Transcript variants of CELF2 gene as unique prognostic indicators in breast cancer
Breast cancer (BC) is the most common malignant tumor in women around the world. Aside from finding a cure for this disease, it is also critical to identify prognostic biomarkers that can help clinicians intervene with the appropriate treatment and prevent BC progression. Current biomarker identification methods rely primarily on multi-gene prognostic signature models. However, due to the tumors' high heterogeneity, the accuracy of these multi-gene signatures is questionable. As a result, our main objective was to conduct a comprehensive analysis to identify a reliable prognostic biomarker in BC. Previously, a group of eight carnitine metabolites and SAH were linked to a poor prognosis in BC (Dr. Waqas Akbar, Unpublished Data). We discovered the genes associated with these metabolites using correlation analysis, and then we identified CELF2 as a good prognostic biomarker in BC. We validated CELF2's prognostic role in RNA-Seq and Microarray datasets in-silico. We demonstrate that the CELF2 1554569_a_at probeset is more consistent in its association with a favorable prognosis direction than the 202157_s_at probeset. When compared to the other probeset, CELF2 – 202157_s_at is expressed at higher levels and in a broader range of tissues. We were unable to find a clear significant association between CELF2 expression and prognosis during the in-vitro immunohistochemistry validation experiments. We hypothesized that this could be because our polyclonal anti-CELF2 antibody also recognized the less consistent 202157_s_at CELF2 probeset. We discovered that there are probeset-specific CELF2 transcript variants that are associated with different prognosis while testing this hypothesis. We created a Risk Score model by combining the expression levels of good and less-favorable CELF2 prognostic transcripts to improve prognosis prediction accuracy. The model successfully stratified the patients and predicted a higher overall survival in the Low-Risk group versus the High-Risk group. Overall, our findings suggest that each unique transcript variant of a gene can be associated with different prognosis directions. Therefore, we propose that studying prognostic associations at a gene transcript level could be a rich resource for the development of more robust biomarkers and therapeutics in cancer in the future.