Immunochip Meta-Analysis of Inflammatory Bowel Disease Identifies Three Novel Loci and Four Novel Associations in Previously Reported Loci

Abstract

Background and Aims

Recent meta-analysis of genome-wide association studies have identified over 241 inflammatory bowel disease susceptibility loci. However, the known variants only account for a fraction of inflammatory bowel disease heritability. To identify additional susceptibility loci, we performed a trans-ethnic meta-analysis as well as an Asian-specific meta-analysis, using all published Immunochip association results of inflammatory bowel disease.

Methods

An inverse-variance fixed-effects meta-analysis was carried out across Korean and East Asian Immunochip datasets of 4156 cases and 4904 controls [Asian ancestry]. A trans-ethnic meta-analysis of inflammatory bowel disease was performed together with the European datasets of 38 155 cases and 48 485 controls genotyped on the immunochip using a Bayesian approach, Meta-Analysis of Trans-ethnic Association studies [MANTRA].

Results

We identified seven novel associations, including three novel susceptibility loci at MYO10-BASP1, PPP2R3C/KIAA0391/PSMA6/NFKB1A and LRRK1 as well as four novel secondary associations within previously known loci at NCF4, TSPAN32, CIITA and VANGL2. The new loci further implicate alterations in B cell biology in Crohn’s disease pathogenesis. The effects of five loci were universal across European and Asian ancestries, whereas the NCF4 and CIITA loci showed significant heterogeneity between European and East Asian populations. In addition, 103 previously known IBD loci showed supporting evidence of association with nominal significance [p < 0.05] in Asians.

Conclusions

Our findings of new loci not previously associated with IBD support the importance of studying inflammatory bowel disease genetics in diverse populations.

1. Introduction

Inflammatory bowel disease [IBD] is a chronic inflammatory disorder of the gastrointestinal tract. Crohn’s disease [CD] and ulcerative colitis [UC] are the two major subtypes of IBD. IBD is believed to develop due to dysregulated mucosal immune responses to gut flora in genetically susceptible individuals.¹

Although the incidence of IBD is lower in Asian populations than in Western populations, the incidence of IBD is rapidly increasing throughout Asia due to environmental changes.^2–5 Epidemiological and clinical studies indicate that the phenotype and clinical course of IBD differs between Asians and Europeans.^2–7 First, there is a male predominance of CD in Asia, with male to female ratios ranging from 1.67:1 to 2.9:1. Second, ileocolonic disease is predominant in Asian cases, comprising about two-thirds of CD cases, while colonic disease is only noted in ~10% of cases. It is unclear whether these Asian-specific clinical characteristics of IBD are solely due to the difference in the environments between Asia and the West, which underscores the necessity of genetic studies of IBD in Asian populations.

Recent large-scale studies on populations of European ancestry have markedly advanced our understanding of IBD-related genetics. A meta-analysis by the International IBD Genetics Consortium [IIBDGC] of genome-wide association studies [GWASs] and Immunochip data from 96 486 individuals with multiple ancestries [including Asian samples] identified >200 susceptibility loci for IBD, and reported an overlap in the directionality of the odds ratios [ORs] between the cohorts with European and Asian ancestry.⁸ The latest genome-wide meta-analysis performed on populations with European ancestry reported 241 susceptibility loci for IBD.⁹ However, it is clear that additional studies are needed to expand our understanding of the genetic architecture of IBD. Despite the differences in the clinical characteristics of IBD between different ethnicities, the number of studies in non-European populations is limited. Several GWASs on CD or UC have been performed in Asian populations.^10–15 These studies provided insights into the genetic architecture of each disease, but not a broad view of the genetic basis of IBD in Asians. In the recent largest-to-date Asian-specific GWAS of IBD, we identified two new susceptibility loci to IBD and confirmed associations with 28 established IBD loci in Koreans.¹⁶ In the present study, we aimed to identify [i] novel genetic variants associated with IBD in Asians through a meta-analysis of Korean and East Asian studies and [ii] additional IBD susceptibility loci through a trans-ethnic meta-analysis of Korean, East Asian, and European studies. Here we present the findings of the meta-analyses, including seven novel associations identified at the level of genome-wide significance.

2. Methods

2.1. Study population

The samples of the three Immunochip datasets are described in Supplementary Table 1. Samples in the Korean dataset were used in our previous published Immunochip studies of CD and UC,^12,13 respectively. The summary statistics of Korean IBD data can be downloaded from a webpage [https://drive.google.com/drive/folders/1L1Zu4G0yzVuB0Ea11HkF-XQqr9pX15-0]. Their clinical characteristics are described in Supplementary Table 2. Previously published summary statistics of the East Asian and European IBD Immunochip datasets available at the IIBDGC [https://www.ibdgenetics.org/downloads.html] were used. The East Asian study comprised 2054 cases from Japan, 453 cases from Korea, 317 cases from China, and 3719 controls. The European dataset comprised 38 155 cases and 48 485 controls.

2.2. Quality control

The quality control [QC] procedures of the Korean, East Asian and European Immunochip datasets are described in previous publications.^8,12,13 As Korean Immunochip studies were performed on CD and UC separately, standard QC procedures were applied using the PLINK v1.9 software [https://www.cog-genomics.org/plink2] as described previously for the Korean IBD dataset.^12,13 Briefly, single nucleotide polymorphisms [SNPs] with missingness rates of >2%, a minor allele frequency [MAF] of <1%, or failing the Hardy–Weinberg equilibrium [p < 1 × 10^–5] test were excluded. Samples with a high proportion of missing genotypes [>4%] and identity by descent [PI_HAT >0.2] were removed. QC was conducted on each dataset separately and the combined set of samples using a common approach. To control for population stratification, genotypes that passed QC filters were merged with HapMap Phase III data [194 individuals] from three populations: European [CEU], Asian [CHB + JPT], and African [YRI]. A principal-components analysis was subsequently performed on the merged dataset, and population outliers were excluded [Supplementary Figure 1]. For the meta-analysis, 87 overlapping cases [79 CD and 8 UC cases that were present in both Korean and East Asian datasets] were excluded from the Korean dataset. Finally, the Korean dataset consisted of 1332 cases and 1185 controls, with 89 051 SNPs [Supplementary Table 3].

2.3. Statistical analysis

Association tests were performed on the Korean dataset of IBD, CD and UC using a logistic regression analysis in PLINK v1.9. A quantile–quantile [Q–Q] plot was generated using R [3.2.0] [http://www.r-project.org/] to evaluate the overall significance of the genome-wide associations and the potential impact of population stratification. The impact of population stratification was also evaluated by calculating the genomic control inflation factor [λ_GC] and the genomic inflation factor for 1000 cases and 1000 controls [λ_GC1000] using a set of 3120 “null” SNPs that are not associated with autoimmune diseases [Supplementary Table 1]. After QC, 2117 SNPs were used as null markers to generate the Q–Q plot shown in Supplementary Figure 2 [λ_GC = 0.98]. Both the Q–Q plots and genomic inflation factor [λ_GC] of Immunochip test statistics showed that the three Immunochip association analyses had negligible inflation due to population stratification. A Manhattan plot was generated with −log₁₀p by using R [3.2.0]. For regional plots of novel associations that were identified in the Asian meta-analysis, imputation was performed using ImpG version 1.0 [https://github.com/huwenboshi/ImpG]¹⁷ and the Asian reference data [JPT + CHB] from the 1000 Genomes Project [February 2012 release] [http://www.1000genomes.org/]. To assess whether candidate novel signals were due to long-range linkage disequilibrium [LD] with variants in previously reported loci, a conditional analysis in the regions identified in trans-ethnic meta-analysis was performed. For all variants in candidate loci that were <3 Mb away from a known locus, conditional analysis was performed on each of the three datasets separately (East Asian/European dataset, GCTA [cnsgenomics.com/software/gcta/]; Korean dataset, PLINK v1.9) followed by a meta-analysis, or on the combined dataset. Secondary SNPs with conditional p < 5 × 10^–8 or Log₁₀BF > 6 were assumed to be independent from the reported lead SNP in the region.

2.3.1. Asian-specific meta-analysis

For the Asian-specific meta-analyses, Korean and East Asian association results were combined using the inverse-variance method under the assumption of a fixed effect as implemented in METAL.¹⁸ Between-study heterogeneity was quantified using the I² heterogeneity score, and statistical significance was assessed using the Q test statistic. For the fixed-effects model, significance was defined as p_meta < 5 × 10^–8.

2.3.2. Trans-ethnic meta-analysis

For the trans-ethnic meta-analyses, the three independent datasets were combined using two approaches. First, a meta-analysis was performed using the METAL software assuming fixed effects across studies and using inverse-variance weighting. Second, the ethnic-specific Immunochip summary statistics were combine using the MANTRA [Meta-Analysis of Trans-ethnic Association Studies] package, a meta-analysis software tool allowing for heterogeneity in allelic effects caused by differences in LD structure in diverse populations.¹⁹ MANTRA results are reported as Log10 Bayes’ factors [Log₁₀BF]. A Log₁₀BF > 6 was considered to be a genome-wide significant threshold value, and SNPs with posterior probability of heterogeneity [p_het] > 0.5 were interpreted as having significant heterogeneity. Both fixed effects results and MANTRA results are reported.

2.4. eQTL and bioinformatics analysis

To gain more insight into the potential functional roles of the novel IBD loci, a cis-eQTL analysis was performed by searching publicly available expression data generated from eQTL Blood Browser,²⁰ the Genotype-Tissue Expression [GTEx] database²¹ and Geuvadis/1000 Genomes resources.²² Whole blood, small intestine, transverse colon, and sigmoid colon were selected in the GTEx browser because they are the most important tissues in mucosal immunity. To explore epigenetic profiles of genomic locations associated with IBD, ENCODE histone modification data, HaploReg and Regulome DB were used to examine whether any of the SNPs or their proxies [r² ≥ 0.8 in the 1000 genomes of JPT + CHB reference panel] were annotated as transcription factor binding or enhancer elements. Evidence of prior association signals with autoimmune diseases or other immune-related phenotypes was searched for in the Ensembl, UCSC Genome Bioinformatics, and GeneCards databases. When the SNP was not directly typed, a proxy SNP was used [r² ≥ 0.8].

3. Results

3.1. Asian-specific meta-analysis

To identify additional IBD risk loci in Asians, we first performed a meta-analysis using the Korean IBD Immunochip dataset and the summary statistics from the East Asian Immunochip dataset.^8,12,13 The combined dataset consisted of 4156 IBD cases and 4904 controls, with 1332 cases and 1185 controls from the Korean population and 2824 cases and 3719 controls from the East Asian population. Association analysis of the Korean dataset alone did not show any novel IBD loci with genome-wide significance [Supplementary Figure 3]. In a meta-analysis of the combined dataset, 12 of the previously reported regions exceeded statistical significance [p_meta < 5 × 10^–8], including two novel associations in Asians at rs2072711 in NCF4-CSF2RB at 22q13 [OR = 1.21; 95% confidence interval [CI], 1.13 – 1.29; p_Asian-meta = 5.07 × 10^–9] and at rs12928665 in CIITA at 16p13 [OR = 1.19; 95% CI, 1.13 – 1.25; p_Asian-meta = 3.93 × 10^–9] [Supplementary Table 4, Supplementary Figure 4]. We also examined the 245 previously established European IBD-associated loci [276 independent SNPs] in the Asian meta-analysis.^8,9 Of 193 SNPs in 174 loci [including 36 proxy SNPs with r² ≥ 0.8] available, 111 SNPs from 102 loci were replicated at nominal p < 0.05 in association analyses for IBD, CD or UC [excluding rs864745 and rs10995235 with effects in opposite directions; Supplementary Table 5]. An additional locus [JAK2], not in LD with the European SNP, reached a genome-wide significant association in the Asian meta-analysis [Supplementary Table 4], resulting 112 SNPs being replicated at 103 loci in Asians.

3.2. Trans-ethnic meta-analysis

We performed a trans-ethnic meta-analysis of IBD using three independent Immunochip datasets. The combined dataset consisted of 42 311 IBD cases and 53 389 controls, with 1332 cases and 1185 controls from Korean population, 2824 cases and 3719 controls from East Asian population, and 38 155 cases and 48 485 controls from the European population. After applying stringent quality controls, we tested 89 051, 106 681, 126 098 and 80 291 SNPs for association in the Korean, East Asian, European and combined cohort, respectively [Supplementary Table 1]. We performed a meta-analysis using [i] inverse-variance based meta-analysis in METAL software and [ii] the MANTRA package [Supplementary Figure 5]. Following meta-analyses, the SNPs within 241 known loci⁹ in the MHC region [25−34 Mb, hg19] and known genes were removed.^9,23 Twenty-two loci showed significant associations with IBD in the MANTRA analyses [Log₁₀BF > 6], of which 19 loci also showed significant associations in the fixed-effects trans-ethnic METAL analyses [p < 5 × 10^–8, Supplementary Table 6]. As these loci were <3 Mb away from known loci except for rs2624435 near MYO10-BASP1, we performed a conditional analysis for 21 variants [Supplementary Table 7]. Only three loci [CIITA, TSPAN32 and NCF4] showed genome-wide significant associations following conditional analyses on the reported European lead SNPs, suggesting the presence of three novel secondary associations within previously reported loci [Table 1]. The MANTRA analyses of CD showed eight loci with Log₁₀BF > 6 [Supplementary Table 6]. Only two of these loci, rs57275892 [PSMA6-NFKBIA] and rs7170683 [LRRK1], were novel [Table 1]. Of the three loci with Log₁₀BF > 6 in the MANTRA analyses of UC [Supplementary Table 6], rs17371986 [VANGL2] showed an independent association within a previously known locus following conditional analysis [Supplementary Table 7]. Of seven novel associations, only rs12928665 [CIITA] and rs2072711 [NCF4] showed evidence of heterogeneity between Asian and European populations [p_het ≥ 0.91; Table 1, Supplementary Table 6].

Table 1.

Seven novel associations identified through Asian and trans-ethnic meta-analysis on inflammatory bowel disease.

Chr	SNP	Position	Effect allele	Study	Effect allele frequency		Inflammatory bowel disease
		[hg19]				Controls	Cases	OR	p	p random	p Q
						[N]	[N]	[95% CI]
22	rs2072711	37 268 555	A	Korean	0.32	1 185	1 332	1.21	0.002
								[1.07–1.36]
[NCF4, CSF2RB]				East Asian	0.26	3 719	2 824	1.21	6.55 × 10–7
								[1.13–1.28]
				European	0.17	48 485	38 155	1.00	0.954
								[0.97–1.03]
				Combined1 †		4 904	4 156	1.21	5.07 × 10 –9**	5.07 × 10 –9**	0.97
								[1.13–1.29]
				Combined2‡		53 389	42 311	1.03	0.027	0.121	4.42 × 10–7
								[1.00–1.05]
16	rs12928665	10 971 474	A	Korean	0.43	1 185	1 332	0.86	0.008
[CIITA]								[0.75–0.97]
				East Asian	0.46	3 719	2 824	0.83	1.29 × 10–7
								[0.76–0.90]
				European	0.76	48 485	38 155	0.94	1.29 × 10–8
								[0.91–0.96]
				Combined1 †		4 904	4 156	0.84	3.93 × 10 –9**	3.93 × 10 –9**	0.61
								[0.79–0.89]
				Combined2 ‡		53 389	42 311	0.92	1.23 × 10 –13**	0.006	0.002
								[0.90–0.94]
11	rs2074023	2 325 581	A	Korean	0.20	1 185	1 332	0.88	0.061
[TSPAN32]								[0.74–1.01]
				East Asian	0.25	3 719	2 824	0.87	8.45 × 10–4
								[0.79–0.95]
				European	0.40	48 485	38 155	0.95	1.71 × 10–7
								[0.93–0.97]
				Combined1†		4 904	4 156	0.87	1.29 × 10–4	1.29 × 10–4	0.92
								[0.81–0.94]
				Combined2 ‡		53 389	42 311	0.94	1.05 × 10 –9**	0.005	0.09
								[0.92–0.96]
5	rs2624435	17 095 269	A	Korean	0.14	1 185	1 332	0.92	0.287
[MYO10, BASP1]								[0.76–1.08]
				East Asian	0.14	3 719	2 824	0.89	0.035
								[0.79–1.00]
				European	0.23	48 485	38 155	0.94	1.93 × 10–7
								[0.92–0.96]
				Combined1†		4 904	4 156	0.90	0.019	0.019	0.79
								[0.83–0.98]
				Combined2 ‡		53 389	42 311	0.94	1.80 × 10 –8**	1.80 × 10 –8**	0.62
								[0.92–0.96]
15	rs7170683	101 488 582	A	Korean	0.48	1 185	1 332	0.98	0.748
[LRRK1]								[0.87–1.09]
				East Asian	0.47	3 719	2 824	0.98	0.673
								[0.91–1.06]
				European	0.52	48 485	38 155	0.96	1.98 × 10–4
								[0.94–0.98]
				Combined1†		4 904	4 156	0.98	0.597	0.597	0.97
								[0.93–1.05]
				Combined2 ‡		53 389	42 311	0.97	2.11 × 10–4	2.11 × 10–4	0.81
								[0.95–0.98]
14	rs57275892	35 747 271	A	Korean	0.35	1 185	1 332	0.80	1.98 × 10–4
[PPP2R3C, KIAA0391, PSMA6, NFKBIA]								[0.69–0.92]
				East Asian	0.33	3 719	2 824	0.92	0.025
								[0.84–0.99]
				European	0.17	48 485	38 155	0.95	1.50 × 10–4
								[0.92–0.98]
				Combined1†		4 904	4 156	0.88	9.25 × 10–5	0.029	0.06
								[0.83–0.94]
				Combined2 ‡		53 389	42 311	0.94	5.55 × 10–7	0.013	0.02
								[0.92–0.96]
1	rs17371986	160 402 259	A	Korean	0.83	1 185	1 332	0.88
[VANGL2]								[0.73–1.02]	0.080
				East Asian	0.79	3 719	2 824	0.95	0.205
								[0.86–1.03]
				European	0.79	48 485	38 155	0.95	5.18 × 10–5
								[0.93–0.98]
				Combined1†		4 904	4 156	0.93	0.047	0.047	0.39
								[0.86–1.00]
				Combined2 ‡		53 ,389	42 311	0.95	8.27 × 10–6	8.27 × 10–6	0.55
								[0.93–0.97]

Chr	Crohn's disease					Ulcerative colitis
	Cases	OR	p	prandom	pQ	Cases	OR	p	prandom	pQ
	(N)	(95% CI)				[N]	[95% CI]
22	637	1.29	6.95 × 10–4			695	1.14	0.082
[NCF4, CSF2RB]		[1.11–1.49]					[0.98–1.31]
	1 690	1.26	2.17 × 10–7			1 134	1.15	0.009
		[1.17–1.35]					[1.04–1.26]
	20 550	0.96	0.008			17 647	1.05	0.008
		[0.92–0.99]					[1.01–1.08]
	2 327	1.27	6.01 × 10 -10**	6.01 × 10 -10**	0.83	1 829	1.15	0.002	0.002	0.89
		[1.18–1.37]					[1.05–1.25]
	22 877	1.00	0.986	0.218	1.31 × 10-10	19 476	1.06	3.20 × 10–4	0.017	0.15
		[0.97–1.03]					[1.03–1.09]
16	637	0.90	0.140			695	0.82	0.004
[CIITA]		[0.76–1.04]					[0.69–0.96]
	1 690	0.81	1.07 × 10–6			1 134	0.85	8.92 × 10–4
		[0.73–0.90]					[0.76–0.95]
	20 550	0.94	6.61 × 10–6			17 647	0.94	2.36 × 10–5
		[0.91–0.97]					[0.91–0.97]
	2 327	0.84	7.53 × 10–7	3.48 × 10–4	0.22	1 829	0.84	1.38 × 10–5	1.38 × 10–5	0.64
		[0.78–0.90]					[0.78–0.91]
	22 877	0.93	2.22 × 10 –9**	0.020	0.005	19 476	0.93	4.50 × 10 –8**	0.007	0.03
		[0.90–0.95]					[0.90–0.95]
11	637	0.82	0.026			695	0.93	0.361
[TSPAN32]		[0.65–0.99]					[0.77–1.09]
	1 690	0.87	0.004			1 134	0.88	0.017
		[0.77–0.96]					[0.77–0.98]
	20 550	0.95	1.04 × 10–5			17 647	0.94	1.73 × 10–5
		[0.92–0.97]					[0.92–0.97]
	2 327	0.86	2.94 × 10–4	2.94 × 10–4	0.58	1 829	0.89	0.013	0.013	0.56
		[0.79–0.93]					[0.81–0.98]
	22 877	0.94	4.18 × 10–7	0.012	0.07	19 476	0.94	1.44 × 10–6	1.44 × 10–6	0.41
		[0.92–0.96]					[0.92–0.96]
5	637	1.05	0.628			695	0.80	0.028
[MYO10, BASP1]		[0.86–1.24]					[0.61–1.00]
	1 690	0.90	0.099			1 134	0.89	0.093
		[0.78–1.02]					[0.74–1.03]
	20 550	0.96	0.003			17 647	0.93	3.64 × 10–6
		[0.93–0.99]					[0.91–0.96]
	2 327	0.94	0.262	0.525	0.19	1 829	0.86	0.008	0.008	0.42
		[0.85–1.04]					[0.76–0.96]
	22 877	0.96	0.001	0.001	0.41	19 476	0.93	2.80 × 10–7	0.005	0.26
		[0.93–0.98]					[0.90–0.96]
15	637	0.86	0.027			695	1.11	0.117
[LRRK1]		[0.72–0.99]					[0.98–1.25]
	1 690	0.93	0.073			1 134	1.08	0.097
		[0.84–1.01]					[0.99–1.18]
	20 550	0.94	4.97 × 10–7			17 647	0.98	0.212
		[0.92–0.96]					[0.96–1.01]
	2 327	0.91	0.007	0.007	0.34	1 829	1.09	0.024	0.024	0.76
		[0.84–0.97]					[1.01–1.18]
	22 877	0.94	1.99 × 10 –8**	1.99 × 10 –8**	0.39	19 476	0.99	0.608	0.345	0.04
		[0.92–0.96]					[0.97–1.02]
14	637	0.82	0.006			695	0.79	7.59 × 10–4
[PPP2R3C, KIAA0391, PSMA6, NFKBIA]		[0.67–0.96]					[0.65–0.93]
	1 690	0.89	0.014			1 134	0.96	0.399
		[0.81–0.98]					[0.86–1.06]
	20 550	0.93	4.89 × 10–6			17 647	0.98	0.355
		[0.90–0.96]					[0.95–1.02]
	2 327	0.87	4.34 × 10–4	7.98 × 10–4	0.30	1 829	0.90	0.008	0.167	0.03
		[0.81–0.94]					[0.82–0.97]
	22 877	0.92	2.38 × 10 –8**	3.91 × 10 –4	0.20	19 476	0.97	0.064	0.145	0.01
		[0.89–0.95]					[0.94–1.00]
1	637	0.83	0.046			695	0.92	0.341
[VANGL2]		[0.66–1.01]					[0.74–1.10]
	1 690	0.99	0.913			1 134	0.88	0.037
		[0.89–1.10]					[0.77–1.00]
	20 550	0.99	0.576			17 647	0.92	1.95 × 10–8
		[0.96–1.02]					[0.89–0.95]
	2 327	0.95	0.276	0.368	0.09	1 829	0.89	0.023	0.023	0.74
		[0.87–1.04]					[0.81–0.98]
	22 877	0.99	0.384	0.414	0.17	19 476	0.92	1.58 × 10 –9**	1.58 × 10 –9**	0.83
		[0.96–1.02]					[0.89–0.94]

Chr, chromosome; CI, confidence interval; OR, odds ratio; Position, chromosome position. Genome-wide significant results from the combined analysis are indicated in bold.

**Association results at the genome-wide significance level [p < 5.0 × 10^–8].

†Combined1: Korean+East Asian dataset.

‡Combined2: Korean+East Asian+European dataset.

3.3. Novel loci

All three newly discovered SNPs are non-coding variants, rs2624435 in MYO10-BASP1 at 5p15 for IBD, rs57275892 in PPP2R3C/KIAA0391/PSMA6/NFKB1A at 14q13, and rs7170683 in LRRK1 at 15q26 for CD. The SNP densities across the MYO10-BASP1 and LRRK1 loci were sparse [Figure 1A, 1B]; however, upon examining the eQTL database, rs2624435 showed an association with BASP1 expression levels in blood, and rs7170683 showed an association with both LRRK1 and ALDH1A3 expression levels in lymphoblastoid cell lines of European ancestry [Supplementary Table 8]. A novel IBD susceptibility locus at rs2624435 is located between MYO10 and BASP1. The biological function of the Brain Acid Soluble Protein 1 [BASP1] gene is not well known, except for its involvement in nephrogenesis as a transcriptional co-suppressor,²⁴ whereas an unconventional myosin, Myosin X [MYO10], is implicated in phagocytosis via its role in filopodia induction. Macrophages lacking Myo10 showed markedly reduced filopodia formation.²⁵ A novel CD susceptibility locus at rs7170683 is located in the Leucine-rich repeat kinase 1 [LRRK1]. LRRK1 belongs to a member of the ROCO family of proteins with multiple functional domains including ankyrin-like repeats, leucine-rich repeats [LRRs], a Ras-like GTPase domain [ROC] and an adjacent C-terminal domain [COR], and a serine–threonine kinase domain. Its homolog LRRK2 was shown to be associated with CD²³ and Parkinson’s disease.²⁶ LRRK1 is involved in a variety of functions including autophagy and osteoclast differentiation. A recent knock-out mice study reported that LRRK1 plays a critical role in B cell development and antibody production by regulating NF-kB signaling.²⁷ The other novel CD susceptibility SNP [rs57275892] at 14q13 is located ~500 bp upstream of PSMA6 [Figure 1C] in a LD region of 645 kb [35.212−35.857 Mb] that includes BAZ1A, SRP54, FAM177A1, PPP2R3C, KIAA0391, PSMA6 and NFKB1A. Although rs57275892 is located ~500 bp upstream of PSMA6, eQTL analyses showed that it was associated most significantly with the expression of a nearby gene, protein phosphatase 2 regulatory subunit B″ gamma [PPP2R3C] [p = 2.00 × 10^–22] in whole blood [Supplementary Table 8]. PPP2R3C is a regulatory subunit of a serine/threonine phosphatase, protein phosphatase 2. Knock-out and transgenic mice studies show that PPP2R3C is involved in both the survival of germinal center B cells and the differentiation of peritoneal B cells into autoantibody-producing plasma cells, suggesting possible roles in inflammatory diseases.²⁸

Figure 1.

Signal plots for the three novel IBD loci: [A] rs2624435 at 5p15, [B] rs7170683 at 15q26 and [C] rs57275892 at 14q13. The left panels represent signal plots from the fixed-effects meta-analyses of Korean and East Asian–ancestry individuals. SNPs are plotted according to their chromosomal positions [NCBI Build 37] with –log₁₀p values from the Asian meta-analysis in the region flanking 750 kb on either side of the marker SNP. The circles indicate the genotyped SNPs, and the squares indicate the imputed SNPs. The most strongly associated SNP in the discovery stage is shown as a small purple circle. Linkage disequilibrium [LD; r² values] between the lead SNP and the other SNPs are indicated using colours. The relative location of the annotated genes and the direction of transcription are shown in the lower portion of the figure. The estimated recombination rates of the Asian samples from the 1000 Genomes Project [Nov 2014] are plotted to reflect the local LD structure. Plots are generated using LocusZoom. The right panel presents signal plots for the MANTRA association signal after trans-ethnic meta-analysis of Korean, East-Asian, and European Immunochip data. Each point represents an SNP that passed the QC in the MANTRA analysis, plotted with their BF [on a log₁₀ scale] as a function of genomic position [NCBI Build 37]. Plots were generated using R.

3.4. Novel secondary associations within previously known loci

The association signal [rs2074023] at chromosome 11p15 is located within intron 1 of Tetraspanin32 [TSPAN32, Log₁₀BF = 7.49, p_meta = 1.05 × 10^–9] in a LD region of 117.8 kb that includes C11orf21, TSPAN32, CD81 and TSSC4 [Table 1, Supplementary Figure 6]. rs2074023 is located ~452 kb away from the previously reported rs907611^8,9 [r² < 0.2]; however, conditional logistic regression analysis on rs907611 supported the independent effects of the two SNPs [Supplementary Table 7]. Examination of the eQTL database for rs2074023 showed that it is associated with the expression levels of both C11orf21 and TSPAN32 in lymphoblastoid cell lines and small intestine, whereas previously reported rs907611 is associated with the expression levels of CTSD in blood [Supplementary Table 8], supporting their independent effects. A search of HaploReg v4 for rs2074023 showed a Regulome DB score 1b, indicating that it is likely to affect binding and to be linked to expression of a gene target [Supplementary Table 9]. TSPAN32 is a member of the tetraspanins of integral membrane proteins with functional roles in cell motility, membrane fusion, proliferation and immunity. Tetraspanins, of which over 30 have been identified in humans, can associate with one another and with other molecules such as integrins or proteins of the immunoglobulin superfamily to form a network on the surfaces of many different cell types.²⁹ Knock-out mice experiments have shown that TSPAN32 may play a role in the negative regulation of peripheral T-lymphocyte proliferation,³⁰ suggesting that it may have roles in inflammatory diseases. The biological function of the other gene, C11orf21, is unknown; however, rs7944004, located in 5.7 kb 3' of C11orf21 [14.4 kb away from rs2074023 with r² = 0.75], is associated with chronic lymphocytic leukemia.³¹ The association signal [rs17371986] at chromosome 1q23 is located at 3.8 kb from the 3' end of VANGL planar cell polarity protein 2 gene [VANGL2] in a LD region of 110.7 kb that includes NCSTN, NHLH1 and VANGL2 [Table 1, Supplementary Figure 6]. The previously reported rs4656958 [~2 kb 5' of ITLN1] is located ~454 kb away from rs17371986 [r² < 0.2]; however, conditional logistic regression analysis of rs4656958 supported the independent effects of the two SNPs [Supplementary Table 7]. Examination of the eQTL database for rs17371986 showed that it is associated with VANGL2 expression levels in the sigmoid colon in the GTEx database [Supplementary Table 8]. As VANGL2 is involved in the regulation of planar cell polarity, its involvement in UC susceptibility is not obvious.

3.5. Loci showing heterogeneity between Asian and European populations

Of the two novel IBD associations that showed significant heterogeneity between Asian and European populations, association effects of NCF4 were mapped to two independent SNPs in two populations [p_het = 1.0, Supplementary Table 6 and Supplementary Figure 6]. The top SNP rs2072711 identified in the Asian-specific meta-analyses showed much stronger association with CD than with UC, whereas the recently reported European top SNP rs4821544 [10 kb away from rs2072711, r² < 0.2 in ASN] showed a significant association with CD [OR = 1.07, p = 6.75 × 10^–8] only [Supplementary Table 10]. rs2072711 showed weak associations with CD and UC in the European dataset with different effect directions, but rs4821544 failed to show associations with IBD in the Asian samples. The conditional analysis on rs4821544 did not abolish the association at rs2072711 in East Asian [p = 6.55 × 10^–7, p_condition = 6.15 × 10^–7] and Korean samples [p = 2.16 × 10^–3, p_condition = 1.95 × 10^–3], suggesting two independent associations in the NCF4 locus. rs2072711 is located on chromosome 22q12 in a LD region of ~60.3 kb, which includes two genes, NCF4 [neutrophil cytosolic factor 4] and CSF2RB [colony stimulating factor 2 receptor beta common subunit]. Examination of eQTL databases for rs2072711 showed that it had a much stronger association with the mRNA expression levels of NCF4 [p = 6.20 × 10^–15 in lymphoblastoid, p = 4.35 × 10^–55 in blood] than with those of CSF2RB [p = 1.10 × 10^–5 in lymphoblastoid] [Supplementary Table 8].^20,32,33NCF4 encodes a cytosolic regulatory component of superoxide-producing phagocyte NADPH oxidase, a multicomponent enzyme system important for host defense. The risk alleles for both rs4821544 and rs2072711 were associated with decreased expression of NCF4 in the blood eQTL database, p = 3.24 × 10^–28 and p = 4.35 × 10^–55, respectively.²⁰

rs12928665 in intron 1 of the MHC class II transactivator gene [CIITA] at chromosome 16p13 showed significant heterogeneity between Asian and European populations [Log₁₀BF = 7.76, p_het = 1.00; p_{trans-ethnic meta} = 1.23 × 10^–13p_Q = 2.14 × 10^–3] [Supplementary Table 6 and Supplementary Figure 6]. Of note is the significant difference in frequency of rs12928665 between Asian and European populations [Supplementary Table 11]. Previous studies have reported two independent association signals, rs529866 in the SOCS1 locus^8.9 [401.8 kb away from rs12928665] for CD and rs11641184 in the LITAF locus⁸ [733.1 kb from rs12928665] for IBD. The conditional analysis on rs529866, rs11641184 and both did not abolish the association at rs12928665, suggesting three independent associations in the CIITA-SOCS1-LITAF locus [Supplementary Table 11]. Previously, rs4781011 of CIITA [intron 2, 3838 bp away from rs12928665] was reported to be associated with UC in Europeans, but not to be present at 200 IBD susceptibility loci or present at the level of genome-wide significance.³⁴ LD between rs12928665 and rs4781011 in Asians [JPT + CHB] was low [r² < 0.2], whereas it was almost complete in Europeans [r² = 0.95]. Indeed, association of rs12928665 with IBD [p = 1.29 × 10^–8] was stronger than that of rs4781011 in the European dataset [p = 4.21 × 10^–8], suggesting that the association previously reported might have been due to an indirect association caused by the LD between the two SNPs. Examination of the eQTL database for the three SNPs showed that they were associated with the expression levels of different genes in blood, indicating that their effects are independent [Supplementary Table 8]. CIITA is an important transcription factor for the expression of HLA class II molecules and is involved in the expression of HLA class I molecules.^35,36

4. Discussion

Here we present the largest Immunochip meta-analysis of IBD, using all published Immunochip association results of IBD. Despite the limitation of using only 80 291 common SNPs for the analyses, we identified seven novel associations, including three novel susceptibility loci at MYO10-BASP1 for IBD, PPP2R3C/KIAA0391/PSMA6/ NFKB1A and LRRK1 for CD, and four novel secondary associations within previously known loci at NCF4, TSPAN32, CIITA and VANGL2. In addition, we were able to replicate 103 loci of known IBD susceptibility loci in Asian samples [Supplementary Table 5]. The comparison of the effect sizes for all 193 SNPs from known loci and seven novel associations identified in the present study between Asians and Europeans revealed a positive correlation in the direction of effects for both CD and UC [Figure 2] [r² = 0.40 and p = 1.54 × 10^–23 for CD; r² = 0.57 and p = 1.84 × 10^–37 for UC]. This observation is consistent with a previous large-scale study that reported substantial genetic overlap between Europeans and Asians.⁸

Figure 2.

Comparison of odds ratios [ORs] for the 193 reported and 7 newly identified Crohn’s disease and ulcerative colitis risk variants in Europeans and East Asians [based on the minor allele in Europeans]. The 193 reported SNPs are listed in Supplementary Table 5 and 7 newly identified SNPs in Table 1. Each dot represents the ORs [on a log scale] for each SNP in [A] Crohn’s disease and [B] ulcerative colitis. Colour denotes the range of association p values for Crohn’s disease or ulcerative colitis in the discovery dataset. The red line refers to linear regression, which was weighted by the inverse of the variance of the log[ORs] in the discovery dataset. The correlation coefficients and p values of the linear model are shown at the bottom-right corner.

Of the two novel IBD associations showing significant heterogeneity between Asian and European populations, the association effects of NCF4 were mapped to two independent SNPs in two populations. The fact that the risk alleles for both SNPs were associated with decreased expression of NCF4 in blood eQTL database suggests that the heterogeneity of the most significant signals might be due to the LD difference between Asians and Europeans [Supplementary Figure 7]. The NCF4-CSF2RB region showed genome-wide significant associations with both IBD and CD in Asians, but with CD only in Europeans. Previous studies have shown association of rs4821544 with ileal CD in Europeans^37–39 and its effects on reactive oxygen species production following stimulation with GM-CSF.⁴⁰ Stronger association of NCF4 in Asians could be due to the fact that ileocolonic disease is the most common type of CD in Asian populations, whereas ileal, colonic and ileocolonic disease occur in equal proportions in the CD of western populations.

In this study, we conducted the largest Immunochip trans-ethnic meta-analysis of IBD and discovered seven novel associations, including three novel susceptibility loci and four novel independent associations within previously known loci. The new loci suggest that additional factors are involved in the T cell biology and B cell immunity of IBD. In conclusion, this trans-ethnic study advances our understanding of the genetic architecture of IBD susceptibility by discovering novel associations and revealing allelic heterogeneity between Asian and European populations.

Web Resources

The URLs for data presented herein are as follows:

METAL, http://csg.sph.umich.edu/abecasis/metal/

The 1000 Genome Project, http://www.1000genomes.org/

UCSC Genome Browser, http://genome.ucsc.edu/

GCTA, http://cnsgenomics.com/software/gcta/

IIBDGC, https://www.ibdgenetics.org/

RegulomeDB v2, http://www.broadinstitute.org/mammals/haploreg/haploreg.php

Genotype–Tissue Expression [GTEx] project, http://www.gtexportal.org/home

eQTL Blood Browser, http://www.genenetwork.nl/bloodeqtlbrowser/

Geuvadis/1000 Genomes resources, http://www.ebi.ac.uk/Tools/geuvadis-das/

Funding

This work was supported by a Korean Health Technology R&D Project grant through the Korea Health Industry Development Institute to S-KY [A120176], funded by the Ministry of Health & Welfare, as well as a Mid-career Researcher Program grant through the National Research Foundation of Korea to KS [2014R1A2A1A09005824, 2017R1A2A1A05001119], funded by the Ministry of Science, Information & Communication Technology and Future Planning, the Republic of Korea. MH was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education [NRF-2016R1A6A3A01013069].

Conflict of Interest

None declared.

Supplementary Data

Supplementary data are available at ECCO-JCC online.

Author Contributions

KS and S-KY obtained financial support. KS conceived and designed the study and supervised the data analysis and interpretation. BDY and BH participated in the study design and supervised the data analysis and interpretation. S-KY, BDY, SHP and HSL recruited subjects and participated in the diagnostic evaluation. TH, KDT, JIR, JL and DPBM, participated in genotyping and supervised the data analysis. SYB and T-HK provided additional control data. MH, SJ, JH, JB, WL, YL, BMK and SBL performed data analyses. KS and MH drafted the manuscript. KS revised the manuscript.

Glossary

Abbreviations:

CD

Crohn’s disease

CI

confidence interval

GWAS

genome-wide association study

IBD

inflammatory bowel disease

LD

linkage disequilibrium

MAF

minor-allele frequency

OR

odds ratio

SNP

single nucleotide polymorphism

UC

ulcerative colitis