Genome-Wide Evaluation of Epistasis with APOL1 Risk Variants in Sickle Cell Disease Nephropathy

BACKGROUND: Two functional variants (G1 and G2) in the APOL1 gene have been implicated as contributing to the high rate of chronic kidney disease in African Americans (Genovese et al., 2010; Kopp et al., 2010). We subsequently demonstrated that the APOL1 G1 and G2 variants, as well as non-coding genetic variation in the neighboring gene MYH9, were associated with proteinuria in a cohort of sickle cell disease (SCD) patients (Ashley-Koch et al., 2011). When controlling for the effect of the functional variants in APOL1, the effect of MYH9 on proteinuria remained, suggesting that both genes were independent risk factors in SCD nephropathy. A statistical interaction between the two genes predicting glomerular filtration rate (GFR) was also observed, suggesting a possible role of epistasis. In addition, we have demonstrated that myh9 and apol1 functionally interact in an in vivo zebrafish model, particularly in the context of anemic stress and more specifically that the APOL1 G2 risk allele modulates myh9 expression in the zebrafish (Anderson et al., 2015). In an effort to identify additional genes that may interact with APOL1 to influence risk for SCD nephropathy, we have performed a genome-wide evaluation of epistasis with APOL1 G1 and G2 variants.

METHODS: Adults with SCD were recruited from sickle cell centers at Duke University, University of North Carolina at Chapel Hill, Emory University, and East Carolina University. Medical history, laboratory values and DNA for genotyping were collected. GFR was estimated using the definition from the 'Modification of Diet in Renal Disease' study (Levey et al., 1999). Genotyping was performed using the Illumina Human610-Quad BeadChip (Illumina, San Diego, CA) and missing genotypes were imputed using IMPUTE2 (Howie et al., 2009) with a global reference panel from the 1000 Genomes project. Principal component (PC) analysis was conducted using EIGENSOFT (Patterson et al., 2006) to obtain measures of population substructure. APOL1 coding variants G1 (rs73885319) and G2 (rs71785313) were genotyped using Taqman assays (Applied Biosystems, Foster City, CA). To test for interactions between each SNP and the G1 or G2 alleles, linear regression was used to predict GFR and logistic regression for proteinuria. All models included main and interaction effects of SNPs (coded additively), as well as two genome-wide PCs to control for population stratification. False discovery rate (FDR) q-values were generated using PROC MULTTEST in SAS (SAS Systems, Cary, NC).

RESULTS: A region on chromosome 6 significantly interacted with the APOL1 G2 allele to predict GFR (FDR q's < 0.036); the most significant SNP was rs6916880 (p=6.4x10^-8). We also identified SNPs in a region on chromosome 4 which significantly interacted with APOL1 G2 to predict GFR (FDR q's=0.075). Finally, rs10790180 in CEP164 showed a significant interaction with the APOL1 G2 allele to predict GFR (FDR q=0.0932). No significant interactions were observed with the APOL1 G1 allele to predict GFR, or with either APOL1 risk allele to predict proteinuria (FDR q's > 0.9).

DISCUSSION: We have identified several genome-wide significant gene-gene interactions involving the APOL1 G2 allele influencing GFR in an adult SCD cohort. Although the most significant interactions were observed with intergenic regions on chromosomes 6 and 4, both regions are characterized by histone marks in ENSEMBL, indicating that they may have important regulatory function. We also observed an interaction between a SNP in CEP164 and the APOL1 G2 allele. CEP164 is a basal body protein required for assembly of the primary cilium (Graser et al., 2007), and mutations in CEP164 have been associated with nephronophthisis-related ciliopathies (Chaki et al., 2012). Moreover, suppression of cep164 in the developing zebrafish resulted in a ciliopathy phenotype that included pronephric tubule cysts (Chaki et al., 2012). We did not replicate previous reports of epistasis with APOL1 in ESRD (Divers et al., 2014; Bostrom et al., 2012), but our work has shown that SCD nephropathy may be a unique form of kidney disease with distinct genetic drivers, due to the context of anemic stress (Anderson et al., 2015; Ashley-Koch et al., 2011).