AXL-Driven EMT State as a Targetable Conduit in Cancer

AXL, which stems from the Greek word for uncontrolled, “anexelekto,” is a receptor tyrosine kinase (RTK) belonging to the tumor-associated macrophage (TAM) family, comprising of TYRO-3, AXL, and MER (Fig. 1). Structurally, the TAM receptors comprise two Immunoglobulin-like (Ig) domains, two fibronectin type III (FNIII) moieties in their extracellular domain, and the conserved amino acid sequence KW(I/L)A(I/L)ES in their kinase domain (1). Among the RTK family, TIE and TEK are also known to contain both Ig and FNIII motifs on their ectodomains. The Ig domains are common to the FGF, VEGF, and platelet-derived growth factor (PDGF) receptor families, whereas the FNIII is prevalent in the ephrin and insulin families. The MET RTK family, comprising of cMET and RON, share high degree of sequence similarity in the kinase domain (2). The human AXL gene resides on chromosome 19q13.2 and is encoded by 20 exons to form the full-length protein comprising of 894 amino acids. Though the expected molecular weight is 98 kDa, it is posttranslationally glycosylated to form either a 120 kDa (partially glycosylated form) or 140 kDa protein (complete glycosylation; ref. 3). It is activated by its ligand growth arrest specific 6 (Gas6; ref. 4), which becomes biologically active upon vitamin K–dependent posttranslational modifications (5). AXL was originally identified in 1991 as a transforming gene in chronic myeloid leukemia (CML; ref. 3). Under normal physiologic conditions, it is ubiquitously expressed in several tissues and organs, including but not limited to the hippocampus, cerebellum, macrophages, platelets, endothelial cells, heart, liver, kidney, and skeletal muscle. It was found to be overexpressed in several cancers (6) such as breast, lung, liver, colon, gastric, ovarian, pancreatic, and glioblastoma.

Gas6 binds to the ectodomain of AXL, causing receptor dimerization with a 2:2 stoichiometry of Gas6 and AXL (7). The proximity of the kinase domains of two AXL moieties in the RTK–ligand complex enables trans-autophosphorylation of the residues on the cytoplasmic tails, where signaling molecules like phospholipase C-γ (PLCγ), PI3K, and growth factor receptor-bound protein 2 (Grb2) can dock (8, 9). The tyrosine residues 698, 702, and 703 in the human sequence of AXL are conserved among the TAM receptors and are involved in the functional activity of the kinase. The tyrosine residues 779, 821, and 866 are also potential autophosphorylation sites in the C-terminal AXL domain (9). Gas6-independent AXL phosphorylation can also occur when the RTK is overexpressed, resulting in RTK dimerization (10, 11).

In a cell type–dependent context, Gas6/AXL signaling can mediate growth, survival, proliferation, motility, and invasion by harnessing a diverse repertoire of signaling networks such as the Ras/Raf/MEK/ERK cascade, PI3K/Akt signaling pathways. The MAPK/ERK cascade is usually involved in proliferation, whereas PI3K activation signaling converges on cell survival through the Akt/ribosomal s6 kinase (S6K) axis (12). Gas6/AXL signaling also causes increased expression of antiapoptotic proteins such as B-cell lymphoma 2 (Bcl-2) and B-cell lymphoma-extra large (Bcl-xL), phosphorylation and activation of NF-κB, phosphorylation and stabilization of Bad, and inhibition of proapoptotic proteins such as caspase 3, to induce prosurvival signaling (13, 14). Signaling through the PI3K/Ras/Rac axis causes actin reorganization and enables migration (15, 16). The adapter protein Nck2, which comprises of three tandem SH3 and one SH1 domains, is involved in linking AXL with other signaling complexes. In particular, the AXL–Nck2 interaction facilitates AXL-mediated modulation of the integrin-linked kinase (ILK), a major component of signaling platforms at focal adhesions, thereby enabling AXL to regulate cytoskeleton dynamics (6). Furthermore, AXL has also been directly linked to control of contractility by its ability to phosphorylate tropomyosin 2.1 (17). AXL signaling in cancers has also been implicated in processes such as epithelial–mesenchymal transition (EMT; refs. 6, 18, 19), wherein polarized epithelial cells lose their junctional integrity to become motile, invasive, mesenchymal cells (20, 21).

The modulation of AXL required to regulate signaling has not been extensively documented. The phosphatase C1-TEN, which is known to prevent signaling downstream of Akt (22), can interact with AXL (23). There is evidence to suggest that the suppressor of cytokine signaling (SOCS-1) could serve to negatively modulate AXL signaling (24, 25). The E3-ubiquitin ligase Cbl-b has also been reported to counteract AXL signaling by ubiquitination and subsequent degradation of AXL (26, 27), enhancing the antitumor efficacy of NK cells in the immune system. The schematic representation of the Gas6/AXL signaling is illustrated in Fig. 2.

In a panel of 643 human cancer cell lines, elevated AXL expression correlated positively with the mesenchymal phenotype, particularly in NSCLC and breast cancer. It was further demonstrated that AXL inhibition was synergistic with antimitotic drugs in EMTed systems that presented with resistance to tyrosine kinase inhibitors (TKI; ref. 28).

Given the relationship between drug resistance, changes in signaling, and emergence of an invasive phenotype is well appreciated, EMT state-mediated rewiring of the RTK signaling nodes has identified AXL as a key player. In triple-negative breast cancer (TNBC), AXL is transactivated by EGFR and is shown to diversify EGFR-mediated signaling and confer resistance to EGFR-TKI (29).

In ovarian cancer, the EMT state has been shown to modulate AXL signaling, with mesenchymal systems preferentially amplifying the Gas6/AXL signaling node (30). Exclusively in mesenchymal systems, AXL coclusters with, and activates EGFR, HER2 and tyrosine-protein kinase Met (cMET), resulting in protracted downstream phosphorylated ERK (pERK) temporal response, motility, and invasion. This addiction to the Gas6/AXL signaling node sensitizes the mesenchymal system to AXL inhibition. The epithelial systems present with linear AXL signaling axis, are under regulation of cellular phosphatases such as dual specificity protein phosphatase 4 (DUSP4) that curtail pERK response, and are therefore less sensitive to inhibition of AXL (Fig. 3; ref. 30).

EMT gradient-mediated rewiring of signaling circuitries has also been documented for the PI3K pathway. Epithelial systems rely on HER3 signaling to activate PI3K signaling, and mesenchymal systems rely on growth factor stimulation, p110α, and upregulated PI3K catalytic 110-KD alpha (PIK3CA) to harness the PI3K signaling node (31).

The role of AXL in EMT has been documented in literature for breast, ovarian, non–small cell lung, pancreatic, cancer, and glioblastoma among others. In breast cancer, AXL overexpression confers worse prognosis in patients, with AXL expression being elevated in metastases. AXL was also identified as a downstream effector of EMT, which facilitated tumor formation and invasiveness. Furthermore, AXL knockdown prevented the dissemination of highly metastatic breast cancer cell lines from mammary glands to the lymph nodes, and other organs, thereby increasing survival. This study provided the first in vivo evidence that directly links AXL to metastasis (32). It has been demonstrated that AXL acts in a positive feedback loop to induce EMT in normal and immortalized human mammary epithelial cells, and regulates self-renewal and chemoresistance in breast cancer stem cells (18). Furthermore, EMT-driven breast cancer cell motility is dependent on AXL upregulation (33).

AXL inhibition in ovarian cancer drastically abrogated invasion and matrix metalloproteinase activity. Furthermore, inhibiting AXL prevented initiation of metastatic dissemination and also prevented progression of established metastatic lesions (34). The Gas6/AXL signaling node has also been shown to sustain the EMT state in ovarian cancer by inducing motility in the mesenchymal subtype (30), as well as invasion by converging on the integrin β3 pathway (35). Importantly, AXL has also been identified as a prognostic marker for ovarian cancer patients by several groups (30, 35, 36).

In esophageal squamous cell carcinoma, AXL overexpression correlated positively with tumor progression, and consequently adverse prognosis and distant metastasis (37). AXL knockdown in pancreatic cancer resulted in significant reduction in EMT transcription factors SNAI1, SNAI2, and TWIST, and thereby attenuated EMT-mediated invasiveness, migration, and metastases (38). In malignant gliomas, AXL is implicated in brain tumor growth, invasion, and prolonged survival (39).

In non–small cell lung cancer (NSCLC), AXL is shown to confer mesenchymal cells with increased resistance to EGFR targeted therapy, with inhibition of AXL resensitizing the cells (40). Also, in EGFR-mutant lung cancer models, AXL overexpression conferred erlotinib resistance even in the absence of EGFR T790M mutation or cMET activation. Genetic or pharmacologic inhibition of AXL restored sensitivity to erlotinib in these tumor models (41). It is of interest to note that AXL confers resistance to EGFR-targeted therapies (29, 41, 42) by diversifying signaling, as well as to PI3K inhibitors due to association with EGFR (43). Inhibition of upstream regulator of AXL, such as the YAP/Hippo pathway, has also been shown to restore sensitivity to EGFR inhibitors in otherwise EGFR-TKI–resistant lung cancers (44). Apart from NSCLC, AXL confers resistance to EGFR-targeted therapy in head and neck cancer (42).

AXL has also been implicated in acquired resistance to HER2-targeted therapy in breast cancer through EMT-mediated mechanisms (45). Gas6/AXL signaling has also been shown to confer chemoresistance in breast cancer cells through the Akt/GSK-3β/β-catenin, which converges on ZEB1 to regulate DNA damage repair pathways and EMT (46). AXL-mediated EMT also results in resistance to anaplastic lymphoma kinase (ALK) inhibition in neuroblastoma (47) and sunitinib resistance in renal cell carcinoma (48).

Aside from their pro-oncogenic signaling in cancer cells, the TAM receptors have also been previously described for their role in modulating immune homeostasis. They function as negative immune regulators primarily by dampening activation of innate immune responses as well as clearance of apoptotic cells. Hence, deregulation of TAM signaling, particularly AXL, has been coupled with autoimmune and inflammatory diseases. In the cancer setting, AXL is involved in tumor initiation and progression, as well as negatively regulating antitumor immune responses (49). Tumor cells have been shown to harness AXL signaling to directly promote growth and metastasis, as well as to dampen NK-cell activation and reduce innate cell-mediated antiimmune response (49). At present, there are many bottlenecks of immunotherapy to target immune checkpoints such as CTLA-4 and PD-1/PD-L1 with limited clinical responses in solid tumors. T-cell exclusion is a common mechanism of immune resistance employed by tumors, and AXL has been implicated in sustaining oncological pathways that contribute to this phenomenon. Indeed, targeting the Gas6/AXL pathway has been shown to enhance anticancer immune response following radiotherapy (50). Therefore, targeting AXL would have the dual benefit of being an anticancer therapeutic as well as synergizing the antitumor immune response.

Several small-molecule inhibitors have been developed to target AXL, and R428 (51), which is an AXL-specific inhibitor, was shown to block tumor dissemination and prolong survival in metastatic breast cancer mouse models. It is based on a trisubstituted triazole core with nanomolar range potency against AXL in biochemical assays. R428 prevented Gas6-mediated AXL and subsequently Akt phosphorylation, and also suppressed invasion of cancer cells, in a dose-dependent manner. In mouse models, R428 suppressed angiogenesis and abrogated breast cancer metastasis, correlating with attenuation of Akt and ERK phosphorylation. Pharmacokinetic profiles in mice showed good plasma stability with a half-life of 13 hours. R428 has also shown to reverse AXL-mediated resistance to ALK inhibitors in neuroblastoma (47). A summary of R428 and other AXL small-molecule inhibitors is given in Table 1 (52).

Another small-molecule inhibitor n-butylidenephthalide, loaded and delivered through an intracerebral gliadel wafer, has shown remarkable efficacy in down modulating AXL expression and tumor invasion in glioblastoma multiforme (53). In glioblastoma, another small-molecule inhibitor BMS-777607 has been shown to effectively reduce growth, migration, and invasion, both in vitro and in vivo (54). Recently, a series of 4-Oxo-1,4-dihydroquinoline-3-carboxamide derivatives were synthesized as highly potent AXL inhibitors. The lead compound 9im showed inhibition against the AXL kinase domain at low nanomolar concentrations and suppressed TGFβ1-induced EMT, migration, and invasion in breast cancer cells (55).

Aside from small-molecule inhibitors, biologics to target AXL are in preclinical development. The YW327.6S2 antibody inhibits AXL activation by abrogating its binding with Gas6 as well as down regulating the receptor (56). Also, AXL inhibition with YW327.6S2 enhanced the efficacy of ERGR inhibitors in tumors resistant to EGFR-targeted therapy. YW327.6S2 also showed synergy with carboplatin/paclitaxel in the lung cancer xenograft model (56). The anti-AXL monoclonal antibody 20G7-D9 induced AXL degradation in TNBC cell lines as well as in patient-derived xenografts, and curtailed Gas6/AXL-dependent signaling events, particularly the EMT-related genes SNAIL, SLUG, and VIM, which are mediated by the Gas6/AXL/FRA-1 axis (57). Although both these antibodies are promising therapeutic strategies, a potential drawback of this approach would be binding of the antibody to soluble AXL (sAXL) in the blood stream; this might severely reduce efficacy of the treatment. MicroRNA sponges could also perform an alternate form of AXL targeting, given miR-432 has been shown to down modulate AXL in lung adenocarcinoma.

Other biological approaches would include engineering sAXL to sequester Gas6 to prevent Gas6-mediated AXL activation (58). The high-affinity AXL decoy-receptor MYD1-72 has been shown to bind Gas6 with femtomolar affinity and drastically inhibit disease progression in several preclinical models (59). Although this novel approach would ensure high specificity and limited toxicity, Gas6-independent AXL dimerization due to receptor overexpression could still promote tumorigenesis (60).

Given the heavy dependence of the mesenchymal state on AXL signaling, it would be beneficial to target specifically these AXL-driven states. A potential design could be to select for patients having tumors with mesenchymal features using gene expression profiling. Gene expression profiling on formalin-fixed, paraffin-embedded diagnostic blocks has been validated in the clinical setting such as ovarian cancer (61). Intratumoral heterogeneity would need to be taken into consideration when applying this strategy. Therefore, multiple core biopsies taken from the tumors would need to be adopted. Another strategy could be to stratify patients based on the notion of observed AXL/RTK clustering in the mesenchymal tumors, thereby enriching for tumors amplifying the AXL signaling nodule. Proximity ligation assays could be deployed in patient biopsies to identify the presence of these clusters and thereby enrich for AXL-driven tumors (62). Downstream signaling molecules such as pERK could serve as surrogate pharmacodynamic readouts for the response of anti-AXL therapy.

EMTed systems represent a challenge in the field of cancer biology wherein they equip cancer cells with aggressive phenotypes, yet they cannot be effectively targeted. AXL represents a novel class of RTKs, which is linked to, and reciprocally regulated by, the EMT state. Targeting AXL reverts the EMT phenotype and resensitizes cancer cells to several small-molecule inhibitors and chemotherapeutics. Given the increasing evidence that underscores the intrinsic links between EMT and the AXL-addicted state, it would be beneficial to enrich the mesenchymal subtype patients in a prospective AXL inhibitor trial. Using AXL signatures or expression state could greatly benefit in stratifying patients for potential AXL inhibitor trials, as well as biomarkers for disease status and progression.

No potential conflicts of interest were disclosed.

This work was supported by the National Medical Research Council (NMRC) of Singapore under its Center Grant scheme to National University Cancer Institute (NCIS) under the EMT Theme (R.Y.-J. Huang). J. Antony was funded by the Ovarian Cancer Action Centre and NUS Graduate School for Integrative Sciences and Engineering.