Targeting Fc Receptor-Mediated Effects and the “Don't Eat Me” Signal with an Oncolytic Virus Expressing an Anti-CD47 Antibody to Treat Metastatic Ovarian Cancer

Ovarian cancer is the most lethal gynecologic malignancy (1–3). Although the majority of women with ovarian cancer initially respond to surgery and platinum-based chemotherapy, over 80% develop platinum-resistance, relapse, and die. Women with platinum-resistant tumors have poor prognosis as current treatments are usually ineffective. Increasing evidence from preclinical and clinical data indicates that immunotherapeutic approaches that harness and enhance antitumor effector cells, such as immune checkpoint blockade and specific mAb therapy, have led to clinical benefits for ovarian cancer (4).

CD47 was first identified as a tumor antigen on human ovarian cancer in the 1980s, and its overexpression is associated with a poor prognosis in ovarian cancer (5–7). CD47 provides a “don't eat me” signal when it binds to its receptor, signal regulatory protein alpha (SIRPα), on macrophages, thereby suppressing phagocytosis (8–10). Anti-CD47 antibodies, including the human IgG1 scaffold and IgG4 scaffold, are being actively tested in clinical trials (11, 12).

The traditional methods of delivering therapeutic mAb or mAb–drug conjugates often result in a limited distribution to tumor tissue and frequently create systemic toxicity due to off-tumor, on-target effects, or Fc receptor-mediated activation of innate immune effector cells (13, 14). However, oncolytic virus (OV) is an ideal carrier that is able to specifically deliver a payload into the tumor environment after administration (15). More importantly, the infection of OV can dramatically activate immune responses at the local tumor microenvironment, in turn, improving the efficacy of the antibody or other payloads delivered with the OV (15–18).

In this study, we generated novel oncolytic herpes viruses (oHSV) termed OV-aCD47-G1 encoding a full-length anti-CD47 antibody on a human IgG1 scaffold (αCD47-G1) and OV-aCD47-G4 encoding a full-length anti-CD47 antibody on a human IgG4 scaffold (αCD47-G4). Both OV-αCD47-G1 and OV-aCD47-G4 blocked the “don't eat me” signal of macrophages, but only OV-αCD47-G1 induced antibody-dependent cellular phagocytosis (ADCP) of macrophages and antibody-dependent cellular cytotoxicity (ADCC) of NK cells targeting ovarian cancer cells. In ovarian cancer mouse models, the administration of OV-aCD47-G1 or the mouse counterpart, OV-amCD47-G2b significantly inhibited local tumor growth as well as had an abscopal effect in an ovarian cancer metastasis mouse model.

All experiments and animal handling were conducted under federal, state, and local guidelines and with approval from the City of Hope Animal Care and Use Committee. Peripheral blood cones were collected from healthy donors with a written informed consent under a protocol approved by the City of Hope Institutional Review Board.

The human ovarian cancer cell line A2780, the Chinese hamster ovary cell line CHO, the monkey kidney epithelium-derived Vero cell line, and the mouse ovarian cancer cell line ID8 were cultured with DMEM supplemented with 10% FBS, penicillin (100 U/mL), and streptomycin (100 μg/mL). CHO cells were used for producing αCD47-G1 and αCD47-G4 antibodies. Vero cells were used for viral propagation and plaque-assay-based viral titration. The A2780 cell line was a recent gift from Dr. Edward Wang's lab at City of Hope. The CHO cell line was recently purchased from ATCC. All cell lines used in this study were not authenticated upon receipt but are routinely tested for the absence of mycoplasma by using the MycoAlert Plus Mycoplasma Detection Kit from Lonza.

OV-αCD47-G1, OV-αCD47-G4, OV-αmCD47-G2b, and OV-αmCD47-G3 were generated by using a published protocol with slight modifications (15, 19). For expressing full-length αCD47-G1 and αCD47-G4 simultaneously in the viral vector, a DNA sequence encoding a T2A self-cleaving peptide was used to link the light chain and heavy chain coding genes. For expressing αmCD47-G2b and αmCD47-G3 in the viral vector, a mouse anti-CD47 antibody single-chain variable fragment was fused with different mouse IgGs. The engineered viral vectors were recombined with fHsvQuik-1 for engineering the corresponding oHSV. Virus titration was performed using plaque assays using Vero cells, as we reported previously (15, 18). The viral particle purification and concentration were conducted also following the published protocol (15, 18).

A2780 cells were subjected to a saturated infection with OV-Q1, OV-αCD47-G1, or OV-αCD47-G4. Two hours after infection, the infection media were changed with DMEM supplemented with 10% FBS. The supernatants from each group were collected at 6, 12, 24, 48, and 72 hours after infection. The antibody concentrations of the indicated timepoints were measured by ELISA as previously reported with slight modifications (20). The standard curve was constructed with αCD47-G1 and αCD47-G4 antibodies purified from CHO cells with known concentrations with the recombinant human CD47 protein (Abcam, ab174029) used as a coating reagent. The anti-human Fc antibody (sigma, MAB1307) was the detecting antibody.

To compare the concentration of αmCD47-G2b at tumor sites in the abdomen and the systemic bloodstream, 6- to 8-week-old female C57BL/6 mice were intraperitoneally injected with 2 × 10⁶ wild-type ID8 cells at day 0. At day 4, mice were randomly divided into groups that were intraperitoneally injected with either 1 × 10⁶ PFU OV-Q1 or the same amount of OV-αmCD47-G2b in 100 μL saline. Saline alone was used as a placebo control. Mice were euthanized at days 5, 6, 7, and 10. The peripheral blood and tumors in the abdomen were collected. Sera were prepared from the peripheral blood and the tumors were homogenized for ELISA assay to detect the concentrations of αmCD47-G2b. The ELISA standard curve was constructed with αmCD47-G2b antibodies purified from CHO cells with known concentrations determined by NanoDrop One (Thermo Fisher Scientific). The recombinant mouse CD47 protein (Abcam, ab231160) served as a coating reagent and the anti-mouse Fc antibody (Invitrogen, 31439) was the detecting antibody.

The in vitro oncolysis assay was measured by real-time cell analysis (RTCA) using xCELLigence RTCA MP instrument. Briefly, A2780 cells were seeded onto 96-well plates at densities of 5,000 cells/well and allowed to attach for 24 hours. Cells were then treated with graded concentrations of OV-Q1, OV-αCD47-G1, or OV-αCD47-G4. The oncolysis assay was monitored and analyzed with RTCA software (ACEA Biosciences, Agilent).

For isolating and culturing human primary macrophages, human monocytes were isolated from peripheral blood of healthy donors and enriched by using the RosetteSep Human Monocyte Enrichment Cocktail Kit (Stemcell, Catalog No. 15068). RPMI 1640 medium containing 20 ng/mL human M-CSF (PeproTech, Catalog No. 300-25-50UG) and 2% human serum was used to differentiate monocytes into macrophages.

Murine bone marrow-derived macrophages (BMDM) were isolated from BALB/c mice and induced differentiation by RPMI 1640 medium containing 10 ng/mL murine M-CSF (PeproTech, Catalog No. 315-02-50UG) and 10% FBS.

For the phagocytosis assay of human primary macrophages, A2780 cells prelabeled with CFSE (Thermo Fisher Scientific, C34554) were cocultured with human macrophages at a ratio of 2:1 for 4 hours in the presence of vehicle control, αCD47-G1 or αCD47-G4 at the dose of 5 μg/mL or supernatants of OV-infected cells at 37°C in ultra-low-attachment 96-well U-bottom plates (Corning) in serum-free 1640 (Life Technologies). Then the cells were harvested and stained with antihuman CD45 (BD Biosciences, 552850) to identify macrophages. All flow cytometry data were collected using a Fortessa X20 flow cytometer (BD Biosciences). Phagocytosis was calculated as the number of CD45⁺CFSE⁺ macrophages, quantified as a percentage of the total CD45⁺ macrophages.

For the phagocytosis of mouse macrophages, ID8 or A2780 cells prelabeled with CFSE were cocultured with murine macrophages at a ratio of 2:1 for 6 hours with 5 μg/mL antibodies or supernatants of OV-infected cells at 37°C in ultra-low-attachment 96-well U-bottom plates (Corning) in the serum-free 1640 media (Life Technologies). The cells were harvested. An anti-mouse F4/80 antibody (BD Biosciences, Catalog No. 565787) was used to identify macrophages. Phagocytosis was measured as the number of F4/80⁺CFSE⁺ macrophages and quantified as a percentage of the total F4/80⁺ macrophages.

For evaluating the effect of αCD47-G1 and αCD47-G4 on activating transcription of typical human macrophage cytokine genes, human macrophages and A2780 cells were cocultured at a ratio of 1:1 for 6 hours with or without 5 μg/mL αCD47-G1 or αCD47-G4. Then human macrophages were FACS-sorted by staining with an anti-CD45 antibody using a BD Aria Fusion flow cytometer. The total RNA was extracted for measuring the relative transcription of human IL1B, IL6, IL10, IL12A, Arg1, TNF, CCL2, CD206, and NOS2 genes with a pair of corresponding primers for each gene. 18s rRNA was used as an internal control (Supplementary Table S1).

A2780 cells prelabeled with ⁵¹Cr were used as target cells. Effector cells were primary human NK cells isolated from peripheral blood cones using an NK-cell Isolation Kit (MACSxpress Miltenyi Biotec) and an Erythrocyte Depletion Kit for the removal of red blood cells (Miltenyi Biotec). The target cells were incubated with 1 μg/mL αCD47-G1 or αCD47-G4 antibodies or vehicle for 30 minutes. Then the target cells were cocultured with isolated human primary NK cells at different effector: target ratios at 37°C for 4 hours. Release of ⁵¹Cr was measured with a MicroBeta² microplate radiometric counters (Perkin Elmer). The cytotoxicity was calculated as previously reported (15, 21). At least three technical replicates with NK cells from different donors were performed. The expression of CD69 and granzyme B of NK cells were measured 12 hours after coculturing with A2780 cells with or without 1 μg/mL αCD47-G1 or αCD47-G4 antibodies or 100 μL supernatants derived from OV-Q1-, OV-αCD47-G1-, and OV-αCD47-G4-infected A2780 ovarian cancer cells at 48-hour after infection at a ratio of 1:1 by flow cytometry after being stained with the anti-CD56 antibody (BD Biosciences, Catalog No. 557919), anti-granzyme B antibody, anti-CD69 (BD Biosciences, Catalog No. 562883) antibody. Other NK-cell activation markers were measured 12 hours after coculturing with A2780 cells at a ratio of 1:1 in the presence or absence of 1 μg/mL αCD47-G1 or αCD47-G4 antibodies or 100 μL supernatants derived from OV-Q1-, OV-αCD47-G1-, or OV-αCD47-G4-infected A2780 cells, harvested 12 hours post infection, by flow cytometry after being stained with the anti-CD56 antibody (BD Biosciences, Catalog No. 557919), anti-4–1BB antibody (Biolegend, Catalog No. 309804), anti-CD107a antibody (BD Biosciences, Catalog No. 555800), anti-CD27 antibody (Biolegend, Catalog No. 356404), anti-CD11b antibody (Biolegend, Catalog No. 301306), and anti-Ki67 antibody (BD Biosciences, Catalog No. 564071).

Six- to eight-week-old female athymic nude mice were purchased from The Jackson Laboratories. For survival studies in an immunodeficient model, mice were anesthetized and subcutaneously injected with 5 × 10⁶ A2780 cells into the right flank. One day after A2780 cells injection, animals were subsequently randomly divided into groups that were injected intratumorally either with 2 × 10⁵ PFU oHSV (OV-Q1, OV-αCD47-G1, or OV-αCD47-G4) in 10 μL of saline or with saline as control. Mice were subsequently monitored for ovarian cancer progression. Mice were euthanized when the diameter of tumors was over 15 mm.

For establishing the immunocompetent mouse ovarian tumor model, 6- to 8-week-old female C57BL/6 mice were purchased from The Jackson Laboratories. The mice were anesthetized and intraperitoneally injected with 2 × 10⁶ wild-type ID8 cells or ID8-hCD47 cells, which express the human CD47 gene. After the cells grew for 3 days in vivo, mice were randomly divided into groups that were intraperitoneally injected either with 1 × 10⁶ PFU oHSV (OV-Q1, OV-αCD47-G1, OV-αCD47-G4, OV-αmCD47-G2b, or OV-αmCD47-G2b) in 100 μL of saline. Saline alone was used as a placebo control. Mice were subsequently monitored and weighed frequently to monitor ovarian cancer disease progression. Luciferase-based in vivo images were taken after tumor implantation to evaluate the tumor development. Mice were euthanized when they became moribund.

Mononuclear cells in the tumor site, liver, and kidney were extracted with Percoll and stained with anti-NKp46 (Biolegend, Catalog No. 137618), anti-CD3 (BD Biosciences, Catalog No. 553066), anti-CD45 (BD Biosciences, Catalog No. 559864), anti-CD11b (BD Biosciences, Catalog No. 552850), anti-F4/80 (Thermo Fisher Scientific, Catalog No. 12–4801–82), anti-CD4 (BD Biosciences, Catalog No. 557308), anti-CD8 (BD Biosciences, Catalog No. 553030), anti-CD86 (BD Biosciences, Catalog No. 740900), anti-CD206 (Biolegend, Catalog No. 141734), anti-CD69 (BD Biosciences, Catalog No. 564683), anti-IFNγ (BD Biosciences, Catalog No. 563773), and/or anti-granzyme B (Thermo Fisher Scientific, Catalog No. 48–8898–82) antibodies for flow cytometric analysis of immune cell infiltration and activation. H-2K(b) herpes simplex virus type 1 glycoprotein B tetramer was provided by the NIH Tetramer Core Facility. The flow cytometric assessments of murine immune cells were performed with at least three independent animals. All flow cytometry data were collected using the Fortessa X-20 flow cytometer except cell sorting experiments that were performed using FACS Aria II cell sorter (BD Biosciences).

For continuous endpoints that are normally distributed, data are presented as mean ± SD. Student t test was used to compare two independent conditions, and one-way ANOVA was used to compare three or more conditions. For data with repeated measures from the same subject/donor, a linear mixed model was used to account for the underlying variance and covariance structure. P values were adjusted for multiple comparisons by Holm procedure or the Bonferroni method, and a P value of 0.05 or less was defined as statistically significant. Mouse survival data were analyzed by the Kaplan–Meier method and compared by log-rank test. All tests were two-sided. Statistical software GraphPad, R.3.6.3., and SAS 9.4 were used for the statistical analysis.

We first demonstrated a high surface expression of CD47 on the human ovarian cancer cell line A2780 (Fig. 1A). We generated IgG1 and IgG4 versions of the humanized anti (α)-human CD47 antibody. We reconstructed αCD47-G4 by following the previously reported method (22). The human IgG4 constant region of αCD47-G4 was replaced by the human IgG1 for constructing αCD47-G1. CHO cells were transduced with lentiviral vectors carrying the corresponding constructs to produce αCD47-G1 or αCD47-G4. To assess the specificity of these mAbs, A2780 cells were first incubated with the increasing concentrations of αCD47-G1 or αCD47-G4 antibodies purified from the lentiviral-infected CHO cells, followed by staining with a BV786-conjugated anti-CD47 antibody (clone: B6H12). Results showed a similar dose-dependent decrease of BV786 signal when the cells were preincubated with increasing concentrations of either αCD47-G1 or αCD47-G4, indicating that both αCD47-G1 and αCD47-G4 have similar recognition of CD47 that blocks CD47 binding to SIRPα (Fig. 1B). OV-αCD47-G1- and OV-αCD47-G4 were constructed by inserting the coding genes of αCD47-G1 and αCD47-G4 into the oHSV backbone (OV-Q1), respectively, driven by the immediately early gene promoter IE4/5 of the virus to get a high-level expression after infection. The construction of OV-αCD47 is shown in Supplementary Fig. S1. When the human ovarian cancer cell line A2780 was infected by OV-αCD47-G1 or OV-αCD47-G4 at a multiplicity of infection (MOI) of 2, anti-CD47 antibodies were detected in supernatants of infected cells as early as 6 hours after infection, reaching a concentration of >5 μg/mL at 24 hours (Fig. 1C). We next assessed whether the inserted gene encoding the anti-CD47 antibody into the oHSV adversely affected the tumor lysis capability of OV-αCD47-G1 or OV-αCD47-G4 when compared with the parental oHSV, OV-Q1. When A2780 cells were infected with varying titers of OV-αCD47-G1, OV-αCD47-G4, or OV-Q1, there was no obvious difference in cell death or oncolysis among the three groups tested (Fig. 1D). Therefore, the insertion of αCD47-G1 and αCD47-G4 coding genes does not affect the oncolytic ability of OV-αCD47-G1 and OV-αCD47-G4 in vitro.

We next determined the effect of αCD47-G1 and αCD47-G4 purified from the lentiviral-infected CHO cells on human macrophage phagocytosis. Both αCD47-G1 and αCD47-G4 effectively induced human macrophage phagocytosis against A2780 ovarian cancer cells compared with control, but αCD47-G1 performed significantly better than αCD47-G4 (Fig. 2A and B). Next, we repeated this experiment with supernatants from OV-αCD47-G1-, OV-αCD47-G4-, or OV-Q1-infected A2780 cells. The results showed that supernatants from OV-αCD47-G1-infected cells significantly enhanced macrophage phagocytosis against A2780 ovarian cancer cells compared with supernatants from OV-Q1- and OV-αCD47-G4-infected cells (Fig. 2C and D). Similar results were obtained with BMDM isolated from BALB/c mice using the antibodies purified from lentiviral-infected CHO cells as well as supernatants from OV-infected A2780 cells (Fig. 2E–H). For both murine and human macrophages, the increased phagocytotic effect of αCD47-G1 compared with vehicle control should result from both the blockade of “don't eat me” signaling and ADCP, whereas the increased effect compared with αCD47-G4 should be mainly from ADCP.

The effects of CHO-derived αCD47-G1 and αCD47-G4 on regulating the cytokine expression of human macrophages were assessed. Human macrophages were incubated with A2780 cells in the presence of αCD47-G1, αCD47-G4, or vehicle for 6 hours, followed by total RNA extraction and quantitative assessment of cytokine transcripts. Compared with αCD47-G4, αCD47-G1 dramatically increased the transcription of macrophage cytokine genes such as IL6, IL10, NOS2, TNF, and IL12A, which are also macrophage M1 markers; however, the transcription of IL1B and CCL2 was similar in both treatment groups. Furthermore, αCD47-G1 and αCD47-G4 had no regulatory effect on M2 macrophage markers, including Arg1 and CD206 (Fig. 2I; Supplementary Fig. S2).

NK cells have an antitumor function because of their natural cytotoxicity and ADCC in the presence of some antibodies. To determine the effect of αCD47-G1 and αCD47-G4 on NK-cell antitumor activity, we performed NK-cell cytotoxicity assays using freshly isolated human NK cells as effector cells and A2780 ovarian cancer cells as target cells. The results showed that αCD47-G1 but not αCD47-G4, both of which were purified from lentivirus-infected CHO cells, induced strong NK-cell ADCC against ovarian cancer cells (Fig. 3A). Next, supernatants from OV-Q1-, OV-αCD47-G1-, and OV-αCD47-G4-infected A2780 ovarian cancer cells were used to repeat this experiment, and comparable results were obtained, that is, only supernatants from OV-αCD47-G1-infected A2780 cells induced strong NK-cell cytotoxicity compared with the OV-Q1 or OV-αCD47-G4 supernatants (Fig. 3B). Furthermore, in the presence of tumor cells, αCD47-G1 but not αCD47-G4 significantly increased the surface expression of the activation marker CD69 on NK cells (Fig. 3C and D), and similar results were seen when using supernatants from OV-Q1-, OV-αCD47-G1-, and OV-αCD47-G4-infected A2780 ovarian cancer cells (Fig. 3E and F). Stronger effects were shown in all OV supernatant conditions than purified antibodies likely because 1 μg/mL antibody concentration was used in Fig. 3C and D whereas the concentration in Fig. 3E and F was >5 μg/mL (Fig. 1C). Both αCD47-G1 and αCD47-G4 significantly upregulated the production of granzyme B in NK cells compared with the control, but αCD47-G1 treatment showed a significantly stronger activated effect on the production of granzyme B in NK cells compared with αCD47-G4 treatment (Fig. 3G and H). We also analyzed Ki67, NKG2D, 4–1BB, CD27, CD11b, CD107a, and IFNγ expression of NK cells. In the presence of tumor cells, αCD47-G1 but not αCD47-G4 significantly increased the surface expression of CD27, 4–1BB, and CD107a on NK cells but had no effect on regulating Ki67, NKG2D, CD11b, and IFNγ expression (Supplementary Figs. S3A–S3C). Similar results were seen when using supernatants from OV-Q1-, OV-αCD47-G1-, and OV-αCD47-G4-infected A2780 ovarian cancer cells (Supplementary Figs. S3D and S3E).

To evaluate the efficacy of OV-αCD47-G1 and OV-αCD47-G4 in an ovarian cancer mouse model, we utilized a previously described xenograft model of ovarian cancer by subcutaneously injecting 5 × 10⁶ A2780 cells into nude mice (23). One day after tumor implantation, animals received an intratumoral injection with OV-αCD47-G1, OV-αCD47-G4, or OV-Q1 at the dose of 2 × 10⁵ plaque-forming unit (PFU) per mouse, or saline as a placebo control. Tumor progression was monitored by measuring the tumor size with a caliper. OV-αCD47-G1 was significantly more effective than OV-αCD47-G4 and OV-Q1 at inhibiting the progression of tumors and prolonging the survival of mice in vivo (Fig. 4A; Supplementary Fig. S4). OV-Q1 moderately slowed tumor progression compared with vehicle control (Fig. 4A–C).

Metastatic ovarian cancer is an advanced stage malignancy with tumor cells in the ovaries migrating to distant areas of the body, including the abdomen, liver, and kidney. To evaluate the role of OV in an ovarian cancer metastasis mouse model, we first detected the virus distribution after OV-Q1 intraperitoneal administration. The virus titer in the liver and kidney was measured by qPCR. Compared with the saline group, the OV group had high titers of the virus in the liver and kidney (Supplementary Fig. S5A), suggesting that OV might have an abscopal effect in controlling metastatic tumors. We next sought to evaluate the efficacy of OV-αCD47-G1 in an immunocompetent ovarian cancer metastasis mouse model. Mouse firefly luciferase (FFL) gene-expressing ID8 ovarian cancer cells were modified to express human CD47 (referred to as ID8-hCD47-FFL cells), so that OV-αCD47-G1 or -G4 can bind to hCD47 on tumor cells. ID8-hCD47-FFL cells also expressed the GFP marker for a tracing purpose. For establishing a mouse model that can be used to test OV-αCD47-G1 and -G4, ID8-hCD47-FFL cells were injected intraperitoneally into immunocompetent wild-type C57BL/6 mice on day 0. On day 4, mice were randomly grouped and treated intraperitoneally with 1 × 10⁶ PFU OV-Q1, OV-αCD47-G1, OV-αCD47-G4, or saline. Tumor progression was monitored by luciferase-based imaging on days 10 and 20 posttumor implantation. Mice treated with OV-αCD47-G1 significantly prolonged their median survival when compared with mice treated with OV-αCD47-G4 or OV-Q1 (Fig. 4C). However, there was no significant difference between OV-αCD47-G4 and OV-Q1 treatments, mainly because of the murine macrophages still receiving the murine-CD47-mediated “don't eat me” signal from the ID8-hCD47-FFL cells, which cannot be blocked by human αCD47-G1 or αCD47-G4. Bioluminescent imaging data and sizes of representative tumors in the abdomen confirmed this result (Supplementary Figs. S5B and S5C). Tumor cells were also detected in the liver and kidney in this ID8-hCD47 immunocompetent metastatic ovarian cancer model. The injections of each OV reduced the tumor burden in the liver and kidney, with a significantly more profound effect for OV-αCD47-G1 than OV-αCD47-G4 and OV-Q1 (Fig. 4D; Supplementary Figs. S5D and S5E). For evaluating the immune responses, cells isolated from the tumor, liver, and kidney were collected then subjected to FACS analysis. Compared with the saline group, more NK cells and T cells infiltrated into tumors in the abdomen of OV-treated mice, but there were no differences among the three OVs (Supplementary Figs. S6A and S7A). It has been reported that anti-CD47 antibody treatment stimulates phagocytosis of glioblastoma by promoting macrophages polarizing towards M1 type in vivo (24). Therefore, we detected the M1 (as a marker of CD86) and M2 (as a marker of CD206) macrophages in the tumor. Flow cytometry data showed that OV-αCD47-G1 treatment significantly decreased the number of M2 type macrophages but increased M1 type macrophages, when compared with the other treatments (Fig. 4E; Supplementary Fig. S8A). OV-αCD47-G4 decreased the M2 type macrophages but showed a negligible effect on regulating M1 type macrophages when compared with OV-Q1 treatment (Fig. 4E; Supplementary Fig. S8A). Our data indicated that OV-αCD47-G1 treatment promoted macrophages polarizing towards M1 type in vivo. The activation status of NK cells in vivo, indicated by the percentages of CD69⁺ NK cells, was also analyzed. Compared with saline, OV-Q1, and OV-αCD47-G4, OV-αCD47-G1 treatment increased the percentages of CD69⁺ NK cells, indicating that OV-αCD47-G1 administration activated NK cells in vivo (Fig. 4F; Supplementary Fig. S8A). Significantly more macrophages, NK cells, and T cells infiltrated into the liver in the OVs-treated mice versus saline-treated mice (Supplementary Figs. S6B and S7B). Although there was no difference in percentages of infiltrated immune cells in the liver among three viruses, M1 type macrophages and activated NK cells were increased in the OV-αCD47-G1 group, compared with the OV-Q1- or OV-αCD47-G4-treated group (Fig. 4G and H; Supplementary Fig. S8B). Similar results for immune cell infiltration and cell activation were found in the kidney (Supplementary Figs. S6C, S7C–S7E, and S8C). Collectively, our data suggest that OV-αCD47-G1 is superior to OV-αCD47-G4 or OV-Q1 for improving oncolytic virotherapy in an immunocompetent ovarian cancer metastasis mouse model as it more effectively polarizes macrophages toward the M1 type and activates NK cells.

To confirm our data, we generated the murine counterpart of OV-αCD47-G1 and OV-αCD47-G4 using the sequences of a murine anti-CD47 antibody (25). According to the binding affinity between mouse IgG and its Fc receptors, αmCD47-IgG2b version and αmCD47-IgG3 version were engineered, corresponding to human αCD47-G1 and OV-αCD47-G4, respectively (26). For assessing the specificity of αmCD47 antibody, ID8 cells were first incubated with increasing concentrations of αmCD47-G2b antibodies purified from the lentiviral-infected CHO cells, followed by staining with an APC-conjugated anti-mouse CD47 antibody. Similar to human αCD47-G1 (Fig. 1B), flow cytometry analysis showed a dose-dependent decrease of the APC signal when cells were preincubated with increasing concentrations of αmCD47-G2b (Fig. 5A). Murine macrophage phagocytosis and NK-cell cytotoxicity against ID8 cells were measured. The results showed that supernatants from OV-αmCD47-G2b- and OV-αmCD47-G3-infected cells significantly enhanced macrophage phagocytosis against ID8 cells compared with those from OV-Q1-infected cells, yet supernatants from OV-αmCD47-G2b-infected cells were superior to those from OV-αmCD47-G3-infected cells (Fig. 5B). These results suggest that the increased phagocytotic effect of αmCD47-G2b over control groups is dependent on both blockade of “don't eat me” signaling and ADCP, whereas the effect of αmCD47-G3 may be mainly from blockade of “don't eat me” signaling. For NK-cell cytotoxicity assay, only supernatants from OV-αmCD47-G2b-infected cells activated murine NK-cell ADCC against ID8 cells (Fig. 5C). For evaluating the virotherapy efficacy of OV-αmCD47-G2b and OV-αmCD47-G3, an immunocompetent ovarian cancer metastasis mouse model was established. Similar to the human version of OV-αCD47-G1, administration of OV-αmCD47-G2b significantly prolonged the median survival of mice when compared with OV-αmCD47-G3 or OV-Q1 (Fig. 5D). Different from the result of human versions in the ID8-hCD47 model where the murine CD47-SIRPα axis-mediated “don't eat me signal” could not be blocked (Fig. 4B), OV-αmCD47-G3 treatment showed a significantly stronger antitumor effect compared with OV-Q1 treatment, exhibiting the effect of blocking the murine CD47-SIRPα axis-mediated “don't eat me” signal (Fig. 5D; Supplementary Fig. S9).

To clarify the role of NK cells and macrophages in OV-αmCD47-G2b virotherapy, we repeated the survival studies with NK-cell depletion and macrophage depletion separately in the ID8 immunocompetent mouse model treated with OV-αmCD47-G2b. NK-cell depletion significantly shortened the survival of OV-αmCD47-G2b-treated mice compared with OV-αmCD47-G2b-treated mice without NK-cell depletion (Fig. 5E). Macrophage depletion also significantly shortened the survival of OV-αmCD47-G2b-treated mice when compared with OV-αmCD47-G2b-treated mice without macrophage depletion (Fig. 5F). These data indicate that both macrophages and NK cells mediate the effects of the OV expressing αmCD47-G2b in improving survival of mice with ovarian cancer.

The cells isolated from the tumor, liver, and kidney were also analyzed. Compared with saline, OV-Q1, and OV-αmCD47-G3, OV-αmCD47-G2b significantly reduced the tumor cell numbers in the liver and kidney (Fig. 6A and B). Different from the result of human versions (Fig. 4D; Supplementary Fig. S5D), OV-αmCD47-G3 treatment also decreased the tumor cell numbers in the liver and kidney compared with the OV-Q1 (Fig. 6A and B), which validated the survival data (Fig. 5D). The immune responses were also analyzed in the tumor, liver, and kidney. Similar to the human version of OV-αCD47, administration of OVs promoted the infiltration of macrophages, NK cells, and T cells into the tumor microenvironment. However, there was no difference among the three OVs in tumors in the abdomen, liver, and kidney (Supplementary Figs. S10A–S10C). We also detected the percentages of M1 and M2 types of macrophages. In tumors from the abdomen, liver, and kidney, OV-αmCD47-G2b and OV-αmCD47-G3 treatments significantly decreased the number of M2 type macrophages and increased the number of M1 type macrophages compared with saline and OV-Q1 treatment (Fig. 6B, D, and F; Supplementary Fig. S11). Similar to the human version of OV-αCD47-G1 (Fig. 4F and H; Supplementary Fig. S7E), the percentages of CD69⁺ NK cells in tumors from the abdomen and in the liver as well as kidney were significantly increased in the OV-αmCD47-G2b group, indicating that OV-αmCD47-G2b administration activated NK cells in vivo (Fig. 6C, E, and G; Supplementary Fig. S11).

We also performed a flow cytometric assessment of the amount of CD4⁺, CD8⁺, and Treg T cell in the spleen of mice treated with OV-Q1, OV-αmCD47-G2b, OV-αmCD47-G3, or saline control in the fully immunocompetent model. We found that after treatment with each of the three oHSVs (OV-Q1, OV-αmCD47-G2b, and OV-αmCD47-G3), there was no difference of CD4⁺ T cells and Tregs. However, CD8⁺ T cells were significantly increased following treatment with OV-αmCD47-G2b compared with the groups treated with OV-Q1 or OV-αmCD47-G3 (Supplementary Fig. S10D). We also assessed IFNγ and granzyme B secretion of CD8⁺ T cells in the spleen. The results showed that after treatment with each of the above three oHSVs, IFNγ and granzyme B secretion were increased without differences among those three oHSV treatments (Supplementary Fig. S10E). We also analyzed antiviral immune responses after OV- αmCD47-G2b treatment by measuring anti-HSV-1 specific CD8⁺ T-cell percentages in the spleen and anti-HSV-1 specific antibodies in the peripheral blood in this model. The anti-HSV-1 specific CD8⁺ T cells were assessed by flow cytometry using the HSV-1 glycoprotein B tetramer. We found that after treatment with each of the above three oHSVs, anti-HSV-1-specific CD8⁺ T cells were increased, while there was no difference among these three groups (Supplementary Fig. S12A). Furthermore, the levels of anti-HSV-1 specific antibodies were similar among the three groups (Supplementary Fig. S12B). Therefore, our data indicate that OV-αmCD47-G2b has no effect on enhancing antiviral effects and clearance when compared with OV-Q1. We measured the αmCD47-G2b antibody concentration at the tumor sites in the abdomen and the peripheral blood of the mice treated with 1 × 10⁶ PFU OV-αmCD47-G2b by ELISA. Results showed that αmCD47-G2b antibody concentrations were significantly much higher at the tumor sites than in sera at each of the indicated time points (Supplementary Fig. S13).

Collectively, OV-αmCD47-G2b showed the strongest antitumor effect compared with OV-αmCD47-G3 or OV-Q1 in an immunocompetent ovarian cancer metastasis mouse model, correlating to promoting macrophages polarizing towards M1 type and NK-cell activation but lacking substantial adaptive immune responses.

We compared the ability of αmCD47-G2b and OV-αmCD47-G2b on prolonging survival of mice bearing ID8 tumors. For this purpose, wild-type C57BL/6 mice were intraperitoneally injected with 2 × 10⁶ ID8 cells. Four days later, mice were intraperitoneally injected with 1 × 10⁶ PFU of OV-αmCD47-G2b or 150 μg of purified αmCD47-G2b. Saline was injected as control. The amount of antibody was calculated on the basis of the data from Fig. 1C to make certain the two groups had a similar amount of antibodies. Both αmCD47-G2b and OV-αmCD47-G2b treatments prolonged the survival of mice compared with saline treatment. However, prolongation of survival was significantly greater in tumor-bearing mice treated with OV-αmCD47-G2b when compared with tumor-bearing mice treated with αmCD47-G2b (Fig. 7A), which strengthens our hypothesis that OV-αmCD47-G2b treatment leads to: (i) direct tumor lysis by oHSVs; and (ii) the critical role of the innate immune cell infiltration and activation within the tumor microenvironment by oHSVs.

To enhance the adaptive immune responses in our system, we performed combination therapy with anti-PD-L1 antibody in the immunocompetent mouse model in vivo. For this purpose, we analyzed PD-L1 expression of ID8 ovarian tumor cells after OV-Q1 or OV-αmCD47-G2b infection in vitro. Our data showed that PD-L1 expression was increased after oHSV infection (Fig. 7B). Therefore, in an in vivo assay, anti-mPD-L1 mAb or control vehicle was injected 24 hours after OV-αmCD47-G2b treatment. Our data showed that compared with single reagent treatments, the OV-αmCD47-G2b and anti-PD-L1 mAb combination therapy significantly prolonged survival in mice bearing with ID8 ovarian tumor (Fig. 7C).

In this study, we combined an mAb and oncolytic virus into a single agent to develop an innovative antibody delivery system to improve the efficacy of ovarian cancer therapy. OV-αCD47-G1 not only directly lyses tumor cells but also activates NK-cell ADCC and macrophage ADCP function in vitro. Furthermore, OV-αCD47-G1 also blocked the “don't eat me” signal normally mediated by the interaction between SIRPα on macrophages and CD47 on ovarian tumor cells, respectively. Therefore, OV-αCD47-G1 treatment significantly improved the virotherapy efficacy against ovarian tumors in vivo.

In our study, the anti-CD47 antibody and herpes simplex oncolytic virus were selected to construct the OV-αCD47-G1 and OV-αCD47-G4, which are designed to secrete αCD47-G1 and αCD47-G4, respectively, to the ovarian cancer tumor microenvironment after OVs administration. CD47 is highly expressed on surfaces of numerous types of tumor cells, including ovarian cancer, which binds to its receptor, SIRPα on the surface of macrophages. This binding results in a “don't eat me” signal to macrophages and protects tumor cells from being killed by the former (5). Consistent with this concept, clinically used anti-CD47 antibodies have been developed to enhance the phagocytosis of cancer cells and are actively being tested in numerous clinical and preclinical studies (5, 22, 27, 28). Furthermore, some oncolytic viruses have been engineered by fusing the SIRPα ectodomain to human IgG1 or IgG4 Fc to block the CD47- SIRPα axis (29, 30).

Macrophages play key roles in tumor progression. Because of the different origins and local tumor microenvironment, tumor-associated macrophages (TAM) act in diverse roles. Therefore, macrophages are potential therapeutic targets for antitumor therapy. Indeed, talimogene laherparepvec (T-VEC), the first FDA-approved oHSV for virotherapy, expresses GM-CSF colony-stimulating factor to increase tumor-antigen presentation by dendritic cells and promote macrophage accumulation (31). By secreting several cytokines or growth factors, TAMs could induce angiogenesis and promote tumor cell invasion (32). The chemokines producing by TAMs can enhance different immune cell infiltration into the tumor microenvironment (33). Because of the heterogeneity and plasticity, macrophages are identified into M1 and M2 types macrophages, which play extremely different roles in regulating inflammation and tumor development (34–36). M1 type macrophage could mediate resistance against tumors by secreting IL6, iNOS, whereas M2 type macrophage work as a tumor promoter by secreting IL10, Arg 1 as well as TGFβ and VEGF that induces angiogenesis (37). It has been reported that anti-CD47 treatment increases M1-polarizied macrophages in glioblastoma treatment (24). Our study showed that OV-αCD47, especially OV-αCD47-G1 and the murine counterpart OV-αmCD47-G2b, can decrease the percentages of M2 type macrophage and increase the M1 type macrophage percentages, which restrain tumor progression.

Many studies focused on the role of T cells and NK cells on tumor clearance. However, macrophages play an important role in recognizing and depleting damaged and aged cells (38). Phagocytosis by macrophages initiates one type of innate immune responses. There are a number of phagocytic receptors, such as mannose receptors, complement receptors, recognizing the pathogens (39). Fc receptors binding with specific antibodies can also stimulate a strong phagocytosis effect (40). Furthermore, studies found that SIRPα is an important negative receptor on macrophage phagocytosis by binding to CD47 on tumor cells (5). Thus, many studies work on inhibiting the SIRPα–CD47 signal pathway to control tumor progression (5, 12, 41, 42). In our study, we combined the SIRPα–CD47 signal pathway with Fc receptor function to activate macrophage phagocytosis locoregionally. Our results demonstrated that OV-αCD47-G1 and OV-αmCD47-G2b, both of which can block the SIRPα–CD47 signaling pathway and stimulate Fc receptors, are more effective in inhibiting tumor progression than OV-αCD47-G4 and OV-αmCD47-G3, respectively, both of which can only affect the SIRPα–CD47 signaling pathway. Like PD-1 on T cells, SIRPα is a checkpoint on macrophages. Although our data show that OVs expressing a full-length CD47 antibody does not substantially modulate adaptive immune responses, combination of the virus expressing such an antibody with an Fc function and a PD-L1 antibody to block PD-1 checkpoint on T cells shows an improved outcome compared with the corresponding monotherapy. Therefore, the oHSV expressing a full-length anti-CD47 antibody with an Fc function should likely be considered for translation into the clinic for the treatment of ovarian cancer as a single agent or combined with another effective therapy. For this purpose, we have manufactured OV-αCD47-G1 in our institutional good manufacturing practices (GMP) facility.

In the last 20 years, mAb-based drugs have been proven to be effective and widely used for treating hematologic malignancies and solid tumors. However, some limitations exist, which restrain its applications (43, 44). One limitation is the poor intratumoral distribution after administration. A mAb drug administrated through a traditional mAb delivery route, such as intravenous, subcutaneous, or intramuscular will mainly stay in the peripheral circulatory system with low levels of accumulation in the tumor microenvironment. In addition, peripheral circulation of the antibodies results in systemic off-target side effects. If an mAb has long half-life, it would worsen toxicities. Therefore, anti-tumor mAbs were developed on an IgG4 scaffold to minimize the Fc-dependent effect to limit the potential toxicity, but this scaffold diminishes their anti-tumor effects due to limited Fc receptor-mediated anti-tumor immune responses (45, 46). Similarly, several anti-CD47 antibodies are constructed on the IgG4 scaffold to minimize the Fc receptor-dependent effect and to avoid the side effect from off-targeting (5). Using our oncolytic virus-based antibody delivery system, the anti-CD47 antibody can be directly and accurately delivered to the tumor microenvironment. This should enhance the availability of mAbs and limit the potential toxicity, thus providing an opportunity to use a stronger scaffold such as IgG1 to harness and maximize Fc receptor-mediated antitumor effects.

In summary, we have developed an effective oHSV-based approach via engineering a transgene expressing a full-length anti-CD47 Ab on an IgG1 scaffold to treat ovarian cancer. The approach has multifaceted functions, combining direct tumor lysis, innate immune infiltration and activation, immune checkpoint inhibition of macrophages, and Fc-dependent innate immune cell cytotoxic functions. Our approach significantly enhances the overall efficacy of both oncolytic virotherapy and mAb therapy for the treatment of ovarian cancer in preclinical mouse models.

M. Feng reports a patent for stimulating TLR/BTK signaling to promote CRT in macrophages licensed to Forty Seven, Inc. B. Kaur reports grants from NIH during the conduct of the study; nonfinancial support from Mesoblast outside the submitted work. M.A. Caligiuri reports grants from NIH during the conduct of the study; other support from CytoImmune Therapeutics outside the submitted work; also has a patent for OV-anti-CD47 pending. J. Yu reports grants from NIH during the conduct of the study; other support from CytoImmune Therapeutics outside the submitted work; also has a patent for OV-CD47 pending. No disclosures were reported by the other authors.

L. Tian: Supervision, writing–original draft, project administration, writing–review and editing. B. Xu: Supervision. K.-Y. Teng: Investigation. M. Song: Investigation. Z. Zhu: Investigation. Y. Chen: Validation. J. Wang: Validation. J. Zhang: Software. M. Feng: Supervision. B. Kaur: Supervision. L. Rodriguez: Supervision. M.A. Caligiuri: Supervision, funding acquisition, project administration. J. Yu: Supervision, funding acquisition, project administration, writing–review and editing.

This work was supported by grants from the NIH (NS106170, AI129582, CA247550, CA163205, CA223400, CA265095, CA210087, CA20175, and CA255250), the Leukemia & Lymphoma Society (1364–19), California Institute for Regenerative Medicine (DISC2COVID19–11947), the Breast Cancer Alliance, Markel Friedman Accelerator Fund, and the V Foundation for Cancer Research V Scholar Award. The authors thank the NIH Tetramer Core Facility for providing the herpes simplex virus type 1 glycoprotein B tetramer.

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.