Emerging Roles for AKT Isoform Preference in Cancer Progression Pathways

Abstract

The phosphoinositol-3 kinase (PI3K)-AKT pathway is one of the most mutated in human cancers, predominantly associated with the loss of the signaling antagonist, PTEN, and to lesser extents, with gain-of-function mutations in PIK3CA (encoding PI3K-p110α) and AKT1. In addition, most oncogenic driver pathways activate PI3K/AKT signaling. Nonetheless, drugs targeting PI3K or AKT have fared poorly against solid tumors in clinical trials as monotherapies, yet some have shown efficacy when combined with inhibitors of other oncogenic drivers, such as receptor tyrosine kinases or nuclear hormone receptors. There is growing evidence that AKT isoforms, AKT1, AKT2, and AKT3, have different, often distinct roles in either promoting or suppressing specific parameters of oncogenic progression, yet few if any isoform-preferred substrates have been characterized. This review will describe recent data showing that the differential activation of AKT isoforms is mediated by complex interplays between PTEN, PI3K isoforms and upstream tyrosine kinases, and that the efficacy of PI3K/AKT inhibitors will likely depend on the successful targeting of specific AKT isoforms and their preferred pathways.

Background

AKT (also known as protein kinase B (PKB)) is a serine/threonine kinase first identified in 1987 as the oncogene of the acute transforming retrovirus, AKT8, corresponding to AKT1 and AKT2 human proto-oncogene homologues (1). A third isoform, AKT3, was identified a decade later (2). Further research identified phosphoinositide 3-kinase (PI3K) as the upstream signaling activator of AKT (3) through its ability to generate PIP₃ (4), and that the phosphatase and tensin homologue (PTEN) functions as a negative regulator of this PI3K/AKT pathway based on the ability of its lipid phosphatase activity to revert PIP₃ to PIP₂ (5,6) (Figure 1). The notion that AKT isoforms play non-redundant roles was made evident even at the initial discovery publication, which showed gene amplifications of AKT1, but not AKT2, in gastric adenocarcinomas (1). These non-redundant roles were further evidenced by the findings that Akt1^−/− mice suffer embryonic lethality, Akt2^−/− mice develop severe diabetic phenotypes, and Akt3^−/− mice predominantly exhibit reduced brain sizes rather than cognitive deficiencies (7–11). Akt1^−/− that do survive also exhibit impaired body growth (7). Only Akt2^−/−;Akt3^−/− intertypic crosses were viable, suggesting both compensatory and non-redundant roles in development (9).

Figure 1: AKT pathways and mediators.

The activation of PI3K via receptor tyrosine kinases (RTK) or G-protein coupled receptors (GPCR) facilitates AKT-regulated cell growth, proliferation, survival, migration, and metabolisms mediated by the phosphorylation of substrates such as Palladin (PALLD), p21^Waf1 (CDKN1A), S6K (RPS6KB1), FOXO family members or GSK3β. Actuation of maximal AKT serine/threonine kinase activity is mediated by a feedback cycle involving activation of mTORC2 followed by its phosphorylation of AKT on Ser473, and by PDK1 phosphorylation of AKT on Thr308.

AKT activation is regulated by a multilayer process, beginning with PI3K activation by receptors tyrosine kinases or G-protein coupled receptors. A major role for PI3K is as a lipid kinase that phosphorylates phosphoinositol 4,5-bisphosphate (PIP2) to produce PIP3. AKT is then recruited to plasma membrane sites through the interaction of its Pleckstrin Homology (PH) domain to PIP3 islands in the plasma membrane (12). Indeed, roughly 1% of all cancers (cBioPortal analysis of MSK-IMPACT pan-cancer cohort) have an E17K mutation in the AKT1-PH domain, resulting in increased association with PIP2 and oncogenicity in embryo fibroblasts (13). Following this, AKT is phosphorylated at two sites, T308 and S473, both of which are required for full activation: T308 is phosphorylated first by PDK1,which triggers a positive feedback mechanism mediated by the mTORC2 complex which phosphorylates AKT at S473, resulting in its maximal activation (14–16). The finding that AKT-S473 phosphorylation can be carried out by other kinases, including DNA-PK and PKCγ (17,18) complicates the use of AKT^poS473 as a marker of PI3K activation¹.

Shared and Distinct Roles of AKT Isoforms in Cellular Biology and Cancer

AKT is responsible for a variety of key parameters of normal or oncogenic cell behavior, including cell migration, proliferation, invasion, survival and metabolism (19–22). Many of the roles for AKT isoforms are, in fact, redundant (Figure 2). These include negative regulation of i) Bcl-2 family proteins, ii) GSK3 isoforms responsible for cell survival and cellular metabolism, or iii) TSC2, a known tumor suppressor (23). There has been increased attention on defining both shared and isoform-specific AKT roles in these biologies. Noting that AKT activation regulates cellular energetics through the control of intracellular ATP levels (24,25), multiple studies identified key roles for AKT2 in mediating metabolic regulation induced in response to insulin or glucose stimulation, lipogenesis or GLUT4 trafficking (8,26–28). However, there is also a role for AKT1 in metabolism including insulin production in pancreatic islet cells and glucose metabolism in breast tissue during lactation (29,30). Similarly, although AKT2 was shown to play a more dominant role in tumor migration and invasion, some overlapping roles for AKT1 are also identified (31,32). Indeed, a demonstrated E17K PH domain oncogenic mutation in AKT1 has been identified in 3–5% of all cancers in cBioPortal, and equivalent putative oncogenic mutations for AKT2 and AKT3, though less frequent, have also been identified, with the note that all these mutations cause increases in relative AKT kinase activity (33). This may lead to conditions such as Proteus syndrome with the E17K mutation (33). In addition, presence of the poS473 residue on AKT by immunohistochemistry (IHC) has been correlated with poor outcome in prostate cancer patients (34,35). This led to the notion that mutations that disrupt auto-inhibitory PH-kinase interactions can facilitate oncogenic progression by inducing the kinase activities of AKT isoforms (36). Many cancers with copy number losses or function-negating mutations in PTEN have increased AKT activation, as assessed by relative increases in poT308 and poS473 levels, or have gene amplifications of AKT isoform genes (37), which correlate with increased expression levels of AKT isoform proteins (38). Importantly, high AKT activation levels correlate with worse overall survival in hematological and solid cancers (39).

Figure 2: Roles for PI3K and AKT isoforms in specific parameters of oncogenesis.

Specific and sometimes conflicting roles have been attributed to AKT1 or AKT2 in parameters of oncogenic growth, proliferation, survival motility or metabolism. Preferred upstream control of AKT1 or AKT2 is likely controlled PI3K containing p110α or p110β kinase domains, respectively. Though less clarified, several roles for AKT3 in oncogenic progression have been identified recently.

Even though all of the three AKT isoforms are considered to be oncogenic drivers, based on cases of AKT isoform gene amplification or kinase-activating mutation, recent data strongly suggest that they play overlapping and non-redundant roles depending on the specific oncogenic progression parameter studied. For example, in mouse transgenic models of Erbb2-driven mammary carcinoma, the expression of activated Akt1, in which the T308 and S473 residues are switched to phospho-mimetic Asp residues (Akt1^T308D,S473D), increased primary tumor development but did not increase metastatic or invasive potential (31,40). In contrast, Akt2^T309D,S474D had little effect on the latency of mammary tumor growth in transgenic mice induced by the expression of activated ErbB2 or polyomavirus middle T-ag, but it induced marked increases in pulmonary metastases and in tumor cell invasiveness (31,41). Further, in non-tumor biology, AKT1^S473D, but not AKT2^S474D, can induce lipogenesis in mature brown adipocytes (Martinez Calejman et al., 2020). These experiments are important in that they show that these phospho-mimetic mutations retain AKT1- or AKT2-specific signaling functions. In contrast, fusions of AKT isoforms to the N-terminal Src myristylation signal (“myr”- MGAG residues) induces constitutively-active AKT kinase activity presumably by bypassing the need for PH domain-mediated membrane association by AKT (42,43). Myr-AKT isoform variants can oncogenically transform embryo fibroblasts, and when expressed in transgenic mice via the mammary-specific MMTV promoter, increase susceptibility to carcinogen-induced mammary carcinomas (43–45). In contrast, these same studies showed that overexpression of wt-AKT isoforms, whose intrinsic kinase activities were relatively much lower, were >100-fold poorer at transforming embryo fibroblasts. Lastly, the notion that forcing membrane association of AKT is sufficient to induce its oncogenic activity is underlined by the fact that the v-akt oncogene contains an N-terminally myristoylated GAG fusion that promotes membrane association (46,47). Taken together, these data strongly suggest that pro-oncogenic roles for AKT isoforms direct relate to both kinase activities and membrane association.

In contrast to the phospho-mimetic activated AKT isoform variants, however, there is no formal demonstration that the myr-AKT variants retain isoform-specific signaling, leading to the hypothesis that constitutive membrane association abrogates isoform-specific effects. Indeed, AKT isoforms may manifest their signaling specificities by acting in different cellular compartments. For example, Gonzalez and McGraw showed that although insulin induced equal activation of Akt1 and Akt2 in mouse adipocytes (based on relative poT308/9 levels), activation of the GLUT4 glucose transporter was facilitated by greater accumulation of Akt2 at the plasma membrane; expression of the Akt1-E17K variant, whose PH domain mutation facilitates constitutive plasma membrane association, overcame the Akt2 dependence for GLUT4 activation (48). Santi and Lee showed that in MDA-MB-231 human breast cancer cells, AKT1 was mainly found in the cytoplasm, AKT2 in the mitochondria and AKT3 in the nucleus and at the nuclear membrane, and that isoform-specific knockdown did not alter the localization of the other isoforms (49). They also showed whereas insulin induced the selectively accumulation of AKT2 to plasma membrane sites, swapping the AKT2-PH to AKT1 did not facilitate its movement to the plasma membrane. Indeed, swapping the AKT2-PH produced an AKT1 that was incapable of inducing cell proliferation and G1/S transition (50)Complicating this further are multiple, sometimes conflicting findings in breast cancer tissue samples that show enrichment of activated AKT1 and AKT2 in the nucleus (reviewed in (51)). These data suggest that AKT isoform-specific pathways could be facilitated by non-redundant interactions with signaling mediators enriched in particular cell compartments, and that some of these may be regulated by isoform-specific PH domains. An example of the latter is how the selective binding of TCL1b to AKT3 is dependent on its PH domain, a funciton that cannot be replacedby the AKT1 PH domain (52).

Although less is known about AKT3-specific roles in cancer progression, starting with the initial observation by Nakatani et al. in 1999 that AKT3 expression and phosphorylation are upregulated in ER+ breast cancers and androgen-independent prostate cancers, multiple studies have elucidated AKT3-specific oncogenic roles, as reviewed by Hinz and Jücker (2,51). These include mediating endocrine therapy resistance in ErbB2-driven mammary carcinomas and AKT inhibitor resistance in breast cancers, or suppressing chemotaxis and metastasis of triple-negative breast cancer cell lines (53–55).

Non-redundant roles for AKT1 and AKT2 have also been shown in prostate cancer using genetically-engineered mouse models and tetracycline-regulated AKT isoform shRNA knockdowns: AKT1 promotes primary tumor growth, whereas AKT2 is responsible for development of distant metastasis and disease aggressiveness (56,57). More recent research has suggested that this isoform plasticity is dependent on other factors within the PI3K/AKT pathway. For example, PTEN-deficient tumors rely more on AKT2 than on AKT1 signaling for maintenance and survival under 3D growth conditions (56). A significant body of data indicate that the preferential activation of AKT isoforms can be influenced by which upstream PI3K isoforms are activated. For example, multiple studies show increased activation of PI3K-p110β in cancers with PTEN loss, corresponding to increased relative activation levels of AKT2 (56,58). In contrast, activation of Src-family kinases through the stimulation of specific growth factor receptor tyrosine kinases leads to the preferential activation of PI3K-p110α and the subsequent activation and dependence on AKT1 (59). Nonetheless, some plasticity exists within this axis: treatment of PTEN-null breast or prostate cancers with a PI3K-p110β inhibitor induces drug resistance by upregulating p110α and HER3/ERBB3 (59–61). Thus, optimal cancer suppression requires therapeutically targeting multiple nodes upstream of AKT activation, such as multiple p110 isoforms and/or activated tyrosine kinases (61). Lastly, it is important to note the existence of AKT-independent PI3K cancer progression pathways (62,63), suggesting that AKT activation markers (e.g.- poT308 and poS473), though important, are insufficient to fully assess the role of activated PI3K in cancer progression.

AKT Isoform-Preferential Substrate Identification: Challenges and Successes

From a logical standpoint, it is reasonable to assume that non-overlapping or opposing roles for the AKT isoforms involve distinct sets of isoform-specific substrates, whereas the overlapping roles utilize shared substrates. As will be described below, limited attempts have been made to identify purely isoform-specific substrates, and even in these cases, it is unclear whether the isoform-specificity is only cell type- or context-specific. This has led to the likelihood that AKT isoform-dependent biologies utilize “preferential” substrates, yet to fully define these, cell-type and cancer-specific contexts will need to be considered. As an example, whereas AKT1 is pro-migratory and AKT2 anti-migratory in untransformed fibroblasts, opposite roles are manifest in breast cancer cell lines (64).

The quest to identify a consensus AKT substrate phosphorylation motif was heavily influenced by the initial substrates studied and subsequent experimental techniques that heavily focused on this canonical motif. The first AKT substrate validated was glycogen synthase kinase 3 beta (GSK3β), based on the ability of wortmannin- a PI3K inhibitor, to block insulin-induced GSK3β phosphorylation (65) and on the ability of immunoprecipitated PKB/AKT to phosphorylate GSK3β in vitro (66). The current list of >100 AKT substrates includes protein and lipid kinases, signaling proteins, transcription factors, ubiquitin ligases, metabolic enzymes and cell cycle regulators (66). The initial description of a possible consensus AKT substrate motif was developed by comparing the AKT/PKB phosphorylation sequence shared between GSK3β, MAPKAP kinase-1 (RPS6KA1/p90Rsk) and p70 S6Kinase (RPS6KB2), and then validated on a limited set of 7- to 10-mer peptides in in vitro kinase assays containing HA-tagged AKT1 immunopurified from HEK-293 cells, eventually defining the canonical substrate motif as RxRxxS/Tφ, with S/T representing the phosphorylated residues and φ representing hydrophobic amino acids (16,23,67). More recently, Balasuriya et al. (68) sought to assess how partial AKT1 phosphorylation (poS473 or poT308) affected substrate preference and maximal kinase activity when compared to fully activated AKT1 (containing both poS374 and poT308). They used three peptide libraries, one enriched for the AKT substrate motif found in 84 known substrates, plus two “oriented peptide array libraries” containing ~10¹¹ peptides each, one of which (OPAL1) had Arg fixed at the −3 position, based on its known preference in canonical AKT substrate motif. They showed less dependence on Arg at the −5 position for AKT1^poS473/T308, consistent with a Proline at the −5 position in the AKT phosphorylation sites on AMPK and ATP-citrate lyase (ACLY) (69). However, this study did not attempt to identify potentially non-canonical AKT1 substrate motifs using the more random OPAL2 peptide library. Indeed, one drawback to substrate site/motif identification using in vitro assays is that they lack the context of cellular compartmentalization, and thus, “substrates” that might never interact with AKT isoforms might be identified artefactually. In contrast, a problem for substrate identification in cells is that other kinases, such as RSK, PIM2 or protein kinase A (PKA), can phosphorylate versions of the AKT substrate motif (67,70–72). PIM2 and AKT can phosphorylate overlapping sites on the anti-apoptotic protein, BAD, and on the cell cycle regulator, p21, and the AKT phosphorylation site on ACLY (S455) can be also phosphorylated by PKA, mTOR or BCKDK (73–78). This points to the complication that substrates identified by phosphoproteomics analyses in cells cannot readily distinguish between direct and indirect substrates. Thus, even when using a canonical AKT substrate motif, confidence in the growing list of shared and preferential AKT isoform substrates requires combining data from in vitro kinase assays and from cellular proteomics data using either isoform-specific knockouts, knockdowns or overexpressions, as described below, or AKT inhibitory drugs (79)

The importance of alternative AKT substrate motifs increases with the discussion of how specific pathways or biologies are controlled by specific AKT isoforms (51). Thus, in addition to shared substrates that encode the RxRxxS/Tφ motif, phosphorylation of AKT isoform-preferred substrates is likely facilitated via either non-redundant, non-canonical motifs or substrate conformations. For example, Palladin, was identified because the phosphorylation motif surrounding S507 conformed to the canonical AKT motif (80). Newer data show that Palladin is a preferred substrate of AKT1 but not AKT2, whereas AKT2, but not AKT1, regulates its expression (32,81). Girardi et al. suggest that the ability of casein kinase 2 to phosphorylate AKT1, but not AKT2, helps drive the kinase specificity of AKT1 for palladin (82).

There have been a small number of phosphoproteome analyses to identify AKT isoform-preferential substrates. For example, (83) performed LC-MS/MS on phospho-serine or -threonine peptides from Akt1/2/3 knockout (TKO) lung fibroblasts (from transgenic mice) vs. TKO cells re-expressing Akt1, Akt2 and/or Akt3. Sixteen new substrates were identified, including IWS1, an RNA processing regulator, which appeared as an Akt1- and Akt3-preferred substrate. Although this analysis minimized the possibility of isoform compensation- a phenomenon described previously (9), it is restricted to basal AKT isoform activities in a single cell-type context (fibroblasts), and moreover, there was no analysis of non-canonical substrate motifs. Cenni et al. identified Ankrd2/ARPP as a novel AKT substrate based on combining proteomics analysis of peptides captured using beads coated with antibodies that recognize the canonical phosphorylated RxxS/T motif (84). Using AKT isoform-specific siRNAs as well as in vitro kinase assays with immunoprecipitated HA-tagged AKT isoforms, they showed that Ankrd2 is an AKT2-preferred substrate. Likewise, the insulin-induced phosphorylation of MYO5A could be abrogated by siRNA to AKT2 but not to AKT1 (85). Nonetheless, the vast majority of the 167 AKT substrates listed on the Cell Signaling site (https://www.cellsignal.com/contents/resources-reference-tables/pi3k-akt-substrates-table/science-tables-akt-substrate) are listed as AKT1-specific, with only seven listed as AKT2-preferred (CD34, EZRIN, GAPDH, MYO5A, QIK/SIK2, SFRS5 and STXBP4) and none as preferred by AKT3 alone. Indeed, the true AKT isoform specificity of this list needs reevaluation: many of the AKT1- or AKT2-specific substrate designations may be due to the fact that these were the only kinases tested, or because “total” AKT activity was attributed to AKT1, or because in a given cell context, only one isoform is predominantly expressed. An example of the latter is EZRIN, which was identified as an AKT2-specific substrate in Caco-2 cells, even though these cells expressed little AKT1 or AKT3 (86). This suggests that some previous identifications of isoforms-specific substrates might actually reflect cell type-preferential AKT isoform expression.

A continuing confounding issue is the contribution of non-AKT kinases (e.g., PKA or RSK) in the phosphorylation of a putative AKT isoform-specific substrate under specific biological conditions. For example, based on the assumed role for AKT2 in insulin-mediated inhibition of gluconeogenesis (87,88), Dentin et al. showed that purified AKT2 could phosphorylate SIK2/QIK in vitro at a canonical AKT substrate site, Ser358, that was shown to be induced by insulin (87,88). However, subsequent work showed that whereas insulin induced Akt2 activation in mouse liver cells, it could not induce Sik2^poS358. In contrast, Sik2^poS358 could be induced by PKA activated by glucagon or a cell-permeable cAMP analogue (Bt₂-cAMP), even in Akt2^−/− hepatocytes or Akt2^+/+ hepatocytes treated with the pan-AKT inhibitor, MK-2206 (89). Thus, although it is listed as such, there remains no direct evidence that SIK2 is a bona fide AKT2 substrate in cells.

It is important to note at this point that while the focus is on canonical AKT phosphorylation as the mechanism for determining substrate preference, there are other factors that may contribute to this phenomenon. These factors, highlighted in the review by Hinz and Jücker, include post-translational modification, miRNA regulation, extracellular activation, and others (51). While these other factors may indeed contribute to some of the specificity, there is limited research into what roles these different cellular processes have on AKT isoform preferentiality.

Translating AKT Isoforms and Substrate Preferentiality to the Clinic

Given the growing evidence of AKT isoform-specific roles in cancer progression, the lack of clarity in reference to isoform-preferred substrates and pathways adds to the lack of understanding of how to therapeutically target the AKT axis. This is especially evident in PTEN-deficient cancers, which are assumed to be driven by increased AKT signaling. Consistent with our lack of full understanding of the oncogenic roles for PI3K and AKT, drugs targeting PI3K or AKT have performed poorly as monotherapies in solid tumors even though this axis is one of the most mutated in cancers (90,91). One pan-AKT inhibitor, Ipatasertib/RG7440, has been used in combination with endocrine and CDK4/6 targeted therapy in metastatic breast cancer (92). Indeed, this combination arose from evidence that resistance to chronic treatment of estrogen receptor-positive breast cancers with the CDK4/6 inhibitor, Palbociclib (Ibrance), caused a dependence on upregulated PI3K/AKT signaling as a driver of cyclin E/CDK2-mediated proliferation (93). In theory, targeting all three AKT isoforms with a pan-inhibitor such as Ipatasertib should override both isoform-specific dependence and compensatory isoform activity. However, even within the class of so-called pan-AKT inhibitors, there can be >10-fold selective inhibition of one AKT isoform over others (Table 1). As examples, using in vitro assays containing purified AKT1, AKT2 or AKT3 plus a commonly recognized GSKα 15-mer peptide substrate, Dumble et al. showed that GSK2142795 inhibits AKT2 8.6-fold or 4.7-fold more potently than AKT3 or AKT1, respectively; for GSK2110183, inhibition of AKT1 is 25-fold or 32.5-fold more potent than for AKT2 or AKT3, respectively (94). Moreover, even though cancer cell lines with either mutant PIK3CA or PTEN loss were relatively more sensitive to these two drugs than those expressing WT genes, the drugs levels to needed to establish 50% growth inhibition (GI₅₀) in 2D cultures were typically >100-fold higher than those needed to inhibit AKT isoform kinase activities. Even in cancer lines where mutant PIK3CA or PTEN loss was the only identified oncogenic driver, GI₅₀ values varied >50-fold. Thus, in addition to not understanding how this class of drugs is targeting AKT isoforms, there is a disconnect between an assumed dependency on oncogenic AKT signaling and drug potency. As an example of this disconnect, Li et al., show that whereas AKT1 knockdown was sufficient to inhibit the invasiveness of MCF-7 and MDA-MB-231 human breast cancer cells, treatment with the “pan-AKT” inhibitor MK2206 (not a clinical candidate inhibitor) induced increased experimental lung metastasis formation in vivo (95).

Table1:

Specificity of pan-AKT inhibitors for AKT isoforms

Drug	IC50 (nM)			reference
	AKT1	AKT2	AKT3
uprosertib/GSK2142795	180	328	38	Dumble et al., 2014
afuresertib/GSK2110183	0.08	2	2.6	Dumble et al., 2014
MK2206	5	12	65	Hirai et al., 2010
capivasertib/AZD5363	3	7	7	Davies et al., 2012
Ipatasertib/GDC-0068	5	18	8	Lin et al., 2011

The only AKT1- or AKT2-specific inhibitors described to date are A-674563 and CCT128930, respectively, both of which are ATP-competitive inhibitors, yet neither of which is in clinical trials. No AKT3-specific small drug inhibitors have been described. A-674563 (IC₅₀=11nM against AKT1) is also a potent PKA and CDK2 inhibitor (IC₅₀’s of 16nM and 46nM, respectively), whereas CCT128930 (IC₅₀=6nM against AKT2) requires 28-fold more drug (IC₅₀=168nM) to inhibit PKA (96). However, no data exist comparing the selectivity of these two drugs against purified AKT isoforms. Moreover, the mechanism-of-action of A-674563 in tumor cells remains unclear in that treatment of A427 lung carcinoma cells decreased the phosphorylation of some substrates assumed to be shared by all AKT isoforms, such as FOXO1, yet paradoxically increased (PRAS40) or had no effect (GSK3β) on the phosphorylation of other common substrates (97). A recent study describes AKT1-, 2- or 3-specific nanobodies, however, these have not been tested for their ability to block AKT isoform-specific activation in cells (98).

The needed development of improved AKT isoform-specific inhibitors and the elucidation of isoform-preferred substrates will undoubtedly allow for the ability to dissect the molecular pathways controlled by each AKT isoform during normal development or cancer progression. These data will also identify new, potentially targetable cancer biomarkers involved in specific progression parameters. Lastly, they will also help explain how possible compensatory effects by other AKT isoforms, upstream PI3K-p110 isoforms and other oncogenic mediators such as receptor and non-receptor tyrosine kinases regulate response to drugs targeting of the PI3K-AKT axis (59–61,99). Clarification of these important issues will help improve the therapeutic targeting of this critical signaling pathway in solid tumors.