Oncolytic Herpes Virus Armed with Vasculostatin in Combination with Bevacizumab Abrogates Glioma Invasion via the CCN1 and AKT Signaling Pathways

Gliomas represent about 30% of primary brain tumors. Despite numerous efforts to develop new treatments for malignant gliomas, therapeutic options remain limited and the prognosis is still poor (1, 2). Temozolomide is the only agent validated for its effectiveness on overall survival, and its concomitant use with radiotherapy is the standard therapy for malignant glioma (3). Many investigators continue to seek novel therapeutic approaches for glioma including surgery, chemotherapy, radiotherapy, immunotherapy, and combination therapies.

Antiangiogenic therapy is one of the strategies used to treat glioblastoma. Glioblastoma cells secrete high levels of VEGF. Bevacizumab binds to all VEGF isoforms, causing reduced tumor vascularization, reduced vascular permeability, and the inhibition of tumor growth (4). Bevacizumab, which targets proangiogenic VEGF, is a recombinant humanized mAb that was approved as a chemotherapeutic agent for primary and recurrent glioblastoma in Japan. Its clinical use is increasing, even though its advantages on overall survival were lacking in previous trials (5, 6). Recent studies indicated that anti-VEGF therapy induced glioma invasion via several mechanisms including the integrin-related pathway (7, 8), indicating it is important to test the potential uses of bevacizumab in combination therapies.

Oncolytic viral (OV) therapy has appeared as a promising treatment modality that utilizes the tumor-specific properties (9). Oncolytic herpes simplex viruses (HSV) is designed to replicate and have cytotoxicity selectively in tumor cells, but not in normal tissues. Oncolytic HSVs include genetically engineered viruses such as talimogene laherparepvec, and a spontaneously mutated virus without the insertion of foreign genes, such as HF10 (10). Intralesional talimogene laherparepvec administration improved durable response rates in a randomized phase III trial (11), for which the accelerated FDA approved to use oncolytic HSVs for patients with recurrent melanoma. Phase I and II trials of HF10 in patients with recurrent metastatic breast carcinoma, recurrent head and neck squamous cell carcinoma, advanced pancreatic carcinoma, refractory and superficial cancers, and melanoma have been successfully conducted (10). There are several challenges regarding oncolytic HSVs, such as their rapid clearance by host immune responses, and limited intratumoral spread of the virus. To overcome these challenges, genetic engineering of OVs or combination therapy with OVs and systemic treatments such as molecular targeting drugs have been suggested (12–17).

Vasculostatin (Vstat120), the extracellular fragment of brain-specific angiogenesis inhibitor 1 (BAI-1), is a potent antiangiogenic and antitumorigenic factor (18, 19). Vasculostatin contains an integrin-antagonizing RGD (Arg-Gly-Asp) motif, 5 thrombospondin type 1 repeats, a GPS (G-protein–coupled receptor proteolytic site) domain, and 7-transmembrane domains (18, 20). The BAI-1 expression is negatively correlated with pathologic grading, angiogenesis, and brain edema in gliomas (21). A vasculostatin-armed oncolytic HSV-1, termed rapid antiangiogenesis mediated by oncolytic virus (RAMBO), significantly suppressed intracranial and subcutaneous glioma growth in mouse glioma models compared with control virus (12, 13). Furthermore, Fujii and colleagues reported the efficacy of combination therapy with cyclic RGD peptide and RAMBO for malignant glioma (12). We hypothesized that bevacizumab and RAMBO combination therapy has a synergic effect, because vasculostatin expressed by RAMBO might antagonize integrin-related pathways induced by bevacizumab.

In this study, we evaluated RAMBO and bevacizumab combination treatment of glioma. RAMBO reduced bevacizumab-induced glioma invasion with vasculostatin expressed by RAMBO-infected glioma cells. Evaluation of the invasive mechanism revealed the decreased activation of AKT signaling pathways in cells treated with combined RAMBO and bevacizumab.

U87ΔEGFR was initially engineered by the Cavenee laboratory at the Ludwig Institute for Cancer Research, New York, NY. U251MG was obtained from Dr. Balveen Kaur at Ohio State University, Columbus, OH. U87MG was obtained from the ATCC. Vero cells were purchased from ATCC and used for viral replication. Glioma cells and Vero cells were prepared and maintained as described previously (14). MGG23 was provided by Dr. Hiroaki Wakimoto and cultured as described previously (22, 23). All cells were cultured at 37°C in an atmosphere containing 5% CO₂. U87ΔEGFR, U251MG, and U87MG were authenticated by Promega via short tandem repeat profiling in December 2016. Mycoplasma was negative in all cells.

Bevacizumab was purchased from Genentech/Roche/Chugai Pharmaceutical Co.

The construction and efficacy of HSVQ, a first-generation OV deleted for both copies of ICP34.5 and disrupted for ICP6, and RAMBO, a Vstat120-expressing OV within the context of HSVQ1, have been previously described (13, 14, 17, 24, 25). HSVQ1 was engineered by the Chiocca laboratory, and RAMBO was originally engineered by the Chiocca and Kaur laboratories.

The cytotoxicity of U87ΔEGFR, U251MG, U87MG, and MGG23 glioma cells were analyzed using the water-soluble tetrazolium (WST)-1 according to the manufacturer's instructions (Roche Molecular Biochemicals). We performed WST-1 quantitative colorimetric assay for cell survival as previously described (12).

U87ΔEGFR, U251MG, and U87MG glioma cells were infected with RAMBO or HSVQ dissolved in DMEM with 0.1% FBS at MOI 2, and conditioned medium (CM) was harvested 14 hours later by centrifugation, as reported previously (12).

The scratch wound assay was performed as previously described (12, 26). Glioma cells were exposed to bevacizumab from 72 hours before assessment. Medium was changed to CM or DMEM with 0.1% FBS, and the indicated concentration of bevacizumab was added. Glioma cells were assessed by counting migrating cells in the area of the gap every 6 to 24 hours (Keyence).

An in vitro migration assay was performed using a 24-well plate and ThinCert (8 μm pore, 24-well format, Greiner Bio-One) according to the manufacturer's instructions, as reported previously (26, 27).

The in vitro invasion assay was performed using a BioCoat Matrigel invasion chamber (24-well format; Corning Incorporated) according to the manufacturer's instructions. A total of 5 × 10⁴ cells were seeded in CM or DMEM with 0.1% FBS in the upper chamber, followed by treatment with bevacizumab or PBS, as described previously (26, 27).

In another in vitro invasion assay, MGG23 cells were seeded in a 96-well ultra-low attachment plate (Costar; Corning Incorporated) at a density of 1.0 × 10³ cells/well in 25 μL of medium, followed by treatment with viruses and bevacizumab to the indicated wells. After centrifugation to assemble all the cells to the center, matrigel (25 μg/insert; Becton Dickinson) was added to each well. Digital photomicrographs of the midplane of spheroids were taken daily with a BZ-8100 microscope (Keyence). Core and invasive diameter were measured using ImageJ (http://rsb.info.nih.gov/ij/) and the radius of invasion was calculated, as previously described (28).

All experiments were conducted in accordance with the guidelines of the Okayama University Animal Research Committee. All procedures and animal protocols were approved by the Committee on the Ethics of Animal Experimentation at Okayama University, as described previously (27). U87ΔEGFR cells were injected into athymic mice (CLEA Japan Inc.), and MGG23 cells were injected into severe combined immunodeficiency mice (Charles River Laboratories Japan, Yokohama, Japan), respectively. Glioma cells (2 × 10⁵ cells) were stereotactically injected into the right frontal lobe, as previously described (7). Five days after implantation of the glioma cells, mice were treated with bevacizumab at the indicated concentration or PBS intraperitoneally twice a week. Seven days after inoculation of the glioma cells, anesthetized mice were stereotactically injected with the indicated plaque forming units of RAMBO at the same location as the tumor.

In both mouse glioma models, the survival time was assessed with a Kaplan–Meier survival analysis. U87ΔEGFR harboring mice were sacrificed 18 days after tumor implantation or if they showed signs of morbidity for pathological analysis, qRT-PCR, and Western blotting. MGG23 harboring mice were sacrificed 50 days after tumor implantation for pathologic analysis.

Surgically excised brains from mouse glioma models were fixed with 4% paraformaldehyde, embedded in paraffin, and 4-μm sections were prepared. IHC analyses were carried out as described previously (7, 25). Antihuman leukocyte antigen mAb (1:100 dilution; Abcam Inc.) was used for the staining, and mouse immunoglobulin was used as a negative control. The sections were stained with Dako Envision + System-HRP Kit in accordance with the manufacturer's protocol (DakoCytomation), and were counterstained with hematoxylin. IHC samples were observed with a BZ-8100 microscope.

We isolated total RNA from the cell lines or tumor specimens. Syntheses of cDNA and qRT-PCR procedures were conducted as described previously (26, 29). As an internal control, we used GAPDH mRNA. The primer sequences used were as follows: Human C-terminal Src kinase (CSK) primers: forward, gacgtgtggagtttcggaat; reverse, agctgctctcggagctgtag. Human SHC (Src homology 2 domain containing) transforming protein 3 (SHC3) primers: forward, agagtgtggaaggctcagga; reverse, gtgctttttcagcgagaacc. Human protein tyrosine kinase 2 (PTK) primers: forward, cttctgcagtttccccagag; reverse, ccaggtggttggctcactat. Human Caveolin 3 (CAV) primers: forward, tttgccaagaggcagctact; reverse, accctttactggagccacct. Human Son of sevenless homolog 1 (SOS1) primers: forward, ccttgcttgaggttttctgc; reverse, gcagatgctgatgaaccaga. Human cysteine rich protein 61 (CCN1) primers: forward, cctcgcatcctatacaacccttta; reverse, gattctgacactcttctcccttgt. Human GAPDH primers: forward, gacctgccgtctagaaaaacc; reverse, gctgtagccaaattcgttgtc.

We prepared cell lysates and proteins using RIPA buffer and phenyl-methylsulfonyl fluoride (Cell Signaling Technology), as described previously (28). Then, we performed Western blotting as described previously (15, 27). After blocking, membranes were incubated overnight with primary antibodies (anti-CYR61, 1:100, Novus Biologicals; anti-AKT, 1:1,000, Cell Signaling Technology; anti-p-AKT, 1:1,000, Cell Signaling Technology; and anti-GAPDH, 1:1,000, Cell Signaling Technology; anti-BAI1, 1:200, WuXi Biosciences) at 4°C. The secondary antibodies used were horseradish peroxidase (HRP)–conjugated antimouse IgG and HRP-conjugated anti-rabbit IgG (Cell Signaling Technology; 1:5,000). HRP signals were analyzed by the VersaDoc molecular imaging system (Bio-Rad).

The changes in cell death, migration, and invasion were analyzed using 1-way ANOVA followed by Tukey post hoc test. Kaplan–Meier survival curves were compared using the log-rank test. Data on mRNA expression obtained by quantitative real-time PCR were analyzed by 1-way ANOVA followed by Scheffe post hoc test. Data on protein expression obtained by Western blotting were analyzed using ANOVA followed by Tukey post hoc test. All statistical analyses were performed using SPSS statistical software (version 20; SPSS, Inc.).

The cytotoxic effect of combined bevacizumab and RAMBO on glioma cells was investigated by WST-1 proliferation assay. Glioma cell lines and glioma stem cells were incubated with the indicated concentrations of bevacizumab or RAMBO at the indicated MOI. Treatment with RAMBO decreased viable cells compared with saline as a control in a time-dependent manner. After treatment with RAMBO, U87ΔEGFR cells were aggregated and floated from the dishes, whereas MGG23 cells were dissociated and adhered to the dishes (Fig. 1A). There was a significant decrease in viable cells treated with RAMBO compared with saline treatment of each cell line at 48 and 72 hours (U87ΔEGFR, P < 0.001; U251MG, P < 0.001; U87MG, P < 0.001; MGG23, P < 0.001). However, bevacizumab had no cytotoxic effect against glioma cells and did not increase the cytotoxicity of RAMBO against glioma cells (Fig. 1B).

To examine the in vitro effect of vasculostatin on GBM cell migration over time, we performed a scratch wound assay using bevacizumab and CM. The supernatant of malignant glioma cells infected by RAMBO was centrifuged and filtrated to eliminate virus and cell lysates, then it was used as RAMBO-CM. Infection of each cell line by oncolytic virus was detected by the expression of GFP implanted into the viral sequence (Supplementary Fig. S1A). In the RAMBO-infected glioma cells, the expression of vasculostatin was detected by Western blotting (Supplementary Fig. S1B). Vasculostatin in CM had no cytotoxic effect against glioma cells similar to fresh medium (Supplementary Fig. S1C). The rate of migrating cells was assessed every 6 hours after scratch formation and we performed Giemsa staining 24 hours after scratch formation (Fig. 2A; Supplementary Fig. S2A and S2B). RAMBO-CM significantly reduced the rate of migration of each cell line compared with saline control (U87ΔEGFR: P < 0.001, U251MG: P < 0.001, and U87MG: P < 0.001). Furthermore, the rate of migrating cells induced by bevacizumab treatment was reduced by RAMBO-CM (U87ΔEGFR: P < 0.001, U251MG: P < 0.001, and U87MG: P < 0.001; Fig. 2B). We also performed another migration assay using ThinCert for an enhanced quantitative analysis (Fig. 2C). Bevacizumab significantly increased the migration of each cell line compared with saline control (U87ΔEGFR: P < 0.001, U251MG: P < 0.001, and U87MG: P < 0.001). Furthermore, the rate of migrating cells induced by bevacizumab treatment was reduced by RAMBO-CM (U87ΔEGFR: P < 0.001, U251MG: P < 0.001, and U87MG: P = 0.010; Fig. 2D).

To examine the in vitro effect of vasculostatin on GBM cell invasion, we performed a matrigel invasion assay with a Corning chamber using bevacizumab and CM. The supernatant of malignant glioma cells infected by RAMBO or HSVQ was centrifuged and filtrated to eliminate virus and cell lysate, then they were used as RAMBO-CM or HSVQ-CM. The expression of vasculostatin was detected by Western blotting in RAMBO-infected glioma cells but not in HSVQ-infected glioma cells (Supplementary Fig. S1B). Giemsa staining was performed 24 hours after seeding glioma cells into the upper chamber, and then cells invading through the membrane were counted (Fig. 3A). RAMBO-CM significantly reduced the number of invading cells of each cell line compared with saline control (U87ΔEGFR: P < 0.001, U251MG: P < 0.001, and U87MG: P < 0.001). Furthermore, the invading cells induced by bevacizumab treatment were reduced by RAMBO-CM (U87ΔEGFR: P < 0.001, U251MG: P < 0.001, and U87MG: P < 0.001; Fig. 3B).

To examine the in vitro effect of vasculostatin on GBM stem cell invasion, we performed Matrigel invasion assays (P < 0.01; Fig. 3C). After measurement of the core and invasive diameter, the proportion of invasion was calculated. Bevacizumab significantly increased the proportion of glioma cell invasion compared with saline controls (P = 0.001). Combination therapy with bevacizumab and RAMBO significantly inhibited bevacizumab-induced glioma cell invasion of MGG23 cells (P = 0.001), whereas combination therapy with bevacizumab and HSVQ did not inhibit bevacizumab-induced glioma cell invasion of MGG23 (P = 0.062; Fig. 3D).

The antitumor effect of combination with bevacizumab and RAMBO was tested in mice harboring intracerebral U87ΔEGFR glioma cells. Seven days after tumor cell implantation, we injected RAMBO or HSVQ into the brain tumor at the indicated pfu. Five days after tumor inoculation, bevacizumab was injected into intraperitoneal twice a week (Fig. 4A). The survival of mice in each group (7 mice per group) was compared by Kaplan–Meier analysis.

We assessed the efficacy of combination with RAMBO or HSVQ at 1.0 × 10⁵ pfu and bevacizumab at 10 mg/kg. Mice bearing U87ΔEGFR glioma cells treated with saline, bevacizumab at 10 mg/kg, RAMBO at 1.0 × 10⁵ pfu, HSVQ at 1.0 × 10⁵ pfu and bevacizumab at 10 mg/kg, and RAMBO at 1.0 × 10⁵ pfu and bevacizumab at 10 mg/kg were compared. Control mice treated with PBS had a median survival of 17 days after tumor cell implantation, and mice treated with RAMBO had a median survival of 28 days after tumor cell inoculation that was similar to that of PBS-treated mice. Mice treated with bevacizumab had a median survival of 37 days. Mice treated with HSVQ and bevacizumab combination had a median survival of 46 days, which did not reach statistical significance compared with bevacizumab monotherapy (P = 0.075). However, mice treated with bevacizumab and RAMBO combination had a median survival of 64 days, which was significantly longer than mice treated with PBS, RAMBO alone, bevacizumab alone, and bevacizumab and HSVQ combination (log-rank test: P < 0.001, P < 0.001, P < 0.001, and P = 0.001, respectively; Fig. 4B).

Next, we performed survival analysis using glioma stem cells. We compared immunodeficient mice bearing MGG23 cells treated with saline, bevacizumab at 10 mg/kg, HSVQ at 1.0 × 10⁵ pfu as monotherapy, HSVQ at 1.0 × 10⁵ pfu, and bevacizumab at 10 mg/kg, and RAMBO at 1.0 × 10⁵ pfu and bevacizumab at 10 mg/kg. Control mice treated with PBS had a median survival of 62 days, and mice treated with bevacizumab had a median survival of 61 days. Mice treated with RAMBO as monotherapy had a median survival of 65 days after tumor cell inoculation, which was significantly longer than mice treated with PBS (P = 0.001). Mice treated with HSVQ and bevacizumab combination had a median survival of 65 days after glioma cell implantation, which reached statistical significance compared with bevacizumab monotherapy (P = 0.001). Furthermore, mice treated with bevacizumab and RAMBO combination had a median survival of 70 days, which was significantly longer than mice treated with bevacizumab monotherapy, RAMBO monotherapy, HSVQ and bevacizumab combination, or untreated mice (P = 0.001, P = 0.005, P = 0.001, and P = 0.001, respectively; Fig. 5A).

To address the therapeutic effect against glioma invasion, we evaluated combination therapy with RAMBO at 1.0 × 10⁵ pfu and bevacizumab at 10 mg/kg. RAMBO and bevacizumab were administered using the same schedule as for the survival analysis (Fig. 4A).

Eighteen days after tumor inoculation athymic mice with U87ΔEGFR glioma were sacrificed. IHC staining using antihuman leukocyte antigen was performed, and then glioma invasion was assessed by the distance between the mass edge of tumor and invasive area (Fig. 4C). After treatment with bevacizumab, the tumor border showed tumor invasion. Anti-VEGF therapy with bevacizumab significantly increased cell invasion compared with saline controls (P = 0.010). However, combination therapy with bevacizumab and RAMBO significantly decreased the depth of glioma invasion induced by bevacizumab (P = 0.006, Fig. 4D).

Next, immunodeficient mice harboring MGG23 glioma stem cells were sacrificed at 50 days after tumor implantation, and IHC staining with antihuman leukocyte antigen was performed. MGG23 cells treated with bevacizumab as monotherapy showed a greater invasion to the ipsilateral cerebral cortex adjacent to the injection site and to the contralateral corpus callosum compared with saline controls or the bevacizumab and RAMBO treated group (Fig. 5B). We assessed invasion activity with the number of cells in the ipsilateral or contralateral cerebral cortex, as previously reported (27). There was a significant increase of glioma cells invading into the cerebral cortex in the MGG23 cell treated with bevacizumab group compared with saline controls (ipsilateral cortex: P = 0.016, contralateral cortex: P < 0.001). However, combination therapy with bevacizumab and RAMBO significantly decreased the depth of glioma invasion induced by bevacizumab (ipsilateral cortex: P = 0.002, contralateral cortex: P < 0.001, Fig. 5C). These results indicated that RAMBO reduced invasion with bevacizumab.

To investigate the mechanism of the antitumor effect of combination therapy with bevacizumab and RAMBO, we performed qPCR analysis. We chose the integrin-related cell adhesion pathway and hepatocyte growth factor receptor signaling pathway because we previously reported its relationship to bevacizumab-induced invasion (7). Relative expression levels of CSK, SHC3, PTK, CAV, SOS1 and CCN1 in the U87ΔEGFR mouse model with bevacizumab were upregulated 1.84-, 1.35-, 2.35-, 6.98-, 3.95- and 3.34-fold, respectively compared with the control group. In particular, only CCN1 expression was significantly reduced in tumors treated with bevacizumab and RAMBO as combination therapy compared with those treated with bevacizumab alone (Fig. 6A–F, P < 0.05).

Western blotting was performed to investigate the relationship between CCN1 and the AKT pathway (Fig. 6G). Tumors treated with bevacizumab showed significantly higher CCN1 activation than those treated with saline (P = 0.013) and those treated with bevacizumab and RAMBO as combination therapy (P = 0.001). In addition, tumors treated with bevacizumab showed significantly higher p-AKT at Ser473 than those treated with saline (P = 0.024), but bevacizumab and RAMBO as combination therapy significantly reduced AKT phosphorylation compared with bevacizumab (P < 0.001, Fig. 6H). Full scans of the Western blotting are shown in Supplementary Fig. S3.

These results demonstrated that vasculostatin expressed by RAMBO and ENVE34.5 reduced CCN1 expression and AKT phosphorylation induced by bevacizumab.

In 2009, the FDA conditionally approved bevacizumab for patients with recurrent glioblastoma. Lately, prospective 2 phase III trials of newly diagnosed patients, AVAglio and RTOG 0825, showed that overall survival did not reach statistical significance although these studies decreased the risk of progression-free survival in patients (5, 6). Our data showed that U87ΔEGFR-bearing mice treated with bevacizumab had significantly longer survival than those treated with saline. Although U87dEGFR has a poor-invasive phenotype in contrast to clinical glioblastomas, this cell line has been used in several experimental studies to evaluate glioma invasion. In contrast to U87ΔEGFR, bevacizumab had no significant antitumor effect against MGG23-bearing mice compared with saline, which was similar to the results of multiple phase III clinical trials. A study using a mouse model reported showed that bevacizumab significantly reduced tumor growth (30). Our results showed that invasive activity increased by bevacizumab seemed to counteract the effectiveness of bevacizumab in the diffuse invasion glioma model. Moreover, our experiments using 2 different mouse glioma models indicated that RAMBO inhibited glioma cell invasion induced by bevacizumab, resulting in a synergistic effect.

Previous reports indicated that tumor invasiveness was increased by anti-VEGF therapy (7). de Groot and colleagues described 3 patients who, during bevacizumab therapy, developed infiltrative lesions visible by MRI and reported pair imaging features seen on MRI with histopathologic findings (31). In this report, we showed that glioma migration and invasion were increased by bevacizumab, similar to previous reports (7, 32). Interestingly, our data also showed that invasive activities of glioma cells were increased by bevacizumab both in the poor-invasive model using U87ΔEGFR and in the diffuse invasive model using MGG23, indicating that bevacizumab increased glioma cell invasion regardless of the original invasive activity.

RAMBO is composed of cDNA encoding for human vasculostatin (Vstat120) within the backbone of HSVQ (13). Vasculostatin was reported to enhance the antitumor effect of oncolytic HSV-1 (13, 33). Vasculostatin is an extracellular fragment of brain angiogenesis inhibitor 1, whose expression is reduced in several malignancies (20, 24, 34–36). The reexpression of vasculostatin had an antiangiogenic effect, which enhanced antitumor therapeutic efficacy (9, 37). Vasculostatin was expressed only from RAMBO-infected glioma cells, which indicated that the effect of vasculostatin was only seen in cells or mice treated with RAMBO. Interestingly, combination therapy with RAMBO and bevacizumab but not HSVQ reduced bevacizumab-induced migration and invasion, and prolonged the survival time of glioma-bearing mice compared with combination therapy with HSVQ and bevacizumab. These results indicated that vasculostatin increases antitumor effects by reducing glioma migration and invasion.

The integrin-related cell adhesion pathways were reported to be involved in the mechanism of glioma invasion. DeLay and colleagues revealed a hyperinvasive phenotype, a resistance pattern of glioblastoma, after bevacizumab therapy and which was upregulated with integrin α5 and fibronectin 1 (38). Jahangiri and colleagues showed that c-Met and β1 integrin were upregulated in bevacizumab-resistant glioblastomas (32). We previously reported that bevacizumab treatment led to increased cell invasion via an integrin signaling pathway (7).

Oncolytic HSV-1 therapy increases integrin-activating CCN1 protein in the tumor extracellular matrix. Kurozumi and colleagues reported that the oncolytic HSV-1 infection of tumors induced angiogenesis and upregulated CCN1 (9). Haseley and colleagues reported that CCN1 limited the efficacy of OV therapy via an integrin signaling pathway that mediated activation of a type-I antiviral interferon response (39). RAMBO contains vasculostatin in its construct and has 5 thrombospondin type 1 domains within its N terminal sequence and an integrin antagonizing RGD motif (13, 17, 19, 40, 41). Here, we report that CCN1 expression was upregulated by bevacizumab, and that its upregulation was suppressed by RAMBO. Previous reports showed that HSV-1 without vasculostatin increased CCN1 expression in glioma cells (9). Our results showed that HSV-1 expressing vasculostatin decreased CCN1 expression, indicating that the expression of vasculostatin by oncolytic HSV reduced CCN1 induction by HSV-1 itself and by bevacizumab.

The relationship between CCN1 and the AKT pathway was evaluated previously. In tumor cells, high CCN1 expression was related to high Akt phosphorylation (42). Several reports indicated that targeting CCN1 expression might mediate AKT phosphorylation and tumor cell migration (43, 44). From our data, combination therapy with bevacizumab and RAMBO significantly decreased the phosphorylation of AKT. Paw and colleagues previously reported a relationship between the PI3K–AKT pathway and MMP9 expression, which induced glioma cell invasion (45). Therefore, glioma cell invasion via the CCN1–Akt pathway was reduced by vasculostatin expressing oncolytic virus but induced by bevacizumab.

The efficacy of combination viral therapy and chemotherapy has been reported previously. Cyclic RGD peptide had a synergistic effect with viral therapy including adenovirus and HSV-1 (12, 16). Ikeda and colleagues showed that cyclophosphamide substantially increased herpes viral survival and propagation, leading to neoplastic regression (46). Regarding anti-VEGF therapy, several reports described enhanced viral distribution in tumors (30, 47). In our study, the mechanism of the synergistic effect observed with bevacizumab and RAMBO involved the bevacizumab-enhanced distribution of RAMBO in the tumors, and RAMBO-induced reduction of glioma cell invasion promoted by bevacizumab.

CCN1 interacts with integrins, such as αvβ3, α6β1, αvβ5, and αIIβ3, leading to a wide range of biological activities, including cell adhesion, migration, and invasion (48). In addition, exogenous CCN1 in the glioma ECM orchestrated a cellular antiviral response that reduced viral replication and limited the efficacy of the oncolytic virus (39). In this article, we showed the synergistic effect of combined bevacizumab and RAMBO combination against glioma cells. This synergetic effect might not be clinically relevant because we only used cell lines without heterogeneity, although our survival analysis indicated bevacizumab and RAMBO combination therapy was effective even against a diffuse invading model using glioma stem cells. In the future, we plan to evaluate the effectiveness of bevacizumab or RAMBO combinations using several types of glioma stem cells or primary cultures from patients with glioblastoma, that will be more relevant to clinical trials.

Bevacizumab monotherapy or combination treatment with radiation and/or temozolomide is well tolerated and exhibits modest antitumor activity (6, 49). Although bevacizumab has not been shown to extend overall survival, it may have additional benefits in the setting of immunotherapy (50). Recently, Currier and colleagues reported that the combined effect of oncolytic HSV virotherapy and anti-VEGF antibodies was in part due to the modulation of a host inflammatory reaction to virus (51). In addition, Oka and colleagues reported that CD8- and CD11c-positive cells infiltrated tumors treated with adenovirus vector (15). We intend to evaluate the other combination therapies of bevacizumab and other oncolytic viruses, molecular targeted therapy, and immunotherapy.

Our results indicate that combination therapy with bevacizumab and RAMBO had additional therapeutic effects compared with monotherapy using bevacizumab or oncolytic virus. RAMBO-infected glioma cells significantly reduced glioma migration and invasion induced by bevacizumab both in vitro and in vivo. Combination therapy with bevacizumab and RAMBO significantly increased the antitumor effect in a mouse glioma model. CCN1 expression was modulated by RAMBO to activate or inhibit AKT phosphorylation, which promotes cell migration and invasion.

Our results indicated that vasculostatin-expressing OV therapy enhanced chemotherapy with bevacizumab for malignant glioma by suppressing bevacizumab-induced glioma invasion via the AKT signaling pathway. This may be a potential combination therapy for clinical use in patients with malignant glioma.

No potential conflicts of interest were disclosed.

Conception and design: Y. Tomita, K. Kurozumi, J.Y. Yoo, Y. Hattori, Y. Otani

Development of methodology: Y. Tomita, K. Kurozumi, J.Y. Yoo, T. Ichikawa, Y. Otani

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Tomita

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Tomita, K. Kurozumi, T. Ichikawa, Y. Matsumoto, A. Uneda, Y. Hattori

Writing, review, and/or revision of the manuscript: Y. Tomita, K. Kurozumi, Y. Matsumoto, A. Uneda, T. Oka, B. Kaur

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Hattori, T. Oka

Study supervision: K. Kurozumi, K. Fujii, T. Ichikawa, I. Date

We thank M. Arao and Y. Ukai for their technical assistance. We thank Nancy Schatken, BS, MT (ASCP), from Edanz Group (www.edanzediting.com/ac) for editing a draft of this manuscript. This study was supported by Japan Society for the Promotion of Science to K. Kurozumi (nos. 26462182 and 17K10865).

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.