Recent genome-wide association studies (GWAS) have identified hundreds of signals associated with type 2 diabetes (T2D) , most (∼90%) of which are found in noncoding regions of the genome enriched in tissue-specific regulatory regions. However, most GWAS signals span hundreds of tightly linked variants, and predicting each of their functional effects is difficult. Consequently, progress to translate associations to mechanisms has proceeded slowly. Here, we have designed a massively parallel reporter assay (MPRA) to systematically screen enhancer activity across a panel of genomic fragments, including (1) islet-specific TSSs defined by cap analysis of gene expression (CAGE) and (2) allelic pairs for over 10k T2D- and metabolic trait-associated variants. We cloned this library up- or downstream of a reporter gene driven by a synthetic housekeeping promoter (SCP1) or the human insulin (INS) promoter, together enabling enhancer activity measurements across different promoter and enhancer contexts. We examined expression bias within our library (FDR < 0.05) and found that 43% of CAGE-defined oligos were marked by positional bias (n = 565/1,305) , skewed towards higher expression in the upstream position (n = 539/565) . In contrast, two unique subsets of GWAS oligos displayed either positional (n = 702/11,656) or promoter (n = 698/11,656) bias. Oligos with positional bias were evenly split between the two cloning positions, while most promoter-biased oligos were more highly expressed when paired with the INS promoter (n = 512/698) . To identify sequence features associated with promoter bias, we performed Lasso regression using a set of 562 genomic annotations. This analysis revealed that presence of HNF1 motifs was associated with higher enhancer activity in the context of the INS promoter. Together, these results indicate that reporter construct design clearly influences assayed regulatory activity, particularly with use of tissue-relevant promoters.


A.Tovar: None.


National Institutes of Health (R01DK117960, T32DK101357)

Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at