AXL Is an Essential Factor and Therapeutic Target for Metastatic Ovarian Cancer

Epithelial ovarian cancer (EOC) is a devastating disease that affects 1 in 72 women in the United States. It is the fourth leading cause of cancer–related deaths among women in the United States and is the primary cause of all gynecologic cancer–related deaths worldwide. Approximately 70% of patients diagnosed with EOC present with stage III or IV disease in which the tumor has disseminated beyond the ovaries and pelvic organs to the peritoneum and mesothelial lining of abdominal organs, including the diaphragm, stomach, omentum, liver, and intestines. Despite current surgical and cytotoxic therapies, 80% of patients diagnosed with advanced EOC develop recurrent disease and only 30% of patients survive 5 years after diagnosis (1). These findings underscore that tumor metastasis remains a major clinical challenge in the treatment of EOC. Thus, the identification of molecular targets regulating ovarian metastasis is critical for the development of more effective therapeutic agents for the treatment of EOC.

Ovarian tumor metastasis is a multistep process that is influenced by cell intrinsic factors, tissue microenvironment, and immune activity (2, 3). The demonstration of premetastatic niches that can direct metastatic tumor cell behavior and the reseeding of primary tumors by circulating tumor cells has reinforced the importance of the dynamic relationship between the primary tumor and its metastatic derivatives (4). The interplay of these factors has been proposed to explain why there has been no successful targeting of metastatic disease to date. Intrinsic cellular activities associated with the metastatic cascade include invasion/migration, changes in cellular adhesion, survival, and proliferation (5). Invasion and migration are particularly important for metastatic ovarian tumor cells as these processes are involved in both early and late stages of metastasis and are found to be specifically upregulated in high-grade tumors and metastases (6, 7). Therefore, the development of therapeutic agents targeting invasion/migration may be a useful strategy to inhibit metastasis and may provide clinical benefits to patients with advanced ovarian cancer.

The receptor tyrosine kinase AXL has recently been identified as a critical factor driving tumor cell invasion and migration. AXL is the founding member of the TAM family of receptor tyrosine kinases, which include Tyro3 (or SKY), AXL, and MER (8, 9). TAM receptors are characterized by an extracellular domain containing two immunoglobulin-like domains and two fibronectin type 3–like domains. The ligand growth arrest specific gene-6 (GAS6) is the common ligand for all three receptors. GAS6 has highest affinity for AXL and is the only known ligand for AXL (10). GAS6 binding to AXL results in AXL dimerization and autophosphorylation of tyrosine residues 779, 821, and 866, which serve as docking sites for a variety of intracellular signaling molecules (11). AXL signaling has been shown to promote cellular adhesion, invasion, migration, proinflammatory cytokine production, antiapoptosis, proliferation, and survival (11).

In tumor cells, AXL plays an important role in regulating cellular invasion and migration. Overexpression of AXL in tumor cells with low metastatic potential promotes migration and invasion (12–14). In addition, blockade of AXL signaling in metastatic tumor cells is sufficient to suppress tumor cell migration in vitro (12–17). In vivo, a role for AXL in invasion and metastasis has recently been described in breast and glioma tumor models (15, 17–19). MMP-9 has recently been identified as an important effector of AXL-mediated invasion. MMP-9 expression is affected by both overexpression and downregulation of AXL in breast and pancreatic tumor cells (14, 16). These studies show that AXL is an important factor for tumor cell invasion and metastasis; however, the role of AXL in ovarian cancer growth and progression remains unknown and importantly the efficacy of a therapeutic agent specifically targeting AXL in metastatic disease remains largely untested.

In this study, we test the hypothesis that AXL is a critical factor for ovarian tumor metastasis and that therapeutic blockade of AXL signaling may be an effective treatment for metastatic ovarian cancer. We use both genetic and therapeutic approaches to directly assess the role of AXL in the initiation and progression of metastatic ovarian cancer. This study shows that AXL is a therapeutic target for human ovarian cancer and shows that a specific therapeutic agent targeting AXL is sufficient to significantly inhibit metastatic tumor progression without normal tissue toxicity.

Ovarian SKOV3, SKOV3ip.1, and HEYA8 cells were obtained as a gift from Dr. Gordon Mills (M.D. Anderson Cancer Center, Houston, TX). Ovarian ES-2 and MESOV cells were a gift from Dr. Branimir Sikic (Stanford University, Stanford, CA). OVCAR-3 cells were purchased from the American Type Culture Collection. IGROV-1 and OVCAR-8 cells were purchased from the National Cancer Institute-Frederick DCTD tumor cell line repository. All cell lines were authenticated from the original source and were used within 6 months of receipt. Additionally, cells were tested upon receipt for viability, cell morphology, and the presence of Mycoplasma and viruses, including MPV, MVM, MHV, Ectromelia, MAV-1, MAV-2, Reo 1&3, LCMV, LDH, and TMEV (Charles River Laboratories). Cells were cultured in the appropriate medium supplemented with 10% heat-inactivated fetal bovine serum and 1% penicillin and streptomycin at 37°C in a 5% CO₂ incubator.

Oligos for AXL shRNA were synthesized as previously described (20), 5′-GATTTGGAGAACACACTGA-3′. A scramble sequence was used as a nontargeting shRNA, 5′-AATTGTACTACACAAAAGTAC-3′. Oligos were cloned into the pSiren RetroQ (BD Bioscience) vector. Infected cells were selected in puromycin (Sigma), and polyclonal populations were tested for decreased AXL expression levels by Western blot analysis.

The AXL ectodomain (amino acids 1–451) was amplified from the human AXL cDNA (Open Biosystems) and cloned into the cytomegalovirus-driven pADD2 adenoviral shuttle vector. Transient transfections with control vector or AXL ectodomain were performed with Lipofectamine 2000 (Invitrogen) into HCT116 cells. Conditioned medium was collected 48 to 72 hours following transfection.

The AXL ectodomain (amino acids 1–451) was cloned into the E1 region of E1⁻E3⁻ Ad strain 5 by homologous recombination followed by adenovirus production in 293 cells and CsCl gradient purification as previously described (21). The production and purification of the soluble AXL (sAXL) adenovirus and the negative control virus expressing murine IgG2α-Fc immunoglobulin fragment was performed as previously described (21).

All procedures involving animals and their care were approved by the Institutional Animal Care and Usage Committee of Stanford University in accordance with institutional and NIH guidelines.

For genetic studies, shSCRM and shAXL SKOV3ip.1 and OVCAR-8 cells were injected i.p. with 1 × 10⁶ and 5 × 10⁶ cells, respectively, in 0.5 mL of PBS into female nude mice. After sacrifice, the presence of ascites was identified, metastatic lesions were counted, and all visible lesions were dissected and removed to determine total tumor weight.

For therapeutic studies, SKOV3ip.1 and OVCAR-8 parental cells were injected i.p. with 1 × 10⁶ and 5 × 10⁶ cells, respectively, in 0.5 mL of PBS into female nude mice. Seven (SKOV3ip.1) or 14 (OVCAR-8) days following tumor cell injection, mice were injected with sAXL or control 1.9 × 10⁸ adenoviral plaque-forming units in 0.1 mL PBS into the tail vein.

For subcutaneous tumors, 5 × 10⁶ cells in 0.1 mL of PBS were implanted s.c. into the flanks of nude (nu/nu) 6-week-old female mice. Tumors were measured with calipers over a 45-day time course. Volume was calculated using the following formula: width² × length × 0.5.

Blood was collected from tumor-bearing Ad-Fc– or Ad-sAXL–treated mice. Comprehensive metabolic panel and complete blood count analysis was performed by the Department of Comparative Medicine at Stanford University. Tissue samples were collected from all major organs, including liver, kidney, brain, and spleen; fixed in 10% formalin; embedded in paraffin; sectioned; and counterstained with hematoxylin and eosin.

Statistical analysis of AXL expression in human tissue samples was performed using the Fisher's exact test. All other statistical tests were performed using Student's t test. P values of <0.05 were considered statistically significant.

We first compared AXL expression in normal ovarian tissue, primary tumor, and metastases. AXL expression was first examined in specimens of normal ovarian surface epithelium (OSE) taken from patients with benign ovarian disease because the majority of ovarian tumors are thought to arise from OSE (22). AXL was expressed in 0% (0 of 10) of specimens with normal OSE (Table 1; Fig. 1). In contrast, 73% (219 of 297) of all ovarian tumor samples including both type I and type II tumors were positive for membranous AXL staining in the epithelium, demonstrating that AXL expression is significantly higher (P < 0.0001) in ovarian carcinomas than normal ovarian epithelium (Fig. 1; Table 1). Additionally, we found that the number of cases with positive AXL staining (score of 2 or 3) was significantly higher in type II (high-grade serous and high-grade endometrioid) than type I (low-grade serous and low-grade endometrioid) primary tumors (P < 0.0001, Table 1). Type I tumors usually present at low stage and are not aggressive, whereas type II tumors present at advanced stage and are highly aggressive, indicating that AXL expression correlates with aggressive behavior in type II ovarian tumors (23). Furthermore, metastatic tumor samples collected from the omentum and peritoneum of serous ovarian cancer patients showed high AXL expression in 75% (24 of 32) and 90% (27 of 30) of specimens, respectively (Fig. 1; Table 1). These findings show that AXL expression is significantly induced in aggressive ovarian tumors and its expression is maintained at high levels in ovarian tumor metastases.

To examine the functional role of AXL in metastasis, we used a genetic approach to inhibit AXL in mouse models of ovarian metastasis. For this purpose, we screened a panel of human ovarian cancer cell lines for AXL protein expression to identify metastatic cell lines with high levels of AXL expression. Similar to our clinical findings, AXL was highly expressed in the majority of metastatic ovarian (SKOV3, OVCAR-8, ES-2, MESOV, HEYA8) cell lines, whereas AXL was expressed at undetectable or low levels in cell lines with low metastatic potential (IGROV1 and OVCAR-3; Fig. 2A). AXL-deficient metastatic ovarian (SKOV3ip.1 and OVCAR-8) cell lines were generated using previously described AXL shRNA-targeting sequences (20). Western blot analysis confirmed that cells expressing shAXL-targeting sequences expressed less than 5% of AXL protein compared with cells expressing the scramble control shRNA-targeting sequence (shSCRM; Fig. 2B).

We first examined if AXL plays a general role in the regulation of ovarian tumor cell proliferation and growth. To address these questions, we performed in vitro proliferation assays in which total cell numbers between AXL wild-type (shSCRM) and AXL-deficient (shAXL) cells were counted over a 10- to 14-day period. We found no significant difference in cellular growth curves between shSCRM and shAXL SKOV3ip.1 or OVCAR-8 cells in vitro (Fig. 2C). Similarly, no significant difference was observed in subcutaneous growth of shSCRM and shAXL SKOV3ip.1 cells (Fig. 2D). These findings indicate that AXL is not required for ovarian tumor cell proliferation or subcutaneous growth.

To determine whether genetic inactivation of AXL affects the ability of ovarian cancer cells to metastasize in vivo, we compared the ability of shSCRM and shAXL SKOV3ip.1 and OVCAR-8 cells to form metastases using the peritoneal xenograft model of ovarian cancer. This model recapitulates the peritoneal dissemination of human ovarian metastases in which nude mice develop rapidly progressive disease consisting of ascites and >100 small metastatic lesions attached to the mesentery, diaphragm, liver, and other peritoneal surfaces following peritoneal injection of metastatic human ovarian tumor cells (24). Immunohistochemical analysis of AXL expression in SKOV3ip.1 peritoneal metastases revealed that similar to human ovarian metastases, AXL is highly expressed in SKOV3ip.1 metastatic lesions, confirming that this is a relevant model system to investigate the role of AXL in ovarian metastasis (Fig. 3A). Twenty-eight days following peritoneal injection of shSCRM or shAXL cells, shSCRM mice displayed signs of severe ascites and morbidity, necessitating us to sacrifice the mice and investigate changes in tumor burden between the shSCRM- and shAXL-injected mice. Whereas mice injected with shSCRM cells developed ascites and >100 peritoneal metastases, mice injected with shAXL cells developed very few metastases (Fig. 3B). The average number of peritoneal metastases >5 mm in size was significantly reduced from 13.4 ± 4.3 in shSCRM-injected mice to 0.8 ± 0.5 in shAXL-injected mice (Fig. 3B). Similarly, the average weight of these tumors was significantly reduced from 236 ± 74 mg in shSCRM-injected mice to 39.2 ± 18 mg in shAXL-injected mice (Fig. 3B). In support of these findings, knockdown of AXL expression in OVCAR-8 cells significantly inhibited total ovarian peritoneal tumor mass and tumor number in mice injected with these cells (Fig. 3C). Collectively, these findings show that AXL is a critical factor for the establishment of ovarian tumor metastasis.

To determine a mechanism for AXL-mediated metastasis, we took an unbiased approach and directly compared the role of AXL in the critical cellular functions associated with the metastatic cascade, including proliferation, invasion, migration, adhesion, and survival (5). We found that shAXL SKOV3ip.1 and OVCAR-8 cells were significantly impaired in the ability to invade through type I collagen (Fig. 4). We also observed a modest decrease in cellular migration in shAXL cells; however, we were unable to find a difference in adhesion to ECM proteins or survival following serum withdrawal, indicating that AXL predominantly affects invasion in the metastatic cascade (Supplementary Fig. S1).

Matrix metalloproteinase-1 (MMP-1), MMP-2, and MMP-9 are key contributors to ovarian tumor cell invasion and progression (25). Therefore, we investigated whether AXL regulates the expression and/or activity of these enzymes in human ovarian cancer cells. Using shSCRM and shAXL OVCAR-8 and SKOV3ip.1 cells, we found that AXL regulates the expression of all three MMPs in human ovarian cancer cells. Loss of AXL significantly reduced the expression of MMP-1 and MMP-2 in SKOV3ip.1 cells (Fig. 5A). Similarly, in OVCAR-8 cells, MMP-1 and MMP-9 expression was significantly reduced in AXL-deficient cells (Fig. 5A). A recent report showed a role for MMP-9 in AXL-mediated invasion in breast cancer cells (14). Therefore, we focused on the role of AXL in the regulation of MMP-2 in ovarian cancer. Luciferase reporter assays revealed that MMP-2 promoter activity was significantly decreased in shAXL cells compared with shSCRM cells, indicating that AXL regulates MMP-2 at the transcriptional level (Fig. 5B). Gelatin zymography assays indicated that MMP-2–secreted protein levels were also significantly reduced in shAXL cells compared with shSCRM SKOV3ip.1 cells (Fig. 5C). Collectively, these findings suggest a role for AXL as an upstream regulator of MMP expression and activity in human ovarian cancer cells.

We next sought to elucidate the signaling pathways involved in AXL-mediated MMP expression. Activation of AXL by GAS6 has been reported to directly induce a number of intracellular signaling pathways, including phosphoinositide 3-kinase (PI3K), RAS, mitogen-activated protein kinase (MAPK), SRC, and PLC (for a recent review, see ref. 11). Among these pathways, the PI3K/AKT signaling pathway has been shown to regulate MMP-2 expression and invasion in ovarian cancer cells (26, 27). To determine whether PI3K signaling is affected by loss of AXL in SKOV3ip.1 cells, we performed Western blot analysis for phospho-AKT at Ser473 (P-AKT) in AXL wild-type and AXL-deficient SKOV3ip.1 cells. We found a profound inhibition of P-AKT expression in shAXL cells compared with shSCRM SKOV3ip.1 cells (Fig. 5D). Additionally, GAS6 stimulation of starved SKOV3ip.1 cells resulted in a PI3K-dependent induction of P-AKT as treatment with the PI3K inhibitor Ly294002 completely abrogated GAS6-induced P-AKT expression (Fig. 5D). The MAPK/extracellular signal-regulated kinase (ERK) pathway has also been shown to regulate MMP expression; however, inactivation of AXL in shAXL SKOV3ip.1 cells did not affect phospho-ERK1/2 levels, indicating that AXL may regulate MMP-2 expression through the PI3K/AKT signaling pathway (Supplementary Fig. S2).

Our findings thus far show that AXL is a critical factor for ovarian metastasis and support the hypothesis that therapeutic blockade may be an effective treatment for metastatic disease. To test this hypothesis, we used the soluble human AXL ectodomain as a therapeutic strategy to inhibit AXL signaling. The sAXL ectodomain functionally acts as a decoy receptor and has previously been shown to bind GAS6 with nanomolar affinity (refs. 28, 29; Supplementary Fig. S3). We first examined whether treatment with sAXL ectodomains is sufficient to inhibit AXL signaling and invasion in metastatic tumor cells. We and others have shown that the PI3K/AKT signaling is regulated by AXL in a variety of cell types (Fig. 5D; ref. 11). Treatment with sAXL ectodomains was able to reduce PI3K/AKT activation in GAS6-treated SKOV3ip.1 cells (Supplementary Fig. S3B). Additionally, treatment of metastatic tumor cells in collagen with sAXL was sufficient to dramatically reduce cellular invasion, demonstrating that sAXL treatment inhibits AXL signaling and invasion in vitro (Supplementary Fig. S3C).

We next examined whether sAXL treatment would affect metastatic tumor progression in highly metastatic models of ovarian cancer. We first established SKOV3ip.1 metastatic lesions in nude mice (day 1) and began treatment with sAXL at day 7 following verification of macroscopic lesions. sAXL therapy was delivered using an adenoviral system in which the liver releases systemic production of sAXL protein into the serum of mice for up to 28 days following injection (Fig. 6A; ref. 30). Macroscopic analysis of tumor burden at day 28 revealed that mice receiving sAXL therapy had a significant (P < 0.01) reduction in tumor burden compared with mice treated with the Fc control therapy. In the SKOV3ip.1 tumor model, total tumor weight and tumor number were decreased by 63% in mice treated with sAXL compared with Fc-treated mice (Fig. 6B). Similarly, in the OVCAR-8 tumor model, total tumor weight and tumor number were significantly decreased by 47% and 42% with sAXL therapy, respectively (Supplementary Fig. S4). We examined MMP-2 expression levels in SKOV3ip.1 tumors by real-time PCR analysis and found that MMP-2 levels were significantly decreased in the tumors of sAXL-treated mice compared with Fc control–treated mice (Fig. 6C). Given that previous antimetastatic inhibitors that target MMPs have been shown to have significant effects on normal tissue toxicity, we performed a comprehensive analysis of normal tissue toxicity in mice treated with sAXL therapy for 21 days. We observed no behavioral, macroscopic, or microscopic abnormalities in nude mice treated with sAXL or Fc therapy (Supplementary Figs. S5 and S6; data not shown). These results show that a single-agent AXL therapy is sufficient to significantly reduce metastatic tumor burden in mice with established disease without normal tissue toxicity. In addition, our findings suggest that the therapeutic effect of AXL on metastatic tumor growth may involve the inhibition of invasion at least in part through the regulation of MMP activity.

In this report, we show that advanced type II and metastatic ovarian tumor cells dramatically increase AXL receptor surface expression and are critically dependent on the GAS6/AXL signaling cascade for successful metastatic colonization. Our findings indicate that the PI3K/AKT signaling pathway may be an important component of the GAS6/AXL signaling cascade regulating ovarian tumor metastasis. Although genetic mutations in the PTEN/PI3K pathway are rare in type II ovarian tumors (23), phosphorylation and activation of AKT1 and AKT2 are common events in high-grade and advanced-stage ovarian tumors (31, 32). Proposed mechanisms for AKT activation in these tumors include increased protein expression, activated RAS, and activated receptor signaling pathways (31, 33). Our study identifies GAS6/AXL signaling as a novel mechanism for AKT activation in advanced ovarian tumors. PI3K/AKT controls metastasis by increasing the expression of proteolytic enzymes, including MMP-2 and MMP-9, resulting in the destruction of the extracellular matrix and enhanced tumor cell invasion (26, 27). Given the important role of PI3K/AKT and MMP pathways in ovarian tumor progression and metastasis, the development of therapeutic agents targeting these pathways is important for the treatment of cancer. The PI3K inhibitors wortmannin and LY294002 have been developed and are commonly used to inhibit tumor cell proliferation and tumor growth. However, given the poor solubility and the high toxicity due to the broad range of PI3K-related enzymes targeted by these compounds, these agents have had very limited clinical applications to date (34). AKT and MMP inhibitors have been tested in clinical trials for a variety of human cancers; however, the clinical outcomes from these studies were not satisfied and tissue toxicities were observed (for a recent review, see ref. 35). Here, we show that therapeutic blockade of GAS6/AXL signaling using soluble receptors is a highly specific and nontoxic strategy to inhibit PI3K/AKT, MMP-2, invasion, and metastasis in metastatic ovarian tumor cells.

Analysis of AXL expression in normal tissues shows that AXL is expressed in neural, lymphoid, myeloid, vascular, and reproductive tissues (36–41). However, analysis of Axl and Gas6 germline knockout mice shows that the GAS6/AXL signaling cascade is not required for embryonic development or normal tissue function (41, 42). Under pathophysiologic conditions of acute anemia, Gas6- and Axl-deficient mice have impaired erythropoiesis, indicating a role for GAS6/AXL signaling in the generation of erythroid progenitors and erythroblasts under these conditions (43). Although these studies show that genetic inactivation of GAS6/AXL signaling has minimal effects on normal tissue function, the effects of anti-AXL therapy on normal tissue function have not been tested in preclinical models of metastasis. Our study shows that therapeutic blockade of GAS6/AXL signaling using sAXL receptors does not induce tissue or hematologic toxicity in mice.

We show that single-agent AXL therapy is sufficient to inhibit metastatic tumor progression in highly metastatic models of metastatic ovarian cancer. These findings have important clinical implications for the treatment of ovarian cancer. Currently, there are no Food and Drug Administration–approved targeted therapies for the treatment of ovarian cancer, although several classes of inhibitors targeting epidermal growth factor receptor, vascular endothelial growth factor, platelet-derived growth factor receptor, and Src are in clinical trials for the treatment of advanced and recurrent ovarian cancer (ref. 44; http://clinicaltrials.gov). Standard therapy for ovarian cancer includes surgery with optimal debulking of disease followed by cytotoxic platinum-taxane combination therapy. Despite these efforts, 80% of patients diagnosed with ovarian cancer develop recurrent disease and only 30% of these patients survive 5 years following diagnosis. Our data suggest that AXL therapy may be an effective adjuvant therapy for the treatment of advanced and recurrent ovarian cancer. The model of metastatic ovarian tumor progression used in our studies resembles the development of recurrent disease in human patients following surgical debulking. We found that AXL therapy was able to reduce metastatic tumor burden in mice with established disease by 63%. Although the growth of established tumors was only modestly affected by AXL therapy, the establishment of new metastatic lesions during the progression of disease was significantly reduced. Taken together, our results indicate that AXL seems to function primarily as an antimetastatic agent and may be most effective as a combination therapy with current cytotoxic agents. Interestingly, AXL has also been linked to resistance to chemotherapeutic drugs, suggesting that anti-AXL therapy in combination with cytotoxic drugs may not only inhibit metastatic spread but may also enhance the efficacy of chemotherapy (45, 46).

In summary, AXL is an important therapeutic target for metastatic ovarian cancer. This study shows, using human sAXL receptors, that an AXL-specific therapy is sufficient to significantly reduce metastatic tumor progression in vivo using aggressive models of ovarian cancer. These findings have direct clinical implications for the treatment of ovarian cancer for which current therapies are ineffective and show that an AXL-specific inhibitor is an effective and nontoxic strategy for the treatment of metastatic disease.

No potential conflicts of interest were disclosed.

Grant Support: NIH grants CA67166 and CA116685, Elizabeth Wolf Foundation, and Silicon Valley Foundation (all to A.J. Giaccia), and NIH National Research Service Award T32 CA09151 from the National Cancer Institute (E.B. Rankin).

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.