SEARCH

SEARCH BY CITATION

Keywords:

  • methylation;
  • epigenetic;
  • Crohn's disease;
  • ulcerative colitis;
  • mucosa;
  • microarray

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES
  8. Supporting Information

Background:

DNA methylation constitutes a key epigenetic mechanism by which cells regulate gene transcription. Among its roles are the dynamic regulation of gene expression, for example, as part of an evolving immune response, and cell differentiation in specialized tissues. Here our aim was to study the impact of differences in methylation patterns in the intestine with regard to inflammatory bowel disease (IBD) susceptibility and activity.

Methods:

Having extracted DNA from rectal biopsies, we conducted genome-wide methylation profiling using the HumanMethylation27 BeadChip microarray to identify genes showing evidence of differential methylation between cases of ulcerative colitis and Crohn's disease and healthy controls. Selected methylation signals were validated in an independent replication panel by pyrosequencing. Correlation with gene expression was sought by quantitative real-time polymerase chain reaction (RT-PCR).

Results:

Multiple genes showed significant evidence of differential methylation, several appearing in both ulcerative colitis and Crohn's disease comparisons including THRAP2, FANCC, GBGT1, DOK2, TNFSF4, TNFSF12, and FUT7. Many more than expected by chance overlapped with genes previously implicated as playing a role in IBD susceptibility in genome-wide association scans, including CARD9, ICAM3, and IL8RB (P < 0.001). Correlation between methylation and gene expression was identified for selected transcripts.

Conclusions:

Consistent differences in DNA methylation between IBD cases and controls at regulatory sites within these genes suggest that their altered transcription contributes to IBD pathogenesis. (Inflamm Bowel Dis 2012;)

The last decade has seen substantial progress in characterizing pathways underlying the pathogenesis of inflammatory bowel disease (IBD), in particular driven by the new technologies of genome-wide association scanning (GWAS).1 Use of such hypothesis-free approaches has delivered major new insights regarding pathogenic mechanisms, including identifying the contribution of defects in autophagy as associated with Crohn's disease (CD) susceptibility and the role of Th17 pathways in both CD and ulcerative colitis (UC).2–5 Another key message to have emerged from GWAS studies is that more than two-thirds of confirmed susceptibility loci confer increased risk of IBD by affecting transcriptional regulation, with many of the risk alleles demonstrably correlating with expression of nearby candidate genes.6, 7

In parallel with the developments that allow high-throughput single nucleotide polymorphism (SNP) genotyping have come new technologies that enable interrogation of other forms of genetic variation on a genome-wide scale—among these genome-wide methylation microarrays. These are important, as methylation represents one of the key epigenetic mechanisms by which eukaryotic organisms regulate the transcription of genes. Increased methylation of cytosine residues within the regulatory sequence of a specific gene will impair binding of transcription factors and result in reduced gene expression. Increasingly, it is recognized that much of the complexity of human biology derives not from variation in the DNA coding sequence, but from the complex, networked regulation of gene transcription by epigenetic mechanisms.8

Methylation signatures are heritable, being transmitted at mitosis, for example, to maintain tissue-specificity of gene expression, and their rapid adjustment by DNA methyl-transferase enzymes allows activation or silencing of genes in a manner increasingly implicated in the dynamic regulation of immune responses.9 An important question is whether variation in DNA methylation affects susceptibility to IBD. That GWAS studies have highlighted the importance of altered transcriptional regulation plus the fact that one of the loci identified in the most recent CD GWAS meta-analysis encodes a key de novo DNA methyltransferase (DNMT3A) strongly implicate this mechanism,6 but to date relatively few studies have investigated it directly. Most have focused on the development of neoplasia in the context of chronic UC, based on the known role of altered methylation in sporadic colorectal cancer pathogenesis.10 A study of DNA methylation from nonneoplastic mucosal samples provided evidence of differential methylation between IBD cases and controls, but focused on a limited number of candidate genes only,11 and a recent genome-wide survey identified multiple genes showing differential methylation between CD cases and controls, with significant overlap between these signals and the findings from GWAS studies in CD, but was based on analysis of DNA from peripheral blood rather than mucosal samples.12

Here our aims were to study the impact of differences in methylation patterns in the intestinal mucosa with regard to IBD susceptibility. Specifically, to 1) use DNA derived from rectal biopsies and conduct a genome-wide survey to identify genes showing evidence of differential methylation between cases of UC and CD compared with age- and sex-matched healthy controls; 2) correlate these findings with expression of the implicated genes; and 3) identify overlap with signals from IBD GWAS studies.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES
  8. Supporting Information

Sample Cohort and Collection

Rectal biopsies were collected at the time of planned endoscopic examination from patients with UC or CD and from healthy controls being colonoscoped for investigation of rectal bleeding or anemia but who had no evidence of IBD or any other colonic pathology. Inflammatory status was defined endoscopically and confirmed by histology of a paired biopsy sample—the inflammation in each case being assessed by a consultant histopathologist (A.I.). Eight patients were recruited for each of five groups for methylation microarray analysis (group 1 = patients with active UC in the rectum; group 2 = patients with quiescent UC; group 3 = patients with CD with active rectal inflammation; group 4 = patients with CD but without rectal involvement; group 5 = healthy controls). A replication panel consisted of a further eight patients for each group as defined above (Supporting Table 1a,b). This study was approved by the Cambridge 4 Research Ethics Committee and all study subjects provided informed consent.

Methylation Microrrays and Analysis

DNA was extracted from rectal biopsies using the DNeasy blood and tissue kit (Qiagen, Chatsworth, CA), and then bisulphite converted using EZ DNA Methylation Gold Kit (Zymo Research, Orange, CA), according to the manufacturer's instructions. Bisulfite-converted genomic DNA was analyzed using the Illumina Infinium HumanMethylation27 BeadChip array. This contains 27,578 CpG loci covering more than 14,000 human RefSeq genes at single-nucleotide resolution. The BeadChip protocol comprises six steps (whole-genome amplification, fragmentation, hybridization, washing, counterstaining, and scanning) and was carried out at Cambridge Genomic Services, University of Cambridge, UK.

Infinium Human Methylation27 raw data were subjected to quality control using standard libraries in Bioconductor.13 The data were extracted and summarized using BeadStudio. After passing quality control, sample data were analyzed to determine differentially methylated genes. Genes were ranked in terms of evidence for differential methylation using empirical Bayes moderated t-statistics.14 Resulting P values were adjusted with the Benjamini–Hochberg false discovery rate correction method.15 Results are shown as adjusted P values indicating significance of differential methylation and the B value (the log odds of differential methylation vs. no differential methylation). Pathway analysis was performed by two approaches: signaling pathway impact analysis, which measures the perturbation on a given pathway under a given condition,16 and the global test approach as described by Goeman et al.17 The identified pathways were visualized with the Exploratory Gene Association Networks tool.18 A heatmap showing the methylation profile in each comparison was generated with modified functions within the R statistical environment.19

Methylation and GWAS Interval Analysis

In order to determine whether the overlap observed between GWAS associated regions and differentially methylated genes occurred significantly more frequently than expected by chance, we carried out a permutation analysis. Unlike genetic association signals, epigenetic regulation is not constrained by meiotic recombination hotspots. Association intervals as defined in the recent GWAS meta-analysis reports were thus extended by 25 kb p and q telomeric to reflect this.6, 7 Overlap between genes showing differential methylation in our “inflamed UC vs. control” analysis and genes mapping within the extended UC GWAS meta-analysis intervals was then sought; the CD datasets were interrogated for overlap in an equivalent manner. The permutation tests to ascertain whether the observed overlap exceeded that expected by chance were carried out in the R statistical environment using the Hybrid method of Huen and Russell in the Cooccur package.20 In all, 1000 permutations were used for each disease phenotype.

Pyrosequencing

To validate the findings of differential methylation observed on the Infinium microarray with an alternative quantitative method we used pyrosequencing. We targeted four genes that had shown among the strongest evidence for differential methylation between IBD cases and controls in the microarray analysis and sought replication in a new panel comprising eight cases for each disease phenotype group and eight more controls. DNA was bisulphite-converted as described earlier. Sequencing and polymerase chain reaction (PCR) primers were designed using the Pyrosequencing Assay Design 2.0 software (Qiagen/Biotage, Uppsala, Sweden) (Supporting Table 2). The methodology for the pyrosequencing was as described previously.21 Methylation quantification was performed using PyroQ CpG software, which quantifies each assayed CpG site based on the ratio between methylated and nonmethylated forms of DNA.

Quantitative Real-time PCR (qRT-PCR)

Expression of selected differentially methylated genes was analyzed by qRT-PCR using the Applied Biosystems (Foster City, CA) 7500 fast platform. RNA was extracted from whole biopsies using a DNA/RNA All Prep minikit (Qiagen) and RNA was converted to cDNA using a high-capacity reverse transcription kit (Applied Biosystems) according to the manufacturer's instructions. Taqman assays (DOK2-Hs00182758, FUT7-Hs00237083, and TAP1-Hs00388675; Supporting Table 3) were run using the fast method. Samples were analyzed in triplicate, and genes of interest were compared to GAPDH and 18s rRNA as endogenous references. Relative quantification of gene expression was calculated using the ΔΔCT method.22

Cell Separation and Isolation from Whole Biopsies

To separate the epithelial cell fraction, fresh whole biopsies were collected in phosphate-buffered saline (PBS) and transferred to an enzyme solution containing 1 mg/mL Liberase DH (Roche, Nutley, NJ) and 20 μg/mL hyaluronidase (Calbiochem, La Jolla, CA) in Accumax buffer (Sigma, St. Louis, MO). The biopsy containing solution was incubated at 37°C for 30 minutes with shaking until complete dissociation of the tissue occurred and the solution was then filtered through a 70-μm sieve. The cells were resuspended in red blood cell lysis buffer (Roche) and incubated at room temperature for 10 minutes, then collected by centrifugation, counted, and resuspended in MACs buffer; 5 × 107 cells were then separated using the MACs MicroBeads separation method (Miltenyi Biotec, Auburn, CA) according to the manufacturer's instructions. Briefly, cells were magnetically labeled with CD326 (EpCAM) MicroBeads, and the cell suspension then loaded onto a MACs MS column. Unlabeled (nonepithelial) cells ran through the column and were collected. Labeled cells, retained on the column by exposure to a magnetic field, were then washed, removed from the magnetic field, and eluted from the column.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES
  8. Supporting Information

In the analysis of the microarray data multiple loci showed evidence of differential methylation between inflamed UC and controls, noninflamed UC and controls, and inflamed CD and controls, with no genes showing significant differential methylation in the comparison between noninflamed CD and controls. The genes showing strongest evidence for differential methylation for each of the above comparisons are shown in Table 1 (for full listings, see Supporting Table 5). Inflamed and noninflamed UC and inflamed CD cases could readily be distinguished from controls based on the signatures provided by the methylation profiles (Fig. 1). Of interest relating to their potential use as biomarkers, it was possible to define distinct signatures for noninflamed UC vs. noninflamed CD but not for inflamed UC vs. inflamed CD. From these analyses it is evident, first, that there is significant overlap in the genes identified in the different comparator groups, second, that many of these genes have immunoregulatory functions, and third, that many of the signals fall within regions implicated by recent IBD GWAS studies. Each of these themes was explored in additional downstream analyses and experiments.

Table 1. Methylation Array Data; Genes Showing the Most Statistically Significant Evidence of Differential Methylation
inline image
thumbnail image

Figure 1. Heatmap of the probes and samples for each comparison. Microarray heatmaps use a colored grid linked by a dendrogram of the samples to hierarchically cluster genes. Each column represents a different sample, each row a different gene probe. Each cell in the heatmap is colored based on the level of expression of that probe in that sample. Green cells signify an increase and red a decrease in methylation. The figure highlights the probes showing the most consistent discrimination between cases and controls. UCI, inflamed ulcerative colitis; UC, noninflamed ulcerative colitis; CDI, inflamed Crohn's disease; CD, noninflamed Crohn's disease. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

While the appearance of several genes (such as DOK2, COG8, TNFSF4, and GBGT1) at the top of more than one comparator list provided indirect validation of the microarray data, particularly given the agnostic testing of 27,000 probes, direct validation required replication of signals in an independent case-control panel using a different technology for assessing locus-specific methylation. This was provided by the targeted pyrosequencing experiments for a small panel of genes (FUT7, DOK2, and TAP1). These were selected because they were among the top differentially methylated genes with known functional relevance to make them interesting candidates, and robust pyrosequencing primers could be designed for the same CpG dinucleotides as probed by the methylation microarrays. For each group we studied biopsies from eight additional patients. The results exactly replicated the microarray findings for each gene tested, in demonstrating differential methylation between groups with the same direction of methylation change observed (Supporting Fig. 1).

We undertook pathway analysis using two different approaches to explore the possible functional connectivity between loci identified as showing differential methylation. The results are shown in Supporting Table 4 and Supporting Figure 2a,b. Among the many innate and adaptive immune pathways implicated were the regulation of cell adhesion and leukocyte migration, epithelial cell–cell junctions, antigen presentation, T- and B-cell receptor signaling, and JAK-STAT signaling. Several of these were highlighted in both CD and UC.

Many genes implicated by GWAS studies also showed evidence of differential methylation between cases and controls, including CARD9, IL8RA/B, MUC1, CDH1 (which encodes e-cadherin), and ICAM3 (Table 2). Formal permutation testing of the correlation between loci showing differential methylation in the current study and loci implicated by previous GWAS analyses demonstrated significantly greater overlap than predicted by chance (P < 0.001). Results of the full analysis are presented in Supporting Table 6.

Table 2. Methylation Events in GWAS Association Intervals
Focal SNPSNP LocationAssociation IntervalMethylated GeneIllumina ProbeComparisonCpG IslandB Value
  1. Association interval is in NCBI Build 36 coordinates. Overlap between loci showing both evidence of association with IBD in GWAS studies and strongest evidence of differential methylation, evidenced by a positive B value and a corrected P < 0.05, in the current study. The 71 CD GWAS intervals and 48 UC GWAS intervals 6,7 were each interrogated for the presence of loci showing differential methylation in their respective comparator groups (i.e., inflamed CD vs. control and inflamed UC vs. control). Overlap was observed for 10 loci in the CD analysis, 13 in the UC analysis and 10 in both. This was substantially greater than would have been expected by chance (P < 0.001), as determined using permutation testing (see Materials and Methods).

rs69473911q1363.58–64.05KCNK4cg13412615inflamed UC vs. Con 5.51
rs1272035619p1310.26-10.50ICAM3Cg14145194inflamed UC vs. Con 4.94
rs92688536p2131.49–33.01C6orf25cg14437986inflamed UC vs. Con6:31798937-317992134.68
rs107814999q34138.27–138.55CARD9cg24793265inflamed UC vs. Con9:138388961-1383891814.58
rs116763482q35218.58–218.97IL8RBcg13739417inflamed UC vs. Con 4.16
rs116763482q35218.58–218.97IL8RAcg21004129inflamed UC vs. Con 3.99
rs92688536p2131.49–33.01C6orf27cg12939547inflamed UC vs. Con6:31851630-318520333.79
rs90761111q151.82–1.93TNNI2cg25623459inflamed UC vs. Con 2.66
rs649918816q2266.98–67.40CDH1cg24765079inflamed UC vs. Con16:67328348-673300002.50
rs107814999q34138.27–138.55CARD9cg02516189inflamed UC vs. Con 2.43
rs601734220q1342.49–42.70ADAcg20622019inflamed UC vs. Con20:42713127-427145402.30
rs649918816q2266.98–67.40SMPD3cg19297232inflamed UC vs. Con16:67039271-670407302.08
rs17999646p2131.49–32.98TAP1cg16853860inflamed CD vs. Con 4.16
rs1187180117q2137.57–38.25PTRFcg14070162inflamed CD vs. Con17:37827092-378292033.70
rs17999646p2131.49–32.98PSMB8cg16890093inflamed CD vs. Con 2.98
rs116763482q35218.58–218.97IL8RBcg13739417inflamed CD vs. Con 2.26
rs31800181q22153.24–154.39PKLRcg02280309inflamed CD vs. Con 1.74
rs17999646p2131.49–32.98C6orf25cg14437986inflamed CD vs. Con6:31798937-317992131.68
rs69473911q1363.58–64.05KCNK4cg13412615inflamed CD vs. Con 1.11

Having obtained evidence for the presence of differential methylation between IBD cases and controls, important questions relate to the functional impact on gene transcription and the cell of origin of such signals. The former was addressed by undertaking qRT-PCR for selected gene targets that had shown evidence of marked differential methylation on the array experiment and that had been validated in the pyrosequencing replication experiments. FUT7, DOK2, and TAP1 were studied, with relative quantification data revealing large fold changes between cases and controls for the first two with reduced methylation observed in biopsies from inflamed mucosa in cases compared to controls correlating with the expected increase in mRNA levels (evidenced by reduced ΔCT values) (Fig. 2). In formal statistical testing using Pearson's rank correlation coefficient the P values for each gene were 0.0001, 0.002, and 0.055, respectively, suggesting that methylation at these sites is indeed impacting gene expression, at least for DOK2 and FUT7. As a preliminary proof of principle we undertook additional methylation analysis using pyrosequencing on biopsy material following separation of epithelial cells from the nonepithelial fraction. The results are presented in Supporting Figure 3 for two target genes of interest, FUT7 and ATG12. It is clear that there is significant differential methylation between the epithelial and nonepithelial fraction for each group tested. Furthermore, for FUT7 the reduced methylation seen in biopsies from inflamed CD and UC cases compared to controls seemed to partition into the nonepithelial fraction, comprising lymphocytes and stromal cells, while for ATG12 most of the increase in methylation seen in cases compared to controls appeared attributable to the epithelial fraction.

thumbnail image

Figure 2. Quantitative RT-PCR data for genes assayed by pyrosequencing to correlate evidence for differential methylation with altered gene expression in rectal biopsies. The three boxplots in the upper panel show expression data for FUT7, DOK2, and TAP1, with lower ΔCT values denoting higher levels of expression (seen for all genes tested in each disease group compared to control, apart from CD in the DOK2 analysis). P values were calculated using Student's t-test. *P < 0.0001; **P < 0.001; ***P < 0.05. The middle panel shows relative quantification (RQ values) for each group, the lower panel shows ΔCT values plotted against percent methylation for each group. Spearman rank correlation revealed statistically significant association between methylation status and gene expression for FUT7 and DOK2, with a nonsignificant trend seen for TAP1. CD, noninflamed Crohn's disease; UC, noninflamed ulcerative colitis biopsies. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES
  8. Supporting Information

This study examined the epigenetic modifications associated with IBD, the impact these factors have on gene expression, and the relationship between methylation status and disease-associated genetic markers. Methylation of specific DNA motifs, particularly CpG islands in gene promoters, represents a key mechanism by which cells regulate gene transcription and pass such signals from mother to daughter cells at mitosis. It is tightly regulated, and can be influenced in different cell types by diverse environmental factors such as smoking, infection, and diet.23

The GWAS evidence implicating altered regulation of gene transcription in IBD pathogenesis is compelling but mostly circumstantial.6, 7 In the current study we provide direct evidence that altered methylation at specific genes in intestinal biopsies correlated with the development of IBD, and in so doing highlight the contribution of these specific genes to IBD pathogenesis. By using a genome-wide methylation microarray we pursued a hypothesis-free approach akin to GWAS studies to obtain new insights regarding pathogenically important pathways. Given the risk of type 1 error inherent in testing thousands of loci, we used stringent statistical correction of microarray data and validated our findings in an independent replication panel. The differences observed between IBD cases and controls thus appear robust and real.

The proteins encoded by the genes showing differential methylation (Table 1; Supporting Table 5) have a wide range of functions. DOK2, for example, acts as a scaffolding protein for the assembly of multimolecular signaling complexes such as modulate the cellular proliferation induced by interleukin (IL)-4. Tap1 is involved in the transport of antigens from the cytoplasm to the endoplasmic reticulum for association with MHC class I molecules. CD28 is a costimulatory molecule essential for CD4+ T-cell proliferation, Th2 development, and IL-2 production. ICAM3 encodes an intercellular adhesion molecule that plays a key role in leukocyte adhesion and migration as well as signaling. Members of the tumor necrosis factor (TNF) superfamily, such as TNFSF4 and TNFSF12 show signal in several comparator groups, thus highlighting the important role played by the TNF cytokine family in IBD. ULK1, by contrast, appears only on the CD vs. control list. It plays an important role in regulating autophagy, and has been implicated in earlier genetic studies.24

The overlap between the results of our current study and previous IBD GWAS findings is intriguing—including here CARD9, IL8RA, ICAM3, and CDH1. A similar observation was made in a recent study comparing methylation profiles between CD cases and controls but in DNA from peripheral blood.12 That entirely different molecular techniques consistently implicate overlapping regions of the genome is a clear indication that these loci encode genes that play an important role in IBD pathogenic pathways. Indeed, the methylation results strongly complement data derived from GWAS studies in implicating specific genes—something not always possible in association studies where resolution may be critically limited by the presence of strong linkage disequilibrium.

Particularly interesting was the evidence of increased methylation we observed at CDH1. This gene encodes e-cadherin, which plays a central role in epithelial cell–cell adhesion, and has been reported to be downregulated in areas of UC inflammation.25 Our data provide a mechanistic explanation for this. Furthermore, in GWAS studies the locus encoding CDH1 has been associated with both UC and colorectal cancer,26 and differential methylation at this site has been identified in DNA extracted from UC-associated colorectal cancers.27 Thus, the increased methylation seen in mucosal biopsies in our study in UC cases vs. controls suggests one potential mechanism by which UC might predispose to bowel cancer.

For a more complete view of the functional impact of the genes showing differential methylation we subjected our genome-wide data to gene set analysis (Supporting Table 4 and Supporting Fig. 2). Many pathways appeared on more than one comparator list—for example, antigen processing and presentation, cell adhesion, B- and T-cell receptor signaling, JAK-STAT signaling, and transforming growth factor beta (TGF-β) signaling were all flagged in both inflamed CD and inflamed UC vs. control. This underlines the fact that many pathways are likely common to all forms of IBD and mimics the overlap between CD and UC suggested by GWAS studies.7 Interestingly, differential methylation of genes involved in the epithelial “tight junction” and “adherens junction” appeared more prominent in UC than CD. Altered methylation of such genes, either as a precondition to or consequence of colonic inflammation, might affect their subsequent transcriptional regulation, altering the integrity of the epithelial barrier and allowing luminal antigens access to the mucosal cellular immune compartment. It is intriguing that this defect appears more prominent for UC, again mimicking the GWAS data and perhaps correlating with the superficial, mucosa-focused nature of the inflammation in UC as opposed to the deeper, transmural inflammation that characterizes CD.

Of note, no genes showed evidence of differential methylation in the comparison between noninflamed CD and controls. Here, rectal biopsies from the “cases” came from individuals with an established diagnosis of CD but without rectal involvement. This is in contrast to the comparison between noninflamed UC and controls, where the “case” rectal biopsies came from a site of previous inflammation that had resolved prior to the sample being taken (a point confirmed on histological examination of paired biopsy samples). In both cases there was no excess of inflammatory cells in the mucosa and submucosa, but previous inflammation seems to have left a footprint of methylation change that perhaps leaves the epithelium in that area more vulnerable to reactivation of inflammation.

One question is whether there is any direct mechanistic connection between the overlap in methylation and GWAS signals—for example, whether there is any evidence that the “focal” (most associated) single nucleotide polymorphism in any of the GWAS intervals (or SNPs in linkage disequilibrium with them) disrupt CpG sites within CpG islands or shores, and hence affect methylation. Certainly there is colocalization, with evidence of many associated SNPs mapping to important methylation regulatory regions such as islands and shores (Supporting Table 6), and many SNPs disrupting existing CpG sites or making new ones, but little evidence from our current data that the exact same sites that show differential methylation also represent polymorphic sites demonstrating genetic association with IBD susceptibility. This might reflect the relatively low density of CpG sites covered by the 27K BeadChip microarray, and it will be interesting to see whether the next generation of methylation arrays (such as the new Illumina 450K array) identify such sites.

Despite these caveats, we have identified GWAS-associated genes with differential methylation within the CpG island closest to the transcriptional start site—for example, at IL8RB. The methylated sequence within its promoter region corresponds to the putative binding site of the PU.1 (SPI-1) transcriptional activator. IL8RA maps to the same region as IL8RB and also contains a differentially methylated sequence within its promoter. The “focal” SNP within the GWAS interval is rs11676348, which lies between IL8RA and IL8RB. Emphasizing the scope for interaction between associated sequence variants and methylation, rs11676348 alters a CpG residue. While this was not itself probed by the Illumina 27k array, it has CpG sites on either side and both of these demonstrated evidence of differential methylation. Furthermore, the SNP rs11676348 lies in the transcription factor binding sites for STAT3, which forms part of the JAK-STAT signaling cascade important for many cytokine receptors and also previously implicated by IBD GWAS studies.28 These data serve to highlight the potential convergence between GWAS associations and methylation, and echo recent results from type 2 diabetes.29

In addition to identifying genes showing evidence of differential methylation between IBD cases and controls, we were able, as a proof of principle, to demonstrate the functional impact of this change on expression of selected genes using qRT-PCR. Systematic testing of all the other genes showing differential methylation was beyond the scope of the current study, but we have no reason to suppose that these would be different.

The distinct methylation signatures that we observed between CD and UC, particularly in the comparison of “uninflamed” cohorts, suggests possible utility as a biomarker. However, from the point of view of delineating pathogenically important pathways there is a clear concern that our analyses were conducted using whole mucosal biopsies, containing a mixed cell population. Given that methylation signatures are cell-type-specific the question arises as to whether the signals we see arise from the epithelial or cellular immune compartments, and whether they might be confounded by the differing proportions of cells present in the inflamed vs. control biopsies. Our pilot results (Supporting Fig. 3) clearly showed differential methylation between epithelial and nonepithelial fractions for the genes tested, emphasizing the importance of cell separation, particularly for more subtle effects. For FUT7 the differential methylation signal observed in inflamed whole biopsies from IBD cases vs. controls appeared to derive mainly from reduced methylation in the nonepithelial fraction, represented by immune and stromal cells, while for ATG12 the differences were mainly seen in the epithelial fraction. FUT7 encodes a fucosyltransferase that plays a key role in the biosynthesis of sialyl Lewis X, which itself serves as a ligand in selectin-mediated adhesion of leukocytes to activated endothelium. ATG12 encodes a key component of the autophagy machinery. As well as further validating the results of the methylation microarray, these data also confirm that we can detect cell-type-specific methylation signals from a heterogeneous cell population. Future work will need to refine the cell separation protocols to allow detailed analysis of each of the different cellular compartments represented in mucosal biopsies.

In summary, we have identified panels of genes that in rectal biopsies show evidence of differential methylation between CD and controls, or UC and controls, or both comparator groups. This strongly suggests that these genes play an important role in IBD pathogenesis, and gene set analysis has highlighted a number of key pathways. The differential methylation observed affects gene expression at the mRNA level, and our findings are corroborated by significant overlap with the results from previous GWAS studies, with the current data providing complementary evidence to implicate specific genes. The scene is now set for more detailed and complex work to analyze the specific cell subsets present in the intestinal mucosa, to obtain a more granular view of the importance of epigenetic regulation of gene transcription in the predisposition to IBD.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES
  8. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES
  8. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
IBD_22942_sm_SuppFigures.doc556KSupporting Information Figures
IBD_22942_sm_SuppTables.doc2374KSupporting Information Tables

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.