Genetic variation in IBD: progress, clues to pathogenesis and possible clinical utility

Abstract

Epidemiological and clinical studies have suggested that the pathogenesis of inflammatory bowel disease (IBD) is strongly influenced by genetic predisposition. Beyond the limitations of linkage analysis, multiple genome-wide association studies, their meta-analyses, and targeted genotyping array techniques have broadened our understanding of the genetic architecture of IBD. Currently, over 200 single nucleotide polymorphisms are known to be associated with susceptibility to IBD and through functional analysis of genes and loci, a substantial proportion of pathophysiologic mechanisms have been revealed. However, because only a modest fraction of predicted heritability can be explained by known genes/loci, additional strategies are needed including the identification of rare variants with large effect sizes to help explain the missing heritability. Considerable progress is also being made on applying outcomes of genetic research in diagnostics, classification, prognostics, and the development of new therapeutics of IBD.

1. Introduction

The inflammatory bowel diseases (IBD), including Crohn’s disease (CD) and ulcerative colitis (UC), are chronic inflammatory diseases of the gastrointestinal tract of unknown pathogenesis. As described in the landmark article by Crohn et al., familial aggregation of IBD implicates genetic background in the development of IBD.¹ Although not yet fully understood, dysregulated mucosal immune response to microbes in a genetically susceptible individual is thought to be pathogenic mechanism of IBD.² It is hoped that genetic studies will possibly answer questions including which individuals are destined to suffer from IBD and which IBD patients will more likely suffer a disabling course of disease. Unlike classical Mendelian disorders, IBD is genetically complex disorder, where traditional genetic analytics are not able to shape the real features of disease. However, with rapid technologic developments, such as gene chip and computational/statistical techniques over the past a few decades, very significant progress has been made in our understanding of the genetic architecture of IBD. This review will cover both historical and the current status of IBD genetic research, as well as identified genes/loci associated with IBD, and potential clinical application of our knowledge on IBD genetics.

2. History of genetic research for IBD

Epidemiologic observations showing clear familial clustering of IBD and higher risk of CD in Jews, especially in Ashkenazi Jews prompted researchers to be interested in heritability and genetic risk of IBD.^3,4 In addition, when combing 6 twin studies from Europe, the concordance rates of 30.3% in 112 monozygotic twins vs. 3.6% in 196 dizygotic twins for CD and 15.4% in 143 monozygotic twins vs. 3.9% in 206 dizygotic twins for UC supported further the impact of genetics in IBD risk.⁵ According to previous genetic epidemiologic studies, a lifetime risk of developing IBD for first-degree relatives of a CD patient was estimated to be 4.8%–5.2% for non-Jews and 7.8% for Jews.^3,6,7 The corresponding figures for first-degree relatives of a patient with UC are 1.6% for non-Jews and 5.2% for Jews.³ Accordingly, the familial aggregation of IBD, the observed concordance in twin studies, and increased risk of developing IBD in relatives of proband triggered molecular genetic studies on IBD.

3. Linkage studies

Due to the technical limitation and high cost of sequencing, early IBD genetic studies were inevitably restricted to studying a number of genetic variants in a small number of individuals.⁸ Therefore, the earliest studies were designed around understanding the patterns of disease inheritance within a specific family having diseases of interest. By tracing the DNA segments that segregated depending on disease status within families, sections of the genome that were likely to confer risk to the specific disease could be identified.⁸ This approach called linkage analysis was useful for detecting variants with high penetrance that segregate well with disease status.⁸ In 1996, the first linkage study in IBD identified a portion on chromosome 16 (IBD1 locus) with CD,⁹ a finding supported in subsequent studies.^10–15 Further studies also identified and replicated areas of significant linkage on additional chromosomes which were designated as IBD2-IBD9.¹⁶ A natural extension of identifying these regions was to perform fine-mapping and through this approach CARD15/NOD2 was identified as the underlying gene at IBD1 in 2001.¹⁷ At the same time, two independent studies, using a candidate gene approach, also identified the role of CARD15/NOD2 in CD susceptibility.¹⁸ Three single nucleotide polymorphisms (SNPs), R702W, G908R, and L1007fsinsC, were revealed to represent 81% of disease-causing mutations within CARD15/NOD2.¹⁹ Further replication studies confirmed that these three SNPS are independently associated with CD.^20–22 Identification of CARD15/NOD2 remained one of the few robustly replicated genetic risk loci discovered through following up linkage analysis signals, not just in IBD, but across common diseases.⁸ The widespread failure of linkage analyses, in general, strongly suggested that common diseases do not have a single, highly penetrant genetic causes, but that they are likely driven by the accumulation of multiple risk factors of only modest effect (the ‘common disease, common variant’ hypothesis).²³ Discovering genetic associations via linkage analysis under this scenario is very difficult, as the genetic risk may be spread throughout the genome rather than concentrated in a single locus.²³ Therefore, alternative association analyses approaches which test if the population-level allele frequencies of cases and controls are statistically different, were thought to be much more powerful because it would be possible to select the right variant to test among millions of variants existing in human population.^8,23

4. Genome-wide association studies

The development of publicly available databases such as the SNP Consortium and the International HapMap Consortium, that contain increasingly comprehensive information about SNPs across the genome, together with an increased understanding of linkage disequilibrium structures in humans as well as the development of new microarray technologies leading to genome-wide SNP chip, opened the way to genome-wide association studies (GWAS).^8,24–27 GWAS compare the allele frequency of a particular variant between unrelated cases and controls. These new approaches were correctly hailed as the next step to unravel the genetic architecture of complex diseases like IBD. GWAS-based approaches avoid underlying assumptions for biological or positional candidate loci, genes, and variants.²⁸ Therefore, GWAS have been labeled as a “hypothesis-free or unbiased” approach, overcoming the limitations imposed by our incomplete understanding of the pathophysiology of complex diseases.²⁸ In the first GWAS for IBD, Yamazaki and colleagues explored 72,738 SNPs and identified several associated SNPs with CD in Japanese populations at the TNFSF15 gene, a finding replicated in European cohorts.²⁹ In the first European ancestry GWAS study, CD-risk variants were identified in the interleukin 23 receptor (IL23R) gene.³⁰ Further independent replication studies confirmed IL23R CD associations and also extended the association to UC.³⁰ A study adopting a slightly different approach, through studying non-synonymous variants across the genome, identified an association to a protein-coding variant in ATG16L1 with CD, providing the first evidence for the importance of autophagy in CD.³¹ GWAS approaches also revealed a pair of associations on chromosomes 5p13 and 10q21 located in gene deserts, thereby suggesting the important role of regulatory and non-coding elements in CD.^32,33 Further GWAS have shown association of other genes related with innate (TLR4, STAT3, NKX2-3, CARD9) and adaptive (TNFSF15, PTPN2, IL-12B, IRF5) immune response pathways and highlight the role of autophagy and intracellular bacterial handing (CARD15/NOD2, ATG16L1, IRGM) in CD.^8,34 These initial CD studies also suggested a partial overlap between CD and other immune-related disorders. Around 30% of associated variants in these initial studies on CD were shared with UC, while close to 50% of loci were revealed to be shared with at least one other immune-mediated disease such as type 1 diabetes, celiac disease or rheumatoid arthritis.^35,36 Additional GWAS in UC cohorts lead to the discovery of multiple novel UC-specific loci.^37–40 Three loci associated with biologically relevant candidate genes, HNF4A, CDH1, and LAMB1, suggested a possible role of epithelial barrier defects in UC pathogenesis.^8,39 GWAS on UC also confirmed the association between UC and human leukocyte antigen (HLA) locus.⁴¹ In contrast to UC, where several variants in HLA-B make the largest contribution to genetic risk (ORs 1.4–1.5), genes in the HLA region confer only a modest effect on CD risk (ORs 1.1–1.2).^36,42

5. Current status of genetic research for IBD

5.1 GWAS meta-analysis

Although early GWAS identified multiple loci associated with CD and UC, thereby generating new biological hypothesis for IBD, the relatively weak associations only explained a fraction of the heritability expected from twin studies.⁸ This ‘missing heritability’ was partially attributed to types of variation not adequately captured by GWAS (e.g. non-European ethnicity, as well as rare and structural variations).^8,43–45 In addition, it was recognized that a substantial number of additional and yet-unidentified common variants with even smaller effect size (e.g. ORs < 1.2 or even 1.1) than those identified by early GWAS, requiring much larger sample sizes, existed.^8,36,44 The need for large sample sizes re-energized The International IBD Genetics Consortium (IIBDGC) (http://www.ibdgenetics.org/) with the aim of bringing together investigators and GWAS datasets from IBD genetics groups around the world in order to search for variants with small effect size not detected by underpowered GWAS.^8,36 The first meta-analysis combined data from over 13,000 individuals from three previously published GWAS and identified 21 new CD loci including another autophagy gene, LRRK2.⁴⁶ Two years later, a second CD meta-analysis of six GWAS with a total sample size of over 50,000 individuals identified 30 new loci, bringing the total number of CD susceptibility loci to 71, explaining 23.3% of the estimated heritability for CD in European ancestry populations.⁴⁷ The first GWAS meta-analysis of UC patients combined 3 studies in a discovery set and performed replication in the independent population.⁴⁸ As a result, thirteen novel loci were identified and multiple previously reported UC- and CD-associated loci were confirmed, increasing the number of UC loci to approximately 30.⁴⁸ Additional meta-analysis of six UC GWAS datasets, comprising 48,950 individuals, identified 29 additional risk loci with genome-wide significance, increasing the number of UC-associated loci to 47, with an estimated 16% of heritability explained.⁴⁹ The recent meta-analysis of 15 GWAS for CD and/or UC as well as additional typing on the Immunochip (totalling over 75,000 subjects) identified 71 novel IBD loci, increasing the number of IBD-associated loci to 163 (110 associated with both CD and UC, 30 CD-specific, and 23 UC-specific).⁴² Out of 53 disease-specific loci, 43 showed the same direction of effect in both CD and UC, suggesting that nearly all of the biological mechanisms involved in one disease play some role in the other.^4,42 Collectively, multiple pathways were implicated as IBD-related mechanisms, including those involved in innate immunity, JAK/STAT signaling pathway, cytokine production (interferon-gamma, IL-12, tumor-necrosis-factor-alpha, and IL-10 signaling) and lymphocyte activation.^8,42 Approximately 70% of IBD risk loci (113 out of 163) were revealed to be shared with other complex diseases or traits, including 66 loci shared with other immune-mediated disorders (especially ankylosing spondylitis and psoriasis).⁴² Moreover, six out of 8 genes linked to Mendelian susceptibility to mycobacterial disease overlapped with IBD and seven out of 8 loci known to be associated with leprosy by GWAS were also shared with IBD.⁴² These overlaps suggest that selection pressures arising from mycobacterial infection may have shaped the genetic architecture of IBD.^36,42

5.2 IBD genetic studies in non-European populations

Historically, GWAS in IBD had centered around subjects of European ancestry with the exception of the first ever GWAS which was a Japanese CD study.²⁹ Although TNFSF15 was identified as a CD gene in Japanese, previous replication studies have shown lack of common CARD15/NOD2 variants in Japanese CD patients.^29,50,51 This lack of association between CARD15/NOD2 variants and CD in Japanese was replicated in Koreans.^52,53 Replication studies on the association between IL23R and IBD have shown conflicting results. A Japanese study showed that none of the 10 IL23R SNPs from the original study by Duerr et al. was associated with CD.⁵⁴ However, in a Korean study, two variants (rs1004819 and rs1495465) were associated with CD.⁵⁵ Of the 35 known IL23R SNPs studied in Han Chinese cohort, only one non-synonymous SNP, rs11465788 (C>T), was associated with CD.⁵⁶ This association had not been identified in previous studies but was replicated in a subsequent Korean study.⁵⁷ Following the first GWAS in Asians,²⁹ subsequent GWAS on Japanese and Korean CD patients demonstrated that some risk loci for CD are shared between East Asians and European ancestry populations, while some appear to be peculiar to the East Asian populations including a loci on chromosome 4p14, 10q25, in the ATG16L2-FCHSD2 region on11q13, and in the SLC25A15-ELF1-WBP4 region on 13q14.^58,59 GWAS on Asian UC patients have shown a tendency to more extensive genetic overlap with European ancestry populations than that seen in CD.^60,61 Recent GWAS from northern India also identified shared contribution of a proportion of UC-susceptibility genes between northern Indians and Europeans as well as identifying 3 HLA-independent loci.⁶² In a similar fashion to studies performed on European populations, researchers have utilized other platforms in East Asians including a study utilizing the Immunochip in Koreans, which identified six additional CD loci thereby increasing the total explained genetic variance for CD from 5.31% to 7.27%.⁶³ A similar approach in UC has increased the risk loci to 13 including 3 previously reported by GWAS in Koreans.⁶⁴ Immunochip analysis of African Americans showed overlaps with Caucasian and Asian studies, through replication of 5 (IL23R, FCGR2A, PTGER4, CARD15/NOD2, and IKZF3) out of 163 SNPs from the Caucasian study and showing the strongest associations between UC and HLA rs9271366, replicating an association previously observed in both Japanese and Korean UC GWAS.^42,61,65,66 Most recently, the first trans-ethnic association study of both CD and UC by the IIBDGC, with GWAS or Immunochip data on 86,640 Europeans and Immunochip data from 9,846 individuals of East Asian, Indian or Iranian descent was published.⁶⁷ As a result, 38 newly associated loci (27 with both CD and UC, 7 specific to CD, and 4 specific to UC) were identified, increasing the number of known IBD risk loci to 231 independent SNPs within 200 loci.⁶⁷ Among those, the majority were shared across diverse ancestry groups, with only a handful demonstrating population-specific effects driven by heterogeneity in risk allele frequency (for example, CARD15/NOD2) or effect size (for example, TNFSF15-TNFSF8).⁶⁷ Still, the number of genes/loci associated with IBD in Asian population is currently fewer than that of Western patients due, at least in part, to the smaller studies performed to date in Asians and other populations. It was suggested that genetic factors could play a critical role in shaping the microbiota in IBD patients as well as healthy subjects.^68,69 Our speculation for a lower incidence of IBD in Asians is that differences in genetic architecture between Western and Eastern IBD populations could result in differences in microbial milieu, thereby causing different incidences of IBD depending on ethnicity. However, regarding host genetics, microbiome, their crosstalk, and the contribution of other environmental factors, more extensive research is needed, especially in Asians. The CD, UC, or IBD-associated loci revealed by GWAS, GWAS meta-analysis, and Immunochip analysis are summarized in Table 1. The number of IBD-associated loci identified among various ethnic groups over the past 15 years is presented in Figure 1.

Table 1.

Summary of significant CD, UC, and IBD-associated loci by GWAS, GWAS meta-analysis, and Immunochip analysis

SNP	Key genes (+ no. additional genes in locus)*	Associated disease	European P-value	European OR	95% CI	Source(s)
rs2066847	NOD2, ADCY7, (5)	CD	5.86E-209	3.103	1.497–1.618	Jostins et al.†
rs12994997	ATG16L1, INPP5D, (7)	CD	4.14E-70	1.233	1.193–1.274	Jostins et al.
rs9264942	HLA-C, PSORS1C1, NFKBIL1, MICB, (18)	CD	4.96E-28	1.145	1.107–1.184	Jostins et al.
rs9286879	FASLG, TNFSF18, (0)	CD	5.53E-22	1.125	1.083–1.167	Jostins et al.
rs2024092	GPX4, HMHA1, (20)	CD	8.26E-22	1.156	1.112–1.201	Jostins et al.
rs3764147	LACC1, (3)	CD	2.19E-21	1.155	1.112–1.199	Jostins et al.
rs2945412	LGALS9, NOS2, (3)	CD	8.68E-17	1.137	1.1–1.175	Jostins et al.
rs6716753	SP140, (5)	CD	1.17E-16	1.134	1.089–1.18	Jostins et al.
rs6651252	0	CD	1.45E-16	1.185	1.128–1.246	Jostins et al.
rs2284553	IFNGR2, IFNAR1, IFNAR2, IL10RB, GART, TMEM50B, (6)	CD	2.14E-16	1.123	1.086–1.162	Jostins et al.
rs1728918	UCN, (23)	CD	4.86E-16	1.123	1.086–1.16	Jostins et al.
rs516246	DBP, SPHK2, IZUMO1, FUT2, (22)	CD	1.00E-15	1.107	1.071–1.143	Jostins et al.
rs6679677	PTPN22, DCLRE1B, (7)	CD	2.03E-15	1.196	1.129–1.268	Jostins et al.
rs13204742	(2)	CD	8.38E-15	1.173	1.118–1.23	Jostins et al.
rs212388	TAGAP, (5)	CD	3.04E-14	1.105	1.069–1.141	Jostins et al.
rs13126505	(1)	CD	1.84E-12	1.172	1.1–1.248	Jostins et al.
rs10065637	IL6ST, IL31RA, (1)	CD	3.68E-12	1.123	1.079–1.17	Jostins et al.
rs12663356	(3)	CD	4.01E-12	1.095	1.06–1.131	Jostins et al.
rs3897478	ADAM30, (5)	CD	1.97E-11	1.161	1.101–1.224	Jostins et al.
rs11681525	-	CD	4.08E-11	0.86	0.83–0.90	Liu et al.‡
rs3853824	-	CD	1.17E-10	0.92	0.9–0.94	Liu et al.
rs7555082	PTPRC	CD	1.47E-10	1.13	1.09–1.18	Liu et al.
rs727563	TEF, NHP2L1, PMM1, L3MBTL2, CHADL	CD	1.88E-10	1.10	1.07–1.13	Liu et al.
rs4802307	(9)	CD	2.00E-10	1.099	1.06–1.139	Jostins et al.
rs9491697	(3)	CD	3.79E-10	1.077	1.042–1.112	Jostins et al.
rs7702331	(4)	CD	5.63E-10	1.088	1.05–1.126	Jostins et al.
rs10865331	(3)	CD	9.77E-10	1.098	1.062–1.134	Jostins et al.
rs35320439	PDCD1, ATG4B	CD	9.89E-10	1.09	1.06–1.12	Liu et al.
rs7773324	IRF4, DUSP22	CD	1.06E-09	0.92	0.90–0.95	Liu et al.
rs7954567	CD27, TNFRSF1A, LTBR	CD	1.30E-09	1.09	1.06–1.11	Liu et al.
rs9525625	AKAP1, TFSF11	CD	1.41E-09	1.08	1.05–1.10	Liu et al.
rs17391694	(5)	CD	2.96E-09	1.134	1.077–1.194	Jostins et al.
rs864745	CREB5, JAZF1,(1)	CD	3.65E-09	1.087	1.052–1.123	Jostins et al.
rs16967103	RASGRP1, SPRED1, (2)	CD	3.88E-09	1.088	1.045–1.132	Jostins et al.
rs10798069	PTGS2, PLA2G4A	CD	4.25E-09	0.93	0.91–0.95	Liu et al.
rs17695092	CPEB4, (2)	CD	4.68E-09	1.095	1.055–1.136	Jostins et al.
rs7758080	MAP3K7IP2	CD	7.27E-09	1.08	1.05–1.11	Liu et al.
rs7236492	NFATC1, TST	CD	9.09E-09	0.91	0.87–0.94	Liu et al.
rs7015630	RIPK2, (4)	CD	1.42E-08	1.075	1.035–1.116	Jostins et al.
rs6837335	TXK, TEC, SLC10A4, (3)	CD	1.75E-08	1.086	1.049–1.123	Jostins et al.
rs13204048	-	CD	2.89E-08	0.93	0.91–0.96	Liu et al.
rs10486483	(2)	CD	3.48E-08	1.089	1.048–1.13	Jostins et al.
rs1748195	USP1	CD	7.13E-08	1.07	1.04–1.10	Liu et al.
rs9319943	-	CD	9.05E-07	1.08	1.05–1.11	Liu et al.
rs724016	-	CD	3.36E-06	1.06	1.03–1.08	Liu et al.
rs6927022	HLA-DQB1, HLA-DRB1, HLA-DQA1, HLA-DRA, (12)	UC	4.71E-133	1.444	1.387–1.503	Jostins et al.
rs6426833	(9)	UC	2.39E-68	1.265	1.221–1.31	Jostins et al.
rs6017342	ADA, HNF4A, (9)	UC	1.43E-43	1.228	1.185–1.273	Jostins et al.
rs4380874	DLD, (9)	UC	2.07E-26	1.137	1.097–1.177	Jostins et al.
rs2816958	(3)	UC	1.98E-17	1.23	1.161–1.302	Jostins et al.
rs561722	FAM55A, FAM55D, (5)	UC	5.15E-17	1.12	1.079–1.163	Jostins et al.
rs798502	CARD11, GNA12, TTYH3, (4)	UC	6.09E-17	1.127	1.084–1.171	Jostins et al.
rs4728142	IRF5, TNPO3, TSPAN33, (11)	UC	4.37E-14	1.104	1.066–1.143	Jostins et al.
rs17229285	0	UC	1.73E-13	1.117	1.079–1.157	Jostins et al.
rs10797432	TNFRSF14, MMEL1, PLCH2, (8)	UC	2.62E-12	1.078	1.041–1.116	Jostins et al.
rs3774959	NFKB1, MANBA, (2)	UC	3.66E-12	1.118	1.077–1.159	Jostins et al.
rs2189234	-	UC	1.95E-10	1.08	1.06–1.11	Liu et al.
rs11150589	ITGAL, (20)	UC	6.04E-10	1.09	1.052–1.128	Jostins et al.
rs113010081	FLJ78302, LTF, CCR1, CCR2, CCR3, CCR5	UC	9.02E-10	1.14	1.09–1.18	Liu et al.
rs1126510	CALM3, (14)	UC	1.55E-09	1.075	1.037–1.113	Jostins et al.
rs7210086	(3)	UC	1.89E-09	1.111	1.062–1.163	Jostins et al.
rs254560	(6)	UC	2.55E-09	1.056	1.019–1.093	Jostins et al.
rs1077773	AHR	UC	5.96E-09	0.93	0.91–0.95	Liu et al.
rs9847710	PRKCD, ITIH4, (8)	UC	1.05E-08	1.064	1.027–1.102	Jostins et al.
rs483905	JRKL, MAML2, (2)	UC	1.21E-08	1.056	1.017–1.096	Jostins et al.
rs11739663	SLC9A3, (8)	UC	1.81E-08	1.071	1.027–1.117	Jostins et al.
rs4722672	(14)	UC	2.06E-08	1.091	1.043–1.14	Jostins et al.
rs6088765	PROCR, UQCC, CEP250, (8)	UC	2.21E-08	1.079	1.041–1.117	Jostins et al.
rs28374715	ITPKA, NDUFAF1, NUSAP1, (8)	UC	2.43E-08	1.082	1.04–1.126	Jostins et al.
rs1016883	RFTN2, PLCL1, (7)	UC	2.87E-08	1.1	1.051–1.15	Jostins et al.
rs1728785	ZFP90, (6)	UC	3.71E-08	1.075	1.031–1.121	Jostins et al.
rs17736589	-	UC	4.34E-08	1.09	1.05–1.12	Liu et al.
rs11583043	SLC30A, EDG1	UC	6.05E-08	1.08	1.05–1.11	Liu et al.
rs7011507	-	UC	6.40E-08	0.90	0.87–0.94	Liu et al.
rs3116494	ICOS, CD28, CTLA4	UC	1.30E-07	1.08	1.05–1.11	Liu et al.
rs616597	NFKBIZ	UC	9.34E-06	0.93	0.91–0.96	Liu et al.
rs11209026	IL23R, IL12RB2, (4)	IBD	8.12E-161	2.013	1.885–2.15	Jostins et al.
rs11742570	PTGER4, (1)	IBD	1.81E-82	1.198	1.164–1.234	Jostins et al.
rs10781499	CARD9, PMPCA, SDCCAG3, INPP5E, (19)	IBD	4.38E-56	1.188	1.154–1.222	Jostins et al.
rs4409764	NKX2–3, (6)	IBD	1.03E-54	1.182	1.149–1.217	Jostins et al.
rs2188962	IRF1, IL13, CSF2, SLC22A4, IL4, IL3, IL5, PDLIM4, SLC22A5, ACSL6, (8)	IBD	1.35E-52	1.158	1.125–1.191	Jostins et al.
rs2836878	(3)	IBD	4.62E-48	1.18	1.142–1.219	Jostins et al.
rs3197999	MST1, PFKFB4, MST1R, UCN2, GPX1, IP6K2, BSN, IP6K1, USP4, (56)	IBD	1.01E-47	1.18	1.144–1.216	Jostins et al.
rs10761659	(3)	IBD	6.37E-46	1.166	1.134–1.2	Jostins et al.
rs10758669	JAK2, (4)	IBD	7.88E-45	1.174	1.139–1.209	Jostins et al.
rs6871626	IL12B, (3)	IBD	1.43E-42	1.181	1.146–1.216	Jostins et al.
rs3024505	IL10, IL20, IL19, IL24, PIGR, MAPKAPK2, FAIM3, RASSF5, (3)	IBD	6.66E-42	1.208	1.163–1.254	Jostins et al.
rs1801274	FCGR2A, FCGR2B, FCGR3A, HSPA6, FCGR3B, FCRLA, (9)	IBD	2.12E-38	1.124	1.092–1.157	Jostins et al.
rs12946510	IKZF3, ZPBP2, GSDMB, ORMDL3, GSDMA, (12)	IBD	4.10E-38	1.157	1.124–1.19	Jostins et al.
rs11741861	TNIP1, IRGM, ZNF300P1, (8)	IBD	2.94E-37	1.249	1.186–1.314	Jostins et al.
rs2155219	(5)	IBD	4.24E-36	1.151	1.119–1.185	Jostins et al.
rs2413583	ATF4, TAB1, APOBEC3G, (16)	IBD	4.40E-33	1.209	1.163–1.257	Jostins et al.
rs7554511	KIF21B, (6)	IBD	1.24E-32	1.164	1.128–1.202	Jostins et al.
rs4246905	TNFSF8, TNFSF15, TNC, (2)	IBD	2.80E-32	1.142	1.106–1.178	Jostins et al.
rs7134599	IFNG, IL26, IL22, (1)	IBD	8.51E-32	1.096	1.064–1.128	Jostins et al.
rs7608910	REL, C2orf74, KIAA1841, AHSA2, (6)	IBD	8.65E-32	1.138	1.105–1.171	Jostins et al.
rs2823286	0	IBD	9.28E-30	1.157	1.121–1.194	Jostins et al.
rs11564258	LRRK2, MUC19	IBD	6.38E-29	1.334	1.217–1.461	Jostins et al.
rs3091316	CCL13, CCL2, CCL11, (4)	IBD	1.22E-26	1.122	1.087–1.158	Jostins et al.
rs7282490	ICOSLG, (9)	IBD	2.35E-26	1.105	1.072–1.138	Jostins et al.
rs1893217	(6)	IBD	3.05E-26	1.171	1.127–1.216	Jostins et al.
rs11010067	CREM, (3)	IBD	2.49E-25	1.115	1.082–1.148	Jostins et al.
rs6062504	TNFRSF6B, LIME1, SLC2A4RG, ZGPAT, (23)	IBD	1.09E-23	1.104	1.071–1.139	Jostins et al.
rs12942547	STAT3, STAT5B, STAT5A, (13)	IBD	5.51E-22	1.103	1.072–1.136	Jostins et al.
rs26528	RABEP2, IL27, EIF3C, SULT1A1, SULT1A2, NUPR1, (9)	IBD	9.65E-22	1.099	1.067–1.13	Jostins et al.
rs6920220	TNFAIP3, (1)	IBD	1.40E-21	1.102	1.064–1.141	Jostins et al.
rs3749171	GPR35, (12)	IBD	3.07E-21	1.135	1.093–1.177	Jostins et al.
rs1819333	CCR6, RPS6KA2, RNASET2, (3)	IBD	6.76E-21	1.081	1.051–1.113	Jostins et al.
rs395157	OSMR, FYB, LIFR	IBD	2.22E-20	1.10	1.07–1.12	Liu et al.
rs917997	IL1R2, IL18RAP, IL18R1, IL1R1, IL1RL1, IL1RL2, (3)	IBD	3.12E-20	1.103	1.067–1.14	Jostins et al.
rs6568421	(2)	IBD	8.24E-20	1.108	1.074–1.142	Jostins et al.
rs921720	TRIB1, (1)	IBD	8.30E-20	1.081	1.049–1.113	Jostins et al.
rs17085007	(2)	IBD	2.79E-19	1.106	1.065–1.147	Jostins et al.
rs11879191	TYK2, PPAN-P2RY11, ICAM1, (25)	IBD	2.04E-18	1.136	1.096–1.177	Jostins et al.
rs1250546	(5)	IBD	3.15E-18	1.096	1.065–1.128	Jostins et al.
rs12568930	(3)	IBD	1.26E-17	1.095	1.054–1.138	Jostins et al.
rs2266959	MAPK1, YDJC, UBE2L3, RIMBP3, CCDC116, (8)	IBD	1.39E-16	1.105	1.066–1.145	Jostins et al.
rs529866	SOCS1, LITAF, RMI2, (10)	IBD	1.73E-16	1.124	1.085–1.166	Jostins et al.
rs4845604	RORC, (14)	IBD	3.52E-16	1.144	1.098–1.192	Jostins et al.
rs17293632	SMAD3, (2)	IBD	5.97E-16	1.067	1.032–1.102	Jostins et al.
rs6545800	ADCY3, (6)	IBD	6.14E-16	1.109	1.077–1.141	Jostins et al.
rs6586030	TSPAN14, C10orf58, (4)	IBD	9.24E-16	1.115	1.076–1.156	Jostins et al.
rs35675666	TNFRSF9, (6)	IBD	1.12E-15	1.112	1.07–1.156	Jostins et al.
rs4246215	C11orf9, FADS1, FADS2,(12)	IBD	1.93E-15	1.079	1.046–1.112	Jostins et al.
rs925255	FOSL2, BRE, (1)	IBD	2.67E-15	1.092	1.061–1.124	Jostins et al.
rs17694108	CEBPG, (8)	IBD	5.85E-15	1.1	1.065–1.135	Jostins et al.
rs1456896	ZPBP, IKZF1, (4)	IBD	7.28E-15	1.088	1.055–1.123	Jostins et al.
rs4256159	0	IBD	9.00E-15	1.107	1.063–1.152	Jostins et al.
rs941823	(3)	IBD	2.07E-14	1.071	1.036–1.107	Jostins et al.
rs8005161	GPR65, GALC, (1)	IBD	2.35E-14	1.153	1.097–1.211	Jostins et al.
rs9557195	GPR183, GPR18, (6)	IBD	2.37E-14	1.112	1.075–1.151	Jostins et al.
rs2412970	LIF, OSM, MTMR3, (8)	IBD	2.70E-14	1.08	1.049–1.111	Jostins et al.
rs6863411	SPRY4, NDFIP1, (5)	IBD	3.59E-14	1.089	1.057–1.121	Jostins et al.
rs9358372	(2)	IBD	8.66E-14	1.089	1.057–1.121	Jostins et al.
rs1569723	CD40, MMP9, PLTP, (11)	IBD	9.95E-14	1.091	1.056–1.126	Jostins et al.
rs3851228	TRAF3IP2, FYN, REV3L, (2)	IBD	1.08E-13	1.153	1.089–1.219	Jostins et al.
rs1734907	EPO, (21)	IBD	1.67E-13	1.114	1.071–1.158	Jostins et al.
rs7657746	IL2, IL21, (2)	IBD	2.76E-13	1.116	1.08–1.154	Jostins et al.
rs1363907	ERAP2, ERAP1, LNPEP, (2)	IBD	5.62E-13	1.068	1.037–1.099	Jostins et al.
rs12103	TNFRSF18, TNFRSF4, (30)	IBD	7.66E-13	1.099	1.059–1.139	Jostins et al.
rs2488389	C1orf53, (2)	IBD	8.45E-13	1.115	1.077–1.153	Jostins et al.
rs1292053	TUBD1, RPS6KB1, (9)	IBD	8.85E-13	1.076	1.045–1.106	Jostins et al.
rs11230563	CD6, CD5, PTGDR2, (12)	IBD	9.03E-13	1.085	1.053–1.118	Jostins et al.
rs259964	ZNF831, CTSZ, (5)	IBD	1.01E-12	1.085	1.054–1.116	Jostins et al.
rs2382817	SLC11A1, CXCR2, CXCR1, PNKD, ARPC2, TMBIM1, CTDSP1, (8)	IBD	3.70E-12	1.073	1.042–1.104	Jostins et al.
rs10495903	(5)	IBD	8.03E-12	1.086	1.041–1.131	Jostins et al.
rs9297145	SMURF1, (6)	IBD	8.21E-12	1.082	1.047–1.117	Jostins et al.
rs2538470	CNTNAP2	IBD	3.00E-11	1.07	1.05–1.09	Liu et al.
rs17119	0	IBD	3.08E-11	1.071	1.032–1.11	Jostins et al.
rs1517352	STAT1, STAT4, (2)	IBD	3.28E-11	1.077	1.046–1.109	Jostins et al.
rs559928	CCDC88B, RPS6KA4, TRPT1, FLRT1, (20)	IBD	4.19E-11	1.101	1.061–1.142	Jostins et al.
rs670523	UBQLN4, RIT1, MSTO1, (28)	IBD	5.79E-11	1.06	1.028–1.092	Jostins et al.
rs1042058	MAP3K8, (3)	IBD	5.93E-11	1.075	1.044–1.106	Jostins et al.
rs11672983	NLRP7, NLRP2, KIR2DL1, LILRB4, (15)	IBD	6.50E-11	1.087	1.055–1.119	Jostins et al.
rs4703855	-	IBD	7.16E-11	0.93	0.91–0.95	Liu et al.
rs7495132	CRTC3, (3)	IBD	9.48E-11	1.134	1.082–1.189	Jostins et al.
rs1847472	(1)	IBD	1.57E-10	1.06	1.029–1.092	Jostins et al.
rs111781203	CCL20	IBD	2.16E-10	0.94	0.92–0.96	Liu et al.
rs907611	TNNI2, LSP1, (17)	IBD	2.70E-10	1.068	1.035–1.101	Jostins et al.
rs194749	ZFP36L1, (4)	IBD	2.70E-10	1.075	1.039–1.111	Jostins et al.
rs2231884	RELA, FOSL1, CTSW, SNX32, (22)	IBD	2.91E-10	1.083	1.044–1.122	Jostins et al.
rs12722515	IL2RA, IL15RA, (6)	IBD	3.76E-10	1.102	1.06–1.147	Jostins et al.
rs4836519	(1)	IBD	4.24E-10	1.072	1.039–1.106	Jostins et al.
rs6142618	HCK, (10)	IBD	6.05E-10	1.072	1.041–1.103	Jostins et al.
rs2227564	(13)	IBD	6.75E-10	1.082	1.048–1.118	Jostins et al.
rs10896794	CNTF, LPXN, (8)	IBD	6.80E-10	1.08	1.045–1.116	Jostins et al.
rs7404095	PRKCB, (5)	IBD	9.68E-10	1.06	1.03–1.091	Jostins et al.
rs4911259	DNMT3B, (8)	IBD	1.20E-09	1.075	1.044–1.106	Jostins et al.
rs7240004	SMAD7, (2)	IBD	1.31E-09	1.057	1.026–1.088	Jostins et al.
rs10521318	IRF8, (4)	IBD	1.41E-09	1.155	1.094–1.219	Jostins et al.
rs1991866	(2)	IBD	1.65E-09	1.054	1.024–1.084	Jostins et al.
rs4743820	NFIL3, (2)	IBD	3.60E-09	1.056	1.023–1.089	Jostins et al.
rs727088	CD226, (2)	IBD	4.65E-09	1.077	1.046–1.108	Jostins et al.
rs4656958	CD48, SLAMF1, ITLN1, CD244, F11R, USF1, SLAMF7, ARHGAP30, (8)	IBD	6.80E-09	1.061	1.029–1.094	Jostins et al.
rs630923	CXCR5, (17)	IBD	7.07E-09	1.074	1.039–1.11	Jostins et al.
rs11168249	VDR, (8)	IBD	7.78E-09	1.054	1.024–1.084	Jostins et al.
rs2790216	CISD1, IPMK, (2)	IBD	8.07E-09	1.066	1.029–1.104	Jostins et al.
rs34856868	BTBD8	IBD	9.80E-09	0.82	0.77–0.88	Liu et al.
rs2930047	DAP, (2)	IBD	1.03E-08	1.065	1.034–1.096	Jostins et al.
rs11612508	LOH12CR1, (8)	IBD	1.06E-08	1.058	1.025–1.091	Jostins et al.
rs653178	SH2B3, ALDH2, ATXN2	IBD	1.11E-08	1.06	1.04–1.08	Liu et al.
rs4692386	-	IBD	1.21E-08	0.94	0.93–0.96	Liu et al.
rs38904	(6)	IBD	1.31E-08	1.054	1.025–1.085	Jostins et al.
rs12654812	DOK3, (17)	IBD	1.68E-08	1.068	1.036–1.1	Jostins et al.
rs2111485	IFIH1, (5)	IBD	1.93E-08	1.066	1.035–1.097	Jostins et al.
rs12199775	PHACTR2, (5)	IBD	1.99E-08	1.129	1.066–1.195	Jostins et al.
rs2651244	(3)	IBD	2.29E-08	1.015	0.986–1.044	Jostins et al.
rs6740462	SPRED2, (1)	IBD	2.35E-08	1.081	1.046–1.116	Jostins et al.
rs6025	SELP, SELE, SELL	IBD	2.51E-08	0.84	0.79–0.89	Liu et al.
rs2472649	CXCL5, CXCL1, CXCL3, IL8, CXCL6, PF4, CXCL2, PF4V1, (3)	IBD	2.57E-08	1.095	1.046–1.146	Jostins et al.
rs4664304	MARCH7, LY75, PLA2R1	IBD	2.61E-08	1.06	1.04–1.08	Liu et al.
rs4899554	FOS, MLH3, (6)	IBD	2.71E-08	1.083	1.042–1.125	Jostins et al.
rs7911264	(4)	IBD	2.98E-08	1.066	1.035–1.097	Jostins et al.
rs913678	CEBPB, (5)	IBD	4.59E-08	1.056	1.024–1.088	Jostins et al.
rs17057051	PTK2B, TRIM35, EPHX2	IBD	5.50E-08	0.94	0.92–0.96	Liu et al.
rs11064881	PRKAB1	IBD	5.95E-08	1.10	1.07–1.14	Liu et al.
rs3740415	NFKB2, TRIM8, TMEM180	IBD	1.03E-07	0.95	0.93–0.97	Liu et al.
rs2073505	HGFAC	IBD	1.46E-07	1.10	1.06–1.14	Liu et al.
rs564349	C5orf4, DUSP1	IBD	1.54E-07	1.06	1.04–1.08	Liu et al.
rs6856616	-	IBD	9.72E-07	1.10	1.06–1.14	Liu et al.

CD, Crohn’s disease; CI; confidence interval; GWAS, Genome-wide association study; IBD, inflammatory bowel disease;

OR, odds ratio; SNP, single nucleotide polymorphism; UC, ulcerative colitis.

^*Numbers in parentheses refer to the number of additional genes in the locus.

^†Refers to the study by Jostins et al.⁴²

^‡Refers to the study by Liu et al.⁶⁷

Figure 1.

Progress of discovery of inflammatory bowel disease-associated loci in three ethnic groups over the past 15 years.

5.3 Beyond GWAS

With GWAS, Immunochip, their meta-analysis and the recently published trans-ethnic study, the number of identified IBD-associated loci has increased approximately 100-fold during the past 15 years.⁶⁷ However, this approach is still based on the ‘common disease, common variant’ hypothesis,²³ and is mostly capturing variants identified from European ancestry populations.^8,67 The latest estimates suggest that the Immunochip and GWAS identified variants listed above only explain 19% and 26% of the heritability, respectively for CD and corresponding figures of 15% and 19%, respectively, for UC.⁷⁰ This missing heritability could be attributed to the regions overlooked by GWAS, such as the sex chromosomes, as suggested by recent studies which identified ARHGEF6 and XIAP as IBD-related genes.^8,71–75 It is also important to recognize that heritability estimates are prone to error and are an inexact science.

Rare or low frequency variants

Generally, rare variants have a low correlation with the marker SNPs used in the traditional genotyping platforms that are, on the whole, designed to capture common variation with a minor allele frequency over 0.05.^8,76 Therefore, to discover a rare or low frequency disease-associated allele, direct testing of the variants is necessary. In addition, with the infrequency of such alleles in the population, even the largest catalogues of known human variation will not contain all variants of interest as novel variants are discovered every time a human genome is sequenced.⁸ Therefore, sequencing of an entire region, not just the known variable sites would be required to discover rare or low frequency variations. With the introduction of next generation sequencing (NGS) technology, progress in this area will soon be realized. Early NGS studies for IBD were focused on early-onset IBD, under the hypothesis that these are more severe cases and may be closer to single gene or Mendelian disorders than adult-onset IBD.⁸ Exome sequencing in a male child who presented at 15 months with very severe CD-like presentation identified a XIAP mutation leading to immunodeficiency with bowel manifestations that improved significantly after allogenic hematopoietic progenitor cell transplant.⁷² Subsequently, according to a recent German study, private variants in XIAP were commonly observed in about 4% of male patients with pediatric onset CD.⁷⁵ Similar examples of rare mutations found in very early onset IBD (VEO IBD) subjects were identified in genes encoding the IL-10R subunit proteins,⁷⁷ and a deletion mutation in ADAM17 that was homozygous in both of two children born to consanguineous parents.⁷⁸ Further application of NGS technology for searching rare and low frequency variation in candidate IBD loci across cases and controls will likely reveal additional variants. Deep resequencing of GWAS loci in 350 cases of CD and 350 controls, followed by genotyping of 70 rare and low-frequency protein-altering variants in independent case-control series identified four additional independent risk loci in CARD15/NOD2, two additional protective variants in IL23R, and a highly significant association with a protective splice variant in CARD9.⁷⁹ A similar targeted sequencing approach on 55 candidate genes in 200 cases of UC and 150 controls with follow-up genotyping of 42 rare non-synonymous variants in independent case-control cohorts confirmed significant association of rare variants in both IL23R and CARD9, previously identified from sequencing of CD loci and identified a novel association in RNF186.⁸⁰ Ideally, the next step will be the deep sequencing of whole genome across sufficient number of cases and controls, not limited to candidate genes. However, minor allele frequency of a targeted rare variant is very low, thereby needing extremely large number of cases and controls to obtain a meaningfully significant difference unless variants with a very strong effect are identified. In addition to difficulty in enrolling large number of study subjects, the cost of deep sequencing of the whole genome remains a barrier as does the intensity of the subsequent analyses. Before whole genome sequencing becomes common place as costs continue to drop, a compromise approach of whole exome (coding regions in the genome) sequencing will likely become common place, as exomes represent only 1% of the complete genome.⁸¹ One example of successful exome sequencing is the discovery of a novel A-to-C missense variant (c.694A>C) in exon 6 of the FOXP3 gene on chromosome X in VEO IBD.⁸² Interestingly, GWAS-based studies have shown that a substantial proportion of IBD-associated loci are located in non-coding regions, suggesting rare variants regulating gene expression may also be important in IBD pathogenesis.⁸ The roles of non-coding regions on their pathogenic effects through modulation of gene expression are being identified by expression quantitative trait loci (eQTL)-GWAS mapping analysis.²⁷

A further example of regions potentially missed by GWAS-based approaches was also suggested by a recent functional methylome map of UC colonic mucosal tissues which identified sixty-one genomic regions with differential methylations patterns that were also associated with nearby differentially expressed transcripts, implicating epigenetic regulation of gene expression in contributing to UC pathogenesis.^27,83 Various types of micro-RNAs have also been implicated in IBD pathogenesis through the regulation of several IBD-associated genes, such as CARD15/NOD2, ATG16L1, and IL-23R.^84–86 Another suggested approach is sequencing many individuals at low depth and combining data across individuals to generate accurate calls in shared stretches of chromosome.⁸⁷ This approach may also provide useful disease-specific reference panels that can drive imputation into additional samples to increase power further.⁸⁷

5.4 Identifying the causal variants and genes

Although genome-wide scanning, meta-analyses, and Immunochip analyses have identified over 200 SNPs associated with CD and/or UC, the question on ‘functionality’ and ‘causality’ of these findings remains elusive. Increasing resources are being directed to identifying causal variants and thereby implicating genes within known loci.⁸ Traditional approaches to ‘prove’ causality included experimental validation of suggested genes using in vitro cellular or animal models. This approach is difficult and a time-consuming process, but can be complemented by fine-mapping strategies capitalizing on the large number of samples already genotyped by GWAS and Immunochip platforms. It is believed that the majority of GWAS-identified risk variants are not likely themselves to be causally affecting the trait, but rather, are correlated to the true causal variant through linkage disequilibrium.⁸⁸ Therefore, real causal variants could be identified in fine-mapping studies through targeted sequencing followed by prioritization of variants for functional validation.⁸⁸ For prioritization purpose, a novel framework integrating association strength with functional genomic annotation data has been proposed recently.⁸⁸ In addition, an algorithm which is integrating fine-mapped genetic and epigenetic data to identify candidate causal variants for 21 autoimmune diseases has also been suggested.⁸⁹

6. Genes/loci associated with IBD risk

6.1 CARD15/NOD2

The identification of CARD15/NOD2 in the IBD1 locus as a CD-associated gene was a major breakthrough for IBD genetic research. NOD2 consists of two amino-terminal caspase recruitment domains (CARDs), a centrally located nucleotide binding domain and multiple leucine rich repeats (LRRs) at its carboxyterminal end.¹⁷ The LRR serves as a receptor for muramyl dipeptide, a small molecule derived from the cell wall peptidoglycan of gram-positive and gram-negative bacteria.⁹⁰ Three different variants located in or near the LRR region are independently associated with CD in European ancestry populations: the frameshift mutation (L1007fsinsC), which causes a truncated protein transcript and two nonsynonymous polymorphisms (Arg702Trp and Gly908Arg).^17,20–22 Carriage of one copy of the risk allele increases the risk of developing CD 2- to 4-fold and carriage of two risk alleles increases the risk of disease 20- to 40-fold in adults.¹⁶ However, the allele frequency of the 3 SNPs, and thus the population-attributable risk for CD from these mutations varies widely depending on ethnicity, showing no effect in East Asians.^{19,20,51–53,91–96} Studies on the function of CARD15/NOD2 have shown that variants are associated with an impairment in intracellular killing of Salmonella, increased susceptibility to infection with Listeria monocytogenes, and decreased expression of certain cryptdins.^97,98 In addition, CARD15/NOD2-mutant mouse showed elevated NF-κB activation in response to muramyl dipeptide, more severe colitis induced by dextran sodium sulphate, and increased macrophage apoptosis and levels of IL-1β.⁹⁹ Recently, nod1 and nod2 proteins were also implicated as having a role in autophagy by recruiting atg16l1 protein to the plasma membrane at the point of bacterial entry, and by inducing autophagosome formation.^100–102 In CD patients with CARD15/NOD2 variants, ileal expression of α-defensin was more pronouncedly diminished than in those with wild-type.¹⁰³ CARD15/NOD2 variants are not associated with UC and some data suggest that they may decrease the risk of colonic inflammation.^19,104 In summary, CARD15/NOD2 mutations which are functionally linked to a deficient antimicrobial defense, are significantly associated with risk of CD in European ancestry, but not in East Asians.

6.2 Autophagy genes

Autophagy is a natural cellular mechanism removing unnecessary or dysfunctional cellular components for maintaining cellular and tissue homeostasis.¹⁰⁵ The representative autophagy genes, ATG16L1 and IRGM have been shown to play roles as restriction factors for pathogens through autophagy.¹⁰⁵ The discovery of the association of the Thr300Ala (rs2241880) in ATG16L1 with CD was a significant advance in our understanding of the pathogenesis of IBD. In 2007, three GWAS from Germany, UK, and North America identified a significant association of ATG16L1 with CD.^31,33,106 Subsequent studies from multiple European cohorts confirmed this association.^107–111 A Japanese study of 1311 CD cases and 6585 controls did not find any association between ATG16L1 variants and CD.⁹⁶ However, a subsequent Korean study of 1809 cases and 2436 controls, did demonstrate Thr300Ala association with CD.¹¹² Interestingly, a Korean GWAS suggested an association between ATG16L2 and CD.⁵⁹ Atg16l2 protein is a homologue of atg16l1, but its role in autophagy has not been confirmed yet.^59,112,113 The association of another autophagy gene, IRGM, with CD was also reported in European population studies.^106,107,114 However, replication studies on Asians showed conflicting results.^96,115–118 Immunochip and GWAS-based studies in UC have not shown any significant association in Europeans and Asians.^{37,38,116,117} A knock-in mouse model expressing the ATG16L1 Thr300Ala variant showed morphological defects in Paneth and goblet cells.¹¹⁹ Selective autophagy was also reduced in multiple cell types from Thr300Ala knock-in mice compared with wild type mice.¹¹⁹ ATG16L1 Thr300Ala protein showed more susceptibility to caspase 3- and caspase 7-mediated cleavage than wild type atg16l1 protein, resulting in decreased protein stability and effects on antibacterial autophagy and inflammatory cytokine production.¹¹⁹ In a subsequent study, increased proportions of abnormal Paneth cells were associated with the presence of Thr300Ala risk allele.¹²⁰ The cumulative number of ATG16L1 Thr300A and CD-associated CARD15/NOD2 variants had an additive effect on the proportion of abnormal Paneth cells as well.¹²⁰ Abnormal Paneth cells were also associated with a faster time to disease recurrence after surgical resection in CD cases.¹²⁰ Based on evidences from these studies, therapeutic approaches influencing autophagy could be a rational and promising agent for a subset of CD in the near future.¹²¹

6.3 IL23R

The first published European ancestry GWAS identified association between several common variants at IL23R and IBD,³⁰ an association confirmed in subsequent GWAS and replication studies.^106,122 Among multiple associated alleles in the IL23R gene region, the most significant association was observed at rs11209096 (Arg181Gln) where having the minor glutamine allele gives a 2- to 3-fold protection against developing IBD.^27,30 Subsequent functional studies revealed that the glutamine is a loss-of-function allele, and confers attenuated IL-23-mediated Th17 effector function, decrease of circulating CD4+ Th17 cells and CD8+ Tc17 cells, and reduced levels of proinflammatory cytokine secretion.^123–125 These findings suggested that blocking the IL-23 signaling pathway may be effective in treating IBD and studies blocking both the IL-12/23 pathway with monoclonal antibody to p40 subunit of IL-12/23 and targeting IL-23 pathway through blocking p19 (subunit of IL-23) have shown promise in active CD patients, especially for patients who have failed anti-TNF therapy. ^124,127

6.4 TNFSF15

The first ever GWAS on IBD which was performed in a Japanese population identified TNFSF15 as an IBD locus. Tnfsf15_28 (14,340T→C in intron 3 of TNFSF15) was the most significantly associated variant with CD among several associated SNPs at this locus.²⁹ In the UK cohort from the same study, five polymorphic markers (tnfsf15_26, 31, 35, 36 and 41) were associated with IBD.²⁹ Subsequently, the association of TNFSF15 with CD was confirmed in replication studies, GWAS and deep resequencing from both Japan and Korea.^{58,59,128–130} In European ancestry populations, the association of TNFSF15 with CD was confirmed, although the observed effect size was smaller than that seen in East Asians.^46,131,132 Tnfsf15 protein, also known as TL1A is a proinflammatory molecule which stimulates proliferation and effector functions of CD8(+) cytotoxic T cells as well as Th1,Th2, and Th17 cells in the presence of TCR stimulation.¹³³ Currently, TL1A is believed to be involved in the pathogenesis of IBD by stimulation of effector T cells and upregulation of proinflammatory cytokine production, defective generation of peripheral Tregs, and dampening suppressive activity of preexisting Tregs.¹³³ A recent study revealed that monocyte-derived macrophages from rs6478108 A allele CD risk-carriers in the TNFSF15 region showed increased tnfsf15 expression, and increased pattern-recognition receptor-induced signaling and cytokines compared with GG carriers.¹³⁴ TNFSF15 variants have also been associated with stricturing phenotype in CD.¹³⁵ Collectively, these findings suggest that targeting tnfsf15 could be a promising therapeutic area for IBD patients. In a mouse model, antibody to TL1A prevented DSS-induced chronic colitis and T cell-mediated chronic colitis, as well as attenuating established DSS-induced chronic colitis by down-regulating of both Th1 and Th17 activation.¹³⁶ In the near future, it is likely that TL1A blocking agents will be trialed in IBD patients.

6.5 HLA

The HLA region on chromosome 6p21 contains many genes related to immune function. The association between the HLA region and IBD has been studied for many years, and multiple associated loci have been reported including confirming early serological-based associations with HLA. The recent GWAS/Immunochip meta-analysis demonstrated that SNP rs6927022 near the class I gene HLA-DQA1 is the strongest UC-associated locus in the genome, and rs9264942 located in the HLA-B gene locus of the HLA class I region is the strongest CD-associated HLA locus.⁴² The recent high-density genotyping for 7406 SNPs within the HLA region in a total of 66954 individuals (18,405 with CD, 14,308 with UC, and 34,241 controls) identified multiple risk alleles associated with one or both diseases, with HLA-DRB1*01:03 being the most strongly associated with both CD and UC.¹³⁷ They also showed that the contribution of class I and class II HLA variants to disease risk is relatively equivalent in CD, but HLA class II variation has a more important role in UC.¹³⁷ In addition, most associated HLA alleles had a predominant role in either CD or UC, with very few conferring shared IBD risk.¹³⁷

7. Clinical application of genetics in IBD

Considering significant heterogeneity in phenotype, natural history, and therapeutic response, IBD could be an ideal model for personalized or precision medicine, which incorporates individual variability in building prevention and treatment strategies.¹³⁸ The significant advances in understanding the genetic architecture of IBD may be helpful in clinical practice in areas such as diagnostics, prognostics, and therapeutics.

8. Diagnosis of IBD

Despite abundant knowledge acquired from multiple studies on genetic traits associated with IBD, low pre-test probability (i.e. low prevalence of IBD in the general population) and moderate genotype-relative risks of IBD-associated loci limit the utility of genetic testing for new diagnosis of IBD.^27,139 One example of this limitation is poor agreement on direction of risk of CD between two direct-to-consumer (DTC) genetic testing products with 3 out of 5 individuals receiving discordant CD risk-estimates from the 2 DTCs.¹⁴⁰ An alternative strategy of combining serological, genetic, and inflammatory markers to differentiate non-IBD, CD, and UC has been attempted.¹⁴¹ Using a total of 17 markers (8 serological markers, 4 genetic markers, and 5 inflammatory markers) for a diagnostic random forest algorithm, the IBD vs. non-IBD discrimination area under the curve (AUC) was 0.87, which was significantly higher than the AUC for serology-only panel, 0.80 (P = 0.0001).¹⁴¹ Similarly, the AUC for CD vs. UC increased from 0.78 of serology-only panel to 0.93 with the combined panel (P = 0.0001).¹⁴¹ However, the contribution of genetic markers on this model needs to be evaluated further and the validity of this approach remains to be confirmed in additional cohorts.²⁷ Although there is a long way to go for applying genetic tests of common variants as a screening or diagnostic tool of IBD, tests for single genes causing IBD could contribute to both diagnosis and intervention based on pathogenesis as well as genetic counseling.²⁷ Known examples of single genes resulting in VEO IBD are the previously discussed XIAP, IL-10R, ADAM17, and FOXP3.^72,77,78.82 Because 50 or more single genes causing IBD are implicated in VEO IBD, a useful IBD gene panel test including multiple candidate genes may guide clinicians to aid diagnosis and management of VEO IBD.

9. IBD subphenotypes and prognosis

In addition to the genetic variation associated with overall IBD risk, there has been considerable interest in the association between genetic variation and IBD subphenotypes, prognosis, and natural history. There have been multiple studies on the association between IBD susceptibility genes and prognosis, mainly among European ancestry subjects. The CARD15/NOD2 SNPs have been associated with ileal CD, stricturing disease, penetrating disease, familial disease, and earlier onset of disease.^{19–21,92,142–146} CARD15/NOD2 was also significantly associated with both bowel resection and complicated disease course (defined as one or more of the following: stenosing or internal penetrating behavior; perianal disease; or bowel resection) from a recent European multinational, multicenter study.¹⁴⁵ Subphenotypic associations of CARD15/NOD2 for CD was also confirmed in a study involving 49 sites in 16 countries in Europe, North America, and Australia, which suggested that CARD15/NOD2 primary association was predominantly with disease location (ileal) and not with a stricturing phenotype.¹⁴⁷ Among autophagy genes, the ATG16L1 T300A variant has been associated with ileal CD,^{55,110,148,149} and IRGM was associated with internal and perianal fistula in Italian CD patients.¹⁵⁰ IL23R variation was also associated with ileal involvement (rs7517848),¹⁵¹ and with stricturing or penetrating phenotype (rs1004819, rs1495965, and rs11465804).^55,145 For HLA, the DR3 DQ2 haplotype was predictive of extensive UC rather than distal disease.¹⁵² For CD, HLA-DRB1*01:03 was associated with pure colonic CD rather than ileal involvement and also with later age of diagnosis.^144,153 In contrast, HLA-DRB1*07:01 and DRB1*04 were associated with ileal CD.¹⁴⁴ The most recent multicenter study also confirmed the association between HLA and IBD subphenotypes, especially disease location.¹⁴⁷ Three SNPs in the TNFSF15 (rs6478109, rs7848647, and rs4979462) were associated with perianal lesions in Japanese patients with CD.¹²⁸ In Korean CD patients, rs6478108 CC genotype was associated with stricture and non-perianal penetrating complications, and rs4574921 CC genotype with perianal fistula.¹³⁵ The rs4263839 in the TNFSF15 was related with bowel resection in CD patients with European ancestry.¹⁴⁵ Comparing medically refractory UC (MRUC) and controls, the contribution of the HLA region to severe disease was confirmed and 2 additional loci reached a suggestive level of responsibility including TNFSF15.¹⁵⁴

Understanding the limitation of a prognostic test based on a single variant with limited effect size, researchers have combined genetic variants to produce composite or gene-risk scores. Previously, a genetic risk score from the total number of risk alleles (0, 1, or 2) across 46 risk SNPs associated with MRUC ‘explained’ 48% of the variance for colectomy risk.¹⁵⁴ In addition, when UC patients were grouped into four categories based on quarterly-divided risk score, the proportion of MRUC for the four groups was: less than 1%; 17%; 74%; and 100%, respectively (P < 2.2 × 10⁻¹⁶).¹⁵⁴ In the most recent genotype-phenotype association study from the IIBDGC, information from 193 SNPs and 23 HLA types which are known to be associated with IBD was accumulated and used to generate genetic risk scores (CD score and UC score).¹⁴⁷ As a result, CD vs. UC risk score showed very strong correlations with CD location and behavior (P = 1.65 × 10⁻⁷⁸, or P = 9.23 × 10⁻¹⁸ after excluding the individual loci that achieved genome-wide, CARD15/NOD2, HLA, and 3p21).¹⁴⁷ Moreover, predictive models based on genetic risk score strongly distinguished colonic from ileal CD and based on genetic risk score, patients with IBD could be much better characterized into 3 groups (ileal CD, colonic CD, and UC) rather than CD and UC as currently defined.¹⁴⁷ This modeling could be clinically useful for differential diagnosis of IBD, especially for colonic IBD. Genetically, it appears that colonic CD is ‘closer’ to UC than it is to pure ileal CD.¹⁴⁷ Further work using ‘finer’ phenotyping characteristics and classification may be able to delineate this further.¹⁴⁷ Another approach is developing a composite model including genetic variables and other variables such as clinical and serologic markers to predict the course of IBD. For example, a combination of phenotypic, serologic, and genetic variables ‘predicted’ time to first surgery in CD more accurately than clinical only, genetics only, and clinical + serology models.¹⁵⁵ In that study, variation at the IL12B locus showed a consistent association with both the need for surgery and the time to surgery in all models, thereby suggesting IL12B may be a potential target for therapeutic intervention.¹⁵⁵ A group of researchers combined demographics, clinical characteristics such as disease location and perianal involvement, serologic markers, and CARD15/NOD2 genotypes to build a validated, individualized, and web-based tool to visualize individual risks for developing CD complication.¹⁵⁶ As a result, a multivariate model including disease location, serologic markers (ASCA, CBir1, ANCA), the CARD15/NOD2 frameshift mutation was associated with the risk of complication, and a web-based tool based on the multivariate model was able to individualized disease outcomes in a patient-friendly format.¹⁵⁶

10. Pharmacogenetics associated with IBD therapy

Identification of genetic variations related with effects or adverse events related with IBD therapy are recognized as an area that is most likely to benefit from advances in genetic technologies. The thiopurines are commonly-used drugs for treating IBD and it is well established in Caucasians that genetic polymorphisms in the thiopurine S-methyltransferase (TPMT) gene leading to reduced TPMT activity and high 6-TGN concentration are associated with the development of myelotoxicity.¹⁵⁷ However, despite lower frequency of TPMT mutations in Asians including Koreans, Japanese, and Chinese than in European ancestry, the frequency of thiopurine-associated leukopenia was considerably higher in Asians.¹⁵⁸ Moreover, even in Europeans, only around 25% of myelosuppressive episodes during thiopurine therapy are associated with a TPMT deficient genotype.^159,160 These findings suggest that additional factors involved in thiopurine-associated myelosuppression exist. A recent Korean Immunochip-based association study in 978 Korean patients with CD treated with thiopurines identified association between a nonsynonymous SNP in NUDT15, rs116855232 (encoding p.Arg139Cys) and early leukopenia (allele frequency of 55.3% in early leukopenia cases vs. 3.4% in thiopurine-treated cases without leukopenia, OR 35.6; P_combined = 4.88 × 10⁻⁹⁴).¹⁵⁸ The presence of this NUDT15 allele also showed a high sensitivity and specificity for early leukopenia (89.4% and 93.2%, respectively), with an area under the curve value of 0.92 (95% CI 0.88–0.97).¹⁵⁸ This SNP is rare in European ancestry individuals but is also associated with thiopurine-induced leukopenia in Europeans and the association with bone marrow toxicity has also been reproduced in Japanese IBD patients.^158,161,162 Pancreatitis is one of the dose-independent adverse reactions related with thiopurines limiting their use in IBD.¹⁵⁷ In a recent worldwide study to identify genetic markers predicting pancreatitis within 3 months of starting thiopurines in patients with IBD, a GWAS on 172 cases and 2035 controls as well as additional validation was performed.¹⁶³ As a result, strong evidence of association within the class II HLA region, with the most significant association identified at rs2647087 (OR 2.59, P = 2 × 10⁻¹⁶), with further replication in an independent set, was observed.¹⁶³ Clinically, individuals homozygous for the risk allele at rs2647087 had an approximate 17% risk for pancreatitis, and a risk of 9% in rs2647087 heterozygotes was observed.¹⁶³ Therefore, genetic information on the risk of myelosuppression (including both NUDT15 and TPMT variants) and pancreatitis development could be utilized as a screening panel that could aid clinicians to reduce the risk of serious adverse events related with thiopurines for IBD patients through choosing alternative approaches such as methotrexate for those at high-risk for these events.^158,163

Given the failure of anti-TNF agents for a substantial number of IBD patients, multiple attempts to reveal genetic factors in predicting responses to anti-TNF agents have been attempted. Early studies showed inconsistent results and no association was observed between CARD15/NOD2 and non-response.¹³⁹ In a study on 90 patients with UC, who were treated with infliximab as an induction therapy, patients homozygous for IBD risk-increasing IL23R variants (rs1004819, rs2201841, rs10889677, rs11209032, rs1495965) were more likely to have a response than those homozygous for risk-decreasing IL23R variants (rs7517847, rs10489629, rs11465804, rs1343151) (74.1 vs. 34.6 % ; P = 0.001).¹⁶⁴ Based on the finding that infliximab acts partly through inducing apoptosis of activated T cells, whether SNPs involved in apoptosis-related pathway is associated with infliximab response was investigated by the Leuven group. They constructed a novel ‘apoptotic pharmacogenetic index’ (ranging from 0 to 3) by assigning points for 3 SNPs in the apoptosis-related genes (Fas ligand −843 C>T, Fas −670 G >A, and caspase-9 93 C>T) based on their association with response to infliximab in CD.^165,166 Response and remission rates after infliximab administration significantly increased with apoptotic pharmacogenetic index score in both luminal CD and fistulizing CD with some influence by age, medication, or CRP levels in some subgroups.¹⁶⁶ Another approach of searching for genetic variation related with anti-TNF therapeutic response is analyzing gene expression profile in tissues such as bowel mucosa.¹³⁹ A Belgian study performed microarray analysis of total RNA from pre-treatment rectal mucosal biopsy samples from refractory UC patients.¹⁶⁷ As a result, the top five differentially expressed genes (osteoprotegerin, stanniocalcin-1, prostaglandin-endoperoxide synthase 2, IL-13 receptor alpha 2 and IL-11) could separate infliximab responders from non-responders with 89.1% overall accuracy, 95.0% sensitivity, and 84.6% specificity.¹⁶⁷ Using similar methods for CD, the top 5 differentially expressed genes (TNFAIP6, S100A8, IL11, G0S2, S100A9) could completely separate responders and non-responders to infliximab in the colonic CD group, but not in an ileal CD group.¹⁶⁸ Interestingly, predictive genes from UC and colonic CD datasets showed near complete overlap, and the top five differentially expressed gene panel in UC predicted response to infliximab in colonic CD with 94.7% of accuracy.^167,168 These findings provide insight into the molecular mechanisms involved in anti-TNF responsiveness, which appear similar between UC and colonic CD, but not with ileal CD.^167,168 A small study of 94 pediatric IBD patients attempted to build a model to predict primary nonresponse to anti-TNFs.¹⁶⁹ A composite model with three novel loci (rs975664 in the TACR1, rs4855535 in the FAM19A4, rs6100556 in the PHACTR3), one known IBD susceptibility loci (rs2836878 in the BRWDI), pANCA positivity, and UC diagnosis could predict the primary nonresponse to anti-TNFs with an R² of 0.82 and an AUC of 0.98.¹⁶⁹ In conclusion, genetic variations involved in the immune pathways and drug metabolism seem to have a greater effect on TNF responsiveness than IBD susceptibility genes. Moreover, the combination of genotype, serotype, and phenotype may be useful in predicting response to anti-TNFs, resulting in proper use of anti-TNF drugs for right patients.^139,169 However, it is important to stress that these underpowered study findings’ need validation in additional cohorts. More recently it has become apparent that there is considerable pharmacokinetic variation between individuals exposed to anti-TNFs and it is likely that true non-responders can only be determined after adequate drug exposure has been confirmed.

11. Conclusions

During the past decades, the research community has achieved remarkable advances in the understanding of the genetics of IBD and various other related immune-mediated disorders through approaches including linkage analysis, GWAS, and GWAS meta-analysis, etc. Through experience with working on individual cohort GWAS sets, it became clear that large numbers were essential for advancing the field towards an understanding of the molecular architecture of IBD. This recognition led to the collaboration of researchers from multiple centers in Europe, North America, and Australia. This multinational collaborative approach should be a model for other groups such as those studying Asian populations, in whom there is a rising incidence of IBD. Despite multiple susceptibility genes/loci discovered through chip-based technologies such as GWAS and Immunochip, only a modest portion of the expected heritability has been explained, which may be inherent in the basic concept of ‘common disease, common variant’ hypothesis. Therefore, rare variants with greater effect sizes on IBD development need investigating and various strategies are currently being pursued to achieve this. The ultimate goal of genetic studies is to make advances in the management of IBD, including diagnosis, subclassification, predicting course of disease, and the development of new therapeutics. These fields are evolving rapidly in parallel with other ‘omic’ advances which, it is anticipated, will lead to an increasingly personalized management of IBD patients and a realization of precision medicine in the near future.

12. Expert commentary

In European ancestry populations, more than half of predicted heritability of IBD cannot be explained by genes/loci discovered to date, a figure that is considerably higher in Asian IBD populations. Furthermore, the majority of IBD-associated loci involve non-coding variation, which is in contrast to the traditional concept of disease-causing nonsynonymous, coding region variations. Well-designed, adequately powered fine mapping and deep sequencing strategies coupled with gene expression studies in appropriate tissues, will help define the functional role of these multiple non-coding variants. One interesting phenomenon is that the majority of people carrying IBD-associated risk variants remain healthy, others may develop another immune-mediated disease such as spondyloarthropathy, while a third, unfortunate, group may develop more than one condition. These observations suggest that other factors including gene-environment, gene-gut microbome, and even perhaps gene-gene interactions need to be studied to determine why some subjects get disease and others remain healthy (even within a family sharing similar environmental exposures). Furthermore, additional studies on epigenetic regulation of genes are necessary to extend our knowledge on the pathogenesis and natural history of IBD.

13. 5-year view

Our current knowledge on the architecture of IBD genetics has been achieved by embracing advances such as microarray technology, novel analytic programs, and statistical methodologies handling huge volumes of data. The same will be true for future development of IBD genetics, and next generation sequencing technology and further rapidly developing advanced technologies are expected to reveal more variants associated with IBD susceptibility and its prognosis. The collegiate and collaborative nature of the IIBDGC has been pivotal in driving the advances in this arena and it is evident that large sample sizes with well-defined subjects’ characteristics are required for more accurately characterizing the genetic architecture of IBD populations. It is imperative that the potential benefits of these genetic advances are available to all populations and the establishment of an Asian IBD genetic consortium as well as additional efforts in African-American populations together with trans-ethnic studies will be necessary to achieve these aims. Furthermore, designing new genotyping chips containing loci specifically designed to capture common and rare variants in these populations will also be beneficial. These populations with a rapidly expanding prevalence are crucial to study both environmental influences and therefore also gene by environmental associations. Findings from these studies will be of benefit to all IBD populations.

14. Key issues

• Genetic contribution on IBD development was strongly suggested by epidemiologic observation such as familial disease clustering, high concordance rates in twins, and the different prevalences of IBD observed in different ethnicities.

• GWAS opened a new era into IBD genetic research, helping overcome the limitation of linkage analysis for discovering causal variants responsible for complex immune-mediated chronic disorders.

• GWAS, GWAS meta-analysis and Immunochip analyses expanded the number of IBD-associated loci to 163 (110 with IBD, 30 CD-specific, and 23 UC-specific) in European ancestry populations.

• The recent genetic association studies on Asian and African ancestry are showing both overlap and some ethnic specificity in the molecular architecture of IBD.

• The trans-ethnic Immunochip analysis increased the number of IBD risk loci to 231 independent SNPs within 200 loci, with the majority shared across diverse ancestry groups.

• Even with IBD-associated loci from multiple genetic association studies and meta-analyses, only a modest fraction of predicted heritability can be explained in both European ancestry and Asian populations. To overcome this limitation, searching for rare or low frequency disease-associated alleles with deep resequencing, whole genome sequencing, and whole exome sequencing approaches using next generation sequencing technology will extend our knowledge on the genetic architecture of IBD.

• From a screening and diagnostic point of view, the clinical application of IBD genetics is still limited.

• Multiple loci have been known to be associated with specific IBD subphenotypes and prognosis. A genetic risk scoring system could be more helpful for classification, subphenotyping, and prognostication of IBD in the future especially if therapeutic approaches targeting some of the implicated genetic pathways are developed.

• Genetic knowledge on genes associated with thiopurine-related adverse events such as myelosuppression and pancreatitis is expanding, leading to direct clinical benefit in IBD.

• History has taught us that collaborative approaches are needed to assemble adequately powered cohorts for genetic research and these principles should be continued to be followed as investigators study non-European ancestry populations and also extend the studies to other modalities such as sequencing or the study of epigenetics.