Single nucleotide polymorphisms in TNFSF15 confer susceptibility to Crohn's disease

Abstract

The inflammatory bowel diseases (IBDs), Crohn's disease (CD) and ulcerative colitis, are chronic inflammatory disorders of the digestive tract. The pathogenesis of IBD is complicated, and it is widely accepted that immunologic, environmental and genetic components contribute to its etiology. To identify genetic susceptibility factors in CD, we performed a genome-wide association study in Japanese patients and controls using nearly 80 000 gene-based single nucleotide polymorphism (SNP) markers and investigated the haplotype structure of the candidate locus in Japanese and European patients. We identified highly significant associations (P=1.71×10⁻¹⁴ with odds ratio of 2.17) of SNPs and haplotypes within the TNFSF15 (the gene encoding tumor necrosis factor superfamily, member 15) genes in Japanese CD patients. The association was confirmed in the study of two European IBD cohorts. Interestingly, a core TNFSF15 haplotype showing association with increased risk to the disease was common in the two ethnic groups. Our results suggest that the genetic variations in the TNFSF15 gene contribute to the susceptibility to IBD in the Japanese and European populations.

INTRODUCTION

The chronic inflammatory bowel diseases (IBDs), Crohn's disease (CD; MIM 266600) and ulcerative colitis (UC), are chronic conditions characterized by remitting and relapsing inflammation of the small and/or large intestines. The combined prevalence of the two diseases in western countries and Japan has been increasing significantly over the last decade. Epidemiologic and linkage studies suggested that the involvement of genetic factors in the etiology of IBD (1,2). A linkage and fine-mapping approach recently identified CARD15 (NOD2) as a susceptibility gene at the IBD1 locus on chromosome 16 (3,4); three polymorphisms (R702W, G908R and 1007fs) in this gene were confirmed to be independently associated with susceptibility in Caucasian patients with CD (5). It was generally accepted that the clinical profiles of CD are similar in Caucasians and Asians (6), but extensive DNA sequence analysis of this gene in more than 400 Japanese CD patients identified such genetic variations in only a single case (7), indicating that CARD15 variants do not play a significant role in the pathogenesis of Japanese CD patients. In addition, associated variants at a second susceptibility locus, IBD5 (8), were extremely rare in the Japanese populations (9,10). Increased risk of CD is associated with different HLA alleles in Caucasians and Japanese (11). These studies suggest that there may be ethnic differences in genetic susceptibility to CD that are not explained by the minor demographic and phenotypic differences between the populations.

To identify other susceptibility genes to CD, we performed a large-scale case–control study using gene-based single nucleotide polymorphism (SNP) markers and identified very significant associations of SNPs in the TNFSF15 gene with CD in the Japanese population. The association of this gene with IBD was also confirmed in two European IBD cohorts and enabled the identification of common ‘high-risk’ and ‘low-risk’ haplotypes in these two ethnic groups.

TNFSF15 (tumor necrosis factor superfamily, member 15, also called TNF superfamily ligand A, TL1A or vascular endothelial cell growth inhibitor, VEGI) is a novel TNF-like factor expressed primarily in endothelial cells (12). A recent study has demonstrated the up-regulation of mRNA and protein levels of TNFSF15 in macrophages and CD4⁺/CD8⁺ lymphocytes of the intestinal lamina propria of CD patients (13). Our demonstration that genetic variants of TNFSF15 influence risk of CD offers a novel insight into the etiology of disease and provides information that may be important for the development of treatment and/or prevention targeting this molecule.

RESULTS

Large-scale case–control study using gene-based SNP markers

We performed a large-scale genome-wide case–control association study of CD in the Japanese population using gene-based SNP markers selected from the Japanese SNP (JSNP) database (14,15). First, we carried out genotyping of 94 Japanese CD patients by our high-throughput genotyping system that was based on a combination of multiplex PCR and the invader assay described previously (16) and successfully obtained information at 72 738 SNP loci (17–20). Overall success rate of the genotyping assay was 90.3% and their accuracy was 99.97%. The markers were evaluated for deviation from Hardy–Weinberg equilibrium, and allelic frequencies were compared with those of 752 unrelated Japanese individuals. The analyses identified 1888 SNPs with P-values of <0.01 in at least one test comparing allele frequencies, genotype frequencies, frequency of carrier status for the major allele or frequency of carrier status for the minor allele in patients vs. controls and without significant deviations from Hardy–Weinberg equilibrium in either group (P=0.01) (Supplementary Material, Table S1A and B). We subsequently performed a second screening of these 1888 SNPs with an increased number of CD patients (n=484) and identified 22 SNPs that showed P-values of <1×10⁻⁴ (Supplementary Material, Table S2). Of these 22 SNPs, seven were located within a region of ∼280 kb on chromosome 9q32 containing TNFSF15 and TNFSF8 loci; one SNP was outside of any genes (tnfsf15_1) and the remaining six (tnfsf15_54, 55, 78, 80, 85 and 87) were in TNFSF8. The remaining SNPs are distributed across other genome regions (Supplementary Material, Table S2).

Association of TNFSF15 variants with CD in Japanese patients

To further characterize the relationship between the 9q32 locus at LOC389786-TNC and CD, we constructed a high-density SNP map of the 450 kb region including the original seven SNPs. We screened for relevant SNPs by re-sequencing of four gene loci, LOC389786, TNFSF15, LOC402376 and TNFSF8, contained in this region and found a total of 143 SNPs including 28 novel SNPs. These markers were genotyped in 484 CD patients (the original 94 CD patients and 390 others) and 345 unaffected control individuals (a new control population) (Fig. 1A and B) (Supplementary Material, Table S3). Success rates of the genotyping assay for the case and control populations were 99.1 and 99.4%, respectively. The most significant association with CD was observed at an SNP locus in TNFSF15, tnfsf15_28 (14,340T→C in intron 3 of TNFSF15), with χ²=58.8, P=1.71×10⁻¹⁴ and odds ratio of 2.17 (95% CI, 1.78–2.66). In addition, several other SNPs in the TNFSF15 gene revealed strong evidence of disease association with P-values of <10⁻¹⁰ (Table 1). None of the SNPs outside of the region around TNFSF15 revealed such significant results (Supplementary Material, Table S3). The first and second control series were not significantly different for markers genotyped in both (data not shown).

Next, we constructed a high-density LD map of this region using 100 SNPs with minor allele frequencies >15% (Fig. 1C). In the case and control populations, we identified one extended block with a high degree of LD; D′ declined near tnfsf15_13 (61 T>C in exon 1 of LOC389786) and tnfsf15_48 (−10534 C>A in 3′ flanking region of LOC402376). This LD block included two genes, LOC389786 and TNFSF15. Although the LD block contains a part of LOC389786 (exon 1 only), the SNPs in this gene are much less strongly associated with CD than those in TNFSF15. To analyze which gene variants in LOC389786 or in TNFSF15 were responsible for association with CD, we performed logistic regression. After accounting for tnfsf15_28, which indicated the most significant association in TNFSF15, the other SNPs in LOC389786, namely tnfsf15_12, which was the most strongly associated marker in LOC389786, and tnfsf15_13, which was located in this LD block, were not significantly correlated with disease (P=0.52 or 0.67, tnfsf15_12 or tnfsf15_13, respectively). In contrast, the association with tnfsf15_28 was always significant, even after adjusting for other SNPs in LOC389786 in a logistic regression (P=8.36×10⁻¹², 7.78×10⁻¹²). Therefore, we concluded that the TNFSF15 gene was a susceptible gene to CD in 9q32.

Confirmation of association between CD and TNFSF15 variants in UK cohorts

To corroborate our finding of association between TNFSF15 variants and CD, we examined 10 markers spanning the region of significant association in two independent Caucasian disease cohorts from the UK. These cohorts were composed of a sample of 347 IBD trio families (161 CD, 175 UC and 11 with indeterminate colitis) and 233 IBD multiplex families (containing 263 individuals affected with CD and 196 individuals affected with UC) and an independent sample of 363 CD patients and 372 healthy controls. Five of the markers were found to be monomorphic or nearly monomorphic in the UK samples (tnfsf15_19 T allele, tnfsf15_28 T allele, tnfsf15_33 G allele, tnfsf15_38 G allele and tnfsf15_40 C allele).

The five polymorphic markers (tnfsf15_26, 31, 35, 36 and 41) were associated with the IBD phenotypes in the UK families (Table 2), and three also revealed suggestive evidence for CD association in the UK cases and controls (Table 3). The same alleles showed increased risk of CD in the UK families and unrelated cases. Similar results were obtained when the combined data from UK trio probands and cases were compared with the controls (Table 4).

As shown in Table 5, the five SNPs that were polymorphic in both the Japanese and UK samples formed two frequent disease-associated haplotypes. The most frequent haplotype (A) was a ‘high-risk’ haplotype, whereas the other (B) was a ‘low-risk’ haplotype in both Japan and UK. The patterns were consistent across all three cohorts except Japanese controls and the strongest association was indicated in Japanese [P=7.8×10⁻¹¹ and 9.3×10⁻¹¹, respectively, haplotype (A) and (B)]. The (B) haplotype was significantly associated with the disease in both UK cohorts (P=0.02 for both). Only one other haplotype (C) was observed with the frequency of >0.05, and this was not significantly associated with disease either in Japan or the UK.

DISCUSSION

In this study, we applied the strategy of the gene-based genome-wide SNP analysis for CD using a total of 72 738 SNPs (17,18). On the basis of simulations, our approach has a power to include true positive loci in the first screening at ∼50% probability if the relative risk associated with an SNP is >1.9 and the risk-allele frequency is 0.20 in case of the dominant-inheritance model. We also stimulated an expected number of false positive loci at the first screening by random permutation test. Among the 72 738 SNPs analyzed, the expected number of false positive loci was calculated to be 826 loci (1.14%). The confidence of obtaining one or more true positives in a genome-wide analysis also depends on a total number of susceptibility loci.

TNFSF15 is a novel TNF-like factor expressed primarily in endothelial cells (21). This region including TNFSF15 was reported as a locus which indicated a nominal multipoint evidence near D9S2157 (MLOD=1.41, P=0.0054) in CD family (22). Our case–control study provides additional confirmatory evidence for 9q32 and the first evidence to indicate a significant genetic association between SNPs in the TNFSF15 gene and IBD populations from different ethnic backgrounds. The determinants of a majority of complex genetic disorders are currently poorly understood, but the few examples that do exist (e.g. CARD15) demonstrate clinically important differences in gene frequency between separate racial and ethnic groups (23). Several of the TNFSF15 SNPs found to be associated with CD in Japan were not polymorphic in the UK samples. Markers that were polymorphic in both ethnic groups show similar patterns of allelic and haplotypic association with the disease. However, disease risk ratios were substantially higher in Japanese due to frequency differences for the high-risk haplotype (A), which is observed at a similar frequency in patients from the two ethnic groups but a markedly lower frequency in Japanese controls compared with UK controls. We also investigated whether these variants in TNFSF15 interact with variants in another gene(s) associated with CD, but found no evidence of interaction between TNFSF15 and CARD15 [maximum χ²=1.59 (NS) at tnfsf15_31] or between TNFSF15 and IBD5 susceptibility locus [maximum χ²=2.82 (NS) at tnfsf15_41]. Substantial differences have been described in allele frequencies for the known susceptibility-associated variants in CD in CARD15 (7), IBD5 (9,10) and MHC (11). While lack of association in one ethnic group could not be used to exclude a relationship with disease in the second ethnic group, our positive results from both Japanese and Caucasian populations provide strong evidence for the involvement of genetic variants of TNFSF15 in CD (and potentially in UC).

The novel TNFSF superfamily, TNFSF15 is abundantly expressed primarily in endothelial cells (12). The function of TNFSF15 remains to be completely elucidated, but a recent study revealed that it is a ligand for the receptors, TNFRSF25 (tumor necrosis factor receptor superfamily, member 25, also called death domain receptor 3, DR3) and TNFRSF6B (tumor necrosis factor receptor superfamily, member 6b, decoy, also called decoy receptor 3, DcR3), and that it activates NF-κB and induces apoptosis in TNFRSF25-expressing cell lines (21). It has also been recently reported that mRNA and protein levels of TNFSF15 are up-regulated in macrophages and CD4⁺/CD8⁺ lymphocytes of the intestinal lamina propria of CD patients (13).

Responses to luminal bacteria have been implicated in the pathogenesis of IBD (24), and LPS is widely used as endotoxin in Gram-negative infections. Recently polymorphisms in two lipopolysaccaride (LPS) receptor genes, CD14 and TLR4, were reported to be associated with IBD (25,26). To investigate a possible role of TNFSF15, we stimulated peripheral blood mononuclear cells isolated from six volunteers with and without LPS for 24 h. Expression of TNFSF15 and its receptor genes was measured by quantitative real-time PCR (Supplementary Material, Fig. S1). Under the LPS stimulation condition, the expression level of TNFSF15 was significantly up-regulated (P=0.031), whereas those of the receptor genes, TNFRSF25 and TNFRSF6B, were unaffected (data not shown). These preliminary results on the transcriptional induction of TNFSF15 by LPS stimulation in white blood cells imply that TNFSF15 could play an important role in protection of the intestinal barrier via the innate immune system.

The existence of disease susceptibility variations in TNFSF15 further adds to the evidence implicating multiple pathological pathways for IBD. The common disease susceptibility variants that we described may influence the expression level of TNFSF15, possibly leading to the modification of cytokine production and imbalance of the immune response in the bowel. Other factor(s) involved in the TNFSF15 signaling pathway may also be related to etiology of IBD. On the basis of our results, it is now of interest to undertake further functional studies of TNFSF15 variants to obtain a better understanding of the biochemical mechanisms that are linked to the pathogenic process in IBD. Potentially, they offer an approach for the development of novel therapies targeting this molecule or its receptor.

MATERIALS AND METHODS

Subjects and DNAs

Blood samples were obtained from 484 Japanese CD patients at the Social Insurance Chuo General Hospital [male:female ratio 73.1%; mean±SD age at diagnosis (in years) 22.4±7.7; 23 individuals with family history to IBD]. We selected 94 of the patients at random for the first stage of our large-scale association study [male:female ratio 79.6%; mean±SD age at diagnosis (in years) 21.4±7.7; four individuals with family history to IBD]. The first set of Japanese control subjects were made up from members of the general population consisting of other common-disease groups (n=752) [male:female:unknown ratio 31.4:45.4:23.1%; mean±SD age at sampling (in years) 58.4±14.8], and the second set was recruited from 345 unaffected control individuals belonging to the Osaka–Midosuji Rotary Club [male:female ratio 80.9%; mean±SD age at sampling (in years) 51.3±14.8]. All Japanese CD cases were diagnosed at the Inflammatory Bowel Unit of the Social Insurance Hospital by clinical, radiological, endoscopic and histological findings according to the Lennard-Jones' criteria (27). Patients with indeterminate colitis were excluded. DNAs were prepared from these samples according to standard protocols.

Patient and family collections from the UK (Oxford patient and family collection) have been previously described (28). The Oxford case–control panel consists of 363 CD ascertained through the John Radcliffe Hospital, Oxford IBD clinic and 372 healthy controls. The family panel is composed of 347 IBD trios (161 CD, 175 UC and 11 with indeterminate colitis), and 233 multiplex IBD families comprising 263 individuals affected with CD and 196 individuals with UC. Control samples were obtained from healthy individuals attending ‘well-person’ clinics in Oxfordshire, UK. The family collection was ascertained through probands attending the Oxford IBD clinic and others referred from Gastroenterologists around the UK. Both family and case–control cases were diagnosed as having IBD according to standard clinical, endoscopic and histological findings.

All individuals included in the study gave their written consent, and approval was obtained from the relevant ethical committees.

SNP discovery and genotyping

Approximately 280 kb of genomic sequence spanning the four genes, LOC389786, TNFSF15, LOC402376 and TNFSF8 (corresponding to reference sequences, AL160275, AL39024 and AL133412 from the GenBank database), was re-sequenced in 24 Japanese control individuals to identify SNPs. This provides complete genomic coverage of these genes with the exception of repetitive sequences. Registration of one of the putative genes, LOC402376, was subsequently eliminated in build 35 of the human genome sequence database. The Japanese samples were genotyped by PCR amplification of multiple genomic fragments with 20 ng of genomic DNA (16) followed by characterization with the invader assay (29). Genotyping of the UK Caucasian DNA samples was undertaken using the Sequenom MALDI-TOF system (30).

Large-scale association study for genes susceptible to CD

For a genome-wide association study for Japanese populations, we selected a total of 72 738 gene-based SNPs for genotyping from the JSNP database after application of validation criteria (31). Most of the markers used in this study were same as the markers used in our previous reports (17,18); a part of these SNPs that showed a low success rate of genotyping (<90%) was excluded from the data. The detailed information of these markers has been described in our previous reports (14,15).

In the first screening step, we carried out genotyping of 80 592 SNPs using genomic DNAs from 94 Japanese CD patients and 752 unrelated Japanese individuals. We then compared allele frequencies in CD patients and controls and evaluated for deviation from Hardy–Weinberg equilibrium at each of these loci. We subsequently performed the second screening of these selected SNPs showing P-value of <0.01 in at least one of comparisons using a larger number of CD patients (n=484) and a second control series. None of the markers genotyped in the second screening showed significant deviations (P=0.01 level) from Hardy–Weinberg expectations in the patient or control series (data not shown).

Statistical analysis

Genotype distributions and allele frequencies of each of the SNPs were compared, respectively, between the cases and the controls in Japanese populations as described elsewhere (32). Haplotype frequencies in Japanese samples were estimated using the expectation–maximization algorithm (33), and haplotype blocks were defined by methods as previously described (34). To analyze residual effect among SNPs, we performed logistic model by using R. Analysis of the UK family samples was conducted using standard TDT approaches for trios (35). The FBAT (36) programs were used for combined association analysis of the trios and multiplex families, using all affected individuals. The haplotype frequencies for the trio founders were estimated using the Merlin program (37) and companion program fugue that accounts for linkage disequilibrium between markers. All other haplotype frequencies were estimated using the E–M algorithm as implemented in the SNPHAP program (see URL subsequently). None of the markers at the TNFSF15 locus exhibited significant deviation from Hardy–Weinberg equilibrium (P=0.05 level) in the Japanese or Caucasian cohorts.

URLs

The JSNP database is available at http://snp.ims.u-tokyo.ac.jp/index.html. The National Center for Biotechnology Information's SNP database is available at http://www.ncbi.nlm.nih.gov/SNP. R is available at http://lib.stat.cmu.edu/R/CRAN/. FBAT is available at http://www.biostat.harvard.edu/~fbat/default.html. MERLIN and fugue are available at http://www.sph.umich.edu/csg/abecasis/. SNPHAP is available at http://www.gene.cimr.cam.ac.uk/clayton/software/.

SUPPLEMENTARY MATERIAL

Supplementary Material is available at HMG Online.

ACKNOWLEDGEMENTS

We thank H. Bando, H.M. Nan, M. Kubo, J. Hata, S. Arai and Y. Onouchi for their advice and assistance; H. Kawakami and M. Yamaguchi for their expertise in computer programming. This work was supported in part by a ‘Research for the Future’ Program Grant no. 00L01402 of the Japan Society for the Promotion of Science to Y.N. and S.M. L.R.C. was supported by a Wellcome Trust principal fellowship. D.McG. was supported by the Medical Research Council, UK, the International Organization for the Study of IBD and the Digestive Diseases Foundation (Core)/Belmont Trust. The authors declare that they have no competing financial interests.

Conflict of Interest statement. None declared.

Figure 1. Case–control association and linkage disequilibrium at the chromosome 9q32 locus associated with CD. (A) Transcriptional and SNP maps in this region. Arrows indicate the spanning region of each gene. Vertical lines indicate the positions of SNPs used in this study. (B) Case–control association plots (−log₁₀(P)) for a comparison of allelic frequencies between the Japanese case and control groups at each of the SNP loci. (C) Pairwise LD between SNPs, as measured by D′ in each of the case and control groups.

Table 1.

Summary of association of TNFSF15 between case and controls in the Japanese populations

ID	Positiona	Alleleb		Genotype of case				Genotype of control				Allele 1 vs. 2		Genotype 11 vs. 12+22		Genotype 11+12 vs. 22
		1	2	11	12	22	Sum	11	12	22	Sum	χ2 (P-value)	OR (95% CI)	χ2 (P-value)	OR (95% CI)	χ2 (P-value)	OR (95% CI)
41	−12 506	A	G	303 (63.9%)	152 (32.1%)	19 (4.0%)	474	147 (42.6%)	157 (45.5%)	41 (11.9%)	345	44.0 (3.27×10−11)	2.11 (1.69–2.64)	36.6 (1.42×10−9)	2.39 (1.80−3.17)	18.2 (1.95×10−5)	3.23 (1.84–5.67)
40	−10 642	A	C	190 (39.8%)	209 (43.8%)	78 (16.4%)	477	74 (21.4%)	153 (44.3%)	118 (34.2%)	345	52.9 (3.46×10−13)	2.09 (1.71–2.55)	31.0 (2.54×10−8)	2.42 (1.77−3.32)	35.1 (3.08×10−9)	2.66 (1.91–3.70)
38	−7505	T	G	190 (39.4%)	212 (44.0%)	80 (16.6%)	482	71 (21.0%)	152 (45.0%)	115 (34.0%)	338	51.4 (7.64×10−13)	2.07 (1.69–2.53)	31.0 (2.52×10−8)	2.45 (1.78–3.37)	33.3 (7.96×10−9)	2.59 (1.86–3.60)
36	−638	G	A	264 (56.4%)	171 (36.5%)	33 (7.1%)	468	128 (37.1%)	153 (44.3%)	64 (18.6%)	345	43.4 (4.43×10−11)	2.03 (1.64−2.50)	29.7 (5.16×10−8)	2.19 (1.65–2.92)	25.0 (5.75×10−7)	3.00 (1.92–4.69)
35	−358	C	T	273 (57.0%)	173 (36.1%)	33 (6.9%)	479	128 (37.2%)	152 (44.2%)	64 (18.6%)	344	46.0 (1.18×10−11)	2.06 (1.67–2.55)	31.4 (2.13×10−8)	2.24 (1.68–2.97)	26.4 (2.73×10−7)	3.09 (1.98–4.82)
33	1396	A	G	191 (39.9%)	208 (43.4%)	80 (16.7%)	479	72 (20.9%)	155 (45.1%)	117 (34.0%)	344	53.0 (3.38×10−13)	2.09 (1.71–2.55)	33.0 (9.00×10−9)	2.51 (1.82–3.44)	33.0 (9.46×10−9)	2.57 (1.85−3.57)
31	9706	A	G	274 (57.0%)	182 (37.8%)	25 (5.2%)	481	122 (35.6%)	157 (45.8%)	64 (18.7%)	343	56.5 (5.62×10−14)	2.24 (1.81–2.76)	36.7 (1.36×10−9)	2.40 (1.80–3.19)	37.7 (8.43×10−10)	4.18 (2.57–6.80)
28	14 340	C	T	189 (39.9%)	198 (41.8%)	87 (18.4%)	474	65 (18.8%)	157 (45.5%)	123 (35.7%)	345	58.8 (1.71×10−14)	2.17 (1.78–2.66)	41.3 (1.32×10−10)	2.86 (2.06–3.96)	31.3 (2.17×10−8)	2.46 (1.79–3.40)
26	15 524	G	A	256 (53.7%)	185 (38.8%)	36 (7.5%)	477	117 (35.0%)	149 (44.6%)	68 (20.4%)	334	43.7 (3.89×10−11)	2.02 (1.64–2.49)	27.5 (1.59×10−7)	2.15 (1.61–2.87)	28.8 (7.84×10−8)	3.13 (2.03–4.82)
25	16 806	T	A	195 (40.7%)	207 (43.2%)	77 (16.1%)	479	76 (22.3%)	154 (45.2%)	111 (32.6%)	341	49.0 (2.53×10−12)	2.03 (1.66–2.48)	30.6 (3.25×10−8)	2.39 (1.75–3.28)	30.6 (3.17×10−8)	2.52 (1.81–3.51)
20	30 075	A	G	334 (70.2%)	124 (26.1%)	18 (3.8%)	476	169 (49.7%)	135 (39.7%)	36 (10.6%)	340	42.3 (7.81×10−11)	2.17 (1.71–2.74)	35.1 (3.10×10−9)	2.38 (1.78–3.18)	14.9 (1.15×10−4)	3.01 (1.68–5.40)
19	30 886	C	T	187 (39.5%)	200 (42.2%)	87 (18.4%)	474	65 (19.0%)	156 (45.5%)	122 (35.6%)	343	56.8 (4.90×10−14)	2.15 (1.76–2.62)	39.2 (3.81×10−10)	2.79 (2.01–3.86)	31.0 (2.61×10−8)	2.46 (1.78–3.38)

CI, confidence interval.

^aThe first nucleotide of the exon 1 start site is designated as position 1 based on the reference sequence GenBank NM_005118.2.

^bAllele 1 is as risk allele.

Table 2.

Summary of the association results in UK cohorts for five polymorphic markers in TNFSF15: single-marker association results for UK families

Marker	Over-transmitted	TDT in trios				Trios+multiplex families
	allele	T	U	χ2	P-value	Z	P-value
IBD
tnfsf15_26	G	155	121	4.19	0.0407	2.30	0.0215
tnfsf15_31	A	178	120	11.29	0.0008	3.32	0.0009
tnfsf15_35	C	154	105	9.27	0.0023	2.85	0.0044
tnfsf15_36	G	160	107	10.52	0.0012	3.21	0.0013
tnfsf15_41	A	168	105	14.54	0.0001	3.68	0.0002
CD
tnfsf15_26	G	77	64	1.20	0.2736	1.20	0.2307
tnfsf15_31	A	95	65	5.62	0.0177	2.59	0.0095
tnfsf15_35	C	77	53	4.43	0.0353	1.89	0.0590
tnfsf15_36	G	78	55	3.98	0.0461	1.96	0.0501
tnfsf15_41	A	85	58	5.10	0.0240	2.56	0.0103
UC
tnfsf15_26	G	73	55	2.53	0.1116	1.66	0.0970
tnfsf15_31	A	79	53	5.12	0.0236	1.77	0.0761
tnfsf15_35	C	73	50	4.30	0.0381	1.92	0.0544
tnfsf15_36	G	77	50	5.74	0.0166	2.25	0.0246
tnfsf15_41	A	78	44	9.48	0.0021	2.38	0.0173

Table 3.

Summary of the association results in UK cohorts for five polymorphic markers in TNFSF15: results from UK case–control analyses

SNP ID	Allele frequencies		Allele 1 vs. 2		Genotype 11 vs. 12+22		Genotype 11+12 vs. 22
	Cases (n)	Controls (n)	χ2 (P-value)	OR (95% CI)	χ2 (P-value)	OR (95% CI)	χ2 (P-value)	OR (95% CI)
tnfsf15_26 (G→A)	0.74 (348)	0.68 (368)	5.73 (0.02)	1.32 (1.05–1.66)	6.57 (0.01)	1.47 (1.09–1.97)	0.74 (0.39)	1.24 (0.76–2.02)
tnfsf15_31 (A→G)	0.71 (354)	0.66 (370)	3.50 (0.06)	1.24 (0.99–1.55)	3.94 (0.05)	1.34 (1.00–1.80)	0.59 (0.44)	1.20 (0.75–1.94)
tnfsf15_35 (C→T)	0.70 (359)	0.67 (370)	2.36 (0.13)	1.19 (0.95–1.49)	2.26 (0.13)	1.25 (0.93–1.67)	0.71 (0.40)	1.23 (0.76–2.00)
tnfsf15_36 (G→A)	0.72 (348)	0.67 (372)	3.60 (0.06)	1.23 (0.99–1.56)	4.26 (0.04)	1.35 (1.02–1.82)	0.46 (0.50)	1.18 (0.73–1.92)
tnfsf15_41 (A→G)	0.73 (345)	0.69 (372)	1.27 (0.26)	1.14 (0.91–1.43)	1.46 (0.23)	1.20 (0.89–1.61)	0.18 (0.67)	1.12 (0.66–1.89)

Table 4.

Summary of the association results in UK cohorts for five polymorphic markers in TNFSF15: results from UK trio probands and CD cases compared with UK controls

SNP ID	Allele frequencies		Allele 1 vs. 2		Genotype 11 vs. 12+22		Genotype 11+12 vs. 22
	Cases (n)	Controls (n)	χ2 (P-value)	OR (95% CI)	χ2 (P-value)	OR (95% CI)	χ2 (P-value)	OR (95% CI)
tnfsf15_26 (G→A)	0.73 (524)	0.68 (368)	5.84 (0.02)	1.29 (1.05–1.59)	6.78 (0.009)	1.41 (1.09–1.85)	1.85 (0.17)	1.35 (0.87–2.08)
tnfsf15_31 (A→G)	0.71 (532)	0.66 (370)	3.73 (0.05)	1.22 (1.00–1.49)	4.72 (0.03)	1.33 (1.03–1.72)	0.93 (0.33)	1.23 (0.80–1.89)
tnfsf15_35 (C→T)	0.71 (526)	0.67 (370)	2.92 (0.08)	1.19 (0.98–1.46)	2.76 (0.10)	1.25 (0.96–1.62)	1.30 (0.25)	1.28 (0.83–1.98)
tnfsf15_36 (G→A)	0.71 (519)	0.67 (372)	4.04 (0.04)	1.23 (1.00–1.52)	4.12 (0.04)	1.32 (1.01–1.69)	1.01 (0.31)	1.25 (0.81–1.92)
tnfsf15_41 (A→G)	0.72 (516)	0.69 (372)	1.42 (0.23)	1.13 (0.92–1.40)	1.36 (0.25)	1.31 (0.83–2.08)	0.78 (0.38)	1.13 (0.87–1.46)

Results for the monomorphic or near-monomorphic SNPs are not reported. The TDT outcomes were obtained as in reference (35). The Z-scores and P-values for the analyses of trios and multiplex families were obtained using the program FBAT (36). Sample of 347 IBD trios (161 CD, 175 UC and 11 with indeterminate colitis) and 233 multiplex IBD families comprising 263 individuals affected with CD and 196 individuals with UC were analyzed. The case–control samples consisted of 363 CD patients and 372 healthy controls. The bracketed values (n) in the allele frequency columns of Tables 3 and 4 represent the number of individuals at each marker for whom genotypes were called.

Table 5.

Comparison of haplotype frequencies and association results for the five markers that were polymorphic in the three cohorts

						Haplotype frequencies and P-values in three study cohortsa
						Japanese case–control			UK families			UK case–control
Haplotype ID	26	31	35	36	41	Case	Control	P-value	Trio affected offspring	Trio founders	P-value	Case	Control	P-value
A	G	A	C	G	A	0.69	0.53	7.8×10−11	0.67	0.62	0.0005	0.65	0.62	NS
B	A	G	T	A	G	0.19	0.33	9.3×10−11	0.23	0.25	0.02	0.21	0.26	0.02
C	A	G	T	A	A	0.04	0.06	NS	0.03	0.03	NS	0.03	0.03	NS

NS, no significant.

^aAll haloptypes with estimated frequencies >0.05 in at least one of the groups are shown. P-values are reported for UK families as obtained from the FBAT program (36) using the combined trio and multiplex families. Similar results were obtained with the transmit program (data not shown).