Identification of additional type 2 diabetes susceptibility loci through large-scale replication using "Metabochip": early result

Background and aims: Over 30 common variant signals influencing type 2 diabetes (T2D)-susceptibility have been identified so far. However, replication efforts to date have focused on limited numbers of the most strongly-associated signals and have failed to exploit fully the information provided by the well-powered genome wide association (GWA) meta-analyses available for stage 1 discovery. The Metabochip - a custom iSELECT array containing ~ 195,000 SNPs - has been designed to support large-scale follow-up of putative associations for T2D and other metabolic and cardiovascular traits, and to enable the fine-mapping of known loci.


Materials and methods: This analysis is based on Metabochip data from 3,185 T2D cases and 3,569 controls (from the UK T2D Genetics Consortium sample recruited in Tayside, Scotland) that were successfully called (Gencall v1.1) and passed quality control (QC). Of 185,802 Metabochip SNPs passing QC, 4,821 were included in this analysis because they capture the top ~5,000 independent autosomal signals from the DIAGRAM (v3) GWA meta-analysis (12,057 T2D cases, 56,071 controls, all of European-descent). Association analysis was performed under an additive model with adjustment for 3 principal components to account for sample substructure.


Results: We observed directional consistency for all published T2D-loci (and for 11 novel autosomal loci derived from the most recent DIAGRAM+ consortium effort) including variants at TCF7L2 (P<10-14); SLC30A8, KCNQ1, FTO (P<10-4 ); KCNJ11 and IRS1 (P<5x10-3). We compared overall patterns of replication between the DIAGRAM stage 1 discovery data and the UKT2DGC follow-up results for 4,333 independent SNPs represented in both samples. Of these, 2,468 SNPs showed directionally consistent effects (binomial p<10-19) with 217 of the 2,468 also showing nominal replication (i.e. p<0.05, same effect direction). Only 79 of 4,333 SNPs had p<0.05 in the opposite direction (binomial p<10-18). Despite the modest size of this first follow-up sample (compared to the discovery set), joint analysis of DIAGRAM+Stage 1 and UKT2DGC Stage 2 data revealed several signals near to or exceeding genome-significance, including a locus near ARL15 (ADP-ribosylation factor-like 15) on chromosome 5p15 (UKT2DGC P=10-4, meta-analysis P=3.2×10-8, OR=1.11[95%CI 1.07-1.16]).


Conclusion: These preliminary data obtained from the Metabochip are consistent with a long tail of common variant association signals of modest effect contributing to T2D susceptibility. Metabochip-genotyping is currently underway in over 30,000 T2D cases and 50,000 controls of European descent. Follow-up on this scale should add considerably to the tally of proven T2Dsusceptibility loci. Even in this modestly scaled initial follow-up effort, we have identified one novel T2D-susceptibility signal mapping to chromosome 5p15.