Effects of ErbB2 Overexpression on the Proteome and ErbB Ligand-specific Phosphosignaling in Mammary Luminal Epithelial Cells

Abstract

Most breast cancers arise from luminal epithelial cells, and 25-30% of these tumors overexpress the ErbB2/HER2 receptor that correlates with disease progression and poor prognosis. The mechanisms of ErbB2 signaling and the effects of its overexpression are not fully understood. Herein, stable isotope labeling by amino acids in cell culture (SILAC), expression profiling, and phosphopeptide enrichment of a relevant, non-transformed, and immortalized human mammary luminal epithelial cell model were used to profile ErbB2-dependent differences in protein expression and phosphorylation events triggered via EGF receptor (EGF treatment) and ErbB3 (HRG1β treatment) in the context of ErbB2 overexpression. Bioinformatics analysis was used to infer changes in cellular processes and signaling events. We demonstrate the complexity of the responses to oncogene expression and growth factor signaling, and we identify protein changes relevant to ErbB2-dependent altered cellular phenotype, in particular cell cycle progression and hyper-proliferation, reduced adhesion, and enhanced motility. Moreover, we define a novel mechanism by which ErbB signaling suppresses basal interferon signaling that would promote the survival and proliferation of mammary luminal epithelial cells. Numerous novel sites of growth factor-regulated phosphorylation were identified that were enhanced by ErbB2 overexpression, and we putatively link these to altered cell behavior and also highlight the importance of performing parallel protein expression profiling alongside phosphoproteomic analysis.

Fig. 1.

Study workflow for protein expression (A) and phosphoproteomic (B) profiling. Independent cell cultures were reciprocally labeled with heavy and light lysine/arginine as shown for at least six passages. Cells were serum-starved and then stimulated with EGF or HRGb1 for 10 min (or left unstimulated) as shown. Cells were lysed, and equal amounts of protein were mixed from each condition, and then these were mixed to generate reciprocally labeled biological duplicate pools. These were separated by SDS-PAGE, and 50 gel slices per lane (n = 100) were excised and digested with trypsin, prior to LC-MS/MS. For phosphoproteomic profiling, a common reference pool was generated by pooling equal amounts of protein from light-labeled cultures, and this was used in singlet comparisons for each heavy-labeled condition. The six heavy/light mixtures were digested, separated into five SCX fractions, and each was subjected to SIMAC phosphopeptide enrichment generating three fractions each. Thus, 90 samples were generated and analyzed by LC-MS/MS. Protein identification, phosphosite identification, and SILAC-based quantification were performed for both datasets using MaxQuant software.

Fig. 2.

A, graph showing correlation between normalized protein abundance ratios common to biological replicate experiments. B, plots showing normalized protein abundance ratios versus ion intensity for each replicate experiment. Individual protein groups are gray-scaled according to significant B value. C, comparison of protein expression (SILAC) and mRNA (microarray) ratios between HB4a and C3.6 cells. Microarray data was taken from Ref. .

Fig. 3.

Protein interaction analysis of significantly up/down-regulated proteins. Significantly up/down-regulated proteins (significance B <0.05) were mapped to STRING interaction networks with a confidence score cutoff of ≥0.55 (intermediate confidence). Up-regulated gene-products are colored in dark gray and down-regulated proteins are in light gray. The number of edges per node represents the different types of evidence for the interaction. The interaction network was imported into Cytoscape, and densely connected protein clusters were identified using the graph theoretical clustering algorithm MCODE. Clusters with an MCODE score of ≥4 were analyzed for enriched GOSlim terms using the Cytoscape plugin Bingo as described. The table lists the top scoring GOSlim term for each cluster.

Fig. 4.

Western blotting confirmation of differential expression in HB4a and C3.6 cells. EGFR, ErbB2, and ErbB3 expression were also compared. Actin served as a loading control.

Fig. 5.

A, Western blotting showing suppression of IRF9 expression in C3.6 cells that could be induced by IFNβ or IFNγ treatment (24 h) and blocked by co-treatment with EGF or HRGβ1. CDK4 expression served as a loading control. B, IRF9 mRNA levels in IFNγ and IFNγ plus EGF (γ+E) co-treated HMLECs measured by qRT-PCR. C, IRF9 expression and ERK signaling in HMLECs treated with IFNγ and IFNγ plus EGF with or without pretreatment with protein synthesis inhibitor cycloheximide (CHX), MEK inhibitor PD098059 (PD), proteasome inhibitor PS341 (PS), or ErbB receptor kinase inhibitor AG1478 (AG). CDK4 expression served as a loading control. D, IRF9 protein expression is decreased by proteasome inhibitor (PS341) treatment in IFNγ-stimulated C3.6 cells (upper panel) and randomly growing HB4a cells (lower panel).