Expression QTL analysis of a gene expression signature which predicts advanced non-alcoholic fatty liver disease

Abstract

BACKGROUND: Non-alcoholic fatty liver disease (NAFLD) is associated with obesity and the metabolic syndrome and affects over one third of American adults. A subset of NAFLD patients develop advanced disease associated with liver injury, scar tissue (fibrosis), and cirrhosis, increasing risk for liver-related mortality, including primary liver cancer. Several genes have been associated with NAFLD, but determination of their function in the disease process has been difficult. We hypothesized that expression quantitative trait loci (eQTL) analysis would improve the identification of causally related genetic variants by integrating single nucleotide polymorphisms (SNPs) with gene expression. METHODS: Patients with biopsy-proven NAFLD were selected based on fibrosis stage (mild NAFLD: stage F0-F1 vs. advanced NAFLD: stage F3-F4). Genomic data, including genotype (Cardiometabolic SNP array, Illumina), methylation (Human Methylation450 BeadChip, Illumina) and expression (Human U133 Plus 2.0 GeneChip, Affymetrix) were available for 69 patients. We identified a novel 64-gene expression signature differentiating advanced and mild NAFLD, confirmed in a validation data set. Using ANOVA, cis-eQTL analysis was conducted with transcript levels from this signature and SNPs within 1 MB of the expression probe midpoint, while controlling for gender, age at biopsy, diabetes and BMI. SNPs falling directly within expression probes were excluded. Empirical pvalues were generated from permutation testing. RESULTS: 28 SNPs significantly predicted transcript levels (p<0.001), includ ing those for CD24P2 (3 independent SNPs), COL1A2 (1 SNP), CYBRD1 (1 SNP) and ID4 (23 SNPs in 3 haplotype blocks). CD24P2 is a pseudogene which does not appear to produce a functional protein. COL1A2 encodes the pro-alpha2 chain of type 1 collagen, a fibrillar collagen found in most connective tissues. CYBRD1 is a ferric-chelate reductase present at the brush border of duodenal enterocytes. All 23 SNPs predicting ID4 expression lie distally within or near the CDKAL1 gene, previously associated with type 2 diabetes. Both ID4 and CKDAL1 are involved in regulation of cell cycle transitions. Interestingly, COL1A2, CDKAL1 and ID4 also exhibit reduced methylation and increased gene expression in advanced versus mild NAFLD patients. CONCLUSION: The identified eQTLs provide important insight into pathologic processes contributing to advanced NAFLD and are significantly enriched for tissue repair, insulin resistance and cell fate decisions.