- Top of page
- MATERIALS AND METHODS
Background: Inflammatory bowel disease (IBD) is a chronic gastrointestinal condition without any known cause or cure. An imbalance in normal gut biota has been identified as an important factor in the inflammatory process.
Methods: Fifty-eight biopsies from Crohn's disease (CD, n = 10), ulcerative colitis (UC, n = 15), and healthy controls (n = 16) were taken from a population-based case-control study. Automated ribosomal intergenic spacer analysis (ARISA) and terminal restriction fragment length polymorphisms (T-RFLP) were used as molecular tools to investigate the intestinal microbiota in these biopsies.
Results: ARISA and T-RFLP data did not allow a high level of clustering based on disease designation. However, if clustering was done based on the inflammation criteria, the majority of biopsies grouped either into inflamed or noninflamed groups. We conducted statistical analyses using incidence-based species richness and diversity as well as the similarity measures. These indices suggested that the noninflamed tissues form an intermediate population between controls and inflamed tissue for both CD and UC. Of particular interest was that species richness increased from control to noninflamed tissue, and then declined in fully inflamed tissue.
Conclusions: We hypothesize that there is a recruitment phase in which potentially pathogenic bacteria colonize tissue, and once the inflammation sets in, a decline in diversity occurs that may be a byproduct of the inflammatory process. Furthermore, we suspect that a better knowledge of the microbial species in the noninflamed tissue, thus before inflammation sets in, holds the clues to the microbial pathogenesis of IBD.
Inflammatory bowel disease (IBD) comprises Crohn's disease (CD) and ulcerative colitis (UC) and is a chronic lifelong gastrointestinal condition that results in considerable morbidity including surgery in many patients.1, 2 It is also an economically important disease because it normally affects adults in their prime working years.3 The prevalence of this disease has increased dramatically in the Western world over the last 20–30 years, and has even begun to increase in westernizing Asian populations in Singapore and Japan.4 Epidemiological data suggest that the incidence rate of IBD in the developed world is 3–4 times higher than developing countries.5
It has long been suspected that the cause of IBD lies at the intersect of a susceptible human genome, a dysfunctional immune system, and a microbial pathogen(s).6–9 The fact that the incidence of IBD in the Singapore population has increased faster during the post-war years10 cannot be accounted for by natural genetic drift in the human genome, and gives credence to the hypothesis that environmental factors, like microorganisms, must be involved.11–14 Absence of colitis in germ-free animal models,15 and protective effects of probiotics in IBD patients,16, 17 are additional reasons in support of the role of microbes in the pathogenesis of IBD. A number of microorganisms have been linked to IBD, the most common being Mycobacterium paratuberculosis,18–22 but no definitive cause and effect relationship has been proved.
The pitfalls of research into microbial etiologies of IBD have been: 1) biases introduced by culture-dependent investigations, and 2) the lack of case-control studies that are population-based. In the first instance, the majority of microbial studies in IBD have been based on cultivation of microorganisms associated with biopsy tissue,23, 24 with the related biases of not being able to culture the majority of biopsy bacteria.25 In the second case, when culture-independent studies have been conducted, the control biopsies were not true controls, in the sense that the non-IBD subjects were individuals who were negative for IBD, but were usually undergoing treatment for conditions like colorectal cancer, celiac disease, or irritable bowel syndrome.26–28
In this article we describe the similarities between controls, UC, and CD biopsies from the cecum and rectum using terminal restriction fragment length polymorphisms (T-RFLP)29 and automated ribosomal intergenic spacer analysis (ARISA).30 Biopsies used were generated from a unique population-based case-control study in Manitoba, Canada.
- Top of page
- MATERIALS AND METHODS
We analyzed a total of 58 biopsies isolated from 41 individuals, including controls (Table 1). Diagnosis of CD or UC was based on histological examination and 8 noninflamed CD and 10 noninflamed UC tissues were included in the sample collection. Four patients with ulcerative proctitis were grouped with UC patients. Four samples from UC patients were collected from the descending colon and were grouped with rectum and 2 from the ascending colon were also grouped with cecum.
Universal primers amplified the intergenic spacer region between the 16S and 23S rDNA genes and separation by capillary electrophoresis (ARISA) indicated that noninflamed and inflamed tissues had different profiles (Fig. 1). A total of 114 different sized fragments, with an average of 11 peaks per profile, were observed. The presence or absence of unique peaks from ARISA was used to construct a matrix of incidence values (0 or 1) that were numerically analyzed to produce a dendrogram of relationships (Fig. 2). ARISA profiles from controls, UC (including noninflamed UC tissues), or CD (including noninflamed CD tissue), did not cluster as tightly as noninflamed versus inflamed. If a clustering distance of 80% was used the majority of inflamed or noninflamed tissues aggregated into a single cluster for 14 of the 17 clusters. However, for the largest cluster, constituting 29 observations, a match of only 55.1% was obtained (Fig. 2).
T-RFLP profiles were generated when 16S rDNA fragments amplified with universal primers (27f and 342r), digested (HhaI), and separated by capillary electrophoresis. Sixty-eight distinct fragments were obtained in total, with an average of 9 fragments per sample. Noninflamed tissue and diseased tissue had different profiles (Fig. 3). As was the case with ARISA, clustering was not robust when tissues were grouped by their clinical diagnosis, and a much higher level of clustering was attained if tissues were separated based on inflammation state (Fig. 4). At a 70% distance cutoff, more than 90% of tissues grouped into inflamed or noninflamed clusters.
Species richness and diversity indices were used to describe the microbial population in the biopsy tissue. When control, CD, and UC tissues were compared using ARISA and T-RFLP output data, richness tended to be lower in the rectum than the cecum (Table 2). However, when the same calculations were done when designating tissues as inflamed or noninflamed a much higher level of discrimination was obtained (Table 3). When diversity was calculated no significant difference were obtained with the Shannon index. Using the Simpson index diversity with ARISA data was control > CD > UC, but with T-RFLP data the diversity was CD > control > UC. Classifying tissues according to their inflammation state, the diversity was noninflamed > control > inflamed (Table 3). Similarity indices were calculated and compared normal tissues of controls and patients, as well as noninflamed and inflamed IBD tissues (Table 4).
T-RFLP fragments generated and matched to GenBank accession numbers produced peak identities that in many cases fell into multiple phyla (Table 5). To resolve this conflict we matched multiple accession numbers associated with single peaks at the phylum, class, order, family, and genus level based on Bergey's taxonomy. At the order level more than 95% of accession numbers were matched to a single taxonomic order (Table 5). Thus, we assigned fragments generated from T-RFLP only to the order level. Using this approach we observed fewer (P < 0.05) members of the Bacteroidetes, but more Firmicutes in the inflamed tissue (Table 6).
- Top of page
- MATERIALS AND METHODS
In molecular microbial analysis of bacterial populations associated with IBD, we are faced with a 3-fold challenge. First, analyses of microbiota should preferably be done on biopsy tissue rather than stool. Although it is much easier to obtain fecal samples, they obscure the differences between the sites.31, 46–49 Second, identification of adequate controls is difficult. The ideal control needs to be an individual that voluntarily submits to colonoscopy, but does not have any gastrointestinal symptoms that could potentially bias the analysis. Third, there is significant microbial variation between individuals, and a major portion of this variation can be accounted for by the individual's genotype.50–52 This means that to obtain reliable results, sizable numbers of individuals must be sampled. The first 2 of the above criteria were met because analyzed biopsy tissues were taken from a case-control population-based IBD tissue bank held at the University of Manitoba.18 The highest resolution technique based on 16S rDNA is the use of clone libraries, but this is an expensive and time-consuming procedure.31, 53 Consequently, few samples can be processed with clone libraries, so to meet our third criteria, large numbers of samples, we used ARISA and T-RFLP.
Multivariate analysis of the ARISA (Fig. 2) and T-RFLP data (Fig. 4) did not clearly differentiate between controls, CD, and UC when disease type was the clustering criteria. A far higher level of clustering was obtained using the inflamed versus noninflamed criteria. Closer inspection of the dendrograms (Figs. 2, 4) indicates that the clustering was better with T-RFLP (>90%) than ARISA (78%). T-RFLP is based on the 16S rDNA structural gene, which has a relatively low level of taxonomic resolution compared to the ribosomal intergenic spacer region, which not only has significant sequence variation, but also a length polymorphism.54, 55
ARISA and T-RFLP are both accepted methods for assessing complex microbial communities.56, 57 Recently, Danovaro et al56 showed that ARISA is more efficient than T-RFLP for estimating the diversity of aquatic communities. However, our study indicated that neither ARISA nor T-RFLP could group tissues into phenotypically meaningful clusters. For example, a majority of tissues clustered into inflamed or noninflamed sets only if a very liberal cutoff was used (Figs. 2, 4). A more conservative cutoff resulted in clustering that according to our definition (based on phenotype of tissue or disease) was not meaningful. The inability to clearly cluster tissues by disease state is most likely the consequence of interindividual microbiota variation.50–52
The effects of host genotype can be most clearly demonstrated in homozygotic twins.51 DGGE profiles of the 16S genes indicated that the variation in population was more influenced by host genome than the environment, even though the environmental factors were obviously still very important. This concept can be further supported by the fact that spouses (different genotypes) sharing the same diet, housing, etc. (same environment) were less similar in gut microbiota composition than the siblings of the spouses (same genome) living in a different environment. Eckburg et al31 made 16S rDNA clone libraries from biopsy tissue of 3 healthy individuals consisting of over 13,000 clones. There were significant differences between individuals as well as between different sites of the same subject, but the intersubject variance was greatest.
Despite its limitations, recognition of the extent of diversity in gut ecology is essential to our understanding of the association between a community's structure and function,58 as well as the relationship between the microbial flora and the human host.59 The gut microbiota, in general, develops over the first few years of life and then remains remarkably stable, except in the case of certain diseases,8 although some healthy individuals can show significantly greater population fluctuations.60 Prolonged alteration of this balanced composition may result in the chronic stimulation of the mucosal immune system and, consequently, loss of tolerance to the commensal bacteria.61
Given the limitations in analyzing the ARISA and T-RFLP data using hierarchical clustering methods, we employed statistical indices of ecological diversity and similarity. Diversity indices and similarity indices are often used interchangeably in the gut microbiology literature and it is useful to define these terms explicitly. Diversity indices are made up of richness and individual abundance. Species richness is simply the number of different species, or operational taxonomic units (OTU) present, and diversity is a means of weighting OTU abundance. Similarity, on the other hand, is a measure of shared species. For example, two microbial communities can have identical species richness, but zero similarity, because the same species are not present in both communities.
Previous studies have demonstrated a reduction in diversity in both feces53, 62 and mucosal associated63 microbiota of IBD tissues as compared to controls. Using two approaches (ARISA and T-RFLP) we have shown similar results (Table 2), but have gone further and examined noninflamed tissue in IBD patients (Table 3). Our results indicate that there is a clear difference in both richness and diversity between the microbiota of the inflamed and noninflamed sites, but this is less clear with disease state as the comparator. This is very interesting because one of the clinical features of CD are the skip-regions, which are areas of mucosa that are not inflamed and lie adjacent to inflamed tissue.14, 49 In UC, largely a large bowel disease, there is usually a gradient from noninflamed to inflamed, and over time the inflammation progresses into the noninflamed areas. Organisms present in the noninflamed or preinflamed tissue in IBD patients are of interest because biological processes in these regions potentially represent pathogenic entities. Our results clearly indicate that there is an increase in diversity from controls to the noninflamed tissues, and then when the inflammation sets in the diversity of microbiota declines again. We hypothesize that there is a recruitment phase in the noninflamed tissue and species involved in this phase may well hold the keys to pathogenesis.
Results of recent studies suggest similar microbial composition for inflamed and noninflamed IBD tissues.64–66 Bibiloni et al64 and Seksik et al65 used DGGE and TTGE to calculate similarity indices, not species richness or diversity, for inflamed and noninflamed tissues for a given patient. Gophna et al66 compared 16S clone libraries from inflamed and noninflamed sites in the same patient, based on a modified similarity index. In our study, we pooled all inflamed and noninflamed biopsy data, then made comparisons regardless of the subject, gut location, or disease. We observed less similarity between the inflamed and noninflamed libraries (64%–76%, Table 4) than previous studies (90%–97%).64, 65
Assignment of T-RFs to hierarchical taxonomic groups indicates that statistically significant differences, or near-significant differences, occur in the phyla Bacteroidetes and Firmicutes (Table 6). In the Bacteroidetes there is a putative loss of species as the tissue becomes inflamed, while in the class Clostridia of the phylum Firmicutes there is an increase in species. In particular, we observed a highly significant increase in a number of unclassified Clostridia in the inflamed tissue.
Recently, the composition of feces microbiota of CD patients was assessed in separate studies.53, 62, 67 Manichanh et al53 constructed metagenomics libraries from feces taken from CD patients while Scanlan et al62 generated DGGE profiles for different microbial groups. Both studies found significant changes in the composition of the microbiota of CD patients, mainly in the Firmicutes (decreased diversity), and Bacteroidetes (less diversity).53, 62 In our study we observed a putative increase in the diversity of Firmicutes, most significantly in unclassified Firmicutes. Given that the unclassified Firmicutes represent less well understood bacteria, it is likely that the bacteria represented by these T-RFs are good targets for future studies in IBD.
In summary, we have demonstrated that diversity of microbial species in noninflamed IBD tissue likely forms an intermediary community of organisms in transition to the inflamed state. We believe that further analysis of the organisms involved in this transitional state may lead to significant advances in our understanding of the pathogenesis of IBD. Future studies will focus on isolating these species and trying to address their relationship to the pathophysiology of CD and UC.