Inflammatory bowel disease (IBD) encompasses both Crohn's disease (CD) and ulcerative colitis (UC), chronic inflammatory disorders of the gastrointestinal tract with developed world predominance and an incidence that has risen dramatically in the post-war period [1]. IBD manifests with symptoms such as severe diarrhoea, weight loss and debilitating abdominal pain, resulting in substantial morbidity and impairment in quality of life [2]. In both diseases visibly inflamed and non-inflamed areas of intestine can be identified at assessment by colonoscopy.

The cause of both conditions is still speculative. Host genetics play a key role, with genetic factors more important for development of CD than UC [3, 4], but genetic defects cannot wholly explain the increasing prevalence of IBD in recent years, suggesting that environmental factors are also involved [5]. The current generally accepted disease hypothesis is that the chronic inflammation of IBD results from a genetically dysregulated host immune response directed at the gut microbiota [6–8].

The human gut microbiota is a highly diverse and abundant community of microbes that under normal circumstances is either commensal or beneficial to human health [9]. Bacteria in the gut contribute to host nutrition via production of short chain fatty acids and vitamins, and play integral roles in maintaining human health by preventing colonisation by pathogens and by shaping and maintaining normal mucosal immunity [10]. The microbiota is also, however, a major source of antigens, including lipopolysaccharides, peptidoglycan, lipoproteins, flagellin and unmethylated CpG-containing DNA, all of which can activate both innate and adaptive immune responses [11, 12]. A balanced relationship, therefore, must exist between bacteria and their human hosts. A disruption in this homeostasis threatens the state of immune tolerance and may result in gut inflammation.

Several lines of evidence suggest a role for gut bacteria in the pathogenesis of IBD. Faecal stream diversion induces remission in CD [13], animal models of colitis require the presence of gut bacteria to initiate inflammation (reviewed in [14]), an increased mucosal bacterial load is observed in IBD patients [15, 16], genome-wide IBD association studies have identified polymorphisms in genes involved in bacterial recognition and clearing (reviewed in [17]) and broad-spectrum antibiotics have some efficacy in the treatment of CD [18, 19].

With CD in particular, individual species such as Mycobacterium avium subspecies paratuberculosis or Escherichia coli have been implicated in disease aetiology [20, 21] while the emerging "dysbiosis" hypothesis implicates multi-species assemblages in an overall imbalance between harmful and protective bacteria [22, 23]. Numerous studies have attempted to characterise the microbial communities in IBD and to compare these with healthy individuals. Results indicate that individuals with IBD have reduced bacterial diversity, temporal stability and cluster separately when compared to healthy controls [24–28]. Compositional comparisons have generated inconsistent results but have generally identified reductions in components of the Firmicutes phylum in IBD, often, but not always, with concurrent increases in Bacteroidetes and facultative anaerobes such as Enterobacteriaceae [12, 22, 29–31].

Faecal/luminal bacterial communities have repeatedly been shown to be distinct from mucosal communities [32–37], meaning that study of the IBD mucosa-associated microbiota and comparison with those from healthy individuals should provide the best insight into whether or not a particular microbial signature is disease specific. In addition, within IBD-affected intestines disease-causing agents might be enriched at sites of active inflammation relative to comparatively unaffected mucosa. We have therefore used in-depth bacterial 16S rRNA gene cloning and sequencing technology to compare the mucosa-associated microbiota from inflamed and non-inflamed sites of the colon in CD and UC patients and in non-IBD controls. Our findings indicate that mucosal microbial diversity and composition is disturbed in IBD and that there are significant differences in microbial community structure between inflamed and non-inflamed mucosa.

To our knowledge, this is one of the largest clone library studies investigating the microbiota in IBD. In contrast to an earlier study by Frank et al., [30], which examined a smaller number of clones from a large number of patients, we sought instead to add to current knowledge by obtaining a higher resolution of the IBD-associated microbiota with particular emphasis placed on observing differences between inflamed and non-inflamed colon sites in the same patients. This was inevitably done in a smaller number of patients and samples because of the depth of molecular analysis required for each sample. Our in-depth clone library analysis, utilizing the resolving power of near full-length 16S rRNA gene sequences, revealed significant differences in diversity and composition between the mucosal microbiota of healthy patients and IBD sufferers. The results also suggest a tendency towards a reduction in Firmicutes and an increase in Bacteroidetes species in IBD patients compared to controls and also indicate that there is an increase in Enterobacteriaceae in CD. Similar shifts in composition, in either one or all of these groups, have been reported by other investigators using both culture [22] and a variety of molecular techniques [29, 31, 45–55]. A previous large-scale clone library analysis by Frank et al., [30], however, reported a decrease in proportions of Bacteroidetes and the Firmicutes family Lachnospiraceae in a subset of, but not all, IBD patients and an increase in Proteobacteria. The observed discrepancies between these two large-scale clone library studies may in part be explained by different disease phenotypes, dietary or other environmental differences, the effect of inter-individual variation between patients or the differing number of samples studied and the depth of sequencing between each study.

We also demonstrated a reduction in bacterial diversity within IBD patients compared to controls and this is in agreement with several previous studies [24–27, 56, 57]. Our data shows, however, that despite the differences between IBD and non-IBD patients in both bacterial composition and diversity that samples clustered predominantly by individual rather than disease.

Using both culture and molecular methods, many studies have demonstrated that the mucosal community along the length of the colon is largely stable, in healthy and IBD patients, and distinct from that recovered in faeces [32–37]. Here we provide evidence instead for the development of localised differences in mucosal microbiota structure in IBD. Our community comparison results suggest that there may be differences between inflamed and non-inflamed tissue, with significant changes in the composition of the bacterial communities at these sites. A number of prior studies have also attempted to establish whether or not there is localised dysbiosis in IBD between inflamed and non-inflamed tissue. While two of these studies indicated that there is a dysbiosis [58, 59], the majority have suggested that this is not the case [29, 48, 60–62]. Discrepancies between these results and ours may result from the use of differing molecular methodology and/or the greater sequencing depth we employed. DGGE/TGGE and FISH are useful tools but the resolving power of these methods is much lower than that for in-depth clone libraries covering the full length of the 16S rRNA gene [63]. In addition, DGGE/TGGE cannot accurately describe quantitative differences between dominant bands or describe qualitative differences in sub-dominant species and single bands on the gel may contain DNA from more than one species [64].

While our results suggest that localised changes in the mucosal microbiota do exist in IBD we were not able to identify a bacterial species or cluster that was consistently associated with the inflamed gut and therefore, potentially, with IBD aetiology. Other large-scale clone library analyses have also failed to identify specific pathogens [29, 30]. While their absence may indicate that potential pathogens may simply form a very minor component of the microbiota, these results do not support the hypothesis that a particular bacterial agent causes IBD. Clone libraries generate inherent biases, however, and it is possible that they are unable to detect certain species due to methodological artefacts. Indeed, this may be important with Mycobacterium avium subspecies paratuberculosis, a member of the often underrepresented Actinobacteria phylum [65, 66]. The absence of bifidobacteria from our dataset indicates that our clone libraries also suffer from this same bias against Actinobacteria. It is also worth noting that our analysis would not detect any viral, archaeal or eukaryotic aetiological agents. This may be important given recent evidence suggesting a role for viruses in the induction of at least some models of IBD [67].

Sequence-based microbiota comparisons such as ours can of course only demonstrate associations and do not provide information regarding mechanism or causation. It is also difficult to differentiate between compositional changes that may play a role in disease pathogenesis and those which may simply have occurred as a result of disease. However, given the absence of a specific and recurring aetiological agent in the cumulative data across all published IBD studies, which incorporate both culture- and molecular-based methodologies, it is possible that the alterations in bacterial composition and diversity seen between healthy and IBD patients and between inflamed and non-inflamed mucosa may be, to at least some extent, the result of the disturbed gut environment rather than the direct cause of disease. Indeed, there are a number of reasons why IBD is likely to result in altered conditions for bacterial growth. For example, the gut in IBD is likely to be a less stable environment than that of healthy individuals, with more exposure to antibiotics and other drug regimes, and alterations in transit time. Microscopy studies have suggested that there is a higher penetration of bacteria and a greater bacterial load in the mucosal layer in IBD patients [47, 68] and the resulting inflammation drives the localised release of antimicrobial compounds [69]. In addition, in UC there is a reduced mucus layer in inflamed relative to non-inflamed regions [70].

Despite proportional increases in Enterobacteriaceae and Bacteroidetes within IBD patients, if these organisms were directly responsible for disease we might expect them to be elevated at sites of inflammation and this was not shown in our analysis. Taking into account all of the above factors, the observed increases in these bacterial groups in IBD patients as a whole may therefore simply reflect the adaptation of the individual microbiota to the IBD gut environment. Bacteroides thetaiotaomicron, for example, can adapt to inflammation in an experimental mouse model by inducing genes that metabolise host oxidative products [71] and inflammation per se has also been shown to promote the growth of Enterobacteriaceae in mouse models [72, 73]. Clearly, further similar studies are required on a far greater range of gut bacterial species so that we can better understand the response of the gut microbiota to alterations in environmental conditions.

This work demonstrates a dysbiosis, or imbalance, in microbial community structure and composition in inflammatory bowel disease patients relative to non-IBD controls. It also indicates that inflamed tissue differs from non-inflamed tissue, but not in a consistent or predictable manner. Indeed, despite general trends such as a reduction in diversity, the response to IBD may be to some extent specific to the individual. This lends support to the emerging hypothesis that IBD is combinatorial in aetiology, with many different combinations of genetic and environmental causes leading to similar therapeutic responses [67], and highlights the importance of interconnection between the environment, the microbiota and the host in health and disease.

Despite this, even if particular bacteria are not the specific cause of IBD, altered immune responses may act to select particular bacterial species through creation of favourable microenvironments and might therefore cause the outgrowth of potentially pathogenic commensal species [74]. Shifts in the microbiota may therefore still impact gut health by altering the antigenic exposure to the gut mucosa or by reducing its exposure to beneficial microbes and/or their metabolic products, thereby initiating a cycle that favours recruitment and growth of more pro-inflammatory species [17, 75]. The observed reduction in Firmicutes proportions, for example, might lead to an undesirable affect on gut health. Recent work describing the anti-inflammatory properties of one Firmicutes species, Faecalibacterium prausnitzii [42] illustrates this point.

Finally, results from metagenomic studies indicate that, regardless of species composition, the collective genomes of each individual's microbiota appear to encode a remarkably conserved set of functions [28]. If similar, and potentially aggravating, factors are encoded by multiple species, it is possible that we will be better served in the future by looking at the complete gene complement of the microbial community as a whole, not just species composition. With this in mind, it is hoped that further analysis of the complex interplay between host and microbes will yield important insights into the pathogenesis of IBD.