The composition of the human intestinal microbiota is recognized to be important in maintaining an individual's health,1, 2 and the dysbiosis of the enteric microbiota in human inflammatory bowl disease (IBD) may be a result of defective regulation of mucosal immune interactions.3 The interaction between gut microbes and host immune response has been recognized as a potential factor in the development of IBDs such as Crohn's disease (CD),4 which may derive from an underlying genetic susceptibility that results in an impaired mucosal immunity, and it has been reported that 17–25% of CD patients have mutations in the NOD2/CARD15 gene, which regulates host responses to bacteria.5 However, the precise microbial species, metabolites, or functional genes involved in initiation or sustaining the disease remain unknown. Recently, Manichanh et al6 reported that the fecal microbiota of healthy subjects contained a markedly increased diversity and abundance of Firmicutes, and the Clostridium leptum phylogenetic group was in significantly higher numbers in healthy subjects compared with CD patients. Some of these bacteria produce short-chain fatty acids, including butyrate, which is a source of energy for colonic enterocytes and ameliorates inflammation.7 Therefore, shifts in the population of certain beneficial or harmful gut bacteria in patients with IBD may result in potentiation or suppression of the disease.
Oligonucleotide microarray is a powerful tool that can be used for simultaneous analysis of thousands of genes or target DNA sequences on a single slide,8 but until recently most 16S rDNA based microarrays were employed with a restrictive number of probes.8–10 The technology has now evolved to high-density probe platforms (10,000–300,000 probes per microarray) that provide quantitative information on the broader taxonomic composition of diverse microbial populations.11, 12 However, there are still opportunities to improve the specificity of detection at the species level for bacteria resident in the human gastrointestinal tract because intestinal microbes from this ecosystem are extremely diverse at low taxonomic ranks. In this study we adopted a unique approach to develop and validate a new phylogenetic custom microarray. To increase specificity and sensitivity, we employed the GoArray program to design the majority of oligonucleotide probes used on the microarray.13 In addition, 40 mer oligonucleotide probes which were reported in the literature as species-specific were also used.12, 14 To validate the microarray we used pooled DNA from a group of six healthy subjects and six CD patients (ileum) whose microbial diversity and abundance had been characterized previously using 16S rDNA recovered from metagenomic clones of the DNA pools as well as fluorescent in situ hybridization (FISH) analysis of the respective fecal samples.6 Furthermore, major differences revealed by the microarray in bacterial species between the healthy and CD subjects were verified by quantitative real-time polymerase chain reaction (PCR) analysis using specific primers.
- Top of page
- MATERIALS AND METHODS
- Supporting Information
In this study we developed a phylogenetic microarray for analysis of human intestinal microbiota. The microarray was validated by hybridization with 48 sequenced fosmid clones (Supporting Tables 1, 2) and comparison analysis with quantitative real-time PCR (Fig. 4). The microarray analysis was consistent with the published findings from the fosmid libraries of the same pooled community DNA samples, which showed that Firmicutes was higher in abundance in the healthy subjects than in the CD patients.6 Since Guschin et al21 described the use of phylogenetic microarray to detect and quantify microbes in environmental samples, the technique has gained recognition and acceptance as an additional tool for high-throughput microbial diversity analysis. Palmer et al12 developed a DNA oligonucleotide microarray composed of 10,462 probes to target 1590 bacterial and 39 archaeal species for analysis of commensals and medically relevant microbes from the human gastrointestinal tract and they provided quantitative information on the taxonomic composition of diverse microbial populations. More recently, a phylogenetic microarray comprising 4800 oligonucleotide “HITChip” has been developed which provides broad coverage of the human gastrointestinal microbiota.22 DeSantis et al11 reported a novel microarray containing 297,851 probes that target the 16S rDNA of 842 prokaryotic subfamilies and demonstrated its utility on 16S rRNA gene amplicons generated from aerosol, soil, and water. However, it is still challenging to develop robust microarrays that have species-level specificity. Such microarrays are especially useful for analysis of bacteria resident in the human gastrointestinal tract, which contains great species richness within several predominant phyla. The microarray described in this study does not have the depth of taxonomic coverage described for the “HITChip”22 because there is a physical limitation of 2240 probes on the Combimatrix 4× 2K microarray chip. Because of this limitation we chose to include a higher proportion of Proteobacteria probes on the microarray relative to gut diversity due to the potential role of this phylum in IBD. Furthermore, several species including Anaerostipes, Anaerotruncus, Subdoligranulum were not included because sequence information unavailable in public databases when the probe design was being performed. Although the GoArray probes comprised 21% of the total probes on the microarray, this approach facilitated the design of species-specific probes which could not be achieved for some target species through the conventional design of a continuous 40 nt probe. This microarray does not generate truly quantitative data on bacterial population size, limiting the comparisons which may be made with data generated using other molecular approaches.
Since Palmer et al12 found specificity increased greatly when they hybridized labeled cRNA instead of DNA, we used labeled single-strand cRNA rather than double-strand DNA for hybridization on the microarray. In addition, we examined whether fragmentation of labeled cRNA improved sensitivity but there was no significant difference from unfragmented material, which was in accord with the findings of Hoen et al.23 They reported that there was no significant difference between average signals from fragmented cRNA (40–200 nt) and signals from intact cRNA (500–2000 nt) on their oligonucleotide microarray. However, in the case of native rRNA, it has been reported that direct labeling and fragmentation of native rRNA (20–100 nt) by using NaOH and ZnCl2 catalyzed methods not only enhanced hybridization signal intensities but also reduced false-negative signals.24 Based on our studies and these published results we used fragmented single-stranded cRNA to hybridize samples on the microarray. It is possible that bias could have occurred in the PCR amplification of different bacterial species but comparisons of differences in relative population abundance between CD and healthy samples were not invalidated by any PCR bias.
Manichanh et al6 showed that the fecal microbiota of patients suffering from CD contained a markedly reduced abundance and diversity of Firmicutes, and in particular, the C. leptum group was significantly less abundant in CD patients than in healthy subjects. Using the microarray, we found that some bacteria of Firmicutes were 5–10-fold more abundant in the healthy group than in the CD patients. These included Eubacterium rectale of the Lachnospiraceae, Ruminococcus albus, R. callidus, R. bromii, and F. prausnitzii of the Ruminococcaceae. The metagenomic data of Manichanh et al also showed that the C. coccoides group (includes E. rectale) was significantly higher in the healthy group. Based on a report by Duncan et al,25 these bacteria were included in the clostridial clusters IV and XIV, many of which can utilize and metabolize a wide range of carbohydrate substrates including starch and/or produce butyrate in the colon. This finding is interesting because butyrate is considered beneficial to colonic health as the main energy source for colonic epithelial cells26 and it inhibits proinflammatory cytokine mRNA expression in the mucosa.27 Also recently, Sokol et al28 reported that the species F. pausnitzii from the C. leptum group was significantly underrepresented in colitis compared with the fecal microbiota composition of patients with colitis to that of healthy subjects. This result should be another crucial evidence that lower counts of F. prausnitzii are consistently associated with a reduced protection of the gut mucosa.
In contrast, the CD patients carried Enterococcus sp., Clostridium difficile, Lactobacillus fermentum, E. coli, Shigella flexneri, and Listeria sp. (Proteobacteria) in greater abundance (more than 5-fold) than the healthy subjects. It is interesting that L. fermentum and bifidobacteria appeared to be in high abundance in some CD patients, as these bacteria are generally regarded as beneficial and used as probiotics. L. fermentum has some peculiar properties compared to other probiotics bacteria, such as the ability to produce antioxidants (glutathione) and induce the growth of other lactobacilli species and prevent colitis.29 However, this should be interpreted with caution because the probiotic strains may possess beneficial attributes that differ from the bacteria detected in this study. Also Enterococcus sp., particularly E. faecium, which belongs to the same cluster as L. fermentum, has not been identified previously as being in high abundance in CD patients. Over the last two decades E. faecalis has emerged as a pathogen, especially of nosocomial infections.30 Although E. faecalis can suppress the proliferation of Listeria sp.31, 32 and infant intestinal E. faecalis led to downregulation of inflammatory responses in human intestinal cell lines,33 many reports have shown that E. faecalis can induce IBD in interleukin-10 knockout mice34 and activate macrophages by releasing proinflammatory cytokines (TNF-α, IL-6, and IL-12 p70).35 Moreover, Huycke and Wang36, 37 reported that the intestinal commensal E. faecalis is unique in producing extracellular superoxide (O), and the superoxide can induce macrophage cyclooxygenase-2 (COX-2). Eventually, chromosomal instability (CIN) in mammalian cells was promoted by COX-2, leading to colorectal cancer. Indeed, Balamurugan et al38 reported that E. faecalis increased in the feces of colon-cancer patients and butyrate producers were decreased, these shifts in colonic bacterial population could be associated with or potentially lead to epithelial cell damage. In our microarray analysis, we found both E. faecalis and E. faecium were in higher abundance in CD patients. Interestingly, these two species are phylogenetically close based on 16S rRNA gene sequences (96% similarity) but E. faecium was present in higher numbers than E. faecalis using both microarray analysis and species-specific real-time PCR (data not shown). Although E. faecium has not been implicated in any intestinal immune response, this species also has been shown to account for the majority of infections caused by Enterococcus spp.39 and to produce extracellular superoxide.40 In our real-time PCR analysis of several bacterial species, E. faecium and C. difficile showed very high abundance in one of the six CD patients (CD5). C. difficile has been reported previously as being present in recurrent CD patients.41 In the PCA plot (Fig. 3), the CD5 sample was also significantly different than the other five CD samples. This may suggest that bacterial populations in CD patients could be different based on the stage and severity of the disease and give crucial information about the higher variability in microbial composition between CD patients and healthy individuals (Fig. 3). The microarray and qPCR analysis indicated that the B. fragilis group was higher in the healthy subjects, which is similar to the metagenomic data from Manichanh et al,6 although their FISH analysis of the total Bacteroides showed no difference between the groups. Nevertheless, future studies are required to determine the implication of E. faecium in CD. Our preliminary results suggest that F. prausnitzii and E. faecium might serve as “indicator” bacteria of a dysbiosis in patients. It should also be noted that the microarray approach does not generate truly quantitative data on bacterial population size but can be effectively used to focus the attention on discrete populations for quantitative analysis such as real-time PCR.