APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Feb. 1980, p. 430-435 Vol. 39, No. 20099-2240/80/02-0430/06$02.00/0

Isolation and Analysis of Molds from Soy Sauce Koji inThailand

A. BHUMIRATANA,' * T. W. FLEGEL,' T. GLINSUKON,2 AND W. SOMPORAN3Departments ofMicrobiology' and Physiology,2 Faculty of Science, Mahidol University, and Department of

Science, Ministry of Science, Technology and Energy,3 Bangkok, Thailand

Five different isolates of Aspergillus and one of Mucor were compared with aJapanese commercial strain of Aspergillus oryzae for proteolytic activity onwheat bran substrate. One isolate of Aspergillus with superior protease produc-tion, identified as Aspergillus flavus var. columnaris, showed no detectableaflatoxin production on glutinous rice or soybean substrate. Preliminary testsusing this fungus as a koji mold in a traditionally operated factory resulted in asoy sauce superior in quality to that usually produced.

Soy sauce fermentation in Thailand is usuallycarried out by small producers following tradi-tional methods handed down through the gen-erations. The process is of the Chinese ratherthan the Japanese type (3) and, briefly, it con-sists of the following steps. Whole soybeans aresoaked for various periods (usually overnight)before being cooked (3 to 6 h) in boiling wateror with steam. They are then drained and cooledbefore being mixed with a quantity of wheat orrice flour (Fig. 1). Amounts vary and, in somecases, additions are made until the mixture"looks right." However, the total weight addedvaries roughly from 15 to 50% of the dry weightof the soybeans used. In some establishments,the flour is roasted before being added. The beanand flour mixture is next placed in woven bam-boo trays which are incubated in racks (Fig. 2)for 4 to 7 days (koji stage). Fungi inoculatedfrom the air and from the bamboo trays (theseare not sterilized between batches) overgrow themixture, which is eventually dumped into largeearthenware jars (Fig. 3) with sea salt brine(final concentration of salt is approximately 18%[wt/vol] and results from mixing approximately20 kg of koji with 60 liters of 22% salt brine).Brine incubation (mash, or moromi, stage) usu-ally lasts for 30 to 90 days in open yards in thesun, but some hasty processers use incubationsof only 2 weeks. The original brine is drained offand constitutes the first grade of soy sauce. Morebrine may then be added, and further incubationfor 1 or 2 months may follow to yield lower-grade sauces. In other cases, the solid bean sed-iment is separately bottled and sold as anothertype of condiment (called tau jeow). The totalprocess is summarized in Fig. 4 and is character-ized by no control over the inoculum or theculture conditions. As a result, the quality of theproduct is variable, and some batches are decid-

edly inferior. This contrasts markedly with Jap-anese manufacturers who control the inoculum,the culture conditions, and the ageing condi-tions.The microbiology of soy sauce fermentation

has been studied in some detail (12), and one ofthe most important features of the koji step isthe growth of Aspergillus strains with high pro-tease activity. In this regard, interviews withseveral Thai factory owners revealed that thegreen kojis common during cool weather werepreferable to the black or grey kojis commonduring hot weather. Our examinations of greykojis showed that they contained a predomi-nance of zygomycetous fructifications (e.g., Rhi-zopus spp., Mucor spp., Absidia spp., and aSyncephalastrum sp.), whereas the green onesshowed a predominance of Aspergillus conidio-phores. The work in this study was thereforedirected predominantly towards Aspergillusstrains, although a representative of Rhizopuswas included for comparative purposes.With the aim of eventually improving the

reliability and constancy of local soy sauce man-ufacture by the introduction of controlled kojiinoculum, we have isolated several molds fromlocal factories and analyzed them for suitabilityas koji fungi. Our criteria for selection were highprotease activity, no toxin production, and agood taste and aroma of the final product. Thisreport describes these analyses, which culmi-nated in the recommendation ofone ofthe moldsfor factory use.

MATERIALS AND METHODSMold strains. Samples of soy sauce koji of various

ages were obtained from a number of factories in theBangkok area. They were collected in plastic bags andtaken to the laboratory within 1 h, at which time theywere either processed immediately or stored at 4°C

430

on April 25, 2019 by guest

http://aem.asm

.org/D

ownloaded from

MOLDS FROM SOY SAUCE KOJI IN THAILAND 431

FIG. 1. Factory workers mixing wheat flour withcooked soybeans on a wooden mixing table. Theamount of flour added was not weighed but ratheradded until the mixture looked right.

FIG. 2. Koji stage of soy sauce fermentation con-sisting of uncleaned, unsterilized bamboo trays towhich the soybean flour mixture illustrated in Fig. 1was transferred. The koji was not purposely inocu-lated with microorganisms but must have becomerandomly inoculated by spores on the trays, in theair, and in the uncooked flour. These trays wereincubated in the open with no temperature or humid-ity control.

and processed the following day. One gram of eachsample was thoroughly mixed with 4.0 ml of sterilized0.05 M phosphate buffer (pH 7.0), after which thesupematant suspension was serially diluted in thesame buffer. Then, 0.1 ml of each dilution was platedon Sabouraud dextrose agar (Difco Laboratories), andthe plates were incubated at 30°C for 2 days. Predom-inant mold colonies were transferred by streakingspores on new plates, from which stock cultures werein turn obtained. Six of these were selected for detailedanalysis, and they were designated 2a, 4a, 5a, 7a, 8a(Aspergillus spp.), and lc (Mucor sp.). One Japanesecommercial strain of Aspergillus oryzae (designated9a) was used for comparative purposes.

Evaluation of protease production. One sub-strate used to assess protease production by stockcultures consisted of 5 g of wheat bran with 5 ml of0.5% NaCl solution mixed and autoclaved in 500-mlErlenmeyer flasks. After cooling to room temperature,the flasks were inoculated with 5.0 ml of spore suspen-sions (107 spores per ml), which had been prepared asfollows. Fungus cultures were grown on 50 ml of Sa-

bouraud dextrose agar in 250-ml Erlenmeyer flasks at30°C for 7 days. Thirty milliliters of 0.05 M phosphatebuffer (pH 7), containing 2 to 3 drops of Tween 20,was added, and the flasks were shaken vigorously tosuspend the spores. Numbers were determined bydirect count in a Bright-line hemacytometer (Ameri-can Optical Corp.) or by plate counts, and appropriatedilutions were made with phosphate buffer beforeinoculation. The wheat bran cultures were incubatedat room temperature (30 to 35°C). Another substrateused consisted of soybean-flour mixtures (describedbelow) of 300 g each, which were sterilized by auto-claving before inoculation and incubation on bambootrays in the open at ambient temperatures (30 to35°C). These cultures were used for comparison withthe relative enzyme activities obtained using wheatbran substrate.At daily intervals up to 4 days, two flasks for each

tested mold strain were individually extracted by stir-ring for 15 min with 50 ml of 0.04 M phosphate buffer(pH 6.5) for wheat bran cultures and with 200 ml of0.04 M phosphate buffer for soybean cultures. Theextract was then filtered through Whatman no. 2 filterpaper, and the filtrate (crude proteolytic enzyme ex-tract) was stored in sealed vials at -20°C until assayed.The proteolytic activity of these crude extracts was

determined at 40°C by the casein digestion method ofSpies (11). One unit of protease activity was definedas 1 ug of tyrosine equivalent liberated per min.

Protease production by the six fungi isolated in

f

FIG. 3. Factory worker transferring 5- to 7-day-oldkoji to an earthenware jar before the addition of saltbrine and incubation in the sun.

VOL. 39, 1980

on April 25, 2019 by guest

http://aem.asm

.org/D

ownloaded from

432 BHUMIRATANA ET AL.

Soybeans

soaked in water overnight

cooked(by boiling or by steam)

drained and cooled

Flour(wheat and/or rice)

optional roasting

mixed undernonsterile conditions

Incubated 4-7 days on uncleaned,previouslyused bamboo trays in the open; no temperatureor moisture control

(Koji stage)

Transferred to earthermarejars

221 sea salt brine added

beans

nore brine added andincubation in the su

lower grade soy sauc

or

incubated in the sun for 60 days(moronm stage)

beans and liquidseparated by draining

liquid

further incubated in th! sun for a

n for further period of variable:e duration

filtered and bottled as firstgrade ;(y sauce

bottled as a condement calledtao Jemow (low grade)

FIG. 4. Schematic outline of the process of soysauce manufacture in a traditionally operated factoryin Thailand.

Thailand was compared with that of a culture ofAspergillus oryzae sold commercially as a koji moldin Japan.

Toxicity assays. Glutinous rice (250 g) mixed with250 ml of water was autoclaved in two 2,800-ml Fern-bach flasks for 15 min at 121°C. The flasks were theninoculated with an appropriate fungus and incubatedat room temperature (30 to 35°C) for 10 to 14 days.Then both the whole cultures were extracted twicewith 500 ml of chloroform in a Waring blender. Thetwo extracts were pooled (1 liter) and filtered severaltimes through Whatman no. 1 filter paper before beingcleared and dehydrated by filtration through anhy-drous sodium sulfate. On a vacuum rotary evaporator,the volume was reduced to an oily residue which was

taken up in petroleum ether. Insoluble matter was

removed by filtration and dried in vacuo. This methodwas modified from that of Glinsukon et al. (2) forpreparing crude toxin, and the final residue was la-beled "crude toxin." It was tested for aflatoxin contentand toxicity to rats as outlined below.

(ii) Duplicate 100-ml samples of soy sauce were

thrice extracted for 30 min with 250 ml of chloroformon a wrist action shaker. The pooled extracts (750 mleach sample) were then filtered through anhydroussodium sulfate to remove water. The filtrate was thenreduced to a small volume in a rotary vacuum evapo-

rator, transferred to a small vial, and evaporated to

APPL. ENVIRON. MICROBIOL.

dryness under N2 gas. This crude extract was thenassayed for aflatoxins as described below.

(iii) Toxicity to rats was tested for in weanlingfemale rats (40 to 50 g, Fischer-derived strain; AnimalProduction Centre, Faculty of Science, Mahidol Uni-versity) with five animals each in all test and controlgroups. Crude toxin was dissolved in dimethyl sulfox-ide and administered intraperitoneally at a dose levelequivalent to 40 g of mold rice (-30 mg of crude toxinper rat). Animals were sacrificed 7 days after dosing,and autopsies were performed for gross examinationof visceral organs.

(iv) The aflatoxin assay was modified from themethods of Eppley (1). Dried extracts [(i) and (ii)above] were dissolved in 0.1 ml of chloroform, and100- to 200-Ml samples were applied to thin-layerchromatographic plates along with standard aflatoxinsB,, B2, G, and G2. The plates were developed in twosolvent systems and monitored for aflatoxin underultraviolet light. Sensitivity was down to 25 ng.

Preparation of semicontrolled koji. Commercialdried soybeans (44 kg) were soaked overnight in anexcess of tap water. After draining, the beans weresteam cooked at 15 lb/in2 for 15 min, cooled to roomtemperature, and mixed thoroughly with 6 kg of riceflour previously roasted in a hot-air oven. The kojiwas then inoculated with spores (106 per g [dry weight]of koji). The inoculated koji was then spread on roundbamboo trays (-60-cm diameter) like those used inthe factories, except for being washed before use. Themaximum depth of the koji did not exceed 4 cm. Thetrays were incubated in the open at room temperaturefor 7 days before being transported to a factory forincubation in salt brine, parallel to regular factory koji.The object of this experiment was not to work understerile conditions, but rather to follow a proceduresimilar to that in the factory except for the care takento inoculate the koji with spores from a specific mold.

Miscellaneous analytical methods. Protein wasroutinely determined by the method of Lowry et al.(7) with tyrosine as the standard. Soluble nitrogen wasdetermined by the method of the Association of Offi-cial Analytical Chemists (5), and reducing sugar wasdetermined by the method of Somoggi (10). Gustatoryand olfactory judgments were made by the soy saucefactory owners.

RESULTSThe proteolytic activities of wheat bran cul-

ture extracts for six Thai isolates of koji fungiare shown in Table 1, along with the activity ofa commercial strain of A. oryzae (9a) from Ja-pan. Strain 4a gave the highest sustained activityover the 4-day test period. Similar tests on soy-bean substrate with fewer isolates also ranked4a highest in activity (Table 2). As a conse-quence, 4a was tested for toxicity as a prelude toits use as inoculum in factory-produced koji.

Glutinous rice cultures of strain 4a yielded anaverage of 0.67 g of residue from crude toxinextracts. However, no aflatoxins were detectablein the residue, and toxicity tests with rats werenegative (there were no deaths or significant

on April 25, 2019 by guest

http://aem.asm

.org/D

ownloaded from

MOLDS FROM SOY SAUCE KOJI IN THAILAND 433

changes in body weight, and visceral organs werenormal).For identification, 4a was grown on Czapek

solution agar and malt extract agar (8) with thefollowing results. Growth on Czapek agar was

4.5 cm in 5 days with colonies that were yellow

green, not turning brown with age (2 weeks).Conidiophores arose mostly from the agar andwere evenly produced but not densely packed;many conidial heads were columnar. In olderparts of the colonies there was occasional over-

growth by aerial mycelium bearing smaller co-

nidiophores. On malt extract agar, growth was

6.5 cm in 5 days with green colonies. Virtuallyall conidial heads were tightly columnar, arisingdirectly from the agar. Conidiophores were nottightly packed but showed slight zonation. Co-nidiophores on Czapek agar were not over 300,im long, and on malt extract agar they were notover 250 ,Am long. The following macroscopicfeatures were the same on both media. Co-nidiophores, 6.5 to 10.4,um wide with vesicles 13to 28 ,um in diameter, were at first clavate butlater more or less globose. Sterigmata weremostly on the upper two-thirds to one-half ofthe vesicles, uniseriate and 7 to 12 ,um long.Spores were globose, 3.5 to 7.5 ,um in diameter,and minutely sculptured (Fig. 5).The isolate was identified as Aspergillus fla-

vus var. columnaris Raper and Fennel (8), andthis identification was kindly confirmed by K. B.Raper.

Results of tests made on koji and soy saucefrom a regular factory run and from one usingstrain 4a as inoculum are shown in Fig. 6 andTable 3. Figure 6 shows the protease activity oftwo regular factory runs and a third one usingstrain 4a as inoculum. The activity with 4a was

approximately 4 times higher from days 2 to 5.Analysis of the soy sauce product of these kojis

TABLE 1. Proteolytic activity in wheat brancultures of six isolates compared with A. oryzae (9a)

from Japan

Tyrosine equivalent released (jug/g [dry wt]Strain of substrate per min) at following no. of

days of incubation:

1 2 3 4

lc 0 103 187 68 1702a 0 696 1,582 811 2244a 0 3,313 11,360 6,978 2,4885a 0 500 3,077 1,109 4327a 0 6,830 6,410 811 9478a 0 5,950 3,624 892 4469a 0 1,353 974 122 439Control' 0 0 0 0 0

'The control consisted of equally treated but un-

inoculated substrate.

TABLE 2. Proteolytic activity of three Aspergillusstrains on soybean substrate

Tyrosine equivalent released (,ug/g [dry wt] of substrate per min) at

Strain following no. of days of incuba-tion:

1 2 3 4

4a 728 8,008 6,344 5,6065a 270 520 1,352 1,1449a 208 728 2,264 2,288Controla 624 1,040 832 208

a The control consisted of equally treated but un-inoculated substrate.

after 45 days' incubation in salt brine is shownin Table 3. Several features are apparent. Levelsof the tyrosine equivalents, sugar, and solublenitrogen were all higher in the soy sauce pro-duced from 4a-inoculated koji; the soluble nitro-gen, in particular, was much higher.

In the opinion of the factory owners, the or-ganoleptic qualities of the soy sauce resultingfrom the semicontrolled koji using strain 4a asinoculum made it superior to that resulting fromthe regular factory koji.

DISCUSSIONBecause of its superior protease production

and flavor-inducing qualities, we recommendstrain 4a for use as a koji mold in Thailand. Thefact that it is a locally occurring fungus may lendit some ecological advantage over imported kojimolds, and this could be particularly importantin factories where the control over koji cultureis limited. To our knowledge, this is the firstreport of the use of a variety of A. flavus as akoji mold in the soy sauce industry. This maycause some concern to those who are worriedabout the danger of aflatoxin production, but wefeel there are several reasons to allay any fearsand justify its use. First, A. flavus and A. oryzaeare closely related, and the distinctions betweenthem are not sharply defined. If one considers arange of characteristics between two extremes,each representing one of these species, some-where in the shady middle one must choose onespecies over the other. In the absence of geneticdata, one is forced to use limited morphogeneticand physiological criteria, and the authority wechose to follow considers the strain we haveselected as a variety of A. flavus. However, alook at the synonyms of this taxon (8) showsthat others have called it various things, includ-ing A. oryzae. Second, we have tested this strainfor aflatoxins on the substrate (i.e., glutinousrice) considered ideal for their production byspecies of A. flavus, and the tests were negative.So were tests with soybean koji and soy sauce.

VOL. 39, 1980

on April 25, 2019 by guest

http://aem.asm

.org/D

ownloaded from

434 BHUMIRATANA ET AL.

i..

.tIt

A': -1o. uI

B *

DFIG. 5. Photomicrographs of strain 4a identified as A. flavus var. columnaris. (A) Young conidiophores

which arose directly from the agar- (180x), (B) old conidiophores and mature spores (180x), (C) mature spores(1,660X), (D) young conidiophore showing uniseriate sterigmata (1,660X).

Third, soybean has been shown to be a poorsubstrate for aflatoxin production, even forstrains of Aspergillus which are high producerson other substrates (4, 9; C. W. Hesseltine, in P.MatangKasombut, ed., Proc. Fifth Int. Conf.Global Impacts Appl. Microbiol.). On thestrength of these arguments, we feel confidentin recommending 4a as a koji mold and in sub-jecting it to further improvement by mutagene-sis. Large-scale tests are presently being con-

ducted at a new factory where inoculum and

culture conditions can be controlled. Analysis ofthese tests will be the subject of further reports.In addition, work is under way to select a highamylase-producing mutant of strain 4a that canbe co-inoculated with the high protease strain.Our preliminary tests with laboratory koji in-

oculated on bamboo trays and then removed toa factory for the moromi, or mash, stage gavesuperior results to regular factory runs (judg-ment ofthe manufacturer). However, an attemptto inoculate koji at the factory site (inoculation

APPL. ENVIRON. MICROBIOL.

r

r

11

i-ey

4-s

Ji t...I..." 4f: ",

on April 25, 2019 by guest

http://aem.asm

.org/D

ownloaded from

MOLDS FROM SOY SAUCE KOJI IN THAILAND 435

TABLE 3. Qualities of soy sauce produced from semicontrolled koji inoculated with strain 4a comparedwith those of soy sauce produced from regular factory koji

Protease Tyrosine Soluble nitro- Sugar (g of Toxicity (afla-Soy sauce prepared from: (U/mI) (g/liter) gen (g/liter) liter) toxin, bioas-Acceptabylity

Semicontrolled kojiAfter filtering 96 38.7 10.2 45.9 NDc OKAged 103 37.8 10.8 40.7 ND OK

Factory-prepared kojiAfter filtering 64 22.9 4.65 32.1 ND OKAged 126 28.2 4.9 35.4 ND OK

a Sensitivity was down to 25 ng.b The bioassay was done in weanling female rats with a dose level equivalent to 40 g of mold rice per rat.'ND, Not detectable.

0

Y 4~ +Pure Cultured Koji

D 2-

0)

E0a. tFctory Prepard Koji

0

0 2 4 6 8 10

DAYSFIG. 6. A comparison ofprotease activity in labo-

ratory-prepared and factory-prepared koji. The lab-oratory-prepared koji differed from the factory kojiin the care taken to clean the bamboo trays and topurposely inoculate the substrate before incubation.Protease activity is given in kilounits per gram (dryweight) of koji.

at the flour-mixing step) did not produce thedesired green koji but an undesired black one.

This may have resulted because of higher tem-peratures at the factory (the temperature in thekoji was measured as high as 50°C on the secondday of incubation, whereas the optimum tem-perature for growth with this fungus as meas-

ured by colony diameter tests on agar was 30°C)or because of contamination arising from thepreviously used bamboo trays. It may also haveresulted from improper moisture content or poorventilation. In any case, it was recommended tothe owner that some form of control be exertedover temperature and moisture during the kojiculture. The influence and importance of thesevariables have been reviewed by Yong and Wood(12).We have not yet made a detailed study of the

nonmold microorganisms also isolated from thekoji in local factories. However, in spite of an

earlier study recommending co-inoculation offungus, yeast, and bacteria (6), it appears thatgood results can be obtained by using only mold

inoculum (12). For this reason, it may be moreprofitable to concentrate on the yeasts and bac-teria in the mash, or moromi, stage.Because the amylase production by strain 4a

is relatively low, work is currently under way toselect a high-amylase mutant by ultraviolet mu-tagenesis. Such a mutant will then be tested asa co-inoculant with wild-type 4a.

ACKNOWLEDGEMENTWe acknowledge the excellent technical assistance of Wa-

tanalai Panbangred and Attawut Impolsap.This work was supported by a research grant from the

Association of South East Asian Nations-Australian EconomicCooperation Programme Protein Project.

LITERATURE CITED1. Eppley, R. M. 1968. Screening method of zearalenone,

aflatoxin and ochratoxin. J. Assoc. Off. Anal. Chem. 51:74-78.

2. Glinsukon, T., S. S. Yuan, R. Wightman, Y. Kitaura,G. Buchu, R. C. Shaule, G. N. Wojan, and C. M.Christensen. 1974. Isolation and purification of Cyto-chalasin E. and two tremorgens from Aspergillus cla-vatus. Plant Food for Man 1:113-119.

3. Hesseltine, C. W. 1965. A millennium of fungi, food andfermentation. Mycologia 57:149-197.

4. Hesseltine, C. W., 0. L. Shotwell, J. J. Ellis, and R.D. Stubblefield. 1966. Aflatoxin formation by Asper-gillus flavus. Bacteriol. Rev. 30:795-805.

5. Horowitz, W. (ed.). 1975. Total nitrogen-kjeldahlmethod, p. 15-16. In Official methods of analysis, 12thed., Association of Official Analytical Chemists, Wash-ington, D.C.

6. Lockwood, L. B. 1947. The production of Chinese soyasauce. Soybean Dig. 7:10-11.

7. Lowry, 0. H., N. J. Rosebrough, A. L. Farr, and R. J.Randall. 1951. Protein measurement with the Folinphenol reagent. J. Biol. Chem. 193:265-275.

8. Raper, K. B., and D. L. Fennel. 1965. The genus Asper-giUlus. The Williams & Wilkins Co., Baltimore.

9. Shotwell, 0. L., E. E. Vandergraft, and C. W. Hessel-tine. 1978. Aflatoxin formation on sixteen soybean va-rieties. J. Am. Oil Chem. Soc. 61:574-577.

10. Somoggi, M. 1945. A new reagent for the determinationof sugars. J. Biol. Chem. 160:61-68.

11. Spies, J. R. 1957. Colorimetric procedures for aminoacids. Methods Enzymol. 3:467-477.

12. Yong, F. M., and B. J. B. Wood. 1974. Microbiology andbiochemistry of soy sauce fermentation. Adv. Appl.Microbiol. 17:157-194.

VOL. 39, 1980

on April 25, 2019 by guest

http://aem.asm

.org/D

ownloaded from