INFCTON AND IMMuNrTY, Oct. 1970, p. 376-386Copyright © 1970 American Society for Microbiology

Vol. 2, No. 4Printed in U.S.A.

Efficiency of Various Intestinal Bacteria in AssumingNormal Functions of Enteric Flora After

Association with Germ-Free MiceSALAM A. SYED, GERALD D. ABRAMS, AND ROLF FRETER

Departments of Microbiology and Pathology, The University of Michigan, Ann Arbor, Michigan 48104

Received for publication 20 March 1970

Strictly anaerobic and facultatively anaerobic bacteria were isolated from theintestinal tract of normal mice. Germ-free mice were associated with mixtures ofvarying complexity of pure cultures of these bacteria. The development of nor-mal features in these animals was then determined on the basis of the followingcriteria: (i) size of the cecum, (ii) size of the Escherichia coli population in thececum, (iii) histology of the intestinal tract, and (iv) development of a mucosa-associated flora in stomach and large intestine. Germ-free mice contaminatedwith cecal contents from conventional mice were used as controls to establish nor-mal values for these parameters. Some strictly anaerobic bacteria could be im-planted into germ-free mice only after prior implantation of an E. coli strain. E.coli was found in large numbers in stomach and cecum of mice monocontaminatedwith this organism. Use of restraining devices indicated that the E. coli populationin the stomach was maintained by coprophagy and did not contribute signifi-cantly to the size of the cecal population. A mixture of 50 strictly anaerobic bac-teria plus 80 facultative anaerobes rendered recipient animals normal with respectto the criteria tested. Other, less complex bacterial mixtures reduced the cecal sizeand the intestinal E. coli population to levels intermediate between those foundin normal and germ-free mice. With all bacterial mixtures tested, the intestinal E.coli population decreased, if at all, within a period of about 10 days after intro-duction of other bacteria, and remained stable thereafter. This suggests that theintestinal E. coli population is controlled by a mechanism which reduces popula-tion size without affecting the growth rate.

It is well known that the normal enteric floraof the mammalian body contributes significantlyto the host's resistance to infectious diseases.Furthermore, changes in the composition of thisflora are often associated with (and possibly maybe the cause of) diseases such as nonspecificdiarrheas, malabsorption syndromes, etc. (3). Itwould therefore be highly desirable to have somemeans of controlling the composition of theenteric flora so as to establish a microbial popu-lation of maximal benefit to the host. Unfor-tunately, this is not possible at the present time,because the ecology of mammalian intestinalflora remains one of the poorly understood fieldsin microbiology. As pointed out in earlier publi-cations (2, 15), one of the primary reasons for thisstate of affairs is the inadequacy of conventionalculture methods for the isolation of the strictlyanaerobic bacteria, i.e., of those microorganismswhich form the predominant flora of the largeintestine.

As a necessary preliminary to a study of theecological control mechanisms which determinethe composition of normal enteric flora, wedescribed (2) an anaerobic glove-box culturetechnique which allows the recovery of a signifi-cant fraction (20 to 50% of total microscopiccounts) of the cecal flora of the mouse. Althoughthis constitutes a considerable improvement overconventional anaerobic culture methods (such asthe anaerobic jar technique), one cannot dis-regard the fact that 50 to 80% of the total bac-terial count is still not cultivated by our method.This may mean that a number of bacterialspecies cannot be cultivated in the glove box.Alternatively, a substantial fraction of the cecalbacteria may be nonviable, which, in turn, wouldimply that the culture method is sufficient for cul-tivating at least those bacteria which predominatein the intestinal population.

In the studies described below, the abovequestion was investigated by testing whether

376

on June 8, 2020 by guesthttp://iai.asm

.org/D

ownloaded from

ENTERIC FLORA OF GNOTOBIOTIC MICE

bacteria isolated from the normal mouse intestinewould form a "normal" flora when implantedinto germ-free mice. If this were possible, onecould be confident that the bacterial culturesinvolved include at least those species which aremost important in the control of the ecologicalbalance of the intestinal tract. It would then bepossible to proceed with a study of the mecha-nisms of bacterial interactions in the intestine.Similar attempts to "normalize" germ-free micehave been made in the past by others (e.g., 4, 7,9). To our knowledge, all of these attempts havebeen unsuccessful in that a completely "normal"flora could be implanted into germ-free or strep-tomycin-treated mice only by feeding fecalmaterial. Introduction of mixtures of bacteriaisolated in pure culture from normal mice onlypartially restored the animals to their normalstate.Much of the significance of the present work

depends on the criteria used for describing anexperimental animal as "normal." It is obviouslydesirable to test for a variety of parameters whichmay be assumed to be mediated by a variety ofdifferent mechanisms. On the other hand, it isnecessary to keep the required technical efforts ata manageable level. In view of these considera-tions, the following parameters were selected: (i)the reduction of the population in the cecum ofan Escherichia coli strain to normal levels [thestrain selected, E. coli C25, had been shown inearlier studies to suppress invading pathogenssuch as Shigella and Vibrio (5)]; (ii) reductionin size of the typically enlarged cecum of thegerm-free mouse to normal levels; (iii) develop-ment of normal histological features in the gas-trointestinal tract; and (iv) the development of amucosa-associated layer of bacteria in the stom-ach and large intestine, as described by Savageet al. (13), Plaut et al. (11), and others. In accord-ance with common practice in germ-free research(7), ex-germ-free animals, contaminated withcecal homogenate from normal mice, were usedas normal standards for the above parameters.The experiments reported below define the

necessary conditions for the implantation ofstrictly anaerobic bacteria into germ-free miceand describe a collection of pure cultures of bac-teria which, when implanted into germ-free mice,will render the animals normal with respect to allcriteria tested. The data presented also allowsome preliminary conclusions as to the mecha-nisms involved in the control of the E. colipopulation by the other elements of the normalenteric flora.

MATERIALS AND METHODSMice. Bacteria of the normal flora were isolated

from mice of strain BALB/wm, maintained in this

department by William Murphy. Normal cecumhomogenates were also prepared from organs of thesemice. The germ-free mice used were Charles Riverstrain CD-1. These were maintained in Trexler-typeflexible vinyl isolators, sterilized with peracetic acid.Autoclaved powdered Lobund diet L-356 was fed inall experiments, except that shown in Table 2, inwhich the mice received pelleted sterile diet obtainedfrom Charles River Mouse Farms. (This accounts forthe relatively high E. coli population observed in thatexperiment.) Plastic cages containing autoclaved woodshavings were used inside the germ-free isolators.

Microorganisms. Bacteria were isolated from nor-mal mice in an anaerobic glove box. Enriched Trypti-case Soy Agar with palladium black overlay, asdescribed in an earlier publication (2), was used forthe isolation of strict anaerobes. Facultative anaerobeswere isolated on the same medium. In addition, blood-agar and L B S Medium (BBL) were used for isolatingfacultative anaerobes and microaerophilic lacto-bacilli. Plates for the isolation of the latter organismswere incubated inside the anaerobic glove box as wellas aerobically in a conventional incubator.The taxonomy of anaerobic bacteria is still "in a

state of flux" (14). As pointed out recently by Moore(10), most of the bacterial population of the intestineof man and animals "has not been thoroughly charac-terized." We can certainly confirm this statement. Allstrains of anaerobic bacteria used in the present studyhave been characterized according to most of thecriteria published recently by Smith and Holdeman(14), but many strains do not fit into the taxonomicschemes proposed by these authors, by Prevot andFredette (12), or by the Virginia Polytechnic InstituteAnaerobe Laboratory (16). Since the taxonomic prob-lem of anaerobes is pursued vigorously by otherworkers, no attempt has been made at this time toassign species names to the bacteria used. The anaero-bic bacteria were of different morphological types,ranging from short to very long filamentous rods,with rounded or tapered ends. Some had vacuoles(spores?) and were resistant to heating at 70 C for 10min. The Gram reaction was generally negative, butmany strains showed gram-positive inclusions ofvarious sizes and for this reason would perhaps beconsidered gram-positive by other workers. All mor-phological types of bacteria which can be seen inGram-stained smears from ceca of normal mice wererepresented among the cultures of anaerobes used inthis study. The facultative anaerobes used in thisstudy included enterobacteria, such as Aerobacter,Proteus, and Pseudomonias species, Staphylococcusalbus, Streptococcus faecalis, Lactobacillus, and bac-teria resembling the Actinomyces-Bifidobacteriumgroup (14).The strain of E. coli C25 was the same one used

in earlier studies in this laboratory (5).Inoculation of animals. Suspensions of bacteria

were prepared inside the anaerobic chamber. Thesewere transferred into ampoules which were then stop-pered, removed from the glove box, and sealed with-out contamination by atmospheric oxygen. After in-troduction into the germ-free isolators, one ampoulewas opened just before inoculation of each animal.The mice were injected with 0.5 ml of suspension via

VOL. 2, 1970 377

on June 8, 2020 by guesthttp://iai.asm

.org/D

ownloaded from

SYED, ABRAMS, AND FRETER

the rectum, through a 0.060-inch (outer diameter)polyethylene tube (Adams Intramedic tubing sizePE100). Preliminary tests had shown that this pro-cedure results in the introduction of some materialdirectly into the cecum. Rectal inoculation was usedin all experiments described in the present paperbecause this procedure had been found superior to oralinoculation by R. W. Schaedler (personal communi-cation). Later experiments, not reported here, suggestthat this precaution may not have been necessarywith the strict anaerobes used in the present study.These microorganisms may be established in germ-free mice via the drinking water. Exposure to theoxygen in the drinking water results in a dramaticdrop in the bacterial (plate) count, but a sufficientnumber of bacteria appear to remain viable longenough to initiate growth in the mouse intestine.

Quantitative bacterial cultures. Organs to be studiedwere homogenized in 100 ml of Trypticase Soy Brothin a Waring Blendor. Serial dilutions of the ho-mogenates were plated on the surface of Deoxycho-late Agar (BBL) containing 1 mg of streptomycin perml. This medium allows the growth of E. coli C25while suppressing all other bacteria used in this study.

Microscopic bacterial counts. Counts were per-formed on organ homogenates prepared as above forquantitative cultures. A Petroff-Hausser chamber wasused as described earlier (15).

Restrining devices to prevent coprophagy. Tubes ofacrylic plastic 1 inch in inner diameter by 3.25 incheslong (2.5 by 8.25 cm), which contained the animalscomfortably but were sufficiently narrow to keep themice from turning around, were used. An opening[0.75 by 0.75 inch (1.9 by 1.9 cm)] in the anal regionallowed fecal material to be dropped. The front ofthe tube led into a food hopper, which also permittedaccess to a drinking tube leading from a water bottleoverhead.

Histopathological techniques. Specimens of thegastrointestinal tract were fixed in 10% Formalin,embedded in paraffin, sectioned at 5 pm, and routinelystained with hematoxylin and eosin for histopatho-logical examination. Segments of small and largeintestine were processed without opening or disturb-ing the included contents, whereas portions of stom-ach and cecum were fixed after opening, but withadherent contents disturbed as little as possible.Some material was also frozen in methylcellulosesolution (13), but, as there appeared to be no particu-lar advantage to the method, the observations re-corded below were derived entirely from paraffin-embedded material.

RESULTSTable 1 shows the results obtained in attempts

to implant a gram-negative anaerobic bacteriuminto germ-free mice, into ex-germ-free micewhich had been exposed to E. coli C25 plus astrain of Lactobacillus 3 days previously, intoex-germ-free mice which had been exposed tolactobacilli (same strain used above) 3 days pre-viously, and into germ-free mice, the inoculumof anaerobes being mixed with the same Lacto-

TABLE 1. Implantation of a strictly anaerobic,gram-negative, rod-shaped bacterium into the

intestinal tract (cecum) of germ-freeand gnotobiotic mice

Implan-Inoculum Recipient tation

aerobes

Anaerobes Germ-free 0/17aAnaerobes Carrying anaerobic 46/57

lactobacillus plusE. coli C25

Anaerobes Carrying anaerobic 0/4lactobacillus only

Anaerobes plus E. Germ-free 0/11coli C25, plusanaerobic lacto-bacillus

a Number of mice in which anerobes becameestablished/total number of animals inoculated.

TABLE 2. Recovery of E. coli C25 on streptomycin-containing Deoxycholate Agar from mice

monocontaminated for 91 days

Mouse no. No. of E. coli cells per Microsco ic counts ofMousno. cecum cultured E. coli cels per cecum

36/5 1.8 X 1010 1.97 X 101036/6 2.0 X 1010 1.46 X 101036/7 1.3 X 1010 1.8 X 101036/8 1.4 X 1010 2.0 X 101036/9 2.2 X 1010 3.0 X 101036/10 2.9 X 1010 2.2 X 101036/11 1.8 X 1010 2.1 X 10'0

bacillus and E. coli strains used above. The micewere killed 5 or more days after inoculation, andthe presence of anaerobic bacteria was deter-mined by anaerobic culture and microscopicobservation of cecal homogenates. Presence ofthe anaerobes at 5 or more days after inoculationwas considered evidence of successful implanta-tion. In all instances, the E. coli or the lacto-bacilli, or both, grew to high levels after the micewere exposed to them in their drinking water.The results in Table 1 show that the anaerobescould be implanted only in animals carrying apreviously established flora of E. coli. The Lacto-bacillus strain used was ineffective, as was simul-taneous inoculation of the anaerobes in mixedculture with E. coli. Consequently, all subsequentexperiments involving implantation of anaerobeswere carried out afterE. coli had been establishedin the mice for at least 2 days.One important feature in the experiments

described below is the determination of the E.coli population in the cecum. To facilitate the

378 INFEC. IMMUN.

on June 8, 2020 by guesthttp://iai.asm

.org/D

ownloaded from

ENTERIC FLORA OF GNOTOBIOTIC MICE

recovery of this organism, the streptomycin-resistant strain C25 was used, because it could berecovered quantitatively on streptomycin-con-taining Deoxycholate Agar, a medium which didnot support the growth of any other bacteriumused in these studies. To investigate the efficiencyof the above culture medium in recovering thisstrain, germ-free mice were monocontaminatedwith E. coli C25. After 91 days, the mice were

killed and their ceca were homogenized and cul-tured in the usual manner. As may be seen inTable 2, there was close agreement between themicroscopic and viable counts within the limitsof error to be expected with the methods em-

ployed. This indicates that E. coli could berecovered quantitatively by viable count fromthe mouse cecum and that there was no signifi-cant overgrowth of streptomycin-sensitive mu-

tants under the conditions of this experiment.It is known that normal mice may harbor a

sizable bacterial population in their stomach(13). Preliminary experiments in the presentstudy had also shown that mice monocon-

taminated with E. coli C25 harbor large numbersof this microorganism in the stomach, but not inthe jejunum or upper ileum. Since subsequentexperiments are to be concerned with the E. colipopulation in the cecum, it was necessary todetermine the degree to which E. coli from the

stomach contributes to the bacterial flora in thececum. Germ-free mice were therefore monocon-

taminated with E. coli C25, and 2 days later were

placed into restraining tubes to prevent coproph-agy, as described under Materials and Meth-ods. Quantitative cultures were made 6 to 10days thereafter of homogenates from stomach or

cecum.The results (Table 3) show that all mice kept

in restraining devices for more than 6 days hadsignificantly reduced E. coli counts in theirstomachs. Most likely, this reduction occurred as

late as the 6th day, since two mice (in experiment1) had normal counts in the stomach and twoother animals (in experiment 3) showed less thannormal counts at this time. The reduction in E.coli counts in the stomach after the 6th day was

consistently in the order of 1,000-fold and is,therefore, statistically highly significant. In con-trast to these findings in the stomach, there wasno effect of the restraining devices on the E. colipopulation of the cecum. All mice kept in re-

straining devices had normal cecal E. coli countswhich were independent of the size of the E. coli

population in the stomach. (Additional counts ofcecal E. coli in unrestrained, monocontaminatedmice are shown in Table 4.) One must, therefore,conclude that E. coli multiplied in the stomach ata rate which was insufficient to maintain a large

TABLE 3. Effiect of a restraining device which prevents coprophagy on the viable counts of E. coli instomach and cecum of monocontaminated mice

Mice in restraining devices Unrestrained control mice (kept in regular cages)Exptno.

Stomach Cecum Stomach Cecum

1 (6)' 6,000 X 105b Not done (6) 6,300 X 105 Not done(6) 4,600 X 105 Not done (6) 19,200 X 105 Not done

(10) <0.01 X 105 Not done (10) 18,500 X 105 Not done(10) 0.01 X 105 Not done (10) 0.35 X 105 Not done

(10) 6,700 X 105 Not done(10) 11,600 X 105 Not done

2 (7) 10 X 105 Not done (7) 12,000 X 105 Not done(7) 17 X 105 Not done (7) 19,000 X 105 Not done(7) 12 X 105 Not done (9) 11,000 X 105 (9) 46 X 108(9) 1.0 X 105 (9) 90 X 108 (9) 4,500 X 105 (9) 60 X 108(9) 0.5 X 105 (9) 110 X 108 (9) 4,300 X 105 (9) 50 X 108(9) 0.15 X 105 (9) 60 X 108 (10) 3,950 X 105 (10) 90 X 108

3 (6) 0.15 X 105 (6) 60 X 108 Not done Not done(6) 200 X 105 (6) 34 X 108 Not done Not done(9) <0.01 X 105 (9) 40 X 108 Not done Not done(9) 0.9 X 105 (9) 80 X 108 Not done Not done

a Number of days the animal was kept restrained prior to culture or, in the case of the control mice,the number of days that the corresponding experimental animals had been restrained on the day ofculture.

b Figures indicate viable (colony) counts of E. coli per one whole organ. Each figure represents onemouse.

VOL. 2, 1970 379

on June 8, 2020 by guesthttp://iai.asm

.org/D

ownloaded from

SYED, ABRAMS, AND FRETER

TABLE 4. Number of E. coli C25 cells in rhe cecum ofmice associated with different types ofintestinalflora

Flora implanted

E. coli only

E. coli plus one strain ofLactobacillus

As in 2 plus one strain each ofenterococcus, Lactobacillus,and Candida, plus 4 morpho-logically different strains ofgram-negative anaerobes

As in 3 plus two additionalstrains of gram-negativeanaerobes with fusiformmorphology

As in 4 plus 30 additionalstrains of gram-negativeanaerobes

Cecal homogenate fromnormal mouse

No. of E. coli cells per cecum5

1-4 daysb

111 X 10844 X 10830 X 108

41 X 10838 X 10830 X 108

5-8 days

90 X 10838 X 10825 X 108

17 X 10815 X 108

30 X 108 12 X 10820 X 108 12 X 108

50 X 1.0

50 X 108

9-12 days

60 X 10830 X 10820 X 108

40 X 10820 X 108

100 X 10880 X 108

5SX 1085 X 1013 X 1083 X 108

1.9 X 10,1.6 X 108l.OX 1080.9 X 10,0.6 X l08

13-16 days

65 X 10846 X 10837 X 108

45 X 108

53 X 10880 X 10840 X 1085 X 10"

5 X 1082 X 1082 X 108

0.97(X 1080.88 X 1080.87 X 1080.70 X 1080.50 X 1080.40 X 101

17-21 days

60 X 10834 X 108

20>X 108

45 X 10130 X 10820 X 108

26-31 days

60 X 10850 X 10833 X 108

14 X 10828 X 108

20 X 10'

1.0 X 1080.76 X 10O0.70 X 1080.56 X 108

Median = 0.095 X 101 (same data as shown in Table 6)

a Each figure represents one mouse. Counts are viable (colony) counts on Desoxycholate Agar con-

taining streptomycin.bDays after implantation of the flora. In experiments 2-6, E. coli was implanted at least 2 days prior

to introduction of the other bacteria. The E. coli population on day zero of these experiments wastherefore in the order of 109 to 1010 per cecum, i.e., similar to that shown in experiment 1.

population in the absence of corpophagy, andthat the E. coli population in the cecum was notderived from the stomach to any significantdegree. Since there was no large E. coli popula-tion in the jejunum and upper ileum of mono-contaminated unrestrained mice, one must con-clude that the E. coli population of the cecum islargely derived from local multiplication of thisorganism in the cecum and perhaps the lowerileum.The main part of the present study consisted

of experiments in which germfree mice were asso-ciated with mixtures of microorganisms ofvarious complexity. Six representative experi-ments are shown in Table 4. In each instance,E. coli C25 had been established first in germ-free mice, followed 2 days later by the flora de-scribed in Table 4. As may be seen, when E. coligrew as a monocontaminant, it established a pop-ulation ranging from 20 x 108 to 111 x 108 bac-teria per cecum. In animals contaminated with

cecal homogenate from a normal mouse (i.e., inour "normal" standard), this E. coli populationwas reduced by a factor of about 1,000 (experi-ment 6). Association with known bacteriaresulted in E. coli populations which were eitherunaffected (experiments 2 and 3) or reduced byfactors of about 10 (experiment 4) or 100. Itshould be noted that in all instances tested(including the data in Tables 5-7) the E. colipopulation adjusted, if at all, within a period ofabout 10 days after introduction of other bac-teria and remained stable thereafter. Cecal sizewas not determined exactly in the experimentsshown in Table 4, but inspection revealed thatthe ceca of mice in experiments 1 to 5 were con-sistently larger than those of the controls (experi-ment 6).

In a final series of experiments, a mixture of80 facultative anaerobes plus 50 strictly anaerobicgram-negative bacteria were tested in the usualmanner for their effect on germ-free mice. The

Exptno.

1

2

3

4

5

6

380 INFEC. IMMUN.

on June 8, 2020 by guesthttp://iai.asm

.org/D

ownloaded from

ENTERIC FLORA OF GNOTOBIOTIC MICE

results of one representative experiment areshown in Table 5. These data should be com-pared with those of Table 6 for "normalized"controls (i.e., animals fed cecal homogenate fromnormal mice). Values for cecal size of germ-freeanimals ranged from S to 10%0 of body weight.As may be seen in Tables 5 and 6, the complex

mixture of bacteria used did reduce the E. colipopulation and cecal size to levels which werewithin the normal range. A further controlexperiment indicated that the 80 facultativeanaerobes used in the above experiments (Table5) did not, by themselves, render recipient germ-free mice normal.Table 7 shows data for mice contaminated

with homogenates from stomach and jejunum ofnormal mice. As may be seen, the bacteria inthese homogenates did not reduce the cecal

TABLE 5. Effect of introducing SO strains of gram-negative strict anaerobes plus 80 facultativeanaerobes on size of cecum anzd E. coli

population in gnotobiotic mice

Mouse no. iDays after Cecum (% No. of E. coliMouse o. ifloracngof body wt) cells per cecum

33.133.233.333.4

33.533.633.733.8

34B134B334B4

33.933.1033.1133.12

34B534B634B7

33.1333.1433.1533.16

33.1733.1833.1933.20

Median

S5S5

10101010

121212

16161616

191919

27272727

60606060

1.71.51.31.8

1.91.71.3

1.60.91.3

1.61.31.31.3

1.10.91.6

1.71.51.61.3

1.81.00.70.9

1.3

41 X 10629 X 10610 X 106

8.5 X 106

16 X 10616 X 10613 X 106

6.0 X 106

8.4 X 1066.7 X 106n 7 \ 11A6

TABLE 6. Effect offeeding a homogenate of normalmouse cecum on size of cecum and E. coli

population in ex-germ-free mice

Day after Cecum (N/ No. of E. coliMouse no. introducingoCecum wt% No.l ofer ccumflora ofbdwt ceapeccu

32.132.232.332.4

32.532.632.732.8

32.932.1032.1132.12

32.1332.1432.1532.16

32.1732.1832.1932.20

32.2132.2232.2332.2432.2532.26

Median

5555

11111111

15151515

20202020

25252525

606060606060

1.11.51.21.4

1.71.61.41.3

1.71.00.60.7

0.91.1

0.7

0.70.71.01.5

0.51.11.41.20.80.5

1.1

20 X 10612 X 10610 X 10676 X 106

12 X 1063.5 X 106

0.02 X 101110 X 106

52 X 10650 X 10611 X 106

9.7 X 106

4.1 X 1066.2 X 1064.1 X 1066.8 X 106

3.0 X 1062.3 X 10663 X 106

6.5 X 106

0.8 X 10672 X 10611 X 106

9.4 X 10B8.0 X 1062.0 X 106

9.5 X 106

U. I A iV- E. coli population or cecal size to normal levels,

3.1 X 106 thus indicating that the cecal flora was necessary12 X 106 for this effect.11 X 106 Histopathological studies. Histopathological

3.8 X 106 studies were carried out on the following groupsof animals: (group 1) germ-free mice, (group 2)

62 X 106 mice monocontaminated with E. coli C25 for 306.0 X 106 days, (group 3) mice contaminated with cecal45 X 106 homogenate from normal animals for 30 days

7.0 X 106 (taken from the experiment shown in Table 6),5.0 X 106 (group 4) mice contaminated with a complex4.0 X 106 mixture of bacterial isolates for 30 days (taken4.9 X 106 from the experiment shown in Table 5). Four to

six mice from each group were studied.8.3 X 106 Stomachs. The histology of the gastric wall per6.4 X 106 se did not vary significantly from group to5.4 X 106 group. In all animals there were a few scattered2.8 X 106 inflammatory cells in the submucosa, this popu-

7.6 X 106 lation being slightly more dense in the animals ofgroups 3 and 4 than in the E. coli-monoasso-

VOL. 2, 1970 381

on June 8, 2020 by guesthttp://iai.asm

.org/D

ownloaded from

SYED, ABRAMS, AND FRETER

TABLE 7. Effect offeeding a homogenate of normalmouse stomach and jejunum on size of cecum and

E. coli population in ex-germ-free mice

Day after Cecum (% No. of E. coliMouse no. introducing of body wt) cells per cecum

flora

31.1 5 2.6 1800 X 10631.2 5 2.6 3,800 X 10631.3 5 2.8 3,900 X 106

31.4 11 3.2 740 X 10631.5 11 2.7 430 X 10631.6 11 3.3 710 X 106

31.7 16 1.1 2,000 X 10631.8 16 3.3 850 X 10631.9 16 3.3 540 X 106

31.10 30 3.3 910 X 10631.11 30 2.7 280 X 106

Median 2.8 740 X 106

ciated or the germ-free animals (groups 1 and2). In both groups 3 and 4, a dense bacterialpopulation was intimately associated with thesuperficial, keratinized layers of the squamousmucosa (Fig. 1). In contrast, only rare individualbacteria were seen in the E. coli-monoassociatedanimals, no layer being formed. The keratinizedlayer itself appeared slightly more compact in thegerm-free animals (group 1) than in members ofthe other three groups.

Small intestine. As expected, the mucosa of thegerm-free small intestine contrasted sharply withthat of the "conventional" (group 3) animals,particularly with regard to the greater develop-ment and cellularity of the lamina propria in thelatter group. A similar degree of development ofthe mucosa was seen in group 4. In contrast, themucosa of the animals monoassociated with E.coli (group 2) could not be distinguished readilyfrom that of the germ-free mice (Fig. 2).Cecum and colon. The cecal mucosa in germ-

free animals was thin and somewhat villous, witha sparse leukocyte population in the lamina pro-pria. In "conventional" animals (group 3), andin animals harboring the complex mixture (group4), the mucosa was thicker and less villous, andhad a greater population of inflammatory cells

in the lamina propria. Monoassociation with E.coli (group 2) failed to alter significantly theappearance of the mucosa as compared with thegerm-free state. Parallel results were obtainedfor the colon (Fig. 3).Dense layers of bacteria lined the epithelial

surfaces of cecum and colon, apparently in themucous layer, in "conventional" animals (group

FIG. la. Section of nonglandular gastric mucosa ofa mouse harboring a defined flora (group 4). Thegranular appearance of the loose, uppermost epitheliallayers is due to the presence of bacteria intimatelyassociated with the epithelium. Hematoxylin andeosin, X 280.

FIG. lb. Section of nonglandular gastric mucosa ofa mouse of group 4. Numerous bacteria are adherentto superficial squamae. Methylene blue, X 685.

3). With artifacts of fixation and sectioning, it wasdifficult to judge the continuity of this strikinglayer of bacteria, but it appeared somewhat more

complete in colon than in cecum. The bacteriallayers in group 4 were indistinguishable histo-logically from those in group 3. No discrete layersof bacteria could be identified in the E. coli-monoassociated animals of group 2 (Fig. 4).

In summary, then, the animals in group 4resembled the "normal" controls (group 3) withrespect to all histological features examined. Incontrast, animals monoassociated with E. coliC25 histologically resembled germ-free mice.

DISCUSSIONThe data presented indicate that it is possible

with the improved culture methods used in thepresent study to isolate those intestinal bacteria

a -,,. .... ..0.# Al.

... E w ,,^k

i a eW s

tjRo

_S eh.

*8 8'

}'} i

T..: :'.,.

'.-3

r-lt. 11,4.

J`*i

b

INFEC. IMUN.382

on June 8, 2020 by guesthttp://iai.asm

.org/D

ownloaded from

ENTERIC FLORA OF GNOTOBIOTIC MICE

a b

FIG. 2a. Section of ileal mucosa of a germ-free mouse (group 1). The lamina propria of the villi is poorly de-veloped, and the crypts of Lieberkuhn are shallow. Contrast with 2c and 2d. Hematoxylin and eosin, X 215.

FIG. 2b. Section of ileal mucosa ofa mouse harboring E. coli (group 2). The appearance is similar to that shownin 2a. Hematoxylin and eosin, X 215.

FIG. 2c. Section of ileal mucosa of a mouse harboring a conventional flora (group 3). The lamina propria, withits "usual" cellularity, contrasts with that shown in 2a and 2b. The crypts are also deeper. Hematoxylin and eosin,X 215.

FIG. 2d. Section of ileal mucosa of a mouse harborinig a defined flora (group 4). The histological appearanceresembles that of the conventional mucosa illustrated in 2c. Hematoxylin and eosin, X 215.

which are responsible for the "normal" featuresof the conventional mouse, at least within thelimits of the criteria examined. It is unlikely thatall of the bacteria used in the experiment shownin Table 5 are necessary for this effect. In fact,preliminary experiments have shown that thisnumber may be reduced drastically. However,some strict anaerobes appear to be requiredbecause the facultative aerobes used (Table 5)were ineffective when established alone in theabsence of strict anaerobes. It has been demon-strated (Table 1) that the establishment of atleast some strictly anaerobic bacteria requiresthe presence of other bacteria. For this reasonalone it seems likely that more than one speciesof microorganism is necessary to establish"normal" features in a germ-free mouse.The requirement of a previously established

E. coli flora for the successful implantation of a

strictly anaerobic bacterium recalls the findingsof Gibbons et al. (6), who showed that someanaerobic bacteria may be established in germ-free mice only as polycontaminants, not asmonocontaminants. The simplest explanation forthis finding would be that the E. coli strain usedin the present experiments reduced the oxidation-reduction (O-R) potential of the intestinal con-tents to a level permitting growth of strict an-aerobes. It is well known that the O-R potentialin the cecum of germ-free mice is relatively high(7) and that E. coli is very active in reducing itsenvironment (5).One may conclude from the above studies

that the population of E. coli in the gastrointes-tinal tract of the mouse is a function of the mul-tiplication of this bacterium in the cecum, andperhaps the lower ileum, and that the sizable E.coli population in the stomach is largely depend-

VOL. 2, 1970 383

on June 8, 2020 by guesthttp://iai.asm

.org/D

ownloaded from

SYED, ABRAMS, AND FRETERa F *r.S \ 4;:.4: it|i. g,.'

w; X

, .... 6

w ...f. S.

.>iic.

* - R S 9v**A';A

FIG. 3a. Section of cecal wall of a germ-free mouse (group 1). The enttire wall is thin, the mucosa somewhatvillous, and the lamina propria poorly developed. Contrast with 3c and 3d. Hematoxylin and eosin, X 215.

FIG. 3b. Section of cecal wall of a mouse harboring E. coli (group 2). The appearance is similar to that shownin 2a. Hematoxylin and eosin, X 215.

FIG. 3c. Section ofcecal wall ofa mouse harboring a conventionalflora (group 3). The thicker wall, the mucosalarchitecture, and the cellularity of the lamina propria contrast with the appearance of 3a and 3b. Hematoxylinand eosin, X 215.

FIG. 3d. Section of cecal wall of a mouse harboring a defined flora (group 4). The histological appearance re-sembles that of the conventional mucosa illustrated in 3c. Hematoxylin and eosin, X 215.

ent on that of the cecum. The cecal population,in turn, appears to be controlled by the otherbacteria normally present at this site. The datapresented indicate that these "other" bacteriamust include anaerobes in order to be effective.The data presented in Tables 4-6 allow somepreliminary conclusions as to the mechanism bywhich the normal bacteria control the E. colipopulation. Meynell (8) concluded from studieson the growth rate of Salmonella in the normaland streptomycin-treated mouse intestine thatthe mechanism controlling the intestinal popula-tion of enterobacteria was based on the inhibi-tion of these microorganisms by volatile fattyacids. The present findings are at variance withthis hypothesis. Regardless of the type of floraestablished, the population of E. coli C25 becamestable at a density which was usually lower than

that in mice monocontaminated with E. coli C25.If the cecum is looked upon as a mixed continu-ous-flow culture, as was done by Meynell, astable E. coli population would, of course, implythat the growth rate of E. coli was unimpairedeven when its population size was much reduced.In fact, since normal mice have been found toshow a faster intestinal transit time than germ-free animals (1) the growth rate of E. coli in micecontaminated with other bacteria may actuallyhave been somewhat higher than that in mono-contaminated mice (even though the populationdensity was much higher in the latter case). Thisfinding is, of course, incompatible with thepresence of a fatty acid inhibitor which functionsto reduce the overall rate of bacterial multiplica-tion in a completely mixed system. It is, how-ever, compatible with the assumption that the

s t->*E.; b

..:

:. tS: *:w , f - A x2'w'W/.,.: r.< r ..... . =* .Y ^.: <s;E -- ._Ni,. _

INFEC. IMMUN.384

on June 8, 2020 by guesthttp://iai.asm

.org/D

ownloaded from

ENTERIC FLORA OF GNOTOBIOTIC MICE

-4:

FIG. 4a. Section of colon of a mouse harboring a

conventional flora (group 3). The blank space across

the mid-portion of the picture represents a shrinkageartifact. The light granular layer immediately abovethis space represents a feltwork of bacteria. The upper-

most darker layer is fecal. Hematoxylin and eosin,X 190.

FIG. 4b. Section of colon of a mouse harboring a

defined flora (group 4). The dense mass above theshrinkage space is a feltwork of bacteria comparableto that shown in 4a. Hematoxylin and eosin, X 190.

FIG. 4c. Serial section of the bacterial layer shownin 4b. The bacterial population consists oflong, slender,or slightly tapered rods. Methylene blue, X 915.

size of the E. coli population is controlled by alimiting nutrient, a mechanism which has beenfound to function in E. coli-Shigella interactionsin the mouse intestine (5).An alternate explanation for the above findings

may be reached, if one assumes that the cecum isnot analogous to a perfectly mixed continuous-flow culture. Most likely, bacteria associatedwith the mucosa would be washed out at a lowerrate than those in the lumen and, consequently,could maintain a stable population at a lowergrowth rate than those multiplying in the lumen.Thus, reduction of the overall growth rate byfatty acids or some other inhibitory mechanismmay eliminate the population of a given bacterialspecies from the lumen but still allow a constantnumber of bacteria to be shed from the mucosa-associated part of the original population. If thisoccurred in vivo, it would, of course, invalidateany conclusions drawn from the simple mathe-matical model used by Meynell (8). However, inthe case of E. coli, such a situation appears lesslikely because, as discussed above, the E. coliused in the present studies did not form an exten-sive mucosa-associated flora comparable to thatfound with certain other bacteria.

It is obvious from this discussion that thequestion of control mechanisms operating in theintestinal flora requires intensive future study.The data presented in the present paper providethe necessary tools for such studies by showingthat the bacteria involved in the control processcan be isolated in pure culture and can be re-implanted into germ-free mice to form a knownflora with "normal" properties.

ACKNOWLEDGMENTS

This study was supported by Public Health Service grantsAI-07328 and AI-07631 from the National Institute of Allergyand Infectious Diseases.

LITERATURE CITED

1. Abrams, G. D., and J. E. Bishop. 1967. Effect of the normalmicrobial flora on gastrointestinal motility. Proc. Soc.Exp. Biol. Med. 126:301-304.

2. Aranki, A., S. A. Syed, E. B. Kenney, and R. Freter. 1969.Isolation of anaerobic bacteria from human gingiva andmouse cecum by means of a simplified glove box procedure.Appl. Microbiol. 17:568-576.

3. Donaldson, R. M. 1964. Normal bacterial populations of theintestine and their relation to intestinal function. N. Engl.J. Med. 270:938 and 994.

4. Dubos, R., R. W. Schaedler, R. Costello, and P. Hoet. 1965.Indigenous, normal and autochthonous flora of the gastro-intestinal tract. J. Exp. Med. 122:67-75.

5. Freter, R. 1962. In vivo and in vitro antagonism of intestinalbacteria against Shigella flexneri. J. Infec. Dis. 110:38-46.

6. Gibbons, R. J., S. S. Socransky, and B. Kapsimalis. 1964.Establishment of human indigenous bacteria in germ-freemice. J. Bacteriol. 88:1316-1323.

7. Gordon, H. A. 1968. Is the germ-free animal normal? Areview of its anomalies in young and old age, p. 127-150.

VOL. 2, 1970 385

lw...ft'-....n.

I

.1.

k.. 'Irlk....:4 .k.%:,4:.. P.N

on June 8, 2020 by guesthttp://iai.asm

.org/D

ownloaded from

SYED, ABRAMS, AND FRETER

In M. E. Coates (ed.), The germ-free animal in research.Academic Press Inc., New York.

8. Meynell, G. G. 1963. Antibacterial mechanisms of the mousegut. II. The role of EH and volatile fatty acids in the nor-

mal gut. Brit. J. Exp. Pathol. 44:209-219.9. Miller, C. P., and M. Bohnhoff. 1963. Changes in the mouse's

enteric microflora associated with enhanced susceptibilityto Salmonella infection following streptomycin treatment.J. Infec. Dis. 113:59-66.

10. Moore, W. E. C. 1969. Current research on the anaerobicflora of the gastrointestinal tract. In The use of drugs inanimal feeds. Publication 1679, National Academy ofSciences, Washington, D.C.

11. Plaut, A. G., S. L. Gorbach, L. Nahas, and L. Weinstein.1967. Studies of intestinal microflora. Gastroenterology53:868-873.

12. Prevot, A., and V. Fredette. 1965. Manual for the classi-fication and determination of the anaerobic bacteria. Leaand Febiger, Philadelphia.

13. Savage, D. C., R. Dubos, and R. W. Schaedler. 1968. Thegastrointestinal epithelium aiid its autochthonous bacterialflora. J. Exp. Med. 127:67-76.

14. Smith, L. D. S., and L. V. Holdeman. 1968. The pathogenicanaerobic bacteria. Charles C Thomas, Publisher, Spring-field, 111.

15. Spears, R. W., and R. Freter. 1967. Improved isolation ofanaerobic bacteria from the mouse cecum by maintainingcontinuous strict anaerobiosis. Proc. Soc. Exp. Biol. Med.124:903-909.

16. Virginia Polytechnic Institute Laboratory. 1969. Outline ofclinical methods in anaerobic bacteriology. The AnaerobeLaboratory, Virginia Polytechnic Institute, Blacksburg.

386 INFEC. IMMUN.

on June 8, 2020 by guesthttp://iai.asm

.org/D

ownloaded from