Research

To date our work has been following up the results of the genome-wide association studies in breast cancer that we and our colleagues initiated. The chance of an individual developing breast cancer is roughly two-fold greater if that individual has a close relative with breast cancer. Twin studies indicate that this risk is largely genetic. The genes that confer this risk have been sought either by genetic linkage mapping in multiple-case families, or by genome-wide association studies (GWAS). The former have identified rare but higher risk alleles such as those of BRCA1 and BRCA2, while GWAS have identified common variants that each carry only a small risk of cancer. BRCA1 and 2 explain about 15-20% of the estimated total genetic risk of breast cancer, and loci identified through GWAS a further 10%.

One question is how to find the genes that account for the 'missing' 70% or so of heritability. Larger and more powerful GWAS will find some, while genome resequencing will identify an unknown contribution from rare genetic variants. Pending these studies, we are exploring other approaches in breast and in lung cancer.

In breast cancer, Ana-Teresa Maia is making a catalogue of genes in which there are common variants that cause different levels of expression of the two alleles in a heterozygous individual. Since many of the genetic variants so far identified in GWAS studies in general appear to have their effects through altered gene regulation, the subset of genes that show differential allelic expression (DAE) should be enriched for genes involved in susceptibility. If correct, this information would allow prioritisation of genes for further study from the very large numbers of loci that show borderline levels of significance in existing GWAS. We are now testing this hypothesis.

The common genetic variants identified through GWAS individually have very small effects. They can be thought of as causing small perturbations of regulatory gene networks within the cell - the combined effect of many variants produces a greater perturbation that leads to disease. A regulatory variant in the fibroblast growth factor receptor 2 (FGFR2) gene is the common variant with the greatest effect on breast cancer susceptibility. We are using a systems biology approach to understand the function of this predisposing gene. To this end we have treated the well-studied oestrogen dependent cell line MCF-7 with oestrogen and FGF10, an activator of the FGFR signalling pathway. Microarray analysis has identified downstream target genes that are differentially regulated (Figure 1). We will try to identify the network of genes that are coordinately up or down regulated after FGFR and oestrogen receptor signalling in a time-dependent manner. Highly connected genes, forming a hub in the network, are likely to be master regulators and might themselves act as predisposing genes. These hubs may also be promising therapeutic targets that could be modulated in order to correct a de-regulated network.

Figure 1
Differentially expressed genes in MCF-7 cells after activation of the FGFR and oestrogen signalling pathways. The following comparisons were carried out - red circle: untreated versus oestrogen (E2) treated; blue circle: oestrogen treated versus oestrogen plus FGF10 (E2.FGF10); green circle: oestrogen treated versus FGF10 and FGFR inhibitor (E2.FGF10.PD).

We have also continued our work to identify the functional variants within the risk loci identified by GWAS for breast cancer. We have focussed this work on the TNRC9 gene (Udler et al., Hum Mol Genet 2010; 19: 2507) and 8q24 region (Meyer et al., submitted). The latter is a gene-poor region that contains multiple susceptibility loci for prostate, breast, colon and bladder cancers. By studying chromatin conformation and protein-DNA interactions both in vitro and in vivo, we have identified a functional variant for one of the 8q24 prostate cancer predisposition loci. The protective allele is able to bind YY1 and mediates transcriptional repression. Chromatin conformation capture (3C) suggests that this region is able to interact with both the MYC and PVT1 oncogenes, located more than 600kb downstream of the risk interval. Furthermore, we find that increased PVT1 expression correlates with the presence of the risk allele in normal prostate samples.

In lung cancer, we propose to explore a different hypothesis to identify the 'missing genetic heritability' which in this cancer is possibly even greater than in breast cancer. Cigarette smoking is the major cause of lung cancer, but only 15% of heavy smokers develop cancer. We hypothesise that this is not chance, but that the susceptible 15% differ, because of genetic variation, in their response to cigarette smoke injury to the airway cells. If so, analysis of relevant phenotypes in those normal cells (or possibly in a surrogate such as blood lymphocytes) may provide an integrated readout of the effects of genetics and smoke exposure, that determines the likelihood that cancer will develop. We will start by analysing: 

1. Gene expression patterns in airway epithelium of smokers who do, and do not, have lung cancer. We will examine both nasal and bronchial epithelium to see if nasal epithelium can provide an accessible surrogate tissue; and
2. H2AX phosphorylation as a measure of DNA damage response in irradiated cultured lymphocytes. If the results are promising, we have access to a large prospective study for confirmation.