Evaluation of IL12B as a candidate type I diabetes susceptibility gene using data from the Type I Diabetes Genetics Consortium

Abstract

As part of its efforts to identify genes affecting the risk of type I diabetes (T1D), the Type I Diabetes Genetics Consortium commissioned an extensive survey of variants associated with genes reported earlier to have an association with disease susceptibility. In this report, we present the analysis of a set of single-nucleotide polymorphisms (SNPs) within and flanking the IL12B gene, which encodes the p40 subunit of the cytokines interleukin (IL)-12 and IL-23. No SNP showed individually significant association in the population as a whole. Nevertheless, subjects stratified according to genotype at the earlier reported SNP in the IL12B 3′UTR, rs3212227, confirmed small, but significant, differences in age of disease onset with a relative hazard = 0.88 (P = 0.005). The protective effect of rs3212227 allele 2 was gender specific (P = 0.004 overall and P = 0.0003 when unaffected siblings were considered). Among females, the 2.2 genotype was more protective, with relative hazard = 0.75. We conclude that while there was no major effect of IL12B polymorphisms on T1D susceptibility in the entire study group, they have an impact on a subset of at-risk individuals.

Introduction

Type I diabetes (T1D) is a consequence of autoreactive lymphocytes that initiate a response that destroys the insulin-producing beta cells.^1,2 It is not known how these lymphocytes escape appropriate immune regulation, but both environmental and genetic factors contribute. Chief among the genes which influence T1D susceptibility are those encoded in the major histocompatibility complex, especially those encoding the class II HLA-DQ molecules.^1,2 Other non-HLA genes also contribute to risk, although for many of these, it has been difficult to confirm the original reports. This may reflect population-specific differences, failure to replicate because of under-powered studies, or original type 1 errors.

The Type I Diabetes Genetics Consortium (T1DGC) has assembled a resource of over 11 000 individuals from >2000 affected sib-pair families.³ In an effort to resolve which of the reported genes contribute to T1D risk globally, the T1DGC sponsored a comprehensive study using this resource. Where possible, the original single-nucleotide polymorphism (SNP) in each reported gene was tested, supplemented by other SNPs in and around the test gene. The investigators who reported the original data for each gene were invited to join a working group to analyze the data, as was any other T1DGC member with an interest in the relevant gene. In this paper, analyses of the IL12B gene are described.

IL12B encodes the p40 subunit of the heterodimeric cytokines interleukin (IL)-12 and IL-23.^4,5 These cytokines have diverse functions in the immune response. In particular, they have strong effects in determining the nature of the immune response that is to be mounted. IL-12 influences activated T cells to secrete cytokines typical of inflammatory immune response, whereas IL-23 mediates recruitment and activation of cells required for the induction of chronic inflammation.^4,5 These and other properties make these cytokines, and hence IL12B, candidates for involvement in the dysregulated auto-immune response leading to T1D. Indeed, IL-12 was shown to accelerate T1D in the diabetes-prone NOD mouse⁶ that shared an Il12b allele with other strains liable to autoimmune disease.⁷

The first report implicating IL12B in T1D susceptibility came from a study of Australian and UK families.⁸ These families showed linkage to distal chromosome 5, and alleles at the 3′ end of the IL12B gene were differentially transmitted to affected offspring. These results were later confirmed in studies of T1D subjects from the USA,⁹ Japan,¹⁰ and Spain.¹¹ Other studies did not confirm an IL12B effect in different geographic regions and/or from other ethnic groups.^12–15

It is now clear that genetic variants of IL12B contribute to susceptibility of various diseases that have in common an underlying immune dysregulation. Support for this has accumulated from many studies. These diseases include dermatitis^16,17 infections (cerebral malaria, tuberculosis, and leprosy),^18–23 asthma,^24,25 and, most recently, steroid-sensitive childhood nephrotic syndrome.²⁶ In these studies, the major effects of IL12B have been attributable to polymorphisms in the 3′UTR, or in its promoter, or from a combination of both. In a major study of psoriasis, a further SNP, located 60 kb telomeric from IL12B, was also found to contribute independently to susceptibility.²⁷ In addition, the demonstration that polymorphisms in other genes can impact on IL12B expression and IL-12 production emphasizes the complexity of this system.²⁸

In light of the above, this study has evaluated association with T1D risk of a set of 27 SNPs, including the commonly assessed 3′UTR SNP.

Results

Genotypes were determined at 27 SNPs across ~44 kb of the chromosomal region containing IL12B. First, an evaluation of quality control was performed. Among the 11 281 individuals genotyped, there were 12 061 genotypes missing from the Illumina set and 4623 from the Sequenom set, with only 712 genotypes missing from both platforms. A total of 1667 SNP genotypes had conflicting genotypes on these two platforms. Examination of the key reported IL12B SNP, rs3212227 that affects a TaqI site in the 3′UTR,²⁹ revealed that there were 31 discrepancies between the two datasets. These discrepancies were all scored as heterozygotes by Illumina, but were scored as a homozygote (of each genotype) by Sequenom. The Sequenom scoring failed to detect heterozygotes, but did not have a consistent bias as to which allele it failed to detect. Manually checking the genotypes by PCR and TaqI digest showed that all but one of these conflicting samples was a genuine heterozygote. Therefore, we concluded that the Illumina data were more reliable, so used these data for further analyses. Sequenom data were used only for those samples that had no corresponding Illumina data.

Pedigree error checking by SibShipper (Mehta and Morahan, in preparation) discovered a total of 21 Mendelian errors. There were no Mendelian errors in genotypes from the IL12B SNP rs3212227. These genotyping errors occurred when family member(s) were not genotyped by Illumina, and the Sequenom data was used. These erroneous genotypes were set to ‘missing’ before conducting statistical genetic analyses.

Linkage disequilibrium (LD) across the region was calculated and an LD map generated using Haploview.³⁰ The IL12B gene has high LD in the non-coding regions at each end of the gene, but reduced LD within the gene itself (Figure 1). One of the blocks of highest LD included seven markers centromeric of IL12B, whereas the other major block was over 10 kb telomeric from the promoter. These results are consistent with earlier studies showing that there were high frequencies of subjects having all four combinations of the IL12B promoter and 3′UTR polymorphisms.^18,22,24

Figure 1.

LD map across the IL12B locus. The genomic organization of IL12B²⁹ across 45 kb of chromosome 5 is depicted in the top panel, showing location of exons (black bars) and SNPs (asterisks). The IL12B gene is in the reverse orientation with respect to the centromere. Study SNPs 1–14 are colored white, while SNPs 21–27 are black. The 3′UTR SNP rs3212227 (SNP #8) and rs17056705 (SNP #21) are circled. LD was calculated from parental genotypes using Haploview³¹ and color-coded according to the ‘Confidence bounds’ system, whereby D values over 90 are colored black. The two blocks of highest LD include the centromeric markers and those in the 3′ end of the gene (SNPs 1–14), and markers 21–27, which are telomeric of IL12B.

Preferential transmission of IL12B SNP alleles was assessed using the parental transmission–disequilibrium test as implemented in Plink.³¹ Only a few IL12B SNPs had nominal P<0.05 (Table 1); however, there was no significant effect of any SNP, including the initially reported association rs3212227, in the combined dataset after correction for multiple testing. For further analyses, not wishing to incur a penalty for testing SNPs that had shown no prior evidence of association, only the 3′UTR SNP, rs3212227, was considered.²⁹

Table 1.

Transmission disequilibrium test analyses of IL12B SNPs

SNP	BP	A1 A2	T	N	P
rs7730126	158662525	A G	1443	1407	—
rs1549922	158664126	G A	1664	1677	—
rs4921466	158665350	G A	677	640	—
rs6859018	158669570	T C	1097	1122	—
rs17875325	158671385	G C	1007	1010	—
rs3181225	158673201	T C	989	999	—
rs17875324	158674365	C T	638	591	—
rs3212227	158675528	G T	1091	1116	—
rs2853696	158677194	A G	1020	938	0.06
rs919766	158678860	G T	710	667	—
rs2853694	158680526	C A	1669	1687	—
rs2569253	158682193	G A	1593	1693	0.08
rs3181219	158683859	T C	699	652	—
rs17875303	158685525	A T	1347	1451	0.05
rs1433048	158687191	G A	999	937	—
rs2546893	158688858	T C	1614	1675	—
rs2546890	158690524	G A	1700	1649	—
rs10052709	158692190	G C	817	809	—
rs7709212	158693856	G A	1574	1532	—
rs6868898	158695523	C T	1386	1527	0.009
rs17056704	158697189	G C	1323	1481	0.003
rs13188370	158698855	T A	231	216	—
rs17056705	158700521	T C	328	310	—
rs17056706	158702188	T C	1459	1594	0.01
rs953861	158703854	G A	974	927	—
rs4244437	158705520	C T	1504	1433	—
rs11747112	158707187	G C	868	849	—

The position on chromosome 5 of the 27 IL12B single-nucleotide polymorphisms (SNPs) is indicated. The number of transmissions (T) and non-transmissions (N) of each allele (A1, A2) from heterozygous parents was calculated using Plink.³⁰ For clarity, only nominal P-values <0.1 are shown.

The major allele of rs3212227 was found earlier to be associated with an earlier age of onset of T1D in a population-based cohort.³² This effect was examined within the T1DGC set of families. In view of earlier unpublished results, suggesting a stronger association in females, males and females were distinguished in evaluating an effect. Among females, but not males, the rate of disease onset was slower for carriers of the minor allele compared with homozygotes for the major allele (that is the allele earlier reported to be associated with susceptibility). This effect was observed both in an analysis of affected children (relative hazard = 0.88, P = 0.005, N = 4579) and in an analysis of all children from cohorts incorporating age of collection for unaffected siblings (relative hazard = 0.84, P = 0.002, N = 3450; Table 2; Figure 2). For both analyses, there was a significant difference between the sexes in the effect of IL12B genotype (P = 0.004 and 0.0003, respectively).

Table 2.

Relative hazard of IL12B genotypes

Genotype	Females		Males
	Relative hazard	P	Relative hazard	P
1.2	0.85	0.007	1.11	0.06
2.2	0.74	0.03	1.04	0.7

Relative hazards of IL12B genotypes were calculated for males and females, comparing genotypes bearing the minor allele with homozygotes for the major (‘susceptibility’) allele (that is, genotype 1.1).

Figure 2.

Effect of IL12B genotype on age of onset of T1D. Kaplan–Meier survival plots of age of onset, stratified according to genotype, where ‘1’ indicates the most frequently occurring allele. Carriage of the minor allele is associated with significant protection/delayed onset among females, but not males.

Discussion

Earlier studies have shown that the different IL12B genotypes affect the level of expression of this gene, with the major (‘susceptibility’) allele associated with higher expression. This is consistent with the notion that higher amounts of IL-12 would tend to skew T-cell responses to the Th1 pathway. The emerging consensus from studies of IL12B in other diseases is that rather than contributing to disease susceptibility per se, it affects disease severity in predisposed individuals.^16–27 In the context of a binary disease such as T1D, ‘severity’ approximates to disease onset. The analyses of the T1DGC data showing that there is an impact of IL12B genotypes on age of onset in females is consistent with a function in mediating disease severity.

There is evidence that IL12B effects are found in some populations, but not in others.^8–15 It is well established that T1D has a substantial environmental component. Many environmental factors are likely to differ among the populations examined in the various studies cited; an obvious one in which Australia differs from northern Europe and parts of the USA is in sunlight exposure. Exposure to UVB can induce IL-12 production by human keratinocytes in vivo,³³ suggesting one way in which gene-environment variables may impact on T1D susceptibility in different populations.

These results confirm earlier reports that the allele associated with higher IL-12 production, and hence more likely to influence pro-inflammatory immune responses, is associated with earlier disease onset in at least a subset of T1D subjects.

Materials and methods

Subjects

Recruitment of affected sib-pair families was reported elsewhere.³ DNA samples were available for this study from a total of 11 281 individuals from over 2000 families recruited from each of the four T1DGC Networks (Asia-Pacific, Europe, North America, and UK).

Genotyping

The IL12B SNP earlier reported to be associated with T1D susceptibility affects a TaqI restriction site within the 3′UTR;²⁹ this SNP is now designated rs3212227. Other SNPs within and flanking IL12B were chosen from HapMap. Selection of SNPs was completed before the report of the 60 kb telomeric SNP associated with psoriasis,²⁷ so this SNP was not included in this study. It should also be noted that the insertion–deletion polymorphism in the IL12B promoter that is commonly associated with affects on IL-12 production and susceptibility to disease^18,24 was not included on either of the two genotyping platforms. The T1DGC sponsored the genotyping of these SNPs using both the Illumina and Sequenom platforms. Genotyping was carried out at Broad Institute of Harvard/MIT, and data were subjected to quality control and pedigree cleaning by the T1DGC Coordinating Center (Brown et al.,³⁴ this volume) before analysis by the IL12B Working Group.

Analyses

The following programs were used to manage, analyze, and visualize the data: Plink,³¹ Haploview,³⁰ and SibShipper (http://sibshipper.waimr.uwa.edu.au/). Instructions for use of each of these programs are available on their respective web sites. Proportional hazards age-of-onset analyses were carried out using S-PLUS 8.0.4 (Insightful Corp., Seattle, WA, USA). Data were stratified by cohort of origin (to give a ‘common’ effect across cohorts) and had coefficient standard errors adjusted for within-family correlation.