ORIGINAL ARTICLE
Mucosally associated bacterial flora of the human
colon: quantitative and species specific differences
between normal and inflamed colonic biopsies
Shri Pathmakanthan1, John P. Thornley2 and Chris J. Hawkey1
From the1Department of Gastroenterologyand2Microbiology, University Hospital, Queens Medical Centre,
Nottingham NG7 2UH, UK
Correspondence to: ShriPathmakanthan, Department of Gastroenterology, University Hospital Queens Medical
Centre, Nottingham NG7 2UH, UK. Tel: +44 0115 9249924; Fax: + 44 0115 9422232; e-mail:
[email protected]
Microbial Ecology in Health and Disease 1999; 11: 169–174
This study quantified and characterised changes to a species level for aerobic and anaerobic bacteria from colonic biopsies of acute
ulcerative colitis (UC) comparing them with that of a normal control group. Fresh endoscopic biopsies of the recto – sigmoid region were
obtained during flexible sigmoidoscopy from 10 patients suffering an acute attack of UC and a similar number of healthy controls.
Quantitative estimation was carried out on selective media for total aerobic, anaerobic, Bacteroides, Bifidobacterium and Lactobacillus
spp. (Miles – Misra). Species identification involved gram stain, indole, catalase tests and commercially validated anaerobic, gram-positive
and non-enteric fermenting enzyme hydrolysis kits There was a significant quantitative decrease in growth of Lactobacillus spp. in colitic
biopsies (p B0.05). Total aerobic speciation revealed 32 different sub-species with 18 of these found only in UC biopsies. Anaerobic
speciation revealed 41 different sub-species with a mean 6.7 species in normal and 4.7 in UC patients. Bacteroides thetaiotaomicron was
isolated in significantly increased frequency in UC biopsies (8/10) in comparison with normal (4/10). Our data are consistent with the
possibility that a reduction in mucosally associated Lactobacillus bacteria in UC permits proliferation of a large number of potentially
pathogenic aerobic species and anaerobic predominance of B. thetaiotaomicron. Key words: ulcerative colitis, intestinal flora, Lactobacillus, B. thetaiotaomicron.
INTRODUCTION
The human colon is the largest ‘microbial organ’ in the
body whose primary functions involve the dessication and
storage of faeces and water and electrolyte absorption.
Recently, great interest has been expressed in the roles
played by colonic bacteria in health and disease (1). This is
also of growing relevance to idiopathic inflammatory
bowel disease, e.g. ulcerative colitis (UC). This is a relapsing and remitting condition of unknown aetiology characterised by extensive and inappropriate inflammation of the
colonic mucosa.
It is now recognised that commensal colonic bacterial
flora not only play important roles in nutrition, metabolic
and immune functions, but that their luminal or mucosally
associated position may be influential to these functions
(2). The interaction between colonic mucosally associated
bacteria and the mucus barrier serve to limit the presence
and binding of exogenous pathogenic bacteria to the
colonic epithelium. It is these bacteria found within and
below the mucus layer coating the colonic epithelial cell
layer and in the immediate vicinity of the mucosa that are
functionally relevant and the probable source of antigenic
© Scandinavian University Press 1999. ISSN 0891-060X
challenge in UC. Much work in this area has involved the
faecal luminal flora, as this material is readily available
without need for any invasive procedure. Studies have now
demonstrated that differences exist between luminal bacteria and bacterial populations intimately associated with the
mucosal surface (3). These studies have quantified the
numbers and genera of bacteria present, but few identify
these bacteria to species level (4). These studies have also
been used to investigate the colonic flora with regard to
pharmacological regimens (5).
To date, investigations into the colonic bacterial flora in
disease have been undertaken with respect to Crohn’s
disease (6), cancer and polyp tissue (7) and UC (8). All
these studies with the exception of the UC investigations
by Poxton et al. and Hartley et al. showed few bacterial
flora changes from normal. The UC study undertaken by
Hartley et al. (8) showed alterations in the mucosally
associated flora at genus level but fell short of speciating
these differences or including a ‘normal’ control group.
The recent investigations by Poxton et al. (3) which included a control group, concentrated upon alterations
within the genus Bacteroides at species level. This study
Microbial Ecology in Health and Disease
170
S. Pathmakanthan et al.
showed changes in mucosally associated Bacteroides species. It is with these considerations in mind that we chose
to specifically examine mucosally associated bacterial flora
in acute relapse UC by unprepared flexible sigmoidoscopy.
We aimed to quantify and characterise changes to species
level for all aerobic species present and a random selection
of anaerobic flora and to compare these findings with that
of a normal control group.
obe Agar supplemented with 4% horse blood and 1%
freeze-thaw lysed horse blood for total anaerobic counts
(FAA); FAA supplemented with kanamycin 75 mg/ml and
vancomycin 2.5 mg/ml (FAAKV) for Bacteroides spp;
Rogosa agar (Unipath) for Lactobacillus spp. (RA); Reinforced Clostridial Cotton Blue Agar (RCCBA) for
Bifidobacterium.
MATERIALS AND METHODS
Quantification of both aerobic and anaerobic bacteria
employed the Miles – Misra method (9). After complete
homogenisation of biopsy, 0.5 ml of the homogenate was
transferred to 4.5 ml of brain heart infusion and subsequent preparation of 10-fold dilutions (10 − 1 –10 − 8 cfu/
ml) prepared from this concentration.
From each concentration, three separate 5 ml spot volumes were seeded on to the appropriate culture plates for
the genus under investigation. These plates were then
incubated in the appropriate conditions and colonies
counted from 5 ml spots of smallest dilution at 48 and 72
h for aerobic and anaerobic cultures, respectively.
Processing of biopsy material
Endoscopic biopsies were obtained from patients in whom
flexible sigmoidoscopy was being performed in the course
of routine clinical care (n=20). Full written consent was
obtained from patients and permission for obtaining
colonic biopsies was obtained from the hospital ethics
committee. Normal controls were obtained from patients
undergoing flexible sigmoidoscopy for investigation of
haematochezia and familial screening for rectal carcinoma
(n= 10). The control group was age and sex matched with
the colitic cohort (6 male, 4 female age 18–52 years).
Neither control group nor patients with UC had consumed
any antibiotic preparation 3 weeks prior to investigation.
Flexible sigmoidoscopy allowed examination and biopsy
of an unprepared colon without the need for a colonic
lavage or enema. This was important as earlier in 6itro
experiments in our laboratory had shown an inhibition of
bacterial mucosally associated flora in the presence of
Picolax (Ferring Pharmaceuticals, Middlesex) but not
Klean Prep or Fleet lavage preparations
All biopsies were obtained from the recto–sigmoid junction with forceps sterilised in gluteraldehyde (Cidex, Johnson and Johnson, Skipton, Yorkshire) which were rinsed
in sterile saline prior to use. Biopsies were gently washed in
sterile saline to remove adherent faecal material and immediately transferred with a sterile needle to pre-weighed
vials containing 5 ml brain heart infusion broth and 10%
glycerol transport medium. This vial was immediately
snap-frozen in liquid nitrogen. All bacterial analysis was
carried out within 1 week of storage. A representative
sample from adjacent mucosa was fixed in formalin for
routine histopathology.
Frozen vials were transported to an anaerobic cabinet
(37°C, N2 80%, H2 10%, CO2 10%) where they were
allowed to thaw. Biopsy sample and transport medium (5
ml) were transferred into a ‘Stomacher’ sterile sampling
bag (Fischer Scientific, Loughborough, Leicester) and broken up manually. The biopsy homogenate and medium
was then pipetted to ensure complete homogenisation.
Media
The following media were used: Columbia Agar (Unipath)
supplemented with 5% horse blood for total counts of
aerobic and facultative organisms (BA); Fastidious Anaer-
Quantification studies
Speciation studies
For anaerobic speciation studies, 10 ml of biopsy homogenate was transferred and spread on to BA using a
firm 10 ml loop. Following 72 h of anaerobic incubation at
37°C, up to 16 colonies were chosen at random and
subsequently re-plated anaerobically and aerobically (to
exclude facultative anerobic growth e.g. E. coli ) for single
colonies. These were once again anaerobically incubated
for 48 h at 37°C and examined for purity. Of these 16
colonies, those successfully replated and purified were subjected to enzyme hydrolysis kits for final identification.
Bacterial colonies which demonstrated aerobic growth (all
facultative anaerobic E. coli ) were subsequently speciated
and analysed aerobically.
For aerobic speciation, a similar 10 ml amount from
original biopsy homogenate was transferred and spread on
to BA for colony identification. Following 48 h of culture
in an aerobic incubator, all colonies were subsequently
replated and purified. This was possible as much lower
numbers of aerobic colonies were found relative to anaerobic growth from both normal and inflamed mucosa allowing complete identification of all colonies to species level.
All colonies undergoing species identification underwent
gram stain, tests for indole production, catalase activity
and oxidase reactions. All colonies were also cross plated
in aerobic and anaerobic environments to test for facultative growth.
Following replating and purification, anaerobic and aerobic colonies were identified with a commercially validated
technology using Anaerobic, Gram-Positive and Non-enteric fermenting enzyme hydrolysis kits (Becton-Dickinson,
Identification Systems). Briefly, this methodology consisted
of 29 dried biochemical and enzymatic substrates. A pure
Mucosally associated bacteria
bacterial suspension approximating McFarland standard
No 4 was made in the supplied inoculum fluid. These tests
were based on microbial utilisation and degradation of
specific substrates detected by indicator systems. Enzymatic hydrolysis of fluorogenic substrates containing coumarin
derivatives
of
4-methylumbelliferone
or
7-amino-4-methylcoumarin resulted in increased fluorescence that was detected with a UV light source (10).
Chromogenic substrates upon hydrolysis produced colour
changes that were detected visually with a standard light
source. These ID systems detected the ability of an organism to hydrolyse, degrade, reduce or otherwise utilise a
number of substrates with pre-formed enzymes. All reaction results were scored as per the manufacturer’s instruction and entered into a computer database for species
identification. All organism identification included recent
name changes and was based on Bergey’s manual of
determinative bacteriology (9th ed. Williams and Wilkins,
Baltimore).
171
Table II
Aerobic species that were isolated from both normal and UC
samples
Bacteria
% Isolates in normals
% Isolates in
UC
Escherichia coli
Micrococcus spp.
Klebsiella oxytoca
Gardenella 6aginalis
Pantea agglomerans
Lactococcus raffinolactis
Staphylococcus spp.
Stenotrophomonas
spp.
Bacillus spp.
Rothia dentocariosa
Acinetobacter
Fla6imonas oryzihabitans
38
19
7.1
7.1
4.8
4.8
37.5
1.1
1.1
2.3
0
1.1
4.8
4.8
4.5
0
2.4
2.4
2.4
2.4
1.1
0
6.8
2.3
Statistical analysis
Statistics was carried out on SPSS software. For quantitative studies, logged data was analysed using the non paired
t-test and the Mann Whitney U Test was used on unlogged data. In speciation studies, The x 2-test was used to
compare frequency of isolates between normal and inflamed biopsies.
RESULTS
Quantification studies
Total viable counts of aerobic and anaerobic bacteria
isolated from normal and inflamed biopsies are shown in
Table I. Average biopsy weight was 5.5 mg. No significant
differences were detected between total aerobic and anaerobic counts from normal patients (N) and those with UC.
Facultative anaerobes were included in both counts. From
20 biopsy samples, (10 N samples and 10 active UC),
Table I
Bacteroides spp. predominated in the flora of both N and
UC patients (mean: 7.4 × 108 cfu/g and 1.5× 108 cfu/g,
respectively). Bacteroides spp. comprised 46 and 42% of
the total anaerobic counts from N and UC biopsies,
respectively. Bifidobacterium was isolated in all biopsies
(mean values, N: 5.3 ×108 cfu/g; UC: 5.6 ×107 cfu/g).
Growth on Lactobacillus selective agar (N = 4.8× 108,
UC = 1.1× 106) revealed significantly less growth from
UC biopsies in comparison to Ns.
Speciation studies
Every aerobic isolate identified was speciated from N and
UC biopsies. This totalled 130 (42 normal, 88 colitic)
isolates yielding 32 different sub-species (Table II) Eighteen of these 32 sub-species were found only on UC
biopsies (Tables II and III). E. coli was the commonest
isolate grown aerobically being present in large amounts in
both N and UC patients.
Comparison of quantitati6e counts in selecti6e media between normal and UC (cfu/g)
Media
Biopsy
Mean
SD
p-Value
Total anaerobes
N
UC
N
UC
N
UC
N
UC
N
UC
1.6×109
3.6×108
7.4×108
1.5×108
5.3×108
5.6×107
4.8×108
1.1×106
6.4×107
2.6×107
92×109
95.1×108
99.6×108
92.1×108
91.5×109
97×107
91×108
92.4×106
96.4×107
92.6×106
ns
ns
ns
ns
ns
ns
ns
pB0.05
ns
ns
Bacteroides
Bifidobacterium
Lactobacillus
Total aerobes
N, normal, UC, ulcerative colitis. SD, standard deviation.
Table III
Percentage of aerobic isolates found only in UC samples
Aerobic bacteria only isolated in UC
biopsies
% Isolated in
UC
Streptococcus spp.
Sphingomonas paucimobilis
Corynebacterium spp.
Chromobacter 6iolaceum
Pseudomonas spp.
Corynebacterium aquaticum
Oersko6ia spp.
Enterococcus faecium
Enterobacter agglomerans
8
6.8
5.7
4.5
3.4
3.4
3.4
2.3
2.3
172
S. Pathmakanthan et al.
DISCUSSION
Table IV
Percentage of Bacteroides isolates; UC 6s normal
% Isolates in normals % Isolates in UC
Total Bacteroides
B. 6ulgatus
B. splanchnicus
B. uniformis
B. o6atus
B. distasonis
B. thetaiotaomicron
B. fragilis
B. ureolyticus
B. gracilis
B. capillosus
B. caccae
B. praecutus
55.7
31.3
15.6 *
14.1 *
9.4
9.4
7.8
57.7
39.3
1.8
1.8
8.9
5.4
32.1 **
4.7
3.1
3.1
1.6
0
0
1.8
1.8
0
1.8
3.6
1.8
** pB0.01, * pB0.05.
In anaerobic speciation, from the 20 biopsy samples (10
N, 10 UC), 212 isolates (115 N, 97 UC) were successfully
purified and identified to represent 41 different sub-species
(Tables IV and V). For any individual patient there was a
mean 6.7 different anaerobic species in N and 4.7 in UC
samples. The principle species were Bacteroides, which
were present in every biopsy (Table IV). For any individual patient only a few species of Bacteroides (mean: N =
3.1, UC =2.4; range: N = 1–5, UC =1–4) were isolated.
Bacteroides thetaiotaomicron and 6ulgatus were the highest
total number of anaerobic isolates identified. B. 6ulgatus
was isolated in near equal frequency from normal and
inflamed mucosa isolated in 8/10 N patients and 6/10 UC
patients. B. thetaiotaomicron was isolated from 8/10 biopsy
samples of UC patients with active inflammation but in
only 4/10 of N biopsies.
Table V
Percentage isolation of other anaerobic bacteria cultured from
normal and UC biopsies (no significant differences)
Anaerobic bacteria
Frequency in normal Frequency in UC
(%)
(%)
Bifidobacterium
spp.
Clostridium spp.
Actinomyces spp.
Lactobacillus spp.
Peptococcus spp.
Eubacterium spp.
Pre6otella spp.
Fusobacterium
spp.
Mobiluncus spp.
Propinibacterium
spp.
Tissierella spp.
13
7.2
9.6
6.1
5.2
4.3
1.7
1.7
0.9
5.2
4.1
8.2
3.1
0
1
3.1
0.9
0.9
5.2
2.1
0.9
1
The mucosa of the gastrointestinal tract separates 1013
eukaryotic cells from 1014 bacteria with 96 – 99% of this
resident bacterial flora consisting of anaerobes and 1 –4%
facultative aerobes (gram negative Coliform bacteria, Enterococci, and Proteus, Pseudomonas, Lactobacilli, Candida and other organisms). The predominant cultivable
bacteria belong to the genus Bacteroides which accounts
for 20 – 30% of the species isolated. This genus constitutes
B. fragilis, B. distasonis, B. o6atus, B. thetaiotaomicron, B.
6ulgatus, B. eggerthii, B. uniformis, B. caccae, B. merdae
and B. stercoris. B. 6ulgatus is considered the commonest
species in this genus and although B. fragilis is less common it is the most prevalent in clinical specimens of blood
cultures and wounds (11).
Many of the bacteria in the GI tract especially the gram
negative rods are opportunist pathogens capable of eliciting injurious, pro-inflammatory responses on interaction
with the systemic immune system. It is therefore important
to know which species in quality and number are in the
most intimate association with the mucosa.
UC is a disease where genetic and environmental factors
interrelate to induce chronic spontaneously relapsing intestinal inflammation. The clinical features of UC in relapse
are similar to those of dysentery and other infectious
colitides with luminal bacteria being a pre-requisite for
disease development (12).
Bacteria are a prerequisite for the development of UC in
a susceptible individual and this source of antigen is likely
to originate within the mucus sheet adhering to the colonic
epithelial cell. This mucus sheet is abnormal in colitics and
has been shown to express receptors not usually exposed
leading to the potential cultivation of certain potentially
pathogenic organisms (13).
We avoided the use of any colonic preparation prior to
biopsy sampling as preparation for colonoscopy in the
form of purgatives, laxatives, anti-microbial agents and
enemas has the potential to influence the colonic flora.
This was confirmed by earlier in 6itro experiments in our
laboratory showing an inhibition of bacterial growth with
certain preparations.
An earlier study by Poxton et al. had shown that
mucosally associated flora of the rectum and the proximal
colon are similar indicating that information may be obtained from rectal or sigmoid biopsy samples within range
of a flexible or rigid sigmoidoscope (8). The recto – sigmoid
junction was especially chosen to avoid any effect rectal
steroid or aminosalicylate preparations might have on the
rectal mucosa. During flexible sigmoidoscopy it was noted
that rectal sparing was observed in some patients who
were on steroid enema preparations which may have influenced the quality and quantity of the mucosal flora in
biopsies if they were obtained from the rectum.
Mucosally associated bacteria
All UC patients in this study were also on various
preparations of maintenance sulphasalazine. An earlier
study involving this drug has concluded that it has little
effect on the bacteria colonising the colorectal mucosa (5).
The first part of this paper set out to determine quantitative differences between normal and colitic patients. The
significantly low numbers of Lactobaccillus in inflamed UC
biopsies is an important finding. The reduction in numbers
of this particular genus has been reported previously both
in an animal model and in biopsy samples of patients with
UC (14, 15). Lactic acid bacteria constitute an integral
part of the healthy gastrointestinal micro-ecology and are
involved in host metabolism with nutritional and therapeutic benefits (16). Their beneficial influence in gastrointestinal disorders and the recent isolation of a broad spectrum
anti-microbial substance from some Lactobacilli species
demonstrate their protective and stabilising actions in the
gastrointestinal tract (17, 18).
These quantitative studies also revealed a significant
aerobic population resident in mucosally associated flora
with aerobic:anerobic ratios of 0.07 and 0.04 in N and UC
biopsies, respectively. These ratios are much higher than
would be found in faecal samples indicating that aerobic
bacteria may have a greater role to play in the homeostasis
of this environmental niche.
It is important to emphasise that all aerobic growth of
biopsies from N and UC biopsies were speciated. These
data illustrate a clear increase in range of aerobic organisms found in UC. Although it is not possible to interpret
this as a primary or secondary effect, it does illustrate an
increase in potentially pathogenic organisms which probably play at least a perpetuating role in the disease process.
The sixteen anaerobic isolates from each biopsy sample
for speciation were a random representation of the total
number of isolates grown. We chose to cultivate anaerobic
isolates on non selective blood agar to enable a cross
sectional representation of anaerobic flora to be sampled
for speciation. Despite the large number of species present
in the colon with the potential to colonise the mucosally
associated environment, there is clear predominance of
Bacteroides spp. mainly 6ulgatus and thetaiotaomicron in
N and UC biopsies. Interestingly, B. thetaiotaomicron was
more commonly found in acute UC biopsies perhaps
indicating that this organism may have a role to play in
disease pathogenesis. Similar findings were found in a
recent investigation by Poxton et al. (3).
It is tempting to speculate that a lack of lactobacillus
found in UC biopsies shown in the first set of quantitative
data may allow a combination of B. thetaiotaomicron and
potentially pathogenic aerobic species to alter the characteristics of the bacterial environment adjacent to the mucosa towards one of an increasingly pro-inflammatory
nature.
These data warrant further investigation into the protective nature of Lactobacilli in the human colon and the
173
effects of this genus of flora in maintaining intestinal
homeostasis. The potentially pathogenic nature of B.
thetaiotaomicron is unlikely to be found in its lipopolysaccharide as it has previously been shown to be of low
pathogenicity (19).
In conclusion, we have found significant quantitative
and qualitative differences in the mucosally associated
flora of N individuals and patients with UC. The significant decrease in Lactobacillus spp. in patients with UC as
well as the increase in isolation frequency of B. thetaiotaomicron and aerobic species illustrate clear quantitative
and qualitative differences which may contribute to the
pathogenesis of this disease.
REFERENCES
1. Bengmark S. Immunonutrition: Role of biosurfactants, fiber
and probiotic bacteria. Nutrition 1998; 14: 585 – 94.
2. Peach SL, Tabaqchali S. Mucosa associated flora of the
human gastrointestinal tract in health and disease. Eur J
Chem Antibiot 1982; 2: 41 – 50.
3. Poxton IR, Brown R, Sawyer A, Ferguson A. Mucosa associated bacterial flora of the human colon. J Med Microbiol
1997; 46: 85 – 91.
4. Croucher SC, Houston AP, Bayliss CE, Turner RJ. Bacterial
populations associated with different regions of the human
colon wall. Appl Environ Microbiol 1983; 45: 1025 – 33.
5. Hartley MG, Hudson MJ, Swarbrick ET, Grace RH, Gents
AE, Hellier MD. Sulphasalazine treatment and the colorectal
mucosa-associated flora in ulcerative colitis. Aliment Pharmacol Ther 1996; 10: 157 – 63.
6. Hudson MJ, Hill MJ, Elliott PR, Berghouse LM, Burnham
WR, Lennard Jones JE. The microbial flora of the rectal
mucosa and faeces of patients with Crohn’s disease before
and during antimicrobial chemotherapy. J Med Microbiol
1984; 18: 335 – 45.
7. Edmiston CE, Avant GR, Wilson FA. Anaerobic bacterial
populations on normal and diseased human biopsy tissue
obtained at colonoscopy. Appl Environ Microbiol 1982; 43:
1173 – 81.
8. Hartley GM, Hudson MJ, Swarbrick ET, Hill MJ, Gent AE,
Hellier MD, Grace RH. The rectal mucosa-associated microflora in patients with ulcerative colitis. J Med Microbiol
1992; 36: 96 – 1039.
9. Miles AA, Misra SS, Irwin JO. The estimation of the bactericidal power of the blood. J Hyg 1938; 38: 732 – 49.
10. Manafi M, Kneifel W, Bascomb S. Fluorogenic and chromogenic substrates used in bacterial diagnostics. Microbiol
Rev 1991; 55: 335 – 48.
11. Duerden BI, Drasar BS, eds. Anaerobes in human disease.
London: Edward Arnold: 1991
12. Shorter RG. Recent developments in non-specific inflammatory bowel disease. N Eng J Med 1982; 306: 837 – 48.
13. Podolsky DK, Isselbacher KJ. Comparison of human colonic
mucin: selective alteration in inflammatory bowel disease. J
Clin Invest 1983; 72: 142 – 53.
14. Fabia R, Ar’Rajab A, Johansson ML, Andersson R, Willen
R, Jeppsson B, Molin G, Bengmark S. Impairment of bacterial flora in human ulcerative colitis and experimental colitis
in the rat. Digestion 1993; 54: 248 – 55.
15. Molin G, Jeppsson B, Johansson ML, Ahrne S, Noboek S,
Stahl M, Bengmark S. Numerical taxonomy of Lactobacillus
spp. associated to healthy and diseased mucosa of the human
intestines. J Appl Bacteriol 1993; 74: 314 – 23.
174
S. Pathmakanthan et al.
16. Bengmark S. Ecological control of the gastrointestinal tract.
The role of probiotic flora. Gut 1998; 42: 2–7.
17. Reid G, Bruce AW, McGroarty J, Cheng KJ, Costerton
WJ. Is there a role for Lactobacilli in prevention of urogenital
and intestinal infections? Clin Microbiol Rev 1990; 3:
335 – 44.
.
18. Axelsson LT, Chung TC, Dobrogusz WJ, Lindgren SE. Production of a broad spectrum antimicrobial substance by Lactobacillus reuteri. Micro Ecol Health Dis 1989; 2: 131 – 6.
19. Delahooke DM, Barclay GR, Poxton IR. A re-appraisal of
the biological activity of Bacteroides lipopolysaccharide. J
Med Microbiol 1995; 42: 102 – 12.