A three-gene expression signature model to predict neo-chemoradiotherapy response of esophageal squamous cell carcinomas.
e15135 Background: Preoperative chemoradiotherapy (CRT) followed by surgery has been proved to improve survival in comparison with surgery alone. However, the outcomes of CRT are heterogeneous, and no clinical or pathological method could prediction CRT response. In this study, we aim to identify mRNA markers for ESCC CRT-response prediction. Methods: Gene expression analyses were performed on pretreatment cancer biopsies from 28 ESCCs who received neoadjuvant CRT and surgery. Surgical specimens were assessed for the pathological response to CRT. The identified differentially expressed genes were validated by real-time quantitative polymerase chain reaction (qPCR), based on which a classifying model was built up by Fisher’s linear discriminant analysis. The predictive power of this model was further assessed in another set of 32 ESCCs. Results: The profiling of the 28 ESCCs identified 10 differentially expressed genes with more than 2-fold changes between pathological complete responsers (pCRs) and less than pCRs (<pCRs), among which 6 genes (LIMCH1, SDPR, Clorf226, SLC9A9, GSTM3, and IGSF10) were down-regulated and 4 genes (MMP9, MMP1, MMP12 and OASL) up-regulated in pCRs versus <pCRs. A prediction model based on qPCR values of 3 genes was built up, Y=-10.388 - 0.343 × MMP1 + 0.642 × LIMCH1 + 0.921 × Clorf226 with a cut-off value of 0.607. It provided a predictive accuracy of 85.7% with leave-one-out cross-validation. Further, the predictive power of this model was validated in another set of 32 ESCCs, among which a predictive accuracy of 81.3% was achieved. Conclusions: The combination of three genes by qPCR identified by microarrays in our study provides possibility for ESCC CRT prediction, which will facilitate individualization of ESCC treatment. Further perspective validation in larger independent cohorts is warranted to fully assess the predictive power of this prediction model.