Reassessment of the type I diabetes association of the OAS1 locus

Abstract

To reassess the type I diabetes (T1D) association of the OAS1 locus, the Type I Diabetes Genetics Consortium (T1DGC) genotyped 11 tag single-nucleotide polymorphisms spanning ~41 kb from the 5′ to 3′ flanking region. For each sample obtained from over 2000 affected sib-pair families from nine cohorts, the genotyping was performed on both the Illumina Golden Gate and Sequenom iPlex platforms. The data suggest that there may be a weak association with T1D for two OAS1 polymorphisms, rs3741981 and rs10774671, in populations of European descent. The OAS1 locus is close to a recently identified T1D-associated linkage disequilibrium (LD) block in human chromosome 12q24. Extended LD in populations earlier examined may account for the prior observation of an association of T1D with OAS1 variants. This possibility needs to be addressed further by fine mapping of the T1D association represented in 12q24.

Introduction

The 2′,5′-oligoadenylate synthetase genes (OAS1, OAS2, OAS3) that are located in human chromosome 12q24 encode a family of enzymes pivotal in innate anti-viral defense.^1–4 OAS1 has a major function in the total constitutive activity of OAS enzymes.⁵ The first reported association of type I diabetes (T1D) with the OAS1 locus was with the single-nucleotide polymorphism (SNP) rs10774671.⁶ This SNP occurred in a region that involved a splicing site. The minor (G) allele of rs10774671 introduces a splicing site and is associated with increased OAS enzyme activity; further, this SNP was associated with T1D susceptibility.⁶

Although methodological limitations of that study were later identified,⁷ in our independent study on candidate polymorphisms, the highly correlated (r² = 0.69) non-synonymous (Ser162Gly) SNP rs3741981 of the OAS1 gene was significantly associated with T1D risk.⁸ In two large cohorts of European descent, however, no statistically significant association of OAS1 SNPs with T1D was observed.⁷

To reassess the T1D association of the OAS1 locus, the Type I Diabetes Genetics Consortium (T1DGC) genotyped 11 tag SNPs. The 11 SNPs span ~41 kb from the 5′ to 3′ flanking region of the OAS1 gene and capture 95.9% of the available HapMap (http://www.hapmap.org) SNPs within and around OAS1 with minor allele frequency >1% at r²≥0.80 (Supplementary Figure 1). Both the splicing site SNP (rs10774671) and the non-synonymous SNP (nsSNP rs3741981) were included.

Results

All 11 SNPs selected for the OAS1 locus were genotyped by both the Illumina GoldenGate and the Sequenom iPlex platforms. Of the 11159 individuals from 2298 families (5003 affected), a total of 1477 individuals had no genotypes from either platform because of unavailability of samples at the time of genotyping. In addition, 322 individuals had only the Illumina genotypes and 401 individuals had only Sequenom genotypes. As this rate of genotype availability is similar to that of other candidate genes in the T1DGC experiment, these missing genotypes seem to be unrelated to assay quality.

The genotyping quality assessment of each SNP (Table 1) eliminated rs7135579 because of low call rate and significant deviation of the rs7135579 genotype distribution from Hardy–Weinberg Equilibrium expectation. Both of the earlier associated OAS1 candidate SNPs, rs3741981 and rs10774671, have genotype distribution in Hardy–Weinberg Equilibrium and exhibit highly consistent results between the genotyping platforms.

Table 1.

The 11 SNPs in the OAS1 locus studied by the T1DGC

Marker	Physical position	ILMN call rate	ILMN HWE p	SQNM call rate	SQNM HWE p	Concordance rate of ILMN vs SQNM
rs3741982	111791701	0.998	0.070	0.984	0.138	0.958
rs12177	111798145	0.999	0.514	0.982	0.443	0.957
rs2240193	111798381	0.999	0.179	0.985	0.393	0.993
rs2240191	111798451	0.997	0.743	0.970	0.006	0.989
rs12309946	111798975	0.998	0.869	0.992	0.857	1.000
rs4766662	111808419	0.984	0.457	0.981	0.214	0.978
rs3741981a	111811590	0.999	0.166	0.988	0.137	0.999
rs10774671a	111819913	1.000	0.188	0.994	0.162	0.999
rs7135579	111829602	0.930	5.58×10−18	0.989	0.016	0.970
rs3803057	111831454	1.000	0.398	0.990	0.793	0.999
rs7967461	111832479	1.000	0.133	0.993	0.184	0.999

Abbreviations: HWE, Hardy–Weinberg equilibrium; ILMN, Illumina GoldenGate; SQNM, Sequenom iPlex; SNP, single-nucleotide polymorphism; T1D, type I diabetes; T1DGC, Type I Diabetes Genetics Consortium.

^aEarlier reported to be associated with T1D.⁶,⁸

The results of association analysis between the OAS1 SNPs and T1D are shown in Table 2. The discrepancies in results between the two technologies are largely because of incomplete overlap of the DNA sets tested (7% of the samples were genotyped in only one of the two platforms), rather than technical differences between platforms as shown by the high concordance rates in Table 1. In these affected sib-pair families from the T1DGC, there was no evidence of a statistically significant allelic association of OAS1 SNPs with T1D risk.

Table 2.

The T1D association test of the 11 OAS1 SNPs

Marker	Minor allele	MAF	ILMN Z	ILMN P	SQNM Z	SQNM P
rs3741982	A	0.383	0.192	0.848	−0.510	0.610
rs12177	A	0.491	−0.294	0.768	0.699	0.484
rs2240193	T	0.066	−1.138	0.255	−1.734	0.083
rs2240191	A	0.170	−1.157	0.247	−1.134	0.257
rs12309946	A	0.040	0.508	0.611	0.408	0.683
rs4766662	A	0.222	−0.091	0.927	0.797	0.426
rs3741981	G	0.433	0.484	0.628	0.987	0.324
rs10774671a	C	0.365	0.842	0.400	1.925	0.054
rs7135579	A	0.103	−2.058b	0.040b	−0.154	0.877
rs3803057	A	0.049	−0.871	0.384	−1.980	0.048
rs7967461	G	0.370	0.859	0.390	1.784	0.074

Abbreviations: ILMN, Illumina GoldenGate; MAF, minor allele frequency; SQNM, Sequenom iPlex; SNP, single-nucleotide polymorphism; T1D, type I diabetes.

^aThe Type I Diabetes Genetics Consortium (T1DGC) genotyped the DNA antisense strand. The C allele of rs10774671 corresponds to the G allele in the sense strand, which produces a splicing site.

^bPoor genotyping assay.

In our earlier study, the genotypic association with T1D was under a recessive model.⁸ In this model, only homozygosity for the minor allele of each of the three OAS1 SNPs (rs3741981, rs10774671, and rs3177979) conferred an increased risk for T1D. The affected sib-pair family collection assembled by the T1DGC suggests the same result for OAS1 SNPs rs3741981 and rs10774671 (Table 3) in the Sequenom assay. In the Illumina assay, the genotypic association under the recessive model did not reach statistical significance, although a trend in the same direction as that observed with the Sequenom assay was seen. The same pattern of genotypic association was also shown by the Wellcome Trust Case–Control Consortium study (P = 0.020).⁹ The three reported OAS1 SNPs were not genotyped in the Wellcome Trust Case–Control Consortium; however, SNP rs2660 has r² = 1 with rs10774671 (Supplementary Table 1). The genotypic association was not replicated, however, in the 1552 T1D families and 4287 T1D cases vs 4735 controls from the same population.⁷ Thus, the possible genotypic association still lacks statistical validation. The observed borderline significance may be due to the multiple hypotheses implicitly tested.

Table 3.

The genotypic association test of rs3741981 and rs10774671

Marker	Genotype	Frequency	ILMN Z	ILMN P	SQNM Z	SQNM P
rs3741981	A/A	0.327	0.599	0.549	0.441	0.659
	A/G	0.481	−1.541	0.123	−1.858	0.063
	G/G	0.192	1.507	0.132	2.133	0.033a
rs10774671	C/C	0.137	1.634	0.102	2.013	0.044a
	C/T	0.455	−1.087	0.277	−0.363	0.717
	T/T	0.408	0.075	0.940	−1.051	0.293

Abbreviations: ILMN, Illumina GoldenGate; SQNM, Sequenom iPlex.

The genotypic type I diabetes (T1D) association was tested by the family based association test (FBAT) (http://www.biostat.harvard.edu/~fbat/fbat.htm).

^aP<0.05.

Discussion

The above results suggest that association of the two OAS1 SNPs, rs3741981 and rs10774671, with T1D in populations of European descent is extremely weak, if present at all. This weak and inconsistent association may be due to long-range linkage disequilibrium (LD) with the recently replicated association with T1D in the region of human chromosome 12q24, involving the intronic SNP rs17696736¹⁰ following the Wellcome Trust Case–Control Consortium study.⁹ This association was also replicated in Stage 2 of our own genome-wide association study (Hakonarson et al., unpublished data). The T1D-associated SNP rs17696736 is located in a LD block of over 1 Mb in length, which is only ~300 kb from the LD block containing OAS1. The LD between these two blocks is low in European-descent HapMap set (Supplementary Figure 2).

In the Canadian family dataset reporting T1D association of the OAS1 variant,⁸ rs17696736 has D′ = 0.085, r² = 0.005 with the splicing site SNP rs10774671, and D′ = 0.061, r² = 0.003 with the nsSNP rs3741981. However, it is possible that the association signal in the chromosome 12q24 region may originate in an unidentified variant, with higher extended LD with the OAS1 SNPs. This may account for the T1D association in the earlier reports.^6,8 This issue needs to be addressed by fine mapping of the T1D association at 12q24.

Materials and methods

Subjects

The T1DGC created a resource of 2295 affected sib-pair families from nine cohorts for this experiment. SNPs in 21 candidate genes (including OAS1) were genotyped on two platforms (Illumina GoldenGate and Sequenom iPlex). Details of the sample, quality control, and other aspects of the data can be found in this volume (Brown et al.¹¹). The majority of the subjects were Caucasian (81%) with 18% unknown ethnicity and 1% other (Asian, African American, Pacific Islander).

Genotyping

All 11 OAS1 SNPs were genotyped by both the Illumina GoldenGate and the Sequenom iPlex platforms. Of the 9682 individuals, 93% were genotyped in both platforms.

Statistics

Data quality checks, using standard methods were performed by the Coordinating Center of the T1DGC at Wake Forest University. T1D association was tested by the family based association test (http://www.biostat.harvard.edu/~fbat/fbat.htm).¹² As the T1DGC families have multiple siblings (affected sib-pair families), the option of the empirical variance was used in the FBAT statistics to permit a robust, but unbiased test of genetic association.