Alterations in the PI3K/AKT pathway occur in up to 70% of melanomas and are associated with disease progression. The three AKT paralogs are highly conserved but data suggest they have distinct functions. Activating mutations of AKT1 and AKT3 occur in human melanoma but their role in melanoma formation and metastasis remains unclear. Using an established melanoma mouse model, we evaluated E17K, E40K, and Q79K mutations in AKT1, AKT2, and AKT3 and show that mice harboring tumors expressing AKT1E17K had the highest incidence of brain metastasis and lowest mean survival. Tumors expressing AKT1E17K displayed elevated levels of focal adhesion factors and enhanced phosphorylation of focal adhesion kinase (FAK). AKT1E17K expression in melanoma cells increased invasion and this was reduced by pharmacologic inhibition of either AKT or FAK. These data suggest that the different AKT paralogs have distinct roles in melanoma brain metastasis and that AKT and FAK may be promising therapeutic targets.
This study suggests that AKT1E17K promotes melanoma brain metastasis through activation of FAK and provides a rationale for the therapeutic targeting of AKT and/or FAK to reduce melanoma metastasis.
Cutaneous melanoma is the most lethal form of skin cancer due to its propensity to metastasize. The disease exhibits organotropism for specific sites including the lung and brain (1). The predilection of melanoma to preferentially seed and colonize select organs is particularly prominent in the form of brain metastasis. Melanoma has the highest propensity to metastasize to the brain among all adult malignancies (1) with up to 75% of stage IV patients with evidence of brain metastases at autopsy (2). However, the mechanisms that contribute to melanoma brain tropism are not well understood.
Hyperactivation of the PI3K/AKT signaling pathway occurs in most metastatic melanomas and increased PI3K/AKT pathway activity correlates with disease progression (3, 4). Loss or inactivation of the tumor suppressor PTEN is common in cutaneous melanoma and results in activation of the pathway leading to elevated phospho-AKT (P-AKT) levels (5, 6). Interestingly, melanoma brain metastases exhibit higher P-AKT than those of the liver or lung (6). Melanomas with loss of PTEN also have a shorter time to brain metastasis (5) and AKT overexpression results in a shift from radial to vertical tumor growth (7). Taken together, these data suggest that overstimulation of the PI3K/AKT pathway plays a key role in melanoma progression and may be particularly important in brain homing and/or colonization.
AKT1 is one of three closely related serine/threonine kinases (AKT1, AKT2, and AKT3). Although less common than PTEN loss, the “hotspot” mutation E17K in AKT1 and AKT3 (6, 8) also induces PI3K/AKT pathway hyperactivation (9) by disrupting the intramolecular interaction between the pleckstrin homology domain and the kinase domain (10). AKT1Q79K has also been detected in human melanoma and reported to not only result in constitutive activation of the enzyme but to also mediate resistance to mutant BRAF inhibition (11). To date, mutations that result in constitutive activation of AKT2 have not been observed in melanoma; however, one study reported that PTEN loss preferentially activates AKT2 over the other AKT paralogs suggesting that aberrant AKT2 signaling may also influence melanoma progression (12). This finding seems contradictory to a recent report demonstrating that AKT2 preferentially binds phosphatidylinositol-3,4-biphosphate (PI(3,4)P2), which forms from phosphatidylinositol-3,4,5-triphosphate (PIP3) by SHIP2 at the plasma membrane and delivered to early endosomes through clathrin-mediated endocytosis. Conversely, AKT1 and AKT3 preferentially bind PIP3, a phosphatidylinositol sequestered to the plasma membrane (13). These distinct fates of AKT paralogs may contribute to differential exposure to substrates. Studies aimed at defining the role of different AKT paralogs in cancer have provided some insight into paralog-specific functions in disease progression. Riggio and colleagues reported that while AKT1 overexpression promoted primary tumor growth in breast cancer–cell migration, invasion and metastasis were inhibited. Conversely, overexpression of AKT2 had the opposite effect (14). AKT3 has been reported to inhibit migration and metastasis in breast cancer (15), whereas in melanoma, AKT3 has been implicated in tumorigenic potential (16). These studies imply that the different AKT paralogs may have specific and context-dependent functions in cancer, including tumor-suppressive properties, and therefore it is critical that these functions are accurately characterized so effective therapeutic strategies targeting AKT or its effectors can be developed.
Using an established autochthonous mouse model of melanoma, we previously demonstrated that activation of AKT1 through the addition of an N-terminal myristoylation sequence (myrAKT1) reduced survival and increased the incidence of melanoma lung and brain metastasis in the context of BRAFV600E expression and Cdkn2a loss. Loss of Pten cooperated with myrAKT1 and further enhanced brain metastasis in this context (17). In this study, we evaluated the ability of constitutively active E17K, E40K, or Q79K mutants of each AKT paralog to promote tumor progression and metastasis in the context of BRAFV600E expression and loss of Cdkn2a and Pten. Expression of AKT1E17K promoted highly aggressive melanomas that metastasized to the lungs and brain. This metastatic phenotype was not significantly observed in the case of other mutant AKT-positive tumors, suggesting that the AKT paralogs have distinct, nonoverlapping roles in the development of melanoma brain metastases. AKT1E17K-positive tumors showed AKT1E17K-dependent upregulation of multiple focal adhesion (FA) factors, which are key components of focal adhesions and established stimulators of cell motility, as well as phosphorylation of focal adhesion kinase (FAK). Ectopic expression of AKT1E17K in nonmetastatic melanoma cells increased cell invasion, a phenotype abrogated by pharmacologic inhibition of AKT or FAK. These findings strongly suggest that one mechanism by which AKT1 promotes melanoma lung and brain metastasis is through regulation and activation of proteins involved in focal adhesions. This has important implications for the development of therapeutic strategies aimed at preventing or treating disseminated disease.
Materials and Methods
Mice and genotyping
Dct::TVA;BrafCA;Cdkn2alox/lox;Ptenlox/lox mice have been described previously (17). The breeding scheme and genotype procedures are described in Supplementary Materials.
Viral constructs and propagation
The avian retroviral vectors used in this study are replication-competent Avian Leukosis Virus splice acceptor and Bryan polymerase-containing vectors of envelope subgroup A [designated RCASBP(A) and abbreviated RCAS]. Cloning details for RCAS AKT–mutant constructs and the method of viral propagation are described in Supplementary Materials.
The culture conditions for DF-1, YUMM 1.1, and primary tumor cell lines are described in Supplementary Materials. All primary tumor cell lines were derived from Dct::TVA mice.
Viral infections (in vivo and in vitro)
Detailed experimental procedures describing the injection of Dct::TVA mice, as well as the generation YUMM1.1 isogenic variants, and primary tumor cell line variants used in functional experiments are described in Supplementary Materials.
Histology and histochemical staining
Mice were euthanized at their experimental endpoints and subjected to a full necropsy. Brain, lung, and primary tumor tissues were fixed in formalin overnight, dehydrated in 70% ethyl alcohol, and paraffin embedded. Sections were stained with H&E or left unstained for IHC. Expression of mutant AKT in each primary tumor of the mutant AKT cohorts was confirmed by IHC for the HA epitope tag. Mice whose tumors were absent detectable HA were excluded from all analyses. A separate portion of each primary tumor was frozen for use in immunoblot analyses, reverse phase protein array (RPPA), and RNA sequencing.
Detailed experimental procedures and GEO accession numbers for the RPPA and RNA sequencing as well as TCGA data analysis are described in Supplementary Materials.
Spheroid formation, transwell invasion, and transwell migration assays
All three-dimensional cell growth/motility assays were performed using YUMM 1.1 cells. Detailed experimental procedures, including the pharmacologic inhibitors used in the transwell invasion study, are provided in Supplementary Materials.
Tail vein and intracranial injections
NOD scid gamma (NSG) mice were injected via the tail vain or intracranially with primary mouse melanoma cells. Additional experimental details are provided in Supplementary Materials.
IHC, immunoblotting, and immunofluorescence
Detailed experimental procedures for all antibody-based assays are described in Supplementary Materials.
Analytic methods used to determine significant differences between groups with respect to mouse survival, incidence of metastasis, immunoblots, spheroid formation, RPPA, and RNA sequencing are described in Supplementary Materials. The asterisks shown in figures correspond to P values as follows: *, P < 0.05; **, P < 0.01; ***, P < 0.001.
All animal experimentation was performed in Association for Assessment and Accreditation of Laboratory Animal Care International–approved facilities at the University of Utah (Salt Lake City, UT). All animal protocols were reviewed and approved prior to experimentation by the Institutional Animal Care and Use Committee at the University of Utah (Salt Lake City, UT).
For more details, refer to web version of PubMed Central for Supplementary Material.
AKT1E17K promotes melanoma metastasis in vivo
To evaluate the effect of distinct activating mutations in AKT on melanoma growth and metastasis, we used an established autochthonous mouse model of melanoma based on the RCAS/TVA avian retroviral system (17). In this model, transgenic mice express the TVA viral receptor under the control of the dopachrome tautomerase (DCT) promoter, which allows targeting of the virus, and expression of the genes it contains, specifically in melanocytes. Dct::TVA mice were previously crossed to BrafCA, Cdkn2alox/lox, and Ptenlox/lox mice to model the most common alterations in cutaneous melanoma (17). Delivery of RCAS-Cre results in expression of BRAFV600E and genetic loss of Cdkn2a and Pten. Activating mutants (E17K, E40K, or Q79K) of AKT1, AKT2, and AKT3 were created using site-directed mutagenesis, engineered to contain an N-terminal HA epitope tag, and cloned into the RCAS retroviral vector. Prior to in vivo studies, all mutant AKT viruses were shown to infect TVA-positive cells and express the mutant AKT protein in vitro (Supplementary Fig. S1). For in vivo analysis, newborn Dct::TVA;BrafCA;Cdkn2alox/lox;Ptenlox/lox mice were subcutaneously injected with viruses encoding mutant AKT1, AKT2, or AKT3 alone or in combination with RCAS-Cre. All mice injected with viruses containing mutant AKT alone, in the absence of RCAS-Cre, remained tumor-free for the duration of the study (150 days, n = 59; Fig. 1). Conversely, all mice injected with viruses encoding Cre, alone or in combination with mutant AKT, developed primary tumors at the site of injection. Delivery of RCAS-AKT1E17K or RCAS-AKT1E40K in combination with RCAS-Cre significantly reduced survival compared with delivery of RCAS-Cre alone (P = 0.0063 and 0.0004, respectively; Fig. 1A; Supplementary Table S1). Likewise, delivery of viruses containing AKT2E40K, AKT3E17K, or AKT3Q79K in combination with RCAS-Cre significantly reduced survival compared with delivery of RCAS-Cre alone (P = 0.0025, 0.0026, and 0.0058, respectively; Fig. 1B and C; Supplementary Table S1). Interestingly, delivery of viruses containing AKT1Q79K, AKT2E17K, AKT2Q79K, or AKT3E40K in combination with RCAS-Cre did not significantly reduce survival compared with delivery of RCAS-Cre alone (Fig. 1A–C; Supplementary Table S1).
To assess metastasis, full necropsies were performed and all major organs were examined histologically. Brain metastases were detected in 38% (8/21) of mice injected with viruses encoding AKT1E17K and Cre (AKT1E17K cohort: BRAFV600E;Cdkn2a−/−;Pten−/−;AKT1E17K), whereas no brain metastases were detected in mice injected with viruses containing Cre alone (control cohort: BRAFV600E;Cdkn2a−/−;Pten−/−; n = 24; P = 0.0009; Fig. 1D). Lung metastases were detected in 24% (5/21) of mice in the AKT1E17K cohort but this was not significantly different from the control cohort (P = 0.225). No significant difference in the incidence of brain or lung metastases was observed between all other experimental cohorts and the controls (Supplementary Table S1). These data suggest that AKT1E17K confers an enhanced capacity to promote brain metastasis compared with the other AKT mutants tested.
AKT1 kinase activity is required for induction of melanoma metastasis in vivo
Propagation of AKT signaling is accomplished through kinase-dependent phosphorylation of downstream effectors that drive cell survival, growth, and motility. However, AKT1 has been reported to promote cell survival in a kinase-independent manner (18). To assess the requirement for kinase activity of AKT1E17K for tumor growth and brain metastasis in our model, we cloned a validated kinase-dead version of the E17K mutant (AKT1E17K, K179M) into the RCAS vector (18, 19). Mice injected with viruses encoding AKT1E17K, K179M and Cre showed no significant difference in survival (Supplementary Fig. S2) or incidence of metastasis (Supplementary Table S1) compared with the Cre-only control cohort. This was also true for the wild-type AKT1 and Cre cohort compared with the Cre-only control cohort. These data demonstrate that the reduced survival and enhanced brain metastases observed require AKT1 kinase activity, which is activated by the E17K substitution.
Primary and metastatic melanomas are histologically similar to the human disease
Primary tumors generated in Dct::TVA;BrafCA;Cdkn2alox/lox;Ptenlox/lox mice were examined histologically following hematoxylin and eosin (H&E) staining by board-certified pathologists (A.H. Grossmann and K.L. Duffy; Fig. 2A). Tumors displayed characteristics consistent with malignancy, including mitotic figures and high nuclear grade with prominent nucleoli. They consisted primarily of short spindle cells but epithelioid cells were variably observed in addition to nonbrisk inflammation (tumor-infiltrating lymphocytes), coagulative tumor necrosis, and intratumoral hemorrhage. Analysis of tumor sections by IHC revealed heterogeneous detection of mutant AKT (HA epitope tag; Fig. 2B), the presence of proliferation marker Ki67 (Fig. 2C), P-AKT (Fig. 2D), P-ERK (Fig. 2E), and the melanoma marker S100 (Fig. 2F).
Microscopic examination of the lungs revealed that a subset of mice had developed metastatic lesions; these tumors exhibited a malignant morphology consistent with those of the primary tumors (Fig. 2G and H). Lung lesions were evaluated by IHC and exhibited loss of PTEN expression (Fig. 2I) as well as detectable P-AKT (Fig. 2J), and P-ERK (Fig. 2K). Likewise, and similar to the primary tumors, lung metastases exhibited positivity for S100 confirming the melanocytic and metastatic origin of these lesions (Fig. 2L). Similar analyses were performed on brain tissue from injected mice to identify those that had developed brain metastases and to characterize the histopathologic features of these lesions (Fig. 2M and N). These analyses confirmed the melanocytic and metastatic origin of these tumors as well as loss of PTEN and positivity for P-AKT, and P-ERK (Fig. 2O–R).
Expression and activation of focal adhesion factors are altered in melanomas expressing AKT1E17K
The AKT1E17K cohort exhibited the lowest mean survival and highest incidence of brain metastasis of all mutant AKT cohorts tested. To identify key effectors essential for AKT1E17K-dependent metastasis, we performed RPPA on five primary tumors samples from each cohort. For RPPA, protein samples were probed for 245 different epitopes and 15% (36/245) were significantly different (P ≤ 0.05) between AKT1E17K and control tumors (Fig. 3A). Several PI3K/AKT pathway–related proteins were differentially regulated in the AKT1E17K–positive tumors compared with controls (Supplementary Fig. S3A). Of these proteins, c-KIT (downregulated; P = 0.0255), and p70-S6K1 (downregulated; P = 0.0346) were found to exhibit consistent patterns of dysregulation in patient brain-metastatic melanomas compared with non-brain metastatic melanomas (TCGA data; ref. 20). However, downregulation of c-KIT and p70-S6K1 would not be predicted to enhance tumor progression. Interestingly, we found that paxillin, a FA factor known for the ability to stimulate cell motility, was upregulated in both AKT1E17K mouse tumors compared with controls (P = 0.0060), as well as patient brain-metastatic melanomas compared with non-brain metastatic melanomas (P = 0.0138). This finding revealed a potential AKT1 downstream effector that contributes to brain metastasis. Likewise, a second FA factor, phospho-focal adhesion kinase (Y397; P-FAK) was also upregulated in AKT1E17K mouse tumors compared with controls (P = 0.0016); however, P-FAK was not an included epitope in the TCGA RPPA panel. The Y397 phosphorylation event in FAK acts as a molecular switch that promotes focal adhesion turnover and ultimately drives cell motility (21). Paxillin is one of several FA scaffolding proteins recruited to nascent FAs by P-FAK and is a necessary component of the FA turnover process (22). RPPA analysis of tumors from AKT2E17K or AKT3E17K cohorts revealed no differential regulation of P-FAK or paxillin compared with controls (Supplementary Fig. S3B and S3C, respectively).
To validate these findings, protein levels of P-FAK, total FAK, phospho-paxillin (P-paxillin), and total paxillin were measured in five AKT1E17K and five control primary tumors via immunoblot (Fig. 3B). Consistent with the RPPA data, a significant increase in the relative levels of P-FAK and total paxillin was observed in the AKT1E17K tumors compared with control tumors (P = 0.0004 and 0.0030, respectively). P-paxillin (Y31), which was not included in the RPPA analysis, similarly exhibited an increase in AKT1E17K tumors compared with controls (P = 0.0130; Fig. 3C). This is consistent with P-FAK–mediated phosphorylation of paxillin on Y31. A significant increase in the relative levels of P-AKT (T308; P = 0.0227) and (S473; P < 0.0001) was also observed in the AKT1E17K tumors compared with controls (Fig. 3C). IHC analysis of P-FAK, paxillin, and P-paxillin in primary tumors was consistent with the immunoblot results (Supplementary Fig. S4). These data suggest that the AKT1E17K mutant potentiates AKT signaling beyond that of Pten loss alone.
FAK and paxillin have been widely reported to promote cancer cell motility and have been implicated in tumor progression and metastasis (23), including in melanoma where FAK has been shown to have a role in proliferation, motility, and invasion (24, 25). FAK and paxillin are critical for FA disassembly (FA turnover), which involves a FAK–Src–paxillin complex, although the precise mechanism(s) are unclear (26). Consistent with these reports, our data suggest that the increase in P-FAK, paxillin, and P-paxillin by AKT1E17K may contribute to tumor cell spreading. Because only a small fraction of the RPPA platform was dedicated to FA factors, we set out to determine whether additional FA proteins were differentially expressed between AKT1E17K and control tumors. RNA was isolated from the same primary tumors used in the RPPA analysis and subjected to RNA sequencing. Of the 10 primary tumors sequenced, one AKT1E17K tumor did not pass quality control procedures and thus was excluded from further examination. Assessment of global differential gene expression by principal component analysis and unsupervised hierarchical clustering of gene expression profiles revealed that samples within each cohort clustered together appropriately (Supplementary Fig. S5A). The majority of the most highly differentially expressed genes were downregulated in AKT1E17K tumors compared with controls (Supplementary Fig. S5B). In contrast, out of 18 different FA genes that have reported roles in FA assembly and/or disassembly (21), 16 of these genes were upregulated in AKT1E17K tumors compared with controls (Fig. 4A and B). These data strongly suggest that AKTE17K has a critical role in upregulating numerous FA factors.
Expression of AKT1E17K alters integrin expression
Expression patterns of alpha and beta integrin genes were also assessed between AKT1E17K and control tumors. Integrin proteins link the actin cytoskeleton to the extracellular matrix (ECM) by forming transmembrane heterodimers. The proteins not only provide an anchor to the ECM required for cell motility during myosin II contraction of actin filaments, but channel mechanical information from the ECM through intracellular FAs to modulate cell responses to the external environment. Each cohort retained similar integrin gene expression profiles within their respective groups (Supplementary Fig. S6A), but 14 of 27 integrins were differentially expressed between the AKT1E17K cohort compared with the controls with 6 genes significantly upregulated (Itgav, Itga5, Itga8, Itgbl1, Itgb1, and Itgb5) and 8 genes significantly downregulated (Itgad, Itgae, Itgal, Itgax, Itga2b, Itga10, Itga11, and Itgb6) in the AKT1E17K tumors (Supplementary Fig. S6B). These data indicate that AKTE17K also influences the expression of integrins.
AKT1E17K promotes melanoma cell migration and invasion
To test whether expression of AKT1E17K promotes cell motility and/or proliferation in vitro, we utilized Yale University Mouse Melanoma (YUMM) cell line 1.1, which expresses BRAFV600E and is deficient for Cdkn2a and Pten (27). These molecular alterations mirror the tumors within our control cohort. The YUMM 1.1 cells were used to generate isogenic cell lines that stably express either wild-type AKT1 or AKT1E17K (Fig. 5A). Immunofluorescence followed by total internal reflection fluorescence (TIRF) and confocal microscopy was used to further visualize AKT1E17K and FAs in these cells. Distinct paxillin puncta, indicative of FA formation, were located primarily at the cell periphery in both the parental and AKT1E17K cells; in some cases, AKT1E17K colocalized with paxillin (Fig. 5B).
Cell proliferation assays were performed at multiple time points over 5 days in the presence or absence of growth factors. With the exception of an increase in proliferation for cells expressing wild-type AKT1 compared with the parental cells on day 5 (Supplementary Fig. S7A), no differences were observed between conditions (Supplementary Fig. S7B). These data suggest that AKT1E17K does not confer a proliferative advantage compared with controls in this context. When the same cells were plated on Matrigel, an increase in the ability of AKT1E17K and wild-type AKT1-expressing cells to form melanoma spheroids compared with the parental cells was observed (P = 0.0041 and 0.0038, respectively; Fig. 5C). However, because wild-type AKT1 expression does not confer a metastatic phenotype in vivo (Supplementary Table S1), the increased capacity of AKT1-expressing cells to form spheroids in this context does not translate to the prometastatic phenotype seen in vivo by AKT1E17K tumors. This distinction could potentially be explained by differences in invasive capacity. Indeed, we observed almost a 5-fold increase in the number of AKT1E17K-expressing spheroids that invaded the ECM compared with the parental cells (P = 8 × 10−15), but less than a 3-fold increase in invasion with cells expressing wild-type AKT1 compared with the parental cells (P = 3 × 10−5; Fig. 5D and E). These data suggest that AKT1 signaling increases the ability of melanoma cells to invade the ECM and that the oncogenic mutation of AKT1 (E17K) enhances this effect.
To determine whether AKT1E17K promotes cell migration, we performed a transwell migration assay using the isogenic YUMM 1.1 cell lines. AKT1E17K-expressing cells demonstrated a substantial increase in their ability to migrate through transwell pores toward a chemoattractant compared with both the parental cells (P = 0.0006) and cells expressing wild-type AKT1 (P = 0.0046; Fig. 6A and B). These data demonstrate that unlike wild-type AKT1, AKT1E17K promotes migration in this context.
Inhibition of AKT or FAK reduces AKT1E17K-mediated invasion
We next sought to determine whether pharmacologic inhibition of AKT or FAK interferes with cell invasion. To address this question, parental, AKT1, and AKT1E17K-expressing isogenic YUMM 1.1 cells were plated into Matrigel-coated transwell chambers and treated with vehicle alone, 1,000 nmol/L GSK-2141795 (AKT inhibitor), 500 nmol/L PF-573228 (FAK inhibitor), or 50 nmol/L dasatinib (Src inhibitor). Src acts at the intersection of the FAK-initiated FA assembly and disassembly by forming a complex with FAK and phosphorylating several key FAK tyrosine residues, one of which (Y925) stimulates FA turnover (21). Each compound was found to reduce levels of P-FAK (Y397 and Y925) in culture (Fig. 6C). Expression of AKT1E17K significantly enhanced the ability of cells to invade compared with both the parental (P = 0.0010) and AKT1 isogenic cells (P = 0.0186; Fig. 6D and E), consistent with our observations of spheroid invasion and transwell migration. Inhibition of either AKT or FAK significantly reduced the invasion of AKT1E17K-expressing cells (P = 0.0200 and 0.0205, respectively) compared with vehicle-only–treated AKT1E17K-expressing cells (Fig. 6D and E). Although inhibition of Src trended toward decreased invasive capacity of AKT1E17K-expressing cells, this difference was not significant compared with vehicle-only–treated AKT1E17K-expressing cells (P = 0.0847). Collectively, these data indicate that AKT1E17K promotes invasion of mouse melanoma cells in a FAK-dependent manner. Importantly, the invasive phenotype of oncogenic AKT1 can be impaired by pharmacologic inhibition of either AKT or FAK.
Either PTEN loss or AKT1E17K expression is sufficient for brain colonization
While enhanced invasive properties allow the melanoma cells to exit the primary tumor site, it is unclear whether activation of the PI3K/AKT pathway is sufficient to allow circulating tumor cells to colonize distal organs, including the brain, and promote tumor growth in these distant sites. To test this, we generated three primary tumor cell lines with variations in PI3K/AKT activity from Dct::TVA mice: 5610 (BRAFV600E;Cdkn2a−/−), 9678 (BRAFV600E;Cdkn2a−/−;Pten−/−), and 7788 (BRAFV600E;Cdkn2a−/−;Pten−/−;AKT1E17K; Fig. 7A and B). Each of the three cell lines were injected into the tail vein of NOD scid gamma (NSG) mice to assess the ability of each cell line to colonize lungs. All mice, regardless of the cell line injected, developed detectable lung lesions (Fig. 7C). To assess the ability to colonize the brain, each of the three cell lines were injected intracranially into NSG mice. Mice injected with 5610 remained healthy for the duration of the study, whereas all mice injected with either 9678 or 7788 demonstrated signs of illness and were euthanized prior to the experimental endpoint. The brains of all mice were examined post-necropsy, sectioned, and stained with H&E in an effort to detect melanoma growth. All mice injected with 9678 and 7788 were found to have detectable brain lesions, whereas no mice injected with 5610 had detectable lesions (Fig. 7D); these differences were statistically significant (P = 0.0119 and 0.0048, respectively; Fig. 7E). Expression of AKT1E17K in the 5610 cell line (5610E17K; Fig. 7A and B) and intracranial injection of this cell line resulted in illness and early euthanasia for all mice. In addition, all mice injected with 5610E17K were found to have detectable brain lesions (Fig. 7D and E) and this was statistically significant when compared with mice injected with the 5610 parental line (P = 0.0048). These data indicate that while aberrant PI3K/AKT pathway signaling does not appear to impact the ability of tumor cells to promote lung colonization, it does promote melanoma brain colonization either through Pten loss, expression of AKT1E17K, or both.
We performed an objective and comprehensive in vivo analysis of three separate AKT-activating mutations (E17K, E40K, or Q79K) in each of the three AKT paralogs on melanoma progression using an established autochthonous mouse model of melanoma. Each of the AKT mutants used in this study engender constitutive activation of the protein (9, 10, 19, 28). The mechanism of AKT constitutive activation was elegantly described by Carpten and colleagues in 2007 using the E17K mutant, whereby the lysine substitution of glutamic acid at amino acid 17 was shown through X-ray crystallography to disrupt the intramolecular interactions that occur between the PH domain and kinase domain. This perturbation forces a conformational change in the protein resulting in increased affinity of the PH domain for phosphatidylinositols at the plasma membrane where the protein is phosphorylated (10). Since this time, dozens of amino acid substitutions in the PH domain of AKT that are predicted to interfere with the PH-Kinase domain interface have been shown to promote kinase hyperactivity, including those found in cancers (9).
To measure the impact of AKT-activating mutations on melanoma, we used overall survival and the incidence of metastasis as readouts of disease progression. We discovered a mutant-dependent effect on survival for all three AKT paralogs. Although AKT2 and AKT3-mutant tumors reduced the survival of mice compared with BRAFV600E;Cdkn2a−/−;Pten−/− control tumors in some cases, they did not significantly promote lung and brain metastasis. However, an analysis of AKT1 mutants in the same context revealed that AKT1E17K-positive tumors not only reduced the survival of mice compared with controls, but also increased the incidence of brain metastasis to nearly 40% (Fig. 1D). This discovery is particularly noteworthy as AKT1 was reported in one study to be “the most important hub gene” as determined by the PageRank algorithm used to identify top hub genes in melanoma (29), and to exhibit the “second highest centrality score” in melanoma as determined by an algorithm that combines PageRank with closeness and betweenness (30). Furthermore, the E17K substitution in AKT1 is the most common mutation found among all AKT paralogs in human cancer, representing over one-third of AKT1 mutations in cBioPortal (31). Specific to melanoma, cBioPortal indicates that approximately 3% of these tumors have missense substitutions in AKT1. Of those, the most common is the hotspot mutation AKT1E17K, which represents 17% of these mutations. Recent evidence also suggests that this alteration may be a more common occurrence in advanced melanomas that are resistant to MAPK inhibition. Shi and colleagues detected AKT1E17K in a melanoma that had acquired resistance to BRAFi (vemurafenib), but this mutation was not detected in the pre-treated tumor (11).
The differences in tumorigenic potential among E17K, E40K, and Q79K variants of each AKT paralog is likely due to the unique molecular features imparted by each amino acid substitution. Indeed, Parikh and colleagues reported that AKT1 wild-type, E17K, and Q79K variants possessed varying magnitudes of intramolecular interactions between the PH and kinase domains as well as contrasting levels of phospho-AKT (S473; ref. 9). These differences may, in turn, produce a range of conformational modifications that could ultimately influence activity, including PH domain affinity for phosphatidylinositols, intrinsic kinase activity, differential activation of downstream substrates, as well as the capacity to confer differential exposure to substrates as a result of changes in subcellular localization. As such, any potential functional modification between variants of a single AKT paralog would likely translate to corresponding paralogs, for which unique target substrates have already been identified.
To develop mechanistic insight into the molecular and cell biological features that cause AKT1E17K-positive tumors to metastasize to the brain, we used RPPA to screen for differentially regulated proteins and phospho-proteins in control versus AKT1E17K-positive primary tumors. FA factors P-FAK and paxillin were found to be among the most highly upregulated factors in AKT1E17K-positive tumors compared with controls. Functionally, FAK and paxillin are intimately involved in stimulating FA turnover processes that promote cell motility and have been implicated in the metastasis of numerous malignancies through their phosphorylation and activation (23). RNA sequencing analyses of the same primary tumors revealed that the mRNA levels of the vast majority of FA genes we examined were upregulated in AKT1E17K-positive tumors compared with controls. Among all FA factors examined, several have been implicated in driving melanoma progression, including Dnm2 (32), Pin1 (33, 34), Bcar1 (24), Actn4 (35), and Itgav (36).
These results led us to hypothesize that the melanoma brain metastases that emerge in our in vivo model may be at least partially explained by an increase in tumor cell invasiveness during an early step in the metastatic cascade when efficient assembly and disassembly of FA structures are critical determinants of this process. To test this hypothesis, we first utilized YUMM 1.1 (BRAFV600E;Cdkn2a−/−;Pten−/−) mouse melanoma cells that stably expressed AKT1E17K or wild-type AKT1. AKT1E17K-expressing cells displayed enhanced invasion of melanoma spheroids, increased migration, and increased invasion through Matrigel compared with the parental and isogenic cells expressing wild-type AKT1. Second, we performed in vivo studies using primary tumor cell lines derived from Dct::TVA mice. Both BRAFV600E;Cdkn2a−/−;Pten−/− and BRAFV600E;Cdkn2a−/−;Pten−/−;AKT1E17K melanoma readily colonized the lungs of mice when introduced directly into the circulation. This suggests that the failure of BRAFV600E;Cdkn2a−/−;Pten−/− melanoma to metastasize is likely the result of a diminished capacity to enter the circulation.
The FA assembly and disassembly processes that stimulate cell motility in response to external cues are well coordinated but highly complex. Although many FA proteins have been identified, the precise interactions that occur between FA factors and the consequences of those interactions on cell behavior are not completely understood and often context-dependent. To date, nearly 100 FA factors had been identified (37), many of which remain largely uncharacterized. What remains apparent is that activation of FAK is an early and key event that orchestrates FA turnover and cell motility. We previously reported cooperation between Pten loss and activation of AKT1 in the formation of melanoma lung and brain metastases but the mechanism of cooperation was not clear (17). Interestingly, PTEN has been shown to coprecipitate with FAK and paxillin in colorectal cancer cells, and more importantly, to dephosphorylate Y397 on FAK in glioblastoma and breast cancer cells. This dephosphorylation event diminishes the capacity of P-Src to bind and phosphorylate additional tyrosine residues on FAK, resulting in a reduced ability of the cells to form FAs (38, 39). Furthermore, P-FAK has been found to phosphorylate PTEN on Y336, an event that enhances PTEN phosphatase activity in an apparent negative feedback loop that prevents an overabundance of intracellular P-FAK and results in repression of cell motility (38, 40). In our melanomas, deficiency of Pten was not sufficient to induce brain metastases despite a presumed increase in FAK activity compared with wild-type Pten melanomas (BRAFV600E;Cdkn2a−/−). This suggests that further activation of FAK by AKT1 (beyond that of Pten loss alone) is required to promote FA turnover, increase tumor invasion, and ultimately elicit brain metastasis.
Several lines of evidence support the hypothesis that AKT1E17K promotes brain metastasis through activation of FAK. AKT1E17K-positive primary tumors express considerably higher levels of P-AKT than the Pten−/− tumors and this increase in P-AKT correlates with increased P-FAK (Fig. 3C). Although conventionally FAK has been considered an upstream activator of AKT signaling (41), more recently it has been shown to be a direct target of AKT1. In mouse embryonic fibroblasts, AKT1 phosphorylates S695 and T700 on FAK (42) and in human colon cancer cells AKT1 phosphorylates S517, S601, and S695 on FAK (43, 44). These phosphorylation events precede and stimulate FAK autophosphorylation of Y397. These studies provide evidence of a direct link between AKT1 and FAK and assert that FAK activation in this context is not simply an associative event, which co-occurs with P-AKT, but rather the direct result of AKT1-mediated phosphorylation. In addition, pharmacologic inhibition of FAK or AKT in AKT1E17K-positive mouse melanomas not only diminishes the levels of P-FAK (Fig. 6C), but also decreases cell invasion through Matrigel (Fig. 6D). Furthermore, while our findings imply FAK-mediated tumor cell invasion, an independent study suggests that FAK also induces melanoma transendothelial migration. Jouve and colleagues demonstrated that pharmacologic inhibition of FAK reduced transmigration of B16 mouse melanoma cells across a monolayer of lung endothelial cells (45), a finding that is relevant for both intravasation and extravasation. Collectively, these data point to a model consistent with a threshold effect. Pten loss alone does not promote brain metastases; however, it is sufficient for intracranial growth of melanoma cells following direct injection into the brain (Fig. 7D and E). The combination of AKT1E17K expression coupled with Pten loss yields persistently elevated levels of P-FAK. An abundance of P-FAK in turn promotes an increase in cell motility and the ability to transmigrate across endothelial barriers thereby allowing the melanoma cells to more effectively spread to distant sites. Although the cell invasion data using FAKi suggests a causal link between AKT1 and FAK in disease progression, further experiments using pharmacologic or genetic inhibition of FAK in vivo are required to demonstrate that FAK activation drives melanoma brain metastasis in this context.
One obvious question pertains to the physiologic relevancy of AKT1E17K mutations coupled with PTEN loss in human melanoma. AKT1E17K occurs at a low frequency but spans over a large range of cancers, including melanoma (9, 31), whereas a loss or deficiency in PTEN is particularly common in melanoma (4, 5). To our knowledge, a comprehensive analysis evaluating the degree of mutual exclusivity of AKT mutations coupled with PTEN loss in melanoma has not been performed. One study designed to partially address this question was performed using colorectal cancer tissue whereby 2,631 cases of cancer were examined for the presence of AKT1E17K. Of colorectal carcinomas found to harbor this mutation, 33%–38% were estimated to be accompanied by PTEN loss (46). Future studies designed to answer this question in melanoma will be important to establish the physiological relevancy of this cooperation.
While the restoration of tumor suppressor genes such as PTEN remains a colossal therapeutic challenge in cancer biology, numerous compounds designed to molecularly target and inhibit kinases such as AKT and FAK have been developed. Many AKT inhibitors have already demonstrated remarkable specificity and efficacy, with low toxicity (47), and although less well-studied clinically, FAK inhibitors have also shown promise as effective suppressors of FAK activity (23, 48). Because the three AKT paralogs are capable of inducing opposing effects on tumor progression, as has been demonstrated in breast cancer (14), the option to pharmacologically target AKT for patients with melanomas that exhibit hyperactivated PI3K/AKT signaling may require careful consideration. However, because none of the paralog-specific AKT-mutant tumors from our in vivo studies offered a protective effect compared with controls, our results suggest that indiscriminant inhibition of all AKT paralogs using a pan-inhibitor would not be detrimental to melanoma therapy. Whether pan-inhibition of AKT in this context offers a survival benefit remains to be determined.
Historically, the use of AKT inhibitors as monotherapies in clinical trials have either failed or produced modest anti-tumor effects (49). Current clinical trials using AKT or FAK inhibitors as anticancer agents primarily focus on combinatorial treatment regimens. Hirata and colleagues found that while BRAF-mutant patient-derived xenograft melanomas failed FAK inhibitor monotherapy, the combination of BRAF and FAK inhibition led to a prolonged period of tumor control–a phenomenon that persisted even after tumors developed resistance to mutant BRAF inhibitor monotherapy. The mechanism of resistance to single-agent BRAF inhibition in this study was attributed to BRAF inhibitor–dependent melanoma-associated fibroblast-mediated remodeling of the ECM. Interestingly, this event was shown to stimulate FAK signaling in melanoma cells, which in turn reactivated the MAPK signaling pathway (50). Our findings in conjunction with studies such as these highlight the role of FAK as an important proto-oncogene in melanoma, and one that may be particularly relevant to BRAF-mutated melanoma. Further studies will determine whether inhibition of AKT and/or FAK in combination with other pathway-targeted agents is a rational therapeutic strategy in this disease.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: D.A. Kircher, M.R. Silvis, G.M. Fischer, S.L. Holmen
Development of methodology: D.A. Kircher, M.R. Silvis, G.M. Fischer, M.C. Mendoza, M.W. VanBrocklin, S.L. Holmen
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D.A. Kircher, K.A. Trombetti, M.R. Silvis, G.L. Parkman, G.M. Fischer, C.M. Stehn, S.C. Strain, A.H. Grossmann, K.L. Duffy, S.L. Holmen
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.A. Kircher, K.A. Trombetti, M.R. Silvis, G.L. Parkman, G.M. Fischer, C.M. Stehn, K.M. Boucher, M. McMahon, M.A. Davies, M.C. Mendoza, S.L. Holmen
Writing, review, and/or revision of the manuscript: D.A. Kircher, M.R. Silvis, G.L. Parkman, G.M. Fischer, A.H. Grossmann, M. McMahon, M.A. Davies, M.C. Mendoza, S.L. Holmen
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.A. Kircher, K.A. Trombetti, G.M. Fischer, S.N. Angel, S.C. Strain, M.W. VanBrocklin, S.L. Holmen
Study supervision: D.A. Kircher, S.L. Holmen
We thank members of the Davies, VanBrocklin, McMahon, Mendoza, and Holmen labs as well as W. Pavan, M. Bosenberg, A.J. Lazar, C. Stubben, C. Conley, A. Welm, B. Welm, and T. Oliver for providing mouse strains, reagents, vectors, and/or advice. We thank Rowan Arave for assistance with figures. We thank the Huntsman Cancer Institute Vivarium staff for assistance with mouse husbandry and the Preclinical Research Resource (PRR) for assistance with tumor cell injections. We acknowledge the use of the DNA Synthesis Core, the DNA Sequencing Core, the Fluorescence Microscopy Core, and the Flow Cytometry Core at the University of Utah. Microscopy equipment was obtained using a NCRR Shared Equipment Grant # 1S10RR024761-01. We also acknowledge the use of the HCI Shared Resources for Research Informatics (RI), Cancer Biostatistics (CB), High-Throughput Genomics and Bioinformatics Analysis (GBA), and the Biorepository and Molecular Pathology (BMP) Research Histology Section supported by P30CA042014 awarded to HCI from the National Cancer Institute. Flow cytometry research reported in this publication was supported by the National Center for Research Resources of the NIH under award number 1S10RR026802-01. RPPA analyses were performed at the Functional Proteomics Core Facility at the MD Anderson Cancer Center, which is supported by a National Cancer Institute (NCI) Cancer Center Support Grant (P30CA016672). This work was supported by R01CA121118 (S.L. Holmen) from the NCI and 347651 (S.L. Holmen) from the Melanoma Research Alliance.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.