Defining Effective Combinations of Immune Checkpoint Blockade and Oncolytic Virotherapy

The last 5 years have seen the emergence of antibody-mediated blockade of immune checkpoints as a key new weapon in the anticancer arsenal (1, 2). The anti-CTLA4 inhibitor ipilimumab has been approved for the treatment of melanoma (3, 4), while a panel of monoclonal antibodies targeting the interaction of PD1 and PDL1 have also demonstrated promising responses in a succession of clinical trials (5, 6). Together, these trials have demonstrated the clinical need to overcome the tumor's capacity to shut down the T-cell response in the creation of an effective cancer immunotherapy.

The field of oncolytic virotherapy has also recently demonstrated its potential to produce clinically effective cancer treatments, with data from several recent randomized trials resulting in impressive response rates (7, 8). One factor that has united the most successful oncolytic vectors has been the expression of an immune-activating transgene (GM-CSF), an indication that a key determinant of the activity of oncolytic viruses is their capacity to activate and target the immune response (9, 10). This has since been confirmed in a multitude of preclinical studies (11–13), such that the oncolytic virus platform might best be considered an immunotherapeutic.

We have previously developed several oncolytic vectors, primarily focusing on vectors based on vaccinia virus (14–17). These provide several advantages as immunotherapies beyond their long historical use as vaccines: (i) They can induce an adaptive immune response raised against tumor antigens as a result of their selective replication within the tumor microenvironment (18, 19). This in situ vaccination effect results in production of CTL targeting relevant tumor antigens without the need for any prior interrogation of the tumor. (ii) Viral replication within the tumor can at least transiently overcome localized immunosuppression, something that most traditional vaccine approaches fail to achieve. However, in many cases, once the oncolytic virus is cleared by the host immune response, the immunosuppressive environment is apparently restored and the tumor relapses. The combination of oncolytic virus and the blockade of immune checkpoint inhibitor therefore is an appealing strategy.

Although there has been much interest in this combination, including the proposed clinical combinations of the oncolytic HSV T-Vec (Amgen) and ipilimumab (Yervoy, Bristol Myers Squibb) in the treatment of melanoma (Clinical Trials.gov NCT01740297), there have been very little supportive data reported to date. Here, we examine the combination of oncolytic vaccinia with several different immunotargeting monoclonal antibodies.

Renca (murine renal adenocarcinoma) cell line was obtained from ATCC. MC38 cell line (murine colon adenocarcinoma) was a kind gift from Dr. David Bartlett (University of Pittsburgh Cancer Institute, Pittsburrgh, PA). Cell lines were maintained in recommended culture media containing 5%–10% FBS at 37°C, 5% CO₂. Cell lines have not been authenticated by the authors beyond their ability to form tumors in syngeneic mouse models.

All recombinant vaccinia strains used in this work are derived from the Western Reserve (WR) strain (BEI Resources). The double-deleted strains vvDD and WR.B18R-.TK- (B18R- in short) have been described previously (15, 20). These contain deletions in the tk gene and in the vgf or B18R viral genes, respectively. In addition, both strains express the firefly luciferase gene from the synthetic vaccinia promoter pE/L (21), which allows monitoring of luciferase expression as a surrogate indicator of viral replication (22). Viruses were titered, manufactured, and purified as previously described (23).

All animal studies were approved by the University of Pittsburgh Institutional Animal Care and Use Committee. C57/BL6 and BALB/c female mice (6–8 weeks old) were purchased from The Jackson Laboratory. Renca or MC38 tumor cell lines were implanted subcutaneously at 5 × 10⁵ cells per mouse into BALB/c or C57/BL6 mice, respectively. Oncolytic vaccinia viruses were injected intravenously (tail vein) at 2 × 10⁸ pfu/mouse when tumors reached approximately 50 to 100 mm³.

Anti-mouse CTLA4 (9D9) and anti-mouse CD25 (PC-61.5.3) antibodies (BioXCell) were injected intraperitoneally at 100 or 200 μg/mouse/dose, respectively, with treatments consisting of 3 doses each 3 days apart. Mouse IgG2b κ Isotype Control (BioXCell) was used as a control. For depletion experiments, anti-mouse CD8 (2.43), anti-mouse CD4 (GK1.5), and anti-mouse IFNγ (XMG1.2) were purchased from BioXCell, and anti-mouse Asialo-GM1 was purchased from Wako Pure Chemicals (Richmond, VA). Mice were injected intraperitoneally with 500 μg at days 1 and 2 after tumor implantation, followed by 250 μg injection every 5 days till the end of the experiment.

Tumor volume was monitored by caliper measurement and defined by V(mm³) = π/6 × W² × L, where W and L are the width and the length of the tumor, respectively. Data are expressed as tumor size relative to the beginning of the therapy (100%). For Kaplan–Meier survival curves, end point was established at ≥750 mm³. Animals whose tumor size never achieved the threshold were included as right-censored information.

Viral gene expression was determined through bioluminescence imaging of luciferase expression in vivo. A dose of 4.5 mg of d-luciferin (GoldBio) was injected intraperitoneally per mouse before imaging on an IVIS2000 (PerkinElmer; 2% isoflurane). Images were analyzed using LivingImage software (PerkinElmer).

For ELISPOT assays, splenocytes were mixed with tumor cells or splenocytes from naïve mice infected with UV-inactivated vaccinia virus at 5:1 ratio. Naïve splenocytes were used as control. 96-well membrane filter plates (EMD Millipore) coated with 15 μg/mL of monoclonal anti-mouse IFNγ antibody AN18 (Mabtech, Inc.) were used. Cells were maintained for 48 hours at 37 ºC and spots were detected using 1 μg/mL of biotinylated anti-mouse INFγ antibody R4-6A2-biotin (Mabtech). Plates were developed using an ABC kit and an AEC substrate kit for peroxidase (Vector Laboratories, Inc.). Specific spots were counted and analyzed using an ImmunoSpot Analyzer and ImmunoSpot software from CTL.

Tumors were harvested from mice and mechanically disaggregated and digested with triple enzyme mixture (Collagenase type IV, DNase type IV, and Hyaluronidase type V, Sigma-Aldrich)]. Cell surface and intracellular immunostaining analyses were performed using a Gallios Flow Cytometer (Beckman Coulter, Inc.). For intracellular staining, cells were fixed and permeabilized using a Foxp3 Fix/Perm Buffer Set (eBioscience). Tumor-disaggregated cells were stained using PE-Cy7 anti-mouse CD3 (BD Biosciences), eFluor450 anti-mouse NKp46, APC anti-mouse NKg2D, FITC anti-mouse CD4, PerCP-Cy5.5 anti-mouse CD8, PE anti-mouse CD25, and APC anti-mouse Foxp3 antibodies (eBioscience).

Standard Student t tests (two-tailed) were used throughout this work, except for the comparison of survival curves, where a log-rank test was used. In all cases, significance was achieved if P< 0.05.

Mice harboring syngeneic subcutaneous mouse renal adenocarcinomas (Renca cells) were injected with a single intravenous dose of oncolytic vaccinia virus and with three intraperitoneal doses of 100 μg of mouse anti-CTLA4 antibody at days 0, 3, and 6 after virus injection. The schedule and doses of anti-CTLA4 antibody used were determined on the basis of previously published preclinical studies (24–26). Initially, we looked to determine the safety of the combination and the effects of injection of anti-CTLA4 antibody on the replication of vaccinia in the tumors. We monitored viral luciferase transgene expression with bioluminescence imaging as ourselves and others have shown this to directly correlate with viral replication (22, 27). Anti-CTLA4 antibody significantly reduced viral luciferase expression from within the tumors (at days 3 and 5 after virus injection, >5- and 40-fold reduction was detected; Fig. 1A). A similar depletion in viral replication was also observed in a second tumor model (MC38 tumors implanted subcutaneously in C57/B6 mice, Supplementary Fig. S1B), demonstrating that this was not a cell line- or mouse strain–associated effect. Reduced viral replication did, however, correlate with enhanced immune activation, as seen with the increased numbers of CTLs recognizing vaccinia epitopes detected in the spleens of the mice (Fig. 1B). Although increased antiviral CTL appeared as early as day 3 after treatment, it is likely that innate immune responses may also be enhanced with the combination as viral replication is reduced as early as 24 hours after treatment. This indicated a more robust immune response was raised in mice when anti-CTLA4 antibody was injected together with the viral therapy.

A novel schedule for oncolytic vaccinia and anti-CTLA4 antibody combination was therefore designed to permit an initial phase of viral oncolytic activity before anti-CTLA4 antibody administration (Fig. 1C). Anti-CTLA4 doses were therefore injected at days 4, 7, and 10 after virus injection, allowing an initial phase of unhindered viral replication and spread within the tumor (Supplementary Fig. S1C). Whereas simultaneous injection of vaccinia virus and anti-CTLA4 antibody resulted in no significant antitumor benefit (compared with mice treated with single vaccinia therapy), when we delayed administration of the blocking antibody until after the peak of viral replication, a greater than 3-fold reduction (P < 0.04) in tumor volume was observed (isotype control antibody had no effect on tumor growth; Fig. 1D and Supplementary Fig. S1A). In addition, this novel treatment schedule was able to significantly increase survival of mice relative to groups treated either with single vaccinia therapy or vaccinia injected concurrently with anti-CTLA4 (Fig. 1E).

As an alternative to immune checkpoint blockade therapy (anti-CTLA4), we looked to test whether other approaches that also target tumor immunosuppression [such as depletion of regulatory T cells (Treg) with anti-CD25 therapy] also synergized with oncolytic vaccinia therapy. We again initially monitored virus replication in tumors via luciferase bioluminescence imaging after anti-CD25 administration. As with anti-CTLA4 combination, we observed a reduction in viral kinetics when antibody therapy began on the same day as viral treatment; however, differences were not significant (Fig. 2A). Furthermore, when the antitumor effects of vaccinia/anti-CD25 combination therapy were tested, neither regimen (injecting anti-CD25 antibody concurrently with virus or after viral replication peak) resulted in improved efficacy relative to single oncolytic vaccinia therapy (Fig. 2B and Supplementary Fig. S2B). As a further test, anti-CD25 antibody was also added before viral therapy (Supplementary Fig. S2C), however, again no therapeutic advantage was seen (Treg depletion with the anti-CD25 regimen used was also confirmed; Supplementary Fig. S2A). Finally, a direct comparison of the anticancer activity of vaccinia/anti-CD25 versus vaccinia/anti-CTLA4 combination therapies confirmed the enhanced efficacy of combining oncolytic virus with blockade of CTLA4 (Fig. 2B).

As a next step, we looked to examine the importance of the viral vector used in these combination approaches. Two different double-deleted oncolytic vaccinia viruses were compared in combination with anti-CTLA4 antibody therapy. vvDD (vgf and tk double-deleted vaccinia virus) has demonstrated highly tumor-restricted replication (28) that is equivalent in level and selectivity to the B18R- strain. B18R- (B18R and tk double-deleted vaccinia virus) also demonstrated highly tumor-restricted replication but this was coupled with enhanced immunogenicity relative to vvDD (including increased production of cytokines and chemokines within the tumor; ref. 29). This is due to the loss of B18R, that encodes a secreted type I IFN-binding protein (14). When both viral strains were compared for anticancer effects in combination with anti-CTLA4 antibody (Fig. 3), B18R-/anti-CTLA4 treatment induced a more than 3.6-fold (P < 0.009) reduction in tumor size at sacrifice compared with PBS treatment, while in this model vvDD/anti-CTLA4 combination only induced a 1.4-fold inhibition.

We next looked to test in more detail the most effective combination of viral vector (B18R-), antibody (anti-CTLA4) and regimen (antibody treatment beginning 4 days after viral therapy) determined from the previous studies.

Mice carrying either Renca (renal adenocarcinoma) or MC38 (colon adenocarcinoma) tumors were injected with a single intravenous dose of B18R- at 2 × 10⁸ pfu per mouse. At days 4, 7, and 10 after virus injection, an intraperitoneal dose of 100 μg of mouse anti-CTLA4 antibody was administrated. PBS or single therapy treatments were used as controls. At the time of sacrifice, combination therapy resulted in a reduction of more than 2.7- (P < 0.035) and 1.3-fold (P < 0.02) in Renca and MC38 tumor models, respectively, relative to single B18R- therapy (Fig. 4A).

The combination induced a reduction of more than 2.8-fold (P < 0.04) in tumor volume compared with singe anti-CTLA4 therapy at day 42 after treatment in Renca models. Importantly, B18R-/anti-CTLA4 combination therapy induced 3 of 12 complete responses in this model. For MC38 tumors, B18R-/anti-CTLA4 combination therapy did not produce as dramatic an effect, but still reduced tumor volume 1.5-fold (P < 0.045) compared with single anti-CTLA4 therapy, a significant improvement by day 24 after treatment.

To evaluate the mechanisms driving the most effective combination of oncolytic vaccinia and anti-CTLA4 antibody, we examined the immune response raised against and within the tumor. Mice bearing Renca tumors were treated as before. Controls included PBS, single B18R- therapy, or single anti-CTLA4 therapy (injected at days 0, 3, and 6). Mice were sacrificed at day 11 after virus administration and evaluated for specific CTLs in the spleen by the ELISpot assay and for immune cell populations in tumors by flow cytometry. Combination therapy was able to significantly increase the numbers of CTLs recognizing tumor cell antigens compared with any of the controls (Fig 4B). When CD3⁺CD4⁺ populations in tumors were quantified, a significant percentage increase was observed after treatment with B18R-/anti-CTLA4 combination therapy relative to any other treatment (Fig. 5A and C). An increase in the percentage of CD3⁺CD8⁺ cells infiltrating the tumor was also observed, but appeared to be more closely associated with replication of the virus in the tumor (Fig. 5B and C), with both virus-treated groups displaying high levels of these cells. Finally, to ensure that the increased CD3⁺CD4⁺ population infiltrating the tumors did not represent Tregs, additional staining for CD25 and FoxP3 was used (Fig. 5D). We observed that in the control group, about 40% of the CD3⁺CD4⁺ cells present a Treg phenotype (CD25⁺Foxp3⁺). Anti-CTLA4 treatment barely reduced this percentage, but treatment with B18R- virus dropped amounts to 17%, and this improved further to only 13% when anti-CTLA4 was combined with oncolytic virus. Although very few NK or NK-T cells were detected in the tumor (<0.01% of cells), this number was also significantly increased only when the combination of B18R- virus and anti-CTLA4 antibody was used in combination (Fig 5E and Supplementary Fig. S4).

To define the host factors critical for the therapeutic advantage seen with the B18R-/anti-CTLA4 combination, viral replication and antitumor effect experiments were repeated in Renca tumor-bearing mice depleted for CD4⁺ T cells, CD8⁺ T cells, or NK cells (Fig. 6). It was seen that both CD8⁺ T cells and NK cells were required for the therapeutic advantage (while antitumor effects were maintained after depletion of CD4⁺ T cells; Fig. 6B). Depletion of CD8⁺ T cells but not NK cells or CD4⁺ T cells also significantly enhanced viral replication, indicating this cell lineage was responsible for both reduced viral replication and enhanced antitumor effects during B18R-/anti-CTLA4 combination (Fig. 6A). Interestingly CD8⁺ T cells appeared responsible for reduced viral replication in the tumor, even at times as soon as 1 day after viral treatment. The importance of CD8⁺ T cells was supported through depletion of IFNγ, which also resulted in loss of therapeutic advantage and enhanced viral tumor-specific replication (Supplementary Fig. S5). NK cells appear to be required for the antitumor effect, but do not limit viral replication.

It is evident that the blockade of immune checkpoint alone is rarely curative, but has the capacity to synergize with other therapies that selectively activate the immune response. As such, the realization that the immune response raised by oncolytic viral therapies is a critical mechanism mediating their therapeutic activity means that the combination of these two platforms would be logical and appealing (30). However, despite the fact that clinical trials have been proposed combining the oncolytic HSV T-Vec with ipilimumab, little preclinical data have been reported on such combinations (31).

Here, we examine approaches to combine oncolytic vaccinia viruses with different monoclonal antibodies that target cancer-mediated immunosuppression. Significantly improved antitumor responses were demonstrated in several mouse tumor models, providing strong support for the clinical translation of this approach. However, it was initially seen that careful consideration was needed to identify the correct combination of antibody, viral strain and especially in the timing of application of the different treatments. Incorrect combination resulted in a loss of benefit and potentially antagonistic effects.

In initial studies combining anti-CTLA4 blocking antibody with vaccinia virus, no therapeutic benefit was seen (Fig. 1D). In these studies the antibody treatment was begun at the same time as viral inoculation, and imaging of viral luciferase transgene expression demonstrated that viral gene expression was reduced by more than 40-fold relative to virus used alone (Fig. 1A). This indicated that a robust antiviral immune response was being raised leading to premature clearance of the virus. Indeed, this combination was also shown to result in a significant increase in the level of antiviral CTL (Fig. 1B) and viral replication was restored after depletion of either CD8⁺ T cells or IFNγ (Fig. 6A and Supplementary Fig. S4A). This is potentially important as several groups are looking to express antibodies blocking immune checkpoints directly from oncolytic vectors (32). We have previously used exogenous regulation of cytokine transgene function to down regulate cytokine function for a period of around 4 days after initial treatment (22, 33). This allowed an initial phase of viral oncolytic activity and unhindered replication within the tumor, prior to a secondary phase of immunotherapeutic activity that could be enhanced through subsequent stabilization of the cytokine function. Using a similar tactical approach, it was felt that addition of anti-CTLA4 antibody at later times after viral therapy could result in improved therapeutic activity.

This was indeed confirmed (Fig. 1D), with initiation of anti-CTLA4 therapy 4 days after viral delivery found to result in significantly improved antitumor effects in mouse syngeneic tumor models.

Several different monoclonal antibody therapies target cancer-mediated immunosuppression, including both blockade of immune checkpoints as well as direct depletion of suppressive immune cell types that are known to accumulate within the tumor microenvironment (34). One example of the latter approach uses anti-CD25 antibody to deplete Tregs within the tumor. It was determined, however, that independent of whether anti-CD25 was added prior to, at the same time as, or after viral therapy there was no therapeutic advantage seen with anti-CD25 antibody combined with oncolytic vaccinia (relative to vaccinia used alone) (Fig. 2B and Supplementary Fig. S2). This somewhat surprising result may be due to the fact that oncolytic vaccinia alone was actually found to be an effective means to reduce the levels of Tregs in the tumor microenvironment (Supplementary Fig. S3A). This also highlights the fact that the important contribution of the anti-CTLA4 antibody to the synergistic combination with oncolytic vaccinia appears to be dependent on its activation of CD8⁺ T cells rather than on depletion of Tregs.

We have also previously demonstrated that some viral mutations result in production of viral vectors with enhanced immunoactivation properties (29). For example, although vaccinia strains carrying the thymidine kinase deletion typically used to mediate tumor selectivity in oncolytic vaccinia vectors such as JX-594 (Pexa-Vec, Jennerex, now part of Sillajen; ref. 10), vvDD (28) or GLV-1h68 (GL-ONC1, GeneLux; ref. 35) did result in immune activation, this could be enhanced if the viral B18R gene (a secreted type I IFN-binding protein; ref. 36) was also deleted. When the oncolytic vaccinia strain vvDD was compared head to head with a comparable strain carrying the B18R gene deletion (WR.TK-.B18R-) in combinations with anti-CTLA4 antibody, it was found that the “immune activation enhanced” virus (B18R-) was significantly more potent (Fig. 3). Therefore, viral strain and backbone are also important considerations when designing combination therapies with blockade of immune checkpoints.

The optimal combination of monoclonal antibody (anti-CTLA4), viral strain (B18R-) and treatment regimen (antibody therapy begun 4 days after viral delivery) resulted in significantly enhanced therapeutic responses in different mouse cancer models (Fig. 4A), including renal cancer and colorectal cancer models in different mouse genetic backgrounds (BALB/c or C57/BL6). The combination was most effective against the Renca tumor model and less effective against MC38 (but still significantly better than either therapy used alone). It is possible the MC38 model may be less immunogenic, and so enhanced induction of antitumor CTL may not be possible or may remain ineffective at enhancing therapeutic responses. The enhanced therapeutic activity of this combination is primarily immune mediated as (i) the combination resulted in significantly greater numbers of CTLs in the spleen that target tumor antigens (relative to either therapy used alone; Fig. 4B); (ii) the combination also resulted in significant increases in the number of CD3⁺CD4⁺ T cells in the tumor relative to either therapy alone (coupled to a decrease in the relative amounts of Tregs), and an increase in CD3⁺CD8⁺ T cells (although this was only significant relative to control or anti-CTLA4 used alone; Fig. 5); and (iii) depletion of either CD8⁺ T cells, IFNγ or NK cells (but not CD4⁺ T cells) resulted in loss of the therapeutic advantage seen with the combination (Fig. 6 and Supplementary Fig. S4). It therefore appears that immune enhanced oncolytic vaccinia strains (such as WR.TK-.B18R-) can activate an adaptive immune response targeting tumor-associated antigens that is dependent on CD8⁺ T cells and requires NK cell involvement during early immune activation (Supplementary Fig. S3B; although NK cells return to close to baseline levels in the tumor by day 10 after viral treatment). Under single viral therapy treatment, this CD8⁺ T-cell immune response is blunted by premature shut down of T-cell proliferation, an effect that can be overcome by adding anti-CTLA4 antibody. If the antibody is added too early, then the virus cannot induce the correct immune response before its immunomediated removal and the benefits are lost.

Together, these data demonstrate how the correct combination of oncolytic virus and anti-CTLA4 antibody results in robust induction of antitumor CTL coupled to targeting of localized immune suppression within the tumor. This combination therefore more efficiently activates the immune response to target the tumor as well as blocking the capacity of the local tumor microenvironment to suppress the resultant immune response, leading to significantly improved therapeutic effects. The clinical examination of these combinations is therefore an exciting prospect.

No potential conflicts of interest were disclosed.

Conception and design: J.J. Rojas, S.H. Thorne

Development of methodology: J.J. Rojas, S.H. Thorne

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.J. Rojas, P. Sampath, W. Hou, S.H. Thorne

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.J. Rojas, W. Hou, S.H. Thorne

Writing, review, and/or revision of the manuscript: J.J. Rojas, S.H. Thorne

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W. Hou

This work was supported by R01CA140215 and R01CA178766, while core facilities funded under the CCSG (P30CA047904) were used, including in vivo imaging, small animal, and flow cytometry facilities.

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.