Background: Epidermal growth factor receptor (EGFr) is a transmembrane tyrosine kinase expressed on many different tumor types. Preclinical and clinical evidence suggests that blocking the EGFr signaling pathway can provide clinical benefit to patients whose tumors express EGFr. Panitumumab, a fully human antibody, binds to the EGFr with high affinity (5x10-11 M) preventing ligand-induced activation resulting in arrest of tumor cell proliferation and apoptosis in some cases 1,2. The objective of this study was to find a set of genes whose expression levels could be used to predict responsiveness to panitumumab monotherapy. Methods: Responsiveness to panitumumab was determined in breast, colon, lung and pancreatic xenograft models. Animals were treated twice per week with 20, 100, 200, and 500 μg/mouse per dose and response was determined as a 40% reduction of tumor volume (versus control). Responsive and non-responsive models from each tissue type were used for the analysis. Untreated xenograft samples were arrayed on the Affymetrix human U133A gene chip. After adjusting for tissue, combinations of genes were selected using a multivariate analysis that could predict responsiveness to panitumimab. The process was validated using a full leave-one-group out analysis. Results: Treatment of 300 mm3 established xenografts determined 8 responsive models, including NSCLC lines with EGFr kinase domain mutations, and 12 non-responsive models to the treatment panitumumab. An initial unsupervised cluster analysis demonstrated that the tissue type had greater influence on the clustering of genes than the responsiveness to panitumumab. A supervised multivariate classification technique was used to identify gene sets that could predict responsiveness to panitumumab independent of the known connection to the EGFr pathway. The gene set could predict treatment outcome in a leave-one-out validation. Conclusion: These data suggest panitumumab can inhibit the growth of breast, colon, lung and pancreatic tumor xenografts and that the tissue type has more influence on the clustering of the models than the responsiveness (or lack of) to panitumumab. Using a supervised analysis, gene sets, regardless of their known association to EGFr signaling, can be generated from microarray data that can predict response in xenograft models. This approach may aid in the selection of genes that could stratify patients that respond to panitumumab.
[Proc Amer Assoc Cancer Res, Volume 46, 2005]