Successful diagnosis and targeted treatment of cancer requires a detailed understanding of the signaling pathways being targeted. However canonical signaling pathways are typically constructed from multiple experiments performed on a variety of cell and tissue types and are thus highly non-specific. To address whether canonical pathway models should be applied to diverse cancer sub-types our studies focus on a panel of 51 breast cancer cell lines which reflect the genomic and biologic heterogeneity of primary tumors. To develop a functional map of the ERK-pathway we assessed the relative levels of RAF, MEK and ERK isoforms and phosphorylation states using western blotting and a nano-capillary immunoassay technology. Considering each of these measurements as nodes in a network, we used a Bayesian Network model to determine the connectivity of the ERK pathway from these highly specific data. This approach allowed us to capture combinatorial influences between nodes in the network in a statistically rigorous manner. Our results clearly showed distinct ERK-pathway connectivity in luminal compared to basal tumors. We found that MEK1 phosphorylated on S298 in the absence of T292 phosphorylation is associated with ERK2 (T185/Y187) phosphorylation in luminal but not basal tumors. Conversely, in basal tumors MEK1 phosphorylation on S298 or T386 (but not both) is associated with ERK1 (T202/Y204) phosphorylation. When we expanded our analysis to include PKC isoforms we found PKCα (T638) phosphorylation influences the ERK pathway more strongly in basal tumors compared to luminal tumors.
Commonly, isoforms of signaling proteins such as MEK and ERK are treated synonymously when describing activation of the ERK pathway. In our cell lines, there is a good correlation between levels of ERK1 (T202/Y204) and ERK2 (T185/Y187) phosphorylation which might support this notion. However, our Bayesian analysis shows that phosphorylation of ERK1 and ERK2 are the result of distinct associations between different isoforms and phosphor-states of MEK, ERK and PKC in different breast cancer sub-types. This finding has profound implications for conceptualizing signaling networks, suggesting that signal propagation through the ERK pathway is context-dependent. From a translational perspective, our data provide a mechanistic basis for understanding why inhibiting MEK1 results in very different outcomes in different tumors. These context-specific features could be exploited by rational design of isoform-specific therapeutic agents or combinatorial approaches directed at tumors whose phosphor-protein profiles have been comprehensively mapped.
98th AACR Annual Meeting-- Apr 14-18, 2007; Los Angeles, CA