Cabozantinib (XL184) Inhibits Growth and Invasion of Preclinical TNBC Models

This article is featured in Highlights of This Issue, p. 785

Triple-negative breast cancer (TNBC) accounts for 15% to 20% of breast cancers and is associated with advanced stage at diagnosis and poorer outcome compared with other breast cancer subtypes (1). TNBC is characterized by the lack of estrogen receptor (ER) and progesterone receptor (PR) expression and HER2 receptor amplification. Characteristic clinical features of TNBC include a peak in recurrence risk within the first 3 years, a peak of cancer-related death in the first 5 years, and a weak relationship between the tumor size and lymph node metastasis (2). At the molecular level, TNBC has significant overlap with the basal-like subtype with approximately 80% of TNBCs being classified as basal-like (1). Recent studies involving comprehensive gene expression analysis of TNBC cases revealed extensive molecular heterogeneity within TNBC and identified four to six distinct molecular TNBC subtypes (3, 4). These subtypes have unique expression signatures and ontologies and are defined as basal-like, mesenchymal, and luminal androgen receptor subtypes.

There is an urgent need for effective targeted therapeutics for TNBC patients; however, new treatment strategies are challenged by the significant intertumoral heterogeneity of TNBCs. Currently, TNBCs are treated with cytotoxic combination chemotherapy, with platinum-based therapies having the highest response rates. Only 22% of TNBC patients have a complete response to neoadjuvant chemotherapy (5) and therefore require additional therapeutic approaches. Receptor tyrosine kinases (RTK) and growth factors are highly expressed in several TNBC subtypes (3) and are attractive therapeutic targets. The success of trastuzumab in HER2⁺ breast cancer underscores the promise of targeting tyrosine kinases, yet several tyrosine kinase inhibitors (TKI) have had only limited success in the clinic due to diverse mechanisms of resistance. In breast cancer and most other cancers, multiple RTKs are frequently activated and contribute to resistance by providing functional redundancy of critical signaling networks (6, 7). Another component of breast cancer that has been revealed to play a significant role in progression and therapeutic resistance is the tumor microenvironment (TME; refs. 8–10). Despite the overwhelming data on the influence of the TME, cancer therapeutics are directed primarily at the tumor cells. The design of successful TNBC treatment strategies will need to take into consideration both the intertumoral heterogeneity of TNBCs and the signaling pathways that promote both progression and resistance.

The RTK MET drives several oncogenic processes, including invasion, proliferation, and survival, and is involved in the progression and metastasis of most solid human cancers (11). In breast cancer, MET is overexpressed in 20% to 30% of cases and is a strong, independent predictor of poor clinical outcome (12–16). We previously demonstrated that MET is expressed in all molecular subtypes of breast cancer, but we observe the highest expression in basal-like (TNBC) breast cancers (17, 18). These findings have been supported by several other studies on MET in basal-like breast cancers [for review, see (19, 20)]. Recently, we demonstrated that MET is coexpressed in the majority of HER2⁺ breast cancers and may be involved in therapeutic resistance to HER2-targeted therapies (21). These findings demonstrate that MET overexpression commonly occurs in the more aggressive breast cancer subtypes (i.e., TNBC) and may be a novel therapeutic target. In cancer, aberrant MET signaling can occur through overexpression of MET or HGF, amplification, mutation, or autocrine signaling. MET signaling is also frequently elevated in tumors due to increased secretion of HGF by cancer-associated fibroblasts (CAF). The exact mechanism by which MET signaling is dysregulated in TNBC has not been elucidated.

The TME is composed of a complex network of stromal cells, immune cells, extracellular matrix, and cytokines/chemokines that is also influenced by pH and hypoxia. The paracrine interactions between the tumor epithelium and TME have been shown to be critical for the invasive, metastatic, and resistant tumor phenotypes. A recent study found that cocultures with CAFs induce HGF signaling in basal-like, but not luminal-like breast cancer cells (22). We have shown that mammary fibroblasts engineered to secrete high levels of HGF (MF:HGF) enhance proteolysis and invasiveness of a preinvasive TNBC cell line (MCF10.DCIS) in 3D cocultures and progression to invasive carcinomas in vivo (8). These studies underscore the influence of the TME on RTK signaling and breast cancer progression.

In this study, we examined the efficacy of MET inhibition on TNBC progression in vitro and in vivo. We utilized the small-molecule inhibitor cabozantinib (XL184), which potently inhibits MET and VEGFR2 (23). Clinical activity with cabozantinib has been demonstrated in a range of cancers. A striking clinical result of cabozantinib is the significant decrease in bone metastases seen on bone scans in castrate-resistant prostate cancer patients (24, 25). Cabozantinib is being tested in a phase II clinical trial for metastatic TNBC and phase III clinical trials for metastatic renal carcinoma and hepatocellular carcinoma and has been FDA approved for treatment of metastatic medullary thyroid cancer. To analyze the temporal and dynamic interactions between the epithelial tumor cells and CAFs, we employed a 3D model referred to as MAME (Mammary Architecture and Microenvironment Engineering) and quantitative image analysis. We have previously shown that MAME cocultures are able to recapitulate tumor-TME interactions, the invasive phenotype that occurs in vivo, and model the ability of antagonists to reduce invasion (8, 26, 27; Sameni and Sloane, unpublished). For investigation of MET inhibition in vivo, we utilized a novel xenograft model that expresses human HGF on an immunocompromised SCID background. Because mouse HGF binds human MET with low affinity (28), traditional immunocompromised mouse models do not effectively activate human MET signaling. The hHGF^tg-SCID model enhances the growth of MET-dependent human tumors, defines the HGF contribution to tumor growth, and is a valuable model for preclinical testing of inhibitors targeting the MET signaling pathway (29, 30). Using these unique strategies, we are able to interrogate the efficacy of MET inhibition in environments in which both autocrine and paracrine signaling may be driving tumor progression.

Here, we observed that cabozantinib (XL184) inhibits growth and invasion of TNBC cells in both monocultures and cocultures with HGF-overexpressing fibroblasts or CAFs. This efficacy was observed for TNBC cell lines representing various TNBC molecular subtypes. Even though cabozantinib inhibits kinase activity of MET and VEGFR2, cabozantinib was ineffective against MET^neg TNBC cells. These findings imply that MET signaling is a crucial signaling node in TNBC growth and invasion. Additional in vivo studies demonstrated that cabozantinib treatment was highly effective in inhibiting TNBC progression and metastasis. In summary, these results reveal that cabozantinib inhibition may be a highly effective treatment strategy for TNBC patients.

Cabozantinib (XL184) was kindly provided by Exelixis. Reconstituted basement membrane (rBM; Cultrex) was purchased from Trevigen. Alexa Fluor 546 phalloidin, DAPI, Slow Fade Reagent, FITC-conjugated affinity-purified donkey anti-rabbit IgG and normal donkey serum, Click-it^REDU assay, Live/Dead Viability/Cytotoxicity Kits and Leibovitz L15 medium were from Life Technologies. Ki67 antibody was from Abcam. Cignal Lenti-RFP was purchased from Qiagen and Lenti-YFP from Lentigen. Protease/phosphatase inhibitor cocktail (#5872) was purchased from Cell Signaling Technology. Mammary epithelial basal medium (MEBM) without phenol red and MEGM was purchased from Lonza. Fetal bovine serum (FBS) was from HyClone. Bovine serum albumin, antibiotics, Triton X-100 and all other chemicals, unless otherwise stated, were purchased from Sigma. See Immunoblot Analysis and Experimental Lung Metastasis for additional antibodies.

All cell lines were cultured in basal medium supplemented with 10% fetal calf serum (MDA-MB-231, DMEM; MDA-MB-468, Liebovitz L-15 media; HCC70, DMEM; and Hs578T, DMEM + 10 μg/mL bovine insulin). Cells were originally purchased from American Type Culture Collection (ATCC). The WS12Ti fibroblasts are a human breast carcinoma associated fibroblast cell line that was isolated at Karmanos Cancer Institute (31, 32). MF:HGF is a normal human breast fibroblast line that has been engineered to express GFP and secrete high levels of HGF (a kind gift of Dr. Kuperwasser, Tufts University Boston, MA; ref. 33). MDA-MB-231 and HCC70 were transduced with Cignal Lenti-RFP and WS12Ti with Lenti-YFP in order to distinguish among the cell types in 3D cocultures. Cell lines were authenticated using the STR PowerPlex 16 system (Promega).

MDA-MB-231 and HCC70 cells were grown as monocultures or in cocultures with CAF or MF:HGF cells. Details for establishing and analyzing MAME cultures are described in detail in Results and Supplementary Data (34).

HCC70 and MDA-MB-231 cells were grown on glass coverslips to 40% to 50% confluency in triplicate. Media were changed to serum-free MEGM 24 hours prior to treatment. Cells were treated with DMSO or 2 μmol/L XL184 or vehicle (DMSO) for 25 minutes at 37°C. Then 100 ng/mL HGF (R&D Systems) was added to the cells in the presence or absence of XL184 and incubated at 37°C for 5 minutes. Cells were fixed with 4% formaldehyde in PBS for 15 minutes at room temperature (RT), followed by washing in PBS, and then incubated in blocking buffer (5% normal donkey serum/0.3% Triton X-100 in PBS) for 1 hour at RT. The cells were incubated with primary antibody, phospho-Met (Tyr1234/Tyr1235)(D26) XP Rabbit mAb (Cell Signaling Technology, 1:50), overnight at 4°C, and then incubated with Alexa Fluor 488-conjugated secondary antibody (Life Technologies) to enable fluorescent detection for 2 hours at RT. Nuclei were labeled with DAPI and cells were mounted with Slowfade gold antifade reagent (Invitrogen). Images were visualized on a Zeiss LSM 780 confocal microscope.

Whole-cell lysates were collected in a RIPA buffer containing protease/phosphatase inhibitor cocktail (Cell Signaling Technology). HCC70, MDA-MB-231, MDA-MB-468, Hs578T, CAFs, and MF:HGF were grown for 6 days, treated with 2 μmol/L of XL184 for 30 minutes, washed with PBS containing protease/phosphatase inhibitor cocktail and harvested in RIPA buffer plus inhibitor cocktail. For blots shown in Supplementary Figs. S1 and S2, cells were treated with HGF and/or XL184 for 24 hours. For 3D cultures, the cells were grown for 6 days in rBM-coated dishes, treated as above, washed with sterile PBS containing protease/phosphatase inhibitor cocktail, scraped in cold PBS 5 mmol/L EDTA plus inhibitor cocktail, placed on ice, washed 2 times for 1 hour each to remove rBM, and then RIPA buffer containing inhibitor cocktail was added to the cell pellets. Lysates were loaded based on DNA concentrations.

Lysates (20–40 μg) were resolved on a 4% to 20% TGX SDS-PAGE gel (Bio-Rad) and transferred to a PVDF membrane (Invitrogen). After blocking for 1 hour with 5% dry milk in TBST buffer (20 mmol/L Tris-HCl pH 7.4, 150 mmol/L NaCl, 0.1% Tween-20), blots were probed overnight at 4°C with the following primary antibodies from Cell Signaling Technology: Met (DC12), pMET (Y1234/Y1235; D26), AKT (#9272), pAKT (S473; #9271), MAPK (#9102), pMAPK (Thr202/Tyr204; #9101), and β-tubulin (#2146). Blots were reacted with peroxidase-conjugated antibody for 1 hour, followed by visualization of the proteins using ECL detection reagents (Amersham).

MDA-MB-231 alone or together with MF:HGF or Ws12Ti cells were seeded on rBM with 2% rBM overlay. Vehicle control or 2 μmol/L XL184 were added at day 0 and replaced every other day. At day 6 or 12, the Live/Dead assay was performed as previously described (35).

All animal studies were conducted with protocols approved by the Institutional Animal Care and Use Committee of the Van Andel Research Institute. The breeding and genotyping of the hHGF^tg-SCID mice has been previously described (29). Female mice ages 4 to 8 weeks were used for the tumor growth kinetic experiments. Age-matched C3H-SCID littermates from the same mating cages as the hHGF^tg-SCID mice were used as experimental controls. For orthotopic tumor formation, subconfluent cells were harvested and resuspended in serum-free DMEM or RPMI at concentrations of 1 × 10⁶ cells/mL. Each animal received 100 μL of cell suspension injected s.c. into the mammary fat pad. Tumor-bearing mice for each mouse strain (either hHGF^tg-SCID or C3H-SCID) were randomized into two groups (10 mice/group) for treatment with vehicle or XL184 (30 mg/kg, once daily). Following cell injection, mice were monitored for tumor formation twice weekly and tumor volume was measured with manual calipers [volume (mm³) = length × width × depth]. Treatment started when the average tumor size reached 150 mm³ and drugs were administered for 21 days. For the established tumor study, only hHGF^tg-SCID animals were used and treatment started when the average tumor size reached 500 mm³. Linear mixed-effects models were used to test for significant differences in drug response across treatment arms; the XenoCat modeling framework (36) was leveraged to increase statistical power for poorly growing xenograft.

For lung metastasis formation, 5.0 × 10⁵ MDA-MB-231 cells were washed and harvested in HBSS and subsequently injected into the lateral tail vein in a volume of 0.2mL. Mice were treated with vehicle or XL184 (30 mg/kg, once daily) starting the following day. Endpoint assays were conducted at 14 days after injection unless significant morbidity required that the mouse be euthanized earlier. Pulmonary metastases were identified both by perfusing with 15% India ink solution and bleaching the collected lungs in Fekete's solution. For quantitation, fixed sections were cut every 200 μm and immunostained for human MHC Class I alpha antibody (EP1395Y; Abcam). Stained sections were scanned on the Aperio XT ScanScope and lung metastases were quantitated using multispectral image analysis. Positive pixel percentage was determined by quantitating the percentage of pixels positive for human MHC Class I alpha staining per total tissue area in each section.

To understand how XL184 treatment affects MET signaling, we first examined MET expression and activity in multiple TNBC cell lines that are representative of several TNBC subtypes identified by Lehmann and colleagues (3). These cell lines correspond to the two major basal-like subtypes and a mesenchymal-like subtype that were identified by the Pietenpol laboratory. This included Hs578t (mesenchymal stem-like; basal B), MDA-MB-231 (mesenchymal stem-like; basal B), MDA-MB-468 (basal-like 1; basal A), and HCC-70 (basal-like 2, basal A) cells. We observed high MET expression in Hs578t and HCC-70 cells, moderate expression in MDA-MB-231 cells, and low expression in MDA-MB-468 cells (Fig. 1A). In 2D cultures, we observed MET phosphorylation (pMET) in HCC70 cells, which expressed the highest levels of MET. MET activation was inhibited with XL184 treatment (Fig. 1A); however, MET phosphorylation was unaffected by MEK inhibition via U0126 (Fig. 1D). To confirm that MET phosphorylation was inhibited by XL184, we treated HCC70 cells with HGF and examined MET Y1234/1235 and Y1349 phosphorylation (Supplementary Fig. S1A and S1B). Both pMET Y1234/1235 and pMET Y1349 were activated by exogenous HGF and inhibited by XL184 even in the presence of HGF. We also examined MET activation and inhibition in HCC70 and MDA-MB-231 cells using immunofluorescence (Fig. 1B; Supplementary Fig. S1D). Here we observed low basal levels of pMet Y1234/Y1235 in vehicle-treated HCC70 cells (Fig. 1B, top left). Treatment with HGF induced increased pMET levels and membrane localization (Fig. 1B, bottom left). Upon XL184 treatment, there was a significant decrease in pMET and membrane localization in both HGF- and non-HGF–treated cells (Fig. 1B, right). A similar pattern was observed in MDA-MB-231 cells, yet membrane localization was stronger in HCC70 cells (Supplementary Fig. S1D). We also examined MET expression and activation in two fibroblast lines that were utilized in the 3D coculture experiments (Fig. 1A, right). The normal breast fibroblast line MF:HGF exhibited very weak MET expression, whereas the CAF line WS12Ti expressed moderate levels of MET. HGF expression was observed to be slightly higher in the MF:HGF fibroblasts than in the WS12Ti CAFs (Supplementary Fig. S1C). Even though HGF is not expressed in normal epithelial cells, low to moderate expression was observed in TNBC cells (Supplementary Fig. S1C).

The PI3K/AKT and MAPK/ERK pathways are critical mediators for maintaining proliferative and invasive signaling during RTK activation. It is hypothesized that signaling plasticity among RTKs is necessary for maintaining PI3K/AKT and MAPK/ERK signaling during tumor progression. MET strongly activates both the PI3K/AKT and MAPK pathways and these signaling effectors may reflect MET inhibition and signaling mechanisms of resistance. We evaluated the effect of XL184 treatment on these pathways and observed only a small decrease in AKT phosphorylation and an increase in total AKT expression in MDA-MB-231 and Hs578t cells (Fig. 1C). The opposite effect was observed with MEK inhibition where AKT phosphorylation increased in the presence of U1026 in both MDA-MB-231 and HCC70 cells (Fig. 1E). Interestingly, we observed a significant decrease in pAKT and an increase in total AKT in both the MF:HGF and WS12Ti fibroblast cells in the presence of XL184 (Fig. 1C). For the MAPK pathway, we observed a moderate increase in pMAPK in XL184-treated HCC70 and MDA-MB-468 cells (Fig. 1C). Again, the opposite affect was observed in the fibroblast lines, where both MF:HGF and WS12Ti cells had a decrease in MAPK activity when treated with XL184. In comparing MET inhibition to MEK inhibition, we observed that U0126 treatment caused a significant decrease in MAPK activation (Fig. 1E). In 3D monocultures of MDA-MB-231 and HCC70 cells, we observed similar MAPK signaling changes with XL184 treatment (Fig. 1F). In the MDA-MB-231 cells, there was no change in pMAPK with MET inhibition, whereas in HCC70 cells MET inhibition led to an increase of pMAPK activity. No change in total MET was detected with XL184 treatment. We also evaluated the effect of MET inhibition on ERK5 expression and activity (Supplementary Fig. S2). We observed a minimal increase in total ERK5 expression in MDA-MB-231 cells treated with XL184, yet did not see any significant changes in pERK5 in HCC70 or MDA-MB-231 cells. Overall, we observed distinct effects of MET and MEK inhibition in TNBC cells. These results also revealed a divergent response to MET inhibition in TNBC cells versus fibroblasts.

To recapitulate the architecture of invasive carcinomas and their associated signaling networks, we utilized 3D MAME coculture models and performed live-cell imaging in real-time to examine responses to XL184 treatment. The MAME cocultures are a tractable system based on the 3D reconstituted basement membrane culture models used by the Bissell and Brugge laboratories [for review, see (37, 38)]. Our MAME models allow for coculture of multiple cell types in the context of a 3D microenvironment and have been optimized for live-cell imaging of real-time interactions among the cells (34). Because MDA-MB-231 and HCC70 cell lines represent two different TNBC subtypes, are highly invasive in vitro, and are capable of developing tumors in vivo, these lines were used in the following assays. We first examined the effects of XL184 treatment on the TNBC cells grown in 3D reconstituted basement membrane overlay cultures using protocols we have described (39). For cocultures, MF:HGF or WS12Ti CAF cells were seeded in addition to TNBC cells. Cellular structures were allowed to develop for 4 days before the addition of XL184, which was replaced every 2 days with structures being imaged on day 10 (Fig. 2A). In monocultures of MDA-MB-231 and HCC70 cells, we observed significant dose–response reductions in volume of the multicellular 3D structures with XL184 treatment (Fig. 2B and E). Even though we saw greater inhibition at 4 to 6 μmol/L, the drug had a tendency to precipitate in 3D cultures at these higher concentrations (dosing ranges were based on studies in Yakes and colleagues; ref. 23). For this reason, subsequent experiments were performed with 2 μmol/L XL184. Next, we examined the effect of XL184 treatment on TNBC growth both alone and in coculture with MF:HGF cells. XL184 caused a significant decrease in volume and a reduction in the number of structures formed by MDA-MB-231 cells grown either alone or in coculture with MF:HGF cells (Fig. 2C and D). XL184 similarly reduced the volume and number of structures formed by HCC70 cells grown either alone or in coculture with MF:HGF cells (Fig. 2F and G). We also evaluated the effect of MEK inhibition on MDA-MB-231 growth (Supplementary Fig. S3). Even though MEK inhibition reduced MDA-MB-231 volume, we observed a greater reduction in structure volume with XL184 treatment. This suggests that while targeting downstream effectors of the MET signaling pathway may be effective in TNBC, greater efficacy is observed by targeting the MET receptor. These observations parallel the results we observed with the selective MET inhibitor SU11274 in MAME cocultures of triple-negative DCIS cell lines and MF:HGF fibroblasts (8).

XL184 is known to have activity against MET and VEGFR2. To verify that the efficacy of XL184 on TNBC growth is mediated through the MET pathway, we examined the effect of XL184 on both MET^pos (MDA-MB-231 and HCC70) and MET^neg (T47D) cell lines in coculture with MF:HGF cells. T47D is a luminal breast cancer cell line that is commonly used as a negative control for MET expression. XL184 treatment reduced invasive outgrowths and significantly decreased the volume of structures formed by MET^pos and MF:HGF cells, but not invasive outgrowths or volume of the structures formed by MET^neg (T47D) and MF:HGF cells (Fig. 3A and B). We also evaluated VEGF and VEGFR2 expression in the TNBC cells. Cytokine analysis revealed no VEGF expression and by Western blot analysis we observed minimal levels of VEGFR2 only in MDA-MB-231 cells (data not shown). These results verify that MET signaling is a vital pathway in the MET^pos TNBC cells that is abrogated with XL184 treatment.

As noted above, XL184 treatment appears to reduce the invasive phenotype of structures formed by MET^pos TNBC and HGF-overexpressing fibroblasts (see Fig. 3A). To further analyze possible effects of XL184 on the invasive phenotype, we used Volocity software to segment the 3D structures into cores (pseudocolored purple) and invasive outgrowths (pseudocolored green; Fig. 4A). A dramatic change in invasive outgrowths is illustrated by the reduced green fluorescence in structures treated with XL184 (Fig. 4B). Quantification of the volume of the structure cores and multicellular invasive outgrowths revealed that XL184 significantly reduced the volume of both cores and invasive outgrowths of MDA-MB-231 structures (Fig. 4C). XL184 also significantly reduced the volume of invasive outgrowths formed by MDA-MB-231 cells in coculture with MF:HGF cells, yet a significant decrease in core volume with XL184 treatment was not observed (Fig. 4D). These results suggest that HGF-expressing cells within the tumor microenvironment may reduce the efficacy of MET inhibitors against TNBC proliferation.

To further understand the effect of cabozantinib on TNBC proliferation we examined the effect of XL184 on live cells in TNBC monocultures or cocultures with fibroblasts. We observed that XL184 treatment reduced the number of live 3D structures (Fig. 5A) and volume of live structures (Fig. 5B) formed by MDA-MB-231 cells in monoculture and coculture with Ws12Ti cells (Fig. 5). A similar effect was observed in cocultures with MF:HGF cells (data not shown). Because tumor recovery after drug removal is a concerning clinical issue, we evaluated the persistence of XL184 treatment on growth suppression. In this experiment, cells were treated for 6 days with XL184 and then the drug was removed for 6 days before imaging (Supplementary Fig. S4). When treatment with XL184 was stopped, some recovery in growth was observed, consistent with XL184 inducing cytostasis in a subpopulation of the cells (40). XL184 does suppress proliferation as evidenced by a decrease in cells staining for the proliferation marker Ki67 in structures formed by MDA-MB-231 cells alone or in coculture with either CAFs or MF:HGF cells (data not shown).

To examine the efficacy of inhibiting MET in vivo, we injected MDA-MB-231 and HCC70 cells into the mammary fat pads of hHGF^tg-SCID and C3H-SCID animals. Because the hHGF^tg-SCID mice express human HGF, they are able to mimic the high HGF-expressing microenvironment of most cancers. When the tumors reached 150 mm³, animals were treated with XL184 for 3 weeks. Both cell lines showed enhanced growth rates in the hHGF^tg-SCID mice (Fig. 6A) as compared with the C3H-SCID control mice (Supplementary Fig. S5A and S5B). MET inhibition with XL184 resulted in a significant decrease in tumor growth as compared with the vehicle (P < 0.001) for both MDA-MB-231 and HCC70 xenografts. Evaluation of MET activation in vehicle and XL184-treated tumors showed that MET is highly active in vehicle-treated tumors (Fig. 6B and D), whereas MET activity is only present in a thin layer of cells at the periphery in XL184-treated tumors (Fig. 6C and E). XL184 treatment caused significant necrosis in the majority of the tumor whereas the cells on the tumor periphery are proliferating (data not shown). These pMET positive cells at the tumor edge are surrounded by viable mammary fat pad and exposed to XL184 therefore it is likely that these cells represent a MET-positive resistant subpopulation. To examine the efficacy of XL184 against established TNBC growth, we started treatment after the MDA-MB-231 tumors were approximately 500 mm³ (Fig. 6F). Even though XL184 significantly inhibited tumor growth (P < 0.005), the established tumors treated with XL184 have an upward growth trend likely due to resistance mechanisms being present in the late-stage tumors. The effect of MET inhibition on metastasis of MDA-MB-231 cells was determined in an experimental lung metastasis model. We observed that XL184 treatment significantly inhibited metastasis (P < 0.02) of TNBC cells to the lung (Fig. 6G). These in vivo studies validate our in vitro results and indicate that XL814 treatment is an effective therapeutic strategy for TNBC.

TNBC comprises 15% to 20% of all breast cancers yet accounts for a disproportionate percentage of breast cancer deaths. This depressing statistic is largely attributable to the aggressive nature of the disease, its high molecular heterogeneity, and the lack of targeted therapies for TNBC (1, 2). Because TNBC often recurs within 3 to 5 years of diagnosis, there is a short timeframe in which to achieve a complete response in TNBC patients (2). The high expression of RTKs and growth factors across several TNBC subtypes makes targeting RTK signaling a promising therapeutic option for TNBC patients (3). In this study, we targeted the MET oncogene as it is a RTK that is highly expressed across several TNBC subtypes and numerous MET inhibitors are currently being tested in clinical trials.

The tumor microenvironment is a substantial component of a tumor that can significantly influence both breast cancer progression and treatment efficacy (41, 42). Nonetheless, the human breast cancer TME is often excluded in preclinical testing of targeted therapies. In this study, we used distinct in vitro and in vivo models that recapitulate the TNBC epithelium and TME. Our 3D MAME cocultures are able to model both progression to an invasive phenotype and the ability of antagonists to reduce that invasive phenotype. By employing MAME cultures of TNBC cell lines that represent different TNBC molecular subtypes in coculture with mammary fibroblasts and CAFS, we were able to test the efficacy of XL184 inhibition in a model that recapitulates the diverse tumor epithelium–stromal interactions of TNBC. This allowed us to evaluate the efficacy of cabozantinib against paracrine MET signaling as both the mammary fibroblasts and CAFs expressed high levels of HGF. We found that cabozantinib reduced the volume of 3D structures including their invasive outgrowths even in the presence of HGF-expressing CAFs.

The exploration of paracrine MET signaling was continued by in vivo analysis of MET inhibition through the use of the hHGF^tg-SCID model. Here we observed that MET^pos TNBC cells are significantly inhibited with cabozantinib. To model the treatment of established human tumors, we delayed XL184 dosing until the tumors had reached 500 mm³ (compared with 100 to 150 mm³ in typical in vivo studies). Even though we observed a significant decrease in tumor growth, there is a clear upward growth trend that begins after the first week of treatment (Fig. 6F). This suggests that there may be therapeutic resistance to targeted MET inhibition in established or late-stage tumors. Resistance to TKI monotherapy is often observed in late-stage cancers. To avoid this therapeutic resistance it will be necessary to identify combination therapies that will target signaling nodes that are essential to tumor survival. One possibility is combination therapy of MET and MEK inhibitors. An increase in MAPK activity was observed in two TNBC cell lines after XL184 treatment (Fig. 1) and previously we observed increased MAPK activity in MET^pos cells treated with MET shRNA in HER2⁺ breast cancers (21). MAPK activity may be increased in response to MET inhibition through several mechanisms, including activation of other RTKs (6, 7). Further studies are necessary to determine which receptors may be compensating for MET inhibition and if targeting MET and MEK in combination in TNBC is more effective.

Our findings and other studies suggest that MET inhibitors are most effective against high MET-expressing tumors (43, 44). Those studies focused on the primary tumor environment, yet MET may also be a critical therapeutic target in metastatic disease and chemotherapy-treated patients. Hypoxia and HIF1 expression are known to induce MET expression and activation of the downstream MET signaling cascades (45, 46). Because most primary and metastatic tumors contain hypoxic, avascular regions, MET inhibition also may be advantageous in combination with chemotherapy or in recurrent tumors that moderately express MET, but have high levels of hypoxia (47).

The present results are the first time that MET inhibition has been evaluated as a targeted therapeutic approach in vitro and in vivo in TNBC. Both our in vitro and in vivo preclinical models demonstrate that cabozantinib treatment is an effective therapeutic strategy in TNBC. Considering that TNBC is a highly heterogeneous disease, it is unlikely that a single targeted therapy will be effective for all TNBC patients. Therefore, it is critical that we develop multiple targeted therapeutic approaches in concert with diagnostic tests in order to identify TNBC patients who will benefit from specific therapeutic strategies. The necessity of companion diagnostics in the clinic has been underscored by the failure of TKIs in clinical trials, particularly MET inhibitors (48). These failures demonstrate that not only do we need clear diagnostic thresholds defined by receptor expression or activation, but we also need to understand the overall complexity of the kinase signaling networks. Because these signaling networks are heavily influenced by the TME, it is essential that preclinical studies utilize models, such as MAME, that can recapitulate this milieu. The combination of enhanced in vitro and in vivo models, companion diagnostics, and rationale clinical trial design will accelerate our development of successful targeted therapies.

C.R. Graveel reports receiving commercial research grants from Mirati Therapeutics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: C.R. Graveel, B.F. Sloane

Development of methodology: M. Sameni, E.A. Tovar, A. Chalasani, A. Anbalagan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Sameni, E.A. Tovar, C.J. Essenburg, A. Chalasani, E.S. Linklater

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Sameni, A. Borgman, D.M. Cherba, M.E. Winn, C.R. Graveel, B.F. Sloane

Writing, review, and/or revision of the manuscript: M. Sameni, E.S. Linklater, M.E. Winn, C.R. Graveel, B.F. Sloane

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Borgman, A. Anbalagan, M.E. Winn

Study supervision: C.R. Graveel, B.F. Sloane

The authors thank Lisa Turner for her excellence in histopathology and Bryn Eagleson, Sylvia Timmer, and the VARI Vivarium for their continuous dedication. The authors also thank Dafna Kaufman for her assistance with this study. Confocal microscopy was performed in the Microscopy, Imaging and Cytometry Resources Core of Wayne State University, which is supported by NIH Center grant P30 CA022453 to the Karmanos Cancer Institute, and the Perinatology Research Branch of the National Institutes of Child Health and Development. Cell line authentication was performed through the Biobanking and Correlative Sciences Core and the Applied Genomics Technology Center of Wayne State University, which are supported in part by NIH Center grant P30 CA022453 to the Karmanos Cancer Institute. The authors are also grateful to Dan DeSantis for his assistance with image processing.

This study was funded by the Department of Defense (DoD) W81XWH-11-1-0050 (B.F. Sloane and C.R. Graveel) and the Breast Cancer Research Foundation (C.R. Graveel).

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.