Complementing qPCR with SNP data achieves largest case-control analysis of KIR3DL1/S1 copy number variation and reveals no association in type 1 diabetes

Killer Immunoglobulin-like Receptors (KIR) are important surface receptors of Natural Killer cells. KIRs mediate the fate of target cells based on the composite inhibiting/activating signal generated on binding to their corresponding HLA (Human Leukocyte Antigen) class I ligands. Type 1 diabetes (T1D) is an autoimmune disease strongly associated with genetic variation in the HLA region, primarily with Class II genes, but also with HLA Class I loci including HLA- Bw4, an epitope grouping both HLA-A and HLA-B alleles into HLA-Bw4-80I and HLA-Bw4- 80T subgroups. Amongst the 17 known genes of the KIR region, characterised by alternative haplotypes of variable gene copy numbers, KIR3DL1 is the only one proven to bind in-vitro with both HLA-Bw4 epitopes. Its activating counterpart, KIR3DS1, putatively binds only to the HLA-Bw4-80I epitope. KIR3DL1/S1 is therefore a suitable T1D candidate gene but has not been studied in GWAS as SNP arrays can only genotype normal copy number regions. So far, insufficiently powered PCR-based genotyping studies of less than 600 individuals, have failed to detect convincing association of T1D with presence or absence of KIR3DL1/S1.

For the first time, we are able to test for association in a sample size of 6,058 cases and 4,733 controls, twenty-fold larger than any previous study, by using KIR3DL1/S1 copy number calls from a recently developed KIR qPCR assay to impute copy number from ImmunoChip, a custom Illumina genotyping chip with dense coverage of autoimmune regions. The KIR3DL1/S1 copy numbers from qPCR were predicted by clustering jointly on the ΔCt values of KIR3DL1 and KIR3DS1. The imputed KIR3DL1/S1 copy numbers from ImmunoChip were predicted from raw SNP signals, Log R Ratio and B Allele Frequency, in the KIR3DL1/S1 region, using a nearest neighbour classifier trained on 749 case and 727 control samples analysed both with qPCR and on Immunochip (2.04% misclassification error rate in leave-one-out cross-validation). The association of KIR3DL1/S1 copy number with T1D was tested in all samples and on the appropriate HLA-Bw4 subsets. The interaction effect between KIR3DL1/S1 and HLA-Bw4 was tested with a case-only analysis. We can confirm from our study with a substantially increased sample size, that copy number variation in KIR3DL1/S1 is not an important modifier of T1D risk (OR < 1.1 at 95% confidence for common genotypes), either alone or dependently on HLA-Bw4 epitope, and that furthermore, there is no evidence of statistical interaction effect between HLA-Bw4 and KIR3DL1/S1 in T1D.