Human tumors almost invariably harbor mutations in a multitude of oncogenes and tumor suppressor genes. Mutations that result in activation of the PI3K/AKT/mTOR and RAS/RAF/MEK/ERK pathways are especially frequent. Moreover, mutations that activate these two pathways often coexist in certain tumors; thus RAS
mutation, and variant EGFR expression and PTEN
mutation occur together in colorectal carcinoma, thyroid carcinoma, melanoma, and glioblastoma, respectively (Liu et al., 2008
; Mellinghoff et al., 2005
; Nosho et al., 2008
; Simi et al., 2008
; Tsao et al., 2004
). Tumors with activation of PI3K/AKT signaling in the absence of EGFR
mutation tend to be dependent on the pathway and sensitive to selective inhibition of AKT (She et al., 2008
; Yu et al., 2008
). Similarly, tumors with RAS
mutation tend to be sensitive to MEK inhibition if PI3K
are not also mutated (Hoeflich et al., 2009
; Solit et al., 2006
; Wee et al., 2009
and Figure S2A
). RAS-dependent tumorigenesis in animal models has been reported to require PI3K activation by RAS, but the growth of established tumors with RAS mutation is insensitive to PI3K inhibitors (Engelman et al., 2008
) and, as shown here, to AKT inhibitors.
The usual rationalization for coexistent oncogenic mutations in components of the signaling apparatus is that they mediate different aspects of the transformed phenotype that allow for their co-selection. However, in this paper and elsewhere, we and others have shown that tumors with coexistent mutation of both pathways tend to be insensitive to inhibition of either alone, but sensitive to their combined inhibition (Engelman et al., 2008
; Hoeflich et al., 2009
; She et al., 2005
; Wee et al., 2009
; Yu et al., 2008
). These results suggest that neither pathway alone subserves a key function or that the original selective advantage of the first mutation has been lost. In this paper, we provide an explanation for the loss of dependence of these tumors on either pathway alone. In tumors sensitive to AKT inhibition, phosphorylation of certain downstream targets such as S6 and 4E-BP1 and cap-dependent translation are dependent on AKT signaling. In contrast, in tumors with co-activation of both AKT and ERK, inhibition of either is insufficient to adequately inhibit these processes; inhibition of both is required. Moreover, deletion of the oncogenes responsible for activation of either pathway is sufficient to confer dependence on the other. The results suggest that PI3K/AKT and MEK/ERK signaling converge on a common set of targets that integrate their function. Activation of either pathway is sufficient to affect these integrators, thus the second mutation eliminates the dependence of both the target and the tumor cell on either.
AKT and ERK signaling affect many common downstream targets and processes, including regulators of cell cycle progression, apoptosis, transcription and translation (Manning and Cantley, 2007
; McCubrey et al., 2007
). In normal cells, these functions are regulated by a complex signaling network, but, in tumor cells, ‘oncogene addiction’ implies that they have become dependent on a single, dominant, oncoprotein activated pathway. Mutational activation of the second pathway would then serve to reduce dependency on either. The downstream convergence of PI3K/AKT and ERK signaling may account for the significant frequency of coexistent mutations in these pathways. The selective advantage for the second mutation is not certain; it may lie in divergent effects of the second pathway but it is also possible that the dependence of key processes such as translation on a single oncogene-activated pathway may result in decreased fitness of the cell in certain environments. In support of this possibility, the growth of tumor xenografts with mutant RAS
is slowed in calorie restricted mice and this effect is rescued by coexistent PIK3CA
mutation (Kalaany and Sabatini, 2009
). This interpretation is consistent with that of Ericson et al. who report that in tumors with coexistent RAS
mutations, AKT was required for growth only in challenging microenvironments, such as growth factor depletion and during the metastatic process (Ericson et al., 2010
). Whatever the mechanism of selection, it is clear that the second mutation reduces or eliminates the dependency or ‘addiction’ of the tumor to the first mutation. Whether this loss of dependency is responsible for the selection or is a neutral byproduct of the second hit, it has important clinical and biologic implications.
The data reported here support recent studies that show that activation of cap-dependent translation plays an important role in induction and maintenance of the transformed phenotype (Mamane et al., 2004
; Polunovsky and Bitterman, 2006
). The phosphorylation of two components of the translation machinery, S6 and 4E-BP1 was shown to be dependent on AKT signaling in tumors in which the PI3K/AKT pathway is dysregulated, but not in those in which there is coexistent mutational activation of ERK signaling. In such tumors, combined inhibition of both pathways is required to affect their phosphorylation and to significantly inhibit cap-dependent translation. Thus, these two proteins are candidate integrators of AKT and ERK signaling that may play a role in mediating transformation and oncoprotein dependency.
In particular, 4E-BP1 is identified as a key downstream target of both mutant PI3K
-activated signaling in human cancer cells. Knockdown of this inhibitor of translation in tumor cells markedly reduces their dependence on activated signaling for translation and survival. This is somewhat surprising, given that these pathways also activate the phosphorylation of the S6K, S6 ribosomal protein and other regulators of translation, including other members of the 4E-BP family (Pause et al., 1994
). However, in the experiments reported here, knockdown of either S6K, S6 or 4E-BP2, alone or in combination with 4E-BP1 has more than a marginal effect. This suggests that 4E-BP1 inhibition is responsible for much of the activation of translation by RAS and PI3K/AKT in these cells and this in turn plays an important part in mediating the effects of these pathways in the tumor. It is consistent with recent clinical findings that expression of high levels of phosphorylated 4E-BP1 are associated with poor prognosis in several tumor types, independent of specific upstream oncogenic alterations (Armengol et al., 2007
). The AKT dependence of phosphorylation of 4E-BP1 and of tumor growth is closely correlated. These data suggest that this relationship is causal. This is supported by our finding that a dominant negative 4E-BP1 incapable of being phosphorylated in response to upstream pathways is sufficient to suppress the growth of HCT116 tumors in vivo. Others have found that the non-phosphorylated (activated) 4E-BP1 is capable of suppressing tumorigenesis in PTEN
-mutant breast cancer (Avdulov et al., 2004
) and KRAS
mutant non-small cell lung cancer (Jacobson et al., 2006
). We thus show that tumor cells in which both pathways are activated are insensitive to inhibition of either, but sensitive to their combined inhibition or to dominant activated 4E-BP1. Furthermore, tumors in which eIF4E is overexpressed or 4E-BP1 expression is knocked down lose dependence on AKT and ERK signaling. Taken together, these data support the conclusion that inhibition of 4E-BP1 function by activation of AKT and ERK signaling plays a crucial role in maintaining the transformed phenotype and add support to the idea that the eIF4E complex represents a valid and intriguing target for drug development (Graff et al., 2008
The mTOR kinase is another downstream target of both AKT and ERK signaling that integrates their function. This may occur via phosphorylation of TSC2 and perhaps other proteins by both pathways (Ma et al., 2005
; Manning et al., 2002
). The mTOR-containing TORC1 complex is responsible for phosphorylation of S6K and 4E-BP1 by the enzyme. Rapamycin is a selective inhibitor of the TORC1 complex, but is much less effective than combined inhibition of AKT and MEK in downregulating 4E-BP1 phosphorylation and its binding to eIF4E, or inducing apoptosis in tumor cells with coexistent RAS
mutations (data not shown). This suggests that the effects of AKT and MEK inhibition are mediated by other targets in addition to mTOR. However, this result is complicated by the recent report that rapamycin is only a modest inhibitor of TORC1 activity and that mTOR kinase inhibitors are much more efficient downregulators of 4E-BP1 phophorylation (Feldman et al., 2009
; Thoreen et al., 2009
). However, the TORC2 complex is also an upstream activator of AKT (Sarbassov et al., 2005
) and T70 phosphorylation of 4E-BP1 is sensitive to AKT/MEK inhibition and reported to be insensitive to mTOR kinase inhibition (Thoreen et al., 2009
). Furthermore, phosphorylation of 4E-BP1 and its activity have also been shown to be regulated by the PP2A phosphatase and other kinases independent of mTOR (Herbert et al., 2002
; Imai et al., 2008
; Michlewski et al., 2008
). Thus, it is still unclear whether all of the effects of AKT and ERK signaling on 4E-BP1 are integrated by mTOR.
It is likely that some of the effects of combined inhibition of AKT and ERK are mediated by other targets, including components of the apoptotic machinery (She et al., 2005
). We have previously shown that BAD is a downstream target that can integrate EGFR/ERK and PI3K signaling in PTEN-negative/EGFR amplified tumors and that knocking down BAD significantly (~50%) attenuates the effects of combined pathway inhibition in MDA-468 breast cancer cells (She et al., 2005
). In HCT116 cells, knockdown of BAD expression reduces induction of apoptosis in response to combined pathway inhibition by approximately 25% (data not shown). How activation of cap-dependent translation interacts with regulation of apoptotic regulators to mediate oncogenic survival signaling is likely to be complex and a matter for further investigation.
These are important questions because relatively selective inhibitors of RAF, MEK, PI3K, AKT and mTOR kinases are now available and many are in early clinical testing. This work suggests that the tumors from patients in these trials should be evaluated for mutations in components of both pathways and tumors with coexistent mutations in both pathways will not respond to inhibition of one alone. This hypothesis should now be tested in these clinical trials. Furthermore, dephosphorylation of 4E-BP1 in response to drug should be an important biomarker for predicting response to therapy.
The tolerability of the combined inhibition of AKT and ERK and its synergistic effects on cap-dependent translation and on tumor growth suggest that this strategy might be useful in the variety of metastatic tumors in which these pathways are co-activated. There is currently no therapeutic agent that directly and effectively inhibits RAS function. Since RAF and PI3K are two of the key effectors of the transforming activity of mutant RAS
, the combined inhibition of MEK and AKT may constitute an anti-RAS therapeutic strategy as well, of potential utility in diseases (pancreatic, colon, lung carcinoma) with mutated RAS
for which there are few and only marginally effective therapies. Given the importance of 4E-BP1 in integrating the effects of AKT and ERK on protein translation and apoptosis, mTOR kinase inhibitors currently in development may also be useful for treating these tumors. However, these inhibitors release the feedback inhibition of receptor tyrosine kinases and activate both ERK and PI3K/AKT in tumors (N Rosen, unpublished data and Carracedo et al., 2008
). Combined inhibition of ERK and AKT both effectively inhibits 4E-BP1 phosphorylation and prevents reactivation of ERK and AKT and thus may have a therapeutic advantage.