SNPS in putative regulatory loci controlling gene expression in cancer

Given the increasing clinical importance of microarray expression classification of breast tumours and the different biology it may reveal [1], identifying an associated SNP profile may be of considerable value for pharmacogenetics, early diagnostics and cancer prevention. Studying the promoter composition of the genes that strongly predict the patient subgroups, we observed clear separation of the gene clusters based solely on their promoter composition, making feasible the hypothesis that SNPs in the regulatory regions of genes that create or abrogate transcription binding sites have the potential to influence the expression profiles. Morley and colleagues [2] reported linkage analysis of expression levels of 3554 genes and 2500 SNPs in 14 CEPH families (retrieved online [3]), and found significant evidence for the existence of regulation hot spots, suggesting both cis and trans regulatory effects. We report similar observations from a study with a different design, performing actual genotyping of 49 unrelated breast cancer patients, whose tumours have previously been analysed by genome-wide expression microarrays leading to a robust tumour classification with strong prognostic impact [4]. These patients were a part of a pharmacogenetic study of 193 patients who had received radiation therapy or chemotherapy. A high-throughput solid-phase, array-based method using primer extension chemistry has been used to perform the genotyping (GenomeLab™ SNPstream genotyping system; Beckman Coulter, Fullerton, CA, USA). A total of 583 SNPs in 203 selected genes (1–19 SNPs/gene) were genotyped and tumour genome-wide expression was studied in 49 patients. Association in both cis and trans was detected for SNPs in 42 genes. SNP–expression associations with the top 0.25% best P values (9.81 × 10 ^-6 <P < 0.001) revealed regulatory SNPs in 115 genes in trans. The subsets of transcripts that were observed to have significantly many associations in common with a set of SNPs were further analysed using the gene ontology (GO) annotations. The GO terms of the unselected mRNA transcripts found associated to the SNPs in the selected candidate genes were often similar, suggesting that the observed associations are within the same functional pathway. Taken together these data suggest that the observed SNP–expression associations do exist and are observable even in a small set of unrelated individuals. A given expression profile of the tumour may be potentially associated and predicted by the genotype of the patient.