Importantly, erlotinib was able to effectively block ligand-induced EGFR INCB18424 phosphorylation in all cell lines tested, indicating that the ability of erlotinib to block EGFR activation was not impaired even after cells developed resistance to its growth inhibitory effects. To further investigate the relationship of p-AKT, p-ERK1/2 and Mig6 to the sensitivity of erlotinib, we again blotted the 26 cell line panel and plotted protein kinase inhibitors expression level against the IC50 of erlotinib. Our data showed that Mig6 expression was associated better with p-AKT than p-ERK1/2, which suggested that p-AKT pathway might be playing bigger role in regulating Mig6. Interestingly, introduction of Mig6 to H292 cells significantly increased resistance to erlotinib when concomitantly decreased basal EGFR phosporylation was seen. However, it did not affect sensitivity to erlotinib in SCC-S cells where EGFR phosporylation was not affected. Taken together, our data suggested that cellular dependence on EGFR, which can be predicted by basal Mig6/EGFR ratio, underlie the response of cancer cells to erlotinib rather than the absolute expression level of Mig6. This was further supported by our observation that Mig6/EGFR demonstrated a high degree of accuracy as the predictor of EGFR activity in a large panel of head and neck, bladder and lung cancer cell lines examined. In addition, in review of data from a published report, the relative expression of Mig6 and EGFR also correlates well with basal EGFR activity in a panel of breast cancer cells examined. To understand whether Mig6 knockdown in combination with p-AKT inhibition sensitize cells to erlotinib, we knocked down Mig6 and treated cells with AKT inhibitor. We found that AKT pathway inhibition could be detrimental to the resistant cells over the period of a few days. However, co-treatment with low dose of AKT inhibitor did sensitize cells to erlotinib in H1703 cells. To investigate whether our observations with tumor cell lines could be validated in tumor samples from patients, we analyzed directly xenografted low passage human tumors that have been shown to retain the key features of the original tumor, including drug sensitivity, and that accurately represent the heterogeneity of the disease. We obtained 4 human NSCLCs, and 18 pancreatic tumors that were directly xenografted into nude mice. No erlotinib-sensitizing mutations in EGFR were detected in any of these tumors. We initially tested the response of the 4 patient-derived lung xenografts to erlotinib. Among them, BML-5 showed a better response to erlotinib than the other 3 tumors. Analysis of Mig6 expression in tumor xenografts showed that BML-1 and BML-5 expressed less Mig6 than BML-7 and BML-11. In addition, BML-5 expressed higher total EGFR as well as higher basal EGFR phosphorylation than the other tumors. We next characterized and plotted erlotinib responsiveness of 18 directly xenografted pancreatic tumors. Tumor growth inhibition data are displayed with the most sensitive tumors on the far left 
and the most resistant on the far right. Tumor characteristics, including KRAS mutation status as well as EGFR expression and phosphorylation levels, have been reported previously. No EGFR sensitizing-mutations were found in any of these tumors and there was no correlation of KRAS mutation with erlotinib response in pancreatic tumors. EGFR negative tumors tended to cluster on the right side of the map, indicating that they were more resistant to erlotinib. However, in EGFR-positive tumors we saw little association between erlotinib sensitivity and EGFR expression. Instead, we found that in these pancreatic tumors, as Mig6 expression increased, tumors exhibited a more erlotinib-resistant phenotype. For example, the erlotinib-resistant tumor PANC420 expressed markedly higher Mig6 than the erlotinib-sensitive tumor PANC410, even though they expressed comparable amounts of EGFR protein. In keeping with their Mig6 expression status, PANC410 displayed heavy EGFR phosphorylation whereas PANC420 harbored no detectable EGFR phosphorylation. Interestingly, in the 3 erlotinib-resistant pancreatic tumors studied that displayed significantly lower Mig6 expression, IHC labeling revealed that 2 of these 3 xenograft lines did not express EGFR. Studies have suggested a weak association between EGFR protein expression levels and responsiveness to EGFR TKIs. Although the erlotinib-sensitive tumors studied here generally displayed high EGFR levels, our data suggested that it was the activity of EGFR.