Abstract
Transforming growth factor beta (TGFβ) plays a key role in regulating epithelial-to-mesenchymal transition (EMT). A gene expression signature (TGFβ-EMT) associated with TGFβ-induced EMT activities was developed using human Non-Small Cell Lung Carcinoma (NSCLC) cells treated with TGFβ-1 and subjected to Affymetrix microarray analysis. The final 105-probeset TGFβ-EMT signature covers 77 genes, and a NanoString assay utilized a subset of 60 of these genes (TGFβ-EMTN signature). We found that the TGFβ-EMT and TGFβ-EMTN gene signatures predicted overall survival (OS) and metastasis-free survival (MFS). The TGFβ-EMT signature was validated as prognostic of 5-year MFS in 3 cohorts: a 133 NSCLC tumor dataset (P = 0.0002), a NanoString assays of RNA isolated from formalin-fixed paraffin-embedded samples from these same tumors (P = 0.0015), and a previously published NSCLC MFS dataset (P = 0.0015). The separation between high and low metastasis signature scores was higher at 3 years (ΔMFS TGFβ-EMT = -28.6%; ΔMFS TGFβ-EMTN = -25.2%) than at 5 years (ΔMFS TGFβ-EMT = -18.6%; ΔMFS TGFβ-EMTN = -11.8%). In addition, the TGFβ-EMT signature correlated with whether the cancer had already metastasized or not at time of surgery in a colon cancer cohort. The results show that the TGFβ-EMT signature successfully discriminated lung cancer cell lines capable of undergoing EMT in response to TGFβ-1 and predicts MFS in lung adenocarcinomas. Thus, the TGFβ-EMT signature has the potential to be developed as a clinically relevant predictive biomarker, for example to identify those patients with resected early stage lung cancer who may benefit from adjuvant therapy.