Correlation of the Gene Expression Profile of Primary Breast Carcinoma with Clinical Outcome
SM Kennedy, D Tracey, P Doolan, JP Mehta, M Clynes, L O'Driscoll. St Vincents University Hospital, Dublin, Ireland; Dublin City University, Dublin, Ireland
Background: Optimal treatments for breast cancer rely on classic clinicopathologic parameters, which are not specific enough to identify all subgroups at high risk of treatment failure. DNA microarray technology allows analysis of the RNA expression of several thousand genes simultaneously. Discovery of new clinically useful prognostic subgroups, currently not found by conventional parameters, is made possible by use of gene expression profiling. Our aim was to investigate the global molecular profile of invasive primary breast cancer with 15 year follow up to discover prognostically important genes associated with survival and potential new biomarkers for therapeutic targets.
Design: RNA (100ng) extracted from 104 tumors and 17 normal breast samples was analyzed using Affymetrix whole genome microarrays. Following normalization using dCHIP, statistical filters were used to identify significant differentially-expressed genes between tumor and normal; lymph node positive vs node negative tumors; patients who relapsed or died within 5 years vs survivors. Principal component analysis was performed on resulting genelists; literature mining software was used to identify pathways affected; individual mRNAs were extensively investigated for associations with clinicopathologic characteristics. qRT-PCR was used to validate results.
Results: 7448 transcripts were differentially expressed (P=0.0068) between tumor and normal. 998 transcripts (P=0.0009) and 1369 (P=0.0013) between tumors associated with relapse or death within 5 years vs those that did not. 12 mRNA associated with unfavorable factors and 24 with favorable prognostic import were identified. Patent office searches showed that this panel of 36 transcripts had not previously been associated with cancer and many do not appear to be represented on earlier microarrays where the information is publicly available. Amongst these transcripts, multivariate Cox regression analysis of validating qRT-PCR results showed PRAME to be independently unfavorable in terms of overall survival (OS) (P=0.02) and relapse free survival (RFS) (P=0.026); SNIP to be independently unfavorable for OS (P=0.005); and TMEM25 to be independent in terms of both OS (P=0.001) and RFS (P=0.011).
Conclusions: We have identified a panel of 36 gene transcripts independently associated with favorable or unfavorable outcome in primary breast carcinomas by multivariate analysis. These results identify new prognostic subgroups that may provide new potential therapeutic targets.
Monday, March 9, 2009 2:30 PM
Platform Session: Section B, Monday Afternoon