Tumor-associated stromal gene expression signatures predict therapeutic response to erlotinib/bevacizumab in non-small cell lung cancer (NSCLC)

Background: Tumor-associated stromal cells have an important role in tumor growth, disease progression and drug resistance. Gene expression analysis (GEA) of various cancers identified molecular subtypes and predictive models for clinically relevant outcomes. This information however, is often derived from heterogeneous clinical samples (tumor and stromal cells). We aimed to identify gene expression signatures (GES) of stromal cells from microarray data of NSCLC patients treated with erlotinib/bevacizumab, exploiting the transcriptome characteristics of these distinct cell types. Methods: We performed gene enrichment analyses for identifying which tumor compartment is most likely to predict patients' outcome using Affymetrix Exon-Arrays of 42 biopsy samples (from SAKK1905 trial) using time-to-progression (TTP) under therapy as endpoint. Results: Our analysis revealed that genes predicting time-to-progression under therapy are mostly derived from endothelial compartment (mean-rank=7006, P =0.04). No compartmental predictive effects were observed for the other endpoints. These findings support recent reports demonstrating that tumor-specific endothelial cells participate in the resistant phenotype of tumors, providing a chemotherapeutics/tumor cells protective barrier. Conclusions: Our method revealed that GES derived from the endothelial stromal compartment has significant predictive value of time-to-progression for NSCLC under therapy. This underlines the importance of statistical methods directed at maximizing the information derived from heterogeneous tissue GEA, which could lead to the identification of clinically relevant biomarkers.