Abstract P6-04-11: Quantitative Differences among Spectral Counts Based on LC-MS/MS Spectra Accurately Cluster Basal, Her2-Overexpressing and Luminal A Breast Carcinomas

Abstract

Abstract Characterization of breast cancer subtypes at the protein-level is largely unexplored and is a powerful approach to identify novel biomarkers that may have diagnostic and prognostic utility. Shotgun proteomics analyzes mixtures at the peptide-level to generate MS/MS spectra that are then used to identify the peptides and proteins from which they were derived. We developed a method in our laboratory utilizing SDS-PAGE followed by in-gel digestion coupled with reversed-phase (RP) liquid chromatography-tandem mass spectrometry (RP-LC-MS/MS) for proteomic analysis of LCM-acquired cells. This method was then applied to a clinical sample set consisting of basal, Her2-overexpressing and luminal A frozen tumor tissues as determined by prior microarray studies. Three tumors per group were dissected in triplicate, for a total of 27 samples. Each dissection consisted of approximately 10,000 cells, corresponding to 3-5 μg protein. Following LCM, the thermoplastic membranes containing the captured cells were peeled from the cap, suspended in SDS loading buffer and gently heated for 10 minutes. The solubilized proteins were electrophoresed 2 cm into a 10-20% Tricine gel followed by in-gel trypsin digestion and peptide extraction. Replicate LC-MS/MS analysis of the tryptic peptides from each of the 27 samples was performed on a Thermo LTQ XL ion trap mass spectrometer. The resultant tandem mass spectra were searched against the human IPI database using the Myrimatch search algorithm and the results filtered using IDPicker. A spectral counting approach was applied to the data-dependent data to compare quantitative differences among the identified proteins. A total of 91,646 spectra were confidently identified, corresponding to 1671 protein groups across all biological and technical replicates. An average of 545, 543, and 610 protein groups were identified in the basal, ERBB2, and luminal A subtypes, respectively, from an equivalent of approximately 800 cells. Spectral count differences between datasets were analyzed using both a Quasi-likelihood Poisson regression model, developed in-house, and the Limma package in Bioconductor. Using a False Discovery Rate < 0.05, a total of 90 proteins showed statistically significant differences in expression among the breast tumor subtypes. Hierarchical cluster analysis generated a heat map of the protein expression patterns. Novel proteins as well as proteins corresponding to genes previously shown by gene array studies to be specific to each tumor type, including NES, ERBB2 and LGALS3BP, were represented among the differentially expressed proteins. These results demonstrate the feasability of this method to identify differential expression of protein biomarkers among breast cancer subtypes. These biomarkers should yield new insights into the pathogenesis of breast cancer and could be rapidly adapted into clinical diagnostics to guide therapeutic decisions for individual patients. The possible role in breast cancer for the novel proteins identified is currently under investigation. Citation Information: Cancer Res 2010;70(24 Suppl):Abstract nr P6-04-11.