Background

Type 2 diabetes (T2D) is a complex disease influenced by environmental and genetic factors. Genetic association studies have thus far identified at least 20 replicating T2D susceptibility loci of modest to small effect, which together explain less than 10% of the genetic component of disease [1, 2]. Several genome-wide association scans (GWAS) have been carried out for T2D [3–10]. These have used a variety of genotyping platforms with different SNP content, typically capturing over 80% of common variation in European-descent populations. Although this extent of coverage, in combination with imputation approaches [11], reduces the need for candidate gene studies, in-depth investigation of variation at loci of interest can conceivably prove useful in characterising them further.

The TNF/LTA locus has been a long-standing T2D candidate gene. T2D and obesity have been hypothesised to have an inflammatory basis [12, 13]. Insulin resistance is associated with increased plasma levels of proinflammatory cytokines such as TNF and IL6, and with interactions between TNF and NFkappaB that lead to an increase of oxidative stress [14–16].

The genes coding for TNF and LTA reside in the class III MHC region on chromosome 6p21.3. TNF and LTA are members of the TNF ligand superfamily, bind the same TNF receptors and mediate similar pleiotropic effects [17, 18]. Of the multiple SNPs in the TNF/LTA gene region, the rs361525 (G-238A) and rs1800629 (G-308A) TNF promoter variants, and the rs1041981 (T60N) LTA variant have been the most frequently studied in T2D. The majority of studies of TNF/LTA SNPs have been small-scale, with some notable exceptions [17], and have produced no convincing evidence for association with the disease [19–27].

The Wellcome Trust Case Control Consortium (WTCCC) T2D GWAS examined 17 directly typed and imputed SNPs from the TNF/LTA gene region and detected no association with T2D in 2000 cases and 3000 controls from the UK [6, 28]. In addition, a GWAS meta-analysis for T2D carried out by the DIAGRAM consortium, which examined the same 17 directly genotyped and imputed SNPs in the TNF/LTA region in samples from three sources (Diabetes Genetics Initiative (DGI), Finland-United States Investigation of NIDDM Genetics (FUSION) and WTCCC) also found no association between TNF/LTA SNPs and T2D [28]. However, the WTCCC genotyping platform (Affymetrix 500k) and HapMap-based imputation do not provide exhaustive coverage of common variation in this gene region. To increase coverage, we carried out a genetic association study of the TNF/LTA loci in a total of 5359 samples from the UK by typing additional SNPs, selected on the basis of sequence data to better capture variation in the region.

Methods

Subjects

This study comprised a case-control and a family-based component. The case-control dataset included 1520 cases from the Diabetes UK Warren 2 Sib Pair Repository (61.5% males) and 2570 control samples from the 1958 British birth cohort (n = 2027, 50.6% males) [29], and the HRC control collection (n = 543, 49.9% males) derived from UK blood donors and available from the European Centre for Cell Culture (ECACC, CAMR, Salisbury, UK). The family-based dataset comprised 423 parent-offspring trios (58.5% male probands) from the Diabetes UK Warren 2 Trios (W2T) Repository. W2T probands were selected by strict clinical, immunological and genetic criteria as previously described [30]. All cases included in the present study had T2D diagnosed according to the World Health Organization criteria and were selected for early diabetes onset and/or positive family history. Importantly, autoimmune diabetes was excluded based on GAD antibody typing, age of disease onset above 25, insulin independence following diagnosis, no ketoacedosis and no first degree relatives with type 1 diabetes [30, 31]. Clinical characteristics of the cases are provided in Table 1. 58.2% of WTCCC cases and 35.3% of WTCCC controls overlapped with the samples examined as part of our study. All subjects were exclusively of UK/Irish origin and provided signed informed consent prior to blood sampling. Reported investigations have been carried out following the principles of the Declaration of Helsinki as revised in 2000. Ethical oversight for collection and use of the T2D cases was provided from MREC 00/6/55, Peterborough and Fenland LREC 05/Q0106/78 and from over 100 individual local research ethics committee approvals. Use of the 1958 Birth Cohort samples is in accordance with Joint UCL/UCLH Research Ethics Committee A approval 08/H0714/40 and South-East Multi-Centre Research Ethics Committee approval MREC 01/1/44. The HRC samples are a commercially available set of anonymised DNA samples from blood donors sourced from the Health Protection Agency Culture Collection and approved for research use only.

Table 1 Clinical characteristics of T2D cases
Full size table

SNP Selection and Genotyping

Eleven haplotype-tagging SNPs (tag SNPs) (rs928815, rs909253, rs746868, rs1041981 (T60N), rs1800750, rs1800629 (G-308A), rs361525 (G-238A), rs3093662, rs3093664, rs3093665, rs3093668) were selected across the TNF and LTA loci from sequencing and genotype data generated in 32 Caucasian trios as previously described [32, 33] (Table 2). Genotypes for this study were determined using a fluorescence-based competitive allele specific assay (Kaspar, Kbioscience, UK). Genotyping and SNP selection details are available from the authors on request. When we investigated linkage disequilibrium (LD) between the SNPs, we found SNPs rs928815 and rs746868, and SNPs rs909253 and rs1041981 to be in tight LD with each other (r2 > 0.8). SNP pairs rs361525-rs3093662, rs361525-rs3093668 and rs3093662-rs3093664 displayed moderate LD (0.6 < r2 < 0.8), while all other investigated SNPs showed no or very weak pairwise LD (0 < r2 < 0.52) (Additional file 1, Figures S1a and S1b). Eight of the 11 studied tag SNPs have been genotyped in the HapMap (rs928815, rs909253, rs1041981, rs1800750, rs1800629, rs3093662, rs3093665, rs3093668). These 8 variants capture 61.1% of common (MAF > 0.05) variation across the TNF/LTA gene region (overall 18 SNPs in the HapMap database) at an r2 threshold of ≥0.8, based on the HapMap (CEU population, Rel 24/PhaseII Nov 08 [34]). The proportion of common variation across the region captured by the typed tag SNPs was calculated on a multimarker tagging basis using Tagger [35]. However, this is an underestimate of the capture of common variation, since the 3 remaining non-HapMap variants were excluded from this calculation. To asses coverage further we performed another calculation on the basis of the 1000 genomes project data (http://www.1000genomes.org) April 2009 release).

Table 2 Characteristics of 11 TNF/LTA tag SNPs.
Full size table

Statistical analysis

Quality control (QC) of the obtained genotypes was performed prior to association analysis. The SNP genotyping success rates ranged from 93.3% to 98.6%. We evaluated the comparative rate of missing genotypes between cases and controls using Plink (version 1.00) [36] and excluded rs3093662 from the case-control association analysis due to low call rate. The tag SNPs were tested for deviation from Hardy-Weinberg equilibrium (HWE) in affected and healthy individuals separately using Stata v. 8 (Stata Corporation, College Station, TX, USA) and Plink (version 1.00) [36]. No deviations from HWE were observed. Minor allele frequencies (MAFs) of controls in both studies were compared with the National Center for Biotechnology Information SNP database (NCBI dbSNP) MAFs for the CEU population and showed no significant differences. Testing of Mendelian inheritance using Plink and Haploview [36, 37] identified inconsistencies in one family, which was excluded from further analysis. After QC, 10 tag SNPs were taken forward to case-control association analyses and 11 tag SNPs were included in family-based association analysis.

Single-point case-control association analyses were carried out using Stata v. 8 (Stata Corporation, College Station, TX, USA). Multi-point case-control association analyses of fixed haplotype sizes (sliding windows of 2-10 SNPs shifting 1 SNP at a time) were performed using the expectation-maximisation algorithm-based approach implemented in Plink [36]. Single-point and multi-point (sliding windows of 2-11 SNPs) family-based association analyses were carried out using implementations of the transmission disequilibrium test (TDT) in Plink [36]. 10,000 permutations were run for each association analysis. r2 and D' measures of pairwise LD were calculated for all SNPs using Haploview [37]. Power was calculated under the log-additive model for a range of effect-sizes (1.1<OR>2) at α = 0.05 using Quanto [38]. All association analyses are unadjusted (e.g. for BMI, blood pressure and other environmental variables), as these data were not available to us. We did not investigate gene-environment interactions.

Results

Genotype distributions for the 11 TNF/LTA tag SNPs in the case-control and parent-offspring datasets are shown in Additional file 1, Table S1 and Table S2, respectively. Overall, we did not identify any consistent significant SNP associations with disease. The most frequently studied SNPs, rs1800629 (G-308A), rs361525 (G-238A), and rs1041981 (T60N) did not show robust association with the disease in any dataset (Table 3). Exhaustive multimarker case-control analyses did not identify any strong haplotypic associations (data not shown). There were no statistically significant deviations in the transmission of alleles from parents to affected probands by single-point (Table 4) or haplotype-based analysis (data not shown). Our study had 80% power at α = 0.05 to identify modest/large effect sizes (OR > 1.3) at common loci (Additional file 1, Tables S3a and S3b).

Table 3 Case-control association analysis results for the 10 TNF/LTA tag SNPs
Full size table
Table 4 Transmission disequilibrium analysis of 11 TNF/LTA tag SNPs in T2D parent-offspring trios
Full size table

In the WTCCC GWAS [5, 28], a total number of 17 (one directly genotyped, rs1799964, and 16 imputed) SNPs from the TNF/LTA gene region were investigated and showed no association with T2D. The case-control association results of the most frequently studied SNPs, rs1800629 (G-308A), rs361525 (G-238A) and rs1041981 (T60N) from the present study and from the WTCCC dataset (across which there is considerable overlap) is shown in Table 5. Five of the tag SNPs from the present study (rs746868, rs1800750, rs361525, rs3093664 and rs3093665) were not directly typed or imputed in the WTCCC GWAS. To assess the extent of additional coverage these 5 SNPs offer, we examined LD based on our genotype data in T2D cases and controls. SNPs rs746868 and rs361525 are in high LD (r2 = 0.99 and r2 = 0.79) with two of the WTCCC-typed SNPs, rs928815 and rs3093668 respectively. The remaining 3 SNPs that have not been examined in the WTCCC (rs1800750, rs3093664 and rs3093665) demonstrate low LD with WTCCC-typed and other HapMap SNPs (0 < r2 < 0.52) (Additional file 1, Figures S1a and S1b). Therefore these polymorphisms capture additional variation missed by the WTCCC study.

Table 5 The comparison of association results for T60N, G-308A and G-238A between the present study and the WTCCC T2D GWAS
Full size table

We investigated capture further on the basis of the 1000 genomes project data. Four of our 11 tag SNPs (rs909253, rs1800750, rs3093662 and rs3093665) were not found in the 1000G dataset and the remaining 7 tag SNPs capture 60.6% of common variation (overall 33 TNF/LTA SNPs in the 1000G dataset) on a multimarker tagging basis at an r2 threshold of ≥0.8. This is again an underestimate of the TNF/LTA common variation capture by our tag SNPs.

Discussion

In this study of 11 tag SNPs, we find no consistent evidence for association between TNF/LTA region variation and T2D. The present study was designed to analyse a set of tag SNPs specifically selected to capture the majority of common variation in the TNF/LTA gene region based on proprietary sequence and genotype data [32, 33]. Although a proportion of the investigated variants had been examined as part of the WTCCC GWAS [5, 28], this study provides further capture of common variation across the region. However, the overall conclusion remains unchanged - there was no evidence of association with disease.

This is one of the largest studies to date, showing no association between TNF/LTA variation and T2D. A recent meta-analysis (2106 cases and 2920 controls) of the rs361525 (G-238A) variant did not detect a significant association with T2D [23]. Similar meta-analyses of all reported association studies for the rs1800629 (G-308A) and rs1041981 (T60N) SNPs, which have been widely investigated with respect to T2D, may boost power to detect possible small effects at these loci.

T2D is a complex disease caused by complex interplay between environmental and genetic factors. A limitation of our study is that we have not been able to adjust for or investigate interaction of SNPs with BMI, age, gender, blood pressure, serum lipid levels etc. as these data were unavailable to us. In addition, even though our study examined the majority of common variation across the region, it is possible that causal, associated variants may have been missed.

Conclusions

The purpose of this study was to examine if genetic variation in the genes encoding inflammatory proteins TNF and LTA alter the risk of developing T2D. We tested a carefully selected set of haplotype tagging SNPs that capture the majority of common variation in the TNF/LTA gene region in case-control and parent-offspring samples and find no robust evidence for association. Large-scale meta-analyses will be required to investigate the presence of smaller effects at polymorphic sites in the TNF/LTA gene region.