Gut, 1980, 21, 701-705

Faecal bacterial flora in cancer of the colonD VARGO, M MOSKOVITZ, AND M H FLOCH

From the Department of Medicine, Section of Gastroenterology, Norwalk Hospital and the YaleUniversity School of Medicine, Norwalk and New Haven, Connecticut, USA

SUMMARY Selective aerobic and anaerobic plate media were employed to isolate the predominantfaecal flora of patients with cancer of the colon (CC), cancer with non-gastrointestinal involvement(NGI), and with non-malignant diseases (N). The CC and N groups did not differ significantlyin either total aerobic or anaerobic counts. The CC group did have a significantly lower anaerobic/aerobic ratio compared with the N group (2.42 vs. 2.96, p < 0.05). This was the result of a greaterpredominance of aerobic bacteria and a decrease in anaerobic cocci, Eubacterium and Fusobacteriumin the CC group. Previous studies report that aerobic organisms have a greater ability to produceamines than non-spore forming anaerobes. If the intestinal flora can produce carcinogenic nitro-samines in vivo from amines and nitrites, the aerobic bacterium in the faeces may be of importancein supplying the amine substrate for nitrosation. The comparison of the NGI group with the Ngroup showed a significant variation in the total anaerobic count (11.02 vs. 11 -41, p < 0.05) and inthe composition of the faecal flora. This indicates that discretion must be used in analysing the dataobtained from cancer patients, as the presence of a carcinoma may be responsible for changes inbacterial flora.

The relationship between various environmentalfactors and cancer of the colon has received muchattention since epidemiological studies correlated theincidence of the disease with the degree of economicdevelopment.1 2 The main observable differenceamong the populations with varied economicdevelopment has been their dietary pattern. Thehypotheses relating a dietary factor to colon cancerfocus on the enzymatic activity of intestinal bacteria.The enzymes supposedly activate an ingestedcarcinogen, or may produce a carcinogen fromdietary components or intestinal secretions.3

Significant differences have been observed in thefaecal flora and bacterial enzyme activity betweenindividuals residing in countries with a high in-cidence of colon cancer and individuals in countrieswith low incidence.4 5 This investigation was under-taken to study groups of subjects within a populationhaving a high incidence of carcinoma of the colon,and to observe if similar differences occurred in thefaecal flora composition of control groups comparedwith a group with definite colon carcinoma.

Methods

The groups studied consisted of 26 control patients(N), 17 cancer of the colon patients (CC), and

Received for publication 18 February 1980

20 cancer patients with non-gastrointestinal in-volvement (NGI). This study was approved by ourhospital Committee on Human Investigation andinformed consent was obtained from all subjects.The diagnoses of the NGI group of patients con-sisted of nine lung cancer, seven breast cancer,and one each of myeloma, ovarian cancer, reticulumcell sarcoma, and an unknown primary site. TheN and NGI patients were selected to reflect the agedistribution of the CC group. A stool specimenand medical and dietary histories were obtained onall patients. Specimens were collected after initialbowel examinations and before bowel preparationfor surgery. No specimens were taken after thepatient had received a laxative or enema. Becauseof the short time interval before surgery, it wasnecessary to obtain some specimens by digitalremoval of faeces.The stool specimen obtained from each patient

was transferred to an anaerobic chamber in thelaboratory within 15 minutes of elimination. Thecomplete sample was homogenised in a Waringblender and approximately 4 g removed for wet/dryweight determination and I g for serial dilutions.The remainder of the sample was freeze-dried fordetermination of acid and neutral sterols and isreported elsewhere.6

Fourteen selective aerobic and anaerobic plate701

Vargo, Moskovitz, and Floch

Table 1 Aerobic and anaerobic selective plate media

Media Dilutions Primary genus isolated

Aerobic mediaBrucella blood agar" 10-5, -6 -7 Total aerobesPhenyl ethyl alcohol agar lo-6. -6, -7 Streptococci,

staphylococciTergitol-7 agar 10-5 -6, -7 EnterobacteriaXLD agar Undiluted Enteric pathogensEnterococcus M mediag 10- -6, -7 Streptococci

Anaerobic mediaWAL-APM media6 10-7, -9, -10 Total anaerobesNeomycin blood agar10 10-7. -9 -10 ClostridiumKanamycin-vancomycin

blood agar" 10-5' -7' -9 BacteroidesLactobacillus S agar" 10-3 5 -7 LactobacillusF-M media 10-6' -7' -9 FusobacteriumEgg yolk agar" tO-3 -5, -7 ClostridiumRifampin blood agar' 10-5 -7' -9 EubacteriumBifidobacterium medium8 10-7. -, -10 BifidobacteriumVeillonella neomycin agar" 10-7. -6, -10 Veillonella

media were streaked with appropriate dilutionsusing a calibrated loop as indicated in Table 1.Aerobic plates were incubated for two days at 37°C.Egg yolk plates were inoculated from dilutiontubes which were heated in a water bath for 10minutes at 80°C in order to isolate spore producers.Anaerobic plates were incubated in an anaerobicchamber from four to seven days.7 The Lactobacillusplates were incubated in a candle jar at 37°C.For each of the selective media, colonies with

identical morphology were counted. One isolateof each colony type was picked for identificationusing a magnification of x 3.5. Whenever therewas a suspicion of variation in colonial morphologyall varieties were isolated for identification. Anaero-bic identifications were made according to thecriteria of Holdeman and Moore, based on cellularmorphology, Gram reaction, biochemical tests, andanalysis of metabolic fermentation end-product.10Biochemical tests were done using the API 20Anaerobic System (Analytabs Products, Inc.), inconjunction with a micro pH probe (Microelect-rodes, Inc.), to record pH. Aerobic identificationswere made according to conventional methods.The API 20 E System was employed for identi-fication of facultative Gram negative bacilli. Finalcounts were obtained from correlating the platecount with colony identification. This permittedthe greatest recovery of organisms.

Student's t test or median test were used todetermine statistical significance and were per-

formed on the logarithmic transformation of thebacterial counts.1

Results

Total aerobic and anaerobic bacterial counts, theratio of the bacterial counts, and the wet/dry ratio

of the faeces in the three groups are shown in Table2. The total anaerobic mean counts (logl0/g of dryfaeces) of the NGI group differ significantly fromthe N group (11-02 vs. 11.41, p<0.05). No significantvariation is observed in the total aerobic mean.The CC and N groups do not significantly differin either total aerobes or anaerobes.The anaerobic/aerobic ratio was calculated from

plate count totals. The CC group has a significantlylower anaerobic/aerobic ratio compared with the Ngroup (2-42 vs 296, P<005), indicating that theCC patients have a greater predominance of aerobicbacteria.During the laboratory work-up of the samples

it was evident that the NGI and CC stools wereusually less formed. The increased moisture contentof the stools indicated by the wet/dry weight ratiois statistically significant in both groups whencompared with the N group (NGI vs. N, P<005;CC vs. N, P<001), but not when the CC is com-pared with the NGI.The bacteria recovered are placed into categories

in Table 3. For each group the recoveries are listedas (a) percentage frequency of recovery, (b) re-covered range of the counts, (c) the mean withinthe range. The frequency of recovery of all thebacterial categories analysed is grossly similar. Therange of the recoveries for each category is verywide for each patient group. Comparison of theCC and NGI bacterial categories with those of theN group indicates decreased recovery of (1)anaerobic cocci in the CC group (P<0.05) and (2)Eubacterium in the NGI group (P<001).The groups of bacteria were subdivided into

their respective genus and species, and the incidenceof recovery, range and mean studied for genus andspecies. However, no significant differences wereobserved in regard to various organisms amongthe patient groups.

Table 2 Characteristics offaeces and bacterial countsofnormal (N), non-gastrointestinal cancer controls(NGI), and colon cancer subjects (CC)*

N NGI CC

Wet dry ratio 3.99±0 23 5 09±0 541 7-53±1.402Percent H2O 749 80-4 86-7Total aerobic meant 8-46±0 25 8-80±0t24 8-96±0.21Total anaerobic meant 11-41±0 10 11-02±0-183 11-36±015Anaerobic/aerobic ratio 2.96±0-22 2-30±0-23 2.42±020'Percent aerobes 0-90±0 44 5 30±2.80 220± 1-10Percent anaerobes 99-10±0-44 94-70±2.80 97-80±1-10

Results reported as mean±standard error.tLogarithmic transformation of the mean/g dry faeces.N vs NGI P< 05.N vs CC P<001.aN vs NGI P<005.'N vs CC P<0.05.

702

Faecal bacterial flora in cancer of the colon

Discussion

The comparison of the predominant faecal floraof the cancer of the colon patients (CC) and cancerpatients with non-gastrointestinal involvement(NGI) with control patients (N) yields many interest-ing trends and some statistically significant results.As reported by other investigators, we found greatdiversity in the composition of the faecal florabetween individual subjects. 12-14

Hill et al. reported that aerobic bacteria arerecovered in greater numbers in stool from subjectsresiding in countries exhibiting a low incidence ratefor colon cancer.'5 In high incidence countriesBacteroides, Bifidobacterium, and Fusobacteriumare more predominant and Lactobacillus, Entero-cocci and Enterobacteria are less predominant.515It has been postulated that (1) the stool flora fromsubjects in low incidence countries have less bacterialenzymatic activity for activating or producingcarcinogens and that (2) the diet may be the in-fluencing factor causing the variation in intestinalflora and enzyme activity between countries.4

In our study, no differences were observed be-tween the three groups with respect to dietaryintake of protein, fat, carbohydrates, fibre, choles-terol, or total calories. All individuals consumedconventional 'western' type diets.6 The bacterialcounts indicated that similar totals of anaerobicand aerobic bacteria existed between the CC andN groups. However, the ratio of anaerobic toaerobic bacteria of individual subjects indicatedthat the aerobic bacteria made up a greater per-centage of the bacterial flora in the cancer of thecolon patients. This increase in aerobes was producedby the Enterobacteriaceae family having a higherpercentage of the total flora in the CC group,

while the Eubacterium, Fusobacterium, andanaerobic cocci had a lower percentage.The hypothesis that the stool flora of individuals

in high incidence countries have a greater abilityto produce a carcinogen has been advanced byinvestigations of bacterial enzymatic reactions. Theanaerobic bacteria can perform in vitro the types ofreactions necessary to produce a carcinogen frombile acids.'6-'8 Lecithinase negative clostridium,particularly strains of Clostridium paraputrificum,are the only bacteria identified in the normal floracapable of some of the required nuclear dehydro-genation reactions.19 Hill and Drasar stated thatthese organisms are rare in the faeces of peopleliving in areas with low incidence of colon cancerbut much more common in faeces from peopleliving in high incidence areas.20 We did not observeany differences in the frequence of any of theClostridium species in the cancer of the colongroup. Other studies comparing high risk to low riskcolon cancer groups also have failed to showdifferences in the levels of Clostridium paraputrificumand related species.2122Although our data do not support the hypothesis

of a positive correlation between Clostridium,Bacteroides, and other anaerobes with an increasedrisk of colon cancer, our findings are similar tothe results of another study involving the faecalflora of colon cancer patients. Johnson, in a pre-liminary investigation on secondary amine pro-duction by gut bacteria, reported a greaterlogarithmic count of aerobic organisms in faecalspecimens obtained from a small sampling of coloncancer patients before surgery as compared withcontrol subjects.23 It was also reported that aerobicbacteria have a greater ability for metabolisingcholine to secondary and tertiary amines than do

Table 3 Major categories of bacteria: frequency, range, and mean of recovery

N NGI CC

%F* ranget mean: %F range mean %F range mean

Facultative and aerobic Gram-bacilli 88-5 4-11 8.56 100 0 5-11 8-56 88-2 5-9 8-41Facultative and aerobic cocci 76-9 5-10 7-38 65-0 4-9 7.95 76-5 5-10 8-38Facultative and aerobic Gram+bacilli 34-6 4-9 7-11 40-0 4-8 7-41 29-4 6-10 8-15Anaerobic cocci 50-0 5-11 9-07 65-0 6-11 9-78 41.21 6-11 10-07Clostridium 88-5 5-11 9-41 100 0 6-11 9-58 88-2 7-12 9-76Eubacterium 88-5 5-12 10-41 50-02 8-11 10-27 76-5 8-11 10-06Bifidobacterium 65-4 6-12 10-14 55-0 8-11 9-73 58-8 8-11 9-70Propionibacterium 11-5 10-11 10-66 15-0 8-10 9-63 0 - -Bacteroides 96-2 9-12 10-97 90-0 8-11 10-52 100.0 8-12 10-72Fusobacterium 30-8 9-11 10-25 20-0 9-10 9-72 17-6 8-11 9-91Lactobacillus 76-9 6-11 7-95 80-0 5-11 8-80 88-2 5-10 7.90Unidentified Gram+non-sporing bacilli 19-2 8-11 10-08 25-0 8-11 9-38 35-3 9-11 9-96Unidentified Gram-non-sporing bacilli 30-8 5-11 9.05 10-0 8-11 9-91 23-5 10-11 10-27

*Frequency in percentage.tRange rounded off to nearest Log1o.$Logl./g dry weight of faeces.

IN vS CC p<0-052 N vs NGI P<0-01

703

Vargo, Moskovitz, and Floch

non-spore forming anaerobes.23 Choline is presentin the gut from the bacterial degradation of cholinephosphoglyceride (lecithin), a major component ofanimal fat. The intestinal flora can produce carcino-genic nitrosamines in vitro by an enzymatic reactioninvolving nitrites and secondary amines. Thepresence of aerobic bacteria in the colon may beimportant if nitrosamines production occurs in vivo.

Dietary fibre has been mentioned as possiblyhaving a protective effect against colon cancer.2425The 'western' diet of high incidence countriescontains far less fibre than the diets of low incidencecountries.2 We have previously reported the effectsfrom fibre supplementation of 'western' diets ondietary patterns and on faecal bacterial flora.13 26Individuals who increase their fibre intake usuallydecrease their consumption of fat and cholesterol.26Along with these dietary changes may be a decreasein the aerobic bacteria of the stool.'3 This raisesthe possibility that high fibre diets may influence theproduction of nitrosamines by (1) decreasing theaerobic bacteria population capable of producingsecondary amines and/or (2) decreasing the intakeof animal fat.Comparing the NGI faecal flora with the N

group also revealed statistically significant differ-ences. The decrease in the anaerobic count andEubacterium recovery may be related to metabolicchanges caused by the presence of cancer. Cautionmust be observed when interpreting differences incancer patients, as they may be the result, ratherthan a causative factor, of the disease. For example,stool from both the CC and NGI groups had ahigher wet/dry weight ratio, indicating a greaterwater content than the control group. Historiesof our colon cancer patients revealed that loosebowel movements were not common during theearly stages of the disease. Whether this increasein the water content influenced the flora compositionis not known.Sampling bias must also be considered when

drawing an inference for the general colon cancerpopulation from the CC group data. Many pro-spective colon cancer patients were not used inthis study, as they passed no stool during the fewdays available for study between the time of diag-nosis and preparation for surgery. Therefore, wewere forced to eliminate some of the constipatedcancer patients from our colon cancer group.

Another problem in establishing the relationshipbetween bacterial flora and an individual's riskof colon cancer is the natural variation in the normalfaecal flora between individuals. In our study amuch larger sample would be needed to determineconclusively the existence of statistically significantvariation between our CC and N groups. Un-

fortunately, the procedure for bacterial identificationto the species level involves a great deal of mediaand reagent preparation, technician time, andexpense. Because of these factors the qualitativeand quantitative analysis of the intestinal flora maynot be the most efficient method for investigatingthe role of bacteria in colon cancer carcinogenesis.It has been suggested that future investigations ofintestinal bacteria should stress the role of theirmetabolism within the host.27 28 Research alongthis line has already produced interesting results.Mastromarino et al. have reported that stoolsfrom colon cancer patients have significantly morefaecal microbial 7x-dehydroxylase than controls.29Dietary factors have been found to alter bacterialmetabolism. Rats on a meat diet have higherlevels of nitroreductase, azo-reductase, and 3-glucuronidase activity in the stool than those foundduring grain feeding.30 In human volunteers theeffect of a mixed western diet produced higher3-glucuronidase activity in the stool than whenon a non-meat diet.31 Hill et al. reported that agreater number of clostridium with steroid nucleusdehydrogenase are present in the stool from coloncancer patients than control patients.32

In conclusion, the diversity of bacterial florawithin the groups studied makes it difficult to detectrelationships between intestinal bacteria and coloncancer. Our results suggest that there are alterationswithin the bacterial flora of cancer patients com-pared with normal subjects. The colon cancerpatients exhibit an increase in aerobic bacteriashown by the anaerobic/aerobic ratio, and adecrease in anaerobic cocci, Eubacterium andFusobacterium. The results may have clinicalsignificance in regard to the possible productionof nitrosamines. However, patients with non-gastrointestinal involvement acting as a control forthe colon cancer patients also showed alterationsin faecal flora compared with the N group, adecrease in total anaerobic bacteria and Eubact-erium. This indicates that discretion must be usedin analysing data obtained from cancer patients,as the presence of a carcinoma or even the emotionalstress33 may be responsible for changes in bacterialflora.

This study was supported in part by the NationalCancer Institute Grant No. CA-16090. The authorsgratefully acknowledge the technical assistance ofMrs Edith Russell; Dr C White and Dr R Barnett,for statistical analysis of data, and Dr H M Fuchsand R Ramos for specimen collection and gatheringclinical data.

704

Faecal bacterial flora in cancer of the colon 705

References

1Wynder EL, Shigematsu T. Environmental factors ofcancer of the colon and rectum. Cancer 1967; 20:1520-61.

2Burkitt DP. Epidemiology of cancer of the colon andrectum. Cancer 1971; 28: 3-13.3Weisburger JH. Colon carcinogens: Their metabolismand mode of action. Cancer 1971; 28: 60-70.4Drasar BS, Hill MJ. Intestinal bacteria and cancer.Am J Clin Nutr 1972; 25: 1399-404.5Aries V, Crowther JS, Drasar BS, Hill MJ, WilliamsREO. Bacteria and the aetiology of cancer of the largebowel. Gut 1969; 10: 334-5.6Moskovitz M, White C, Barnett RN, et al. Diet, fecalbile acids and neutral sterols in carcinoma of the colon.Dig Dis Sci 1979; 24: 746-51.'Sutter VL, Vargo VL, Finegold SM. Wadsworthanaerobic bacteriology manual, 2nd ed. WadsworthHospital Center: Veterans Administration Los Angeles,1975.8Floch MH, Gershengoren W, Freedman LR. Methodsfor the quantitative study of the aerobic and anaerobicintestinal bacterial flora of man. Yale J Bio Med 1968;41: 50-8.9Finegold SM. Use of selective media for isolation ofanaerobes from humans. Appl Bacteriol Tech Ser1971; 5: 99-108.

'0Holdeman LV, Moore WEC. Anaerobic laboratoryManual, 3rd ed. Blacksburg, Va: Virginia PolytechnicInstitute and State University, 1976.

"Best WR. On the logarithmic transformation ofintestinal bacteria counts. Am J Clin Nutr 1970; 23:1608-9.

'2Moore WEC, Holdeman LV. Human fecal flora: Thenormal flora of 20 Japanese Hawaiians. Appl Microbiol1974; 27: 961-979.

13Fuchs HM, Dorfman S, Floch MH. The effect ofdietary fiber supplementation in man. II. Alterationsin fecal physiology and bacterial flora. Am J Clin Nutr1976; 29: 1443-7.

'4Finegold SM, Sutter, VL, Sugihara PT, Elder HA,Lehmann SM, Phillips RL. Fecal microbial flora inSeventh Day Adventist populations and controlsubjects. Am J Clin Nutr 1977; 30: 1781-92.

'.Hill MJ, Crowther JS, Drasar BS, Hawksworth G,Aries V, Williams REO. Bacteria and aetiology ofcancer of large bowel. Lancet 1971; 1: 95-100.

16Aries VC, Goddard P, Hill MJ. Degradation of steroidsby intestinal bacteria III. 3-0 55-steroid Al-dehydro-genase and 3-0-55-steroid A4-dehydrogenase. BiochimBiophys Acta 1971 ; 248: 482-8.

11Goddard P, Hill MJ. Degradation of steroids byintestinal bacteria. IV. The aromatisation of ring A.Biochim Biophys Acta 1972; 280: 336-42.

18Hill MJ. Steroid nuclear dehydrogenation and coloncancer. Am J Clin Nutr 1974; 27: 1475-80.

19Goddard P, Fernandez F, West B, Hill MJ, Barnes P.The nuclear dehydrogenation of steroids by intestinalbacteria. J Med Microbiol 1975; 8: 429-35.

20Hill MJ, Drasar BS. The normal colonic bacterial flora.Gut 1975; 16: 318-23.

21Finegold SM, Flora DJ, Attebery HR, Sutter VL.Fecal bacteriology of colonic polyp patients andcontrol patients. Cancer Res 1975; 35: 3407-17.

22Crowther JS, Drasar BS, Hill MJ, et al. Faecal steroidsand bacteria and large bowel cancer in Hong Kong bysocio-economic groups. Br J Cancer 1976; 34: 191-8.

23Johnson KA. The production of secondary animes bythe human gut bacteria and its possible relevance tocarcinogenesis. Med Lab Sci 1977; 34: 131-43.

21Walker APR. Colon cancer and diet, with specialreference to intakes of fat and fiber. Am J Clin Nutr1976; 29: 1417-26.

25Freeman HJ, Kim YS, Spiller GA. Effect of chemicallydefined dietary fiber on rat colonic neoplasmia inducedby 1,2-dimethylhydrazine (abstract). Gastroenterology1978; 74: 1036.

26Kahaner N, Fuchs HM, Floch MH. The effect ofdietary fiber supplementation in man. I. Modificationof eating habits. Am J Clin Nutr 1976; 29: 1437-42.

2lWynder EL, Reddy B. Studies of large bowel cancer;human leads to experimental application. J NatlCancer Inst 1973; 50: 1099-106.

28Moore WEC, Holdeman LV. Discussion of currentbacteriological investigations of the relationshipsbetween flora, diet, and colon cancer. Cancer Res 1975;35: 3418-20.

29Mastromarino A, Reddy BS, Wynder EL. Metabolicepidemiology of colon cancer: Enzymic activity offecal flora. Amer J Clin Nutr 1976; 29: 1455-60.

30Goldin B, Gorbach SL. Alterations in fecal microfloraenzymes related to diet, age, lactobacillus supplementsand dimethylhydrazine. Cancer 1977; 40: 2421-6.

31Reddy BS, Weisburger JH, Wynder EL. Fecal bacterial3-glucuronidase: control by diet. Science 1974; 183:

416-7.32Hill MJ, Drasar BS, Williams REO, et al. Faecal bile-acids and clostridia in patients with cancer of the largebowel. Lancet 1975; 1: 535-9.

33Holdeman LM, Good IJ, Moore WEC. Human fecalflora: Variations in bacterial composition withinindividuals and a possible effect of emotional stress.Appl Microbiol 1976; 31: 359-75.