Armed Oncolytic Adenovirus–Expressing PD-L1 Mini-Body Enhances Antitumor Effects of Chimeric Antigen Receptor T Cells in Solid Tumors

Intratumoral treatment with oncolytic adenoviral vectors expressing an immunomodulatory molecule (Armed Onc.Ads) is safe and has shown some clinical benefit in patients with solid tumors (1). However, local treatment with Armed Onc.Ad has limited antitumor effect against metastasized tumors (1). In addition, Onc.Ads have low transgene capacity (2, 3), limiting the potential to enhance antitumor immunity by adding multiple genetic modifications. We have shown that tumor cells co-infected with Onc.Ad and Helper-dependent Ads (HDAd), which have a cargo capacity of up to 34 kb, and therefore can express multiple immunomodulatory molecules in a single vector, replicate both Onc.Ad and HDAd. Infection with this dual Ad gene therapy (CAd-VEC) leads to multiple cycles of production and release of both the oncolytic and the immunogenic components (4). Although CAd-VEC significantly suppressed tumor growth compared with treatment with either Onc.Ad or HDAd alone in an immunocompetent mouse model (4), it was insufficient to cure bulky or metastasized tumors.

Chimeric antigen receptors (CAR) usually combine the extracellular antigen recognition domains of a monoclonal antibody and a T-cell receptor signaling domain (CAR T cells; ref. 5). CAR T cells can be systemically administered and home to both primary and metastasized tumors (5), overcoming the limited systemic antitumor effects of locally administered Ad-based cancer immunotherapies (1). Striking clinical successes against B-cell malignancies have been reported when CAR T cells are directed to target antigen CD19, which is highly expressed on both malignant and normal B cells (6). Solid tumors have proven trickier, because many express a range of inhibitory cytokines (7) and immune checkpoint ligands (8) that impair the recruitment and sustained activation of effector T cells. Thus, additional immunomodulation is likely required to increase CAR T-cell efficacy against solid tumors.

Recent clinical trials with immune-checkpoint inhibitors have improved tumor-specific T-cell responses (9). PD-L1 expression on solid cancer cells is induced or increased in the presence of Th1 cytokine IFNγ (10), one of the cytokines expressed by activated CAR T-cells (11). CAR-dependent activation of CAR T cells at the tumor site, therefore, may increase the expression of PD-L1 on target cancer cells, decreasing the antitumor effect of CAR T cells through the PD-1:PD-L1 interaction (12).

As there are toxicities associated with systemic infusion of anti-PD-L1 antibody (13), we hypothesized that local secretion of our functional checkpoint blockade through a single combination agent, CAd-VEC, would be simpler, safer and perhaps more efficacious than combining three separate treatment modalities—oncolytic viruses, checkpoint inhibitors and CAR T cells. We hypothesized that a CAd-VEC–expressing anti-PD-L1 mini-antibody (PD-L1 mini-body) could block the PD-1:PD-L1 interaction between CAR T cells and cancer cells locally while lysing tumor cells, and that combining these treatment modalities would yield potent anti-tumor effects in solid tumors. Here, we demonstrate that CAd-VEC expressing a PD-L1 blocking mini-antibody (CAd-VECPDL1) enhances the antitumor effect of CAR T cells against human solid cancer cells in vitro and in vivo. Our “all-in-one” strategy proved more potent than the combination of anti-PD-L1 IgG and CAR T cells with or without additional Onc.Ad pretreatment in vivo.

The HDAd HDΔ28E4EGFP construct containing an EGFP transgene driven by the CMV promoter (HDAdeGFP) was produced as described elsewhere (4). HDAd without transgene (HDAd0) was produced as described elsewhere (14). To generate the PD-L1 mini-body, the anti-human PD-L1 scFv encoding human IgG signal peptide and the single chain variable region of the YW243.55.S70 was fused with a hinge, CH2 and CH3 regions of human IgG1 with a C-terminal HA tag. The PD-L1 mini-body complementary DNA was inserted into the CMV promoter with polyA signal sequences. After confirmation of sequence and expression, this expression cassette was inserted into the pHDΔ28E4 vector, and HDΔ28E4 PD-L1 mini-body (HDAdPD-L1) was rescued as described elsewhere (15). Onc.Ad5Δ24 was produced as described elsewhere (4, 16).

Human prostate cancer cell line PC-3, human non–small cell lung carcinoma cell line A549, human hepatocellular carcinoma cell line HepG2, and human squamous cell carcinoma line SiHa were obtained from the ATCC in 2014. Cell lines were authenticated by using short tandem repeat (STR) profiling by the ATCC. Cells were cultured under the recommended conditions.

Human PBMCs were isolated using Ficoll-Paque Plus according to the manufacturer's instructions (Axis-Shield). For preparation of mature dendritic cells (mDC), PBMCs were cultured in dendritic cell medium (Cell Genix) for 2 hours at 37°C, and non-adherent cells were removed. The remaining monocytes were cultured in DC medium supplemented with 400 U/mL of IL4 and 800 U/mL of GM-CSF. Fresh media with cytokines were supplemented every 3 days. The mDCs were induced by addition of TNFα, PGE-1, IL1β and IL6 on day 6 and cultured for 48 hours (17). CD4⁺ T cells were isolated from PBMCs using MACS column according to the manufacturer's instructions (Miltenyi Biotec).

CD4⁺ T cells were cultured in 96-well round bottom plates together with allogeneic mDCs at a ratio of 10:1, using CTL medium (18). Anti-human PD-L1 IgG, Isotype IgG (Biolegend), medium of A549-infected with HDPD-L1 or HDeGFP were added, as described in Figure legends. Supernatants were collected at 5 days after coculturing CD4⁺ T cells with allogeneic mDCs, and IFNγ levels in media were measured using the BD cytokine multiplex bead array system (BD Biosciences) according to the manufacturer's instructions. Cells were labeled with ³H-thymidine for an additional 18 hours to measure T-cell proliferation.

Human cancer cells genetically modified to express EGFP were seeded in 12-well plates and infected with 1,000 viral particles (vp) per cell of HDAds or treated with 10 μg/mL of anti-human PD-L1 IgG, Isotype IgG (Biolegend). HER2.CAR T-cells with an effector to target ratio of 1:20 were added 48 hours after infection and cultured for 5 additional days. Residual live EGFP⁺ cancer cells and T cells were counted on the basis of EGFP and CD3 expression with Counting Beads (Life Technologies). Cell numbers were calculated per 5,000 microbeads.

An Immulon 2 high binding 96-well plate (VWR) was coated with 500 ng/well of recombinant human PD-L1 (BioVision). After blocking the plate with PBS-T containing 3% BSA, serially diluted media of A549-infected with 1,000 vp/cell of HDegfp or HDPD-L1 mini were added and incubated at 4°C for 24 hours. Serially diluted anti-human PD-L1 antibody starting from 10 μg/well (BioLegend) was used as a positive control. After washing the plate with PBS-T, HRP-labeled anti-human IgG for PD-L1 mini-body detection or HRP-labeled anti-mouse IgG (Bio-Rad) for anti-human PD-L1 and isotype antibody detection were added and incubated at room temperature for 1 hour. Then we developed the washed plate. Absorbance was measured using Tecan reader (TECAN).

Amounts of PD-L1 mini-body in media of cancer cells infected with 10 vp/cell of HDPD-L1 alone, Onc.Ad alone or with CAd-VECPD-L1 (Onc.Ad:HDAd=1:10) were quantified with an ELISA-based assay. Media of cancer cells infected with Ads were added to an Immulon 2 96-well plate (VWR). Recombinant HA-tag fusion protein (Alpha Diagnostic Intl Inc.) was used as a standard. After blocking with PBS-T containing 3% BSA, anti-HA monoclonal antibody (Clone 5B1D10; Thermo Fisher Scientific) was added to the plate and incubated at 4°C for 24 hours. After washing, HRP-labeled anti-mouse IgG (Bio-Rad) was added and incubated at room temperature for 1 hour, and then we developed the washed plate.

After counting, 1×10⁶ harvested cancer cells (PC-3) were re-suspended in 100 μL of PBS and subcutaneously injected into 5-to 6-week-old NU/J male mice. After the tumor size reached 100 mm³, 1×10⁸ of Onc.Ad, HDPDL1 or CAd-VECPD-L1 (Onc.Ad:HDAd;1:20) were intratumorally injected in a volume of 20 μL. Tumor size was followed and volumes were calculated using the formula: Width² x Length x 0.5. For PD-L1 mini-body detection in tumor and serum samples, tumors and serum were collected at the dates described in Results. Tumors were homogenized with micropestles (VWR) at 4°C, and supernatants of tumor lysates were isolated via centrifugation at 2,000 rpm for 5 minutes. Total protein concentration of serum was measured using the Micro BCA Protein Assay Kit (Thermo Scientific).

A total of 1×10⁶ human cancer cells (PC-3 or SiHa) were re-suspended in a volume of 100 μL of PBS and subcutaneously injected into 5-to 6-week-old NSG mice (PC-3: male mice, SiHa: female mice). After the tumor size reached 100 mm³, a total of 1×10⁷ of Onc.Ad or CAd-VECPD-L1 (Onc.Ad:HDAd;1:20) were intratumorally injected in a volume of 20 μL. Three days after injection of Ads, mice received 1×10⁶ HER2.CAR T-cells intravenously. Tumor size was followed and volumes were calculated using the formula: Width² x Length x 0.5. To track the migration and survival of HER2.CAR T-cells in vivo, T cells were genetically modified to express eGFP.FFLuc (19). Biodistribution of HER2.CAR T-cells was assessed using the In Vivo Imaging System (Xenogen; ref. 19).

After rinsing the collected tumors with PBS, tumors were minced and incubated in RPMI media containing Collagenase type IV (5 mg/mL) and type I (1 mg/mL; Thermo Fisher Scientific) at 37°C for 2 hours (20). Cells were passed through a 70-μm cell strainer (BD Pharmingen) and stained with antibodies described in Supplementary Material.

Cells were infected with 10 Vp/cell of Onc.Ad, HDPD-L1 or CAd-VECPD-L1 (Onc.Ad:HDAd; 1:10) and harvested 48 hours after infection. Tumors were injected with a total of 1×10⁸ Vp of Onc.Ad, HDPD-L1 or CAd-VECPD-L1 (Onc.Ad:HDAd;1:20) and harvested at the indicated time points. Total DNA was extracted from infected cells or tumors, and vector copies were quantified as described in Supplementary Material.

Data were analyzed by one-way ANOVA followed by Ranks protected least significant difference test (SigmaPlot).

Cancer cells upregulate PD-L1 in the presence of IFNγ (10), which is produced by activated T cells. We therefore cocultured T cells expressing second-generation HER2-specific CARs with CD28.ζ-signaling domains (HER2.CAR T cells), which were recently reported to be safe in patients with sarcoma (11), with cancer cells. We evaluated the levels of PD-L1 on the HER2 positive human prostate cancer cell line PC-3 and the squamous cell carcinoma cell line SiHa (Supplementary Fig. S1A), and the expression of PD-1 on HER2.CAR T cells at different time points (Fig. 1A; Supplementary Fig. S1B). When cocultured with HER2.CAR T-cells, PC-3 (Pre: 60%, Post: 100%) and SiHa cells (Pre: 20%, Post: 99%) upregulated PD-L1 expression to levels similar to those induced by recombinant IFNγ treatment (Supplementary Fig. S1A). HER2.CAR T cells also expressed PD-1 within 24 hours of coculture (Pre: 2%, Post: 30%), suggesting HER2.CAR T cells express PD-1 upon activation, though PD-1 is one of multiple exhaustion markers expressed by T-cells (21).

To test whether blocking the PD-1:PD-L1 interaction between cancer cells and HER2.CAR T cells increases cytotoxicity locally at the tumor site, we constructed a HDAd encoding PD-L1 mini-body expression cassette (HDPDL1). We confirmed the dose-dependent expression of PD-L1 mini-body in media of non–small cell lung carcinoma A549 cells infected with HDPDL1 (Fig. 1B). We next evaluated whether the PD-L1 mini-body secreted in media of A549 cells binds to recombinant human PD-L1 protein (Fig. 1C). Although there was no binding of supernatant of A549 cells infected with control HDAd (HDeGFP), the supernatant of A549 cells infected with HDPDL1 had dose-dependent binding to rPD-L1 similar to anti-human PD-L1 IgG. To evaluate whether PD-L1 mini-body can promote T-cell responses by blocking the PD-1:PD-L1 interaction, supernatant from HDPDL1-infected A549 cells was added to an allogeneic mixed lymphocyte reaction (MLR; Fig. 1D; refs. 18, 22). In the presence of PD-L1 mini-body or anti-human PD-L1 IgG, IFNγ release increased 10-fold compared with an MLR in the presence of isotype IgG or supernatant of A549 cells infected with HDeGFP. The levels of IFNγ release were dependent on the dose of PD-L1 mini-body or anti-human PD-L1 IgG (Supplementary Fig. S2A). PD-L1 mini-body and anti-human PD-L1 IgG also enhanced T-cell proliferation compared to controls (Supplementary Fig. S2B). These results indicate that PD-L1 mini-body secreted from HDPDL1-infected A549 cells blocks the PD-1:PD-L1 interaction similarly to anti-human PD-L1 IgG.

To evaluate whether PD-1:PD-L1 blockade by PD-L1 mini-body enhances HER2.CAR T-cell cancer cell killing, we cocultured HER2.CAR T cells with PC-3 and SiHa infected with HDPDL1 or control HDAd. We also performed coculture experiments in the presence of 10 μg/mL anti-human PD-L1 IgG or isotype IgG as controls (Fig. 2). Live cells were counted after 5 days coculture. Although there was no difference in live cancer cells between untreated and PC-3 cells treated with control HDAd or isotype IgG, PC-3 treated with anti-PD-L1 IgG or infected with HDPDL1 enhanced cancer cell killing by HER2.CAR T cells 2- to 3-fold. PD-L1 blockade also increased HER2.CAR T-cell expansion 1.3- to 2-fold (Fig. 2A) and proliferation compared with controls (Supplementary Fig. S3). During the same coculture experiment with SiHa cells, only those infected with HDPDL1 had 60% lower cell numbers compared with control groups. HER2.CAR T-cells in the presence of PD-L1 mini-body expanded 5-fold more than other groups (Fig. 2B).

Importantly, we observed no toxicity due to HDAd infection, PD-L1 mini-body or IgG treatments compared with untreated cells (Fig. 2). These results indicate that the cancer cell killing seen in our coculture experiments was dependent on HER2.CAR T cells and that PD-1:PD-L1 blockade can enhance the killing effect of these cells despite increased expression of PD-L1 on cancer cells in the presence of CAR T-cells (Fig. 1A).

To test whether coinfection of Onc.Ad with HDPD-L1 could amplify PD-L1 mini-body in transduced human cancer cell lines, as previously shown with HDAd transgenes (4), we coinfected HDPD-L1 with Onc.Ad (CAd-VECPDL1) into human solid cancer cell lines (non-small cell lung carcinoma A549, prostate cancer PC-3, squamous cell carcinoma SiHa, hepatocellular carcinoma HepG2), and evaluated the levels of PD-L1 mini-body in media 48 hours after infection (Fig. 3A). A549, PC-3 and HepG2 cells infected with a CAd-VECPDL1 had 4-fold (A549), 15-fold (PC-3), 10-fold (HepG2) higher expression of PD-L1 mini-body in media compared with cells infected with HDPDL1 alone. We confirmed the amplified production of PD-L1 mini-body by Western blotting (Supplementary Fig. S4A). To test whether increased expression was dependent on the amplification of HDAd vector DNA, we quantified both HDAd and Onc.Ad vector copies using primer sets for each backbone 48 hours after infection (Fig. 3B). Cells infected with the CAd-VECPDL1 had 1,000-fold more HDAd vector copies than cells infected with HDAd alone. To verify whether CAd-VECPDL1 induces both amplification of HDAd and the lytic effects of Onc.Ad, cellular lysis of CAd-VECPDL1 was evaluated with MTS assay 96 hours after infection (Fig. 3C), and the LD50 of each treatment in each cell line was determined (Supplementary Table S1). Although Onc.Ad alone was more toxic than CAd-VECPDL1, CAd-VECPDL1 had dose-dependent lytic effects in infected cancer cell lines.

To examine the PD-L1 mini-body expression and anti-tumor effect of CAd-VECPDL1 in vivo, nude mice were subcutaneously transplanted with PC-3. After the tumor volume reached 100 mm³, we injected 1×10⁸ viral particles (Vp) of Onc.Ad, HDPDL1 or CAd-VECPDL1 intra-tumorally. Tumor samples were collected at different time points after injection, and the presence of PD-L1 mini-body in tumors was evaluated by Western blot analysis (Fig. 3D). PD-L1 mini-body was detected in lysates from tumors injected with HDPDL1 alone or CAd-VECPDL1 over time. PD-L1 mini-body was not detectable in the serum of mice from either group (Supplementary Fig. S4B). There was no difference in tumor growth between untreated mice and those treated with HDPDL1 alone, indicating that PD-L1 mini-body expression at the tumor site alone does not suppress tumor growth. These results are consistent with those from our in vitro experiments (Fig. 2). Mice treated with CAd-VECPDL1 demonstrated 60% lower tumor volume compared to control mice at 35 days after injection, as did those injected with Onc.Ad alone, indicating that the Onc.Ad-dependent lytic effects are maintained in vivo (Fig. 3E). We also quantified Onc.Ad and HDPDL1 vector copies in tumors at 35 days after injection (Fig. 3F) and found that tumors injected with CAd-VECPDL1 exhibited 100-fold more HDAd copies than those injected with HDPDL1 alone.

To test whether pretreatment of CAd-VECPDL1 enhances the antitumor effects of adoptively transferred HER2.CAR T cells in xenograft models, NSG mice were subcutaneously transplanted with PC-3, and 1×10⁷ Vp of Onc.Ad or CAd-VECPDL1 were intratumorally injected after the tumor reached 100 mm³. Control mice were intratumorally injected with vehicle alone. A total of 1×10⁶ HER2.CAR T cells genetically modified to express firefly luciferase (ffLuc) were systemically infused 3 days after Ad injection (19). Although PD-L1 mini-body increased HER2.CAR T-cell expansion compared to HER2.CAR T-cells alone in vitro (Fig. 2A), there were no differences in ffLuc activity at tumor sites in animals treated with HER2.CAR T cells alone or CAd-VECPDL1 with HER2.CAR T cells (Fig. 4A). In addition, we found that mice treated with Onc.Ad and HER2.CAR T cells initially had 50-90% less ffLuc activity post-HER2.CAR T-cell injection than those treated with CAR T cells alone. The decreased activity reversed at 14 days, such that mice from all three treatment groups had similar ffLuc activity. We detected PD-L1 mini-body in tumor samples 10 days after HER2.CAR T-cell injection (Supplementary Fig. S5A). Although we observed both adenovirus and T cells at tumor sites 10 days after HER2.CAR T-cell injection, it was unclear whether they colocalized there (Supplementary Fig. S5B). These results suggest that adoptive HER2.CAR T-cell treatment has a minimal impact on Onc.Ad-dependent oncolysis.

Although we did not observe increased T-cell expansion in mice pretreated with CAd-VECPDL1 compared with the other groups, mice pretreated with Ad gene therapy (either Onc.Ad or CAd-VECPDL1) had suppressed tumor growth compared to mice treated with HER2.CAR T cells alone (Fig. 4B). While mice treated with CAd-VECPDL1 had similar median survival to mice treated with HER2.CAR T-cells (60 days), mice treated with CAd-VECPDL1 and HER2.CAR T cells had 2-fold longer median survival (110 days) than mice treated with a single agent (Fig. 4C). These results indicate that PD-L1 mini-body-dependent HER2.CAR T-cell activation at the tumor site enhances the antitumor effect of HER2.CAR T cells.

To evaluate how Ad gene therapy phenotypically impacts HER2.CAR T cells and cancer cells, we collected tumors from mice 84 days after Ad injections. We phenotyped infiltrated T cells and residual (recurrent) cancer cells (Fig. 4D and E) and found approximately 60% of T cells isolated from tumor samples still expressed HER2.CAR (Fig. 4D). We also phenotyped HER2.CAR T cells at different time points to confirm the T cells retained HER2.CAR expression (Supplementary Fig. S5C). T cells isolated from tumors treated with Ad gene therapy had more CD8⁺ T cells, but there was no difference in memory phenotype (CCR7/CD45RO) of CD8⁺ or CD4⁺ T cells (Supplementary Fig. S5D). Although there were similar expression levels of other exhaustion markers (Tim-3 and LAG-3; Supplementary Fig. S5D), HER2.CAR T cells in mice treated with CAd-VECPDL1 had 30% lower PD-1 expression compared to other groups. Since PD-1⁺ T-cell populations are mostly CD4⁺ (Supplementary Fig. S1C; ref. 23), the reduced CD4⁺ population in CAd-VECPDL1-treated mice may correlate with fewer PD-1⁺ T cells. Because HER2.CAR T cells still expressed the CAR, we analyzed HER2 expression on cancer cells from tumors 84 days after Ad injection (Fig. 4E). We evaluated HER2 expression on human CD47⁺ cells, which is highly expressed in human cancers, including prostate cancer, as tumor cells also contained mouse stroma cells (Supplementary Fig. S5E; refs. 24, 25). Human CD47⁺ cells had 80-90% lower HER2 expression compared with PC-3 cells cultured in vitro, indicating that PC-3 cells downregulated HER2 expression. Interestingly, cancer cells in a mouse pretreated with CAd-VECPDL1 showed 50% less PD-L1 expression compared with other groups (Fig. 4E). These results suggest that constitutive blockade of PD-L1 by PD-L1 mini-body leads to downregulation of PD-L1 expression on remaining cancer cells.

We performed the same experiments with NSG mice subcutaneously transplanted with SiHa cells (Fig. 5). Although 1×10⁶ HER2.CAR T-cell alone minimally improved median survival (30 days) compared to untreated mice (24 days), mice treated with CAd-VECPDL1 and HER2.CAR T cells had 2-fold longer median survival (42 days) than untreated mice (Fig. 5C). We collected tumors from mice treated with HER2.CAR T-cells with and without Ad treatments and phenotyped both the T cells and tumor cells (Fig. 5D and E). Tumor cells showed reduced HER2 expression in the presence of HER2.CAR T cells, and only the tumor sample pretreated with CAd-VECPDL1 showed 50% less PD-L1 expression compared with other groups, similar to what we see in the PC-3 model.

As anti-PD-L1 IgG is FDA approved for use in humans (9), we compared local blockade of the PD-1:PD-L1 interaction by PD-L1 mini-body with systemic anti-PD-L1 IgG treatment in NSG mice with PC-3 tumors. We systemically injected anti-PD-L1 IgG into mice treated with Onc.Ad and HER2.CAR T cells and mice treated with HER2.CAR T cells alone (Fig. 6A). Because pretreatment of CAd-VECPDL1 leads to PD-L1 mini-body expression at the tumor site before HER2.CAR T-cell infiltration, we infused anti-PD-L1 IgG or isotype IgG before HER2.CAR T-cell injection. Intratumoral injection of CAd-VECPDL1 showed significantly better anti-tumor effects than systemic anti-PD-L1 IgG in mice treated with both Onc.Ad and HER2.CAR T-cell and those treated with HER2.CAR T-cell alone (Fig. 6B). Overall, mice that received CAd-VECPDL1 had 2-fold longer median survival (110 days) compared with the IgG groups (59 days; Fig. 6C), indicating local expression of PD-L1 mini-body is more effective than systemic anti-PD-L1 IgG treatment. Next, we examined HER2.CAR T-cell expansion at tumor sites using ffLuc (Fig. 6D). Mice pretreated with anti-PD-L1 IgG had 30-90% lower ffLuc activity at tumor sites compared to mice injected with CAd-VECPDL1 or mice injected with isotype IgG at 1, 3 and 7 days post-HER2.CAR T-cell injection. Mice pretreated with anti-PD-L1 IgG also had less ffLuc activity at the ventral side 1 day post-HER2.CAR T-cell injection (Supplementary Fig. S6), suggesting that systemic administration of anti-PD-L1 IgG before HER2.CAR T-cell infusion reduces the total number of circulating HER2.CAR T cells. Because intratumoral CAd-VECPDL1 treatment has minimal distribution in the blood (Supplementary Fig. S4B), local PD-L1 mini-body expression may not reduce the number of HER2.CAR T cells before they reach the tumor site, as was seen in mice systemically treated with anti-PD-L1 IgG.

Here, we demonstrate that PD-L1 mini-body expressed by CAd-VECPDL1 can block the PD-1:PD-L1 interaction between HER2.CAR T cells and cancer cells while maintaining cancer cell oncolysis. The combinatorial effect of Onc.Ad-mediated oncolysis and PD-L1 mini-body–mediated blockade augments the antitumor effect of adoptively transferred HER2.CAR T cells. Together, CAd-VECPDL1 and HER2.CAR T-cell treatment significantly prolonged animal survival compared with treatment with HER2.CAR T cells alone and to Onc.Ad with HER2.CAR T cells in a xenograft mouse model of prostate cancer. Local blockade of the PD-1:PD-L1 interaction by CAd-VECPDL1 also induced superior antitumor effects compared to systemic administration of anti-PD-L1 IgG in combination with adoptive transfer of HER2.CAR T cells in our xenograft mouse model.

Immune checkpoint inhibitors (e.g., PD-1, PD-L1 and CTLA-4) have been successful in treating multiple solid tumors, leading to more robust and sustained T-cell responses (9); however, many patients do not respond or subsequently relapse (26). To enhance host immune responses to cancer cells, checkpoint inhibitors have been combined with other therapeutic agents including chemotherapy, radiotherapy and oncolytic viral gene therapy (Onc.Vs) in preclinical models and clinical trials (26). Because Onc.Vs selectively lyse cancer cells and induce proinflammatory responses, combining immune checkpoint inhibitors with Onc.Vs seems to be a natural marriage (1). However, the therapeutic effect of this combination relies on the patient's immune response, and whether this combination is safe remains unknown (1). To overcome these limitations, we aimed to concentrate checkpoint inhibition locally at the tumor site, thereby minimizing off-target toxicities. We tested in two murine xenograft models the combination of Onc.Ad-expressing an anti-PD-L1 checkpoint inhibitor with adoptively transferred CAR-modified T cells, as local checkpoint blockade should increase their potency. We hypothesized that this combination would create a proinflammatory tumor microenvironment through oncolysis by Onc.Ad and block the deleterious PD-1:PD-L1 interaction locally, enabling increased CAR T-cell activity.

We found that constitutive expression of PD-L1 mini-body at the tumor site had superior antitumor effects than systemic treatment of anti-PD-L1 IgG in the presence of HER2.CAR T cells. Ten days after HER2.CAR T-cell injection, some mice that previously received systemic infusion of anti-PD-L1 IgG, but not mice infused with isotype IgG, had transient diarrhea. Similar side effects are seen in patients with renal cell carcinoma receiving atezolizumab (anti-PD-L1 antibody; ref. 13), suggesting that systemic treatment of anti-PD-L1 IgG with HER2.CAR T cells leads to immune-related side effects in mice. In contrast, intratumoral administration of CAd-VECPDL1 followed by systemic HER2.CAR T-cell treatment caused no immune-related side effects (e.g., diarrhea, weight loss). Because PD-L1 mini-body expressed by CAd-VEC is localized at the tumor site with minimal circulation in the blood, our treatment may minimize toxicities related to systemic treatment of anti-PD-L1 IgG. Our results also suggest that systemic administration of anti-PD-L1 IgG before HER2.CAR T-cell infusion reduces the total number of circulating HER2.CAR T cells. Stimulated naïve T cells express both PD-L1 and PD-1, and blocking the PD-1:PD-L1 interaction leads to apoptosis through Fas:Fas ligand (FasL). HER2.CAR T cells in mice may induce the expression of both PD-1 and PD-L1 through their TCR and/or CAR, and therefore blockade of the PD-1:PD-L1 interaction outside the tumor (e.g., in the lung) may induce unwanted Fas:FasL-dependent T-cell apoptosis (27).

Intratumoral administration of Onc.Vs, including Onc.Ads, has limited distribution to metastasized tumors (28), and investigators have been developing viral vectors to target both primary and metastasized tumors systemically. However, patients who received systemic Onc.Vs developed both Th1 and Th2 adaptive immune responses to the viruses after the first treatment, reducing the antitumor efficacy of the second challenge due to rejection of the virus and infected cells (29). Because CAR T cells are generally generated from patients’ own peripheral blood and are therefore better tolerated than Onc.Vs, CAR T cells can be infused repeatedly to cancer patients (30). Our combination of intratumoral CAd-VEC with systemic CAR T-cell therapy could overcome the inherent limitation of Onc.Vs to create enhanced antitumor effects.

In this study, our “all-in-one” strategy attenuated the immunosuppressive effects of the PD-1:PD-L1 interaction on adoptively transferred CAR T cells at tumor sites, leading to superior antitumor effects and prolonged animal survival in a prostate cancer xenograft mouse model. This combinatorial treatment also showed significant effects in mice with subcutaneous SiHa tumors, although the potency was lower, suggesting that blocking the PD-1:PD-L1 interaction alone may be insufficient to maximize the antitumor effect of CAR T cells in particularly aggressive solid tumors. Because CAd-VEC is able to deliver multiple immunomodulatory molecules (up to 34 kb) in a single HDAd vector with Onc.Ad-dependent lytic effect, in the future we can incorporate additional molecules to maximize the anti-tumor effect of adoptively transferred CAR T-cells. A previous study showed that Onc.Ad-expressing chemokine can enhance the infiltration of adoptively transferred GD2.CAR T cells at the tumor site in a neuroblastoma xenograft model, and Onc.Ad-expressing cytokine enhances proliferation of those CAR T cells, leading to superior antitumor effects than the combination of Onc.Ad (without transgene) with GD2.CAR T cells (19). We will investigate whether combining PD-L1 mini-body with chemokines and/or cytokines further enhances the antitumor effects of adoptively transferred CAR T cells.

Overall, we demonstrate that combined CAd-VECPDL1 and CAR T-cell treatment has significant antitumor effects against bulky solid tumors. Because intratumoral injection of Onc.Ads has been tested in patients with a range of solid tumors (31), our concept could readily be applied to other solid tumors and used with CARs targeting different surface molecules (e.g., GD2, PSCA; refs. 19, 32).

M.K. Brenner is an SAB in Tessa Therapeutics, Unum Therapeutics, Torque, Nantkwest, Bluebird Bio., Turnstone, Cellmedica; has ownership interest (including patents) in and is a consultant/advisory board member for Patent. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Rosewell Shaw, M. Brenner, M. Suzuki

Development of methodology: K. Tanoue, A. Rosewell Shaw, C. Porter, M. Suzuki

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Tanoue, A. Rosewell Shaw, N. Watanabe, C. Porter, B. Rana

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Tanoue, A. Rosewell Shaw, N. Watanabe, C. Porter, B. Rana, S. Gottschalk

Writing, review, and/or revision of the manuscript: K. Tanoue, A. Rosewell Shaw, S. Gottschalk, M. Brenner, M. Suzuki

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Rosewell Shaw, S. Gottschalk

Study supervision: M. Brenner, M. Suzuki

The authors would like to thank Dr. Brendan Lee in the Department of Molecular and Human Genetics at Baylor College of Medicine for his support of the project and Catherine Gillespie in the Center for Cell and Gene Therapy at Baylor College of Medicine for her editing of the article.

This work was supported by NIH (R00HL098692 to M. Suzuki), T32HL092332 to A. Rosewell Shaw, BCM Head and Neck Seed Grant, and Concern Foundation to M. Suzuki. This work was also supported by NIH P01 CA094237 to M.K. Brenner and S. Gottschalk, and we appreciate the use of shared resources supported by NIH P30 CA125123. K. Tanoue was supported by Strategic Young Researcher Overseas Visits Program for Accelerating Brain Circulation, Japan Society for the Promotion of Science.

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.