Abstract
Background and objective: The identification of predictive markers for response of breast cancers to therapy is an intensive area of research. We hypothesized that using a physiologically relevant model would identify a signature composed of key, biologically relevant genes and therefore likely to predict therapeutic efficacy across independent datasets. Candidate predictive genes were selected using a well-defined three dimensional (3D) cell culture model of non-malignant human cells that can reacquire the ability to form acini-like structures presenting a lumen, basal polarity and cell cycle arrest in laminin-rich extracellular matrix 3D cultures. These acinar structures replicate many of the characteristics of luminal cell differentiation in the mammary gland. The gene expression profiles of acini formation were determined as a function of time in culture. Gene expression that was modulated during growth arrest and acini formation in 3D cultures (3D-signature) was compared and validated against gene expression measured in cancer and then correlated to patient outcome. Methods and results: We investigated the predictive value of the 3D-signature for response to therapy. We used a published microarray dataset of 24 breast tumors treated with Docetaxel with annotated clinical information (Chang JC et al. Lancet, 2003 362:340-1). The microarray dataset used tissue samples taken before treatment and patients were subsequently found to be resistant to or sensitive to docetaxel treatment. The predictive value of therapeutic sensitivity was analyzed using the 3D-signature. Probesets for 17 of 22 3D-signature genes were present on the HG-U95AV2 microarrays. 2D hierarchical clustering was used to separate patients into 2 groups. The 3D signature accurately predicted drug sensitivity (p=0.0014 Fischer\#8217;s exact test). Conclusions:Validation of the predictive power of the 3D-signature could lead the development of a clinical test for docetaxel sensitivity. These studies together are a proof of principle of the application of physiologically relevant models such as 3D-cultures to predict clinical response to therapy.
Citation Information: In: Proc Am Assoc Cancer Res; 2009 Apr 18-22; Denver, CO. Philadelphia (PA): AACR; 2009. Abstract nr 751.