It is increasingly clear that dysregulation of mitogenic signaling by constitutively activated oncoproteins in cancer cells drives high levels of feedback inhibition of the signaling network (Courtois-Cox et al., 2006
; O’Reilly et al., 2006
; Pratilas et al., 2009
). This may have important phenotypic consequences in the transformed cell. Hyperactivation of signaling by the oncoprotein may depend on its relative insensitivity to negative feedback or to other mutations that inactivate elements of the feedback machinery. Anticancer drugs that inhibit oncoprotein function will relieve this negative feedback and thus reactivate multiple signaling pathways that limit the extent and duration of the anticancer effects.
The PI3K-AKT signaling pathway is a central downstream effector of growth factor receptors and is often dysregulated in cancer. The Insulin and IGF1 receptors exert many of their physiologic effects by activating PI3K, but mutation or overexpression of other receptors, such as HER2 in breast cancer, commonly dysregulate the pathway in tumors (Holbro et al., 2003
; Kooijman et al., 1995
). Activation of PI3K-AKT leads to activation of many downstream targets that together account for the proliferative, antiapoptotic and metabolic effects of the pathway. Of these, the mTOR kinase has attracted much attention because of its central function in integrating nutrient and energy availability and growth signals in the regulation of cell proliferation and size. mTOR functions in two multiprotein complexes, mTORC1 and mTORC2 (Guertin and Sabatini, 2007
). The natural product rapamycin is a specific inhibitor of mTORC1 and leads to dephosphorylation of its two most well characterized substrates S6 kinase and 4EBP1(Brunn et al., 1997
; von Manteuffel et al., 1997
). It thereby inhibits cap-dependent translation and cell proliferation.
Experiments with rapamycin first revealed the extent and clinical implications of oncogene induced feedback. Insulin signaling is feedback regulated in part, by an mTOR/S6K-dependent phosphorylation and downregulation of the major insulin receptor substrate IRS1 (Haruta et al., 2000
). Inhibition of mTOR with rapamycin relieves this feedback, activates insulin and IGF signaling, and thereby activates PI3K and ERK signaling (O’Reilly et al., 2006
). This occurs in vivo
in patients and likely decreases the therapeutic efficacy of the drug (Mellinghoff et al., 2005
; O’Reilly et al., 2006
PI3K and AKT regulate many processes besides mTORC1 activity. We reasoned that, in tumor cells, mutational activation of the PI3K-AKT pathway would induce mTOR independent feedback pathway as well. We used a selective, allosteric inhibitor of AKT to assess AKT-dependent feedback in breast tumor cells in which the pathway is driven by amplification of HER2. We found that inhibition of AKT in these cells induced the expression of HER3. There was a concomitant induction of HER3-HER2 heterodimers and a marked induction of HER3 phosphorylation.
The results are consistent with the idea that AKT activation causes feedback inhibition of HER kinase expression, especially of HER3, which, when phosphorylated docks with and activates PI3K. The induction of HER3 in response to AKT inhibition is associated with an increase in HER2-HER3 heterodimers and leads to increased HER3 phosphorylation. HER3 phosphorylation is blocked by the HER1/2 kinase inhibitor Lapatinib, but the increase in HER3 expression is not. This finding suggests that the increase in HER3 expression is in large part responsible for the observed increase in phosphorylation.
HER3 expression is induced by inhibitors of PI3K or AKT or by knockdown of AKT. That induction of HER3 expression and phosphorylation in response to AKT inhibition represents release of AKT-dependent feedback inhibition of the pathway is supported by the downregulation of HER3 expression that occurs when the AKT inhibitor is washed out of cells. We used phospho-receptor tyrosine kinase arrays to ask whether AKT-induced negative feedback was confined to HER3 or involved other receptors as well. We found that, although HER3 induction was very prominent, the phosphorylation of multiple other receptors was induced as well. Induction of receptor phosphorylation was not confined to HER2-dependent breast cancers, it occurred in tumor cells derived from all lineages tested (breast, prostate, ovary, lung, melanoma). We identified a set of nine RTKs whose phosphorylation is commonly induced after AKT inhibition. Four of these (HER3, IGF1R, Insulin receptor, and EphA7) responded in almost all cells tested.
Phosphorylated HER3 has a high capacity and affinity for PI3K, docking it to the membrane. The most obvious physiologic role of PI3K-AKT signaling is mediating the effects of the Insulin and IGF1 receptors. It seems from our data that these three receptors are coordinately feedback downregulated by AKT when the pathway is activated. AKT inhibition induces the expression as well as the phosphorylation of HER3, IGF1R, and Insulin receptors. Induction of other kinases such as RET and HER4 is confined to phosphorylation; expression is typically unaffected. HER2 is the dominant activated kinase in breast cancers in which it is amplified and, in these tumors, the induction of phosphorylation of the other RTKs is HER2 dependent. Lapatinib blocks their phosphorylation, but not the induction of expression of IGF1R and IR. Previous work by other labs has demonstrated that IGF1R and Insulin receptor kinases are not antagonized by Lapatinib at doses as high as 3μM and the lack of effect of Lapatinib upon IGF1R and Insulin receptor in non-HER2 driven models like H1975 (Fig. S3
) supports that the activity seen here is not due to direct inhibition of those kinases. Whether the HER-kinase dependence of induction of IGF-1R/insulin receptor phosphorylation represents transphosphorylation of these kinases by HER2 or a HER2 dependent activation of autophosphorylation is under investigation. Activation of IGF1-R and Insulin receptor by AKT inhibition does involve both induction of expression and HER-kinase dependent phosphorylation of these kinases. In the non-small cell lung cancer model H292, the induction of phosphorylation of some RTKs like HER3 is HER kinase inhibitor sensitive. Others, such as FGFR and IGF1R are insensitive. It is clear that activation of AKT in tumors induces a complex and broad pattern of feedback inhibition of RTKs that is relieved by inhibition of AKT.
TORC1 inhibition by rapamycin has also been shown to activate signaling and less selective PI3K inhibitors that target both mTORC1 and PI3K have been shown to induce HER3 expression (Amin et al., 2010
; Sergina et al., 2007
). We asked whether AKT inhibition activated signaling via inhibition of mTORC1. Rapamycin partially reproduced the effects of AKT inhibition, inducing the phosphorylation of HER3 along with several other RTKs. However, induction of HER3 was considerably weaker than that observed with AKT inhibition and the phosphorylation of most of the RTKs induced by AKT inhibition was unaffected by rapamycin. The differences between the effects of the AKT inhibitor and rapamycin suggest that there are AKT-regulated feedback pathways that are not mediated by TORC1.
A clue to the nature of these pathways came from studies on the mechanism of induction of expression of HER3, IGF1R and insulin receptor. AKT inhibition results in marked induction of the mRNAs encoding these receptors, whereas rapamycin has either no or marginal effects. AKT has been shown to phosphorylate the FOXO family of transcription factors and thereby prevent their nuclear translocation, thus inhibiting their function (Brunet et al., 1999
). We show that AKT inhibition recruits FOXO proteins to the HER3 promoter and that FOXO1/3/4 knockdown with siRNA suppresses the induction of IGF1-R/IR/HER3 expression and phosphorylation. We note that the knockdown of FOXO proteins has little effect on the basal expression of the RTKs. We postulate that in these cells with activated PI3K/AKT signaling, FOXOs are effectively inhibited and expression of HER3, IGF1R and IR are dependent on other factors in this state. However, AKT inhibition results in activation of these transcription factors enabling them to promote RTK expression. We thus conclude that AKT regulates the expression of these receptors by inhibiting FOXO-dependent transcription.
We propose the following model based on our current understanding to explain the regulation of PI3K-AKT signaling by negative feedback in tumors and how it is affected by targeted drugs (). Receptor activation of PI3K-AKT causes AKT-dependent phosphorylation of FOXO proteins which downregulate the expression of some of the receptors that are tightly coupled to PI3K, including HER3, IGF1R and insulin receptor. In addition, AKT activation leads to activation of TORC1 and S6K which feedback inhibits IRS1 expression and other undefined regulators of receptor signaling. The result is downmodulation of the signal.
AKT inhibition and mTORC1 inhibition relieve feedback inhibition at unique nodes of oncogenic growth factor signaling pathways
Therapeutic inhibition of different components of the pathway reactivates feedback, but by mechanisms specific to the inhibited target (). Thus, AKT inhibition will result in activation of FOXO-dependent transcription of receptors and inhibition of S6K-dependent inhibition of signaling with resultant activation of multiple receptors. The downstream effects of AKT will be suppressed, but other RTK driven signaling pathways will be activated. In contrast, TORC1 inhibition blocks S6K-dependent feedback, activates IGF and HER kinases but not their expression and thus activates both AKT and ERK signaling.
These findings have important basic and therapeutic implications. The enhancement of signaling by autocrine activation or mutation of RTKs that activate PI3K-AKT signaling would be expected to be limited by negative feedback. Selection of oncogenes that encode proteins that overcome or are unresponsive to feedback would be favored (e.g. activating mutation in PI3K or loss of PTEN). All drugs that inhibit components of dysregulated mitogenic signaling pathways would be expected to relieve feedback inhibition of other components of the signaling network. This may reduce the antitumor effects of the drug, but also ameliorate toxicity. Combined inhibition of the oncoprotein and key pathways reactivated by inhibition of negative feedback should have enhanced antitumor activity. This is consistent with our finding that the AKT inhibitor causes tumor regressions when combined with low doses of HER kinase or HSP90 inhibitors that prevent or attenuate induction of receptor phosphorylation. Whether effective inhibition of both PI3K-AKT signaling and feedback reactivated pathways will have an enhanced therapeutic index will have to be evaluated in clinical trials.