The long range regulation of breast cancer associated genes

Abstract

Breast cancer is the most common non-cutaneous cancer of women and a major cause of mortality [1]. Genetic factors are the single biggest risk factor and 75% of the risk derives from common but low penetrance single nucleotide polymorphisms (SNPs) [2]. These are found using genome-wide association studies (GWAS) and the majority localize to non-coding regions of the genome [3]. A 2009 GWAS identified an estrogen receptor (ERa) positive breast cancer susceptibility locus in the 11q13 gene desert [4]. Previous fine mapping of this locus by our group revealed three independent risk signals containing five candidate causal single nucleotide polymorphisms (SNPs). Four of these clustered in an enhancer element (called PRE1) and one in a silencer element (PRE2). Interactions were demonstrated between these elements and the nearby cyclin D1 gene (CCND1), one of the most commonly amplified genes in breast tumours and a known oncogene [5]. Enhancers commonly regulate multiple genes however, suggesting that PRE1 or PRE2 may have other targets than CCND1 [6]. Such targets may be nearby protein coding genes or previously unrecognised noncoding transcripts within the gene desert. The identification of CCND1 as an interacting partner of PRE1 and PRE2 was done using a candidate gene approach with 3C (chromosome conformation capture) targeted confirmation of interaction. To detect other potential targets of PRE1 and PRE2 at the 11q13 locus using an agnostic approach, the 4Cseq and 5C techniques were used. These revealed a number of local interaction regions covering six gene promoters. A novel strategy using knockdown of enhancer RNA transcribed from PRE1 then identified the IGHMBP2 and CPT1A genes as likely additional targets of PRE1. Genome editing with transcription activator-like effector nucleases (TALENS) was also used, creating isogenic cell lines to clarify the effects of the SNPs in their native genomic context. Further functional work is required, however the range of techniques employed has substantially expanded our understanding of the 11q13 breast cancer risk locus and forms a template for future investigations of GWAS risk loci. To identify noncoding RNAs expressed at the 11q13 locus that may be interacting with PRE1 or PRE2 required the use of RNA Capture-seq. RNA Capture-seq involves a targeted enrichment step that greatly increases the sequencing depth at regions of interest and can reveal lowly expressed transcripts that may have not been found by traditional RNA-seq [7]. This identified one novel long non-coding RNA expressed from the (+) strand that was named CUPID1 and a second arising from the same locus on the (n) strand named CUPID2. They were located in the nucleus, oestrogen induced and both expressed relatively highly in ERa positive breast cancer cell lines but not in ERa Preface Page ii negative cell lines or normal breast tissue. 3C revealed that PRE1 exhibits allele specific interactions with the lncRNA promoter and functional assays demonstrated that PRE1 markedly increased promoter activity. The increased activity was partially abrogated following incorporation of the risk SNPs. SiRNA mediated knockdown of both lncRNAs impaired the normal DNA damage response of breast cancer cells and the ChIRPseq technique confirmed that CUPID1 was preferentially bound to promoters of genes involved in DNA repair. Two oestrogen responsive lncRNAs that interact with the 11q13 breast cancer susceptibility locus have thus been identified that may mediate the associated risk of oestrogen responsive breast cancer. The 11q13 locus is amplified in around 20% of breast cancers and is thought to contain a number of driver genes including CCND1 [8]. Given that CUPID1 and CUPID2 are widely over-expressed in ERa positive breast cancer cell lines, it was hypothesized that they may also have a role in driving tumour growth. Publically available RNAseq data showed CUPID2 to be highly expressed in breast and renal cancer but not in normal tissue. Stable cell lines over-expressing the lncRNAs were then generated and subsequent oncogenic assays revealed that CUPID2 increased cellular proliferation and the efficiency of colony formation in breast cancer cells. A murine xenograft model confirmed these findings in vivo by demonstrating a marked increase in tumour size for the mice injected with cells over-expressing CUPID2. There is thus evidence that CUPID2 behaves as an oncogene and may be an additional driver of the 11q13 amplicon in ERa positive breast cancer. In summary, CPT1A and IGHMBP2 were identified as potential mediators of risk at the 11q13 breast cancer susceptibility locus. Further investigations of the locus then identified two novel noncoding transcripts called CUPID1 and CUPID2 that may also mediate the effects of the causal SNPs through downregulation of cellular DNA damage repair pathways. Finally, CUPID2 was shown to function as a putative oncogene in ERa positive breast cancer. These findings add to our knowledge of breast cancer risk and progression and may ultimately lead to more effective treatments and prevention programs to better manage this disease.