Fig 1. Genetic and environmental effects on gene expression. Blood insulin levels represent a cellular environment for tissues such as skeletal muscle. The left panel depicts a single genome with color-coded genomic elements and various heterozygous sites. The right panel shows the relative transcript abundance for the corresponding locus on the left panel. Some genomic elements contain genetic variants. When the variant is the same color as the element, the element is active. In some cases the variant is black, indicating that the variant renders the regulatory element nonfunctional and only basal transcription occurs. The purple element represents a gene with a transcribed SNP (tSNP), shown in the transcripts. Allele specific expression is calculated across both chromosomes and compared to the high and low environment. (A) When regulated by an insulin-responsive element (green), gene expression changes according to insulin concentrations in the extracellular environment. (B) When regulated by an insulin-independent element (orange) containing genetic variation, gene expression changes according to the presence of a genetic variant (eQTL), but not to insulin levels. The tSNP shows allelic bias due to the eQTL effect, but is not associated with the insulin environment. (C) When regulated by both an insulin-responsive element and an insulin-independent element containing genetic variation, the effects of the insulin environment and the genetic variation on gene expression may be additive, although more complex relationships are possible. The tSNP shows some imbalance due to the eQTL effect and is associated to insulin levels. Such cases may be identified as weak reQTLs. (D) When regulated by an insulin-responsive element containing genetic variation, there may exist an interaction effect between the genetic variant and insulin levels such that changes in gene expression across insulin environments depend on the genetic variant. The tSNP shows allelic imbalance associated with insulin levels due to the reQTL effect. One of several possible interaction effects depicted.

Fig 2. GxE signals. (A) Number of reQTLs per clinical variable (10% FDR). (B) Number of tSNP-environment associations per clinical variable (10% FDR).

Fig 3. GTEx replication. Replication rate (y axis) as a function of FUSION reQTL p-value cutoff (x axis). Dashed line represents two standard deviations from the null distribution, calculated using the hypergeometric distribution.

Fig 4. FHOD3 reQTL, rs17746240 (18:33970347). The data for each of the three possible reQTL genotypes are presented in separate plots (columns). The top row plots show the relationship between gene expression (y axis) and the clinical variable (x axis). The bottom row plots show the relationship between the allelic imbalance of the tSNP and the clinical variable (x axis). Note the bottom row has fewer samples because it is limited to samples heterozygous for the tSNP. (A) LDLc GxE effect with rs72895597 (18:34232657) as the tSNP. (B) SBP GxE effect with rs2303510 (18:34324091) as the tSNP.

Fig 5. FHOD3 locus. (A) Top wiggle tracks show ATAC-seq signal in multiple cell types, followed by ChromHMM chromatin state tracks. Beneath are FHOD3 GWAS loci and the SNPs from this study (reQTL and tSNP). The bottom track shows the FUSION FHOD3 RNA-seq signal. (B) ATAC-seq signal highlights potential regulatory regions with the skeletal muscle stretch enhancer. (C) Effects of SNPs overlapping ATAC-seq peaks in the reQTL haplotype on in silico predicted TF binding.