https://orcid.org/0000-0001-6254-4204
Development of efficacious immunotherapy approaches in glioblastoma remains challenging. The induction of antiglioma immune responses utilizing various oncolytic virus strains has been reported from preclinical and early phase clinical trials.1 The optimal design for efficacy studies of oncolytic virotherapy in glioblastoma patients is under debate, but likely such studies will preferentially be performed in patients with recurrent glioblastoma.
In clinical practice, this group of patients is often treated with the anti-angiogenic vascular endothelial growth factor A (VEGF)–neutralizing antibody bevacizumab for its capacity to improve and maintain quality of life, mainly through anti-edematous effects. Beyond promoting angiogenesis, VEGF exerts immunosuppressive effects,2 which are thought to be potentially relevant in the context of immunotherapy in glioblastoma. For example, in the Re-ACT trial, median overall survival of patients with recurrent glioblastoma was prolonged by 3.2 months upon addition of a tumor-specific vaccine to bevacizumab (hazard ratio [HR] = 0.47, P = 0.021) and benefit was associated with vaccine-induced titers within the experimental arm,3 but in the ACT-IV trial the same vaccine had no effect on outcome as an add-on to standard chemoradiation in patients with newly diagnosed glioblastoma irrespective of titer generation in response to vaccination.4
This led us to explore whether bevacizumab should be incorporated into virotherapy clinical trial designs in glioblastoma. We employed a non-immunogenic genetic model of isocitrate dehydrogenase (IDH) wildtype glioblastoma, which is histologically, immunologically, and molecularly highly similar to its human counterpart (N/tv-a;Cdkn2a-/-;Ptenfl/fl:PDGF,Cre).5 Animal research was prospectively approved by the Institutional Animal Care and Use Committee of the Fred Hutchinson Cancer Research Center (animal assurance institutional #A3226-01; protocol #50842). Mouse glioblastomas were treated with oHSVULBP3, an oncolytic herpes simplex virus 1 strain that was armed with a ULBP3 gene expression cassette to boost anticancer immunity (Fig.1A).5 There was a 2-fold increase in VEGF protein levels upon treatment with oHSVULBP3 versus phosphate buffered saline (PBS) in protein lysates from 5 mouse glioblastomas per group (Fig. 1B). A major proportion of VEGF is, however, immobilized in the extracellular matrix. In order to exert its effects, immobilized VEGF isoforms need to be cleaved by matrix metalloproteinases (MMPs), predominantly MMP3 and MMP9.6 Although there was an ~3-fold increase in MMP9 protein levels upon treatment with oHSVULBP3, our model reflects a scenario of overall low MMP expression (Fig. 1B). In order to model effects of MMP expression on the efficacy of oHSVULBP3, we designed an oHSV vector containing an MMP9 expression cassette in addition to the ULBP3 cassette, yielding MMP9 expression specifically in the small virus-infected foci of the tumor. The survival benefit of tumor-bearing mice obtained with oHSVULBP3 versus PBS (HR = 0.27, P < 0.001) was nearly abolished in mice randomized to intratumoral injection with oHSVULBP3-MMP9 (HR = 0.56, P = 0.039; Fig. 1C). Co-treatment of mice with the murine VEGF neutralizing antibody B20 and intracranial injections of PBS, oHSVULBP3, or oHSVULBP3-MMP9 indicated that VEGF was a major driver of reduced therapeutic efficacy of oHSVULBP3 in the presence of MMP9. Consistent with clinical trials of anti-VEGF treatment in glioblastoma patients, there was no effect of B20 on mouse survival (HR = 0.86, P = 0.71), but combining oHSVULBP3-MMP9 with B20 resulted in a marked survival benefit compared with oHSVULBP3-MMP9 alone (HR = 0.32, log rank P = 0.027; Fig. 1D). By contrast, no effect of B20 was observed upon treatment with oHSVULBP3 lacking the MMP9 expression cassette (HR = 1.11, P = 0.83, data not shown). In line with the notion of an immunosuppressive effect of MMP9 and subsequent VEGF release, nCounter gene set analysis revealed downregulation of the gene sets toll-like receptor signaling (P < 0.001) and T-cell activation and checkpoint signaling (P = 0.001) in tumors treated with oHSVULBP3-MMP9 versus oHSVULBP3 (Fig. 1E, F), indicating inhibitory effects on both the myeloid and lymphocytic compartments. Interestingly, disruption of autocrine VEGF signaling in tumor cells results in loss of MMP9 expression, consistent with a positive feedback loop between VEGF and MMP9.7
Anti-VEGF counteracts MMP9-mediated resistance to virotherapy. (A) Experimental setup. (B) Proteomics analysis (R&D ARY028) of pooled tumor lysates (N = 5 per group) from N/tv-a;Cdkn2a-/-;Ptenfl/fl:PDGF,Cre glioblastomas treated by intracranial injection of PBS or oHSVULBP3 as indicated. (C) Symptom-free survival. PBS, N = 20; oHSVULBP3, (1 × 106 plaque-forming units) N = 20; oHSVULBP3-MMP9, (1 × 106 plaque-forming units) N = 12. Kaplan–Meier curves were compared utilizing the log-rank test. (D) Symptom-free survival. PBS alone N = 7, co-treatment with B20 (anti-VEGF; N = 7, 5 mg/kg i.v. every other day), or upon intratumoral injection with oHSVULBP3-MMP9 (1 × 106 plaque-forming units) alone (N = 5) or in combination with B20 (N = 5). Kaplan–Meier curves were compared utilizing the log-rank test. (E) Volcano plot of differentially downregulated genes analyzed by nCounter analysis of the myeloid innate immunity gene set after treatment with oHSVULBP3 or oHSVULBP3-MMP9. (F) Gene set enrichment analysis of genes downregulated upon injection with oHSVULBP3-MMP9 versus oHSVULBP3 (orange in panel E).
In summary our data suggest a mechanism of resistance to virotherapy that involves VEGF release from the extracellular matrix by MMP9 and support a clinical trial design that incorporates the combination of virotherapy and bevacizumab. The unfavorable effect of MMP9 on the efficacy of virotherapy argues against oncolytic vector designs that incorporate MMP9 cassettes to improve viral spreading.
Funding
This research was supported by P2SKP3_158656 from the Swiss National Science Foundation (to HGW), CA60882-01A1 (to ECH), R01 CA195718-01 (to ECH), U54 CA193461-01 (to ECH), R01 CA175052 (to JCG), R01 CA222804 (to JCG), and P01 CA163205 (to E. A. Chiocca, project 1: JCG) by the National Institutes of Health, and by a grant from Oncorus (to ECH).
Conflict of interest statement. JCG and ECH have received honoraria and research funding from Oncorus. CQ is employed by Oncorus. MW has received honoraria and research funding from Roche. All other authors declare no conflicts of interest.
