Association of gene variants in the TGF-beta signalling pathways with invasive breast cancer risk

Many studies of normal cells in vitro, of transgenic mouse models and of somatic mutations in human cancers have provided evidence that the cytokine transforming growth factor beta (TGFβ) acts as a suppressor of primary tumour initiation. However, studies of transformed cells in vitro and of mouse models have also implicated TGFβ as a promoter of the later stages of tumour development. A hypothesis that has been proposed to account for this dual action is that TGFβ acts as a tumour suppressor through the ubiquitous ALK5 receptor signalling via SMAD2 and SMAD3 to inhibit proliferation of primary tumour cells, but acts subsequently through the endothelial-specific ALK1 receptor via SMAD1 and SMAD5 to promote angiogenesis, which is required for tumour progression [1].

In a recent meta-analysis we showed that a single nucleotide polymorphism (SNP) generating a leucine to proline substitution in the signal peptide of the TGFB1 protein is associated with an increase in the risk of invasive breast cancer (OR per additional proline allele = 1.08 (95% CI = 1.04 to 1.11), P _trend = 2.8 × 10^-5) [2]. We have also reported that this SNP increases the amount of TGFB1 protein secreted in vitro by threefold [3]. These data suggest that higher levels of TGFB1 may promote the invasive breast tumour phenotype. To determine the effect of host TGFB1 levels in an in vivo model, Tgfb1 ^+/- and Tgfb1^+/+ mice have been compared in which the mice carry one or two TGFB1 alleles with the ancestral SNP form encoding proline in the signal peptide. These studies have revealed major effects of TGFB1 in controlling the site of metastatic seeding and the number of metastases that develop.

The TGFB1 SNP association with breast cancer suggested that other genes in the TGFβ signalling pathways might be associated with altered risk. We have conducted association studies with SNPs in 16 further genes encoding proteins directly implicated in TGFβ signalling. LTBP1, LTBP2, LTBP4, TGFB1, TGFB2, TGFB3, ALK1, ALK5, TGFBR2, Endoglin, SMAD1, SMAD2, SMAD3, SMAD4, SMAD5, SMAD6 and SMAD7 were analysed. A comprehensive SNP tagging approach was used to select variants for genotyping in a staged study design using up to 6,900 cases and 6,900 controls, all collected from the East Anglia region of the UK (>98% of northwestern European ancestry). From 1,254 common SNPs (minor allele frequency >0.05) in these genes identified from the International HapMap project data, we defined and genotyped 354 tagging SNPs in the East Anglia cases and controls. Statistically significant associations were followed up by genotyping in a Polish set of 2,215 cases and 2,374 controls. Meta-analysis of these results identified associations with cancer susceptibility of a variant in ALK5 (OR per additional rare G allele = 0.88 (95% CI = 0.81 to 0.95), P _trend = 0.001) and in TGFBR2 (OR per additional rare G allele = 0.96 (95% CI = 0.92 to 1.00), P _trend = 0.039). Data from two genome-wide studies have been examined to search further for associations in these genes. The haplotype risks and interactions between two or more loci have been investigated and survival analyses have been conducted.

From this comprehensive study we have identified tagging SNPs in two TGFβ receptor genes that are significantly associated with risk of invasive breast cancer. Identification of the causative SNPs within the LD blocks tagged by the two SNPs and their effects on signalling via the ALK1 and ALK5 pathways remain to be determined.