Abstract Purpose: A common goal in pharmaceutical research is to reveal correlations between signaling pathway activity in intact tissue and clinical outcomes. Correlations support target validation, trial design, patient selection, response assessment, and, if trials are successful, the diagnostic component of theranostics. However, the predictive power of measurements of protein expression depends on the precision and accuracy of tissue analysis tools. For example, many techniques deployed today, such as those based on immunohistochemistry and micro-array detection, or analysis of sample lysates, provide data that are averages from volumes of tissue, including many cells not of interest. These methods blur out key proteomic information that resides at the cellular level, relating to the signaling states of individual cells. The purpose of this study is to demonstrate an effective, practical and reliable platform for cytometric analysis of signaling pathway proteins in intact tissue sections. The goal is to support preclinical and clinical studies through the integration of multiplexed immune-fluorescence labeling strategies with automated high-throughput image acquisition and analysis. Experimental Procedures: The tissue cytometry platform integrates: a) multiplexed immunofluorescence staining protocols, b) an automated slide analysis system (Vectra™) that utilizes multispectral imaging to isolate marker signals from one another and from autofluorescence, and c) a new pattern-recognition-based image analysis package (inForm™) for automatically segmenting images and extracting data from cells-of-interest. For this study, a staining panel was developed targeting phosphoeptitopes of AKT, ERK, and S6, using antibodies of three different isotypes, with secondaries conjugated to Alexa fluorophores (A488, A555, and A647). DAPI was used as a counterstain. Staining protocols were optimized using tissue microarrays (TMA) generated from breast tissue samples prepared using methods designed to preserve phosphoepitopes. Spectral unmixing libraries were generated with single-stained control samples. Image analysis algorithms were trained to segment tissue regions (e.g., malignant and normal epithelia, stroma, necrosis, etc.) and then cells and cell compartments within tumor regions, to extract per-cell data for cytological analysis. Validation of the platform's ability to quantitate changes of phosphoepitope expression was performed with cell blocks of cell lines treated with the relevant inhibitors. Results: Pilot studies of the TMA and cell lines reveal robust and specific signal levels, localized to tissue and cellular structures appropriate for the target molecules, despite being applied in four-plex. Pattern recognition-based, automated image analysis algorithms reliably detected tumor cells and segmented associated cellular compartments, after having been trained on less than 10% of pilot study images. Tissue segmentation accuracy was estimated at greater than 90%, based on visual review by pathologists. Conclusions: Performance of the platform for automated multiplexed tissue cytometry analyses supports its application to routine clinical studies. Presently, protocols and workflow procedures are being implemented for real-time evaluation of patients in clinical trials. Citation Information: Mol Cancer Ther 2009;8(12 Suppl):A48.