IRGM Variants and Susceptibility to Inflammatory Bowel Disease in the German Population

Background & Aims Genome-wide association studies identified the autophagy gene IRGM to be strongly associated with Crohn's disease (CD) but its impact in ulcerative colitis (UC), its phenotypic effects and potential epistatic interactions with other IBD susceptibility genes are less clear which we therefore analyzed in this study. Methodology/Principal Findings Genomic DNA from 2060 individuals including 817 CD patients, 283 UC patients, and 961 healthy, unrelated controls (all of Caucasian origin) was analyzed for six IRGM single nucleotide polymorphisms (SNPs) (rs13371189, rs10065172 = p.Leu105Leu, rs4958847, rs1000113, rs11747270, rs931058). In all patients, a detailed genotype-phenotype analysis and testing for epistasis with the three major CD susceptibility genes NOD2, IL23R and ATG16L1 were performed. Our analysis revealed an association of the IRGM SNPs rs13371189 (p = 0.02, OR 1.31 [95% CI 1.05–1.65]), rs10065172 = p.Leu105Leu (p = 0.016, OR 1.33 [95% CI 1.06–1.66]) and rs1000113 (p = 0.047, OR 1.27 [95% CI 1.01–1.61]) with CD susceptibility. There was linkage disequilibrium between these three IRGM SNPs. In UC, several IRGM haplotypes were weakly associated with UC susceptibility (p<0.05). Genotype-phenotype analysis revealed no significant associations with a specific IBD phenotype or ileal CD involvement. There was evidence for weak gene-gene-interaction between several SNPs of the autophagy genes IRGM and ATG16L1 (p<0.05), which, however, did not remain significant after Bonferroni correction. Conclusions/Significance Our results confirm IRGM as susceptibility gene for CD in the German population, supporting a role for the autophagy genes IRGM and ATG16L1 in the pathogenesis of CD.


Introduction
Crohn's disease (CD) and ulcerative colitis (UC) are chronic inflammatory bowel diseases (IBD) resulting from an inappropriate immune response to microbial antigens in genetically susceptible individuals [1,2,3,4]. Recent genome-wide association studies (GWAS) have provided valuable insights into the genetic architecture particularly of CD, identifying more than 70 CD susceptibility variants with the most significant findings in the gene regions of NOD2, IL23R and ATG16L1 [5,6,7,8,9]. These genetic findings confirm an important role for innate immunity, proinflammatory IL-23/Th17 immune responses as well as autophagy for both gut homeostasis and the development of chronic inflammation in IBD.
In addition to the CD susceptibility gene ATG16L1 involved in autophagy [9,10,11], recent GWAS identified the single nucleotide polymorphism (SNP) rs13361189 -a SNP lying immediately upstream of the autophagy gene IRGM (immunity-related GTPase family M) -and other IRGM SNPs to be strongly associated with CD [12,13]. Since the discovery of the IRGM as a CD susceptibility gene, further studies have investigated IRGM gene variants in both adult and pediatric CD [14,15,16,17,18,19,20,21,22] as well as in UC [14], confirming its role in the IBD pathogenesis.
A functional study suggested a common, 20-kb deletion polymorphism upstream of IRGM, which is in perfect linkage disequilibrium (LD) with rs13361189, as a likely causal variant, since the deletion allele modulated the expression of IRGM in transformed cells [22]. Another recent study implicated a variant in the 59-untranslated region (2308(GTTT) 5 ) to be independently associated with CD [20], while the very recent study by Brest et al. [21] demonstrated functional effects of the synonymous SNP rs10065172 (c.313C.T), which is also in linkage disequilibrium with rs13361189 and the deletion polymorphism. This exonic, synonymous variant rs10065172 in IRGM alters a binding site for certain microRNAs (miR-196) and causes deregulation of IRGMdependent xenophagy of bacteria in patients with CD [21], therefore suggesting rs10065172 as disease-causing variant.
These studies implicate that autophagy plays an important role in human inflammatory disorders by direct elimination of intracellular bacteria and activation of pattern recognition receptor (PRR) signaling which is involved in gut homeostasis and CD pathogenesis [10,23]. The IRGM gene belongs to immunity-related GTPases (IRG), a family of genes in mammalian species induced by interferons (IFNs) and functioning as key mediators of IFN-regulated resistance to intracellular bacteria and protozoa [23]. IRGM has been shown to play a role in the autophagy-targeted destruction of Mycobacterium bovis BCG [23] and the IFN-c-induced host defense against Salmonella typhimurium infection [10]. Interestingly, a recent study in CD patients demonstrated that autophagy limits the replication of intracellular adherent-invasive Escherichia coli (AIEC) associated with ileal CD and that IRGMand ATG16L1-deficient cells had enhanced intracellular AIEC bacteria replication, suggesting a significant impact on the outcome of intestinal inflammation [24].
While several GWAS and replication studies established IRGM as a CD susceptibility gene, its effects on the IBD phenotype are less clear. In addition, epistatic interactions with other IBD susceptibility genes, in particular the second autophagy gene ATG16L1, have not been studied in detail. Therefore, in this study, we aimed to analyze the role of IRGM on CD and UC susceptibility as well as its effect on the IBD phenotype in a large patient-control cohort. In addition, we performed a detailed epistasis analysis of IRGM with the three major CD susceptibility genes NOD2, ATG16L1 and IL23R. In total, six major IRGM SNPs, for which associations with CD were shown in previous studies (see details in Methods), were genotyped in more than 2000 German IBD patients and controls.

Ethics statement
The study was approved by the Ethics committee of the Medical Faculty of the Ludwig-Maximilians-University Munich. Written, informed consent was obtained from all patients prior to the study. Study protocols were based on the ethical principles for medical research involving human subjects of the Helsinki Declaration (http://www.wma.net/e/policy/b3.htm).

Study population and definition of IBD phenotype
The study population (n = 2060) consisted of 1099 IBD patients including 817 patients with CD, 283 patients with UC, and 961 healthy, unrelated controls, all of Caucasian origin. Patient charts were analyzed for demographic and clinical parameters (disease behaviour and anatomic location of IBD, disease-related complications, history of surgery or immunosuppressive therapy) and all patients participated in a detailed questionnaire including an interview at time of enrolment. The diagnosis of CD or UC was determined according to endoscopic, histopathologic and radiological criteria of current international guidelines [25]. Patients with clinical features of both CD and UC (and therefore classified as ''indeterminate colitis'') were excluded from this study. Patients with CD were assessed based on the Montreal classification including age at diagnosis (A), location (L), and behaviour (B) of disease [26]. In patients with UC, anatomic location was also assessed following the Montreal classification analyzing the criteria ulcerative proctitis (E1), left-sided UC (distal UC; E2), and extensive UC (pancolitis; E3) [26]. The demographic baseline characteristics of the study population were collected blind to the results of the genotype analyses and are summarized in Table 1.

DNA extraction
From all study participants, blood samples were taken and genomic DNA was isolated from peripheral blood leukocytes using the DNA blood mini kit from Qiagen (Hilden, Germany) according to the manufacturer's guidelines.

Genotyping of the IRGM variants
Six IRGM SNPs (rs13361189, rs10065172 = pLeu105Leu, rs4958847, rs1000113, rs931058, rs11747270) were genotyped. The selection of these six IRGM SNPs was based on previous studies showing associations for these SNPs in large case-control cohorts. The SNPs rs13361189, rs10065172 = p.Leu105Leu and rs4958847 were selected from the study of Parkes et al. [12], while the SNPs rs1000113 and rs931058 were tested in the study of the Wellcome Trust Case Control Consortium (WTCCC) [13]. The SNPs rs13361189 and rs10065172 = p.Leu105Leu served also as proxies for a common, 20-kb deletion polymorphism upstream of IRGM, since they are in perfect linkage disequilibrium (r 2 = 1.0) with this deletion polymorphism [22]. Additionally rs11747270, which was the most strongly CD-associated SNP within the IRGM region in the meta-analysis of Barrett et al., was included.
IRGM genotyping was performed by PCR and melting curve analysis using a pair of fluorescence resonance energy transfer (FRET) probes in a LightCyclerH 480 Instrument (Roche Diagnostics, Mannheim, Germany) as previously described in detail [27,28,29,30,31]. The donor fluorescent molecule (fluorescein) at the 39-end of the sensor probe is excited at its specific fluorescence excitation wavelength (533 nm) and the energy is transferred to the acceptor fluorescent molecule at the 59-end (LightCycler Red 610, 640 or 670) of the anchor probe. The specific fluorescence signal emitted by the acceptor molecule is detected by the optical unit of the LightCycler 480 instrument. The sensor probe is exactly matching to one allele of each SNP, preferentially to the rarer allele, whereas in the case of the other allele there is a mismatch resulting in a lower melting temperature. The total volume of the PCR was 5 ml containing 25 ng of genomic DNA, 16 Light Cycler 480 Genotyping Master (Roche Diagnostics), 2.5 pmol of each primer and 0.75 pmol of each FRET probe (TIB MOLBIOL, Berlin, Germany). In the case of rs11747270, the amount of the forward primer was reduced to one fifth and in the case of rs4958847 the reverse primer was reduced to one half. In the case of rs10065172 and rs931058, the reverse primers were reduced to one third, respectively. The PCR comprised an initial denaturation step (95uC for 10 min) and 45 cycles (50 cycles in the case of rs10065172) [95uC for 10uC sec, 60uC (55uC in the case of rs10065172) for 10 sec, 72uC for 15 sec]. The melting curve analysis comprised an initial denaturation step (95uC for 1 min), a step rapidly lowering the temperature to 40uC and holding for 2 min, and a heating step slowly (1 acquisition/uC) increasing the temperature up to 95uC and continuously measuring the fluorescence intensity. The results of melting curve analysis have been confirmed by analyzing two patient samples for each possible genotype using sequence analysis. For sequencing, the total volume of the PCR was 100 ml containing 250 ng of genomic DNA, 16 PCR buffer (Qiagen, Hilden, Germany), a final MgCl 2 concentration of 2 mM, 0.5 mM of a dNTP mix (Sigma, Steinheim, Germany), 2.5 units of HotStar Plus Taq TM DNA polymerase (Qiagen) and 10 pmol of each primer (TIB MOLBIOL). The PCR comprised an initial denaturation step (95uC for 5 min), 35 cycles (denaturation at 94uC for 30 sec, primer annealing at 60uC for 30 sec, extension at 72uC for 30 sec) and a final extension step (72uC for 10 min). The PCR products were purified using the QIAquick PCR Purification Kit (Qiagen) and sequenced by a commercial sequencing company (Sequiserve, Vaterstetten, Germany). All sequences of primers and FRET probes used for genotyping and for sequence analysis are given in Tables S1 and S2.

Statistical analyses
For evaluation of data, the SPSS 13.0 software (SPSS Inc., Chicago, IL, USA) and R-2.13.1 (http://cran.r-project.org) were used. Each genetic marker was tested for Hardy-Weinberg equilibrium in the control group. Fisher's exact test was used for comparison between categorical variables and Student's t test was applied for quantitative variables. All tests were two-tailed and pvalues,0.05 were considered as significant. Odds ratios were calculated for the minor allele of each SNP. Correction for multiple testing was performed by Bonferroni correction where indicated. Haplotype analysis was calculated using the -haplogistic command in PLINK (http://pngu.mgh.harvard.edu/ ,purcell/plink/), epistasis analysis was performed with theepistasis option. LD between SNPs was evaluated using the Rlibrary genetics. Genotype-phenotype associations were also tested in R using logistic regression.

Results
The IRGM gene variants are associated with susceptibility to CD The allele frequencies of the SNPs rs13371189, rs10065172 = p.Leu105Leu, rs4958847, rs11747270, rs931058 and rs1000113 of all three subgroups (CD, UC, and controls) were in accordance with the predicted Hardy-Weinberg equilibrium and are summarized in Table 2

IRGM haplotype analysis
Next, we performed a detailed haplotype analysis investigating the role of IRGM haplotypes on CD and UC susceptibility. As demonstrated in tables 3 and 4, several IRGM haplotypes demonstrated an association with CD and UC susceptibility. In CD, the strongest associations were found for haplotypes  containing at least one of the most strongly CD-associated SNP rs13361189 or rs10065172 (Table 3), while in UC, the strongest association was found for rs11747270-rs931058 (omnibus p-value 1.57610 22 ) (Table 4). However, given the large number of haplotypes analyzed, none of these associations withstood Bonferroni correction for multiple testing.

Genotype-phenotype analysis
We further investigated whether IRGM SNPs are associated with certain phenotypic characteristics in IBD patients. Based on the Montreal classification of IBD, the phenotypic data of IBD patients were analyzed for anatomic localization. However, none of the IRGM SNPs investigated were associated with specific disease localization in CD (Table S6) or UC (Table S7). Moreover, a detailed genotype-phenotype analysis in CD patients of the exonic synonymous SNP rs10065172 = p.Leu105Leu, which was in linkage disequilibrium with rs13361189 and with the previously identified 20-kb deletion polymorphism immediately upstream of IRGM (r 2 = 1.0), did not reveal any significant associations with the CD phenotype (Table S8).

Analysis for epistasis of IRGM with other major CD susceptibility genes
Finally, we analyzed potential evidence for gene-gene interactions of IRGM variants with other CD susceptibility genes such as variants in the NOD2, IL23R and ATG16L1 gene including their effect on CD susceptibility. Interestingly, there was evidence for weak gene-gene-interaction between several SNPs of the two autophagy genes IRGM and ATG16L1 (ATG16L1 rs12471449, ATG16L1 rs1441090, ATG16L1 rs4663396), which, however, did not remain significant after Bonferroni correction ( Table 5). The odds ratios of gene-gene interactions, which were significant before Bonferroni correction, are given in Table 6. There was no epistasis between IRGM and the other two major CD susceptibility genes NOD2 and IL23R.

Discussion
This study represents a detailed analysis of IRGM gene variants regarding their role in the susceptibility and phenotype of IBD in a large cohort of more than 2000 Caucasian individuals. In line with previous GWAS and replication studies [12,13,14,15,16,17,18,19], our results confirm an association of the IRGM variant rs13371189 with CD susceptibility. A detailed functional study identified a deletion polymorphism directly upstream of the IRGM locus as a candidate SNP to explain the CD association at this locus [22], affecting the tissue-specific expression level of IRGM [37]. This 20-kb deletion polymorphism is in perfect linkage disequilibrium (r 2 = 1.0) with SNP rs13361189, therefore implicating that rs13361189 is a proxy for this deletion polymorphism.
Similar to the study by McCarroll et al. [22], we demonstrate that the common exonic synonymous SNP rs10065172 = p.Leu105Leu is in linkage disequilibrium with rs13361189 and therefore also with the previously identified 20kb deletion polymorphism. The exonic SNP rs10065172 (c.313C.T) has been previously classified as non-causative given the absence of an alteration in the IRGM protein sequence or splice sites, although this view is challenged by the results of a very recent study [21]. The study by Brest et al. demonstrated that a family of microRNAs (miRNAs), miR-196, is overexpressed in the inflamed intestinal epithelium of CD patients and downregulates the IRGM protective variant (c.313C) but not the CD-associated allele (c.313T) [21]. The same study demonstrated that the resulting loss of tight regulation of IRGM expression compromises the control of the intracellular replication of CD-associated adherent invasive Escherichia coli (AIEC) by affecting the efficacy of bacterial phagocytosis (xenophagy) [21]. Therefore, Brest et al. [21] suggest the synonymous SNP rs10065172 (c.313C.T) as a likely causal variant. rs10065172 has been also shown to be associated with susceptibility to tuberculosis [38] which is of interest, given evidence that certain mycobacteria may play a role in the pathogenesis of CD. Moreover, a functional study demonstrated that IRGM induces autophagy to eliminate intracellular mycobacteria [23].
Overall, the association signal of IRGM with CD found in our study was considerably weaker than that shown by us for the other autophagy gene ATG16L1 in a similar sized cohort [11]. Similarly, the recent CD meta-analyses showed a stronger association signal for ATG16L1 than for IRGM [5]. Anderson et al. performed a very large meta-analysis of CD and UC associated susceptibility loci [6]. In this analysis, the CD case-control cohort included n = 6,333 CD patients and n = 15,056 controls, while the UC case-control cohort consisted of n = 6,687 UC patients and 19,718 controls [6]. In this large meta-analysis, they demonstrated that the IRGM SNP      Table 2) and associations of several IRGM haplotypes with UC (p,0.05), IRGM can be regarded to be weakly associated with UC and as a shared susceptibility gene of both UC and CD, although it has a much more prominent role in the pathogenesis of CD.
In addition, we also performed a detailed genotype-phenotype analysis of IRGM variants in CD and UC patients. In contrast to a recent study of Latiano et al. [40] demonstrating an association of IRGM variants with fistulizing CD, our genotype-phenotype analysis did not reveal any significant association of IRGM variants with the CD phenotype. We were also unable to confirm an association with ileal CD found in a previous study of a smaller CD cohort from New Zealand [18]. Our findings may be related to the rather weak association signal found for IRGM in the German CD cohort, although the results of this genotypephenotype analysis are consistent with the lack of a well-defined phenotype in CD patients carrying risk alleles of the other autophagy gene ATG16L1 [11].
The identification of the two major CD susceptibility genes ATG16L1 and IRGM involved in autophagy has significantly strengthened the importance of autophagy and bacterial xenophagy in the complex and multifactorial etiology of IBD. However, potential epistatic interactions between ATG16L1 and IRGM have not been investigated in detail so far. We therefore analyzed epistasis between these two genes, demonstrating a weak gene-gene-interaction between several SNPs of the two autophagy genes IRGM and ATG16L1 which, however, did not remain significant after Bonferroni correction. Given their close functional relationship, this potential epistasis signal is highly interesting. Very recently, the largest IBD meta-analysis published so far (including 75,000 IBD patients and controls) was made publicly available [39]. Part of this meta-analysis was an epistasis analysis in IBD, UC and CD datasets of the Immunochip study. While the analyses of the CD and UC subsets were inconclusive, the results for the analysis with IBD showed only one suggestive result between SNPs near SLC7A10 (rs17694108) and IL2RA (rs12722515) with a p-value of 3.26610 25 [39]. Therefore, the weak gene-gene interaction found between IRGM and ATG16L1 regarding CD susceptibility in our study, which was not significant after Bonferroni correction, could not been replicated on a significant level in this very large CD cohort. Thus, it is unlikely that epistasis between the two major autophagy genes contributes significantly to CD susceptibility.
There is increasing evidence for important intersections of autophagy and intracellular bacterial sensing (demonstrated by the importance of NOD2 in autophagy induction [41,42]) in the pathogenesis of IBD. Moreover, recent studies identified a new pathway closely linked to autophagy and innate immunity, which is characterized by an unfolded protein response, stimulated by endoplasmic reticulum (ER) stress due to the accumulation of misfolded proteins. Several genes involved in ER stress, including XBP1 and ORMDL3 have been linked to the IBD pathogenesis on a genetic level [43,44]. Interestingly, ATG16L1, NOD2, and XBP1 have been also demonstrated to affect the function of Paneth cells [43,45,46], suggesting a central role for this cell type in the development of IBD.
These recent findings are in line with raising evidence that NOD2 is involved in regulation of autophagy. Dendritic cells from CD patients expressing CD-associated NOD2 or ATG16L1 variants were shown to be defective in autophagy induction, bacterial trafficking and antigen presentation [41]. Most interestingly, a recent study demonstrated that the intracellular sensors NOD1 and NOD2 are critical for the autophagic response to invasive bacteria by recruiting the autophagy protein ATG16L1 to the plasma membrane at the bacterial entry site [42]. In cells homozygous for the CD-associated NOD2 frameshift mutation (p.Leu1007fsX1008), mutant NOD2 failed to recruit ATG16L1 to the plasma membrane and wrapping of invading bacteria by autophagosomes was impaired [42]. This is of particular interest, since we previously demonstrated a very severe stricturing phenotype in CD patients homozygous for the NOD2 p.Leu1007fsX1008 mutation associated with early disease onset, ileal stenosis, recurrent need for surgery and increased prevalence of entero-enteral fistulae [32,33]. However, despite the central functional role of NOD2 in the induction of autophagic processes, our study could not demonstrate gene-gene-interactions between NOD2 and IRGM regarding CD susceptibility. Moreover, we could not identify significant epistatic interactions between IRGM and IL23R, the main IBD susceptibility gene involved in Th17 responses. Of interest, a very recent study demonstrated IL23R variants as susceptibility variants for leprosy and suggested a potential involvement of IL23R in the autophagocytosis of mycobacteria involved in the pathogenesis of leprosy [47].
In conclusion, our results confirm IRGM as susceptibility gene for CD in the German population, while we did not show an Table 6. Odds ratios (ORs) and 95% CI for the gene-gene interaction (epistasis) found to be significant (before Bonferroni correction) for CD susceptibility (shown in Table 5). association with a specific IBD subphenotype. The strongest association signals for CD susceptibility were found for rs13361189 (proxy for the common, 20-kb deletion polymorphism upstream of IRGM) and the exonic synonymous SNP rs10065172 = p.Leu105Leu, supporting previous functional studies that these two SNPs may be the causal variants. However, the strength of the association signal with CD found here was several log-fold weaker than that demonstrated by us for the second autophagy gene ATG16L1 [11], suggesting a more important role for ATG16L1 in the CD pathogenesis. In UC, several IRGM haplotypes were weakly associated with UC susceptibility. This is consistent with recent meta-analyses which found weak associations with UC but very strong disease associations with CD. One might therefore hypothesize that autophagy genes such as IRGM and ATG16L1 play a more important role in the susceptibility to CD than UC. The potential epistasis signal between IRGM and ATG16L1 regarding CD susceptibility found in this study is highly interesting but could not been confirmed in a very large recent IBD meta-analysis [39] arguing against a major role of epistasis between IRGM and ATG16L1 regarding CD susceptibility.