effects on membrane transport systems. Therefore,both the osmotic and ionic stresses placed on bacterial cells byNaCl could affect the ability of those cells to grow. Glycerol is anonionic glass former that passively permeates the cell membrane.Therefore, glycerol would not have a direct effect on ion transportsystems. Unlike NaCl, glycerol forms an aqueous glass. The glasstransition temperature (Tg) is a specific characteristic for eachglass-forming compound. The influence of a solute’s Tg onmolecular mobility directly affects the amount of osmotic stress aspecific solute places on cells. For these reasons, it should beexpected that the limits of growth based on RH would be differentwhen various humectants are used to control RH, and, therefore,comparing studies based on RH alone could lead to falseconclusions.

The differences in modeling approaches also makecomparisons dif. cult. Both the FMM and PMP have very fewkinetic curves generated in which cells would be under stressedconditions, the most dif. cult conditions in which to obtain kineticinformation. Also, there are few true replicates in these areas, withmost of the experimental design similar to a central compositedesign in which most of the replicates are in the middle of thedesign. Kinetic models also model the mean value, which, due to

natural biovariability in the bacterial population, can bemeaningless under stressed conditions. The LIFEREG procedureused to create the boundary models handles this by . tting thedistribution to the error term, hence allowing probabilities to becalculated. This is useful for risk assessment purposes and allowsthe model to be easily used in Monte Carlo simulations. Oneconsequence of modeling this way is that there is a need to havemany replicates equally spaced about the matrix, withapproximately 50% of the conditions allowing and 50% notallowing growth for best-. tting purposes. The log time to growthis used to normalize the variance.

It has generally been accepted that the limits for growth of S.aureus are 85% RH when NaCl is used as the humectant and 89%when glycerol is used. Our models, based on glycerol, show thelimiting RH for growth to be approximately 86% (at pH 7.0),which disagrees with work published by Marshall et al., whofound the growth limits were 89 and 86% RH when glycerol andNaCl were used as the humectants, respectively. This disagreementcould be due to strain variation or to the fact that only quarter-strength BHI broth was used in the work by Marshall et al.,whereas full-strength BHI broth was used in our model system.Shapero et al. reported growth of S. aureus at 88% RH on trypticsoy agar plates with RH adjusted with glycerol but did not conductexperiments with the system at any lower RH values.

Studies with sucrose (a glass former that is not cell membranepermeable and has its own Tg) used as the humectant showedgrowth at 90% RH but not at 87% RH, and Scott reported growthat 88 but not at 86% RH. Broughall et al. developed models forgrowth of S. aureus in UHT milk, with RH adjusted by the additionof glucose (a glass former that is not cell membrane permeable andhas a different Tg). They found growth at 88% RH but did notexplore the RH boundary further, because the additional glucosecaused the system to become saturated. There is published literaturethat describes the ability of S. aureus to grow in various foodsystems, mainly meats and cheeses, but most of these systemsused NaCl as the RH depressant and, therefore, cannot be usedto validate the models created with glycerol. Often in experimentsconducted with food systems, the pH of the product is not reported,


nor is the list of ingredients published, which also leads to dif.culty when trying to compare data from various published studies.

The effectiveness of potassium sorbate to inhibit growth wasconsiderably greater compared with calcium propionate, whichwas especially noticeable at low pH. Potassium sorbate andcalcium propionate have some key similarities that may lead tothe belief that they should act similarly as preservatives; potassiumsorbate has a pKa value of 4.74 and a molecular weight of 150.2,whereas calcium propionate has a pKa value of 4.87 and molecularweight of 186.2.

Both have been reported in the literature to act as weak acids;the molecules remain undissociated at low pH, diffuse throughthe bulk lipids of the cell membrane, then enter and dissociate inthe cytoplasm, where the pH is close to neutral. It is the dissociatedmolecules that decrease the internal pH of the cell, which mayprevent growth by perturbing metabolism. Recent studies haveshown that sorbic acid does have inhibitory effects in theundissociated form, suggesting not only that it may act as a weakacid but also that it may have a secondary mode of action associatedwith membrane-resident sorbic acid that acts as an uncoupler,which may disable active transport. It has also been reported thatfor S. aureus the minimum inhibitory concentration of propionicacid is much greater than sorbic acid. This suggests that the levelsof calcium propionate used in this study that are in line with thetypical use levels in foods may have been too low to observe theinhibitory effect of calcium propionate on S. aureus growth.

In summary, two models describing the growth boundariesfor S. aureus with respect to RH controlled by glycerol, pH, andpreservative were developed. The growth boundaries were exploredwith toxin assays. For the . rst time, these models describe thegrowth/no growth boundaries for S. aureus with respect to RHcontrolled by glycerol and pH in addition to the preservativespotassium sorbate and calcium propionate. These models willallow product developers to visualize the ‘‘safe space’’ for theformulation of shelf-stable intermediate moisture foods using thesepreservation factors, will allow microbiologists to assess risksmore effectively over a wide range of products, and will ultimatelyallow the consumer to have greater assurance of food safety.

Molecular methods for detection of probiotics and intestinalmicrobiota and evaluation of Lactobacillus brevis as a potentialprobiotic dietary adjunct.

The human gastrointestinal (GI) tract harbours an extremelycomplex microbiota mainly composed of fastidious anaerobicorganisms. Because these microbes have a profound impact onhost’s health, modulation of microbiota with probiotic bacteriahas been proposed. However, the mechanisms of action of boththe GI microbes and the proposed probiotics remain obscure. Togain more information, molecular identification methods formembers of the GI microbiota and the probiotic strains are needed.Quick, robust methods are necessary for obtaining an overall imageof changes in GI microbiota, and more sophisticated methods areneeded for following up selected species or strains.

In this study, new methods were tested for their applicabilityin monitoring GI microbiota and probiotic strains. Strain-levelgenetic labelling without introduction of foreign DNA wasdemonstrated with Lactobacillus helveticus CNRZ32 by insertinga site containing silent mutations into the chromosomal pepXgene. Intact phenotypic properties of the mutated strain wereconfirmed with a peptidase assay. The mutated and wild-typestrains could be detected from faeces and milk with the help ofspecific primers. Thus, the labelling method could be used forspecific marking of industrial or probiotic strains in a ‘food-grade’manner provided that a suitable target gene and genetictransformation tools are available.

For analysis of GI microbes and selected intestinal or dairylactic acid bacteria, several species- or group-specificoligonucleotide primers and probes were designed and testedwith three different techniques. A polymerase chain reaction -enzyme-linked immunosorbent assay (PCR-ELISA) applicationwith simultaneous utilisation of multiple species- or group-specificoligonucleotide probes was suitable for detection of predominantmembers present in a mixed bacterial population. Sensitivity ofthe method was improved by using primers selective for the genusBifidobacterium. Comparison of dot blot hybridisation and real-time PCR demonstrated the superior properties of realtime PCR fordetection and quantification of bacterial ribosomal DNA targets.


The final part of this study comprised the evaluation of twoLactobacillus brevis strains ATCC 8287 and ATCC 14869T assupplements in dairy products. The L. brevis strains showedpromising in vitro antagonistic properties towards selectedpotentially harmful microbes and were suitable as supplementarystrains in yoghurt, producing no undesirable side effects on thequality or preservation of products. A small-scale feeding studydemonstrated the survival of L. brevis ATCC 8287 in the humanGI tract, indicating, together with the favourable antagonisticproperties, that this strain could be a candidate for use as aprobiotic supplement in dairy products.

Molecular methods have facilitated culture-independentstudies of gastrointestinal (GI) tract microbes. The GI microbiotais mainly composed of anaerobic organisms and moreover, directmolecular approaches have confirmed the abundance ofuncultivable microbes in intestinal samples. The value of molecularmethods for studying GI tract microbes is therefore immense.

GI microbiota can be influenced by probiotic bacteria. Theprobiotics, defined as “live microbial food supplements whichbeneficially affect the host by improving the intestinal microbialbalance", have been proposed to possess several advantageousproperties, such as antagonistic actions, production ofantimicrobial substances, modulation of immune responses andan impact on the metabolic activities of the gut.

Probiotic bacteria seem to hold great promise for treatment ofgastrointestinal disorders, yet further studies are required to createa more scientific basis for the probiotic action. Furthermore, as nosingle strain is likely to have all the aforementioned beneficialproperties, screening of new strains for probiotic potential isnecessary.

Although several applications, such as dot blot hybridisation,fluorescence in situ hybridisation (FISH), denaturing gradient gelelectrophoresis (DGGE) or thermal gradient gel electrophoresis(TGGE), and polymerase chain reaction (PCR) with species- orgroup-specific primers, are available for the detection andsemiquantitative analysis of GI microbiota, improvements arerequired, especially in relation to sensitivity, cost and quantificationpower of the methods. This is fundamental for better understanding

of the GI tract microbiota and the effects of probiotic bacteria onthese microbes.

REVIEW OF THE LITERATURE

Intestinal Microbiota

An extremely complex microbiota which has a profound impacton host’s health. The normal gut bacterial population of an adultis estimated to comprise more than 400 species, with apredominance of obligate anaerobes. The total number of microbespresent in one gram of intestinal content varies from less than 103microbes in the stomach to 104 – 107 microbes in the small intestineand 1010 – 1012 microbes in the colon. Indeed, the quantity ofmicrobes present in the intestine (about 1014) exceeds 10-fold thetotal number of all human cells. Until recently, analysis of intestinalbacteria has mainly been based on cultivation-dependent methods.Culture-independent studies have however confirmed that only afraction of the organisms present in faeces are cultivable, therefore,the results obtained by cultivation are likely to be biased.

Generally, Bacteroides, Eubacterium, Clostridium, Ruminococcus,Peptococcus, Peptostreptococcus, Bifidobacterium and Fusobacterium arereported to constitute the majority of microbiota. Molecular analyses,however, suggest that most faecal bacteria belong to a fewphylogenetic lineages composed of organisms from several genera.Bacteria related to Bacteroides, Prevotella and Porphyromonas seemto represent one-third of bacteria present in faeces, whereasClostridium leptum subgroup and Clostridium coccoides groups bothaccount for approximately one-fifth of the faecal bacterialpopulations. However, besides being an extremely diversemicrobial ecosystem, the intestinal microbiota appears to be uniquefor each individual. The normal microbiota of the GI tract worksas a barrier against pathogens, contributes to degradation of somefood components, stimulates the host immune system, andproduces certain B vitamins, enzymes and short-chain fatty acids.The microbes can also metabolise potentially carcinogenicsubstances and drugs in either a beneficial or a disadvantageousway. However, the role and action of individual microbial speciesor groups present in the GI tract are poorly known. Starting atbirth, the microbiota develops in a successional manner. The first


colonising microbes include bifidobacteria, enterobacteria,clostridia, enterococci and ruminococci. Recent studies havesuggested the importance of the type of colonising bacteria on thedevelopment of the gut immune system. A role in the developmentof food allergies has been postulated for the intestinal microbiotadue to the occurrence of an adult-like bifidobacterial speciescomposition in the intestine of allergic infants. Significant structuralchanges occur in the microbiota with age, including reduction ofbifidobacteria as well as an increase in the diversity of Atopobiumcluster species present in faeces. Nutritional aspects, however,have a profound effect on the composition of the intestinal microbialpopulation.

The Probiotics

Probiotic bacteria have been defined as “live microbial foodsupplements which beneficially affect the host by improving theintestinal microbial balance." Probiotic bacteria are increasinglyutilised in human food as well as in animal feed products.However, composition of the intestinal microbiota is poorly known,which hinders understanding of the probiotic functions. A probioticstrain should be of host origin, non-pathogenic, technologicallysuitable for industrial processes, acid- and bile-fast, adhere to thegut epithelial tissue, persist in the gastrointestinal tract for shortperiods, produce antimicrobial substances, modulate immuneresponses and influence the metabolic activities of the gut. Theproperties of the strain should be well documented. Althoughsome criteria, such as the non-pathogenic status, technologicalsuitability and careful documentation of the probiotic effects of amicrobial strain, are invariably required, no single strain is likelyto carry all of the abovementioned properties. Moreover, probioticproperties are considered strain-specific, and results obtained withone strain cannot therefore be claimed for another, even a closelyrelated strain. Microbes used in probiotic products include strainsfrom several Lactobacillus and Bifidobacterium species, Enterococcusfaecalis and Enterococcus faecium, Lactococcus lactis, Leuconostocmesenteroides, Pediococcus acidilactici, and Sporolactobacillus inulinus,Streptococcus thermophilus, as well as Bacillus cereus, Escherichia coli,Propionibacterium freudenreichii, Saccharomyces cerevisiae andSaccharomyces boulardii. However, lactobacilli and bifidobacteria

are most common in probiotic products designed for human use;probiotic properties of these bacteria are also the best studied.

Lactobacilli belong to the lactic acid bacteria, which comprisea diverse group of Gram-positive bacteria, most typicallyrepresented by non-sporing, catalase-negative, devoid ofcytochromes, non-aerobic but aerotolerant, fastidious and acid-tolerant cocci or rods producing lactic acid as the major end-product during the fermentation of carbohydrates. Being offastidious nature, lactic acid bacteria require a rich environmentfor growth, such as decaying plant material, food products and amammalian gastrointestinal tract or vagina. The genus Lactobacillusis heterogeneous, containing species with 32-53% G+C of thechromosomal DNA arranged into three groups based on differencesin sugar metabolism caused by the presence or absence of fructose-1,6-diphosphate aldolase and phosphoketolase. Althoughpossessing some phenotypical features common for lactic acidbacteria, the genus Bifidobacterium is actually related to theActinomycetes branch, having a high chromosomal G+C content.Sugar metabolism of bifidobacteria differs from that of lactic acidbacteria; the bifidobacteria lack aldolase and glucose-6-phosphatedehydrogenase, and hexose sugars are exclusively degraded bythe fructose-6-phosphate pathway characterised by fructose-6-phosphate phosphoketolase. Bifidobacteria are predominantmembers of the human intestinal microbiota, with bacterial countsof 109-1011 per gram of stool, with B. bifidum, B. longum, B. infantis,B. breve, B. adolescentis, B. angulatum, B. catenulatum, B.pseudocatenulatum, and B. dentium reported as human isolates.

Several positive effects have been proposed for probioticlactobacilli and bifidobacteria. Antagonism towards intestinalpathogens has been demonstrated for probiotics. Alleviation ofdiarrhoea is a well-documented characteristic of some strains, andparticularly the ability of Lactobacillus rhamnosus GG to shorten theduration of acute rotavirus diarrhoea has been established.Stimulation of the gut immune system by probiotic strains hasbeen reported, and induction of cytokine profiles has been shownto be strain-dependent. Alleviation of allergic reactions in thegastrointestinal tract has also been suggested for some strains:using a mouse model, Murosaki et al. (1998) showed that heat-


killed Lactobacillus plantarum L-137 suppressed production ofantigenspecific IgE by stimulating the production of IL-12. Inaddition, Lactobacillus rhamnosus GG and Bifidobacterium lactis Bb-12 have been demonstrated to relieve symptoms of atopic eczema.Administration of L. rhamnosus GG has been shown to enhancethe production of IL-10, acting thus as an anti-inflammatorymediator in atopic children.

Consumption of probiotics has also been observed to loweractivities of harmful faecal enzymes. Some animal models haveeven suggested the ability of probiotic strains to reduce the incidenceof cancer. Finally, probiotics may have significance for alleviationof intestinal disturbances. For example, ingestion of a mixture oflactobacillar and bifidobacterial strains together with a Streptococcussalivarius ssp. thermophilus was observed to help maintain remissionstatus in ulcerative colitis patients.

Although lactobacilli and bifidobacteria are stated as GRAS(generally recognized as safe) organisms due to their long usageand non-pathogenic status, safety issues have been investigated.Isolation of these organisms from infections has been reportedwith a low incidence, and in most cases, bifidobacteria andlactobacilli were associated with immunocompromised patients.One major concern among lactic acid bacteria is, however,resistance to vancomycin, especially within the genus Enterococcuswhere the resistance has been shown to be transferable.Vancomycin resistance has been reported for lactobacilli by severalauthors, but it is generally considered to be an intrinsic property.Vancomycin resistance expressed by the probiotic strain L.rhamnosus GG has been studied in detail.

Strain GG was not observed to transfer vancomycin resistanceor receive other resistance elements from enterococci, nor were anygenes resembling enterococcal vancomycin resistance genesfound in Lactobacillus GG. Similarly, Klein et al. (2000) reportedno indications for the presence of the vanA gene cluster, the vanBgene or the vanC gene from five L. reuteri strains or L. rhamnosusGG, suggesting that the vancomycin resistance of the strainsstudied is unrelated to the acquired resistance in the Enterococcusspecies.

Genetic Labelling of Lactic Acid Bacteria

Insertion of an extra DNA label into a target strain genomeprovides a way to monitor the specific strain in variousenvironments. However, introduction of foreign DNA is oftenrequired, and in many cases, genetic elements producing aphenotypic change are used to distinguish the marked strain.Genetic elements encoding resistance to an antibiotic have beenutilised for labelling of lactic acid bacteria, but the increasingspread of antibiotic resistance factors between bacterial species orgenera make such approaches unsuitable for strains intended forhuman or animal use. Therefore, usage of various ‘food-grade’markers has been suggested. Labelling of lactic acid bacteria witha plasmid-encoded green fluorescent protein gene placed underan inducible promoter has been reported for Lactococcus lactis andLactobacillus plantarum, whereas Allison and Klaenhammer (1996)suggested the use of a native Lactobacillus gene encoding immunityto Lactacin F as a food-grade genetic marker. A similar role hasalso been demonstrated for ltnI conferring immunity to lacticin3147. Site-specific integration of desired genetic elements intobacterial chromosomes through phage attachment sites has beendescribed for Lactococcus lactis, Lactobacillus delbrueckii andLactobacillus plantarum.

Legislation as well as consumer acceptance limit the usage ofgenetically engineered organisms, and therefore, genetic changesintroduced to target organisms must be carefully considered.Maguin et al. (1996), for example, combined insertion sequenceISS1 with a thermosensitive replicon, enabling a high frequency ofrandom insertion (about 1%) with Lactococcus, Enterococcus andStreptococcus thermophilus, while efficient excision of the plasmidgenerated stable mutants with no foreign markers, leaving only asingle ISS1 copy at the mutated site. However, the smallestdetectable alteration introduced to a target organism is changingone or a few bases in a gene-coding sequence without affecting theamino acid sequence of the corresponding gene product. Inprinciple, such a mutation could be caused by natural processesdue to the degeneracy present in the genetic code.

Silent mutations have been utilised for modification of aLactococcus lactis subsp. cremoris strain plasmid-encoded proteinase


prtP gene by in vitro mutation of the third positions of four adjacentcodons, thus providing a genetic label with no phenotypic effects.Although technically more demanding, genetic marking should,however, preferably be directed to gene replacement on achromosomal locus to ensure maximal stability of the alterationsintroduced.

Utilisation of Nucleic Acid Based Methods for Identification andMonitoring of Bacteria in Population Samples

Utilisation of Methods Independent of Prior Knowledge onSequence Data: Cloning and sequencing of the 16S ribosomalDNA (rDNA) pools in a population sample provides a method forobtaining sequence-level information on uncultivable bacteriaabundantly present in various parts of the gastrointestinal tract.Suau et al. (1999) analysed the sequence of 284 16S rDNA clonesderived from one faecal sample and classified the clones into 82molecular species, using 98% similarity criteria for a species.Importantly, only 24% of the molecular species were derived froma described organism. Similarly, comparative 16S rDNA sequenceanalysis of the intestinal bacterial community in pigs revealedonly a 17% fraction of previously described organisms among atotal of 375 phylotypes. In a large-scale study of subgingival plaquesamples, including analysis of a sequence of 2522 16S rDNAclones, 347 species was observed among the clones studied, whichcorrelated well with a previous estimate of expected speciesdiversity in the oral cavity. Thus, the direct cloning approachseems to give a good idea of the total microbiota present in acomplex population. It also facilitates determination of species-level differences between bacterial populations, as shown by thediscovery of novel phylotypes and species not previouslyassociated with childhood caries by comparing of the oralmicrobiota of a healthy subject and a subject with early childhoodcaries.

Approaches based on cloning are, however, rather tediousand are not optimal for analysis of large numbers of samples. Likeall PCR-dependent methods, construction of clone libraries maybe prone to biasing, possibly leading to falsification of the librarystructure. Nevertheless, with the help of cloning, design ofphylogenetically relevant oligonucleotide probes is enabled.

Denaturing Gradient Gel Electrophoresis (DGGE) and ThermalGradient Gel Electrophoresis (TGGE)

Denaturing gradient gel electrophoresis (DGGE) and thermalgradient gel electrophoresis (TGGE) have become popular methodsfor analysis of microbial populations present in various habitatssuch as water ecosystems, microbial fermentations, and the GItract. These methods allow separation of nucleic acid moleculesbased on their size and sequential differences. Thus, a populationof DNA molecules, such as the 16S ribosomal RNA (rRNA) orPCR-amplified 16S rDNA, can be studied and predominantmembers of the population identified via sequencing of isolatednucleic acid bands. Sensitivity of gradient gel electrophoresis isaffected by the choice of PCR primers used. For example, utilisationof universal 16S primers limits the sensitivity to detection of 1%subpopulations. By contrast, utilisation of species- or group-specific primers may allow detection and identification of bacteriarepresenting a minority of the total population.

Drawbacks of gradient gel electrophoresis include difficulty ofcomparisons between individual gels, requirement of carefuladjustments, and the need for sequencing to confirm identities ofbands seen in a gel. Furthermore, some bacteria may remainunidentified due to low resolution of DNA bands. With bothDGGEand TGGE-based methods, sequencing is required for correctidentification of individual bands seen in the gel. As stated bySchmalenberger et al. (2001), intraspecies operon heterogeneitiesmay significantly contribute to genetic profiles in microbialcommunity analysis, as amplification of one bacterial DNA mayyield several separate bands, which can then be wronglyinterpreted as high microbial diversity. Depending on the 16Sbinding universal primer pairs used, single-strand conformationpolymorphism analysis (SSCP) revealed an average of 1.7 – 2.3bands per pure cultured bacterial organism.

Terminal Restriction Fragment Length Polymorphism (T-RFLP)

Terminal restriction fragment length polymorphism (T-RFLP)is based on endonuclease digestion of PCR-amplified DNA andcapillary electrophoresis analysis of the terminal restrictionfragment (TRF) containing a fluorescent label. Terminal restriction


patterns have been used to analyse marine bacterioplanktoncommunities as well as faecal bacteria. The method is, however,limited by the choice of primers, which can, with their differentaffinities, dramatically change the patterns observed, while anotherproblem is the TRF length overlap by phylogenetically distantbacteria.

Analysis of Community DNA Profiles

Analysis of total bacterial community structure can beaccomplished by measurement of guanosine-cytosine profiles of apopulation DNA sample with DNA reassociation and densitygradient centrifugation. Density gradient fractionation ofcommunity DNA enables analysis of interesting fractions bycloning of PCR-amplified DNA pools, followed by sequencing.Density gradient centrifugation has been applied in analysis ofintestinal microbial guanosinecytosine profiles. However, becausethis approach requires sophisticated and expensive equipment, itis out of reach for most research laboratories.

Utilisation of Specific Oligonucleotide Primers or Probes

Hybridisation: Hybridisation offers a means for direct semi-quantitative monitoring of population samples. A very highsensitivity for detection of DNA targets can be obtained withradioactively labelled probes. However, because of its abundancein bacterial cells, rRNA provides for a more attractive target forhybridisation studies. Indeed, hybridisation assays targeting rDNAhave been verified as 10-fold less sensitive than assays for rRNA.Dot blot hybridisation with rRNA – targeted probes has been usedfor semiquantitative analysis of ruminal microbes and intestinalmicrobiota. Comparison of hybridisation results of a specific probewith a universal probe enables assessment of the target bacterialproportion present in a sample. At best, detection of a 0.1 – 0.01%rRNA subpopulation has been reported, corresponding toapproximately 107 target cells if 1011 bacteria are considered tobe present in a gram of faeces. Use of radioactive isotopes, especiallyin large-scale and long-term analyses, is, however, complicated bythe short half-lives of labels. Nonradioactively labelled probes aremore convenient, but their sensitivity is limited. Preenrichment ofthe target DNA by polymerase chain reaction can be used to

enrich the target DNA. For example, 0.1 pg of B. distasonis or B.thetaiotaomicron DNA (approximately 10-20 cells) or 0.01 pg of B.vulgatus DNA (1-2 cells) applied to a PCR was sufficient to producea positive hybridisation signal with specific digoxigeninlabelledprobes.

An additional factor of concern for filter hybridisations is theneed to label each probe separately and to perform hybridisationreactions after optimisations in separate vessels. However,Ehrmann et al. (1994) described a reverse dot blot hybridisationwith several membrane-bound oligonucleotide probes foridentification of lactic acid bacteria present in mixed populations.Similarly, Becker et al. (2002) used simultaneous detection of PCR-amplified rDNA of 23 oral bacterial species or groups witholigonucleotide probes.

In the fluorescence in situ hybridisation (FISH) method, thebacterial cell samples to be studied are immobilised on microscopeslides and made permeable for fluorescently labelledoligonucleotides with subsequent microscopic observation of thehybridisation signal intensities.

FISH has been used for monitoring faecal microbiotas, anddetection of a 1 – 0.1% bacterial subpopulation has been obtained.Automated analysis of fluorescent signals has also been utilisedto facilitate objective interpretation of results. The major problemwith FISH applications is the different penetration of probes inbacteria with various cell wall types, resulting in a possibleunderestimation of Gram-positive bacteria.

Polymerase Chain Reaction (PCR): Amplification of targetnucleic acids with PCR is an effortless method for detection oftarget DNA from various samples. In principle, the detection limitfor a PCR assay, based on usage of one or two oligonucleotideprimers specific for the target bacteria, is the presence of one copyof the amplified DNA region. Although achievement of such anextremely sensitive assay is unlikely, the PCR remains a verypowerful detection method, typically requiring a minimum of 10-20 target copies for successful amplification. The high sensitivityis also somewhat problematic, as a simple aerosol contaminationmay lead to false amplification.


PCR-based detection of rRNA genes has been used for directdetection of various bacteria present in faeces and pathogens fromclinical samples. Partial or full amplification of the 16S rDNAwith primers specific for a genus or a few closely related generahas also been reported. In addition, utilisation of 16S-23S intergenicsequences as targets of speciesspecific primers for different lacticacid bacteria has been described (Tilsala-Timisjärvi and Alatossava,1997). Good sensitivities have been reported for the PCR detectionassays of faecal bacteria. For example, PCR detection sensitivity offive ruminococcal species with species-specific primers from faecalsamples spiked with the target species did not markedly differfrom the observed detection limit of 4-100 cells from pure cultures.

Although a highly potent detection method, results obtainedby conventional endpoint PCR should not be considered to bedirectly quantitative. PCR may lead to differential amplification oftarget templates originally present in equal amounts. In addition,very low template concentrations may generate random fluctuationsin priming efficiency of population DNA samples with universalprimers, leading to a bias in the end-product concentration.Knowledge of the rRNA gene copy numbers and genome sizes ofbacteria in a mixed DNA sample has been observed to beinsufficient in predicting the final product ratio of a PCRamplification. Suzuki and Giovannoni (1996) also noticed that theaccumulation of end products during mixed-template PCR causedbiasing of the various end-product ratios towards a 1:1 situation,which was hypothesised to be caused by an increase in thehomologous template hybridisation, decreasing the efficiency ofprimer annealing and subsequent amplification.

Some improvements in the reliability of PCR quantificationhave been obtained by competitive PCR approaches. Hahn et al.(1995) created a quantitative PCR assay based on post-amplification differentiation of the internal standard and thesample DNA by selective restriction analysis and digoxigenin-based colorimetric detection. Quantitative detection ofMycobacterium tuberculosis PCR products was performed by enzyme-linked immunosorbent assay by comparison of hybridisationresults with two probes and two IS6110 elements derived fromeither an internal control or a modified template. With co-

amplification of an internal standard, detection of Clostridiumproteoclasticum was linear between 1 × 104 and 5 × 101 cells, witha detection limit of 50 fg or 25 cells, whereas detection of Oxalobacterformigenes from human faecal samples by competitive PCR wasshown to be linear over a range of six logarithmic units, with adetection limit of approximately 100 genomes.

Polymerase Chain Reaction – Enzyme-linked ImmunosorbentAssay (PCR-ELISA) : Polymerase chain reaction – enzyme-linkedimmunosorbent assay (PCR-ELISA) combines utilisation ofpolymerase chain reaction for efficient multiplication of the targetDNA and hybridisation with a detection probe to ensure thespecificity of the reaction. Figure 1 summarises the concept ofPCR-ELISA detection. The PCR-amplified products are labelledwith digoxigenin during or after the amplification reaction andhybridised with the specific biotinylated detection probe. The probeis immobilised in streptavidincoated microtitre plate wells, andhybridised DNA products are detected via digoxigenintargetedantibodies linked with an enzyme capable of producing acolorimetric or fluorimetric signal when brought together with asubstrate. Usage of PCR-ELISA hybridisation for detection of singlepathogens of clinical importance and spoilage bacteria of food hasbeen described. In comparison with filter hybridisation, PCR-ELISAprovides more easily standardised reaction conditions byutilisation of commercially supplied microtitre plates. Otheradvantages of PCR-ELISA compared with several other methodsare its relative simplicity and low costs.

Good sensitivities have been reported for PCR-ELISA assays,and the results can in some cases be interpreted quantitatively.The limit for detection of Bordetella pertussis was 100 targetorganisms, starting from appliance of different target DNA amountsto PCR reaction. Less than ten Epstein-Barr virus genome copiesadded to 750 ng of background DNA were required for a positivePCR-ELISA result. Similarly, a sensitivity of 5 cfu/ml blood wasobtained for detection of C. albicans and A. fumigatus cells,corresponding to the sensitivity of Southern blotting usingdigoxigenin-labelled oligonucleotides. Fletcher et al. (1998) coulddetect Aspergillus fumigatus quantitatively by a PCR-ELISA methodon a log-scale between 100 and 1 pg of target DNA. Detection of


Escherichia coli in oysters was reported to be quantitative in therange of 10-105 cfu/g. Detection of Campylobacter jejuni and C. coliwith PCR-ELISA was shown to be 10- to 100-fold more sensitivethan a gel-based PCR method using the same primers, the smallestamount of C. jejuni template DNA giving a positive signal in theassay being 1.5 fg. Some of the described PCR-ELISA applicationscan be considered to be quantitative competitive PCR approaches.

Real-time PCR: Real-time PCR is based on on-linemeasurement of the amplification reaction, enabling quantificationof the product during the logarithmic phase of PCR. The first andso far most commonly used real-time PCR approach is the 5´-nuclease (TaqMan) assay introduced by Holland et al. (1991). Theoriginal assay operates using a radioactively labelled probe thathybridises in the PCR template region to generate a specific,detectable signal from the amplification reaction. The detectionprobe becomes annealed to one of the DNA strands during theamplification and is cleaved by the Thermus aquaticus DNApolymerase 5´-exonuclease activity during the primer extensionstep. Lee et al. (1993) further extended the 5´-nuclease assay byutilisation of doubly labelled oligonucleotide probes for fluorescentmeasuring of formation of the specific PCR product during primerextension. Introduction of an automated detection method enabledreal-time monitoring of the PCR product formation during theexponential amplification phase. Since the advent of real-timePCR, several techniques have been introduced, including primerswith fluorescent dyes, molecular beacons, dual probes andintercalating dyes such as SYBR Green I. Real-time PCR is asuperior technique for quantification of nucleic acids. Whilecompetitive PCR has been demonstrated to be as reproducible andaccurate as real-time PCR, the latter has the benefit of an easiermethodology, reducing the need for sample DNA treatment.

Although real-time PCR is a relatively new technique, severaldetection or quantification assays targeting various bacteria havealready been described. Target bacteria include carious dentinebacteria, Desulfotomaculum from soil, faecal bifidobacterial species,Helicobacter hepaticus, Staphylococcus aureus, Borrelia burgdorferi sensulato, Campylobacter jejuni, Listeria monocytogenes, Rhodococcuscoprophilus and Mycoplasma genitalium. Some applications for

determination of larger bacterial groups have also been published.Typically, sensitivities of 1 to 100 target genomes per reaction andlinearity ranges of 4 to 8 logarithmic units are obtained in assaystargeted to specific bacteria. Real-time PCR is also utilised as arapid diagnostic tool for detection of pathogenic bacterial speciesor strains present in a sample. With pathogenic bacteria, the targetof choice for real-time PCR is generally a gene associated with thepathogenic traits.

Detection of PCR amplicons with specific probes is oftenfavoured over usage of intercalating dyes due to the former’s bettersensitivity and lack of detection of falsely primed products.However, SYBR Green I has become popular because of thepossibility to use this intercalating dye in virtually any assay.Real-time quantitative PCR with SYBR Green I has been reportedto be 10-fold less sensitive than a corresponding TaqMan assaydue to the formation of non-specific products in reactions startingwith small amounts of template DNA. On the other hand, SYBRGreen I has great potential in situations where a diverse targetpopulation is to be detected with PCR. A probe-based methodologyrequires a binding site for the probe in the vicinity of one of theprimers; however, such a conserved site is likely to be missingfrom a degenerate target DNA population. This should be takeninto account, especially when a bacterial population containingseveral hitherto unknown species is studied.

AIMS OF THE STUDY

The aims of this study were to develop molecular methods foranalysis and monitoring of faecal bacterial populations andputative probiotic strains as well as to characterise the technologicalproperties of two Lactobacillus brevis strains. The following goalswere set:

1. To create a genetic label by introducing silent mutations toadjacent amino acid codons of a genomic peptidase geneof a selected Lactobacillus strain and to confirm theunchanged phenotype of the strain with an availablepeptidase assay.

2. To develop oligonucleotide PCR primers or probes targetingthe 16S ribosomal DNA as species- or group-specificdetection tools.


3. To exploit a PCR-ELISA application with multipleoligonucleotide capture probes for analysis of artificialmixed DNA samples or faecal DNA preparations.

4. To test suitability of real-time PCR for quantification ofselected intestinal or probiotic bacteria in faecal samples.

5. To evaluate the applicability of two Lactobacillus brevisstrains as supplementary strains with potential probioticactions in dairy products.


Microbial Strains, Plasmids, Human Cell Lines and CultureConditions

The microbial strains, plasmids and human cell lines used inthis study have been described in detail in the respective originalpublications I-V. Briefly, Lactobacillus helveticus CNRZ32 was chosenas a model organism for demonstration of genetic labelling (I).Plasmids pUC19 and pSA3 were used to construct plasmidspKTH5052 and pKTH5053, respectively (I). The PCR-ELISAapplication with universal primers targeting 16S and 23S rDNAwas tested with a set of bacteria including 25 species, thatrepresented type strains of lactobacilli, bifidobacteria and otherbacteria reported as members of human intestinal microbiota, orlactobacilli species used in dairy fermentations (II). Extension ofthe PCR-ELISA method with bifidobacteria-targeted 16S rDNAamplification was tested with a set of ten bifidobacterial speciesbelonging to the human intestinal microbiota, suggested aspotentially harmful oral microbes, or used in dairy products (III).With real-time PCR, six bacterial strains, each representing abacterial species or group, present in the human intestinal tract ordairy products, were used (IV).

Dairy technological and probiotic properties of two lactobacillistrains, L. brevis GRL1 (ATCC 8287) and L. brevis GRL62 (ATCC14869T), were determined using dairy starters, positive or negativecontrol strains or target strains for testing the antagonisticproperties (V). The human intestinal cell lines Intestine-407 andCacO2 were used in adhesion studies (V). Culturing of microbesand human cell lines was carried out using media and conditions

as outlined in the Materials and methods sections of originalpublications (I-V).

Basic DNA Techniques

Rapid isolation of genomic DNA from pure cultured bacteriawas performed by cell mill disruption of bacterial samples in thepresence of glass beads, followed by phenol-chloroform extractionand ethanol precipitation (I, II, III, V). For rapid and effectivepurification and isolation of DNA from faecal material, a methoddescribed in Study I was used (I, III). A large-scale DNA isolationmethod described by Apajalahti et al. (1998) was used to producea sufficient amount of template DNA for real-time PCR (IV).

Wizard Minipreps were used for isolation of plasmid DNAfrom E. coli clones (I), whereas the Qiagen Plasmid Protocol Kitwas exploited for isolation of plasmid DNA from L. brevis (V).Restriction enzyme digestions and ligations were carried outaccording to the enzyme manufacturer’s recommendations. Thetransformations of L. helveticus cells were performed as describedby Bhowmik and Steele (1993). Dot blot hybridisation was executedusing standard methods (IV). DNA concentrations were measuredwith a Versafluor fluorometer (Bio-Rad). DNA sequencing wasperformed with an ABI310 DNA sequencer (Applied Biosystems)using BigDye Terminator chemistry (I).

Design of Oligonucleotide Primers and Probes

Oligonucleotide primers and probes were designed with thehelp of published sequence data available in sequence databanks(I, II, III, IV). Generally, the online Internet tools ClustalW andFasta provided by the European Bioinformatics Institute (EBI) wereused for identifying the primers and probes with desired specificitytowards intended target DNA. Oligonucleotides required forcreation of the silent mutation site (I) were planned by utilising thesequence of the L. helveticus CNRZ32 pepX gene (Accession numberU22900). While creating the mutations, major changes in the codonusage frequency were avoided to reduce the likelihood of changesin the expression level of the target gene. The Ribosomal DatabaseProject was utilised for designing the primers and probes used inStudy IV. In addition, signature oligonucleotides published


previously by others were utilised or modified when necessary(II, III, IV).

Polymerase Chain Reaction (I-V)

For end-point detection or production of DNA templates,polymerase chain reactions were carried out in reaction conditionsrecommended by the manufacturer of Dynazyme DNA polymerase.In Study I, the PCR conditions used in detection of wildtype andmutant strains were as follows: the reaction mixture consisted of50 mM Tris- HCl pH 9.0, 15 mM (NH4)2SO4, 0.1% Triton X-100,1.8 mM MgCl2, 360 µM of each deoxynucleotide triphosphate, 1µM of each primer and 0.02 U/µl Dynazyme EXTTM Polymerase,and 1 µl of template or water. Each sample was subjected to aprimary denaturation cycle at 95ºC for 2 minutes followed by 30cycles of denaturation at 95ºC for 30 seconds, annealing at 55ºCfor 30 seconds and elongation at 72ºC for 1.5 minutes. The reactionwas terminated at an elongation step for 5 minutes at 72ºC andfollowed by incubation at 4ºC. In Study IV, gradient PCR was usedto select optimal annealing temperatures for PCR primer pairs.Oligonucleotides were synthesised by commercial suppliers.

PCR-ELISA (II, III)

PCR-ELISA was carried out as described in Studies II and III.Briefly, the PCR products were labelled with the DIG-High PrimeKit according to the instructions of the manufacturer, andconcentrations of the PCR products were measured with theVersafluor fluorometer. For calculation of molecular quantities inthe labelled end products, the amount of DNA (ng/µl) was dividedby the length of the PCR product (base pair) multiplied by theaverage weight of a base pair (ng/base pair). The hybridisationswere performed in commercial streptavidin-coated microtiter plates(Labsystems, Finland). Results were analysed by measuringabsorbance at 450 nm.

Real-time PCR (IV)

Real-time PCR was tested with two different chemistries (5´-nuclease- and SYBR Green I assays) as described in Study IV.Quantitative PCR was performed with an iCycler iQ and iCyclerOptical System Interface software version 2.3. All PCR reactionswere performed in triplicate, in a volume of 25 µl, using 96-well

optical grade PCR plates and optical sealing tape. Optimalconcentrations for various reaction components were tested foreach primer set and chemistry with a dilution series of genomicDNA from the target test species. With SYBR Green I chemistry,effect of the polymerase type on PCR product formation was testedwith a standard polymerase, Dynazyme II, and hot-startpolymerases AmpliTaq Gold® DNA Polymerase, BlueTaq andFastStart Taq DNA Polymerase. The TaqMan assays wereperformed successfully with the Dynazyme II enzyme and thereforeswitching to the hot-start enzyme was not considered.

Peptidase Activity Assays (I)

Activity measurement of the PepX was determined from boththe wild-type and mutated L. helveticus strains according to themethod of El Soda and Desmazeaud (1982).

Evaluation of L. brevis Strains as Dairy Adjuncts

In vitro tolerance of the L. brevis strains to low pH, bile saltsand pancreatic fluid was examined as described in Study V.Antimicrobial properties towards selected food spoilage bacteriaand dairy lactic acid bacteria starters were evaluated by followingthe microbial growth in MRS broth supplemented with 10% filter-sterilized L. brevis GRL1 or GRL62 supernatant at 37ºC for twodays with an automated turbidometer Bioscreen C AnalysingSystem. Adherence to Caco-2 and Intestine-407 cell lines wasstudied as described in Study V. Resistance to selected antibioticswas tested with disc diffusion and microdilution methods, modifiedfrom instructions given by the antibiotic disc manufacturer andNCCLS. In vivo feeding trials with four healthy volunteers wereperformed to see whether the L. brevis strains could survive the GIpassage. Technological suitability of the strains for dairy processeswas also evaluated as described in Study V.

RESULTS

Strain-specific Genetic Labelling of Lactobacillus HelveticusCNRZ32 (I)

Silent mutation positions were introduced in adjacent codonsof the Lactobacillus helveticus CNRZ32 chromosomal pepX genewithout altering the amino acid sequence of the gene product,


creating a strain-specific tag which enabled direct nucleic-acidlevel identification of the changed strain. The mutation site wasintroduced into the pepX gene with a PCR primer containingdesigned mismatched nucleotides. The target gene was amplifiedin two separate reactions to obtain a conserved fragment and amutation site-containing fragment, which were ligated together toform a single mutation site-containing fragment. The mutant genewas first cloned with pUC19 to E. coli to create plasmid pKTH5052and further subcloned to pSA3, resulting in the plasmid pKTH5053,which was then used to transform L. helveticus CNRZ32.

A L. helveticus CNRZ32 transformant strain GRL1021containing the plasmid pKTH5053 was grown at a restrictivetemperature (45ºC) under antibiotic selection to find clones withthe thermosensitive plasmid integrated into the bacterialchromosome through homologous recombination between the wild-type and mutant forms of pepX. An erythromycin-resistant strain,observed to contain the integrated plasmid, was grown atpermissive conditions (37ºC) for approximately 100 generations toenable a second homologous recombination event between thetwo pepX genes, resulting in excision of the plasmid and one ofthe pepX genes. Samples were plated out and colonies that hadlost antibiotic resistance were screened with PCR for replacementof the wild-type pepX with the mutated pepX. A strain with themutated gene was found and designated as GRL1023. Sequencingwas used to confirm the correctness of the integration construct.The intact phenotypic properties of the GRL1023 were verified bymeasuring the activity of the target gene product PepX by utilisingits ability to break down L-glycyl-Lprolyl- p-nitroanilide, withproduction of colour. End-point PCR with PCR primers targetingthe mutated or original pepX sequence was successfully used todetect the wildtype and mutant strains added to faeces or milk.

Suitability of PCR-ELISA for Analysis of Mixed DNA Populations(II, III)

PCR-ELISA with Universal Primers Targeting the 16S and 23SGenes (II)

A PCR-ELISA method was tested with 16 oligonucleotideprobes to simultaneously detect under standardised conditions

selected intestinal bacteria, lactobacilli and bifidobacteria. For thispurpose, species- or group-specific oligonucleotide probes forlactobacilli and selected intestinal bacteria were designed oradapted from the literature. Specificity of the probes was testedwith 25 species. The 16S and 23S region of these test species wasfirst PCR-amplified with universal eubacterial primers, followedby digoxigenin-labelling of the PCR products and addition of 2 ×1010 labelled molecules to each hybridisation reaction. Forhybridisation, the oligonucleotide probes to be tested were boundto the microtiter plate wells via biotin-streptavidin linkage. Underthe hybridisation conditions used, the specificity level expectedwas obtained with most of the oligonucleotide probes chosen.Only the differentiation of closely related L. casei, L. paracasei andL. rhamnosus was partly unsatisfactory. Furthermore, in additionto its intended target DNA, the F. nucleatum probe, fuso16S, alsorecognized E. biforme. Sensitivity of the PCR-ELISA with the probestested varied between detection of 6.77 × 107 and 1.29 × 109 PCR-amplified molecules in a hybridisation reaction. The sensitivityobtained was somewhat lower than expected, which was at leastpartly due to compromises that had to be made in the hybridisationconditions when using several probes simultaneously.

PCR biasing was studied by combining genomic DNA fromdifferent test species into two mixed sample pools, each withseven DNA targets, prior to PCR amplification. PCR was alsoperformed using the pools in the same amplification.

The amplification products were hybridised with 14 differentprobes. Most of the probes tested performed well and could detecttheir target DNA from mixed samples of both seven and fourteenspecies, but the probes for B. adolescentis (ado440, b162), L. brevis(lab86), L. paracasei (par160), L. plantarum (pla448) and L. curvatus(cur150) failed to recognize their target DNAs. Failure in thedetection of B. adolescentis, L. paracasei and L. plantarum in particularcould have been due to either bias during multi-template PCR,where some templates are amplified more efficiently as a result ofdifferences in the specificity of eubacterial primers, or bias bychance. Changing of the total template concentrations tested in thePCR amplification was not observed to affect the hybridisationresults.


PCR-ELISA with bifidobacterial primers targeting the 16S genes(III) The PCR-ELISA method was extended for detection of themost common Bifidobacterium species in humans and applied to afeeding trial including administration of Bifidobacterium lactis Bb-12 and galacto-oligosaccharide -containing syrup as probiotic andprebiotic preparations, respectively.

Oligonucleotide probes based on 16S rDNA sequences weredesigned and tested for specificity and sensitivity with ninedifferent bifidobacterial species, followed by analysis of faecalsamples. Faecal bifidobacteria were monitored for fluctuationsduring and after the feeding trial. Although the bifidobacterialpopulations present in faeces were generally consistent betweensamples taken from different time points, PCR-ELISA resultssuggested that Bifidobacterium longum was partly replaced by B.lactis Bb-12 during the probiotic feeding. This proposition couldnot, however, be verified with statistical methods. Generally, thebifidobacterial species composition of individual subjects wasconsistent and comprised 3-4 bifidobacterial species or groups.The predominant species were Bifidobacterium longum and B.adolescentis.

Comparison of Real-time PCR and Dot Blot Hybridisation forQuantification of the 16S Ribosomal DNA of Target Bacteria(IV)

Real-time PCR was tested with two chemistries: SYBR GreenI and TaqMan probes. Six detection assays were tested for theirsensitivity in detecting target bacterial DNA from pure and mixedsamples with both chemistries. The same target DNAs were alsoquantified with dot blot hybridisation probes. Real-time PCR wasshown to have sensitivity superior to dot blot hybridisation. Thelinear range of amplification of pure cultured target DNA variedbetween 0.1-1 pg and 1-10 ng of specific target genome, whichcorresponds to approximately 20-400 to 2×105–4×106 genomes,depending on the genome size of the target species. Furthermore,a 1 pg addition of target DNA to 20 ng of mixed DNA samplescould be detected reliably with both real-time PCR chemistries. Areconstruction assay with addition of 109, 108, 107, 106 or 105target bacterial cells to faecal samples confirmed that real-time

PCR could detect addition of 106 cells to one gram of faeces,provided that the target bacterium was not included in the originalfaecal sample. Addition of 109 cells of Bifidobacterium longum tofaeces resulted in no difference compared with control sample,which had no added bacterial cells, due to the pre-existence of B.longum in the original sample.

The two chemistries used in real-time PCR were equallysensitive; however, successful usage of the intercalating dye SYBRGreen I required applying of a hot start polymerase to preventformation of primer dimers, whereas with 5´-nuclease chemistry,a similar sensitivity level was reached with a conventional DNApolymerase. Furthermore, the 5´-nuclease chemistry was faster toperform because of its more straightforward incubation protocol,one run taking approximately 80 min.

Suitability of L. brevis Strains as Dairy Supplements (V)

Two Lactobacillus brevis strains, ATCC 8287 and ATCC 14869T,were evaluated for their applicability as putative probiotics indairy products. The strains tolerated well low pH, bile acids andpancreatic fluid under in vitro conditions. They also expressedgood in vitro adherence to human CacO2 and Intestine-407 cells.Adherence of the L. brevis strains was especially strong to thesmall intestinal cell line, Intestine 407. In antimicrobial activityassays, strain ATCC 8287 showed inhibitory properties towardsselected potentially harmful micro-organisms, particularly againstBacillus cereus. Antimicrobial resistance tests revealed that both L.brevis strains were resistant to vancomycin as well as to severalother antibiotics, which is, however, typical for the genusLactobacillus. The L. brevis strains were unable to acidify milk toyoghurt but were suitable as supplement strains in yoghurts. Thiswas demonstrated by producing a set of yoghurt products andanalysing their rheological and sensory properties during a coldstorage period of 28 days. Despite its human origin, L. brevisATCC 14869T did not survive the gastrointestinal tract, whereasL. brevis ATCC 8287 was detected in faecal samples taken duringand immediately after ingestion of the strain. In conclusion, L.brevis ATCC 8287 was considered a promising candidate for useas a probiotic adjunct in dairy products.


DISCUSSION

Molecular biology tools are needed for better understanding ofthe composition and functions of the intestinal microbiota. In thisstudy, different levels of detection were developed for lactic acidbacteria and faecal microbiota.

‘Probiotic studies require careful documentation of theproposed effects. For better monitoring of a potential strain, strain-specific discrimination could be beneficial in, for example, feedingstudies. Furthermore, specific recognition of industrial strains withcommercial value would help in monitoring the usage of thesestrains. Therefore, Lactobacillus helveticus CNRZ32 was chosen asa model organism for demonstration of genetic labelling. Althoughthe strain in question does not fulfil the criteria set for a probiotic,selection of this organism was considered justified for severalreasons. The strain is a starter strain used in the dairy industry,its proteolytic system has been studied extensively and methodsfor its transformation and gene disruption have been developed.Various ‘food-grade’ markers for lactic acid bacteria have beensuggested, often with introduction of new phenotypic properties.However, creation of a label without introduction of anyphenotypic changes into the target strain would be more easilyaccepted by consumers and legislators. Hertel et al. (1992) labelleda Lactococcus lactis strain by introducing silent mutations into aplasmid-encoded gene. By contrast, to ensure the stability of thecreated tag, we used a chromosomal gene for labelling of L.helveticus CNRZ32. As demonstrated in Study I, a silent mutationsite is effortlessly detected from different matrices, such as faecesor milk, by PCR. Alternatively, hybridisation probes could beutilised for detection.

Direct detection of target nucleic acids present in populationsamples removes the need for cultivation steps. Here, PCR-ELISA,dot blot hybridisation and real-time PCR were compared for theirsuitability for analysis of population samples. Similar to PCRELISAmethod described in this paper, DGGE- and TGGE-basedapplications have been found to detect approximately 1-10%subpopulations with PCR-amplified 16S rDNA fragments whenstudying faecal samples, suggesting their use for analysis ofpredominant microbes of population samples. By using DGGE

with groupspecific 16S rDNA primers, however, characterisationof diversity of smaller subpopulations has been reached. Similarly,the bifidobacteria-specific PCR-ELISA could successfully be usedfor species-level analysis of faecal bifidobacteria; previous DGGEanalysis results of the same samples by Satokari et al. (2001b)generally support results obtained with PCRELISA. One drawbackof PCR-ELISA identification has, however, been uncovered. Whenresults were compared with PCR-ELISA analysis, it soon becameevident that ado440, intended to be specific for B. adolescentis, alsorecognized B. ruminantium observed in the faeces of some volunteers.New sequence alignments confirmed that the ado440 probe hadno mismatches with B. ruminantium 16S rDNA. While theadvantage of DGGE is the possibility for further sequence analysisand identification of emerging new amplicons in DGGE profiles,PCR-ELISA seems to be slightly more sensitive to changes in therelative amounts of target species.

Strain-level studies have confirmed the complexity, uniquenessand stability of bifidobacterial populations in the GI-tract. PCR-ELISA results were in accord with earlier reports. The observedspecies distribution correlated well with previous Europeanstudies, and species distribution usually remained unchangedwithin the samples of each subject. However, reduction of intensitiesof B. longum - specific probes was observed frequently in thegroups that ingested the probiotic B. lactis Bb-12, although thiscould not be verified with statistical analysis, mainly due to thesmall size of the test groups. This suggests that B. lactis Bb-12 mayhave partly replaced B. longum during the ingestion period.

Hybridisation allows direct semi-quantitative monitoring ofpopulation samples and thus avoids the problems caused by PCRbias. Dot blot hybridisation with rRNAtargeted oligonucleotideprobes is a well-established method for studying faecal microbes.Ribosomal RNA is abundantly present in the bacterial cells, thusincreasing the sensitivity of the dot blot hybridisation to areasonable level. However, radioactively labelled probes aregenerally needed for successful detection of nucleic acid targetsfrom population samples. Here, dot blot hybridisation wasperformed with rDNA-targeted oligonucleotide probes to quantifythe same starting material as with the real-time PCR applications


tested. Dot blot hybridisation could, at its best, detect a 3% DNAsubpopulation from mixed DNA samples. A similar sensitivitylevel has been reported by Muttray and Mohn (2000) formeasurement of rDNA:rRNA ratios. Thus, the result was asexpected, being approximately 10-fold less sensitive than thatobtained with rRNAtargeted oligonucleotides. Real-time PCR wastested as an alternative for a sensitive detection of target bacteriain faeces. The sensitivity level obtained for the real-time PCRassays was 200-400 target bacteria in pure cultures or mixed DNAsamples, depending on the genome size of a target bacterium.Results obtained from artificial DNA mixtures were confirmed bya reconstruction assay; addition of 106 bacterial cells to faecalsamples containing approximately 1011 bacteria could be detectedwith real-time PCR in a quantitative manner. The acquired level,providing a means for quantifying a 0.01% subpopulation in aDNA sample, was considered sufficient for studying thegastrointestinal tract ecology.

Although nucleic acid based detection methods offer a meansto avoid bias caused by cultivation or phenotypic testing of bacterialisolates, these techniques are not totally devoid of problems. Themain issues for detection of nucleic acids in mixed samples areisolating DNA from different sources with constant efficiency,minimising the occurrence of false-positive reactions due tocontamination or poor planning of the primers or hybridisationprobes used, and avoiding negative reactions due to inhibitingsubstances. Also, any oligonucleotides used as detection probes orprimers should be considered specific only in relation to the currentsequence data. Several methods for faecal sample treatment forisolation of DNA have been described. Nevertheless, this stepremains somewhat problematic because of the presence of PCR-inhibiting substances in the faeces and the overall complexity offaecal microbial populations. Although over 90% recovery of thetotal bacterial DNA present in faeces has been claimed, non-selective isolation of bacterial DNA from population samples isdifficult to achieve or demonstrate. Commercial kits have recentlybecome available for isolation of faecal DNA. However, McOristet al. (2002) observed differences in the relative efficacy of extractionof bacterial DNA from faeces with four commercial kits.

Furthermore, the relative sensitivities could not be extrapolatedfrom DNA extractions performed directly from pure cultures.

Evaluation of possible probiotic properties of a bacterial strainrequires intensive and carefully planned testing of the strain bothin vivo and in vitro. Although L. brevis is not typically used as aprobiotic, Kishi et al. (1996) have shown that oral administrationof live L. brevis ssp coagulans strain significantly stimulated thehost immunity system by increasing IFN-á production in humansubjects. This result cannot, however, be directly extrapolated tothe L. brevis strains studied here. Adherence of lactic acid bacteriato various components of human intestine has been studied withthe human colon adenocarcinoma CacO2 cell line, Intestine-407cells derived from human embryonic jejunum and ileum, andintestinal mucus. Both CacO2 and Intestine-407 cells were used inthe adhesion assays for evaluation of the adhesion ability of L.brevis. These cell lines offer complementary models for adhesionstudies. When compared with the probiotic L. rhamnosus GG,adherence of the L. brevis strains was especially strong to the smallintestinal cell line, Intestine 407, suggesting that the small intestinecould be affected with the strains studied. L. brevis GRL62 has alsobeen shown to have intermediate adhesion properties to humanintestinal mucus.

The two L. brevis strains studied were antagonistic towardssome of the potentially harmful micro-organisms, while they didnot significantly inhibit the growth of yoghurt-starter bacteria. Ina previous study of Koga et al. (1998), L. brevis GRL1 (ATCC 8287),included in a panel of 41 Lactobacillus strains belonging to L.gasseri, L. reuteri, L. casei, L. helveticus, L. brevis, L. fermentum and L.plantarum, weakly inhibited two Vibrio cholerae strains of the eightstrains tested. In the present study, L. brevis GRL1 was shown tostrongly inhibit B. cereus and to some extent also the growth of S.aureus and other harmful micro-organisms chosen for the assay. Invitro adhesion assays and testing for tolerance of low pH, bileacids and pancreatic fluids have often been considered to be goodindicators of the survival of a bacterial strain through the GI tract.In this study, both L. brevis strains performed well in the in vitrotests, and survival through the stomach could therefore besuggested for both strains, especially when simultaneous ingestion


of the bacteria with food, resulting in a higher pH in the stomach,is taken into account. However, L. brevis GRL62 failed to surviveunder in vivo conditions. This was surprising as the strain hasoriginally been isolated from human faeces. In contrast, L. brevisGRL1, originally isolated from fermented olives, could tolerate thegastrointestinal conditions and was found to persist in the humangut.

Although the strains studied were not suitable for fermentingyoghurt, supplementation of yoghurt with either of these strainshad no negative effects on yoghurt taste, appearance, orpreservation. Therefore, L. brevis GRL1, which was able to survivein the GI tract, could be considered for use as a supplementarystrain in yoghurt or other dairy products. To date, severalcommercial probiotics are already available and many others arebeing studied, but no single strain is likely to possess all of thebeneficial properties suggested for probiotics. L. brevis GRL1, shownto be antagonistic towards Bacillus cereus (this study) and Vibriocholerae, could be a valuable addition to the probiotics field. Thesurface (S)-layer of this strain is well-characterized and potentialapplications for the S-layer have been suggested. Recently, the S-layer has been shown to mediate the ability of the L. brevis GRL1strain to adhere to human intestinal, urinary bladder andendothelial cells. Good adhesion properties could assist in thecompetitive exclusion of potentially harmful microbes by L. brevisGRL1, and usage of this strain as an effective mucosal vaccinationtool because of its binding properties may be warranted. Morestudies are, however, required to confirm these hypotheses.

CONCLUSIONS

Creation of a strain-specific nucleic acid tag by labelling of achromosomal target gene with a silent mutation and utilisation ofthe label for strain-specific detection of the target strain wasdemonstrated using Lactobacillus helveticus CNRZ32. Although thestrain in question is not considered to be a probiotic, the methoddescribed here could be used for probiotics. However, this requiresthe pre-existence or development of suitable transformation toolsas well as sequence data of a suitable target gene. From theviewpoint of current legislation and consumer acceptance, the

silent labelling method should be the least unacceptable methodfor making genetic alterations to a bacterial strain used in foodprocessing, especially as the phenotypic properties of the labelledstrain were demonstrated to remain unchanged, thus attesting tothe safety of the method. PCR-ELISA with universal primers wasshown to be suitable for detection of predominant bacteria presentin a mixed population sample. Specificity of the PCR step provedto be critical for the sensitivity of the method; application ofbifidobacterialspecific 16S primers enabled species- or group-specific detection of bifidobacteria present in human faeces. Theoverall sensitivity of PCR-ELISA matched DGGE approaches, withsimilar PCR primer specificities. Furthermore, being technically arather simple method, PCR-ELISA is easily transferred betweendifferent laboratories, and could, in principle, be used in a similarmanner as DGGE for analysis of large sample numbers.

Real-time PCR was established to be a superior method fordetection of single bacterial species or larger groups from mixedDNA samples or faeces. Compared with dot blot hybridisationwith radioactive probes, real-time PCR provided an improvedsensitivity for detection, a means to avoid usage of radioactivelabels, increased speed and volume of sample analysis, an easiermethodology and better reproducibility. Detection of target bacteriaadded to faeces prior to the DNA isolation and purification stepswas demonstrated. In future, multiplexing with different labelsand improvements in hardware and software to allowsimultaneous monitoring of a large number of PCR reactions arelikely to make real-time PCR an even more potent tool forpopulation analyses.

L. brevis GRL1 could be considered to be a potential probioticstrain. Although several probiotics are already on the market, newstrains with beneficial properties are continually being sought.The strain in question has a well-characterized S-layer, and offerspromise as a potential mucosal vaccination strain.

Introductory Food Microbiology Glossary258 259

9GLOSSARY

Acid dyes: Dyes that are anionic or have negatively chargedgroups such as carboxyls. Acid fast: Bacteria like the mycobacteriathat cannot be easily decolorized with acid alcohol after beingstained with dyes such as basic fuchsin.

Acid-fast staining: Staining procedure that differentiatesbetween bacteria based on their ability to retain a dye when washedwith an acid alcohol solution.

Acidophile: Microorganism that has its growth optimumbetween about pH 0 and 5.5.

Actinobacteria: Group of gram-positive bacteria containingthe actinomycetes and their high G 1 C relatives.

Actinomycete: Aerobic, gram-positive bacterium that formsbranching filaments (hyphae) and asexual spores.

Aerobe: The descriptive name given to a microorganism thatcan grow in conditions where oxygen is present. Such organismsare capable of growing in normal air, equivalent to 20% oxygen,or in aerated liquids containing dissolved oxygen. Many aerobesare equally able to grow in the absence of oxygen. These aretermed facultative anaerobes.

Allele: One of two or more alternative nucleotide sequences ata single gene locus which occurs on either of two homologouschromosomes in a diploid organism.

Anaerobe: A microorganism that is capable of growing in thecomplete absence of oxygen. Some of these organisms may also beable to grow in oxygenated conditions (facultative aerobes), whereas

others cannot tolerate oxygen and are killed when exposed to air.Such organisms are termed obligate anaerobes.

Antibody: An inducible immunoglobulin protein produced byB lymphocytes of the immune system, in humans and other higheranimals, which recognizes and binds to a specific antigen moleculeof a foreign substance introduced into the organism. Whenantibodies bind to corresponding antigens they set in motion aprocess to eliminate the antigens.

Antigen: Any foreign substance, such as virus, bacterium, orprotein, which after introduction into an organism (humans andhigher animals), elicits an immune response by stimulating theproduction of specific antibodies; or any large molecule whichbinds specifically to an antibody.

Antimicrobial: A chemical agent that kills microorganisms orinhibits their growth.

Apoptosis: Programmed cell death, the body’s normal methodof ending the lifecycle of cells through the cellular self-destruction.When either heritable or somatic cell mutations cause malfunctionsto occur in the apoptotic pathway, uncontrolled cell growth mayproceed unchecked and cancer may result.

Bacterial growth: Can exhibit at least four different phases:lag phase, growth phase, stationary phase and death phase.

Bacterial strain: Population of bacterial cells all descendedfrom a single pure isolate.

Base pair (bp): Two complementary nitrogenous bases in aDNA molecule, such as the nucleotide coupling of adenine withthymine (A:T) and guanine with cytosine (G:C); also, a unit ofmeasurement for DNA sequences.

Biofilm: Adherent layer of bacteria and/or othermicroorganisms on a solid surface bound together in a bacterially-derived polysaccharide matrix that is protective for the organisms;generally occuring at a liquid/solid interface and often developinginto a complex ecological community (e.g., dental plaque boundtother by dextrans).

Bleaching: The loss of fluorescence usually due tophotochemical reactions.


cDNA (complementary or copy DNA): DNA copiessynthesized from a messenger RNA template using the enzymereverse transcriptase; the single-stranded copy is often used as aprobe to identify complementary sequences in DNA fragments orgenes of interest.

Chromosome: A single DNA molecule that is the self-replicating genetic structure within the cell which carries the linearnucleotide sequence of genes. In humans (or eukaryotes), the DNAis supercoiled, compacted, and complexed with accessory proteins,and organized into a number of such structures. Normal humancells contain 46 chromosomes (except the germ cells, egg andsperm): 22 homologous pairs of autosomes and the sex-determiningX and Y chromosomes (XX for females and XY for males).Prokaryotes carry their entire genome on one circular chromosomeof DNA.

Coccus: (singular): Spherical-shaped cell; cocci (plural).

Coliform bacteria (coliforms): Any fermentative (specificallylactose-fermenting) Gram-negative anaerobic enteric bacilli (E. coli-like).

Commensal: Organisms existing in or on an animal or humanwithout causing disease.

Conjugation (or bio-conjugation): The chemical joining of abiomolecule to another.

Death phase: The death phase occurs when cells are beinginactivated or killed because conditions no longer support growthor survival. Some environmental factors such as temperature cancause acute inactivation. Others may cause mild inactivation aswith growth in the presence of organic acids.

DNA (deoxyribonucleic acid): The nucleic acid moleculeconsisting of deoxyribonucleotide building blocks that encodegenetic information. The genome of most organisms is containedin a double-stranded, double-helical form held together withchemical bonds between each strand of complementary nucleotidebase pairs.

DNA probe: A single-stranded piece of DNA that bindsspecifically to a complementary DNA sequence; the probe is labeled

(e.g., with a fluorescent or radioactive tag) in order to detect itsincorporation through hybridization with DNA in a sample.

Dot blot: A method for detecting proteins by thespecific binding of an antibody or binding molecule to asample spot on nitrocellulose paper. The bound sample isvisualized using an enzymatic or fluorimetric reporter conjugatedto the probe.

Doubling time: The time taken for a population to increase innumber by a factor of two.

Enteric (entero-): Relating to the intestine.

Enterotoxin: Proteins produced by bacteria that are eitheringested as pre-formed toxins or are produced by a pathogen thathas colonised the gastro-intestinal tract. Usually the toxin hasspecific targets and either disrupts cell function or kills the cell.

Eukaryote: (meaning “true nucleus”) An organism whichpossesses a nucleus with a double layer of membrane and othermembrane-bound organelles; includes such unicellular ormulticellular members as all members of the protist, fungi, plant,and animal kingdoms.

Exons: The segment of a gene present in mature mRNAtranscripts that specify the amino acid sequence of a polypeptideduring translation; exons of a gene are linked together by mRNAsplicing.

Exotoxin: Potent toxic substance formed and releasedextracellularly by species of certain bacteria.

Exponential phase: The period in which the cells of a definedbacterial population are growing and dividing continuously.

Extracellular: Produced, then excreted outside the organism.

Facultative: Ability to adapt and live under various conditions.

Facultative anaerobe: Anaerobe that can survive with orwithout oxygen.

Family: Taxonomic level below order and above genus.

Fastidious: Complex nutritional or cultural requirements,making isolation and culture of a fastidious organism moredifficult.


Fermentation: Enzymatic breakdown (catabolism) ofcarbohydrates generally in the absence of oxygen.

Fimbriae: Short, hair-like projections or appendages(organelles) on the outer surface of certain bacteria composed ofprotein subunits (pilin) extending outward from the surface thatact as a virulence factor by promoting adherence; formerly knownas pili; fimbria (singular).

Flagellum: whip-like bacterial locomotory (provide motility)organelles anchored in the cell membranes that are composed ofhelically-coiled protein subunits (flagellin); flagella (plural).

FISH (Fluorescence In Situ Hybridization): A technique thatemploys fluorescent molecular tags to detect probes hybridized tochromosomes or chromatin; useful for genetic mapping anddetecting chromosomal abnormalities.

Flow cytometer: Analytical instrument for flow cytometry.

Flow cytometry: Automated analysis of cells or subcellularcomponents by detection of the fluorescence or light-scatter ofsample fractions passing in narrow-stream droplets through alaser beam.

Fluorophores: Molecules that produce a fluorescentemission when irradiated with light at a suitable excitationwavelength.

Fomite: Inanimate object capable of transmitting infectiousorganisms to a host, e.g., soiled clothes, tissues and handkerchiefs,food processing equipment, dishrags, etc.

Gene amplification: The presence of multiple copies of a geneor segment of DNA; a mechanism by which proto-oncogenes areactivated in malignant cells. A tumor cell amplifies, or copies,DNA segments as a result of cell signals or the effects ofenvironmental insults.

Gene expression: The process by which the encodedinformation of the genome is converted into cellular components.The DNA-coding sequences of expressed genes include those thatare transcribed into mRNA and then translated into proteins, andRNA that is transcribed from DNA, yet not translated into protein(i.e., transfer and ribosomal RNAs).

Gene mapping: A linear map determining the relative positionof genes along a chromosome or plasmid. Distances are establishedby linkage analysis and measured in linkage units.

Gene: A nucleotide sequence of DNA that codes for aprotein, or functional or structural RNA molecule; a locuson a chromosome. The element that determines a trait in anorganism.

Genetic mutation: An alteration in the nucleotide sequence ofa DNA molecule; often from one allelic form of a gene to anotherallele alternative.

Genome: The total amount of genetic material in a cell; ineukaryotes the haploid set of chromosomes of an organism. Thechromosome set is species-specific for the number genes and linkagegroups carried in genomic DNA.

Genomics: The study of genes and their biochemical functionin an organism.

Genotype: The genetic constitution of an organism; or, areference to an individual’s particular allele pair at a specific genelocus in the genome.

Genotyping: Analysis of genotype.

Genus: Taxonomic level below Family and above Species.

Gram-negative or Gram-positive: The classification given tobacteria according to their staining properties as defined by theGram stain procedure.

Growth curve: A graph displaying the behaviour of a bacterialpopulation over time.

Growth phase: During the growth phase, cells growexponentially and at a constant rate. The maximum slope of thecurve is the specific growth rate of the organism. Cell growth isdependent upon the current environment (nutrients, temperature,pH, etc.), but is not dependent upon the previous physiologicalstate. In the field of predictive microbiology, growth rate iscommonly expressed as the change in cell number per time interval.

Growth rate: This is an expression of population increase innumbers expressed as log10 cfu/hour.


Haploid: A cell or individual with a genetic complementcontaining one copy of each nuclear chromosome. (Diploid refersto the condition when a eukaryotic cell possesses two sets ofchromosomes.)

Humectant: A solute that binds free water in a food, reducingthe amount of water available to the microorganisms.

Hybridize (or hybridization): The process where the hydrogenbonding of complementary DNA and/or RNA sequences forms aduplex molecule.

Immunoassays: A technique that detects and measures aspecific antigen or biological substance by employing antibodies(e.g., dot blot, western blot, and ELISA).

in situ hybridization (ISH): Use of a nucleic acid probe todetect and identify specific complementary sequences of DNA inchromosomes or RNA in bacteria, eukaryotic cells, and tissue.

in vitro: (“in glass”) Refers to the recreation of biologicalprocesses in an artificial environment such as a test tube.

in vivo: (“in living”) Refers to biological processes within aliving organism or cell.

Inoculum: A medium containing microorganisms to beintroduced into fresh media or food source in an experiment.

Intron: A nucleotide sequence intervening between exons(coding regions) that is excised from a gene transcript duringRNA processing.

Kinetics: The properties of chemical agents or enzymes in theefficiency and speed of their action upon a chemical reaction.

Klebsiella: Gram-negative rods occur in human feces andclinical specimens, soil, water, grain, fruits, and vegetables. Somespecies are opportunistic pathogens. Kluyvera: Gram-negative rodsoccur in food, soil, sewage, and human clinical specimens. Theyare infrequently opportunistic pathogens. Kurthia: Gram-positiverods are widely distributed in the environment, and are commonin animal feces and meat products.

Lag phase: During the lag phase, cells increase in size but notin number because they are adapting to a new environment, and,synthesis and repair are taking place. The length of the lag phase

depends on the current environment as well as the previousphysiological state of the cells. Cells that are from a very differentenvironment or are damaged from their previous physiologicalstate may require more time to adjust. In some foods a lag phasedoes not exist which results in cells that are ready for immediategrowth.

Lag time: The initial period in a bacterial population lifewhen cells are adjusting to a new environment before commencinggrowth.

Legionella: Fastidious gram-negative rod is isolated fromsurface water, mud, and thermally polluted lakes and streams.There is no known soil or animal source. It is pathogenic forhumans, causing pneumonia (Legionnaires’ disease) or a mild,febrile disease (Pontiac fever).

Ligand: The molecule which binds to a protein molecule (e.g.,receptor). As a ligand binds through the interaction of many weak,noncovalent bonds formed to the binding site of a protein, thetight binding of a ligand depends upon a precise fit to the surface-exposed amino acid residues on the protein.

Listeria: Gram-positive rod widely distributed in theenvironment. Some species are pathogenic for humans and animals(e.g. L. monocytogenes).

Locus: (plural, loci) The specific site of a gene on a geneticmap or chromosome.

Marker (genetic marker): Any genetically derived phenotypicdifference used in the analysis of inheritance patterns or todifferentiate between types of cells. An observable site on achromosome that is heritable and can be either a genetically-expressed region or noncoding segment of DNA (intron).

Maximum population density: Point at which the maximumnumber of bacterial cells can exist in an environment.

Meiosis: Process that allows one diploid celll to divide in aspecial way to generate haploid cells in eukaryotes.

Mesophile: Microorganism able to grow well between 20°Cand 45°C, having an optima of 30°C to 40°C. Many can sustaingrowth at below 10°C albeit very slow growth.


Metaphase: Stage in mitosis or meiosis during whichthe chromosomes are aligned along the equatorial plane of thecell.

Methylobacterium: Mostly isolated from water and leaf surfacemicroflora, and are facultative methylotrophs, that is capable ofgrowing on one-carbon compounds such as formate, formaldehyde,and methanol as the sole source of carbon and energy, as well ason a wide range of multicarbon substrates.

Microaerophilllic environment: Environment with reducedoxygen concentrations, often below 5%. Carbon dioxide levelsmay approach 10%.

Microbacterium: Gram-positive rod is found in dairy products,sewage, and insects.

Microbial load: Total number of living microorganisms in agiven volume or mass of microbiological media or food.

Micrococcus: Gram-positive cocci occur primarily onmammalian skin and in soil, but are commonly isolated from foodproducts and the air.

Microorganism: A living organism too small to be seen withthe naked eye.

Mitosis: Process by which a cell separates its duplicatedgenome into two identical halves. It is generally followedimmediately by cytokinesis which divides the cytoplasm and cellmembrane. This results in two identical daughter cells with aroughly equal distribution of organelles and other cellularcomponents.

Moraxella: Gram-negative rod is parasitic on the mucousmembranes of humans and other warm-blooded animals.

MRNA (messenger RNA): The RNA molecule, transcribedfrom the DNA of a gene, which serves as a template and encodesthe amino acid sequence of a protein.

Multiplexing: Method by which many parameters are testedand processed simultaneously.

Native microflora: Microorganisms that are normally foundwithin a food source (often referred to as spoilage organisms).

Northern blot: Technique used to separate and transfer mRNAfrom a gel to a filter in order to identify and locate mRNA sequencesthat are complementary to and hybridize with a labeled DNAprobe.

Novosphingobium: Gram-negative rods were originallyincluded with Sphingomonas (see Sphingomonas).

Nucleic acid: Molecule composed of nucleotide subunits. SeeDNA and RNA.

Nucleotide: Basic building block of nucleic acids that is amonomeric molecule of DNA or RNA composed of: a pentosesugar (with 5-carbons such as deoxyribose in DNA, ribose inRNA), an organic nitrogenous base, and a phosophate group.DNA consists of the four bases: adenine (A), guanine (G), cytosine(C), and thymine (T); likewise for RNA, except for the substitutionof uracil (U) for T.

Oligonucleotide: (“oligo” means few) A short length of DNAnucleotides, often used as primers for DNA synthesis or probes forarrays and ISH/FISH; usually referred to as “oligo(s).”

Opportunistic: Microorganism that will only cause disease ina patient with a poor or somehow weakened immune system.

Organelle: Microscopic bodies in the cytoplasm of cells thathave distinctive functions (e.g., nucleus, mitochondria,endoplasmic reticulum, etc.).

Pasteurisation: Mild heat treatment process given to foods.The process is designed to eradicate potential vegetative pathogens(not bacterial spores) and reduce other microorganism numbers inan effort to decrease the rate of spoilage.

Pantoea: Gram-negative rods are isolated from plant surfaces,seeds, soil, and water, as well as from animals and human clinicalspecimens. They are opportunistic human pathogens.

Pathogen: Microorganism associated with disease in man.

pH: Measure of the acidity or alkalinity of a solution, definedas the - log10 of the hydrogen ion concentration.

Phenotype: Observable manifestation of a genetic trait,resulting from a specific genotype and its interaction with theenvironment.


Physical map: Map indicating physical locations on a DNAmolecule such as restriction enzyme recognition sites, RFLPs, andgenes; measured in base pairs (bp).

Predictive microbiology: Area of food microbiology that usesmathematical models to define growth kinetics of microorganismsin food.

Primary model: Model that describes changes in microbialnumbers in response to time.

Primer: Short segment of DNA or RNA that anneals toa single strand of DNA in order to initiate template-directed synthesis and extend a new DNA strand by theenzymatic action of DNA polymerase to produce a duplex-stranded molecule.

Probe: Single-stranded DNA or RNA molecule of specific basesequence, either radioactively or fluorescently labeled, that is usedto identify the complementary nucleotide sequence byhybridization to the DNA fragment or gene of interest.

Protein translocation: Spatial movement of protein within acell (e.g., from the cytoplasm to nucleus, or into organelles).

Protein: High-molecular weight biological molecule composedof a polymer of amino acids linked via peptide bonds; may consistof more than one polypeptide molecule that is folded into complexshapes such as helices or sheet-like structures. Proteins are encodedby the specific sequence of DNA nucleotides in a gene and giverise to the structure, function, and regulation of cells, tissues, andorgans within the body. Protein classes include enzymes,antibodies, receptors, hormones, and growth factors.

Proteome: Entire protein make-up of a particular organism.

Proteomics: Study of proteins and their biochemical functionin an organism.

Providencia: Gram-negative rods are isolated from humanclinical specimens and from penguins.

Pseudomonas: Gram-negative rod is widely distributed innature. Some species are pathogenic for humans, animals, or plants(e.g. P. aeruginosa).

Psychrobacter: Gram-negative rod is associated with fish,processed meat and poultry products. Some strains have beenisolated from pathological specimens from humans and animals.

Psychrotroph: Microorganism able to grow well between 0°Cand 7°C, having an optima of 20°C to 30°C.

Rahnella: Gram-negative rods occur in freshwater. They areoccasionally isolated from human clinical specimens, but are notconsidered clinically significant.

Ralstonia: see Pseudomonas.

Raoultella: see Klebsiella.

Rathayibacter: Some of these species are phytopathogens ofterrestial plants. Their main habitats are their respective planthosts.

Receptor: Surface-exposed membrane protein on a cell whichbinds to a specific ligand molecule with high affinity, in order totransmit an extracellular signal and trigger intracellularbiochemical events within the target cell.

Reporter: Gene which codes for an easily measured proteinproduct and is fused downstream of the gene of interest in orderto assess the activity in the region upstream of the reporter gene.Colorimetric and fluorimetric reporters also can be conjugated toprobes to monitor biological events.

Rhodococcus: Aerobic, Gram-positive actinomycetes. Thesewidely-occurring organisms are of considerable environmentaland biotechnological importance due to their broad metabolicdiversity and array of unique enzymatic capabilities, plus theircapacity to degrade hydro-carbons. They are able to survive for along time in soil. They are the most efficient in oil degradationand, relatively speaking, the most abundant in soils and marineenvironments.

Rhizobium: Group of small, rod-shaped, gram-negativebacteria, which are able to produce nodules on the roots, or onsome cases the stems, of leguminous plants.

RNA (ribonucleic acid): DNA-like organic molecule thatconsists of nucleotide subunits—such as adenine, guanine,cytosine, and uracil—which contain ribose sugars linked through


phosphodiester bonds. Different types of RNA function in theprocess of gene expression.

Saprophyte: Microorganism that normally grows on deadmaterial.

Secondary model: Model that predicts changes in primarymodel parameters based on environmental conditions.

Serratia: Gram-negative rods occur in human clinicalspecimens, soil, water, plant surfaces, and other environmentalsites, digestive tracts of rodents, and insects. Some species areopportunistic pathogens.

SNP (Single Nucleotide Polymorphism): Variations in thesequence of DNA among individuals that are present in humanswith a frequency of about once in every 1000 bases, and useful inassessing the patterns of inheritance in genetic linkage studies.

Somatic cell mutation: Mutation in a cell that is acquiredduring the lifetime of an organism and which cannot be geneticallyinherited by offspring.

Somatic cell: Any cell in the body except the germ-line cells(sperm or egg cells).

Southern blot: Procedure which transfers elecrophoreticallyseparated DNA fragments on an agarose gel to nitrocellulosefilters for detection by hybridization with a labeled probecomplementary to the sequence of interest; the position on thefilter of the probe, when exposed to x-ray film, appears as a bandon an autoradiogram.

Sphingobacterium: Gram-negative rod found in soil, on plants,foodstuffs, and in water sources. Sphingobium: Gram negativerods were originally included with Sphingomonas (seeSphingomonas).

Sphingomonas: Relatively new genus derived fromPseudomonas paucimobilis. These organisms are widelydistributed, including having been found in water. OnlySphingomonas paucimobilis is considered clinically significant,and has been isolated from a variety of clinical specimens

Spoilage organisms: Microorganisms naturally found withina food source that cause food spoilage.

Staphylococcus: This gram-positive coccus is mainlyassociated with the skin and mucous membranes of warm-bloodedvertebrates, but they are often isolated from food products, dustand water. Some species are opportunistic pathogens of humansand animals, or produce extracellular toxins.

Stationary phase: The stationary phase occurs at the maximumpopulation density, the point at which the maximum number ofbacterial cells can exist in an environment. This typically representsthe carrying capacity of the environment. However, environmentalfactors such as pH, preservatives, antimicrobials, native microfloraand atmospheric composition as well as depletion of growth-limiting nutrients can affect the maximum population density.

Stenotrophomonas: see Pseudomonas.

Sub-lethal injury: This is cellular damage which results indisruption of metabolic processes which, under ideal conditions,is repairable. Such damage must be repaired before normal growthcan recommence.

Tertiary model: Computer software routines that turn theprimary and secondary models into “user-friendly” programs.

Tsukamurella: Gram-positive rods are isolated from soil,human sputum, and parts of bed bugs. Some strains can bepathogenic.

Toxins: Compounds produced by a microorganism that arepoisonous to other organisms.

Vegetative cell: The vegetative cell state is the form in whichan organism is able to grow and divide continuously, givenfavourable conditions. Unlike endospores, vegetative cells arerelatively poor at surviving environmental stresses such as hightemperature and drying.

Water activity (aW): Water activity is a measure of the amountof free unassociated water molecules in a system. It runs on a scaleof 0 to 1.0, where pure water equals 1.0. This parameter can alsobe expressed as a percentage referred to as the relative humidity.It is influenced by dissolved solutes and insoluble food componentswhich act to bind water, thus reducing the available free waterand hence aW value.

Introductory Food Microbiology272 Bibliography 273

Weeksella: Gram-negative rod is not known in the generalenvironment. It is apparently a parasite, saprophyte, or commensalof the internal surfaces of humans or other warm-blooded animals.

Western blot: A technique which transfers proteinselectrophoretically separated in a polyacrylamide gel to anitrocellulose membrane and uses specific antibodies to bind,locate, and visualize the protein of interest.

Yersinia: Gram-negative rods occur in a broad spectrum ofhabitats, including humans, animals, soil, water, dairy products,and other foods. Some species are pathogenic for humans andanimals; others are opportunistic pathogens, yet others arenonpathogenic.

Xanthomonas: Most Gram-negative rods are plant pathogens,or occur in association with plants. X.maltophilia is the onlyexception, being an opportunistic pathogen of humans.

Zoonotic: Microorganism normally found in or on animals.

BIBLIOGRAPHY

Anderson, Kenneth: International Dictionary of Food & Nutrition. New York: Wiley, 1993.

Ausubel, F.M.: Current Protocols in Molecular Biology, New York,John Wiley and Sons, 1989.

Bernauer, Thomas: Genes, Trade, and Regulation: The Seeds of Conflictin Food Biotechnology, Princeton, Princeton University. 2003.

Bowes, Anna De Planter: Bowes and Church’s Food Values of PortionsCommonly Used. Philadelphia: J.B. Lippincott, 1994.

Broach, J.R.: The Molecular Biology of the Yeast, Cold Spring Harbor,Cold Spring Harbor Laboratory, 1981.

Casado, Matt A: Food and Beverage Service Manual. New York:Wiley, 1994.

Chmelynski, Carol Ann Caprione: Opportunities in Food ServiceCareers. Lincolnwood: VGM Career Horizons, 2000.

Coultate, T.P.: Food: the Chemistry of its Components, BurlingtonHouse, London, 1984.

David Sadava: Plants, Genes and Crop Biotechnology, Sudbury MA,Jones and Barlett Publishers, 2003.

David, Zachary: Essentials of Food Safety and Sanitation. UpperSaddle River: Prentice Hall, 2000.

Fellows, P. and Hidellage, V.: Making Safe Food, IT Publications,London, 1992.

Fellows, P.J.: Food Processing Technology, Woodhead Publishing,Cambridge, 1993.

Ferrando, R.: Traditional and Non-Traditional Foods, FAOPublications, Rome, Italy, 1981.

Gerald, W.: Lodging and Food Service Industry. East Lansing:Educational Institute of the American Hotel & MotelAssociation, 1995.

Introductory Food Microbiology274 Index 275

INDEX

AAccounts, 4.Acid Dyes, 258.Acidophile, 258.Actinobacteria, 258.Administration, 234, 250, 255.Anaerobe, 258, 261.Apoptosis, 84, 259.Application, 17, 26, 28, 69, 176,

229, 244, 257.Approach, 30, 43, 97, 107, 112,

121, 129, 158, 236, 238,242, 266.

Association, 20, 69, 70, 76, 79,175, 178, 196, 272.

Atmospheres, 13.

BBacteria, 1, 8, 11, 12, 13, 14,

26, 30, 39, 46, 49, 50,53, 55, 56, 57, 58, 63,66, 67, 68, 70, 72, 73,76, 81, 84, 85, 86, 87,88, 89, 90, 91, 92, 93,95, 101, 105, 107, 109,115, 116, 119, 120, 123,124, 125, 127, 128, 130,141, 146, 148, 160, 162,165, 167, 168, 170, 171,173, 176, 177, 178, 180,181, 182, 183, 184, 185,186, 187, 188, 200, 201,

202, 209, 211, 215, 216,219, 230, 231, 232, 233,234, 235, 236, 237, 238,239, 240, 241, 242, 243,244, 245, 246, 248, 249,250, 251, 253, 255, 256,257, 258, 259, 261, 262,263, 264, 265, 271.

Bioscreen, 31, 32, 48, 60, 71,84, 93, 95, 96, 97, 100,108, 111, 113, 115, 120,123, 124, 126, 132, 140,147, 148, 149, 152, 159,161, 162, 163, 164, 167,175, 203, 220, 221, 222,247.

Biotechnology, 61.

CCampylobacter, 5, 242.Capacity, 22, 27, 57, 58, 69,

92, 106, 111, 113, 139,269, 271.

Carbohydrates, 8, 10, 11, 19,22, 193, 199, 233, 262.

Carbon Dioxide, 87, 180, 183,184, 266.

Carnobacteria, 8, 9, 10, 11, 12,13, 14, 15, 18, 19, 20,21, 22, 23, 24, 26, 27,166, 167.

Carnobacterium, 8, 9, 10, 11,12, 13, 14, 15, 16, 19,

Hays, Judi Radice: Restaurant & Food Graphics. Glen Cove: PBCInternational, 1994.

Hidellage, V.: Making Safe Food, IT Publications, London, 1992.Hobbs, B. and Roberts, D.: Food Poisoning and Food Hygiene, Edward

Arnold Ltd, London, 1987.Ihekoronye, A.I., and Ngoddy, P.O., : Integrated Food Science and

Technology for the Tropics, Macmillan Press Ltd., London, 1985.Jay, J.M.: Modern Food Microbiology, D van Nostrand, New York,

1978.Low, J.: Fruit and Vegetables, IT Publications, London, 1984.Metting, Jr., F.B.: Soil Microbial Ecology: Applications in Agricultural

and Environmental Management, Marcel Dekker, New York, 1992.Ngoddy, P.O., : Integrated Food Science and Technology for the Tropics,

Macmillan Press Ltd., London, 1985.Roberts, D.: Food Poisoning and Food Hygiene, Edward Arnold Ltd,

London, 1987.Sackler, Warren: Foodservice Cost Control Using Microsoft Excel for

Windows. New York: Wiley, 1996.Shetty, Kalidas: Food Biotechnology, New York, Dekker/CRC Press,

2005.Sims-Bell, Barbara: Career Opportunities in the Food and Beverage

Industry. New York: Facts on File, 1994.Slater, R.J.: Experiments in Molecular Biology, Clifton, Humana Press,

1986.Sprenger, R.A.: The Food Hygiene Handbook, Highfield Publications,

Doncaster, 1996.Susan, R.: Biotechnology: An Introduction, Belmont, Thomson/Brooks/

Cole, 2005.Thomas R.: Bountiful Harvest: Technology, Food Safety and the

Environment, Washington DC, Cato Institute, 2003.Walden, Richard: Genetic Transformation in Plants, England, Open

University Press, 1988.Wallace, L.: Introduction to Professional Foodservice. New York:

Wiley, 1996.Zaccarelli, Herman E.: Food Service Management by Checklist: A

Handbook of Control Techniques. New York: Wiley, 1991.

Introductory Food Microbiology Index276 277

20, 21, 22, 23, 24, 25,26, 166, 169, 170.

Chromosome, 16, 49, 62, 74,248, 260, 263, 264, 265.

Commission, 156, 178.Community, 13, 18, 23, 236,

237, 238, 259.Comparison, 11, 37, 39, 42, 51,

87, 90, 100, 103, 125,133, 136, 138, 143, 145,146, 155, 159, 160, 164,174, 178, 179, 188, 226,229, 238, 240, 241, 250.

Conditions, 12, 13, 18, 19, 22,29, 30, 33, 34, 37, 39,41, 43, 45, 46, 47, 50,52, 53, 54, 57, 58, 59,61, 63, 78, 85, 87, 88,89, 93, 94, 95, 97, 98,99, 100, 101, 102, 103,104, 105, 106, 107, 108,111, 112, 113, 114, 115,116, 117, 118, 119, 120,121, 122, 123, 124, 125,126, 127, 130, 131, 132,134, 135, 137, 138, 140,141, 142, 143, 144, 145,146, 147, 152, 153, 154,155, 156, 157, 158, 162,163, 164, 165, 176, 180,182, 185, 186, 196, 200,201, 203, 207, 209, 210,215, 219, 222, 225, 226,227, 241, 244, 246, 248,249, 251, 256, 258, 260,261, 270, 271.

Conjugation, 260.Conservation, 68, 134.Constitution, 144, 263.Construction, 33, 49, 73, 74,

129, 130, 133, 135, 136,147, 148, 222, 236.

Cultures, 8, 9, 17, 22, 24, 31,47, 48, 49, 52, 53, 60,68, 69, 70, 71, 72, 76,77, 78, 79, 80, 81, 82,85, 95, 96, 97, 98, 100,113, 120, 122, 123, 126,130, 133, 139, 148, 156,167, 181, 191, 196, 197,198, 200, 206, 209, 212,213, 216, 220, 240, 254,255.

DDepartment, 30, 219.Development, 26, 33, 34, 35,

36, 43, 80, 82, 129, 175,200, 219, 224, 232, 256.

Diagnostics, 34, 36.Distribution, 3, 9, 23, 26, 34,

35, 36, 93, 94, 96, 99,101, 102, 103, 105, 111,120, 129, 156, 178, 224,227, 253, 266.

Diversity, 17, 27, 176, 232, 236,237, 253, 269.

DNA, 25, 48, 49, 50, 60, 61,62, 64, 72, 73, 74, 131,168, 169, 171, 173, 174,183, 229, 233, 235, 236,237, 238, 239, 240, 241,242, 243, 244, 245, 246,247, 248, 249, 250, 251,254, 255, 257, 259, 260,261, 262, 263, 264, 265,266, 267, 268, 269, 270.

EEcology, 4, 26, 165, 254.Economic Growth, 2.Ecosystem, 231.Electron Microscopy, 45, 50, 53,

197.

Emission, 50, 179, 262.Energy, 10, 43, 202, 266.Enteric Viruses, 2.Enterotoxin, 4, 6, 28, 34, 39,

217, 220, 222, 225, 261.Environment, 4, 8, 9, 10, 11,

14, 26, 27, 46, 55, 57,58, 59, 83, 84, 94, 101,107, 111, 112, 113, 119,120, 121, 122, 123, 135,136, 138, 139, 142, 143,144, 145, 146, 147, 156,165, 181, 184, 218, 233,263, 264, 265, 266, 267,271, 272.

Escherichia Coli, 6, 10, 47, 59,71, 101, 128, 130, 169,184, 232, 242.

Eukaryote, 261.Evaluation, 131, 135, 145, 178,

185, 229, 230, 247, 255.Evolution, 2, 110, 112, 129, 135,

136, 165, 180, 205.Experiments, 31, 32, 39, 45, 47,

48, 59, 60, 61, 63, 66,76, 87, 95, 96, 98, 105,107, 108, 110, 111, 114,120, 121, 122, 123, 126,127, 128, 130, 132, 133,134, 136, 139, 140, 141,145, 148, 149, 161, 164,165, 167, 174, 175, 193,194, 205, 210, 211, 217,227.

FFarmers, 178.Fermentation, 1, 9, 69, 78, 81,

82, 106, 180, 182, 184,214, 233, 262.

Fermentor, 108, 111, 122, 123,124, 126.

Fluorophores, 262.Food Contamination, 26.Food Microbiology, 1, 94, 136,

268.Food Safety, 1, 5, 21, 27, 43,

44, 107, 112, 128, 141,172, 222, 223, 233.

Foodborne Pathogens, 2, 137,139.

GGenetic Mutation, 263.Genomics, 25, 27, 263.

IIdentification, 9, 31, 45, 57, 60,

62, 130, 133, 176, 177,193, 202, 206, 208, 215,229, 236, 237, 239, 248,253.

Industry, 9, 29, 44, 46, 57,128, 158, 172, 173, 187,214, 252.

Infections, 4, 5, 23, 234.Information, 6, 10, 12, 19, 25,

26, 33, 46, 47, 57, 58,61, 68, 94, 109, 154, 172,187, 193, 194, 226, 229,236, 260, 262.

Injection, 24, 86, 87, 89.Inoculum, 86, 87, 93, 94, 95,

96, 97, 98, 99, 100, 101,102, 103, 104, 105, 127,131, 140, 146, 148, 156,157, 158, 159, 161, 264.

Institute, 25, 37, 151, 152, 245.Institutions, 23.Instruments, 206.Integration, 49, 74, 235, 248.Interests, 112.Isolation, 5, 9, 10, 49, 165, 175,

Introductory Food Microbiology Index278 279

192, 203, 204, 234, 245,254, 257, 261.

LLaboratory, 19, 21, 43, 70, 86,

92, 136, 149, 151, 167.Legislation, 235, 256.Listeria Monocytogenes, 5, 8, 17,

45, 57, 83, 93, 95, 106,107, 112, 113, 122, 136,137, 146, 147, 149, 157,166, 170, 183, 242.

Literature, 29, 38, 174, 227,228, 231, 249.

Low Virulence, 6, 83, 84, 85,88, 90, 91, 92.

MMajority, 26, 39, 147, 155, 157,

158, 178, 198, 231.Management, 180.Manufacture, 1, 68, 69.Measurement, 31, 32, 33, 44,

76, 133, 149, 152, 162,195, 206, 221, 225, 238,242, 247, 254, 259.

Mechanism, 4, 6, 43, 46, 59,99, 171, 172, 184, 190,199, 215, 262.

Media, 12, 19, 21, 30, 31, 32,34, 47, 52, 53, 59, 60,61, 63, 70, 84, 90, 93,96, 99, 100, 108, 109,113, 114, 115, 120, 121,123, 124, 125, 126, 130,135, 139, 141, 148, 157,159, 160, 169, 177, 192,201, 214, 220, 221, 222,225, 244, 264, 266.

Methodology, 26, 132, 149, 242,243, 257.

Methylobacterium, 266.Microbacterium, 266.Microbial Ecology, 26.Microbiology, 1, 48, 59, 71, 94,

106, 112, 116, 119, 122,129, 132, 136, 146, 147,263, 268.

Microorganisms, 1, 2, 8, 17,29, 30, 40, 52, 130, 133,137, 183, 189, 194, 195,197, 198, 200, 211, 212,232, 272.

Microplates, 48, 59, 60, 71, 124,148.

Molecules, 41, 104, 172, 182,196, 228, 237, 249, 262,271.

Mycotoxins, 3.

NNatural Environment, 9, 10, 11,

14.Network, 61, 106, 107, 109,

197.

OObservations, 55, 89, 91, 116,

126, 134, 137, 178.Opportunistic, 22, 24, 264, 267,

270, 271, 272.Organelle, 267.Organisms, 6, 10, 11, 14, 17,

18, 19, 22, 27, 63, 69,83, 101, 106, 107, 112,119, 121, 128, 129, 136,172, 180, 182, 184, 186,201, 203, 209, 215, 220,229, 230, 231, 234, 235,236, 241, 251, 255, 258,259, 260, 262, 266, 269,270, 271.

Osmotolerance, 45, 54, 55, 58,59.

PPathogens, 1, 2, 3, 4, 5, 6, 8,

18, 24, 27, 30, 137, 139,166, 172, 219, 231, 233,240, 241, 264, 267, 270,271, 272.

Performance, 116.Plants, 11, 13, 14, 26, 27, 68,

148, 165, 173, 178, 268,269, 270, 272.

Population, 6, 32, 69, 77, 78,80, 88, 93, 94, 95, 97,100, 101, 102, 103, 104,105, 116, 117, 122, 125,127, 134, 141, 146, 147,151, 153, 154, 155, 157,160, 161, 165, 221, 227,229, 231, 232, 236, 237,238, 240, 243, 252, 253,254, 257, 259, 261, 263,265, 271.

Predictions, 37, 39, 43, 106,107, 109, 111, 112, 113,116, 118, 119, 121, 225.

Preparation, 1, 30, 31, 72, 73,113, 123, 131, 177, 190,199, 220.

Prevention, 17, 218.Production, 3, 4, 8, 14, 16, 17,

19, 20, 21, 23, 28, 34,70, 80, 81, 82, 111, 113,121, 158, 159, 160, 167,169, 172, 173, 179, 180,183, 184, 187, 188, 189,190, 199, 208, 209, 210,214, 215, 217, 218, 222,230, 234, 246, 248, 255,259.

Products, 1, 3, 5, 7, 8, 9, 10,11, 12, 13, 14, 18, 19,20, 21, 22, 27, 28, 39,41, 43, 44, 55, 60, 69,73, 82, 123, 136, 137,138, 139, 145, 146, 147,148, 150, 159, 160, 164,173, 174, 180, 181, 182,184, 185, 187, 189, 196,198, 199, 207, 209, 214,216, 219, 228, 230, 232,233, 240, 241, 243, 244,246, 249, 251, 256, 264,266, 269, 271, 272.

Project, 25, 132, 245.Protection, 11, 18, 68, 210.Proteins, 7, 16, 25, 45, 46, 47,

51, 54, 55, 56, 62, 63,66, 67, 68, 83, 166, 172,182, 192, 197, 204, 208,216, 260, 261, 262, 268,272.

Protozoan Parasites, 2.

RRelations, 44, 216.Research, 24, 27, 151, 178, 200,

238.Resolution, 237.RNA, 45, 47, 49, 50, 52, 237,

253, 260, 262, 263, 264,266, 267, 268, 269, 270.

SSalmonella, 5, 102, 184, 185.Species, 3, 4, 5, 6, 7, 8, 9,

10, 11, 12, 13, 14, 19,20, 21, 22, 23, 24, 25,26, 58, 68, 70, 71, 128,136, 138, 147, 149, 159,160, 163, 164, 165, 166,

Introductory Food Microbiology280 Introductory Food Microbiology 281

CONTENTS

Preface

1. Introduction 1

2. Role of Ctc from Listeria Monocytogenes inOsmotolerance 45

3. Identification of Listeria Monocytogenes Genes 57

4. Experimental Validation of Low Virulence inField Strains 83

5. The Effect of Inoculum Size on the Lag Phase 93

6. Modelling the Growth of ListeriaMonocytogenes in Dynamic Conditions 106

7. Epidemiological Typing of Bacillus spp Isolatedfrom Food 173

8. Modelling the Growth Boundary ofStaphylococcus Aureus 217

9. Glossary 258

Bibliography 273

Index 275

173, 174, 177, 180, 182,185, 186, 187, 189, 211,215, 229, 230, 231, 232,233, 234, 235, 236, 237,239, 240, 242, 243, 244,247, 249, 250, 253, 257,261, 263, 264, 265, 268,269, 270, 271, 272.

Spoilage, 1, 2, 8, 13, 14, 17,18, 20, 21, 22, 26, 27,136, 159, 164, 170, 171,177, 178, 185, 186, 189,202, 205, 220, 246, 254,272.

Stress Conditions, 45, 47, 53, 54,59, 61, 63, 94, 127.

Symptoms, 4, 5, 8, 234.

TTechnology, 30, 73, 74, 220.Temperature, 11, 12, 17, 29,

31, 39, 46, 49, 52, 57,

72, 81, 87, 94, 102, 105,107, 108, 112, 113, 117,119, 120, 121, 122, 127,128, 129, 130, 131, 132,133, 134, 135, 136, 137,138, 139, 140, 142, 143,144, 145, 146, 147, 149,152, 153, 154, 159, 160,162, 163, 164, 165, 166,177, 178, 196, 202, 204,205, 206, 209, 212, 213,216, 218, 219, 220, 226,248, 260, 263, 271.

Treatment, 97, 101, 125, 179,198, 212, 215, 216, 230,242, 254, 267.

VValidation, 39, 64, 83, 110, 111,

129, 219.Velocity, 175.Viruses, 1, 2.

INTRODUCTORYFOOD MICROBIOLOGY

Documents

142 P -Introductory-Food-Microbiology-F