Dual effects of a targeted small-molecule inhibitor (cabozantinib) on immune-mediated killing of tumor cells and immune tumor microenvironment permissiveness when combined with a cancer vaccine

We used the MC38-CEA murine colon carcinoma cell line which expresses both VEGFR2 and MET (Figure 1A, inset), in addition to the human carcinoembryonic antigen (CEA), to examine the in vitro immunomodulatory effects and in vivo antitumor effects of cabozantinib. First, we determined the effect of cabozantinib on the proliferation of MC38-CEA cells. MC38-CEA cells were exposed to 2.5 μg/mL cabozantinib for 24, 72, or 120 h to model the steady-state plasma concentration achievable in humans [3]. After the designated period of treatment, cells were harvested and counted, and their viability was measured. Cabozantinib significantly reduced the proliferation of MC38-CEA cells after 72 and 120 h (Figure 1A). However, despite this reduction, the MC38-CEA cells continued to proliferate and their viability remained >75% at all time points, regardless of treatment. We therefore used this dose of cabozantinib for all subsequent in vitro studies.

It has been previously shown that radiation and chemotherapy can alter the phenotype of tumor cells, rendering them more sensitive to T cell-mediated killing [31],[32],[36]. To determine if cabozantinib could modify the expression of cell-surface markers that influence immune recognition, we treated MC38-CEA cells with cabozantinib for 24 h, then stained and analyzed them by flow cytometry. Cabozantinib significantly upregulated the expression of MHC-I molecules on both the population level (H-2D^b) and a per-cell basis (H-2K^b), increasing the potential for antigen presentation and T cell recognition of the tumor cells (Figure 1B). Cabozantinib treatment also increased the percentage of MC38-CEA cells expressing ICAM-1, Fas, and calreticulin, cell-surface markers that also aid in T cell recognition, adhesion, and stimulation (Figure 1B). Prior studies have suggested that altering the expression of any one of these markers was capable of making tumor cells more amenable to T cell-mediated killing [31],[32],[36]-[38].

To determine if cabozantinib treatment could increase the sensitivity of MC38-CEA cells to T cell-mediated lysis, we treated cells for 24 h, then used them as targets in CTL killing assays. Cabozantinib treatment significantly increased the sensitivity of MC38-CEA cells to T cell-mediated lysis by cytotoxic T lymphocytes (CTLs) specific for CEA (P =0.0004) (Figure 1C) as well as the endogenously expressed tumor antigen gp70 (P =0.0075) (Figure 1D). We next investigated whether cabozantinib treatment made tumor cells more susceptible to killing by other therapies or if their increased susceptibility was specific to T cell-mediated killing. We treated MC38-CEA cells with cabozantinib for 24 h, then exposed them to 0 or 5 Gy external-beam radiation and replated them in the absence of cabozantinib. Cells were counted and assayed for viability 24, 48, and 72 h post-irradiation. Administration of external-beam radiation further reduced the growth of cabozantinib-treated MC38-CEA cells, but did not reduce their viability as the cells remained >75% viable regardless of treatment (Additional file 1). Taken together, these data suggested that cabozantinib was capable of specifically altering tumor cells in ways that made them more amenable to immune mediated attack.

To assess the impact of cabozantinib treatment on the peripheral immune environment, CEA-Tg C57/BL6 mice were fed a cabozantinib-compounded diet for up to 35 days. After 10 days on this diet, mice achieved a cabozantinib serum concentration of 2.77 ± 1.14 μg/mL, which was similar to the steady-state serum concentration achieved at the maximum tolerated dose in a phase I human clinical trial (2.31 μg/mL; Figure 2A) [3]. At 35 days, cabozantinib treatment did not affect the total number of splenocytes in treated mice (Figure 2B), nor did it significantly increase the percentage of CD4⁺ T cells in the spleen (P =0.1585) (Figure 2C). Cabozantinib treatment did, however, increase the percentage of CD8⁺ T cells (P =0.0198) (Figure 2D). In addition, treatment with cabozantinib reduced the percentage of T regulatory cells (Tregs) (P =0.0152) (Figure 2E), and myeloid-derived suppressor cells (MDSCs) (P <0.0001) in the spleen (Figure 2F). As a result of these alterations, CD4:Treg/MDSC ratios improved from 13.34 to 16.43 and 4.14 to 7.59, respectively even in the absence of a significant change in the level of CD4⁺ T cells. CD8:Treg/MDSC ratios improved from 8.24 to 11.91 and 2.54 to 5.50, respectively, due to the significant alteration of CD8⁺ T cells, Tregs, and MDSCs. These changes in immune-cell composition were evident as early as 10 days after initiating cabozantinib treatment. In addition, after 10 days of cabozantinib treatment we observed a transient increase in the percentage of CD4⁺ T cells (19.90% ±0.76 vs. 27.44% ±0.65; P <0.0001); however, this increase waned by day 35.

After determining that cabozantinib could alter the peripheral immune landscape, we next combined it with a poxviral-based cancer vaccine to determine if the combination could alter the function of immune cells as well. Cabozantinib and MVA/rF-CEA/TRICOM were administered to CEA-Tg C57/BL6 mice as indicated in Figure 3A. After 35 days, splenocytes were harvested and Tregs were purified to assay their ability to regulate the proliferation of naïve CD4⁺ T cells. Compared to Tregs from control mice, Tregs from mice treated with cabozantinib alone demonstrated a significantly reduced ability to regulate the proliferation of CD4⁺ T cells (P =0.0183) (Figure 3B). Surprisingly, the regulatory capacity of Tregs from mice treated with MVA/rF-CEA/TRICOM alone was also significantly reduced (P =0.0003) (Figure 3B). However, the combination of cabozantinib and MVA/rF-CEA/TRICOM completely abrogated Treg function, as CD4⁺ T-cell proliferation in the presence of Tregs from these mice was not significantly different from that in the absence of Tregs (P =0.0775) (Figure 3B). Splenocytes from these mice were also analyzed for CEA-specific cytokine production. Splenocytes from mice treated with the combination of cabozantinib and MVA/rF-CEA/TRICOM produced significantly more interferon (IFN)-γ (Figure 3C; P =0.0188) and tumor necrosis factor (TNF)-α (Figure 3D; P =0.0015) than splenocytes from control mice. In addition, splenocytes from mice treated with the combination produced significantly more IFN-γ than splenocytes from mice receiving either cabozantinib alone (P =0.0463) or MVA/rF-CEA/TRICOM alone (P = 0.0020). Splenocytes from combination-treated mice also produced significantly more TNF-α than splenocytes from mice receiving MVA/rF-CEA/TRICOM alone (P =0.0022).

It has been reported that cabozantinib reduces tumor vascularity, primarily through the inhibition of VEGF receptors [1],[39],[40]. We sought to confirm this observation in the MC38-CEA model and determine whether cabozantinib could also affect infiltration of immune cells into the tumor microenvironment when combined with a cancer vaccine. We implanted CEA-Tg C57/BL6 mice with MC38-CEA cells and treated them as indicated in Figure 4A. Immunohistochemical analysis indicated that cabozantinib treatment significantly reduced the vascular density of MC38-CEA tumors compared to tumors from control-treated mice (P =0.0166), and this reduction was maintained when cabozantinib was combined with MVA/rF-CEA/TRICOM (P =0.0249) (Figure 4B and C). In addition, tumors from mice treated with the combination of cabozantinib and MVA/rF-CEA/TRICOM showed significantly increased infiltration of CD3⁺ lymphocytes (P =0.0094) (Figure 4B and D, insert). Tumors from these mice did not show a significant increase in CD4⁺ T-cell infiltration (Figure 4B and E), but did show a significant increase in the number of infiltrating CD8⁺ T cells (P =0.0310) (Figure 4B and F). Flow cytometric analysis revealed that the tumors from mice receiving cabozantinib either alone or in combination with vaccine had a significant increase in Treg infiltration (P =0.01-0.005, P =0.002-0.001, respectively) (Figure 5A). However, cabozantinib and, to a further extent, the combination of cabozantinib and MVA/rF-CEA/TRICOM reduced the infiltration of MDSCs (P =0.01-0.005, P =0.005-0.002, respectively) (Figure 5B). Treatment with either cabozantinib or MVA/rF-CEA/TRICOM alone reduced the percentage of infiltrating tumor-associated macrophages (TAMs) (P =0.01-0.005), but again, the combination further reduced TAM infiltration (Figure 5C). Taken together, these data show that cabozantinib induces a more immune-permissive environment, both in the periphery and at the tumor site, and that this altered environment may be exploited by combining cabozantinib with a cancer vaccine.

We evaluated the antitumor activity of cabozantinib in the MC38-CEA tumor model with and without the addition of MVA/rF-CEA/TRICOM (Figure 6A). Cabozantinib treatment alone significantly reduced the growth rate of MC38-CEA tumors compared to control (Figure 6B and D). In addition, cabozantinib induced tumor regression in one mouse; this tumor reappeared, however, by day 33. The combination of cabozantinib and MVA/rF-CEA/TRICOM significantly reduced the growth rate of MC38-CEA tumors compared to control and MVA/rF-CEA/TRICOM alone. Additionally, 50% of the mice treated with this combination had durable regression of their tumors and remained tumor-free through day 35 (Figure 6E). Administration of cabozantinib prior to MVA/rF-CEA/TRICOM did not further improve the antitumor efficacy of the combination (data not shown). Also, discontinuing cabozantinib after 10 days of treatment, or administering a lower dose (3 mg/kg bw/day), reduced its antitumor activity as a single agent and eliminated its synergy with MVA/rF-CEA/TRICOM (data not shown). Depleting either CD4⁺ or CD8⁺ T cells also abrogated the antitumor activity observed when cabozantinib was combined with MVA/rF-CEA/TRICOM suggesting a role for both T cell subsets in the increased efficacy of the combination treatment (Figure 6F).

Eight- to 12-week-old female C57/BL6 mice were obtained from the National Cancer Institute's Frederick Cancer Research Facility, Frederick, MD. CEA-transgenic (CEA-Tg) mice homozygous for the expression of human CEA were generously provided by Dr. John Shively (Beckman Research Institute, City of Hope National Medical Center, Duarte, CA) and were bred and maintained at the National Institutes of Health (Bethesda, MD) [50].

Murine colon carcinoma MC38 cells expressing human CEA (MC38-CEA) were generated by retroviral transduction of MC38 cells with CEA cDNA, as previously described [51]. MC38-CEA cells were cultured in complete medium (DMEM supplemented with 10% fetal bovine serum, 2 mM glutamine, 1 mM HEPES buffer, 50 μg/mL gentamicin, 100 IU/mL penicillin, 100 μg/mL streptomycin, and 300 μg/mL G418) at 37°C/5% CO₂.

For in vitro studies, cabozantinib malate salt (Exelixis Inc., South San Francisco, CA) was dissolved in DMSO (vehicle) at 1.0 mg/mL and stored at -20 C. A dose of 2.5 μg/mL was used for all in vitro experiments. For in vivo studies, cabozantinib was admixed into standard rodent diet (Research Diets, New Brunswick, NJ) at a concentration of 66.7 mg/kg of diet in order to deliver 10 mg/kg bw/day to the animals. To determine the steady-state serum level of cabozantinib achieved by delivering the drug via rodent diet, CEA-Tg C57BL/6 mice (n =5) were fed control diet or cabozantinib-containing diet for 10 days. Peripheral blood, acquired by retro-orbital bleeding, was analyzed for cabozantinib concentration by LC-MS.

Modified vaccinia Ankara (MVA) and recombinant fowlpox (rF) viruses containing transgenes for the murine costimulatory molecules B7-1, ICAM-1, and LFA-3 (designated TRICOM) in combination with the human CEA transgene (MVA/rF-CEA/TRICOM) have been previously described [52]. For in vivo studies, a priming dose of MVA-CEA/TRICOM was administered s.c., with weekly boosts of rF-CEA/TRICOM, both at 1 × 10⁸ plaque-forming units/mouse/dose.

MET and VEGFR2 expression was determined by western blot using rabbit polyclonal antibodies to MET and VEGFR2 (Abcam, Cambridge, MA). MC38-CEA cells were lysed using Cell Lysis Buffer containing 1 mM PMSF (Cell Signalling, Danvers, MA) and 10 μL/mL HALT Protease/Phosphatase Inhibitor Cocktail (Thermo Scientific, Rockford, IL) according to the manufacturer's protocol, blots were imaged using the Odyssey Infrared Imaging System (LI-COR Biosciences, Lincoln, NE).

To analyze the effect of cabozantinib on in vitro expression of immune-relevant proteins, MC38-CEA cells were treated with cabozantinib or vehicle for 24 h, then stained with the following antibodies: CD66e/CEA-FITC (AbD Serotec, Raleigh, NC), H-2K^b-APC (eBioscience, San Diego, CA), H-2D^b-PE, CD54/ICAM-I-PE, CD95/Fas-FITC (BD Biosciences, San Jose, CA) and calreticulin-APC (R&D Systems, Minneapolis, MN). 7AAD (BD Biosciences) staining was used to determine cell viability. Cells were incubated with the antibodies for 30 min at 4°C, acquired on an LSR II flow cytometer (Becton Dickinson, Franklin Lakes, NJ), and analyzed using FlowJo software (TreeStar, Inc., Ashland, OR).

To evaluate cabozantinib's ability to alter the sensitivity of MC38-CEA cells to lysis by CTLs, cells were treated with cabozantinib, vehicle or left untreated for 24 h, after which they were harvested and used as targets in a standard CTL assay. Cells were labeled with ¹¹¹In-labeled oxyquinoline (Medi-Physics Inc., Arlington Heights, IL) and coincubated in 96-well round-bottom plates at 37°C/5% CO₂ with T572- or gp70-specific effector cells in the absence of cabozantinib at an effector:target ratio of 30:1. The H-2D^b-restricted CEA-specific CD8⁺ CTL line, T572, recognizes the peptide epitope CEA_572-579, as previously described [53]. The H-2K^b-restricted gp70-specific CD8⁺ CTL line, gp70, recognizes the peptide epitope p15e_604-611 of glycoprotein 70 of an endogenous murine retrovirus, as previously described [54]. After 18 h, supernatants were harvested and analyzed for the presence of ¹¹¹In using a WIZARD2 Automatic Gamma Counter (PerkinElmer, Waltham, MA). The percentage of tumor lysis was calculated as follows: % tumor lysis = [(experimental cpm - spontaneous cpm)/(maximum cpm - spontaneous cpm)] × 100.

To assess the effect of cabozantinib and MVA/rF-CEA/TRICOM on immune cell function, CEA-Tg C57BL/6 mice were divided into 4 groups (n =5/group): (a) control, (b) cabozantinib alone, (c) MVA/rF-CEA/TRICOM alone, and (d) cabozantinib + MVA/rF-CEA/TRICOM. Mice receiving cabozantinib were fed cabozantinib-containing diet on days 0-35. Mice receiving MVA/rF-CEA/TRICOM received a priming vaccination with MVA-CEA/TRICOM on day 0, then booster vaccinations with rF-CEA/TRICOM on days 7 and 14. Spleens were harvested on day 35. To examine the function of Tregs, splenocytes were purified by Histopaque gradient (Sigma-Aldrich, St. Louis, MO), then CD4⁺CD25⁺ Tregs were purified using a Regulatory T Cell Isolation Kit (Miltenyi Biotec, Auburn, CA) according to the manufacturer's instructions. Purified Tregs were plated with antigen presenting cells (APCs), irradiated (30 Gy) splenocytes from naïve C57BL/6 mice, CD4⁺ T cells, purified from naïve C57BL/6 mice using a CD4+ T Cell Isolation Kit (Miltenyi Biotec), and anti-CD3 cross-linking antibody (BD Biosciences) for 72 h. [³H]-thymidine was added to the culture for the last 12 h of incubation. Control wells containing Tregs, APCs, and anti-CD3 without CD4⁺ cells were used to determine the background level of Treg proliferation. Control wells containing CD4⁺ cells and Concanavalin A (Sigma-Aldrich) were used to determine the maximum level of CD4⁺ cell proliferation. Cells were harvested using a Tomtec cell harvester (Wallac Inc., Gaithersburg, MD) and [³H]-thymidine incorporation was measured using a Wallac 1205 Betaplate MicroBeta Counter (Wallac Inc.) To examine cytokine production by effector T cells, splenocytes were incubated with CEA_572-579 peptide (GIQNSVSA) (CPC Scientific, Sunnyvale, CA) for 7 days then purified on a Histopaque gradient. Effector cells were plated with APCs in the presence of CEA₅₇₂ peptide for 24 h. Supernatants were collected and analyzed for the presence of cytokines using a mouse Th1/Th2 Cytometric Bead Array (BD Biosciences). Data were acquired using a FACScan flow cytometer and analyzed using BD CBA analysis software (Becton Dickinson).

To investigate the effect of cabozantinib and MVA/rF-CEA/TRICOM on tumor vascularity and immune cell infiltration, CEA-Tg C57BL/6 mice were implanted with 3 - 10⁵ MC38-CEA cells s.c. in the right flank. Fourteen days post tumor implantation, when tumors were established (~300 mm³), mice were divided into 4 groups (n =2/group): (a) control, (b) cabozantinib alone, (c) MVA/rF-CEA/TRICOM alone, and (d) cabozantinib + MVA/rF-CEA/TRICOM. Mice treated with cabozantinib were fed cabozantinib-containing diet on days 14-28. Mice treated with MVA/rF-CEA/TRICOM received a priming vaccination with MVA-CEA/TRICOM on day 14 and a booster vaccination with rF-CEA/TRICOM on day 21. On day 28, tumors were harvested with a portion of each tumor being fixed with Z-Fix (Anatech Ltd., Battle Creek, MI), frozen in OTC (Electon Microscopy Sciences, Hatfield, PA), or dissociated according to the protocol for preparation of single-cell suspensions from implanted mouse tumors (Miltenyi Biotec). Fixed tumor sections were stained with antibodies to von Willebrand factor at 1:1200 and CD3 at 1:1000 (Dako, Carpinteria, CA). Frozen tumor sections were stained with rat monoclonal antibodies to CD4 and CD8 at 1:1500 and 1:700, respectively (Novus Biologicals, Littleton, CO). Control sections were stained with matched isotype antibodies. Entire tumor sections were digitally scanned by an Aperio ScanScope CS scanning system and analyzed by Aperio ImageScope Viewer software (Aperio Technologies Inc., Vista, CA) excluding necrotic regions. Positive tumor regions were determined using the Microvessel Analysis v1 and Positive Pixel Count v9 algorithms, respectively and are depicted as number of vessels or cells/tumor. Dissociated tumor was stained for flow cytometric analysis using the same antibodies and protocol as the splenic immune cell population analysis with the inclusion of CD45-BV605 (ebioscience).

To examine the antitumor effects of cabozantinib and MVA/rF-CEA/TRICOM, CEA-Tg C57BL/6 mice were implanted with 3 x 10⁵ MC38-CEA cells s.c. in the right flank on day 0. On day 4, when tumors became palpable, mice were divided into 4 groups (n =10/group): (a) control, (b) cabozantinib alone, (c) MVA/rF-CEA/TRICOM alone, and (d) cabozantinib + MVA/rF-CEA/TRICOM. Mice treated with cabozantinib were fed cabozantinib-containing diet on days 4-35. Mice treated with MVA/rF-CEA/TRICOM received a priming vaccination with MVA-CEA/TRICOM on day 4, then weekly booster vaccinations with rF-CEA/TRICOM. Tumor dimensions were measured twice weekly and tumor volumes were calculated as follows: (length × width²)/2. For depletion studies, MC38-CEA-bearing CEA-Tg C57BL/6 mice (n =8/goup) were given either anti-CD4 (GK 1.5) or anti-CD8 (Lyt 2.2) hybridomas i.p. on days 1-3, then at weekly intervals, or left untreated. Depletion was confirmed by flow cytometric analysis of peripheral blood. All mice, except for those in the control group, received cabozantinib-containing diet on days 4-33, a priming vaccination with MVA-CEA/TRICOM on day 4, then weekly booster vaccinations with rF-CEA/TRICOM.