Specific targeting of glioblastoma with an oncolytic virus expressing a cetuximab-CCL5 fusion protein via innate and adaptive immunity

Abstract

Chemokines such as C-C motif ligand 5 (CCL5) regulate immune cell trafficking in the tumor microenvironment (TME) and govern tumor development, making them promising targets for cancer therapy. However, short half-lives and toxic off-target effects limit their application. Oncolytic viruses (OVs) have become attractive therapeutic agents. Here, we generate an oncolytic herpes simplex virus type 1 (oHSV) expressing a secretable single-chain variable fragment of the epidermal growth factor receptor (EGFR) antibody cetuximab linked to CCL5 by an Fc knob-into-hole strategy that produces heterodimers (OV-Cmab-CCL5). OV-Cmab-CCL5 permits continuous production of CCL5 in the TME, as it is redirected to EGFR⁺ glioblastoma (GBM) tumor cells. OV-Cmab-CCL5 infection of GBM significantly enhances the migration and activation of natural killer cells, macrophages and T cells; inhibits tumor EGFR signaling; reduces tumor size; and prolongs survival of GBM-bearing mice. Collectively, our data demonstrate that OV-Cmab-CCL5 offers a promising approach to improve OV therapy for solid tumors.

Glioblastoma (GBM) is the most aggressive and fatal primary brain tumor, with its incidence increasing by 3% each year¹. This devastating lethal cancer has a median survival time of ~15 months, a statistic that has not changed substantially in the past two decades. One reason is that the infiltrative growth of GBM reduces the effectiveness of current therapies, such as surgery, chemotherapy and radiotherapy. Although immunotherapy is a promising approach, it is limited by systemic immunosuppression caused by the tumor^2,3.

‘Hot’ tumors, which are characterized by lymphocyte infiltration into the tumor microenvironment (TME), are increasingly recognized as better targets for immunotherapy than ‘cold’ tumors, the latter of which contain few lymphocytes^2,3. GBM tumors are cold because the numbers of intratumoral T lymphocytes, monocytes and dendritic cells are limited, and their anti-tumor functions are suppressed^4–6.

Not surprisingly, there is considerable interest in new approaches to making cold tumors hot. One current focus is taking advantage of chemokines, which attract innate and adaptive immune cells to tumor sites. The chemokine C-C motif ligand 5 (CCL5) is an inflammatory chemokine that promotes chemotaxis of immune cells by interacting with CCR1 and/or CCR5 (refs.^7–9). Tumor evolutionary pressure may lead to methylation and silencing of CCL5 expression in solid tumors¹⁰. Thus, recovering or enhancing CCL5 expression in the TME is a promising therapeutic strategy for turning cold solid tumors hot. However, difficulty with locoregional delivery of CCL5 into the TME and a short half-life have limited the cytokine’s usefulness as a cancer therapy.

One possible solution is to introduce oncolytic viruses (OVs) into immunotherapy. OVs are genetically engineered to selectively replicate in tumor cells, lyse them and enhance immune stimulation. In fact, OVs derived from genetically engineered herpes simplex virus type 1 (oHSV) have already been used in clinical trials¹¹. Most recently, Friedman and colleagues reported that G207 oHSV showed a safe and strong effect on treating pediatric high-grade glioma¹². The success of oHSV-derived therapeutics is thought to depend on both oncolytic destruction of tumor cells and activation of anti-tumor immune responses, the latter of which can potentially lead to long-term cancer remission. We previously found that oHSV treatment dramatically increased the ability of immune cells to infiltrate tumors and destroy tumor cells^13–15. We therefore reasoned that altering oHSV to enhance immune cell infiltration or to increase the specificity of immune infiltration toward tumor cells would allow more robust tumor eradication.

Bispecific antibodies or fusion proteins are engineered to connect two receptor-binding functional domains into a single molecule. This ‘two-target’ functionality can interfere with multiple surface receptors or ligands associated with tumors or other pathological processes¹⁶. Various bispecific antibodies or fusion proteins are in clinical development or already approved for cancer therapy^17,18. Therefore, we engineered an oHSV encoding a bispecific fusion protein containing an IgG1 form of cetuximab (an antibody of EGFR) and CCL5, and we named the virus OV-Cmab-CCL5. This fusion protein can bind or anchor to both wild-type EGFR and the mutant form, EGFRvIII, expressed on tumor cells. Thus, our OV combines chemokine function with antibody function together to improve GBM therapy. It not only targets the delivery of CCL5 to the TME but also functions as an antibody by binding to heterogeneous GBM cells to activate Fc-receptor-mediated NK cell antibody-dependent cellular cytotoxicity (ADCC) and macrophage antibody-dependent cellular phagocytosis (ADCP). In this study, we tested the OV-Cmab-CCL5 for its ability to activate anti-tumor immunity in vitro and in vivo, including an assessment of its ability to prolong survival in vivo.

Results

OV-Cmab-hCCL5 encodes a bispecific protein binding to EGFR

As CCL5 is often epigenetically silenced in tumor cells¹⁰, we found, as expected, that CCL5 is secreted only at low levels by three human GBM cell lines (Extended Data Fig. 1a). Furthermore, we detected high expression of the CCL5 receptors CCR1 and/or CCR5 on human natural killer (NK) cells, macrophages, CD4⁺ and CD8⁺ T cells (Extended Data Fig. 1b–e). To overcome the short half-life of CCL5 and allow high levels of CCL5 to build up in the TME, we generated a soluble bispecific fusion protein, encoded by two corresponding DNA sequences separated by a DNA sequence encoding a T2A self-cleaving peptide, using a knob-into-hole design¹⁹. The knob arm was designed to express cetuximab (a US Food and Drug Administration (FDA)-approved anti-EGFR antibody) to target human EGFR⁺ (hEGFR⁺) tumor cells. The hole arm was designed to contain hCCL5 or mouse CCL5 (mCCL5). To produce heterodimers (Cmab-CCL5) and prevent homodimers, S354C/T366W/K409A mutations of human IgG1 were included to generate the ‘knob’ on the cetuximab-expressing arm, and Y349C/T366S/L368A/Y407V/F405K mutations were included to generate the ‘hole’ on the CCL5-expressing arm (Cmab-hCCL5 or Cmab-mCCL5; Fig. 1a)¹⁹.

Fig. 1 |. Construction and characterization of Cmab-CCL5 and OV-Cmab-CCL5.

a, Schematic of the Cmab-h(m)CCL5 bispecific fusion protein. scFv, single-chain variable fragment. b,c, Detection of purified Cmab-hCCL5 bound to the wild-type EGFR (wtEGFR) U251T2 and EGFRvIII U87ΔEGFR GBM cell lines or the EGFR⁻ A2780 human ovarian cancer cell line, as measured by flow cytometry after staining Cmab-hCCL5-incubated tumor cells with anti-Fc-allophycocyanin (APC) (b) or anti-hCCL5-APC (c). Cmab-hCCL5 was purified from lentivirus-infected CHO cells. IgG1 isotype served as control. Experiments are representative of n = 3 independent experiments with similar results. d, Genetic maps of the OVs used in this study. Wild-type HSV-1 (wtHSV-1; top); control oHSV, OV-Q1, with deletion of two copies of γ34.5, dysfunction of ICP6 and insertion of the GFP gene (middle); and OV-Cmab-h(m)CCL5 showing the insertion of the gene encoding the ‘knob’ and ‘hole’ (bottom). The ‘knob’ and ‘hole’ are linked by T2A sequences and are driven by the viral pIE4/5 promoter. e, Immunoblots of hCCL5 and human Fc from concentrated supernatants of engineered CHO cells and OV-infected U251T2 GBM cells. Experiments were representative of n = 3 independent experiments with similar results. EV, empty vector. f, hCCL5 yields from OV-Cmab-hCCL5-infected GBM cells, as measured by enzyme-linked immunosorbent assay (ELISA). g, Cetuximab (Cmab) yields from OV-Cmab-hCCL5-infected GBM cells, as measured by ELISA. n = 3 independent experiments for panels f and g. Error bars indicate the standard deviation (s.d.) of triplicates (f,g). All data are presented as mean ± s.d.

To validate the construct, we expressed Cmab-hCCL5 in Chinese hamster ovary (CHO) cells, using lentiviral transduction, and evaluated its binding to tumor cells. Because EGFR expression is heterogeneous in the TME, we used U251T2 human GBM cells, which express wild-type EGFR (wtEGFR), U87ΔEGFR human GBM cells, which express EGFRvIII, and A2780 human ovarian tumor cells, which are EGFR⁻. Flow cytometry analysis using an anti-human Fc antibody showed that Cmab-hCCL5 selectively bound to tumor cells expressing wtEGFR or EGFRvIII, but not EGFR⁻ A2780 cells (Fig. 1b). Using an anti-CCL5 antibody, we detected CCL5 on EGFR⁺ tumor cell surfaces (Fig. 1c), suggesting that cetuximab had enabled CCL5 to be indirectly associated with EGFR. We also generated the reverse version of Cmab-CCL5 by switching the order of Cmab and CCL5 (hCCL5-Cmab) and expressed it from CHO cells. We found the two fusion proteins had a similar affinity to bind to U251T2 GBM cells (Extended Data Fig. 2a,b). Therefore, we used Cmab-hCCL5 in all subsequent experiments.

We then generated oHSV-expressing Cmab-hCCL5 using the parental OV-Q1 oHSV, which is doubly attenuated with an inactivated ribonucleotide reductase gene (ICP6) and deletions of both copies of the neurovirulence gene (ICP34.5). These changes limit oHSV’s replication to tumor cells and reduce its neurovirulence²⁰. To generate the OV-Cmab-CCL5 construct, we inserted a DNA sequence encoding Cmab-CCL5 with a knob-into-hole IgG1 Fc into the ICP6 locus of OV-Q1, driven by the promoter of the HSV-1 immediate early gene IE4/5 (Fig. 1d). The supernatants from OV-Cmab-hCCL5-infected U251T2 GBM cells and control (OV-Q1)-infected U251T2 GBM cells were collected 48 h postinfection (hpi) for immunoblotting, along with supernatants from CHO cells with lentiviral expression of Cmab-hCCL5 or empty vector as positive and negative controls, respectively. Both hCCL5 and human IgG heavy chain were detected in the supernatants of the U251T2 GBM cells infected with OV-Cmab-hCCL5 (Fig. 1e). Cmab-hCCL5 produced by OV-Cmab-hCCL5-infected U251T2 GBM cells was confirmed to bind to surface-expressed EGFR using anti-Fc and anti-CCL5 antibodies (Extended Data Fig. 2c,d). The concentrations of hCCL5 and Cmab from OV-Cmab-hCCL5-infected U251T2 GBM cells were quantified by ELISA using CHO cell-derived fusion protein with known concentrations as standards. OV-Cmab-hCCL5-infected U251T2 GBM cells secreted a substantial level of the fusion protein at 12 hpi. The maximum concentration of OV-Cmab-hCCL5 (at 72 hpi) is between 31 to 36 nM according to quantification of hCCL5 and Cmab in the fusion protein (Fig. 1f,g). To determine whether different multiplicity of infections (MOIs) would affect the production of Cmab-hCCL5, we used MOIs of 0.5, 2 and 5 OV-Cmab-hCCL5 to infect U251T2 cells for 72 h. We observed lower MOIs produced more Cmab-hCCL5 than higher MOIs (Extended Data Fig. 2e). This result is likely because tumor cells infected with higher MOI OV-Cmab-hCCL5 are lysed more quickly (Extended Data Fig. 2f), giving them less time to secrete Cmab-hCCL5. Collectively, the data presented thus far demonstrate the successful generation of an OV-Cmab-hCCL5 construct that expresses a Cmab-hCCL5 bispecific fusion protein that can bind to EGFR on tumor cells.

To determine whether adding Cmab-hCCL5 to OV-Q1 would affect the virus’s oncolytic capability, we infected U251T2 GBM cells with OV-Cmab-hCCL5 or OV-Q1 at a MOI from 0.01 to 8. Real-time cell analysis (RTCA) indicated that the two viruses had a similar oncolysis ability (Extended Data Fig. 2g). We also evaluated viral production of OV-Cmab-hCCL5. Results showed that OV-Cmab-hCCL5-infected U251T2 GBM cells produced similar amounts of virus when compared to the same GBM cells infected with OV-Q1 (Extended Data Fig. 2h).

Cmab-hCCL5 promotes migration of human immune cells in vitro

Next, we determined whether purified Cmab-hCCL5 promotes the migration of immune cells in vitro, using a transwell assay with recombinant human CCL5 as positive control. Cmab-hCCL5 purified from supernatants of lentivirus-transduced CHO cells induced a dose-dependent increase in the migration of human NK cells, macrophages, CD4⁺ T cells and CD8⁺ T cells when compared to an IgG1 isotype control (Fig. 2a–d). We also measured the ability of Cmab-hCCL5 produced by OV-Cmab-hCCL5-infected U251T2 GBM cells to regulate immune cell migration, also with a transwell assay. At 48 hpi, supernatants from OV-Cmab-hCCL5-infected U251T2 GBM cells significantly increased the migration of human NK cells, macrophages, CD4⁺ T cells and CD8⁺ T cells when compared to supernatants from OV-Q1-infected U251T2 GBM cells (Fig. 2e–h). We repeated the migration assay by placing U251T2 cells infected with OV-Cmab-hCCL5 into the lower chamber of a transwell. The immune cell migrations were measured by flow cytometry after staining with anti-human CD45 antibody, a pan-leukocyte marker. Similar migration results were obtained for human NK cells, macrophages, CD4⁺ T cells, and CD8⁺ T cells (Extended Data Fig. 3). Thus, we successfully generated Cmab-hCCL5 and its corresponding oHSV that promoted the migration of human NK cells, macrophages and T cells in vitro independent of cell contact.

Fig. 2 |. Cmab-hCCL5 promotes migration of human immune cells in vitro.

a–d, Migration of human NK cells (a), macrophages (b), CD4⁺ T cells (c) and CD8⁺ T cells (d) induced by purified Cmab-hCCL5, as measured with a transwell assay. The indicated amount of recombinant hCCL5 (rhCCL5) in 600 μl media was positive control. e–h, Migration of human NK cells (e), macrophages (f), CD4⁺ T cells (g) and CD8⁺ T cells (h) induced by the supernatant from OV-Cmab-hCCL5- or OV-Q1-infected U251T2 GBM cells, as measured by a transwell assay. A total of 200 ng rhCCL5 in 600 μl media was used as positive control. The results were repeated with three independent donors (b, e and f) or four independent donors (a, c, d, g and h). Error bars indicate s.d., and all data are presented as mean ± s.d. (a–h). Statistical analyses were performed by two-sided Student’s t-test (a–d) or one-way analysis of variance (ANOVA) with P values corrected for multiple comparisons by the Bonferroni method (e–h).

Cmab-hCCL5 promotes NK cell ADCC against EGFR⁺ tumor cells

NK cells have an anti-tumor function, because they possess natural cytotoxicity and can also carry out ADCC in the presence of certain antibodies. To investigate the role of Cmab-hCCL5 in regulating NK cell ADCC, we conducted a standard ⁵¹Cr release assay by co-culturing human NK cells with ⁵¹Cr-labeled U251T2 (wtEGFR), GBM30 (EGFRvIII) or Gli36ΔEGFR (EGFRvIII) GBM cells. The cells were precultured for 30 min in the presence or absence of Cmab-hCCL5 purified from lentivirus-transduced CHO cells. Cmab-hCCL5 significantly induced NK cell ADCC targeting both wtEGFR and EGFRvIII tumor cells when compared to an IgG1 isotype control (Fig. 3a). Similarly, compared to unconcentrated supernatants from OV-Q1-infected cells, those from OV-Cmab-hCCL5-infected U251T2 GBM cells induced significantly more cytolysis targeting wtEGFR and EGFRvIII GBM cells, but not EGFR⁻ cells (Fig. 3b). To validate these data, we co-cultured GBM30 cells with human NK cells treated with Cmab-hCCL5 purified from lentivirus-infected CHO cells for 6 h. Flow cytometry showed that Cmab-hCCL5 significantly increased expression of the two activation markers, CD69 and granzyme B, on NK cells compared to an IgG1 isotype control (Fig. 3c,e). Also, the fractions of CD69⁺ NK cells and granzyme B⁺ NK cells were higher after culturing with unconcentrated supernatants collected from the OV-Cmab-hCCL5-infected U251T2 GBM cells compared to culture with unconcentrated supernatants collected from uninfected or OV-Q1-infected U251T2 GBM cells (Fig. 3d,f).

Fig. 3 |. Cmab-hCCL5 increases human NK cell-mediated ADCC against EGFR⁺ tumor cells.

a, ADCC of human primary NK cells against Cmab-hCCL5-treated U251T2, GBM30 and Gli36ΔEGFR GBM cells at effector/target (E/T) ratios of 40:1, 20:1 and 10:1, measured by ⁵¹Cr release assay. For U251T2, control versus Cmab-hCCL5 5 μg ml⁻¹ P < 0.0001; control versus Cmab-hCCL5 10 μg ml⁻¹ P < 0.0001. For GBM30, control versus Cmab-hCCL5 5 μg ml⁻¹ P = 0.0005; control versus Cmab-hCCL5 10 μg ml⁻¹ P < 0.0001. For Gli36ΔEGFR, control versus Cmab-hCCL5 5 μg ml⁻¹ P = 0.0326; control versus Cmab-hCCL5 10 μg ml⁻¹ P < 0.0001. A linear mixed model was used to account for the underlying variance and covariance structure (n = 3 donors). b, ADCC of human primary NK cells against U251T2, GBM30 and Gli36ΔEGFR GBM cells and A2780 ovarian tumor cells treated with conditioned media from OV-Q1- or OV-Cmab-hCCL5-infected U251T2, as measured by ⁵¹Cr release. For U251T2, OV-Q1 versus OV-Cmab-hCCL5 P < 0.0001. For GBM30, OV-Q1 versus OV-Cmab-hCCL5 P < 0.0001. For Gli36ΔEGFR, OV-Q1 versus OV-Cmab-hCCL5 P < 0.0001. For A2780, there were no significant differences among uninfected, OV-Q1 and OV-Cmab-hCCL5 groups. A linear mixed model was used to account for the underlying variance and covariance structure (n = 4 donors in the GBM30 group and 3 donors in all other groups). c, CD69 expression, as measured by flow cytometry, on human primary NK cells co-cultured with GBM30 cells in the presence of Cmab-hCCL5 or IgG1 isotype control. MFI, mean fluorescence intensity. d, CD69 expression, measured by flow cytometry, on human primary NK cells co-cultured with GBM30 cells in the presence of conditioned media from OV-Q1- or OV-Cmab-hCCL5-infected U251T2. e, Granzyme B expression on human primary NK cells co-cultured with GBM30 cells in the presence of purified Cmab-hCCL5 or IgG1 isotype control, as measured by flow cytometry. f, Granzyme B expression on human primary NK cells co-cultured with GBM30 cells in the presence of conditioned media from OV-Q1- or OV-Cmab-hCCL5-infected cells, as measured by flow cytometry. Error bars indicate s.d., and all data are presented as mean ± s.d. (a–f). Experiments in panels c–f are representative of three independent experiments, and statistical analyses were performed by one-way ANOVA with P values corrected for multiple comparisons by the Bonferroni method (n = 3 donors).

Cmab-hCCL5 enhances human macrophage ADCP against GBM cells

Macrophages play important roles in the TME by mediating ADCP and releasing cytokines. We therefore determined the effect of Cmab-hCCL5 on macrophages. GBM30 cells were co-cultured with human M-CSF-treated primary monocyte-derived macrophages for 4 h with or without Cmab-hCCL5 purified from supernatants of lentivirus-infected CHO cells. Flow cytometry revealed that Cmab-hCCL5 induced a significantly higher level of ADCP against GBM30 cells compared to IgG1 isotype or vehicle control (Fig. 4a). The ADCP assay was repeated with the unconcentrated supernatants from OV-Cmab-hCCL5- or OV-Q1-infected U251T2 GBM cells. The supernatants from OV-Cmab-hCCL5-infected U251T2 GBM cells also dramatically induced ADCP compared to supernatants from OV-Q1-infected or uninfected U251T2 GBM cells (Fig. 4b). Of note, compared to IgG1 isotype control, Cmab-hCCL5 treatment also significantly increased expression of inflammatory genes IL-1B, IL-6, IL-12 and NOS2 in human macrophages, as measured by real-time RT-PCR (Fig. 4c).

Fig. 4 |. Cmab-hCCL5 enhances human macrophage ADCP against GBM cells.

a, ADCP of human macrophages induced by Cmab-hCCL5 targeting GBM30 cells. Carboxyfluorescein succinimidyl ester (CFSE) prelabeled GBM30 cells were co-cultured with human macrophages in the presence of Cmab-hCCL5 or vehicle control. The percentage of human macrophages that phagocytosed labeled tumor cells was measured by flow cytometry (n = 3 independent donors). b, ADCP of human macrophages induced by conditioned media from OV-Cmab-hCCL5-infected U251T2 GBM cells, targeting GBM30 cells. CFSE prelabeled GBM30 cells were co-cultured with human macrophages in the presence of supernatants from OV-Q1- or OV-Cmab-hCCL5-infected U251T2 GBM cells. The percentage of human macrophages that phagocytosed labeled tumor cells was measured by flow cytometry (n = 3 independent donors). c, Cytokine mRNA expression levels, measured by real-time RT-PCR, from human macrophages co-cultured with GBM30 cells at a ratio of 1:1 with or without Cmab-hCCL5 for 6 h (n = 4 independent donors). Error bars indicate s.d., and all data are presented as mean ± s.d. (a–c). Statistical analyses were performed by one-way ANOVA with P values corrected for multiple comparisons by the Bonferroni method.

OV-Cmab-hCCL5 improves OV therapy in xenograft GBM mouse model

To evaluate the efficacy of OV-Cmab-hCCL5 for in vivo treatment of GBM, we used a previously described orthotopic model of human GBM created by intracranial (i.c.) injection of 1 × 10⁵ firefly luciferase (FFL) gene-expressing human GBM30-FFL cells into NOD-scid IL2Rγ^null (NSG) mice²¹. To mimic the human immune system, 4 days after tumor implantation, we intravenously (i.v.) injected each mouse with peripheral blood mononuclear cells (PBMCs) isolated from a single donor. On day 6, we delivered i.v. activated T cells derived from the same donor to each mouse. Seven days after tumor implantation, each mouse received i.c. OV-Cmab-hCCL5 or OV-Q1 or saline as a placebo control. Tumor progression was monitored by luciferase-based imaging (Fig. 5a). This one-cycle treatment with OV-Cmab-hCCL5 was significantly more effective than OV-Q1 or vehicle control at inhibiting the progression of GBM tumors in vivo (Fig. 5b,c). OV-Q1 moderately but significantly slowed GBM progression compared to vehicle, whereas OV-Cmab-hCCL5 significantly prolonged median survival compared to both saline control and OV-Q1 treatment (Fig. 5d). We then performed a similar experiment, but with two instead of one treatment cycles as indicated (Fig. 5e), and evaluated the therapeutic efficacy. This two-cycle treatment with OV-Cmab-hCCL5 also significantly improved the survival of GBM mice compared to the two-cycle OV-Q1 or saline treatment (Fig. 5f). Of note, the two-cycle OV-Cmab-hCCL5 treatment provided significantly more protection to GBM-bearing mice compared to the one-cycle treatment, as the median survival day was prolonged from 39 days to 47 days (Fig. 5f versus Figure 5d). We repeated the experiments with the GBM30-FFL model using nude mice but without infusing PBMCs and activated T cells. We observed that in nude mice, having macrophages and NK cells but lacking both endogenous and infused T cells, OV-Cmab-hCCL5 treatment was also significantly more effective than OV-Q1 or vehicle control at inhibiting the progression of GBM tumors (Extended Data Fig. 4). Collectively, the data demonstrate that OV-Cmab-hCCL5 improves the efficacy of oncolytic virotherapy in vivo in a mouse GBM xenograft model.

Fig. 5 |. OV-Cmab-hCCL5 improves oncolytic virotherapy in a xenograft GBM animal model.

a, Experimental timeline for in vivo studies using a one-cycle treatment. b, Luciferase imaging of GBM30-FFL GBM mice with the indicated treatments on day 15 and day 21. c, Luciferase intensity summary data of the experimental mice shown in panel b (day 21). Error bars indicate s.d., and data are presented as mean ± s.d. Statistical analyses were performed by one-way ANOVA with P values corrected for multiple comparisons by the Bonferroni method (n = 7 independent mice). d, Survival of GBM30 tumor-bearing mice treated with OV-Q1, OV-Cmab-hCCL5 or vehicle control as indicated in panel a. Survival was estimated by the Kaplan–Meier method and compared by log-rank test (n = 7 independent mice). e, Experimental timeline for in vivo studies using a two-cycle treatment. f, Survival of GBM30 tumor-bearing mice treated with OV-Q1, OV-Cmab-hCCL5 or vehicle control, as indicated in panel e. Survival was estimated by the Kaplan–Meier method and compared by log-rank test (n = 6 independent mice).

OV-Cmab-mCCL5 enhances survival in immunocompetent GBM model

To test our oncolytic virotherapy in a fully immunocompetent mouse model of GBM, we replaced human CCL5 in OV-Cmab-hCCL5 with murine CCL5 (mCCL5) to generate OV-Cmab-mCCL5. We also expressed human EGFR in the murine GBM CT2A cell line (CT2A-hEGFR) so Cmab could bind to murine GBM cells. Indeed, Cmab-mCCL5 purified from lentivirus-infected CHO cells did bind to CT2A-hEGFR cells (Extended Data Fig. 5a,b). We also confirmed that Cmab-mCCL5 in supernatants derived from OV-Cmab-mCCL5-infected CT2A-hEGFR cells was detectable by immunoblotting (Extended Data Fig. 5c). The OV-Cmab-mCCL5-infected CT2A-hEGFR cells increased Cmab-mCCL5 release in a time-dependent manner between 12 to 72 hpi and the level at 72 hpi is over 20 nM according to ELISA for detecting mCCL5 and Cmab in the fusion protein (Extended Data Fig. 5d,e). Like our results for Cmab-hCCL5, both Cmab-mCCL5 purified from lentivirus-transduced CHO cells and produced by OV-Cmab-mCCL5-infected CT2A-hEGFR cells boosted the migration of murine NK cells, macrophages, and T cells (Extended Data Figs. 5f and Fig. 6a). The purified Cmab-mCCL5 as well as the secreted Cmab-mCCL5 from OV-Cmab-mCCL5-infected CT2A-hEGFR cells both significantly enhanced the ADCC function of NK cells targeting CT2A-hEGFR cells when compared to the corresponding controls (Extended Data Fig. 6a,b). These data are consistent with the observed increase in CD69 expression on murine NK cells (Extended Data Fig. 6b,c).

Fig. 6 |. Cmab-mCCL5 promotes murine immune cell migration and activation; OV-Cmab-mCCL5 prolongs survival of GBM-bearing mice in an immunocompetent mouse model.

a, Migration of mouse primary NK cells, macrophages, CD4⁺ T and CD8⁺ T cells induced by supernatant of OV-Cmab-mCCL5- or OV-Q1-infected U251T2 GBM cells, measured by transwell assay; recombinant mCCL5 (rmCCL5), positive control (n = 3 independent mice). b, Cytotoxicity of mouse primary NK cells against CT2A-hEGFR cells pre-treated with supernatant of uninfected, OV-Q1- or OV-Cmab-mCCL5-infected CT2A-hEGFR cells, as measured by ⁵¹Cr release assay. OV-Q1 versus OV-Cmab-mCCL5, P = 0.0041. A linear mixed model was used to account for the underlying variance and covariance structure (n = 3 independent mice). c, CD69 expression, measured by flow cytometry, on mouse primary NK cells co-cultured with CT2A-hEGFR cells in the presence of supernatant of OV-Q1- or OV-Cmab-mCCL5-infected CT2A-hEGFR cells. d, ADCP of mouse macrophages, induced by supernatant of OV-Cmab-mCCL5-infected cells, against CT2A-hEGFR cells, as measured by flow cytometry. Error bars indicate s.d., and data are presented as mean ± s.d. Experiments in panels c and d (left) were representative of three independent experiments. One-way ANOVA for P value calculations, corrected for multiple comparisons by the Bonferroni method (n = 3 independent mice (a,c,d)). e, Survival of CT2A-hEGFR-bearing mice treated with OV-Q1, OV-Cmab-mCCL5 or vehicle control with a 3-, 5- and 7-day interval between tumor implantation and treatment. P = 0.0005 for OV-Cmab-mCCL5 versus OV-Q1 (both 3-day interval). P = 0.0008 for OV-Cmab-mCCL5 versus OV-Q1 (both 5-day interval) and P = 0.0045 for OV-Cmab-mCCL5 versus OV-Q1 (both 7-day interval). P = 0.0004 for OV-Cmab-mCCL5 (3-day interval) versus OV-Cmab-mCCL5 (7-day interval). P = 0.0009 for OV-Cmab-mCCL5 (5-day interval) versus OV-Cmab-mCCL5 (7-day interval). P = 0.0295 for OV-Cmab-mCCL5 (5-day interval) versus OV-Cmab-mCCL5 (3-day interval) (n = 7 independent mice in the OV-Cmab-mCCL5 (7-day interval) group and n = 6 mice in all other groups). f, Survival of CT2A-hEGFR-bearing mice treated with OV-Q1 plus Cmab-mCCL5 delivered by osmotic pumps (OPs), OV-Q1 plus mCCL5 delivered by OP, OV-Cmab-mCCL5 or vehicle control. P = 0.0271 for OV-Cmab-mCCL5 versus OV-Q1 plus Cmab-mCCL5 delivered by OP; P = 0.0006 for OV-Cmab-mCCL5 versus OV-Q1 plus mCCL5 delivered by OP. P = 0.0382 for OV-Q1 plus mCCL5 delivered by OP versus OV-Q1 plus Cmab-mCCL5 delivered by OP. Survival data in panels e and f were estimated by the Kaplan–Meier method and compared by log-rank test (n = 6 independent mice in each group).

Cmab-mCCL5 purified from lentivirus-transduced CHO cells or produced by OV-Cmab-mCCL5-infected CT2A-hEGFR cells also significantly increased ADCP of murine macrophages targeting CT2A-hEGFR cells, compared to the corresponding controls (Extended Data Fig. 6c,d). Furthermore, similar to the data in humans (Fig. 4c), purified Cmab-mCCL5 significantly upregulated expression of IL-1b, IL-6, IL-12b, Ccl-2, Ccl-4 and Nos2 in murine macrophages co-cultured with Cmab-mCCL5-treated CT2A-hEGFR cells (Extended Data Fig. 6d).

The murine system that we established consisting of OV-Cmab-mCCL5 and CT2A-hEGFR GBM cells allowed us to evaluate the efficacy of OV-Cmab-mCCL5 in an immunocompetent model in which CT2A-hEGFR cells were injected i.c. into wild-type C57BL/6J mice. Treatment with OV-Cmab-mCCL5 3 days after tumor implantation (or 3-day treatment interval between tumor implantation and OV treatment) significantly prolonged survival compared to OV-Q1 or vehicle control (Extended Data Fig. 7a). The survival data were consistent with tumor size, as determined by microscopy after H&E staining (Extended Data Fig. 7b). Next, we used the same model to compare 3-day, 5-day and 7-day intervals between tumor implantation and treatment initiation. The results showed that each of the three intervals to initiate therapy prolonged the survival of GBM-bearing mice, but the shortest (3-day) interval had significantly better effects than the longer treatment intervals (Fig. 6e).

The survival study was repeated with the CT2A-hEGFR immunocompetent GBM mouse model to compare the effect of OV-Cmab-mCCL5 versus the combination of i.c. treatment of OV-Q1 with continuous release of Cmab-mCCL5 protein versus i.c. treatment of OV-Q1 combined with continuous release of mCCL5 protein. Continuous release of Cmab-mCCL5 or mCCL5 protein was achieved using an osmotic pump. For this purpose, CT2A-hEGFR GBM cells were implanted on day 0. Three days later (day 3), four different treatment groups were set up with i.c. administration: group 1, saline; group 2, a combination of OV-Q1 plus osmotic pump delivery of Cmab-mCCL5 protein; group 3, a combination of OV-Q1 plus osmotic pump delivery of mCCL5 protein; and group 4, OV-Cmab-mCCL5. On days 4 to 7, in groups 2 and 3, Cmab-mCCL5 or mCCL5 was i.c. delivered by the osmotic pump at the rate of 24 nmol per day for 3 days. The delivered amount of each protein by the osmotic pump is similar to the amount produced by the injected virus, per calculation with the data from Extended Data Fig. 5d and the treatment dose of virus (Extended Data Fig. 7c). Among the four different treatment groups, the survival following the administration of the OV-Cmab-mCCL5 was statistically better than the combination of i.c. OV-Q1 and Cmab-mCCL5 delivered by the osmotic pump, as well as the other two groups (Fig. 6f).

OV-Cmab-mCCL5 induces immune cell infiltration and activation

To determine whether OV-Cmab-mCCL5 enhances the infiltration and activation of innate and adaptive immune cells into the TME in vivo, we used our CT2A-hEGFR immunocompetent GBM mouse model. First, we measured the amount of Cmab-mCCL5 in the homogenized brains of mice implanted with CT2A-hEGFR tumor cells on day 0 and treated with OV-Cmab-mCCL5, OV-Q1 or saline on day 5 for 3, 5 or 7 days. The ELISA data showed that Cmab-mCCL5 production was highest when mice were treated with OV-Cmab-mCCL5 for 3 days and decreased but was measurable for 5 and 7 days (Fig. 7a). We also tested whether in vivo Cmab-mCCL5 production correlated with the initial viral dose. Infusing the animals with 1 × 10⁵, 2 × 10⁵ or 4 × 10⁵ plaque-forming units (PFU) OV-Cmab-mCCL5 affected Cmab-mCCL5 production in a dose-dependent manner, suggesting that those doses may not achieve complete infection of all tumor cells in the TME (Extended Data Fig. 8a). We also analyzed mCCL5 expression and the existence of oHSV in the brain by immunohistochemical staining of tumor-implanted mice treated with OV-Cmab-mCCL5, OV-Q1 or saline for 3 days. Compared to saline and OV-Q1 groups, the OV-Cmab-mCCL5 group showed significantly higher mCCL5 production as expected (Extended Data Fig. 8b,c). Next, we determined whether the produced Cmab-mCCL5 could block EGFR signaling in tumor cells. Brains were harvested from mice bearing CT2A-hEGFR cells that were treated for 3 days, we found that the in vivo treatment with OV-Cmab-CCL5 reduced levels of phospho(p)-EGFR and downstream p-AKT compared to OV-Q1 and saline treatments (Extended Data Fig. 9a).

Fig. 7 |. OV-Cmab-mCCL5 induces infiltration of innate and adaptive murine immune cells into the GBM TME.

a, The amount of Cmab-mCCL5 was measured in the brain on days 3, 5 and 7 after OV treatment. C57BL/6J mice were i.c. injected with 1 × 10⁵ CT2A-hEGFR cells. Five days after tumor implantation, mice were received an intratumoral injection of 2 × 10⁵ PFU OVs or vehicle as control. On days 3, 5 and 7 after OV treatments, brains were harvested and homogenized with 500 μl saline for ELISA. b,c, Total numbers (b) and percentages among lymphocytes (c) of NK cells, macrophages and T cells, as measured by flow cytometry. d, Total number of CD4⁺ T cells and CD8⁺ T cells and the ratio of CD4⁺ T cells to CD8⁺ T cells, as measured by flow cytometry. Error bars indicate s.d., and data are presented as mean ± s.d. (a–d). Statistical analyses were performed by one-way ANOVA with P values corrected for multiple comparisons by the Bonferroni method (n = 4 independent mice in the OV-Q1 group or 5 mice in saline and OV-Cmab-mCCL5 groups).

Next, we treated the CT2A-hEGFR immunocompetent GBM mouse model for 2 days, starting on day 5 after tumor implantation on day 0. On day 7, the mice were sacrificed, and their brains were homogenized to analyze the total number of innate (NK cells and macrophages) and adaptive (T cells) immune cells by flow cytometry. The total number of immune cells as well as their subsets, NK cells, macrophages, and T cells, were significantly increased in the OV-Cmab-mCCL5-treated mice compared to either the vehicle control mice or the OV-Q1-treated mice (Fig. 7b,c and Extended Data Fig. 9b–d). Furthermore, compared to the other groups, the OV-Cmab-mCCL5 group had higher numbers of CD4⁺ T cells and CD8⁺ T cells and a lower CD4⁺/CD8⁺ T cell ratio (Fig. 7d). Next, we determined whether infiltration of immune cells into the TME correlated with doses by infusing 1 × 10⁵, 2 × 10⁵ or 4 × 10⁵ PFU OV-Cmab-mCCL5. We indeed observed a dose-dependent effect on the tumor infiltration of NK cells, macrophages and T cells (Extended Data Fig. 9e). The dose effect of immune cell infiltration is consistent with the similar dose effect observed for the survival of mice bearing CT2A-hEGFR GBM cells (Extended Data Fig. 9f). Collectively, our data indicate that OV-Cmab-mCCL5 promotes the infiltration of innate and adaptive immune cells into the GBM TME and prolongs the survival.

We measured in vivo NK cell activation by assessing CD69 expression with flow cytometry. OV-Cmab-mCCL5 treatment significantly increased the percentage of CD69⁺ NK cells compared to OV-Q1 or saline treatment, indicating that OV-Cmab-mCCL5 activates NK cells in vivo (Fig. 8a). To determine how OV-Cmab-mCCL5 treatment affects macrophage function, we performed an in vivo phagocytosis assay with the CT2A-hEGFR immunocompetent mice. Briefly, C57BL/6J mice were injected i.c. with 1 × 10⁵ GFP-expressing CT2A-hEGFR cells. Five days after tumor implantation, mice received i.c. 2 × 10⁵ PFU OVs or vehicle control. Two days after viral injection, mice were euthanized, and cells in brains were isolated. The macrophage phagocytosis of GBM cells was measured by flow cytometry. The results showed that OV-Cmab-mCCL5 significantly enhanced the macrophage phagocytosis against CT2A-hEGFR cells when compared to OV-Q1 or saline (Fig. 8b). We also assessed interferon-γ (IFN-γ) and granzyme B secretion by CD8⁺ T cells in the treated brain. After treatment with OV-Cmab-mCCL5 or OV-Q1, we observed higher IFN-γ and granzyme B secretion, but not after saline treatment. Moreover, OV-Cmab-mCCL5 was significantly more effective in the induction of IFN-γ and granzyme B secretion than OV-Q1 (Fig. 8c,d). Thus, OV-Cmab-mCCL5 was also able to activate the adaptive immune system.

Fig. 8 |. Improvement of OV-Cmab-mCCL5 correlates to immune cell activation and requires NK cells, macrophages and T cells.

CD69 expression in NK cells isolated from mouse brains 2 days after virus injection. Mice implanted with CT2A-hEGFR GBM cells were treated with OV-Cmab-mCCL5, OV-Q1 or saline with a 5-day time interval. Representative flow cytometric plots of each group are shown on the left, and summary data are on the right (n = 4 mice in each group). b, Mice implanted with CT2A-hEGFR cells expressing GFP were treated with OV-Cmab-mCCL5, OV-Q1 or saline with a 5-day interval. Two days after viral injection, the percentage of GFP⁺ macrophages (CD45⁺F4/80⁺CD11b⁺) in the brain, as determined by flow cytometry, indicated the level of phagocytosis (n = 4 mice in each group). c,d, Mice were treated as in panel a. Percentages of IFN-γ⁺ CD8⁺ T cells (c) and GzmB⁺ CD8⁺ T cells (d) from mouse brains 2 days after virus injection were determined by flow cytometry. Error bars indicate s.d., and data are presented as mean ± s.d. (a–d). Statistical analyses were performed by one-way ANOVA with P values corrected for multiple comparisons by the Bonferroni method (n = 4 mice in each group (a–d)). e–g, Survival of CT2A-hEGFR tumor-bearing C57BL/6J immunocompetent mice treated with six conditions: vehicle control, OV-Q1, or OV-Cmab-mCCL5 in the presence or absence of NK cell depletion (e), macrophage depletion (f), or T cell depletion (g). Survival in panels e–g was analyzed by the Kaplan–Meier method and compared by log-rank test (n = 7 mice in each group). P = 0.0002 for the difference between OV-Cmab-mCCL5 with versus without NK cell depletion and the difference of OV-Q1 with versus without NK cell depletion (e). P = 0.0004 for the difference between OV-Cmab-mCCL5 with versus without macrophage depletion and the difference between OV-Q1 with versus without macrophage depletion (f). P = 0.0003 for the difference between OV-Cmab-mCCL5 with versus without T cell depletion and the difference of OV-Q1 with versus without T cell depletion (g).

To determine the contribution of different immune cells toward the efficacy of OV-Cmab-mCCL5, we repeated the survival studies in the immunocompetent GBM mouse model with or without depleting NK cells, macrophages or T cells. Depletion of NK cells or macrophages had no effect on the survival of either the OV-Q1 or vehicle control group. This might be partly explained by the aggressive nature of GBM in that model, as nearly all the untreated mice were dead within ~18 days after tumor implantation. In the OV-Cmab-mCCL5-treated mice, the depletion of NK cells or macrophages resulted in a significant decrease in survival compared to the group that did not undergo depletion of these innate immune cells (Fig. 8e,f). This suggests that these cell types mediate an important anti-tumor effect following treatment with OV-Cmab-mCCL5. These data are also consistent with our in vitro data showing that Cmab-CCL5 produced by GBM cells infected with either OV-Cmab-hCCL5 or OV-Cmab-mCCL5 significantly induced NK cell migration, activation, and ADCC as well as macrophage migration and ADCP (Figs. 2–4 and 6a–d). Depletion of T cells almost completely abolished the protective effect of OV-Cmab-mCCL5, indicating that adaptive immune cell activation is also important to mediate the anti-tumor effect following treatment with OV-Cmab-mCCL5 in this immunocompetent model (Fig. 8g). Collectively, our data show that treatment with OV-Cmab-mCCL5 induces infiltration and activation of innate and adaptive cytolytic immune cells in the GBM TME, both of which mediate the anti-tumor effect following treatment with OV-Cmab-mCCL5. Notably, in this immunocompetent model, T cells of the adaptive immune system seemed to be more important than NK cells and macrophages of the innate immune system for the anti-tumor effects of OV-Cmab-mCCL5, as the depletion of T cells but not NK cells or macrophages resulted in almost complete abrogation of the OV-Cmab-mCCL5 protective effect.

OV-Cmab-mCCL5 enhances the abscopal effect in GBM

The abscopal hypothesis posits that untreated tumors shrink when other tumors are locally treated. To test whether OV-Cmab-mCCL5 has a more pronounced abscopal effect than OV-Q1, we implanted an equal number of GBM cells into both hemispheres of the brain and treated only one hemisphere with OV-Cmab-mCCL5, OV-Q1 or saline. Magnetic resonance imaging (MRI) showed that OV-Cmab-mCCL5 achieved the best inhibition effect of tumor growth on both sides of the brain, whereas OV-Q1 partially inhibited tumor growth on the treated side and had little effect on the untreated side when compared to the saline group (Extended Data Fig. 10a). To determine how OV-Cmab-mCCL5 had affected the untreated side, we used flow cytometry to assess the level of infiltration by innate and adaptive immune cells on the untreated side brain. OV-Cmab-mCCL5 treatment recruited more total immune cells and their subsets of NK cells, macrophages and total T cells to the untreated-side of the brain than either the OV-Q1 or vehicle control treatment (Extended Data Fig. 10b). Furthermore, we repeated the abscopal model and treated the mice with OV-Cmab-mCCL5, OV-Q1 or saline in only one hemisphere of the brain 1 or 3 days after tumor implantation on day 0. The mice treated with OV-Cmab-mCCL5 on one side had a significantly longer survival time than the saline group or the OV-Q1 group (Extended Data Fig. 10c). Collectively, our data suggest that OV-Cmab-mCCL5 has a more pronounced abscopal effect than OV-Q1, correlating with better immune cell infiltration in the former.

Discussion

In this study, we constructed OV-Cmab-CCL5, an oHSV targeting both the EGFR and CCL5 receptors. OV-Cmab-CCL5 produced by infected tumor cells not only produces high and sustainable levels of CCL5 in the TME, thereby increasing infiltration of both innate and adaptive immune cells, but also functions as an IgG1 anti-EGFR monoclonal antibody (mAb) to activate NK cells via ADCC, activate macrophages via ADCP and reduce EGFR signaling in tumor cells. Therefore, local delivery of OV-Cmab-CCL5 induced (1) direct tumor lysis by oHSV, (2) innate and adaptive immune cell infiltration into the TME promoted by CCL5, (3) Cmab-mediated ADCP by bridging Fcγ receptors on activated macrophages with EGFR⁺ GBM cells, (4) Cmab-mediated-ADCC by bridging Fcγ receptors on activated NK cells with EGFR⁺ GBM cells and (5) blocking EGFR signaling in EGFR⁺ GBM cells. In vivo, OV-Cmab-CCL5 improved oncolytic virotherapy in immunodeficient xenograft and immunocompetent GBM models. Furthermore, depletion of NK cells, macrophages or T cells impaired or abolished the anti-tumor effect of OV-Cmab-mCCL5. Thus, our OV provides both in vitro and in vivo preclinical evidence for a compelling therapeutic candidate to be considered for patients with GBM or other EGFR⁺ cancers.

The first and only FDA-approved OV, talimogene laherparepvec (T-VEC), is being used to treat multiple cancers²². T-VEC expresses two copies of GM-CSF driven by the strong cytomegalovirus promoter in the modified HSV-1 genome without γ34.5 expression²³. Recently, the OV G207, with a backbone similar to that of OV-Cmab-CCL5, proved to be safe and effective for treating high-grade pediatric glioma¹². In the current study, we engineered the backbone to express an antibody-based fusion protein, OV-Cmab-CCL5. In experimental systems, OV-Cmab-CCL5 proved suitable for treating GBM, a devastating and highly lethal cancer with a median survival time of approximately 15 months²⁴. We designed OV-Cmab-CCL5 to enhance the immune cell effects of oHSV. Indeed, our data show that OV-Cmab-CCL5 not only increases immune cell infiltration but also activates various immune cells in vitro and in vivo in a fashion that is significantly superior to the control OV-Q1. These enhanced effects include anti-tumor effects mediated by both innate and adaptive immune cells as their individual depletion impaired or abolished the effect of OV-Cmab-CCL5 on improving survival of mice bearing GBM, though the adaptive immune cells had a more profound effect on survival than did the innate immune cells. Furthermore, our virus targeted both wtEGFR and EGFRvIII GBM cells, allowing us to treat GBM with the heterogeneous expression of these two forms of EGFR. Importantly, our OV-Cmab-CCL5 is a single agent that may also be effective if combined with local administration of immune cell modulators such as checkpoint inhibitors or other reagents, as we recently demonstrated for two other oHSVs that we generated^25,26.

The chemokine system not only brings the antigen-presenting cells and naive T cells together to activate an adaptive immune response but also delivers both innate and adaptive immune cells to the TME^27,28. CCL5 is crucial for recruiting various leukocytes into inflammatory sites, including NK cells, macrophages, T cells, eosinophils and basophils^9,29. Its expression by tumor cells is also important for T cell infiltration into tumors¹⁰. Moreover, CCL5 released by T cells induces the activation and proliferation of NK cells to generate C-C chemokine-activated killer cells²⁹. Furthermore, CCL5 can also promote the survival of macrophages⁷. However, CCL5 is often epigenetically silenced in solid tumor cells¹⁰ and so is unable to attract and activate innate and adaptive immune cells that target tumor cells. We therefore examined whether overexpressing CCL5 locoregionally could be a promising approach for solid tumor therapy, as systemic delivery of chemokines often produces severe side effects³⁰ and chemokines have very short half-lives and are challenging to deliver³¹. OVs carrying chemokines can address these challenges, especially for solid tumors in situ, such as GBM, because they can be administered directly into the tumor. This strategy enriches chemokines only in the tumor site and promotes locoregional infiltration of innate and adaptive immune cells into the TME, followed by activation of the local immune response, as we have demonstrated in this study.

Over the past 15 years, anti-tumor therapies based on mAbs have shown remarkable success for certain solid tumors^32–34. Many therapeutic mAbs targeting tumor-associated antigens have been approved by the FDA, including cetuximab^33,35,36. Cetuximab is an EGFR inhibitor that binds the extracellular domain of wtEGFR or EGFRvIII, decreases EGFR signaling and blocks tumor growth and metastasis^37,38. In preclinical GBM models, cetuximab has displayed anti-tumor and radiosensitizing effects on GBM^39,40. Thus, cetuximab therapy is a promising strategy for treating GBM tumors that overexpress wtEGFR and/or EGFRvIII. In our approach, cetuximab produced by tumor cells infected with OV-Cmab-CCL5 was released continuously into the TME to anchor to GBM cells and block EGFR signaling, the latter of which might also contribute to tumor cell growth inhibition.

The TME of GBM is considered immunosuppressive due to cytokines secreted by tumor cells and tumor-associated macrophages that inhibit both the innate and adaptive immune systems^41–43. Thus, NK cell activity is suppressed and T cell proliferation is downregulated⁴⁴. Furthermore, some standard treatments, such as chemotherapy, can kill immune cells and cause immune suppression⁴⁵. Thus, combining standard therapy with immunotherapy has emerged as a key tool for targeting GBM. Immunotherapies, such as checkpoint inhibitors, chimeric antigen receptor T cell therapy and oncolytic virotherapy are all under active consideration for GBM treatment⁴⁶. Our OV-Cmab-CCL5 platform has a broad application, as we can replace part or the entire Cmab-CCL5 with a bispecific antibody, fusion protein, cytokine or any other agent to specifically change the TME from cold to hot or further fine-tune a therapeutic approach. Compared to lentivirus or retrovirus-base gene therapy for GBM, oHSV has the advantage of being neurotropic and having a genome size of ~150 kb, the latter of which allows it to carry a large insert that could include one or multiple transgenes.

Several clinical trials have aimed to specifically activate dysfunctional T cells, which develop senescence, anergy, tolerance, exhaustion and ignorance in GBM^47–49. For example, adoptive T cell transfer has been shown to be safe and effective^50,51. In our study, intratumoral injection of OV-Cmab-CCL5 promoted CD4⁺ and CD8⁺ T cell infiltration, whereas depletion of T cells completely abolished the survival advantage of therapy with OV-Cmab-CCL5. Some studies showed that the CD4/CD8 T cell ratio predicts the outcome of cancer treatment, as CD8⁺ cells play a crucial role in anti-tumor immune responses^52,53. Specifically, CD8⁺ T cell infiltration is associated with prolonged survival in GBM patients⁵⁴. Our data show that intratumoral injection of OV-Cmab-CCL5 not only prolongs survival but also significantly increases T cell infiltration and decreases the CD4⁺/CD8⁺ T cell ratio, indicating that CD8⁺ T cells play an important role in OV-Cmab-CCL5 therapy.

In summary, we generated an oHSV bispecific fusion protein platform expressing cetuximab and CCL5, which combines chemokine activity with mAb function to treat GBM. This platform can deliver a chemokine in a highly targeted fashion, because the second part, such as one encoding a mAb against a tumor-associated antigen, can directly bind or anchor to tumor cells in the TME. Our current platform improves the efficacy of oncolytic virotherapy by combining an immunoregulatory factor with mAb therapy into a single agent that has multifaceted functions for tumor treatment.

Methods

Ethics statement

All experiments using mice were conducted in compliance with federal, state and local guidelines and with approval from the City of Hope Animal Care and Use Committee. Various human cells were isolated from peripheral blood cones of healthy donors, who did not receive compensation, after providing written informed consent under a protocol approved by the City of Hope Institutional Review Board.

Cells

All mouse and human GBM cell lines, human ovarian cell lines A2780 and CHO cells were cultured in DMEM supplemented with 10% FBS, penicillin (100 U ml⁻¹) and streptomycin (100 μg ml⁻¹). CT2A-hEGFR cells were generated by transfecting CT2A cells to express the human EGFR gene. GBM30 spheroid cells, which are patient derived, were modified to express a FFL gene for in vivo imaging and were named GBM30-FFL. Both GBM30 and GBM30-FFL cells were maintained as tumor spheres in basic neurobasal media supplemented with 2% B27, human epidermal growth factor (20 ng ml⁻¹) and fibroblast growth factor (20 ng ml⁻¹) in low-attachment cell-culture flasks. Vero cells, which were derived from monkey kidney epithelium, were maintained in the same media used for the GBM cell lines. Gli36ΔEGFR, U87ΔEGFR, U251T2 and GBM30 cells were authenticated in January 2015 by the University of Arizona Genetics Core using short tandem repeat profiling. The rest cells were not authenticated after receipt. All cell lines were routinely tested for the absence of mycoplasma using the MycoAlert Plus Mycoplasma Detection Kit (Lonza).

Generation and purification of Cmab-hCCL5 and Cmab-mCCL5

Cmab-hCCL5 and Cmab-mCCL5 were purified from the conditional supernatants of lentivirus-infected CHO cells using a protein G column (Thermo Fisher Scientific, catalog no. 89927). For this purpose, the sequence of cetuximab single-chain variable fragment was reconstructed as previously reported⁵⁵ to generate Cmab-h(m)CCL5 on an IgG1 scaffold using a pCDH lentiviral vector with a GFP. The ‘knob’ and ‘hole’, encoding Cmab-Fc and CCL5-Fc, respectively, were linked with a DNA sequence encoding a T2A self-cleaving peptide to express them simultaneously. The lentivirally transduced cells were fluorescence-activated cell sorting (FACS) sorted to be GFP⁺ using a FACS Aria II cell sorter (BD Biosciences).

Determination of Cmab-h(m)CCL5 binding to EGFR on tumor cells

Human U251T2, U87ΔEGFR and A4780 cells or murine CT2A-hEGFR cells were preblocked with 2% BSA and incubated with 1, 5 or 10 μg ml⁻¹ purified Cmab-hCCL5 or human IgG1 isotype control for 30 min. After incubation, all cells were washed twice, and the human cells were stained for 20 min with allophycocyanin (APC)-conjugated anti-human Fc (Jackson ImmunoResearch, catalog no. 209-605-098) or APC-conjugated anti-human CCL5 (BioLegend, catalog no. 515506) antibodies, whereas the murine cells were stained for 20 min with APC-conjugated anti-human Fc or phycoerythrin (PE)-conjugated anti-mouse CCL5 (BioLegend, catalog no. 149103) antibodies. The stained cells were analyzed using a Fortessa X-20 flow cytometer (BD Biosciences).

Generation of OV-Cmab-hCCL5 and OV-Cmab-mCCL5

OV-Cmab-hCCL5 and OV-Cmab-mCCL5 were generated as previously described with some modifications^13,56. Briefly, the Cmab and CCL5 were linked to the knob and the hole, respectively, of an IgG1 scaffold, followed by connecting with the knob and the hole by a T2A. The linked DNA fragment (Cmab-Fc-T2A-hCCL5-Fc for a human version and Cmab-Fc-T2A-mCCL5-Fc for a mouse version) was inserted into a shuttle plasmid downstream of the HSV pIE4/5 promoter to construct the pTransfer-Cmab-hCCL5 and pTransfer-Cmab-mCCL5 plasmids. Each of the two transfer plasmids was recombined with an HSV bacterial artificial chromosome to engineer OV-Cmab-hCCL5 and OV-Cmab-mCCL5. Vero cells were used to propagate and titrate the viruses. Virus titration was performed with plaque assays in which monolayer Vero cells were seeded in a 96-well plate and infected with gradient-diluted viral solutions. Two hours after the initial infection, the infection media were replaced with DMEM containing 10% FBS. Two days after infection, GFP⁺ plaques were counted with a Zeiss fluorescence microscope (AXIO observer 7) to calculate viral titer. To concentrate and purify OV-Q1, OV-Cmab-hCCL5, and OV-Cmab-mCCL5 viral particles, we harvested the culture media containing viruses and centrifuged them at 3,000 × g for 30 min. The supernatants were collected and ultra-centrifuged at 100,000 × g for 1 h. The virus pellets were resuspended with saline as needed.

Immunoblotting assay

The immunoblotting assay was performed as previously described¹³. U251T2 GBM cells were infected with OV-Q1, OV-Cmab-hCCL50 or OV-Cmab-mCCL5 at an MOI of 2. The infection media were replaced with fresh media after a 2-hour infection. Supernatants from each group were harvested at 72 hpi to extract protein for immunoblotting, using a chloroform-methanol method. Cmab-hCCL5 and Cmab-mCCL5 fusion proteins as controls were extracted from lentivirus-infected CHO cells. Rabbit anti-human IgG heavy chain antibody (Sigma, catalog no. MAB1307) and rabbit anti-human CCL5 antibody (R&D, catalog no. MAB278-100) or anti-mouse CCL5 antibody (R&D, catalog no. MAB478-100) with anti-rabbit secondary antibody (LI-COR, catalog no. 925-32210), were used to detect Fc and CCL5, respectively, in supernatants of OV-Cmab-h(m)CCL5 infected GBM cells. To measure EGFR signaling, we injected i.c. 1 × 10⁵ CT2A-hEGFR cells into C57BL/6J mice. Five days after tumor implantation, the mice received an intratumoral injection of 2 × 10⁵ PFU OV or vehicle control. On day 3 after OV treatments, brains were harvested and homogenized for immunoblotting assay. Anti-phospho-AKT antibody (Cell Signaling Technology, catalog no. 9271 L), anti-AKT antibody (Cell Signaling Technology, catalog no. 9272), anti-phospho-EGFR antibody (Cell Signaling Technology, catalog no. 3777), anti-EGFR antibody (Abcam, catalog no. ab52894) and anti-β actin antibody (Proteintech, catalog no. 66009) were used.

Measurement of Cmab-hCCL5 and Cmab-mCCL5 concentration by ELISA

U251T2 GBM cells were infected with OV-Q1, Cmab-hCCL5 or Cmab-mCCL5 at an MOI of 2. Two hours after the infection, the infection media were discarded and replaced with fresh media. The supernatants from each group were then harvested at 12, 24, 48 and 72 hpi to measure Cmab-hCCL5 or Cmab-mCCL5 concentrations by ELISA. Cmab-hCCL5 or Cmab-mCCL5 samples with known concentrations purified from CHO cells, were used as standards for ELISA that was used to detect the concentration of Cmab-hCCL5 or Cmab-mCCL5, for which recombinant human EGFR protein (Abcam, catalog no. ab174029) was used as a coating reagent. Anti-human Fc antibody (Sigma, catalog no. MAB1307) was used to detect Fc within the Cmab-Fc fusion protein. Commercial human or mouse CCL5 Quantikine ELISA Kits (R&D, catalog no. DRN00B, MMR00, respectively) were used to detect hCCL5 or mCCL5 within the Cmab-hCCL5 or Cmab-mCCL5 fusion proteins, respectively, via the interaction between a CCL5 antibody and the CCL5 part of Cmab-hCCL5 or Cmab-mCCL5.

Preparation of immune cells for in vitro studies

Human monocytes were isolated and enriched with the RosetteSep Human Monocyte Enrichment Cocktail Kit (Stemcell, catalog no. 15068) from the peripheral blood of healthy donors. Human NK cells were isolated and enriched using the RosetteSep Human NK Enrichment Cocktail Kit (Stemcell, catalog no. 15065), and the RosetteSep Human T Cell Enrichment Cocktail (Stemcell, catalog no. 15061) was used to isolate and enrich human T cells. The enriched human monocytes were cultured for 7 days with RPMI-1640 media containing 20 ng ml⁻¹ human M-CSF (PeproTech, catalog no. 300-25-50 UG) and 2% human serum to induce macrophage differentiation. The culture media were replaced on day 3 and day 5. Dynabeads Human T-Activator CD3/CD28 for T Cell Expansion and Activation (Thermo Fisher Scientific, catalog no. 11132D) were used to activate human T cells for 3 days with RPMI-1640 media containing 20% FBS, 100 U ml⁻¹ penicillin/streptomycin and 10 ng ml⁻¹ human IL-2. Human NK cells were used immediately for in vitro migration and cytotoxicity assays.

For isolating and culturing mouse macrophages, bone marrow cells of C57BL/6J mice were harvested and extracted from tibias and femurs. Red blood cells were lysed with RBC Lysis Buffer (Thermo Fisher Scientific, catalog no. 00-4300-54). Extracted bone marrow cells were planted in a 100-mm culture dish (BD Falcon) and cultured for 7 days with RPMI-1640 media containing 10% FBS in the presence of 20 ng ml⁻¹ murine M-CSF (PeproTech, catalog no. 315-02). Cultural media were replaced on day 3 and day 5. The EasySep Mouse NK Cell Isolation Kit (Stemcell, catalog no. 19855) and EasySep Mouse T Cell Isolation Kit (Stemcell, catalog no. 19851) were used to isolate mouse NK and T cells, respectively, from spleens of C57BL/6J mice. Freshly isolated NK and T cells were used immediately for in vitro migration and cytotoxicity assays.

Cell migration assay

The migration capability of NK cells, T cells and macrophages was determined using transwell chambers (Corning). Serum-starved NK cells, T cells, or macrophages were seeded into the upper chamber. Recombinant human CCL5, recombinant mouse CCL5, purified Cmab-hCCL5 or Cmab-mCCL5 or supernatant containing Cmab-hCCL5 or Cmab-mCCL5 that had been secreted from OV-Cmab-hCCL5- or OV-Cmab-mCCL5-infected U251T2 GBM cells was added to the lower chamber. The upper and lower chambers were incubated at 37 °C in 5% CO₂ for 12 h. Cells that migrated to the lower face of the porous membrane in the lower chamber were counted. Alternatively, 1 × 10⁵ U251T2 cells were seeded into the lower chamber overnight, followed by infection with OV-Cmab-hCCL5 at an MOI of 2 for 72 h. The lower chamber with the infected cells was co-incubated with the upper chamber containing immune cells for 72 h. The number of immune cells that had immigrated into the lower chamber was determined by flow cytometry after being stained with an anti-CD45 antibody. For detecting T cell and NK cell migration, the pore size of the membrane was 3 μm^57,58, whereas for macrophages, the pore size was 8 μm⁵⁹.

NK cell cytotoxicity and activation assay

NK cell cytotoxicity was evaluated as previously described^13,60,61. U251T2, GBM30, Gli36ΔEGFR and CT2A-hEGFR cells were used as target cells. The target cells were labeled with ⁵¹Cr for 1 h, followed by preincubation with 5 μg ml⁻¹ or 10 μg ml⁻¹ Cmab-h(m)CCL5 fusion protein or IgG1 isotype control for 30 min. The labeled and preincubated target cells were then co-cultured with isolated human or mouse primary NK cells at different effector/target ratios at 37 °C for 4 h. Release of ⁵¹Cr was measured as counts per minute (cpm) with a MicroBeta² microplate radiometric counter (PerkinElmer). Target cells incubated in complete media were used as spontaneous release controls, and those in 2% SDS media were used as maximal ⁵¹Cr release controls. Cell lysis percentages were calculated using the standard formula: 100 × (cpm experimental release − cpm spontaneous release)/(cpm maximal release − cpm spontaneous release). The assays were performed with three or more technical replicates, using NK cells from different donors. Expression of CD69 and granzyme B was measured after co-culturing NK cells with tumor cells for 6 h with 5 μg ml⁻¹ or 10 μg ml⁻¹ Cmab-h(m)CCL5 at a ratio of 1:1. The cells were then stained with anti-CD56 (BD Biosciences, catalog no. 557919), anti-CD69 (BD Biosciences, catalog no. 562883) and anti-granzyme B (BD Biosciences, catalog no. 563388) antibodies (human NK cells) or anti-CD69 (BD Biosciences, catalog no. 564683) with anti-NKp46 (BioLegend, catalog no. 137618) (mouse NK cells) antibodies, followed by analyzing with a Fortessa X-20 flow cytometer (BD Biosciences). The gating strategy is shown in Supplementary Fig. 1.

Phagocytosis assay based on flow cytometry

Primary human or mouse macrophages were incubated with carboxyfluorescein succinimidyl ester (CFSE)-labeled (Thermo Fisher Scientific, catalog no. C34554) GBM30 or CT2A-hEGFR target cells, respectively, at an effector/target ratio of 1:2 for 6 h in the presence 5 μg ml⁻¹ or 10 μg ml⁻¹ Cmab-h(m)CCL5 fusion protein or IgG1 isotype control in a humidified, 5% CO₂ incubator at 37 °C in ultra-low-attachment 96-well U-bottom plates (Corning) in serum-free 1640 media (Life Technologies). The co-cultured cells were harvested by centrifuging them at 400 × g for 5 min at 4 °C and stained with an anti-human CD45 antibody (BioLegend, catalog no. 368516) or an anti-mouse F4/80 antibody (BD Biosciences, catalog no. 565787) to identify human or mouse macrophages, respectively, in flow cytometry assay. Human macrophage phagocytosis was quantified as the percentage of CD45⁺CFSE⁺ macrophages among total CD45⁺ cells. Murine macrophage phagocytosis was quantified as the percentage of F4/80⁺CFSE⁺ macrophages among total F4/80⁺ macrophages. The gating strategy is shown in Supplementary Fig. 2.

Quantitative PCR

For evaluating the effect of Cmab-h(m)CCL5 on mRNA levels of typical human or mouse macrophage cytokine genes, human or mouse macrophages and GBM30 or CT2A-hEGFR tumor cells, respectively, were co-cultured at a ratio of 1:1 for 6 h with or without 5 μg ml⁻¹ Cmab-h(m)CCL5. Human or mouse macrophages in the co-culture were FACS-purified after being stained with anti-human CD45 antibody or anti-mouse F4/80 antibody, receptively, using an Aria Fusion flow cytometer (BD Biosciences). Total RNA was extracted from the sorted macrophages to measure mRNA levels of human IL1B, IL6, IL12A and NOS2 or murine Il1b, Il6, Il12b, Ccl-2, Ccl-4 and Nos2 genes with their corresponding primers (Supplementary Table 1). 18 s rRNA was used as an internal control.

Animal studies

For all mouse study, mice were housed in the City of Hope Animal Facility with a 12-light/12-dark cycle and temperatures of 65–75 °F (~18–23 °C) with 40–60% humidity. GBM-bearing mice were monitored frequently to evaluate disease progression. Mice were euthanized when they were moribund or with substantial neurologic impairments or >20% weight loss, the early removal criteria approved by the City of Hope Animal Care and Use Committee. All experiments were performed in accordance with institutional animal care and use committee requirements for i.c. tumor burden measurements.

For establishing an immunodeficient xenograft GBM model using NSG mice, 6- to 8-week-old NOD.Cg-Prkdc^scid Il2rγ^tm1Wjl/SzJ female mice (strain 005557) were purchased from The Jackson Laboratory. For survival studies in this model, mice were anesthetized and stereotactically injected with 1 × 10⁵ GBM30-FFL cells into the right frontal lobe of the brain (2 mm lateral and 1 mm anterior to bregma at a depth of 3 mm). Four days after tumor implantation, 1 × 10⁶ PBMCs isolated from a single donor were i.v. injected into each mouse. On day 6, 1 × 10⁶ activated T cells derived from the same donor were delivered to each mouse by i.v. injection. On day 7, animals were subsequently randomly divided into groups that were i.c. injected with either 2 × 10⁵ PFU oHSV (OV-Q1 or OV-Cmab-hCCL5) in 3 μl saline or saline only as a control. Luciferase-based in vivo images were taken 15 and 21 days after tumor implantation to assess tumor development. For the immunodeficient xenograft GBM model in nude mice, 6- to 8-week-old female mice (strain 002019) purchased from The Jackson Laboratory were implanted with GBM cells and treated similarly as we performed for the NSG model but without infusing human PBMCs and activated T cells to unravel the roles of NK cells and macrophages. Luciferase-based in vivo images were taken 15 days and 20 days after tumor implantation to assess tumor development.

For establishing an immunocompetent mouse GBM model, 6–8-week-old female and male C57BL/6J mice (strain 000664) were purchased from The Jackson Laboratory. Then, 1 × 10⁵ CT2A-hEGFR cells were injected as described for the immunodeficient xenograft model. For survival studies, 3, 5 or 7 days after tumor implantation, mice were randomly divided into groups that were injected i.c. with either 2 × 10⁵ PFU oHSV (OV-Q1 or OV-Cmab-mCCL5) in 3 μl saline or with saline alone as control. For the in vivo immune cell infiltration and activation study, 5 days after tumor implantation, mice were randomly divided into groups that were injected i.c. either with 2 × 10⁵ PFU oHSV (OV-Q1 or OV-Cmab-mCCL5) in 3 μl saline or with saline alone control. Mice were euthanized 2 days after being treated to evaluate immune cell brain infiltration using flow cytometry as described below.

For continuous delivery of Cmab-mCCL5 or mCCL5 into the tumor site in the CT2A-hEGFR immunocompetent mouse model, osmotic pumps (Alzet, catalog no. 1003D) were filled with 100 μl of desalted Cmab-mCCL5 or mCCL5 at the concentration of 1 mM. The pumps connected to brain infusion kits (Alzet, catalog no. 0008851) were implanted into the mouse brain to continuously deliver proteins at a speed of 1 μl per hour.

The in vivo phagocytosis assay was performed using the GFP⁺ CT2A-hEGFR immunocompetent mouse model. Five days after tumor implantation, mice were treated with OV-Cmab-mCCL5, OV-Q1, or saline for 2 days. Then mononuclear cells extracted from the brain with Percoll were stained with anti-CD45 (BD Biosciences, catalog no. 559864), anti-CD11b (BD Biosciences, catalog no. 552850), and anti-F4/80 (Thermo Fisher Scientific, catalog no. 12-4801-82) antibodies to identify macrophages. Phagocytosis was determined by the percentages of GFP⁺ macrophages among total CD45(high) CD11b⁺F4/80⁺ using a Fortessa X-20 flow cytometer.

Survival studies involving immune cell depletion were performed using the CT2A-hEGFR GBM model. NK cells were depleted by i.p. injection of NK1.1 neutralizing antibody (BioXcell, BE0036, clone PK136; 0.2 mg per mouse) twice, once on the day before and once after virus injection. Macrophages were depleted by i.p. injections of clodronate liposomes (0.2 ml/mouse), using the same schedule as in the NK cell depletion study. CD4⁺ and CD8⁺ T cells were depleted by three i.p. injections of anti-CD4 (BioXcell, BP0003–1, clone GK1.5; 0.2 mg/mouse) combined with anti-CD8 neutralizing antibodies (BioXcell, BE0004–1, clone 53–6.7; 0.2 mg per mouse), once every 3 days starting a day before virus injection. A corresponding isotype or liposomes were used as controls.

For establishing a bilateral CT2A-hEGFR immunocompetent GBM model, 6- to 8-week-old female and male C57BL/6J (strain 000664) mice from The Jackson Laboratory were anesthetized and stereotactically injected with 1 × 10⁵ CT2A-hEGFR cells into both the left and right frontal lobe of the brain (2 mm lateral and 1 mm anterior to bregma at a depth of 3 mm). Three days after tumor implantation, mice were randomly divided into groups that were injected i.c. only into the right hemisphere of the brain with either 2 × 10⁵ PFU oHSV (OV-Q1 or OV-Cmab-mCCL5) in 3 μl saline or with saline alone as control. MRI was used to monitor tumor size on day 9.

MRI

MRI imaging was performed on a 7 T positron emission tomography MRI system from MR Solutions (Bruker Corporation) with T2-weighted fast spin-echo imaging (echo time = 60 ms; repetition time = 5,000 ms; field of view = 35 × 35 mm; matrix = 160 × 160; 18 slices; slice thickness = 1 mm; matrix size, 256 × 238; total imaging time = 11 min 40 s).

Flow cytometry

Brain mononuclear cells extracted with Percoll were 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), and anti-CD8 (BD Biosciences, catalog no. 553030) antibodies for flow cytometric analysis of immune cell infiltration into the brain. Anti-IFN-γ (BD Biosciences, catalog no. 563773), anti-CD69 (BD Biosciences, catalog no. 564683), and anti-GZMB (Thermo Fisher Scientific, catalog no. 48-8898-82) antibodies were used to measure immune cell activation. The flow cytometric assessments were performed with four or more independent animals. Anti-human CD191 (CCR1, BioLegend, catalog no. 362907) and anti-human CD195 (CCR5, BioLegend, catalog no. 313707) antibodies were used to detect their expressions on human NK cells, macrophages and T cells. All flow cytometry data were collected using a Fortessa X-20 flow cytometer. Cell sorting was performed using a FACS Aria II cell sorter (BD Biosciences). The gating strategy is shown in Supplementary Fig. 3. BD FACSDiva and FlowJo 7.6 & 10.0 were used to collect and analyze the flow cytometric data, respectively.

Oncolysis assay

A total of 2 × 10³ per well U251T2 GBM cells were seeded into a special 96-well E-plate (ACEA Biosciences) and incubated at 37 °C to allow cells to attach and form a monolayer. Twelve hours later, different MOIs of OV-Q1 and OV-Cmab-hCCL5 were added to the plate. Oncolysis data were collected and analyzed with RTCA software (ACEA Biosciences).

H&E staining and immunohistochemistry assay

Brains isolated from the experimental mice were placed in 10% neutral buffered formalin for a minimum of 72 h. After paraffin embedding, 4 μM-thick sections were cut from the blocks. H&E staining and immunohistochemical staining with anti-HSV (Cell Marque, catalog no. 361A-15-ASR) antibody and anti-mCCL5 (Thermo Fisher Scientific, catalog no. 701030) antibody were used. Stained slides were mounted and scanned for observation.

Statistical analysis and reproducibility

For continuous endpoints that are normally distributed or logarithm-transformed values, data are presented as mean ± s.d. Two-sided Student’s 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 for group comparisons by accounting for the underlying variance and covariance structure. Survival data were collected with blinding, estimated by the Kaplan–Meier method, and analyzed by log-rank. All tests were two-sided. P values were adjusted for multiple comparisons by Bonferroni method. A P value of <0.05 was defined as statistically significant. Statistical software GraphPad, R.3.6.3 and SAS 9.4 were used for the analysis. No data were excluded from the analyses. No statistical method was used to predetermine the sample size. Sample sizes in this study were estimated based on previous experience that showed significance. All animals were randomized and exposed to the same environment. No blinding was used for in vitro experiments. The majority of in vitro data collected were quantifiable, and blinding would not change any bias in data collection. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Extended Data

Extended Data Fig. 1 |. Concentration of CCL5 secreted by GBM cells, and expression of CCL5 receptors on NK cells, macrophages, and T cells.

(a) Concentration of CCL5 secreted by human GBM cell lines (U251T2, Gli36ΔEGFR, and GBM30), measured by ELISA (n = 3 independent experiments). (b-e) CCR1 and CCR5 expression on human NK cells (b, n = 6 independent donors), macrophages (c, n = 5 independent donors), CD4⁺ T cells (d, n = 3 independent donors), and CD8⁺ T cells (e, n = 3 independent donors), measured by flow cytometry. Error bars indicate the standard deviations (s.d.) and data are presented as mean ± s.d..

Extended Data Fig. 2 |. Comparison between Cmab-hCCL5 and hCCL5-Cmab and characterization of OV-Cmab-hCCL5.

(a, b) Detection of purified Cmab-hCCL5 or hCCL5-Cmab bound to U251T2 cells, measured by flow cytometry after staining Cmab-hCCL5- or hCCL5-Cmab-incubated tumor cells with anti-Fc-APC (a) or anti-hCCL5-APC (b). Cmab-hCCL5 and hCCL5-Cmab were purified from lentivirus-infected CHO cells. IgG1 isotype served as control. (c, d) Detection of Cmab-hCCL5 in the supernatant, collected from OV-Cmab-hCCL5-infected U251T2 GBM cell culture, that was associated with U251T2 cells, measured by flow cytometry after staining supernatant-incubated U251T2 cells with anti-Fc-APC (c) or anti-hCCL5-APC (d). (e) Cmab-hCCL5 produced in the supernatant of OV-Cmab-hCCL5-infected U251T2 cells at different MOIs (n = 3 independent experiments). (f) Cell viabilities of OV-Cmab-hCCL5 infection at three indicated MOIs were measured at 24 hours post infection (hpi), 48 hpi, and 72 hpi (n = 3 independent experiments). (g) The ability of OV-Q1 and OV-Cmab-hCCL5 to induce oncolysis of GBM cells, measured by real-time cell analysis (RTCA) The experiment was performed twice with similar data. (h) U251T2 cells were infected with OV-Q1 or OV-Cmab-hCCL5 at an MOI of 2. The supernatant was harvested at the indicated time points to assess viral production using a plaque assay with Vero cells (n = 3 independent experiments). Experiments in a-d were representative of three independent experiments with similar results. Error bars indicate the standard deviations (s.d.), and data are presented as mean ± s.d. (e, f, h). Statistical analyses were performed by one-way ANOVA with P values corrected for multiple comparisons by Bonferroni method (e).

Extended Data Fig. 3 |. Migration of immune cells induced by OV-Cmab-hCCL5.

NK cells, macrophages, CD4⁺ T cells, and CD8⁺ T cells in the upper chamber induced by OV-Cmab-hCCL5- or OV-Q1-infected U251T2 GBM cells at an MOI of 2 in the lower chamber, measured by transwell assay. 72 hours after the transwell assay was set up, immune cells in the lower chamber were quantified by flow cytometry. Error bars indicate the standard deviations (s.d.) and data are presented as mean ± s.d.. Statistical analyses were performed by one-way ANOVA with P values corrected for multiple comparisons by Bonferroni method (n = 3 independent donors).

Extended Data Fig. 4 |. OV-Cmab-hCCL5 improves oncolytic virotherapy in nude mice bearing GBM30 cells.

(a) Survival of GBM30 tumor-bearing nude mice treated with OV-Q1, OV-Cmab-hCCL5, or vehicle control. Survival was estimated by the Kaplan–Meier method and compared by log-rank test (n = 7 mice). (b) Luciferase imaging of GBM30-FFL GBM mice with the indicated treatments 15 and 20 days post tumor implantation.

Extended Data Fig. 5 |. Construction and characterization of OV-Cmab-mCCL5.

(a, b) Detection of purified Cmab-mCCL5 bound to CT2A-hEGFR cells, measured by flow cytometry after staining Cmab-mCCL5-incubated tumor cells with anti-Fc-APC (a) or anti-mCCL5-PE (b). Cmab-mCCL5 was purified from lentivirus-infected CHO cells. IgG1 isotype served as control. (c) mCCL5 and human Fc levels in Cmab-mCCL5 in the concentrated supernatant from engineered CHO cells or OV-Q1- or OV-Cmab-mCCL5-infected CT2A-hEGFR cells, detected by immunoblotting. (d, e) mCCL5 (d) and Cmab (e) of the Cmab-mCCL5 fusion protein in the supernatant from OV-Cmab-mCCL5-infected CT2A-hEGFR cells, quantified by ELISA. Cmab-mCCL5 purified from engineered CHO cells with known concentrations served as standards (n = 3 independent experiments). (f) Migration of murine NK cells, macrophages, CD4⁺ T cells, and CD8⁺ T cells induced by Cmab-mCCL5 purified from engineered CHO cells, measured using a transwell assay. Recombinant mCCL5 (rmCCL5, 100 ng/ml) served as positive control. Experiments in a-c were repeated with three independent experiments with similar results. Error bars indicate the standard deviations (s.d.), and data are presented as mean ± s.d. (d-f). Statistical analyses were performed by 2-sided Student’s t test (f, n = 3 mice/biologically independent samples; rmCCL5 versus control and Cmab-mCCL5 versus IgG1 isotype are only compared).

Extended Data Fig. 6 |. Cmab-mCCL5 promotes innate and adaptive immunocyte activation in vitro.

(a) Cytotoxicity of mouse primary NK cells against Cmab-mCCL5 pre-treated-CT2A-hEGFR cells, measured by ⁵¹Cr release. Control versus Cmab-mCCL5, P = 0.0009. A linear mixed model was used to account for the underlying variance and covariance structure (n = 3 mice). (b) CD69 expression, measured by flow cytometry, on mouse primary NK cells co-cultured with CT2A-hEGFR cells in the presence of Cmab-mCCL5 purified from engineered CHO cells or IgG1 isotype control (n = 3 independent mice). (c) ADCP of mouse macrophages induced by Cmab-mCCL5 purified from engineered CHO cells, targeting CT2A-hEGFR cells. CT2A-hEGFR target cells were prelabeled with CFSE and then were co-cultured with mouse macrophages in the presence of Cmab-mCCL5 purified from engineered CHO cells or IgG1 isotype control. The percentage of mouse macrophages that had phagocytosed labeled tumor cells was measured by flow cytometry, determined by CFSE⁺ macrophages (n = 3 independent mice). (d) Cytokine RNA expression levels, measured by real-time RT-PCR, of mouse macrophages co-cultured with CT2A-hEGFR cells at a ratio of 1:1 with or without Cmab-mCCL5 for 6 hours. Macrophages were purified by cell sorting to extract total RNA for generating cDNA for real-time RT-PCR (n = 3 independent mice). Error bars indicate the standard deviations (s.d.) and data are presented as mean ± s.d. (a-d). Statistical analyses were performed by one-way ANOVA with P values corrected for multiple comparisons by Bonferroni method (n = 3 mice).

Extended Data Fig. 7 |. OV-Cmab-mCCL5 prolongs the survival of glioblastoma-bearing mice in an immunocompetent mouse model.

(a) Survival of CT2A-hEGFR tumor-bearing mice treated with OV-Q1, OV-Cmab-mCCL5, or vehicle control. Survival was estimated by the Kaplan–Meier method and compared by log-rank test (n = 7 or 8 mice). Of note, tumors remained in the OV-Cmab-mCCL5-treated group on the day of mouse sacrifice (day 50). (b) C57BL/6J mice bearing CT2A-hEGFR cells were intracranially treated with 2 × 10⁵ PFU of indicated oncolytic virus or vehicle control (saline) 3 days post tumor implantation. Eleven days post tumor implantation, brains were collected from mice treated with saline, OV-Q1, or OV-Cmab-mCCL5 for standard H&E staining. Data of three mice in each group are shown. Scale bar, 2 mm. (c) Scheme for main Fig. 6f. Mice bearing CT2A-hEGFR cells were treated with saline, OV-Cmab-mCCL5, OV-Q1 + mCCL5 delivered by an osmotic pump, or OV-Q1+ Cmab-mCCL5 delivered by an osmotic pump.

Extended Data Fig. 8 |. Production of mCCL5 and oHSV in the GBM TME.

(a) Cmab-mCCL5 production in the brains collected from GBM mice treated with the indicated different doses of OV-Cmab-mCCL5. Mice were sacrificed on day 3 post treatment. Error bars indicate s.d., and statistical analyses were performed by one-way ANOVA with P values corrected for multiple comparisons by Bonferroni method (n = 5 mice in the saline and 2 × 10⁵ PFU groups and n = 4 mice in the 1 × 10⁵ PFU and 4 × 10⁵ PFU groups). (b, c) Immunohistochemical (IHC) analysis of mCCL5, oHSV, and H&E staining of the brains collected from mice treated with saline, OV-Q1, or OV-Cmab-mCCL5. Slides with brain tissue isolated from the experimental mice were subjected to H&E and IHC staining, and anti-HSV antibody and anti-mCCL5 antibody are used for detecting mCLL5 production and oHSV (OV-Q or OV-Cmab-mCCL5) existence, respectively. Images with high and low magnification are shown in (b, scale bars, 100 μm) and (c, scale bar, 2 mm), respectively. The boxed images in (c) are shown at higher power in (b).

Extended Data Fig. 9 |. OV-Cmab-mCCL5 inhibits tumor EGFR signaling and promoted the infiltration of innate and adaptive murine immune cells into the GBM TME in a dose-dependent manner.

(a) Immunoblotting of p-AKT and p-EGFR from protein harvested from the brains of mice treated with OV-Q1, OV-Cmab-mCCL5, or saline (n = 3 independent mice). (b) Total number of immune cells in main Fig. 7b–d, measured by flow cytometry (n = 4 independent mice in the OV-Q1 group and n = 5 independent mice in the OV-Cmab-mCCL5 group and the saline group). (c, d) Representative flow cytometry data of immune cells infiltrated into the brains of mice treated with saline, OV-Q1, or OV-Cmab-mCCL5. Summary data are shown in main Fig. 7c. (e) Immune cell infiltration into the brains of mice treated with indicated different doses of OV-Cmab-mCCL5 for two days (n = 5 independent mice in each group). (f) Survival of CT2A-hEGFR GBM-bearing mice treated with indicated different doses of OV-Cmab-mCCL5. Saline served as control. Survival was estimated by the Kaplan–Meier method and compared by log-rank test (n = 7 independent mice in the 2 × 10⁵ PFU group and n = 6 in all other groups). For panels b and e, error bars indicate s.d. and data are presented as mean ± s.d., and statistical analyses were performed by one-way ANOVA with P values corrected for multiple comparisons by Bonferroni method. Experiments in c and d were representative results of one of four to five mice in each group with similar data.

Extended Data Fig. 10 |. OV-Cmab-mCCL5 improves abscopal control of GBM in brain.

(a) Representative MRI images of a bilateral CT2A-hEGFR immunocompetent GBM model treated with saline, OV-Q1, or OV-Cmab-mCCL5 only on the right side of the brain. Data of two out of three to four mice in each group with similar data are shown. (b) The total number of immune cells, NK cells, macrophages, T cells, CD4⁺ T cells, and CD8⁺ T cells as well as the ratio of CD4⁺ T cells to CD8⁺ T cells in the untreated left side of the brain of the mice with the right side of the brain treated as indicated, measured by flow cytometry. Error bars indicate the standard deviations (s.d.) and data are presented as mean ± s.d.. Statistical analyses were performed by one-way ANOVA with P values corrected for multiple comparisons by the Bonferroni method (n = 5 independent mice in each group). (c) Therapeutic effects of OV-Cmab-mCCL5 when treating the two-side model with 1-day or 3-day intervals. Survival was estimated by the Kaplan–Meier method and compared by log-rank test (n = 6 mice).

Data availability

Source data for Figs. 1–8 and Extended Data Figs. 1–10 are provided as Source Data files. All other data supporting the findings of this study are available from the corresponding author upon reasonable request. Source data are provided with this paper.