Association of ATG16L1 and IRGM Genes Polymorphisms with Inflammatory Bowel Disease: a Meta-Analysis Approach

ORIGINAL ARTICLE Association of ATG16L1 and IRGM genes polymorphisms with inflammatory bowel disease: a meta-analysis approach

RJ Palomino-Morales1, J Oliver1,MGo´mez-Garcıá2,MALo´pez-Nevot3, L Rodrigo4, A Nieto5, BZ Alizadeh6 and J Martıń1 1Instituto de Parasitologıá y Biomedicina ‘Lo´pez-Neyra’, CSIC, Granada, Spain; 2Servicio de Digestivo, Hospital Universitario Virgen de las Nieves, Granada, Spain; 3Servicio de Inmunologıá, Hospital Universitario Virgen de las Nieves, Granada, Spain; 4Servicio de Digestivo, Hospital Universitario Central de Asturias, Oviedo, Spain; 5Servicio de Inmunologıá, Hospital Puerta del Mar, Ca´diz, Spain and 6University Medical Center Utrecht, Utrecht, The Netherlands

The aim of this study was to determine the role of the ATG16L1 (rs2241880) and IRGM (rs13361189 and rs4958847) genes polymorphism in Crohn’s disease (CD) and ulcerative colitis (UC). Our study included 557 CD and 425 UC patients and 672 ethnically matched Spanish controls and a meta-analysis with the data published to date. The polymorphisms were genotyped using predesigned TaqMan single nucleotide polymorphism genotyping assays. There was a statistically significant difference in the distribution of the ATG16L1 rs2241880 G allele between CD patients and controls in the Spanish population: P ¼ 6.5 Â 10À9, odds ratio (OR) ¼ 1.62. Although no differences were observed between UC patients and controls in the Spanish cohort, a meta-analysis demonstrated that the ATG16L1 G allele increase significantly risk for UC (P ¼ 0.0003, pooled OR ¼ 1.08). In addition, our meta-analysis data showed that IRGM rs13361189 and rs4958847 polymorphisms were associated with CD (rs13361189 C allele P ¼ 1.07 Â 10À19, pooled OR ¼ 1.34; rs4958847 A allele P ¼ 2.78 Â 10À17, pooled OR ¼ 1.31) and UC (rs13361189 P ¼ 0.0069, pooled OR ¼ 1.16; rs4958847 P ¼ 0.014, pooled OR ¼ 1.13). In conclusion, our results confirm the ATG16L1 rs2241880 and IRGM rs13361189 and rs4958847 polymorphisms as important markers for CD susceptibility and indicate that these variants are also associated with UC. Genes and Immunity (2009) 10, 356–364; doi:10.1038/gene.2009.25

Keywords: ATG16L1; IRGM; autophagy; polymorphism; Crohn’s disease; ulcerative colitis

Introduction involved in the innate immune response to microbial antigens. This fact and evidence derived from animal Inflammatory bowel disease (IBD), which is usually models and clinical studies11–13 suggest that an inappro- classified into two clinical presentations, Crohn’s disease priate innate response to normal constituents of the (CD) and ulcerative colitis (UC), is a chronic relapsing luminal microflora plays a crucial role in the develop- and remitting disease of the intestinal tract. Although the ment of these diseases. aetiology of these diseases is not fully understood, there Recently, autophagy has been shown to be a key is strong support for a genetic component based on process in the innate immune response. Autophagy also findings of familial aggregation, higher concordance in contributes to endogenous major histocompatibility monozygotic twins and ethnic differences in disease complex class II antigen presentation, reflecting its role prevalence.1,2 in the adaptive response. Interestingly, recent genome- To date, different genetic studies have shown several wide association studies (GWASs) of CD have reported genes playing a relevant role in these diseases, including a strong association with two crucial genes in the NOD2/CARD15,3 DGL5,4 SLC22A4 and SLC22A5,5,6 autophagy process, ATG16L1 and IRGM.14–17 TNFSF15,7,8 NOD1/CARD47 and IL23R.9,10 ATG16L1 encodes a protein that interacts with the NOD2/CARD15 is likely to be the major genetic factor Atg12p–Atg5p conjugate during formation of the pre- contributing to CD. This gene encodes a protein directly autophagosomal structure, which is essential for autophagy.18–20 Association of CD with the ATG16L1 T300A (rs2241880) polymorphism was reported simultaneously Correspondence: Dr J Martin, Instituto de Parasitologı´a y Biome- by different GWASs.14–16 A large number of subsequent dicina ‘Lo´pez-Neyra’, CSIC, Parque Tecnolo´gico de Ciencias de la association studies have replicated the strong association Salud, Avenida del Conocimiento s/n, 18100-Armilla (Granada), of this gene with CD.21–34 Spain. E-mail: [email protected] IRGM (immunity-related guanosine triphosphatase, Received 25 September 2008; revised 12 February 2009; accepted 18 the human homologue of mouse Irgm/Lrg47) encodes February 2009 a GTP-binding protein that induces autophagy and plays Association of ATG16L1 and IRGM with IBD RJ Palomino-Morales et al 357 an important role in innate immunity against intra- for the ATG16L1 rs2241880 were observed in genotype cellular pathogens.35,36 Association of two immediately and allele frequencies between UC patients and controls. flanking IRGM gene polymorphisms (rs13361189 and When we combined CD and UC patients, an increased rs4958847) with CD has recently been reported.17,30,37 frequency of the G allele was observed compared with Reduced activity of these genes would be expected to controls (58.9 vs 51.2%, P ¼ 1.3 Â 10À05,OR¼ 1.37, 95% CI lead to persistence of intracellular bacteria, consistent 1.18–1.58), mainly due to the CD subgroup. No correla- with existing models of CD pathogenesis and the recent tion was observed between this genetic variant and ATG16L1 and IRGM association reports. the following variables: disease localization, disease The aim of this study was to determine the potential behaviour, disease severity, need for surgery and age implication of IRGM and ATG16L1 polymorphisms to of onset for both CD and UC patients (Supplementary CD and UC in a Spanish population. In addition, our Table 1). data were pooled with results from other available The distribution of IRGM rs13361189 and rs4958847 studies in a meta-analysis to determine the global role genotypic and allelic frequencies in the Spanish popula- of these gene variants in IBD. tion is shown in Table 2. The minor allele frequency of IRGM rs13361189 shows a clear trend of association with CD (16.7% patients vs 13.8% controls, P ¼ 0.05, OR ¼ 1.25, 95% CI 0.99–1.57), UC (16.8% patients vs 13.8% controls, Results P ¼ 0.06, OR ¼ 1.26, 95% CI 0.99–1.61) and IBD (16.7% Association study patients vs 13.8% controls, P ¼ 0.03, OR ¼ 1.25, 95% CI Deviation from Hardy–Weinberg equilibrium was not 1.02–1.53). For the IRGM rs4958847 variant, we observed observed in the entire cohorts for the ATG16L1 rs2241880 an association with CD compared with controls (22.2% and IRGM rs13361189 and rs4958847 polymorphisms. CD patients vs 18.4% controls, P ¼ 0.02, OR ¼ 1.26, 95% Table 1 shows the ATG16L1 rs2241880 genotype and CI 1.03–1.55). However, no association was detected with allele distribution in healthy controls and in CD, UC UC in our population. and IBD patients. The ATG16L1 rs2241880 G allele was We observed no specific genotype–phenotype strongly associated with susceptibility to CD in the associations for the IRGM gene variants with disease Spanish population (63.0 vs 51.2%, P ¼ 6.5 Â 10À09, odds localisation, disease behaviour, disease severity, need for ratio (OR) ¼ 1.62 95% confidence interval (CI) 1.37–1.91). surgery and age of onset for both CD and UC patients On the other hand, no statistically significant differences (Supplementary Table 1).

Table 1 Genotype and allele distribution of the ATG16L1 rs2241880 polymorphism in healthy controls and in CD, UC and IBD patients

ATG16L1 Controls CD P value OR (95% CI) UC P value OR (95% CI) IBD P value OR (95% CI) T300A n ¼ 666 (%) n ¼ 544 (%) n ¼ 414 (%) n ¼ 958 (%)

AA 167 (25.1) 75 (13.8) 1.1 Â10À6 0.48 (0.35–0.65) 95 (22.9) 0.43 0.89 (0.66–1.20) 170 (17.7) 3.4 Â 10À4 0.64 (0.50–0.83) AG 316 (47.4) 253 (46.5) 0.75 0.96 (0.76–1.22) 194 (46.9) 0.85 0.98 (0.76–1.26) 447 (46.7) 0.75 0.97 (0.79–1.19) GG 183 (27.5) 216 (39.7) 6.8 Â 10À6 1.74 (1.35–2.23) 125 (30.2) 0.34 1.14 (0.86–1.51) 341 (35.6) 5.8 Â 10À4 1.46 (1.17–1.82) A 650 (48.8) 403 (37.0) 6.5 Â 10À9 0.62 (0.52–0.73) 384 (46.4) 0.29 0.91 (0.76–1.09) 787 (41.1) 1.3 Â 10À5 0.73 (0.63–0.84) G 682 (51.2) 685 (63.0) 6.5 Â 10À9 1.62 (1.37–1.91) 444 (53.6) 0.29 1.10 (0.92–1.32) 1129 (58.9) 1.3 Â 10À5 1.37 (1.18–1.58)

Abbreviations: CD, Crohn’s disease; CI, confidence intervals; IBD, inflammatory bowel disease; OR, odds ratio; UC, ulcerative colitis.

Table 2 Genotype and allele distribution of the IRGM rs13361189 and rs4958847 polymorphisms in healthy controls and in CD, UC and IBD patients

IRGM Controls CD P value OR (95% CI) UC P value OR (95% CI) IBD P value OR (95% CI) rs13361189 n ¼ 654 (%) n ¼ 555 (%) n ¼ 425 (%) n ¼ 980 (%)

CC 17 (2.6) 17 (3.1) 0.62 1.18 (0.57–2.46) 13 (3.1) 0.65 1.18 (0.54–2.59) 30 (3.1) 0.58 1.18 (0.62–2.26) CT 147 (22.5) 151 (27.2) 0.06 1.29 (0.98–1.69) 117 (27.5) 0.06 1.31 (0.98–1.75) 268 (27.3) 0.03 1.30 (1.02–1.65) TT 490 (74.9) 387 (69.7) 0.04 0.77 (0.59–1.00) 295 (69.4) 0.05 0.76 (0.58–1.01) 682 (69.6) 0.02 0.77 (0.61–0.96) C 181 (13.8) 185 (16.7) 0.05 1.25 (0.99–1.57) 143 (16.8) 0.06 1.26 (0.99–1.61) 328 (16.7) 0.03 1.25 (1.02–1.53) T 1127 (86.2) 925 (83.3) 0.05 0.80 (0.64–1.01) 707 (83.2) 0.06 0.79 (0.62–1.02) 1632 (83.3) 0.03 0.01 (0.65–0.98)

IRGM Controls CD P value OR (95% CI) UC P value OR (95% CI) IBD P value OR (95% CI) rs4958847 n ¼ 644 (%) n ¼ 539 (%) n ¼ 421 (%) n ¼ 960 (%)

AA 22 (3.4) 27 (5.0) 0.17 1.49 (0.81–2.75) 14 (3.3) 0.94 0.97 (0.47–2.01) 41 (4.5) 0.38 1.22 (0.72–2.21) GA 193 (30.0) 185 (34.2) 0.11 1.22 (0.95–1.57) 136 (32.3) 0.42 1.12 (0.85–1.47) 321 (33.5) 0.14 1.17 (0.98–1.46) GG 429 (66.6) 327 (60.7) 0.03 0.77 (0.60–0.99) 271 (64.4) 0.45 0.91 (0.69–1.18) 598 (62.0) 0.08 0.83 (0.67–1.03) G 237 (18.4) 239 (22.2) 0.02 1.26 (1.03–1.55) 164 (19.5) 0.53 1.07 (0.85–1.35) 409 (21.2) 0.05 1.19 (1.00–1.15) A 1051 (81.6) 839 (77.8) 0.02 0.79 (0.64–0.97) 678 (80.5) 0.53 0.93 (0.74–1.17) 1519 (78.8) 0.05 0.84 (0.70–1.00)

Abbreviations: CD, Crohn’s disease; CI, confidence intervals; IBD, inflammatory bowel disease; OR, odds ratio; UC, ulcerative colitis.

Genes and Immunity Association of ATG16L1 and IRGM with IBD RJ Palomino-Morales et al 358 Meta-analysis marker for UC. The lack of association with UC in all the Figure 1 shows the meta-analysis of all studies with individual studies performed to date (only Prescott published allele frequencies for ATG16L1 rs2241880 in et al.21 reported a modest significant association, CD and UC, including this study.14–16,21–34,38,39 Funnel P ¼ 0.03), including our case–control study, is probably plots and Egger tests were performed to analyse the due to the modest impact of this allele on disease publication bias of the meta-analysis. No statistically susceptibility, reliably detectable only in well-powered significant values were observed in CD studies (P ¼ 0.6) large sample sizes. In this sense, the estimated power of or UC studies (P ¼ 0.4), with symmetric funnel plots for our meta-analysis study comprising 9211 cases and both studies (Supplementary Figure 1). These data 12 863 controls is 99.99% for an OR ¼ 1.3. indicate the absence of publication bias in our meta- There was no evidence that the ATG16L1 rs2241880 analysis. Overall analysis of this variant in CD revealed polymorphism influences disease localisation, disease statistically significant evidence for heterogeneity of the behaviour, disease severity, need for surgery and age OR among the studies (I2 ¼ 50%), and the random-effects of onset for both CD and UC patients (Supplementary model was chosen. The meta-analysis showed that the Table 1). ATG16L1 rs2241880 susceptibility G allele is strongly Although none of the other studies with similar IBD associated with CD (pooled OR ¼ 1.33; combined cohorts have found significant differences with specific P ¼ 8.2 Â 10À67; Figure 1a). With regard to UC, no CD or UC subphenotypes, Prescott et al. found that the heterogeneity was observed, and the fixed-effects model ATG16L1 rs2241880 G allele was associated specifically was chosen (I2 ¼ 9.6%). The combined estimated OR was with the ileal forms of CD. However, we cannot rule out small but statistically significant (pooled OR ¼ 1.08; the possibility that our study is underpowered to detect combined P ¼ 0.0003; Figure 1b). weak associations between ATG16L1 rs2241880 and IBD Figure 2a shows the meta-analysis of IRGM rs13361189 subphenotypes. in CD, which includes the original British cohorts,17 The dynamic balance between microbes, particularly German30 and New Zealander37 replication cohorts, and commensal flora, and host defensive responses at the our Spanish population. We observed highly significant mucosal frontier has an important role in the initiation association of the susceptibility allele frequency (SAF) in and pathogenesis of IBD. This fact is supported by the the combined meta-analysis (pooled OR ¼ 1.34; com- therapeutic benefits of antibiotic and probiotic treatment bined P ¼ 1.07 Â 10À19), confirming previous reports. in IBD patients,40,41 the observation that specific enteric With regard to the meta-analysis of the IRGM flora are found more commonly in patients than in rs4958847 polymorphism, we also observed a highly controls,42,43 and studies in a variety of murine strains in significant association with CD (pooled OR ¼ 1.31; which ‘spontaneous’ colitis seems to be entirely depen- combined P ¼ 2.78 Â 10À17; Figure 2c). For UC, we dent on the presence of a luminal flora.44–46 performed a meta-analysis with the published data30,37,38 Autophagy has an important role in physiological and and our data. In agreement with our results, the meta- pathological processes, such as degradation of cytoplas- analysis of the IRGM polymorphisms detected an mic components, cellular response to nutrient starvation association with susceptibility to UC (rs13361189 pooled and apoptosis.47,48 Moreover, autophagy is also relevant OR ¼ 1.16, combined P ¼ 0.0069; rs4958847 pooled in innate49,50 and adaptive immunity.51,52 Atg16l in OR ¼ 1.13, combined P ¼ 0.014; Figures 2b and d). conjunction with an Atg12–Atg5 conjugate, has a crucial There was no statistically significant evidence for role in the elongation of isolation membranes and the heterogeneity of the OR among the studies for these initial step of autophagosome formation,53 and it is IRGM rs13361189 and rs4958847 polymorphisms in CD essential in the targeting and destruction of pathogens49 or in CU, and the fixed-effects model was chosen (I2 ¼ 0% as well as in foreign (microbial) antigen processing and for all studies). Publication bias in these studies was not presentation52 protecting against viruses, bacteria and found; the Egger test was not statistically significant, other pathogens. The functional role of the ATG16L1 with symmetric funnel plots for all of the studies T300A amino-acid substitution (polar proline to non- (Supplementary Figure 1). polar alanine) is not yet clear, but it lies at the highly evolutionary conserved position 300 in the N terminus of Gene–gene interaction the WD-repeat domain in ATG16L1 and, therefore, may To test gene–gene interaction, CD patients and controls affect protein interactions with other components of the were stratified according to their NOD2/CARD15 phe- autophagosome. This effect could result in functionally notype. There were no differences in allele frequencies altered phagosomes that would kill microbes less for the studied variants of IRGM and ATG16L1 between effectively, resulting in prolonged immune activation or 14 CARD15-positive (defined as carrying at least one incomplete pathogen clearance. IRGM is a member of NOD2/CARD15 variant) and CARD15-negative indivi- the p47 immunity-related GTPase family. It controls duals (Supplementary Table 1). intracellular pathogens by autophagy in mice and humans.54,55 Parkes et al.17 found two immediately flanking IRGM polymorphisms (rs13361189 and rs4958847) associated with CD in two different British Discussion cohorts. Further, they sequenced the whole gene and This study offers several novel insights into the role of found no non-synonymous variants associated with ATG16L1 in susceptibility to IBD. Data obtained for this CD,17 suggesting that the causal variants may lie in the first meta-analysis of ATG16L1 rs2241880 confirm that regulatory or exonic sequences of IRGM in linkage this variant is an important CD susceptibility genetic disequilibrium with the associated variants. Interest- marker. Moreover, the meta-analysis suggests that this ingly, in a very recent study McCarroll et al.56 have variant was prematurely excluded as a susceptibility shown that the causal variant responsible for IRGM

Genes and Immunity Association of ATG16L1 and IRGM with IBD RJ Palomino-Morales et al 359

Figure 1 Pooled data for the ATG16L1 rs2241880 susceptibility allele frequency (SAF). (a) Pooled data of Australian,39 Belgian,16 British,14,21,23,25 Canadian,31 Dutch,28 German,14,29,30 Italian,27,33 Hungarian,34 Japanese,22 New Zealander,26 North American15,24 and Spanish cohorts for Crohn’s disease (CD). (b) Pooled data of Australian,39 British,21,23,37 Canadian,31 Dutch,28 German,14,29,30 Italian,33 Hungarian,34 New Zealander,26,39 North American15 and Spanish cohorts for ulcerative colitis (UC). The odds ratio (OR) estimate of each study is marked with a square. The size of the square represents the weight that the corresponding study exerts in the meta-analysis. The confidence intervals (CIs) of pooled estimates are displayed as a horizontal line. Breslow–Day test, Cochran Q-test, and inconsistency for heterogeneity for the overall analysis of data from all studies: (a) Breslow–Day: w2 ¼ 39.97 (d.f. ¼ 20) P ¼ 0.005, Cochran Q: w2 ¼ 39.921 (d.f. ¼ 20) P ¼ 0.005, I2 ¼ 49.9% (95% CI ¼ 5.4–68.6%); (b) Breslow–Day: w2 ¼ 13.85 (d.f. ¼ 11) P ¼ 0.241, w2 ¼ 13.84 (d.f. ¼ 11) P ¼ 0.24.

Genes and Immunity Association of ATG16L1 and IRGM with IBD RJ Palomino-Morales et al 360 IRGM rs13361189 CD Study SAF Patients SAF Controls

Caucasian Odds ratio meta-analysis plot [fixed effect] British Pakeset al.(cohort1) 0.10 0.07 1.34 (1.18-2.02)

Pakeset al.(cohort2) 0.11 0.08 11.38 (1.16-1.67) German Franke et al. 0.10 0.07 1.44 (1.22-1.72) New Zealander 1.37 (0.97-1.94) Roberts et al. 0.08 0.06 Spanish This study 0.17 0.14 1.25 (0.99-1.57)

Combined (fixed) 1.34 (1.24-1.46)

0.51 2 5 Odds ratio (95% confidence interval)

IRGM rs13361189 UC Study SAF Patients SAF Controls

Caucasian Odds ratio meta-analysis plot [fixed effect] British Fisher et al. 0.09 0.08 1.09 (0.91-1.31) German Franke et al. 0.09 0.07 1.19 (0.97-1.46) New Zealander 1.20 (0.83-1.72) Roberts et al. 0.07 0.06 Spanish This study 0.17 0.14 1.26 (0.99-1.61)

Combined (fixed) 1.16 (1.04-1.30)

0.5 1 2 Odds ratio (95% confidence interval)

IRGM rs4958847 CD Study SAF Patients SAF Controls

Caucasian Odds ratio meta-analysis plot [fixed effect] British Pakes et al. (cohort1) 0.15 0.12 1.27 (1.14-1.30)

Pakes et al. (cohort2) 0.16 0.13 1.27 (1.00-1.47) German Franke et al. 0.15 0.11 1.42 (1.23-1.60) New Zealander 1.35 (1.00-1.82) Roberts et al. 0.1300 0.10 Spanish This study 0.22 0.18 1.26 (1.03-1.55)

Combined 1.31 (1.22-1.40) (fixed) 12 Odds ratio (95% confidence interval)

IRGM rs4958847 UC Study SAF Patients SAF Controls

Odds ratio meta-analysis plot [fixed effect] Caucasian British 1.13 (0.97-1.32) Fisher et al. 0.13 0.12 German 1.14 (0.96-1.35) Franke et al. 0.12 0.11 New Zealander 1.18 (0.87-1.61) Roberts et al. 0.11 0.10 Spanish 1.07 (0.85-1.61) This study 0.19 0.18

Combined (fixed) 1.13 (1.02-1.24)

0.5 1 2 Odds ratio (95% confidence interval) Figure 2 Pooled data for susceptibility allele frequency (SAF) of the IRGM rs13361189 and rs4958847polymorphisms. (a) rs13361189 pooled data of British,17 German,30 New Zealander35 and Spanish cohorts for Crohn’s disease (CD). (b) rs13361189 pooled data of British,38 German,30 New Zealander37 and Spanish cohorts for ulcerative colitis (UC). (c) rs4958847 pooled data of British,17 German,30 New Zealander37 and Spanish cohorts for CD. (d) Pooled data of British,38 German,30 New Zealander37 and Spanish cohorts for UC. Pooled data of the meta- analysis were obtained with both fixed methods. The odds ratio (OR) estimate of each study is marked with a square. The size of the square represents the weight that the corresponding study exerts in the meta-analysis. The confidence intervals (CIs) of pooled estimates are displayed as a horizontal line. OR and 95% CI were estimated using a fixed-effects model. Breslow–Day test for heterogeneity for the overall analysis of data from all studies: (a) Breslow–Day: w2 ¼ 3.28 (d.f. ¼ 4) P ¼ 0.51, Cochran Q: w2 ¼ 3.28 (d.f. ¼ 4) P ¼ 0.5125, I2 ¼ 0% (95% CI ¼ 0–64.1%); (b) Breslow–Day: w2 ¼ 0.88 (d.f. ¼ 3) P ¼ 0.83, Cochran Q: w2 ¼ 0.88 (d.f. ¼ 3) P ¼ 0.83, I2 ¼ 0% (95% CI ¼ 0–67.9%); (c) w2 Breslow– Day ¼ 2.04 (d.f. ¼ 4) P ¼ 0.73, Cochran Q ¼ 2.04 (d.f. ¼ 4) P ¼ 0.73, I2 ¼ 0% (95% CI ¼ 0–64.1%); (d) Breslow–Day: w2 ¼ 0.38 (d.f. ¼ 3) P ¼ 0.94, Cochran Q: w2 ¼ 0.38 (d.f. ¼ 3) P ¼ 0.94, I2 ¼ 0% (95% CI ¼ 0–67.9%).

Genes and Immunity Association of ATG16L1 and IRGM with IBD RJ Palomino-Morales et al 361 association with CD is a 20-kb deletion polymorphism, Table 3 Characteristics of the association studies between immediately upstream of IRGM and in perfect linkage ATG16L1 and IRGM and IBD disequilibrium (r2 ¼ 1.0) with the most significant CD-associated single nucleotide polymorphism Study Year of Disease Patients Controls (rs13361189). The deletion (CD risk) and reference publication (n) (n) (CD protective) haplotypes of IRGM showed distinct expression patterns that modulated cellular autophagy ATG16L1 Caucasian of internalised bacteria. These results suggest that the Australian CD association of IRGM arises from an alteration in Fowler et al.39 2008 CD1 669 1244 IRGM regulation that affects the efficacy of autophagy. CD2 154 420 We confirmed the association of both IRGM rs13361189 UC 543 1244 and rs4958847 polymorphisms with CD. Although these Belgian Libioulle et al.16 2007 CD 1813 1487 variants showed only a trend of association in our British Spanish population, a combined meta-analysis showed a Hampe et al.14 2007 CD 509 656 more significant association for both IRGM genetic Prescott et al.21 2007 CD 727 579 variants than that obtained in previous studies. UC 877 579 Similarly, our meta-analysis data demonstrated that Cummings et al.23 2007 CD 645 1190 ATG16L1 and IRGM variants are associated with UC 676 1190 Burton et al.25 2007 CD 1748 2938 susceptibility to UC. The lack of association of IRGM Fisher et al.38 2008 UC 1841 1470 with UC shown in individual studies performed to date, Canadian including our study, is probably due to the modest Okazaki et al.31 2008 CD 213 310 impact of this gene on disease susceptibility. However, UC 117 310 we recognize that the association found was weak and Dutch Weersma et al.28 2008 CD 311 893 that further studies are needed to clarify the role of IRGM UC 207 893 in UC. German In conclusion, our findings confirm ATG16L1 Hampe et al.14 2007 CD 1233 1400 rs2241880 and IRGM rs13361189 and rs4958847 poly- 2007 UC 788 1032 Glas et al.29 2008 CD 768 1615 morphisms as important markers for CD susceptibility UC 507 1615 and indicate that these ATG16L1 and IRGM genetic Franke et al.30 2008 CD 1850 1817 variants are also associated with UC. UC 1103 1817 Italian Perricone et al.27 2008 CD 163 160 Latiano et al.33 2008 CD 574 749 Patients and methods UC 664 749 Hungarian Study population Lakatos et al.34 2008 CD 266 149 Association study. We performed a case–control study, UC 149 149 comprising 982 IBD patients (557 CD and 425 UC) and North American 15 672 healthy controls recruited from three hospitals in Rioux et al. 2007 CD1 946 977 CD2 625 207 Spain: Hospital Virgen de las Nieves, Hospital Puerta del UC 353 207 Mar and Hospital Central de Asturias. Both patients and New Zealander controls were of Spanish white origin and were matched Robert et al.26 2007 CD 496 591 UC 466 591 for age and sex. Patients were diagnosed according to Scottish standard clinical, endoscopic, radiologic and histopatho- Van Limbergen et al.32 2008 CD 360 345 logic criteria.57 Information on demographics and clinical UC 495 345 parameters of IBD patients has been published else- Spanish where.58 All subjects gave written informed consent for This study CD 557 672 CU 425 672 the study. The study was approved by the local ethics Asian committees of the corresponding hospitals. Japanese Yamazaki et al.22 2007 CD 484 439 Meta-analysis. We searched Medline for all publications IRGM relating to association studies with the ATG16l1 and Caucasian IRGM genes, and we checked the references from the British retrieved publications for additional studies. Family- Parkes et al.17 2007 CD1 1748 8684 based and paediatric IBD studies were excluded from the CD2 1182 2024 meta-analysis. We included 20 studies in the meta- Fisher et al.39 2008 UC 1841 1470 analysis, 19 of which were conducted in Caucasian German Franke et al.30 2008 CD 1850 1817 populations and 1 in Asian. The characteristics of the UC 1103 1817 studies included in the meta-analysis are shown in New Zealander Table 3. Roberts et al.26 2007 CD 507 576 UC 475 576 Spanish Genotyping methods This study 2008 CD 557 672 DNA from patients and controls was obtained from CU 425 672 peripheral blood using standard methods. Samples were genotyped for ATG16L1 rs2241880 and IRGM rs13361189 Abbreviations: CD, Crohn’s disease; IBD, inflammatory bowel and rs4958847 variants using a PCR system with a disease; UC, ulcerative colitis.

Genes and Immunity Association of ATG16L1 and IRGM with IBD RJ Palomino-Morales et al 362 predeveloped TaqMan allelic discrimination assay (Ap- assessed by the method of Egger’s linear regression test, plied Biosystems, Foster City, CA, USA). Allele-specific a linear regression approach to measure funnel plot probes for ATG16L1 rs2241880 and IRGM rs13361189 and asymmetry on the natural logarithm scale of the OR.64 rs4958847 were labelled with the fluorescent dyes VIC The significance of the intercept was determined by the and FAM, respectively. PCR was carried out in a total t-test as suggested by Egger, and a P value of o0.05 was reaction volume of 5 ml with the following amplification considered significant. protocol: denaturation at 95 1C for 10 min, followed by 40 cycles of denaturation at 92 1C for 15 s and finished with annealing and extension at 60 1C for 1 min. Post-PCR, the Acknowledgements genotype of each sample was attributed automatically by We thank all the DNA donors for making this study measuring the allelic-specific fluorescence on ABI PRIM possible. This work was supported by Grant SAF2006- 7900 Sequence Detection Systems using SDS 2.3 software 00398 from Plan Nacional de I þ D þ I, and by the Junta for allelic discrimination (Applied Biosystems). Dupli- de Andalucı´a, Grant CTS-1180. cate samples and negative controls were included to check the accuracy of genotyping. The success rate of genotyping (that is, the percentage that could be References analysed) was 495% for all polymorphisms. 1 Orholm M, Munkholm P, Langholz E, Nielsen OH, Sorensen Data analysis TI, Binder V. Familial occurrence of inflammatory bowel Association study. We used the w2-test for Hardy–Wein- disease. N Engl J Med 1991; 324: 84–88. berg equilibrium and statistical analysis to compare 2 Breslin NP, Todd A, Kilgallen C, O’Morain C. Monozygotic allelic and genotypic distributions. ORs and 95% CIs twins with Crohn’s disease and ulcerative colitis: a unique 59 case report. Gut 1997; 41: 557–560. were calculated according to Woolf’s method using the 3 Hugot JP, Chamaillard M, Zouali H, Lesage S, Cezard JP, Statcalc program (Epi Info 2002, Centers for Disease Belaiche J et al. Association of NOD2 leucine-rich repeat Control and Prevention, Atlanta, GA, USA). P values less variants with susceptibility to Crohn’s disease. Nature 2001; than 0.05 were considered statistically significant. 411: 599–603. 4 Stoll M, Corneliussen B, Costello CM, Waetzig GH, Mellgard Meta-analysis. The meta-analysis was conducted using B, Koch WA et al. Genetic variation in DLG5 is associated with StatsDirect software (StatsDirect, Cheshire, UK). In all inflammatory bowel disease. Nat Genet 2004; 36: 476–480. the studies, allele and genotyping frequencies were 5 Peltekova VD, Wintle RF, Rubin LA, Amos CI, Huang Q, Gu X consistent with Hardy–Weinberg equilibrium for all the et al. Functional variants of OCTN cation transporter genes are Nat Genet 36 variants. In addition to the total group, we classified the associated with Crohn disease. 2004; : 471–475. 6 Newman B, Gu X, Wintle R, Cescon D, Yazdanpanah M, Liu X studies by ethnic and country origin. For each study, the et al. A risk haplotype in the Solute Carrier Family 22A4/22A5 allele frequencies were derived from counting the gene cluster influences phenotypic expression of Crohn’s genotypes in patients and controls. The effects of variants disease. Gastroenterology 2005; 128: 260–269. were assessed by comparing the allele frequencies 7 Yamazaki K, McGovern D, Ragoussis J, Paolucci M, Butler H, between patients vs controls. For each meta-analysis, Jewell D et al. Single nucleotide polymorphisms in TNFSF15 first the Breslow–Day and Cochran Q-tests for hetero- confer susceptibility to Crohn’s disease. Hum Mol Genet 2005; geneity were performed.60,61 In addition, the I2-test was 14: 3499–3506. used to quantify any apparent inconsistency and was 8 Picornell Y, Mei L, Taylor K, Yang H, Targan SR, Rotter JI. interpreted as the approximate proportion of total TNFSF15 is an ethnic-specific IBD gene. Inflamm Bowel Dis 2007; 13: 1333–1338. variation in study estimates that is due to heterogeneity 62 9 Duerr RH, Taylor KD, Brant SR, Rioux JD, Silverberg MS, rather than sampling error. Pooled calculations of ORs Daly MJ et al. A genome-wide association study identifies were obtained using the fixed-effects model (Mantel– IL23R as an inflammatory bowel disease gene. Science 2006; Haenszel) when heterogeneity was not observed (an I2 314: 1461–1463. value lower than 50%); otherwise a random-effects 10 Oliver J, Rueda B, Lopez-Nevot MA, Gomez-Garcia M, Martin model (DerSimonian–Laird) was chosen.63 Fixed-effects J. Replication of an association between IL23R gene poly- models assume that all studies aim to evaluate a morphism with inflammatory bowel disease. Clin Gastroenterol common truth and that results differ by chance alone. Hepatol 2007; 5: 977–981, 981 e1–2. They have the advantage of being relatively insensitive 11 Elson CO, Sartor RB, Tennyson GS, Riddell RH. Experimental Gastroenterology to small studies with exaggerated spurious effects but models of inflammatory bowel disease. 1995; 109: 1344–1367. may be inappropriate if there is genuine heterogeneity in 12 Swidsinski A, Ladhoff A, Pernthaler A, Swidsinski S, Loening- the size of genetic effects across subpopulations. Ran- Baucke V, Ortner M et al. Mucosal flora in inflammatory bowel dom-effects models anticipate that the studies may have disease. Gastroenterology 2002; 122: 44–54. genuine differences in their results; thus, they also 13 Mow WS, Vasiliauskas EA, Lin YC, Fleshner PR, Papadakis incorporate a between-study variance in their estimates. KA, Taylor KD et al. Association of antibody responses to Random-effects models are generally more conservative microbial antigens and complications of small bowel Crohn’s (that is, they provide wider confidence intervals when disease. Gastroenterology 2004; 126: 414–424. there is between-study heterogeneity). However, they 14 Hampe J, Franke A, Rosenstiel P, Till A, Teuber M, Huse K may be influenced more by small studies with exagger- et al. A genome-wide association scan of nonsynonymous 61 SNPs identifies a susceptibility variant for Crohn disease in ated spurious effects. ATG16L1. Nat Genet 2007; 39: 207–211. Publication bias was investigated by funnel plotting, in 15 Rioux JD, Xavier RJ, Taylor KD, Silverberg MS, Goyette P, which the standard error of log(OR) of each study was Huett A et al. Genome-wide association study identifies new plotted against its OR. An asymmetric plot suggested susceptibility loci for Crohn disease and implicates autophagy possible publication bias. Funnel plot asymmetry was in disease pathogenesis. Nat Genet 2007; 39: 596–604.

Genes and Immunity Association of ATG16L1 and IRGM with IBD RJ Palomino-Morales et al 363 16 Libioulle C, Louis E, Hansoul S, Sandor C, Farnir F, 33 Latiano A, Palmieri O, Valvano MR, D’Inca R, Cucchiara S, Franchimont D et al. Novel Crohn disease locus identified Riegler G et al. Replication of interleukin 23 receptor and by genome-wide association maps to a gene desert on autophagy-related 16-like 1 association in adult- and pedia- 5p13.1 and modulates expression of PTGER4. PLoS Genet tric-onset inflammatory bowel disease in Italy. World J 2007; 3: e58. Gastroenterol 2008; 14: 4643–4651. 17 Parkes M, Barrett JC, Prescott NJ, Tremelling M, Anderson 34 Lakatos PL, Szamosi T, Szilvasi A, Molnar E, Lakatos L, CA, Fisher SA et al. Sequence variants in the autophagy gene Kovacs A et al. ATG16L1 and IL23 receptor (IL23R) genes are IRGM and multiple other replicating loci contribute to associated with disease susceptibility in Hungarian CD Crohn’s disease susceptibility. Nat Genet 2007; 39: 830–832. patients. Dig Liver Dis 2008; 40: 867–873. 18 Mizushima N, Noda T, Ohsumi Y. Apg16p is required for the 35 Taylor GA, Feng CG, Sher A. p47 GTPases: regulators of function of the Apg12p-Apg5p conjugate in the yeast immunity to intracellular pathogens. Nat Rev Immunol 2004; 4: autophagy pathway. EMBO J 1999; 18: 3888–3896. 100–109. 19 Suzuki K, Kirisako T, Kamada Y, Mizushima N, Noda T, 36 MacMicking JD. Immune control of phagosomal bacteria by Ohsumi Y. The pre-autophagosomal structure organized by p47 GTPases. Curr Opin Microbiol 2005; 8: 74–82. concerted functions of APG genes is essential for autophago- 37 Roberts RL, Hollis-Moffatt JE, Gearry RB, Kennedy MA, some formation. EMBO J 2001; 20: 5971–5981. Barclay ML, Merriman TR. Confirmation of association of 20 Klionsky DJ, Cregg JM, Dunn Jr WA, Emr SD, Sakai Y, IRGM and NCF4 with ileal Crohn’s disease in a population- Sandoval IV et al. A unified nomenclature for yeast auto- based cohort. Genes Immun 2008; 9: 561–565. phagy-related genes. Dev Cell 2003; 5: 539–545. 38 Fisher SA, Tremelling M, Anderson CA, Gwilliam R, 21 Prescott NJ, Fisher SA, Franke A, Hampe J, Onnie CM, Soars D Bumpstead S, Prescott NJ et al. Genetic determinants of et al. A nonsynonymous SNP in ATG16L1 predisposes to ileal ulcerative colitis include the ECM1 locus and five Crohn’s disease and is independent of CARD15 and IBD5. loci implicated in Crohn’s disease. Nat Genet 2008; 40: Gastroenterology 2007; 132: 1665–1671. 710–712. 22 Yamazaki K, Onouchi Y, Takazoe M, Kubo M, Nakamura Y, 39 Fowler EV, Doecke J, Simms LA, Zhao ZZ, Webb PM, Hata A. Association analysis of genetic variants in IL23R, Hayward NK et al. ATG16L1 T300A shows strong associations ATG16L1 and 5p13.1 loci with Crohn’s disease in Japanese with disease subgroups in a large Australian IBD population: patients. J Hum Genet 2007; 52: 575–583. further support for significant disease heterogeneity. Am J 23 Cummings JR, Cooney R, Pathan S, Anderson CA, Barrett JC, Gastroenterol 2008; 103: 2519–2526. Beckly J et al. Confirmation of the role of ATG16L1 as a 40 Gionchetti P, Rizzello F, Helwig U, Venturi A, Lammers KM, Crohn’s disease susceptibility gene. Inflamm Bowel Dis 2007; Brigidi P et al. Prophylaxis of pouchitis onset with probiotic 13: 941–946. therapy: a double-blind, placebo-controlled trial. Gastroenter- 24 Baldassano RN, Bradfield JP, Monos DS, Kim CE, Glessner JT, ology 2003; 124: 1202–1209. Casalunovo T et al. Association of the T300A non-synonymous 41 Sutherland L, Singleton J, Sessions J, Hanauer S, Krawitt E, variant of the ATG16L1 gene with susceptibility to paediatric Rankin G et al. Double blind, placebo controlled trial of Crohn’s disease. Gut 2007; 56: 1171–1173. metronidazole in Crohn’s disease. Gut 1991; 32: 1071–1075. 25 Burton PR, Clayton DG, Cardon LR, Craddock N, Deloukas P, 42 Barnich N, Carvalho FA, Glasser AL, Darcha C, Jantscheff P, Duncanson A et al. Association scan of 14,500 nonsynonymous Allez M et al. CEACAM6 acts as a receptor for adherent- SNPs in four diseases identifies autoimmunity variants. Nat invasive E. coli, supporting ileal mucosa colonization in Crohn Genet 2007; 39: 1329–1337. disease. J Clin Invest 2007; 117: 1566–1574. 26 Roberts RL, Gearry RB, Hollis-Moffatt JE, Miller AL, Reid J, 43 Darfeuille-Michaud A, Neut C, Barnich N, Lederman E, Abkevich V et al. IL23R R381Q and ATG16L1 T300A are Di Martino P, Desreumaux P et al. Presence of adherent strongly associated with Crohn’s disease in a study of Escherichia coli strains in ileal mucosa of patients with New Zealand Caucasians with inflammatory bowel disease. Crohn’s disease. Gastroenterology 1998; 115: 1405–1413. Am J Gastroenterol 2007; 102: 2754–2761. 44 Elson CO, Cong Y, McCracken VJ, Dimmitt RA, Lorenz RG, 27 Perricone C, Borgiani P, Romano S, Ciccacci C, Fusco G, Weaver CT. Experimental models of inflammatory bowel Novelli G et al. ATG16L1 Ala197Thr is not associated with disease reveal innate, adaptive, and regulatory mechanisms susceptibility to Crohn’s disease or with phenotype in an of host dialogue with the microbiota. Immunol Rev 2005; 206: Italian population. Gastroenterology 2008; 134: 368–370. 260–276. 28 Weersma RK, Zhernakova A, Nolte IM, Lefebvre C, Rioux JD, 45 Onderdonk AB, Hermos JA, Bartlett JG. The role of the Mulder F et al. ATG16L1 and IL23R are associated with intestinal microflora in experimental colitis. Am J Clin Nutr inflammatory bowel diseases but not with celiac disease in the 1977; 30: 1819–1825. Netherlands. Am J Gastroenterol 2008; 103: 621–627. 46 Sellon RK, Tonkonogy S, Schultz M, Dieleman LA, Grenther 29 Glas J, Konrad A, Schmechel S, Dambacher J, Seiderer J, W, Balish E et al. Resident enteric bacteria are necessary for Schroff F et al. The ATG16L1 gene variants rs2241879 and development of spontaneous colitis and immune system rs2241880 (T300A) are strongly associated with susceptibility activation in interleukin-10-deficient mice. Infect Immun 1998; to Crohn’s disease in the German population. Am J Gastro- 66: 5224–5231. enterol 2008; 103: 682–691. 47 Levine B. Eating oneself and uninvited guests: autophagy- 30 Franke A, Balschun T, Karlsen TH, Hedderich J, May S, Lu T related pathways in cellular defense. Cell 2005; 120: 159–162. et al. Replication of signals from recent studies of Crohn’s 48 Shintani T, Klionsky DJ. Autophagy in health and disease: a disease identifies previously unknown disease loci for double-edged sword. Science 2004; 306: 990–995. ulcerative colitis. Nat Genet 2008; 40: 713–715. 49 Nakagawa I, Amano A, Mizushima N, Yamamoto A, 31 Okazaki T, Wang MH, Rawsthorne P, Sargent M, Datta LW, Yamaguchi H, Kamimoto T et al. Autophagy defends cells Shugart YY et al. Contributions of IBD5, IL23R, ATG16L1, and against invading group A Streptococcus. Science 2004; 306: NOD2 to Crohn’s disease risk in a population-based case- 1037–1040. control study: evidence of gene-gene interactions. Inflamm 50 Ogawa M, Yoshimori T, Suzuki T, Sagara H, Mizushima N, Bowel Dis 2008; 14: 1528–1541. Sasakawa C. Escape of intracellular Shigella from autophagy. 32 Van Limbergen J, Russell RK, Nimmo ER, Drummond HE, Science 2005; 307: 727–731. Smith L, Anderson NH et al. Autophagy gene ATG16L1 51 Gutierrez MG, Master SS, Singh SB, Taylor GA, Colombo MI, influences susceptibility and disease location but not child- Deretic V. Autophagy is a defense mechanism inhibiting BCG hood-onset in Crohn’s disease in Northern Europe. Inflamm and Mycobacterium tuberculosis survival in infected macro- Bowel Dis 2008; 14: 338–346. phages. Cell 2004; 119: 753–766.

Genes and Immunity Association of ATG16L1 and IRGM with IBD RJ Palomino-Morales et al 364 52 Paludan C, Schmid D, Landthaler M, Vockerodt M, Kube D, 57 Lennard-Jones JE. Classification of inflammatory bowel disease. Tuschl T et al. Endogenous MHC class II processing of a viral Scand J Gastroenterol Suppl 1989; 170: 2–6; discussion 16-9. nuclear antigen after autophagy. Science 2005; 307: 593–596. 58 Oliver J, Marquez A, Gomez-Garcia M, Martinez A, Mendoza 53 Mizushima N, Kuma A, Kobayashi Y, Yamamoto A, JL, Vilchez JR et al. Association of the macrophage migration Matsubae M, Takao T et al. Mouse Apg16L, a novel WD- inhibitory factor gene polymorphisms with inflammatory repeat protein, targets to the autophagic isolation membrane bowel disease. Gut 2007; 56: 150–151. with the Apg12-Apg5 conjugate. J Cell Sci 2003; 116 (Pt 9): 59 Woolf B. On estimating the relation between blood group and 1679–1688. disease. Ann Hum Genet 1955; 19: 251–253. 54 Collazo CM, Yap GS, Sempowski GD, Lusby KC, Tessarollo L, 60 Breslow NE, Day NE. Statistical methods in cancer research. Woude GF et al. Inactivation of LRG-47 and IRG-47 reveals a Volume I—The analysis of case-control studies. IARC Sci Publ family of interferon gamma-inducible genes with essential, 1980; 32: 5–338. pathogen-specific roles in resistance to infection. J Exp Med 61 Lau J, Ioannidis JP, Schmid CH. Quantitative synthesis in 2001; 194: 181–188. systematic reviews. Ann Intern Med 1997; 127: 820–826. 55 Singh SB, Davis AS, Taylor GA, Deretic V. Human IRGM 62 Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring induces autophagy to eliminate intracellular mycobacteria. inconsistency in meta-analyses. BMJ 2003; 327: 557–560. Science 2006; 313: 1438–1441. 63 DerSimonian R, Laird N. Meta-analysis in clinical trials. 56 McCarroll SA, Huett A, Kuballa P, Chilewski SD, Landry A, Control Clin Trials 1986; 7: 177–188. Goyette P et al. Deletion polymorphism upstream of IRGM 64 Egger M, Davey Smith G, Schneider M, Minder C. Bias in associated with altered IRGM expression and Crohn’s disease. meta-analysis detected by a simple, graphical test. BMJ 1997; Nat Genet 2008; 40: 1107–1112. 315: 629–634.

Supplementary Information accompanies the paper on Genes and Immunity website (http://www.nature.com/gene)

Genes and Immunity