Our sample of 390 genotyped Caucasian SLE trio pedigrees was predominantly female (96%) with a mean±SD age-of-onset of 31±12 years-old. All SNPs had over 90% genotyping success except for rs241453 (86.0%) and rs241425 (88.9%). In the trios analyzed, the genotyped SNPs that met our quality criteria thresholds achieved a mean r2 = 0.87 across 8.76 kb of TAP1 (6 tag SNPs), and a mean r2 = 0.68 across 16.94 kb of TAP2 (9 tag SNPs). From the total of 390 genotyped pedigrees, we had an average of 256 fully genotyped trios for the TAP genes. This sample yields 0.80 power to detect odds ratios (OR) greater than 1.5 under an additive model for alleles whose frequency is greater than 0.20.
We found significant evidence of association between TAP2 and SLE status (), as shown by the increased risk of SLE for individuals with the G allele at rs241453 (P = 1.3 × 10-6, O.R. = 2.39). This effect is considered significant even if we use a stringent Bonferroni multiple testing correction on the total number of tested SNPs (n=15) to set a significance threshold of P = 0.0033. We questioned the genotyping accuracy for this SNP (86.0%), but as explained below, we believe that this association is corroborated by that of its neighbor rs241447, whose genotyping success was 97.5%.
Single Marker Family-Based Association and Conditional Logistic Regression Analysis of TAP1 and TAP2.
More modest effects within TAP2 were also observed, with an increased risk for individuals with the T allele at rs241447 (P = 0.0021, O.R. = 1.46), with the G allele at rs183585 (P = 0.02, O.R. = 1.32), and with the G allele at rs3819714 (P = 0.027, O.R = 1.31). shows the LD structure of TAP2 with the localization of the nine genotyped SNPs. The SNPs rs241453 and rs241447 are 525 bp from each other and are in strong linkage disequilibrium (LD) (r2 = 0.93) in this sample. Haplotype analysis of the TAP2 gene showed that the haplotype defined by rs241453 and rs241447 was significant (P = 0.0011 for the GT haplotype) (data not shown), but it was not as strong an association as rs241453 in isolation. Thus, the association at rs241447 might be due to LD with rs241453.
LD structure of TAP2. The nine genotyped SNPs are indicated by the red lines, numbered as shown in .
Depending on the isoform, rs241453 (P = 1.3 × 10-6, O.R. = 2.39) localizes to the 3′UTR or the last intron of TAP2; it may thus be associated with altered stability of the TAP2 mRNA. Interestingly, rs241447 (P = 0.0021; O.R. = 1.46), which is in a conserved region with regulatory potential, locates to either the last intron or in a predicted last exon, in a human mRNA from GenBank (UCSC browser under URLs).
The rs241453 effect seems to be separate from the HLA haplotypes consistently associated with SLE. Using the HapMap browser (see URLs), we searched for variants in LD (r2 > 0.5) with this SNP in a region of 550 kb around TAP2, from upstream HLA-DRB1 to HLA-DPB1 (). According to the HapMap B35 assembly, rs241453's highest LD (r2 = 1) in this region is with two TAP2 SNPs, one adjacent to it in the 3′UTR (rs241452), and the other in the last intron (rs2857104). A third intronic SNP in TAP2 also shows high LD with rs241453 (rs241442, r2=0.82). Outside of TAP2, the highest LD (r2 = 0.76) is found with an intronic SNP in HLA-DQA2 (rs17500482). Nineteen other SNPs, none in proximity to known genes, show modest LD (r2 ~ 0.5-0.6) with rs241453.
HapMap LD (as measured by the r2) between rs241453 on TAP2 and selected SNPs.
Given the established SLE association with HLA-DRB1
alleles, we specifically focused on the LD with this locus and its HLA-DR2
alleles. Based on the HapMap B35 data (see URLs) the LD between rs241453 and the tag SNPs that capture the HLA-DRB1
alleles is extremely low (r2
< 0.02, ). The LD values are similar between these HLA-DRB1
SNPs and the four aforementioned SNPs in LD with rs241453. The highest LD observed between these HLA-DRB1
and any TAP2
and SNPs is low (r2
< 0.4). These data suggest that the association at rs241453 is not due to LD with HLA-DRB1
We formally tested the hypothesis that the association observed with TAP2 is separate from the HLA-DRB1 locus, namely the DRB1*15xx (HLA-DR2), DRB1*03xx (HLA-DR3) and DRB1*08xx (HLA-DR8) allele groups. In our collection of 191 trios, the conditional logistic regression analysis continued to show differential transmission at rs241453 (P = 1.8 × 10-5, OR = 2.56). The HLA-DRB1*03xx allele group was also significant (P = 1.4 × 10-6, OR = 2.63), but the HLA-DRB1*15xx and HLA-DRB1*08xx alleles were not associated with SLE.
After separately adjusting for either HLA-DRB1*15xx, HLA-DRB1*03xx or HLA-DRB1*08xx, in the TDT conditional logistic regression model, rs241453 remained associated with SLE (P = 1.3 × 10-5 adjusting for HLA-DRB1*15xx, P = 1.6 × 10-4 adjusting for HLA-DRB1*03xx, and P = 6.0 × 10-6 adjusting for HLA-DRB1*08xx, ). This suggests that the TAP2 rs241453 association is a separate effect from HLA-DRB1*15xx, HLA-DRB1*03xx and HLA-DRB1*08xx (). Both rs241453 (P = 1.6 × 10-4) and HLA-DRB1*03xx (P = 2.3 × 10-4) provide evidence for association with SLE in this model. When we include an interaction term in the conditional logistic regression analysis, the P-value for the interaction is highly significant for HLA-DRB1*03xx (P < 1.0 × 10-6), but not for HLA-DRB1*15xx or HLA-DRB1*08xx. These data provide preliminary evidence of a potential interaction between TAP2 and HLA-DRB1*03xx (P < 1.0 × 10-6). Given the modest sample size of 191 trios, this interaction forms an interesting hypothesis in need of a well-powered independent replication study. It is noteworthy that the presence of at least one HLA-DRB1*03xx allele group increases the OR for the TAP2 association. In the 113 trios where the SLE diagnosed offspring were homozygote DRB1*03xx-negative affecteds OR = 2.00 (CI=1.17-3.42), increasing to OR= 3.83 (CI=1.56-9.41) in the 65 DRB1*03xx-heterozygote trios, and to OR= 7.00 (CI=0.86-56.90) in the 13 homozygote DRB1*03xx-positive affected trios. Overall, having at least one HLA-DRB1*03xx allele (78 trios) increased the OR to 4.29 (CI=1.88-9.76), from OR = 2.00 (CI=1.17-3.42) in 113 homozygous DRB1*03xx-negative trios, thus supporting the evidence of an interaction between TAP2 and HLA-DRB1*03xx.
There was no evidence of an association between the SNPs in TAP1 and SLE status (). This may represent a true lack of association or it may reflect that the fact that the allele frequencies are low for the SNPs genotyped and the resulting magnitude of any association in TAP1 is less than we are powered to detect.