Automated Assessment of ER and ki67 in Individual Tumor Cells in Intact Tissue Sections Using Multicolor IHC, Multispectral Imaging, and Pattern-Recognition-Based Image Analysis Software.

Abstract

Abstract Background. A promising area of research is the analysis of multiplexed nuclear proteins on a per-cell basis and the correlation of multi-protein-based cell phenotype statistics with clinical patterns. Many believe that capturing this information from individual tumor cells, rather than average values for proteins across tumor cell populations, as obtained in singly stained serial sections, or as average values for proteins across homogenized tumor samples, may hold key information about disease state, and thus offer valuable information for diagnosis, prognosis and therapy selection. Objective. To demonstrate that multicolor IHC staining and multispectral imaging can be used to quantify reliably multiple nuclear proteins on a per-cell basis, without significant inter-stain interference or cross-talk. Additionally, our objective is to offer an example of the additional information afforded by multiplexed per-cell IHC. ER and ki67 were chosen for this demonstration because of the recent finding that ki67 identifies a subset of ER-positive patients who could be sensitive to docetaxel treatment. (Penault-Llorca et. al. JCO, June 2009)Methods and Materials. Serial sections of a breast tissue micro-array (TMA) were stained singly for ER and for ki67, and a third section was stained for both ER and ki67. Stains used were DAB, Vector Red and hematoxylin as a counterstain. Each TMA was imaged with a CRi VectraTM, an automated multispectral slide analysis system. 4x scans were obtained to detect the TMA and 20x multispectral images acquired from each core. A pattern-recognition-based image analysis software, inFormTM, was then used to automatically detect tumor cells in each core, and 'unmix' the overlapping IHC signals using the spectral signatures of each stain. Per-cell multi-protein data was then exported for data analysis both with Excel and a flow-cytometry analysis package for cell-phenotype classification. Signals from singly stained TMAs were compared with signals that were unmixed from the double-stained TMA.Results. Initial results suggest that if staining is performed carefully, independent and reliable measurement of individual nuclear proteins on a per-cell basis in a multiplexed protocol is possible. Additionally, the flow cytometry analysis software can be used to reveal per-cell multi-parameter expression profiles; in particular percent double negativity, single positivity for each protein, and double positivity. This tissue-cytometry-level detail about protein expression state may offer additional metrics upon which to base correlations with clinical patterns.Discussion. The effectiveness demonstrated in this report supports the application of this powerful new tool to molecular epidemiological investigations, to help reveal correlations between outcomes and complex protein expression patterns in tissue sections. These correlations can inform target validation, trial design, patient selection, and response assessment, and, if trials are successful, can become part of the diagnostic component of theranostics. Citation Information: Cancer Res 2009;69(24 Suppl):Abstract nr 6006.