Targeted resequencing of the microRNAome and 3'UTRome reveals functional germline DNA variants with altered prevalence in epithelial ovarian cancer

Abstract

Ovarian cancer is a major cause of cancer deaths, yet there have been few known genetic risk factors identified, the best known of which are disruptions in protein coding sequences (BRCA1 and 2). Recent findings indicate that there are powerful genetic markers of cancer risk outside of these regions, in the noncoding mRNA control regions. To identify additional cancer-associated, functional non-protein-coding sequence germline variants associated with ovarian cancer risk, we captured DNA regions corresponding to all validated human microRNAs and the 3' untranslated regions (UTRs) of ~6000 cancer-associated genes from 31 ovarian cancer patients. Multiple single-nucleotide polymorphisms in the 3'UTR of the vascular endothelial growth factor receptor/FLT1, E2F2 and PCM1 oncogenes were highly enriched in ovarian cancer patients compared with the 1000 Genome Project. Sequenom validation in a case-control study (267 cases and 89 controls) confirmed a novel variant in the PCM1 3'UTR is significantly associated with ovarian cancer (P=0.0086). This work identifies a potential new ovarian cancer locus and further confirms that cancer resequencing efforts should not ignore the study of noncoding regions of cancer patients.

Figure 1

(a) Workflow for our integrated study of miRNAs and targeted genes. Targeted sequencing of miRNAs and 3′UTRs was performed on 31 ovarian cancer patients. These target regions were also extracted from the whole genome sequencing of the 1000 Genome project was used as controls to detect known and novel variants. To select functional and significant variants, we consider the allele frequency difference between cases and controls, differential expressed genes from gene expression profiling and miRNA-3′UTR-predicted target pairs. Finally, functional variants were validated by Sequenom in a larger corhort. (b) Allele frequency of known SNPs, patient samples versus the 1 KG database (European and all populations).

Figure 2

The SNPs and indels identified in our EOC cases. All novel SNPs and indels identified in EOC cases and only those known variants that were found to be enriched in EOC cases compared with reference European population controls from the 1000 Genomes Project were mapped to their relative positions in the 3′UTR of (a) E2F2, (b) FLT1 (three separate 3′UTR transcripts) and (c) PCM1 genes, respectively. The solid line represents a SNP, dashed line represents an indel, red indicates novel SNP/indel, blue is known SNP/indel and the height of the bar represents the fold enrichment of known SNPs/indels in our EOC cases vs reference European population controls from the 1000 Genomes Project or the allele frequency of novel SNPs in cases. The bar with a circular end represents SNPs within high confidence miRNA complementary sites, which are shown in the black bar. The asterix indicates the novel variant in the PCM1 3′UTR found to be significantly enriched in EOC cases.

Figure 3

(a) The ATTT insertion mediates differential regulation of the PCM1 3′UTR. A luciferase reporter construct containing the PCM1 3′UTR (PCM1) is significantly repressed in CaOV3 and MCF-7 cells, relative to empty vector (psiCHECK2). This repression is maintained in a luciferase reporter containing the PCM1 3′UTR variant containing the ATTT insertion (PCM1 ATTT variant). (b) Luciferase reporters containing ~ 600 nt of the PCM1 3′UTR comprising the region flanking the position of the ATTT insertion were generated, either lacking the ATTT insertion (PCM1 SF1) or containing the insertion (PCM1 SF1 ATTT variant). PCM1 SF1 is significantly upregulated relative to empty vector (psiCHECK2) in CaOV3 and MCF-7 cells. This repression is partially lost in CaOV3 cells and completely lost in MCF-7 cells in PCM1 SF1 ATTT variant. Plotted: mean ± s.d.; n = 3; **P<0.01, Student’s t-test.