Molecular characterization of the bacteria adherent to human colorectal mucosa

Authors


K.D. Bruce, Department of Life Sciences, Franklin Wilkins Building, 150 Stamford Street, King's College London, SE1 9NH, UK.
E-mail: kenneth.bruce@kcl.ac.uk.

Abstract

Aims:  To study large intestinal mucosal bacterial communities by Denaturing Gradient Gel Electrophoresis (DGGE) profiling and sequencing of 16S rRNA gene polymerase chain reaction (PCR) products amplified from DNA extracted from colorectal biopsies taken from healthy individuals. The specific aims were to determine how similar the mucosa-associated bacterial communities are within and between individuals and also to characterize the phylogenetic origin of isolated DGGE bands.

Methods and Results:  Human colorectal biopsies were taken at routine colonoscopy from 33 patients with normal looking mucosa. The DNA was extracted directly from single biopsies and the bacterial 16S rDNA PCR amplified. The PCR products were profiled using DGGE to generate a fingerprint of the dominant members of the bacterial community associated with the biopsy. The reproducibility of this method was high (>98%). Washed and unwashed biopsies gave similar DGGE banding patterns (Median Similarity Coefficient – MSC 96%, InterQuartile Range – IQR 3·0%, n = 5). Adjacent biopsies sampled from the same patient using different forceps gave similar DGGE profiles (MSC 94%, n = 2). Two colorectal biopsies sampled at locations 2–5 cm apart, from each of 18 patients, resulted in very similar profiles (MSC 100%, IQR 2·8%). Biopsies sampled from different locations within the large intestine of the same patient also gave similar DGGE profiles (MSC 98% IQR 3·3%n = 6). Although all patients (n = 33) gave different DGGE profiles, some similarity (c. 34%) was observed between profiles obtained from 15 patients arbitrarily selected. 35 DGGE bands were excised and sequenced. Many were found to be most closely related to uncultured bacterial sequence entries in the Genbank database. Others belonged to typical gut bacterial genera including Bacteroides, Ruminococcus, Faecalibacterium and Clostridium.

Conclusions:  Bacterial communities adherent to colorectal mucosa within a normal patient show little variation; in contrast, mucosal bacterial communities sampled from different patients with normal colorectal mucosa show a high degree of variation.

Significance and Impact of the Study:  This research demonstrates that DGGE profiling of 16S rRNA gene PCR products amplified from DNA extracted directly from mucosal samples offers fresh insight into the bacterial communities that are adherent to colorectal mucosa. These findings are important with respect to further studies on the gastrointestinal tract in health and disease.

Introduction

The microbial community within the human gastrointestinal tract is complex and diverse (Frank and Pace 2001). This microflora plays important roles in maintaining human health, for example by aiding the digestion of food and absorption of nutrients, producing essential vitamins and metabolising exogenous compounds (Guarner and Malagelada 2003). In addition, this flora plays a vital role in the development and function of an effective and competent immune system (Umesaki and Setoyama 2000; Heller and Duchmann 2003). Abnormalities or changes in the gastrointestinal microbial community may contribute to the pathogenesis of a range of diseases including inflammatory bowel disease (IBD), colon cancer and multi-system organ failure (Campieri and Gionchetti 2001; Shanahan 2002; Guarner and Malagelada 2003). Knowledge of the composition, activity and dynamics of the human gastrointestinal microbial flora is therefore fundamental if we are to be able to define better the roles that microbes play in host health and disease.

Previous assessment of the composition of gut bacterial flora has tended to rely on analysis of faecal samples (Wilson and Blitchington 1996; Tannock et al. 2000; Marteau et al. 2001). However, faecal bacterial communities differ substantially from the bacteria that are adherent to intestinal mucosa and may be much less important in influencing gut mucosal interactions and immunity (Pryde et al. 1999; Zoetendal et al. 2002; Lepage et al. 2005). The traditional microbiological techniques based on cultivation by which the gut flora has typically been studied previously may also have given an incomplete result. Culture-independent methods can provide data that are more representative of the total bacterial community than methods that rely on cultivation prior to analysis (Amann et al. 1995; Pryde et al. 1999; Suau et al. 1999). In these culture-independent methods, the polymerase chain reaction (PCR) is typically used to amplify phylogenetically informative ribosomal gene regions from bacterial nucleic acids extracted directly from samples. These PCR products can in turn be separated using Denaturing Gradient Gel Electrophoresis (DGGE) to form a profile of the dominant members of the bacterial community (Muyzer et al. 1993). This is sensitive, with bacterial species at the level of 1% within the total community being detectable (Amann et al. 1995; Murray et al. 1996; Pryde et al. 1999).

By comparing the presence and absence of bands in the generated DGGE profiles in a pairwise manner (Bruce et al. 2000; Leung and Topp 2001), estimates of the extent of similarity between different bacterial communities can be made. Moreover, through the sequence analysis of individual DGGE bands excised from the gel matrix, species identification can be made through comparison with already extensive sequence databases (Bruce et al. 2000; Leung and Topp 2001). Zoetendal et al. (2002) applied a DGGE profiling approach to characterize the bacterial community in DNA extracted directly from samples of colonic mucosa. This study was important because it showed that such analyses were possible for the bacterial community associated with mucosal samples. However, before studies can be performed on bowel diseases, it is important to characterize the extent of variation within and between different individuals and to identify the dominant members of the bacterial community in samples obtained from healthy patients with normal colorectal mucosa.

In this study, we characterized the bacterial community associated with colorectal biopsies taken from patients with normal looking mucosa. Further, the impact on DGGE profiles by washing individual gut mucosal biopsies prior to analysis and sampling from within the same patient using different sets of forceps was also studied. The degree of similarity between the adherent bacterial communities in biopsies from the same patient, taken either close together or in different regions of the large intestine, was determined and DGGE profiles from different patients compared. Finally, bacterial species were identified by sequencing individual bands from DGGE gels and comparing data to sequence databases.

Material and methods

Patients

Colorectal biopsies were collected from 33 patients undergoing routine colonoscopy for a range of indications including rectal bleeding, iron deficiency anaemia, abdominal pain, change of bowel habit and polyp follow up. Twenty-one patients had entirely normal examinations; 12 had sigmoid diverticulosis and/or adenomatous polyps distant from the site of biopsy. No patient had received antibiotic therapy within 8 weeks of their colonoscopy.

Biopsy collection and treatment of samples

Patients were prepared for colonoscopy with two Picolax (sodium picosulphate with magnesium citrate) sachets and one tablet (5 mg) of Bisacodyl. For 3 days before the procedure, they took a low residue diet. At colonoscopy, between two and eight colorectal biopsies beyond those needed for routine diagnostic purposes were taken. For certain samples required in this study, these were taken at sites 2–5 cm apart. In other cases, samples were taken at different locations of the large intestine (rectum, sigmoid colon, descending colon, transverse colon, and caecum). All biopsies were taken using standard 5 mm gape forceps.

Each biopsy specimen, of mass between 3 and 10 mg, was removed from the biopsy forceps using a blunted sterile needle and placed directly into an individually prepared tube. Tubes were prepared by adding a 3 mm diameter tungsten carbide bead (Qiagen, West Sussex, UK) and 75 mg of 0·17–0·18 mm diameter glass beads (B. Braun, Biotech International GmbH, Melsungen, Germany) to a 2·0 ml Safe-lockTM microcentrifuge tube (Eppendorf AG, Hamburg, Germany). The beads were gamma irradiated with a dose ≥ 25 kGy (Isotron, Swindon, UK). Samples were snap-frozen in liquid nitrogen and stored at −70°C prior to total nucleic acid isolation.

To investigate the effects of washing the biopsies prior to analysis, biopsies (n = 4) were taken from adjacent locations in the rectum of five patients. Two of the biopsies were processed as above, and the other two were washed prior to being snap-frozen. Washing involved placing the biopsies initially into separate sterile tubes containing sterile 0·9% saline and repeatedly inverting; they were then transferred to the prepared tubes and snap frozen. To examine the effect on DGGE profiles of using different forceps to take samples from the same patient, adjacent biopsies from the same patient were taken using separate forceps.

DNA isolation

Nucleic acids were extracted from biopsies using a method modified from the manufacturer's guidelines for the MasterPureTM Complete DNA and RNA Purification Kit (Epicentre, Madison, USA). 400 μl of tissue and cell lysis solution (Epicentre) was added to each sample tube and the samples homogenized mechanically using a RetschTM MM300 Mixer Mill (Qiagen) for 40 s at 25 Hz. Following centrifugation at room temperature at 11 500 g for 5 min, 3 μl of proteinase K (50 μμl−1) (Epicentre) was added and the tubes mixed by inversion. Samples were incubated at 65°C for 20 min with mixing every 5 min. Samples were held on ice for 5 min and 200 μl of protein precipitation reagent added (Epicentre). Following mixing by brief vortexing, samples were centrifuged at 11 500 g for 10 min and the supernate transferred to a fresh micro-centrifuge tube. 0·9 volumes of molecular grade isopropanol (BDH, Dorset, UK) was added to the supernatant and the samples inverted for 1 min. Samples were held at −20°C for 1 h. After centrifugation at 11 500 g for 10 min at 4°C, the supernate was removed and the DNA pellet was washed twice with 75% ethanol [100% molecular grade ethanol (BDH) 3 : 1 molecular grade water (Eppendorf)]. DNA was resuspended in 50 μl molecular grade water (Eppendorf) and stored at 4°C overnight before use. On completion of analysis, the remainder of DNA template was stored at −20°C. The extracted DNA was verified by TAE-agarose gel electrophoresis with UV visualization of 1·5% (w/v) TAE agarose gels stained using ethidium bromide (0·5 mg l−1). Images were captured using a Herolab image analyser with e.a.s.y. win 32 software (Herolab, Wiesloch, Germany).

DNA quantification

PicoGreen® dsDNA quantitation kit (Molecular Probes, Leiden, Netherlands) was used to quantify the amount of double-stranded DNA (dsDNA) extracted from the biopsy samples, following the manufacturer's guidelines. Typically 1036 ng (SD ± 5%, n = 4) of dsDNA was extracted per mg of biopsy material (data not shown).

PCR amplification for DGGE

Bacterial 16S rRNA gene regions were amplified from the total nucleic acids extracted by PCR amplification using universal bacterial primers (synthesized and PAGE purified by Invitrogen, Paisley, UK), targeting the bacterial V3 16S rDNA region, as described previously (Muyzer et al. 1993), where primer 2 was (5′ATTACCGCGGCTGCTGG) and primer 3 was (5′CGCCCGCCGCGCGCGGCGGGCGGGGCGGGGGCACGGGGGGCCTACGGGAGGCAGCAG). A 40 base GC clamp region on primer 3 is used to facilitate DGGE analyses. Each 100 μl PCR reaction consisted of both primers at 0·2 μM per PCR reaction, 50 μl Hotstart Taq Master Mix (a premixed solution containing Taq DNA Polymerase, PCR Buffer, and dNTPs. The solution provides a final concentration of 1·5 mM MgCl2 and 200 μM of each dNTP) (Qiagen), and either template DNA or water, as a negative control, and water made up to the final volume. DNA extracted from biopsy samples was used as the DNA template in these reactions. The amount of template typically added to each PCR was 600 ng. The cycling conditions used to amplify the 16S rRNA gene fragments were 95°C for 15 min (Taq activation), followed by 30 cycles of 94°C for 30 s, 55°C for 30 s and 72°C for 1 min. A final extension period of 72°C for 10 min was used. All reactions were carried out using a GeneAmp PCR system (Applied Biosystems, Cheshire, UK). PCR products, approximately 233 base pairs in length, were visualized by UV prior to DGGE, as described above. The concentration of the PCR products was calculated using the e.a.s.y. win 32 software (Herolab) by comparison to a 100 base pair DNA marker of known mass (Invitrogen).

DGGE of PCR amplicons

16S rRNA gene PCR products were separated using a D-Code 16/16 mm gel system (Bio-Rad, Hertfordshire, UK). The DGGE analysis was performed in accordance with the manufacturer's guidelines. Reagents used for DGGE were purchased from Bio-Rad. 80 ng of PCR product was loaded into each well of a 1·0 mm polyacrylamide gel, 8% (w/v). A denaturing gradient ranging between 15 and 55% was formed, where 100% denaturant is defined as 7 M urea and 40% (v/v) formamide (Muyzer et al. 1993). Gels were electrophoresed at 60°C at 180 V for 3 h then stained for 20 min using 1 × SYBR gold (Molecular Probes, Paisley, UK). Gel images were captured using a Herolab image analyser.

DGGE image analysis

DGGE gel images were analysed using phoretix 1d advanced software, Version 5·0 (Nonlinear Dynamics, Newcastle upon Tyne, UK). Dendrograms, showing clustering analysis, for comparisons of DGGE profiles and the resulting similarity coefficients were generated by the unweighted pairwise grouping method with mathematical averages (UPMGA), using the Dice coefficient of similarity (Bruce et al. 2000). For these analyses, PCR products from a single biopsy were electrophoresed in five lanes through the gel to act as an internal marker, to generate Rf lines both for estimating similarity and to allow the matching of bands from different patients. The use of internal markers however meant that only a maximum of 15 samples could be analysed at one time using this computer software program.

Sequencing and analysis of individual DGGE bands

Individual bands were excised from the stained DGGE gels and the sections of polyacrylamide gel were left to diffuse passively overnight in 50 μl molecular grade water at 4°C (Bruce et al. 2000). The resultant solution generated for each excised band was used as template to generate PCR products of the band to be sequenced using the same PCR reaction components and cycling conditions described above, with the exception that the cycle number was reduced to 28 cycles. In every case, DGGE analysis confirmed that these PCR products gave a DGGE band that was electrophoretically indistinguishable from the position of the band that had been excised from the original mixed community profile. The verified PCR products were purified using Qiagen QIAquick PCR purification Kit (Qiagen) following the manufacturer's guidelines prior to being sequenced (Macrogen Inc., Seoul, Korea). The sequences were analysed at the National Center for Biotechnology Information (NCBI) using blast® software (Altschul et al. 1990) and chimeric sequences were checked for using check_chimera (Larsen et al. 1993).

Nucleotide sequence accession numbers

Sequences presented in this study have been deposited in the GenBank database under the following accession numbers; AJ634468 to AJ634472, AJ634752 to AJ634762 and AJ862345 to AJ862376.

Ethical considerations

This study was approved by the East London and City Health Authority Ethics Committee and all patients gave informed consent for additional biopsies to be taken for the purpose of this study.

Results

Reproducibility of DGGE profiling and PCR amplification from single biopsies

DNA was isolated from large intestinal biopsies from 33 patients. The V3 region of the 16S rRNA gene was PCR-amplified from every sample with DGGE profiles formed in turn. To confirm that the PCR and DGGE process was reproducible, nucleic acid extracts from individual biopsy samples from different patients were repeatedly PCR amplified and the predominant bacterial community profiled using DGGE. The 16S rRNA gene PCR product DGGE profiles from the same PCR and independent PCRs of the same template were highly similar (Fig. 1). DGGE profiles (n = 4) from single PCRs (n = 3) gave a Median Similarity Coefficient – MSC of 100% (InterQuartile Range – IQR 1·0%). Independent PCR amplifications (n = 4) of the same template from two different patients (n = 2) gave DGGE profiles that had a MSC of 98·5% (IQR 1·5%).

Figure 1.

SYBR gold stained DGGE gel showing profiles formed from 16S rRNA gene PCR products amplified from DNA extracted directly from a single colorectal biopsy (here, this biopsy was taken from the rectum). Lanes 1–4, PCR products from a single PCR. Lanes 5–7, PCR products from independent PCRs of the same template. Negative gel image.

Effect of washing biopsies prior to analysis

16S rRNA gene PCR product DGGE banding patterns for washed and unwashed biopsy samples were found to be very similar (MSC 96%, IQR 3·0%, n = 5) (Fig. 2).

Figure 2.

SYBR gold stained DGGE gel showing profiles formed from 16S rRNA gene PCR products amplified from DNA extracted directly from colorectal biopsies (here, these biopsies were taken from the rectum). Four biopsies were taken from adjacent locations in the same patient, two were washed (lanes 1 and 2), two unwashed (lanes 3 and 4). Negative gel image.

Effect of using different forceps to take adjacent biopsies

Two pairs of samples were taken from the same patient, from different regions, using different forceps. DGGE profiles of pairs of adjacent biopsies taken from the same patient using separate forceps were very similar. The similarity coefficient of the DGGE profiles of adjacent biopsies for both pairs was 94%.

DGGE of adjacent colorectal biopsies

The variability of the dominant bacterial community in single locations of large intestinal mucosa was investigated by taking two biopsies 2–5 cm apart from 18 patients. The resulting 16S rRNA gene PCR product DGGE profiles from the same patient were very similar (MSC 100%, IQR 2·8%, n = 18) (Fig. 3).

Figure 3.

SYBR gold stained DGGE gel showing profiles formed from 16S rRNA gene PCR products amplified from DNA extracted directly from paired colorectal biopsies. Paired biopsies (lanes 1 and 2, lanes 3 and 4, lanes 5 and 6) were taken from adjacent rectal locations in each of the three patients. Arrows indicate sequenced bands, details given in Table 1. Negative gel image.

Comparison of bacterial communities adherent to different regions of the large intestine

Biopsies taken at different sites of the large intestine (rectum, sigmoid colon, descending colon, transverse colon and caecum) of the same patient had very similar adherent bacterial communities (Fig. 4), similarity coefficients ranging from 96 to 100% (median 98%, IQR 3·3%, n =6).

Figure 4.

SYBR gold stained DGGE gel showing profiles formed from 16S rRNA gene PCR products amplified from DNA extracted directly from biopsies taken at different regions of the large intestine from the same patient. Lanes 1–5, rectum, sigmoid colon, descending colon, transverse colon, and caecum respectively. Negative gel image.

Comparison of bacterial communities adherent to colorectal mucosa of different patients

In contrast to the close similarity seen between the 16S rRNA gene PCR product DGGE profiles from biopsies from the same patient, all 33 patients biopsied gave different profiles (data not shown). DGGE profiles of the 16S rRNA gene PCR products amplified from 15 arbitrarily selected patients were compared. At least 34% similarity in banding patterns was observed for these patients (Fig. 5).

Figure 5.

An UPGMA dendrogram generated from data obtained from DGGE banding patterns of the bacterial community adherent to rectal mucosal biopsy samples taken from 15 patients. DGGE profiles stained by SYBR-gold, negative image.

Bacterial identification

Thirty-five arbitrarily selected 16S rRNA gene PCR product DGGE bands amplified from nine patients were excised, sequenced and matched with entries in the GenBank database (Figs. 3 and 6, Table 1). No evidence for chimeric sequences was identified using CHECK_ CHIMERA. Sequences were characterized in six instances where bands were observed to be of indistinguishable electrophoretic mobility in different lanes. The mean sequence similarity for these bands was 99·5% (SD 0·8, n = 6) (Table 2). As a further control, seven pairs of duplicate bands, taken from separate DGGE separations of PCR products amplified from the same template DNA, were found to be 100% similar at the nucleotide level (data not shown).

Figure 6.

SYBR gold stained DGGE gel showing profiles formed from 16S rRNA gene PCR products amplified from DNA extracted directly from rectal biopsies taken from six individuals. Arrows indicate sequenced bands, details given in Table 1. Negative gel image.

Table 1.  Closest nucleotide match (reported as species identified, accession number and percentage similarity) obtained by blast analysis of the DGGE bands sequenced
BandBase pairsMatchAccession no.Identity (%)
  1. DGGE, denaturing gradient gel electrophoresis.

  2. See Figs. 3 and 6 for band location.

A1114Uncultured bacteriumAY457885 99·1
A298Uncultured bacteriumAY563455100
A3103Faecalibacterium prausnitziiX85022 99·0
B1122Bacteroides uniformisAB050110100
B2154B. uniformisAB050110100
B380Clostridium spp.AF126687100
B4108Clostridium spp.Y10028 98·2
C1121Bacteroides vulgatusAB050111 98·4
C2101Ruminococcus gnavusX94967100
C3111F. prausnitziiX85022100
D1113Uncultured bacteriumAF499894100
D2112Uncultured bacteriumAF530331 99·1
D3100Uncultured bacteriumAB185618100
D4103R. gnavusX94967100
E177Bacteroides thetaiotaomicronAE016937100
E2136Uncultured Bacteroides sp.AB064828100
E3107Uncultured rumen bacteriumAB185618100
E4112R. gnavusX94967100
E5144Escherichia coliU00096100
F1125Uncultured bacteriumAJ408990 99·2
F2128Uncultured bacteriumAJ408988100
F3110F. prausnitziiX85022 99·1
F485Sutterella canisAJ566849100
G1120B. thetaiotaomicronAE016936100
G2128Uncultured bacteriumAJ408988100
G3120Uncultured bacteriumAF132239100
G4113Akkermansia muciniphilaAY271254 99·1
G5105F. prausnitziiAY169429 97·1
H1127B. vulgatusAB050111100
H2104Brachyspira aalborgiZ22781100
H3113B. aalborgiZ22781100
H490F. prausnitziiX85022100
I1100B. thetaiotaomicronAE016931 98·0
I280Clostridium fusiformisAF028349100
I3110F. prausnitziiX85022100
Table 2.  The comparison of DGGE band sequences from bands that were of indistinguishable electrophoretic mobility, taken from different profiles
Band combinationSequence similarity (%)
  1. DGGE, denaturing gradient gel electrophoresis.

  2. See Figs. 3 and 6 for band locations.

A1 and C1 98
E1 and F1100
A3 and C3 99
D3 and E3100
F3, H4 and I3100
D4 and E4100

In terms of species identification, bands A1, A2, D1, D2, D3, E3, F1, F2, G2 and G3 (Figs. 3 and 6) were most closely related to uncultured bacterial sequences in the Genbank database, seven ‘uncultured’ bacterial matches were found for species within the genus Bacteroides and three were found to species within the genus Clostridium. Bacteroides spp. were the closest named matches for bands B1, B2, C1, E1, G1, H1 and I1; the remaining bands gave closest named matches to species within the Ruminococcus, Faecalibacterium, Clostridium, Sutterella, Akkermansia, Brachyspira and Escherichia genera (Table 1).

Discussion

The study of the interactions between the host and microbial communities adherent to human large intestinal mucosa is increasingly recognized as important. The central aim of this research was to examine these communities in patients who had normal-looking mucosa at colonoscopy; focusing on the extent to which these communities were similar within and between patients and the characterization of species that were present in the DGGE profiles generated. Carrying out a thorough analysis of these communities in healthy individuals is essential for subsequent studies in patients with gastrointestinal disease. In order to test the above aims, this study used a culture-independent method to investigate the bacterial community associated with single gut biopsies. We examined the possibility of sampling or other methodological error influencing our findings and demonstrated the reproducibility of our methods. Additionally we characterized the phylogenetic affiliation of a number of DGGE bands present in the bacterial community profiles associated with the colorectal mucosa.

Before carrying out detailed analyses of the adherent bacterial flora, we validated the methodological approach. DNA was extracted successfully from every gut mucosal biopsy studied here. The current focus was not the relative contribution of cells either of human or microbial origin to the DNA extracted, rather the ability to amplify PCR products using oligonucleotide primers specific for the Domain Bacteria (Muyzer et al. 1993). Analysis of the sequence data, discussed below, derived for every individual DGGE band examined confirmed that all were of bacterial origin. PCR products and subsequent DGGE profiles were moreover generated using these primers for every DNA extract processed. This demonstrated that the methodology of DNA extraction and PCR amplification used here was robust, reliable and sensitive, especially given the small quantity (c. 5 mg) of tissue in each biopsy. The reproducibility of the PCR and DGGE profiling approach was also confirmed as PCR product DGGE profiles from the same PCR and independent PCRs of the same mucosal biopsy DNA template were virtually indistinguishable.

The biopsies used in this study were taken from patients who had undergone bowel preparation prior to colonoscopy. By so doing, the sampling of the prepared mucosa supports the contention that these bacteria are truly associated with the gut mucosa and as such are not merely transient. More informatively, the levels of community similarity remained high when DGGE profiles were analysed from biopsies taken from the same subject that either had, or had not, been additionally washed prior to analysis. These results further imply that the bacteria being studied were either strongly adherent to, or present within, the biopsy. This therefore suggests that this method is suitable for the investigation of the mucosa-associated flora of the large intestine. Similarly, little variation was observed in the DGGE profiles generated from biopsies taken from a patient sampled in the same region using different forceps. The implication of this is that any ‘carry over’ artefact from biopsy forceps used to take multiple biopsies is insignificant in the generation of DGGE profiles.

The DGGE profiles of the bacterial community in duplicate colorectal biopsies sampled 2–5 cm apart were shown to be highly similar. Such a stable system gives confidence in the data obtained when using this procedure; it also means that only a single biopsy is required to obtain a representative DGGE profile of the dominant bacterial community associated with the colorectal mucosa.

To date, little work has been done on the regional variation along the human large bowel of the adherent mucosal bacterial flora. Zoetendal et al. (2002) studied the bacterial community using 16S rRNA gene DGGE profiling for nucleic acids extracted directly from single biopsies sampled from different sites from the same patient. They found a high level of similarity (median 96·7%, IQR 12·5%) for the pairwise comparison of DGGE profiles for samples taken from the ascending and descending colon from the five normal patients that they examined. In the aforementioned study, no estimate of the impact of methodology on the relatively large variability in the IQR results obtained could be made, as replicate biopsies from the same region were not analysed. Recently, Lepage et al. (2005) studied the bacterial communities associated with various sites of the gastrointestinal mucosa by using temporal temperature gradient gel electrophoresis, where the primary focus of this study was the analysis of samples collected from individuals suffering from IBDs. They also examined a small number of patients regarded as ‘controls’, and found that these four individuals had highly similar bacterial communities associated with different regions of gut mucosa. By extending the locations and patient numbers studied compared to these earlier works, we feel more confident that virtually indistinguishable bacterial communities are associated with the full length of the large intestine within the same patient. These data are surprising as it suggests the adherent gastrointestinal mucosal bacterial community is uniform along the length of the colon irrespective of differences in the luminal nutrients and pH from the caecum to the rectum (Macfarlane et al. 1992). However, it is possible that the bacterial numbers could vary at different locations along the colon. Also, it is possible that the types of bacteria that are active differ markedly in different locations.

In contrast to the close similarity seen between DGGE profiles from the same patient, the corresponding profiles from all 33 patients studied were distinct. A smaller panel, restricted by software limitations, of 15 patients was arbitrarily selected to study the extent of profile dissimilarity between individuals. It was found that the DGGE profiles generated for these subjects were at least c. 34% similar to each other. It is tempting to speculate from these data that there is a core bacterial community common to all people, that is augmented by other bacterial species to form the gastrointestinal bacterial flora for a given host.

The primary application of DGGE in a study such as this is to explore the structure of the bacterial community within and between samples, not to exhaustively sequence the resulting bands. However DNA sequencing of individual DGGE bands was used to determine whether the bacterial profiles generated were derived from bacterial species considered to be typical of the gut. Again the methodology of this process was rigorously checked. Although the 16S rRNA gene region amplified from the DGGE bands contained rather limited information from which to make strong phylogenetic inferences, some insight was gained into the members of the colorectal flora of these patients.

As mentioned above, the bands sequenced in the present study were all of bacterial origin and were most closely related to database entries from bacteria often associated with the large intestine, including species within the genera Bacteroides, Ruminococcus, Faecalibacterium and Clostridium. This gave further confidence in the methods used in this study. A large number of DGGE bands gave sequences that were most closely related to accession numbers describing bacteria present as database sequences but never cultivated. This further underlines the importance for using molecular methods to investigate such complex bacterial communities, and demonstrates that our knowledge of the bacteria inhabiting the human gastrointestinal tract is far from complete. This is similar to the situation for the bacterial diversity found in other human habitats e.g. the lungs of cystic fibrosis patients (Rogers et al. 2003). In this study a large proportion of the DGGE bands sequenced directly belonged to species within the genus Bacteroides (Table 1). These results are similar to the findings of Wang et al. (2003) who found that approximately 40% of the 16S rRNA gene PCR products amplified from DNA extracted from a mucosal sample taken from a healthy individual were Bacteroides spp. The value of these data are shown when these are compared to findings of a recent study by Ott et al. (2004) who stated that the reduction in diversity of the colonic mucosa associated bacteria that they observed in patients with active IBD was due to a ‘loss of normal anaerobic bacteria such as Bacteroides species, Eubacterium species, and Lactobacillus species’. It is likely therefore that much will be learned by extending comparative studies of individuals with and without IBDs. Although typical gut bacteria have been identified from the DGGE community profiles, it was observed that other characteristic bacteria e.g. species of the Lactobacillus, Eubacterium, Atopobium, Roseburia and Bifidobacteria genera were not recovered. These bacteria may be present in the profiles, but the bands simply not selected in this study.

As with all techniques used to study complex microbial communities, DGGE has limitations. This method is reliant on the DNA recovered from the bacteria being representative of the community from which the sample has been taken (Gelsomino et al. 1999; Lloyd-Jones and Hunter 2001). Moreover, possible biases may be introduced during PCR amplification in the types of sequences that are amplified (Wintzingerode et al. 1997). Although the methodological validation undertaken in this study suggests these are not a limiting factor in this system, these are important topics to be considered. Specific limitations that apply to DGGE are that a single bacterial species may have more than one sequence variant of the 16S rRNA gene thus resulting in multiple bands on DGGE. Alternatively, different strains may have similar electrophoretic mobilities making them difficult or impossible to distinguish (Gelsomino et al. 1999). These factors need to be appreciated when studying bacterial communities analysed using DGGE.

In conclusion, a reliable and reproducible molecular method has been developed to allow the investigation of the dominant bacterial communities associated with single colorectal biopsies. We have shown that the mucosa-associated DGGE bacterial community profiles of healthy individuals are stable along the length of the colon. Each individual has a different bacterial profile, however there is some overlap between the mucosal-associated bacterial communities within individuals. This technique will facilitate further studies to characterize the adherent bacterial flora of the human gastrointestinal tract in health and disease.

Acknowledgements

This work was supported by the Hellenic Society of Gastroenterology and the Christopher Reeves Charitable Trust.

Ancillary