Local Delivery of OncoVEX^mGM-CSF Generates Systemic Antitumor Immune Responses Enhanced by Cytotoxic T-Lymphocyte–Associated Protein Blockade

Understanding of the interactions between tumors and the immune system has resulted in breakthrough oncology immunotherapies, including Blincyto, a CD19-CD3 bispecific T-cell engager (BiTE), Opdivo, Tecentriq, and Keytruda, anti–programmed cell death protein (PD)-1 monoclonal antibodies, and Yervoy, an anti–cytotoxic T-lymphocyte–associated protein (CTLA)-4 monoclonal antibody (1). These agents build on clinical and mechanistic data that demonstrate the central role of tumor-specific CD8⁺ T cells in mediating tumor elimination (2–4). Appropriate T-cell activation is regulated by dendritic cells (DC), which detect peripheral tumor antigen, become activated, and migrate to secondary lymphoid tissues to induce antitumor responses (5, 6). Emergent tumors may develop strategies to evade this immune response, including expression of factors that suppress DC/T-cell activity (7–9).

Currently approved immunotherapies can overcome tumor-induced immune suppression by blocking negative immunoregulatory signals [e.g., CTLA-4 (10) or PD-1 (11)]. Oncolytic viruses may complement immunotherapies by selectively killing infected tumor cells (thereby releasing tumor antigens) and promoting specific anti-tumor immunity (by acting as viral adjuvants and by inducing proinflammatory tumor cell death); such activities are further enhanced by expression of transgenes designed to promote adaptive immune responses (12).

The available clinical evidence supports the safety of intratumorally administered oncolytic viruses, with modest side effects (13–16). Talimogene laherparepvec was designed to kill neoplastic cells by direct lysis and stimulation of a tumor antigen–specific adaptive immune response. Removal of the ICP34.5 gene from herpes simplex virus type 1 (HSV-1) confers tumor-selective replication and markedly reduces neurovirulence (17–20). Insertion of human granulocyte macrophage colony-stimulating factor (GM-CSF) allows local GM-CSF expression to trigger the influx and activation of DCs, which present tumor-associated antigens from lysed tumor cells. These DCs are intended to prime tumor-specific CD4⁺ and CD8⁺ T cells to stimulate a systemic and antitumor adaptive immune response (21, 22). Additional modifications include removal of the ICP47 gene, and immediate early expression of US11 (12).

Studies using mouse syngeneic models have demonstrated antitumor activity of talimogene laherparepvec in injected and contralateral tumors (12). These results support the development of adaptive antitumor immune responses and are consistent with emerging clinical evidence in treated patients (23). To gain further insight into the immune-mediated mechanisms of systemic activity, we have conducted studies in syngeneic murine tumor models using a version of talimogene laherparepvec that expresses the murine GM-CSF gene (OncoVEX^mGM-CSF). Here, we detail the presence of viral particles to injected tumors only, describe longitudinal changes in chemokine expression and immune-cell infiltrates, pinpoint the key role of CD8⁺ cytotoxic cells in systemic efficacy, and characterize the enhanced anti-tumor activity of OncoVEX^mGM-CSF in combination with CTLA-4 blockade. Our data provide definitive evidence of a systemic, tumor-specific adaptive immune response triggered by OncoVEX^mGM-CSF.

The engineering of talimogene laherparepvec (IMLYGIC, Amgen Inc.) is described above and by Liu and colleagues (18). OncoVEX^mGM-CSF is engineered in a manner similar to talimogene laherparepvec, with the exception that the human GM-CSF transgene is replaced with the murine GM-CSF (18). All tumor cell lines were acquired from the ATCC, and cultured as indicated. Cells were plated in a 96-well plate at 2,000 to 10,000 cell per well and incubated overnight at 37°C. Talimogene laherparepvec was added in 1:5 dilutions starting at 10 or 100 MOI. After a 72-hour incubation, the number of cells left in each well was quantified using ATP-Lite (Perkin Elmer).

Female BALB/c mice (Charles River Laboratories), 6 to 8 weeks of age were cared for in accordance with the “Guide for the Care and Use of Laboratory Animals” (24). Animals were housed at Association for Assessment and Accreditation of Laboratory Animal Care International–accredited facilities (at Amgen) in ventilated micro-isolator housing on corncob bedding. All protocols were approved by an Institutional Animal Care and Use Committee. Animals had ad libitum access to sterile pelleted feed and reverse osmosis–purified water and were maintained on a 12:12-hour light:dark cycle with access to environmental enrichment opportunities.

Single-agent and combination efficacy: mouse B-lymphoma cells (cell line A20) or mouse colon carcinoma cells (cell line CT-26) were injected subcutaneously in the right and left flanks of mice (2 × 10⁶ cells). Tumor volume (mm³) was measured using electronic calipers twice per week. Once tumors reached an average of approximately 100 mm³, animals were randomized into groups (10 mice per group) such that the average tumor volume (in both flanks) at the beginning of treatment administration was uniform across treatment groups. Animals were then administered three intratumoral injections of OncoVEX^mGM-CSF or vehicle every third day. Anti-CTLA-4 antibody was administered intraperitoneally twice per week for four doses, beginning on the same day at the same time as OncoVEX^mGM-CSF. Clinical signs, body weight changes, and survival (tumors reached 800 mm³) were measured two to three times weekly until study termination.

CD8⁺ T-cell depletion was achieved by treating animals with antimouse CD8α antibody (clone 53-6.72; BioXCell) intraperitoneally, 500 μg per mouse, 3 days prior to tumor cell implantation and then twice per week until study termination.

Donor immune cells derived from spleen and axillary lymph nodes (LNs) were processed into single-cell suspensions by mechanical dissociation. The splenocytes were then depleted of red blood cells by ammonium chloride lysis before being mixed with LN cells. A total of 2.5 × 10⁷ donor immune cells were transferred intravenously into new Balb/c recipients and then challenged with A20 tumor cells (2 × 10⁶) on the left subcutaneous flank. The appearance of tumor formation was monitored/measured for the ensuing 50 days.

Tumors were either embedded in optimal cutting temperature (OCT) compound or fixed in 10% neutral buffered formalin. Frozen sections were cut at 6 μm and fixed in an acetone/absolute ethanol solution (75%/25%). Paraffin sections were cut at 5 μm, baked at 56 to 60°C for at least 60 minutes, deparaffinized in xylene, and rehydrated through graded ethanol solutions. Epitope retrieval was performed by boiling the slides in DIVA Decloaking solution (Biocare Medical, #DV2004G1).

IHC was performed on frozen sections with the following markers—HSV-1 (Biocare Medical, #APA-3027 AAK), CD3 (clone #145-2C11; BD Biosciences #550275), F4/80 (clone #CI:A3-1; AbD Serotec #MCA497R), CD8 (clone #YTS 169AG 101HL; Thermo Scientific #MA1-70041), p46 (clone #29A1.4; eBiosciences #11-3351), and CD103 (clone #2E7; AbD Serotec #MCA4705GA). IHC was performed on formalin-fixed, paraffin-embedded (FFPE) sections with the following markers—HSV-1, F4/80, Granzyme B (clone #16G6; eBiosciences #13-8822-82), and B220 (clone #RA3-6B2; BD Biosciences #553086).

Morphometric analysis was performed by staining tissue sections via IHC with an anti-B220 antibody to identify the A20 tumor regions. The tumor areas, or region of interest (ROI) were outlined manually (nonviable regions were excluded) and applied to serial images to evaluate staining for cell markers. The area occupied by CD3⁺, CD8⁺, CD103⁺, or Granzyme B⁺ cells is expressed as percentage of the total tumor area.

Total DNA was extracted from pulverized tissue samples (tumor or liver) and whole blood using Qiagen's DNeasy Kit (Qiagen). Digital droplet PCR (ddPCR) was performed using Bio-Rad QX200 System (Bio-Rad). The following sequences were used to amplify HSV-1 thymidine kinase gene—forward primer, 5′-CGATATCGTCTACGTACCCG-3′; reverse primer, 5′-ATATCTCACCCTGGTCGAGG-3′; and probe, 5′-FAM-CGATGACTT/ZEN/ACTGGCGGGTG-IBFQ-3′. Viral copy-number was normalized against mouse hypoxanthine guanine phosphoribosyl transferase (HPRT) gene. The following primers and probe sequences were used to detect murine HPRT—forward primer, 5′-GCTCCACTTTGAAACAGCTG-3′; reverse primer, 5′-CTTTTTCCAAATCCTCGGCA-3′; and probe 5′-HEX-TGCAGATTAGCGATGATGAACCA- BHQ-1-3′. DNA was digested with HaeIII before PCR amplification. Thermal cycling conditions are 95°C for 10 minutes, then 40 cycles at 94°C for 30 seconds, and 60°C for 1 minute, followed by 98°C for 10 minutes.

OCT-embedded or flash-frozen tissues were processed using a TissueLyzer (Qiagen). mRNA was extracted using RNEasy columns (Qiagen) and converted to cDNA using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). qPCR assay primer sets consisting of amplification primers and labeled probes for individual gene assays were synthesized by Integrated DNA Technologies. Pre-amplification and qPCR reactions were carried out using TaqMan PreAmp Master Mix and Universal PCR Master mix (Applied Biosystems), respectively, and then analyzed on Fluidigm Biomark HD using 96.96 integrated fluidics chips (Fluidigm). Expression data for individual samples are normalized to the geometric mean of four housekeeping genes (HKGs) Actb, Gapdh, Ipo8, and Tbp (dCt = Ct_{gene of interest} − Geomean Ct_HKGs) and displayed as normalized gene expression (NGE = 2^−ΔCt with an arbitrary scaling factor of 1,000). IFN response gene signature is comprised of the geometric mean of the NGE values of Ifi27l1, Ifi44, Ifit1, Mx1, and Oas1b.

The tumor-draining LNs (TDLNs) and spleens were isolated, and single-cell suspensions were generated by using protocols and reagents for dissociation of mouse spleen provided by Miltenyi and the GentleMACS Octo instrument (Miltenyi). Lymphocytes were stained with anti-CD3 (clone 145-2C11 or 17A2), anti-CD69 (H1.2F3), anti-CD49b (DX5), anti-CD19 (6D5), anti-Ly6C (HK1.4), and anti-Ly6G (1A8), and AH1_6-14 loaded H2-L^d dextramer (Immudex). Cells were run on either the Miltenyi MACSQuant VYB or a BD LSR Fortessa and analyzed using FlowJo software (FlowJo LLC).

For enzyme-linked ImmunoSpot (ELISPOT) assays, 96-well plates with a nitrocellulose filter base (Millititer HA; Millipore) were coated with purified anti-IFNγ (2 μg/mL) antibody. Enriched T cells (mouse T-cell kit, Miltenyi) from OncoVEX^mGM-CSF- or vehicle-treated A20 tumor-bearing mice on days 4, 7, 11, and 14 after treatment were added to the plates in a 10:1 ratio with either A20 or CT-26 tumor cells.

Alternatively, splenocytes from different treatment groups were used in peptide restimulation assays. Splenocytes (8 × 10⁵) were incubated with control peptides (GFPs) or the AH1 peptide (SPSYVYHQF) at a final concentration of 1 μmol/L for 20 hours at 37°. The AH1 peptide is an immunodominant Ag derived from the envelope protein (gp70) of the endogenous murine leukemia virus presented by the MHC class I L^d molecule (25). Spots were enumerated using a CTLS6 Fluorospot analyzer (CTL).

Mice bearing 150 mm³ A20 tumors in both flanks received an intratumoral injection of 5 × 10⁶ PFU OncoVEX^mGM-CSF (n = 4) or saline (n = 4) into the left flank tumor. Animals received an intravenous injection of 100 μCi [¹⁸F]FHBG (26) into the lateral tail vein 24 hours after OncoVEX^mGM-CSF treatment. Two hours after [¹⁸F]FHBG injection, animals were anesthetized (2–3% isoflurane in 100% oxygen) for anatomical x-ray computed tomography (CT) and 15-minute static PET imaging (Inveon PET/CT; Siemens Medical Solutions). Attenuation-corrected images were reconstructed using the OSEM OP-MAP algorithm, and images were analyzed for tumor [¹⁸F]FHBG uptake using Inveon Research Workplace (Siemens Medical Solutions).

In vivo efficacy data were analyzed by Kaplan–Meier analysis of median survival of mice treated with single agents or single agent versus combination. Unpaired Mann–Whitney nonparametric test was used to analyze morphometric IHC analysis, and P < 0.05 was considered statistically significant. One-way ANOVA with Dunnett's correction was applied for analysis of gene expression differences. Multiple comparison unpaired T-test with Holm–Sidak correction was applied to test the significance of ex vivo T-cell reactivities and flow cytometric analysis of individual populations in a time course manner.

To characterize the oncolytic activity of talimogene laherparepvec against human and murine cell lines, we treated nine cell lines with virus in a 72-hour viability assay. All cell lines tested showed sensitivity to talimogene laherparepvec. IC₅₀ ranged from 0.05 (SK-MEL-5 melanoma) to 12.7 (A20 murine B-cell lymphoma) MOI (Fig. 1). All human melanoma cell lines displayed IC₅₀ below 1 MOI. Syngeneic model cell lines A20 and CT-26 were efficiently lysed by talimogene taherparepvec, although IC₅₀ for CT-26 was 100-fold lower.

Previous studies have demonstrated that local delivery of OncoVEX^mGM-CSF results in oncolysis but the mechanism of antitumor efficacy in noninjected tumors remains unclear (12). We established a syngeneic A20 contralateral tumor model to study systemic efficacy. In OncoVEX^mGM-CSF-treated animals, a majority of injected tumors showed complete regression at all doses. In contrast, contralateral tumors at the lower dose showed no response, whereas the medium and high doses showed complete responses in half the tumors and growth delay in the other half (Fig. 2A–D). Efficacy comparison with backbone OncoVEX (no GM-CSF activity) in this model at the highest dose showed no difference in efficacy in the injected tumors but a significant decrease in cured mice in the contralateral side (Supplementary Table S1). Median survival was significantly increased in the medium- and high-dose groups compared with vehicle (38 days vs. 28 days, respectively; Fig. 2E), and no changes in body weight were observed (Supplementary Fig. S1A). Oncolysis was only observed in the injected tumors (Supplementary Fig. S1B). We monitored the presence of viral DNA and mRNA in injected and contralateral tumors, as well as in the blood, liver, and the respective draining LNs. Viral DNA was detected dose-proportionally only in injected tumors between 2 and 168 hours after injection and did not increase significantly with time. Viral DNA, at a level consistent with a 5 × 10³ PFU dose, was detected in 1/16 contralateral tumors (5 × 10⁵ PFU dose group) and probably represents tissue contamination (Fig. 2F). The expression of three HSV-1 genes, ICP27 (transcriptional regulator, immediate-early gene), VP16 (virion maturation), and thymidine kinase (marker of HSV-1 viral replication), were detected as early as 4 hours in the OncoVEX^mGM-CSF-injected tumor, with a gradual reduction of the signal over time. Conversely, neither the OncoVEX^mGM-CSF contralateral tumor nor the LNs draining any tumor had significant levels of any of the three viral gene products (Fig. 2G and Supplementary Fig. S2A).

We next analyzed tumors for viral antigen and HSV-1 thymidine kinase activity. Histologic analysis of OncoVEX^mGM-CSF-injected tumors revealed a large, well-demarcated central nonviable region that was not present in vehicle-injected tumors. Evaluation of IHC staining in OncoVEX^mGM-CSF-injected tumor tissue sections at higher magnification highlighted a central zone of HSV-1⁺ cell debris, surrounded by a thin band of HSV-1-infected tumor cells undergoing oncolysis (Supplementary Fig. S2B). Ninety-six hours after intratumoral injection treatment, HSV-1 antigen was observed only in the OncoVEX^mGM-CSF-injected tumors (Fig. 2H and Supplementary Table S2). Viral antigen staining was most robust at 72 hours and markedly decreased by 168 hours. Finally, we used [¹⁸F]FHBG PETto detect HSV-1 thymidine kinase activity in live animals injected with vehicle or OncoVEX^mGM-CSF. A very pronounced and significant ¹⁸FHBG tumor accumulation was only measured in the OncoVEX^mGM-CSF-injected tumors (Fig. 2I and Supplementary Fig. S2C). Delivery of OncoVEX^mGM-CSF systemically by intravenous administration did not produce detectable viral DNA in tumor tissue which correlated with lack of efficacy in the A20 syngeneic model (Supplementary Fig. S3).

Type I IFN signaling has previously been shown to be critical in restricting the pathogenesis of HSV-1 in vivo (27). The IFN gene signature (composed of five type I IFN inducible genes) is elevated in OncoVEX^mGM-CSF-injected tumor relative to vehicle 4 hours after injection (Fig. 3A).

GM-CSF can activate antigen-specific immune responses (21) and may play a similar role for OncoVEX^mGM-CSF. We detected high levels of mGM-CSF between 4 and 96 hours, consistent with viral presence in the injected tumor. In addition, significant increase in IFNγ was detected at 4 hours postinjection in OncoVEX^mGM-CSF-injected tumors. Consistent with an inflammatory response, the chemoattractant chemokine (C-X-C motif) ligand (CXCL) 2 and CXCL10 were both highly upregulated in the OncoVEX^mGM-CSF-injected tumor compared with vehicle between 4 and 96 hours (Fig. 3B). Expression of CXCL2, known for its chemoattractant attributes for monocytes and neutrophils, increased in time and matched the increased number of these cell types detected in injected tumors at 24 and 48 hours (Supplementary Fig. S4A).

While evaluating the TDLN for HSV-1 antigen, we observed a marked enlargement of the TDLN proximal to the OncoVEX^mGM-CSF-injected tumor at 48 hours (Supplementary Fig. S4B) but not the contralateral or vehicle-treated TDLN. Enlargement was characterized by an increased number of viable cells. We next investigated the lymphocyte populations by flow cytometry at 2, 5, and 10 days after OncoVEX^mGM-CSF treatment. An increase in the percentage of activated T cells (CD69⁺CD3⁺) in the OncoVEX^mGM-CSF-injected TDLN was observed at days 2 and 5, but not at day 10 (Fig. 3C).

Next, we performed gene expression analysis in tumors harvested from OncoVEX^mGM-CSF-treated and vehicle-treated animals for CD3, CD8, and CD103 at 48 and 168 hours after treatment (Fig. 4A). At 168 hours posttreatment, mRNA expression levels of all three markers were increased in both the OncoVEX^mGM-CSF-injected and OncoVEX^mGM-CSF contralateral tumors compared with the corresponding tumors from vehicle-treated animals. IHC and FACS analysis were performed to support the findings from gene expression analysis. Morphometric analysis of serial tissue sections stained using IHC showed that 5.5% of the ROI area was occupied by CD3⁺ cells in both the vehicle-injected (SD = 3.2%) and contralateral tumors (SD = 1.2%) at 168 hours after injection (Fig. 4B). There was a statistically significant increase in the percent tumor area occupied by CD3⁺ cells in both the OncoVEX^mGM-CSF-injected (14.24% ± 2.29% of ROI, P < 0.01 by Mann–Whitney test) and OncoVEX^mGM-CSF contralateral tumors (11.97% ± 2.89% of ROI, P < 0.05) compared with the respective tumors from vehicle-treated animals. Longitudinal changes in CD3⁺ T cells in contralateral tumors were confirmed by IHC and FACS analysis. Significant increases in CD3⁺ T cells were measured at 168 hours by both techniques (Fig. 4C and D). Similarly, the percent area occupied by CD8⁺ cells was significantly increased in both the OncoVEX^mGM-CSF-injected and contralateral tumors at 168 hours compared with relevant controls (Fig. 4B). The percentage of tumor area occupied by CD103⁺ cells, a marker of tissue resident T cells and a pro-inflammatory subset of DCs (28), was significantly elevated in the OncoVEX^mGM-CSF-injected versus vehicle-injected tumor (P < 0.01) but not in contralateral tumors (Fig. 4B). T-regulatory cells (Tregs) known for their immunosuppressive role in the tumor microenvironment were also significantly decreased in contralateral tumors following treatment with OncoVEX^mGM-CSF (Fig. 4E).

We set out to investigate the ability of the systemic T cells found in OncoVEX^mGM-CSF-treated animals to react to tumor-specific antigens. Pan-T cells were harvested and purified from tumor-bearing animals at days 4, 7, 11, and 14 post-intratumoral OncoVEX^mGM-CSF or vehicle administration and used in an ELISPOT assay. At all time points tested, the number of IFNγ⁺ spots was significantly increased in the OncoVEX^mGM-CSF-treated T cells cultured with A20 tumor cells compared with vehicle–treated T cells (Fig. 5A). In addition, tumor responses were largely specific to A20 and not to CT-26. Reactivity peaked at day 11. HSV-1–derived viral antigens did not appear to play a significant role in the systemic immune response visualized here, as there was no difference in T-cell reactivity when A20 stimulator cell lines were pre-exposed to virus prior to ex vivo restimulation with splenic T cells (Supplementary Fig. S5A).

To further characterize the activity of these systemic T cells in vivo, we performed CD8⁺ depletion experiments and adoptive cell transfer experiments. Depletion of CD8⁺ cells (>95%) using anti-CD8 antibodies (Supplementary Fig. S5B) had profound negative effects on the efficacy of OncoVEX^mGM-CSF (Fig. 5B–F). CD8⁺ cell depletion affected the growth of both injected and contralateral A20 tumors by removing tumor volume variability from the vehicle-treated group (Fig. 5B and C). Most importantly, depletion of CD8⁺ cells from OncoVEX^mGM-CSF-treated mice abrogated all the efficacy driven by OncoVEX^mGM-CSF on contralateral tumors (0/10 vs. 5/10 cures) and also had a profound negative effect on efficacy in the injected tumors (1/10 vs. 9/10 cures) (Fig. 5D and E). The survival benefit produced by OncoVEX^mGM-CSF in this model was not observed when CD8⁺ cells were depleted (Fig. 5F vs. Fig. 1E).

To evaluate the direct role of OncoVEX^mGM-CSF in the specificity of a memory response, we injected mice with immune cells from naïve mice, mice bearing A20 tumors, and mice cured from contralateral tumors through OncoVEX^mGM-CSF treatment. Re-challenged animals that received naïve cells or cells from A20 tumor-bearing mice showed no protection (Fig. 5G–I). Both groups showed established tumors at day 10 in 9/10 mice. However, mice receiving immune cells from OncoVEX^mGM-CSF-cured mice demonstrated resistance to rechallenge with A20 cells with only two out of 10 mice growing tumors at day 32 (Fig. 5J). A similar level of protection was also observed when OncoVEX^mGM-CSF cured mice were directly challenged with A20 cells (Supplementary Fig. S6).

We sought to determine if changes in immune regulatory molecules CD80 and CTLA-4 could be detected after OncoVEX^mGM-CSF treatment. By FACS analysis, CTLA-4 and CD80 were upregulated in the OncoVEX^mGM-CSF-injected TDLN at the early 48-hour timepoint, whereas significant tumor upregulation was observed in injected tumors at 168 hours (Fig. 6A).

Given the observed upregulation of CD80 and CTLA-4, we set to determine whether a combination of OncoVEX^mGM-CSF with anti–CTLA-4 antibodies could result in enhanced efficacy in contralateral tumors. Treatment of A20 tumors with OncoVEX^mGM-CSF in combination with anti–CTLA-4 antibodies resulted in a significant increase in median survival and complete regressions compared with either single agent alone (Fig. 6B–F). The combination produced cures in 90% of the contralateral tumors (Fig. 6E). To extend our observations to other models, we performed a combination study with anti–CTLA-4 antibodies in the CT-26 model (Fig. 6G–K). This model demonstrated some resistance to OncoVEX^mGM-CSF in contralateral tumors, where tumor growth delays may be seen, but tumor regressions were rarely observed in monotherapy (Fig. 6I). The combination of OncoVEX^mGM-CSF with anti–CTLA-4 resulted in complete cures of all injected tumors and regressions in 80% of contralateral tumors (with 6/10 cures) (Fig. 6J). More importantly and similar to what was observed in the A20 model, the median survival in the combination group was more than double that in the vehicle-treated group and was significantly better than either single agent alone (Fig. 6K). Interestingly, treatment in the combination group seemed to significantly delay the dynamics of the OncoVEX^mGM-CSF monotherapy response resulting in large tumors (>500 mm³) suddenly regressing and disappearing (Fig. 6J).

Finally, we used the previously described AH1 antigen in CT-26 cells as a surrogate to assess the ability of OncoVEX^mGM-CSF, CTLA-4, and the combination treatment to release tumor antigens and stimulate antitumor-specific T-cell responses. Quantification of systemic (splenic) anti-AH1 CD8⁺ T cells by ELISpot or by dextramer staining using FACS demonstrated a significant increase in AH1 reactive T cells in mice treated with OncoVEX^mGM-CSF, CTLA4 blockade, or the combination (Fig.6L and M). Quantification of local (tumor) anti-AH1 CD8⁺ T cells showed a significant increase only in the combination group. Consistent with the effect of OncoVEX^mGM-CSF in the A20 model, a significant decrease in Tregs was also observed in the CT-26 model. This effect was greater in combination with CTLA-4 blockade (Fig. 6N).

In OPTiM, a randomized phase III clinical trial of intralesional talimogene laherparepvec versus subcutaneously GM-CSF in unresectable, metastatic melanoma, the primary endpoint of durable response rates was met, with a significant increase in the treatment arm versus control (HR 8.9; 16.3% vs. 2.1%; P < 0.001; ref. 29). At the lesion level, a ≥50% reduction in tumor area was seen in 64% of injected lesions, 32% of noninjected subcutaneously and nodal lesions, and 16% of distant visceral lesions (30). These findings can be explained by direct oncolysis or by enhanced systemic antitumor immunity. Clinical trials to elucidate the mechanism of action (MOA) and the potential use of talimogene laherparepvec in combination with checkpoint blockade (CTLA-4 and PD-1) are underway (NCT01740297 and NCT01740297). However, given the potential benefits for patients, preclinical studies to further elucidate its MOA are critical for optimizing future clinical use in combination with novel therapies that perturb the cancer immunity cycle.

Although evidence indicates systemic efficacy is provided by a talimogene laherparepvec-triggered tumor-specific immune response, data demonstrating that virus does not directly lyse distant tumor cells has been lacking. Using OncoVEX^mGM-CSF, we demonstrate that oncolysis alone cannot account for systemic efficacy. Despite robust tumor growth inhibition in the contralateral tumor, no signals of oncolysis (viral DNA, mRNA, antigen, or thymidine kinase activity) were detected in contralateral tumors, blood, and liver with highly sensitive (31, 32) assays (ddPCR and qPCR). These data support the tumor selective nature of OncoVEX^mGM-CSF replication described previously (12) and point to an immune-mediated MOA as its principal mode of systemic efficacy.

The A20 B-cell lymphoma model used in this study models an immunogenic tumor as evidenced by the diffuse staining of CD3. Despite this infiltration, detection of systemic antitumor reactivity in the absence of treatment is minimal. The localized innate immune response to the injected oncolytic virus and subsequent lysed tumor cells creates type I and type II IFN responses. The notion that such localized inflammatory response is capable of transforming an immunosuppressive microenvironment into an inflammatory one is an attractive therapeutic hypothesis for several reasons. First, altering the tumor immunosuppressive niche may be sufficient to unleash tumor-reactive T cells that are otherwise limited in their function (9). Second, this type of immunogenic cell death is thought to be a key factor in generating novel systemic antitumor responses and enabling efficacy enhancement through checkpoint blockade (23, 33, 34). Therefore, unlike tumor vaccinations, talimogene laherparepvec offers a novel strategy with no requirement for elucidating patient-specific tumor antigens.

Despite the potency of the inflammatory response elicited by OncoVEX^mGM-CSF, these effects appear to be transient and localized to the injected tumor and adjoining LN. Systemic efficacy would theoretically still be governed by the same principles of interaction between the immune system and the tumor. Evidence of this can be seen in the upregulation of T-cell checkpoint molecules concomitant with infiltration of T-cells into the injected and contralateral tumors. In effect, intratumoral administration of OncoVEX^mGM-CSF appears to increase the reactivity of splenic T cells even at early time points. Subsequently, an expansion of tumor-reactive T cells evaluated by secretion of IFNγ and by CTL killing is detected at time points that would be consistent with amplification and/or de novo priming. Without determining antigen specificity of the T cells, it is impossible to distinguish tumor-reactive from virus-specific T cells. However, a preponderance of evidence suggests that the systemic adaptive immune response contains a significant portion of tumor-specific T cells that are not present (or present at lower levels) in untreated tumor-bearing animals: (i) the absence of viral antigens in the contralateral tumor; (ii) ex vivo tumor cell line reactivity; (iii) transplantability of tumor rejecting immune cells; (iv) the loss of efficacy upon CD8⁺ cell depletion; (v) enhanced systemic efficacy in combination with CTLA-4 blockade. The increase in tumor-reactive T cells found in the splenic compartment as early as day 4 posttreatment is consistent with a very rapid mobilization and expansion of tumor-reactive T cells. At later time points, the adaptive response is expanded both systemically, as evidenced by the expansion of splenic tumor-reactive T cells and by the infiltration of lymphocytes into the contralateral tumor. Activation of tumor-reactive T cells may be accomplished by expansion and proliferation of the preexisting tumor-infiltrating lymphocytes (TIL) or by uncovering of novel tumor-derived antigens presented in secondary lymphoid organs. Even though we did not study the uncovering of novel tumor-derived antigens in our current experiments, our assessment of anti-AH1 CD8⁺ T cells in the spleen and tumor following OncoVEX^mGM-CSF, CTLA-4 blockade, or the combination is consistent with the expansion and proliferation of preexisting TILs. The models described here do not allow us to pinpoint if efficacy is driven by a preexisting antigen like AH1, by novel tumor antigens or by the combination. However, previous approaches have shown that release of tumor-associated antigens by localized tumor destruction can lead to enhanced systemic responses and epitope spreading (35–37).

Expression of CTLA-4 (38), which is known to increase upon T-cell activation, is also consistent with an ongoing adaptive immune response. It appears that the inflammatory local microenvironment induced by the administration of OncoVEX^mGM-CSF, evidenced by the increased expression of CD80 and local production of IFN, may provide a critical boost to antigen-presenting cells and thereby lead to the generation of a productive immune response. Therefore, it is likely that combining intratumoral talimogene laherparepvec treatment with blockade of CTLA-4 would further enhance anti-tumor efficacy beyond what can be achieved by monotherapy alone. The first evidence of this type of synergistic effect was described for another oncolytic virus by Allison and colleagues (37). When we combine OncoVEX^mGM-CSF with CTLA-4 blockade the complete regression of many large (>400 mm³) CT-26 tumors confirms this synergistic effect. This level of efficacy is similar to that achieved by the combination of four different therapeutic approaches in a recently published report (39). Talimogene laherparepvec is currently being investigated in combination with the anti–CTLA-4 antibody ipilimumab for the treatment of unresected melanoma, where 56% of patients experienced objective responses (by Immune-Related Response Criteria) of both injected and distant tumors (40).

In summary, here we expand our understanding of the MOA of talimogene laherparepvec in order to optimize its future clinical benefit for patients. Our data clearly demonstrate that intratumoral administration of OncoVEX^mGM-CSF drives the development of systemic antitumor immunity that can be enhanced in combination with checkpoint blockade.

A.K. Moesta and J.B. Rottman hold ownership interest (including patents) in Amgen. R. Ponce reports receiving other commercial research support from Juno Therapeutics. C. Beers is an employee of Tizona Therapeutics and is a consultant/advisory board member for Oncorus. P.J. Beltran holds ownership interest (including patents) in Amgen. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A.K. Moesta, J.B. Rottman, T. Le, C. Glaus, R. Ponce, C. Beers, P.J. Beltran

Development of methodology: A.K. Moesta, K. Cooke, J. Piasecki, J.B. Rottman, K. Fitzgerald, J. Zhan, T. Le, K. Merriam, C. Glaus, K. Ganley, C. Beers, P.J. Beltran

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.K. Moesta, J. Piasecki, P. Mitchell, J.B. Rottman, K. Fitzgerald, J. Zhan, B. Yang, T. Le, B. Belmontes, O.F. Ikotun, K. Merriam, C. Glaus, D.H. Cordover, A.M. Boden

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.K. Moesta, K. Cooke, J. Piasecki, P. Mitchell, J.B. Rottman, K. Fitzgerald, T. Le, B. Belmontes, O.F. Ikotun, C. Glaus, D.H. Cordover, A.M. Boden, R. Ponce, C. Beers, P.J. Beltran

Writing, review, and/or revision of the manuscript: A.K. Moesta, J.B. Rottman, K. Merriam, C. Glaus, R. Ponce, C. Beers, P.J. Beltran

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): P. Mitchell, T. Le, K. Merriam, D.H. Cordover, A.M. Boden

Study supervision: A.K. Moesta, K. Cooke, P. Mitchell, C. Glaus, C. Beers, P.J. Beltran

We thank Angela Chong for technical assistance with gene expression profiling and Annie Mirsoian for assistance with flow cytometry panel design. We thank Emily Plummer who provided medical writing services on behalf of Amgen Inc. Assistance with formatting the manuscript to meet journal specifications was provided by Gurpreet Kaur, CACTUS Communications, on behalf of Amgen Inc.

The study was funded by Amgen, Inc.

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.