An Oncolytic Virus Expressing IL15/IL15Rα Combined with Off-the-Shelf EGFR-CAR NK Cells Targets Glioblastoma

Oncolytic viral (OV) therapy has been recognized as a promising approach for cancer treatment. It differs from traditional gene therapy, where a viral vector only serves to deliver a specific gene. The OV itself acts as an immunostimulatory agent that can also selectively replicate in tumor cells, resulting in their lysis without harming normal tissues. The concept of OV therapy has existed for decades. Findings over the past two decades support a concept that OV can boost systemic immunity and therefore antitumor immune responses (1). For augmenting OV therapeutic efficacy, various transgenes have been incorporated into the virus, as previously done for T-VEC, the first OV approved by the FDA for the treatment of melanoma (2).

IL15 is a pleiotropic cytokine and plays a key role in the development, homeostasis, activation, and survival of T, natural killer (NK), and NK-T cells (3). IL15 receptor alpha (IL15Rα) is one of three receptors required to mediate IL15 signaling (4). With this understanding, researchers have recently generated IL15 super cytokine agonists that include IL15 and complete or partial IL15Rα to improve in vivo antitumor activities (5, 6). Both IL15 and the IL15/IL15Rα complex are actively being tested in phase I and II clinical trials with evidence of immune modulation in patients (7–9). Moreover, engineering viruses to infect tumors and express transgenes, including immunostimulatory molecules such as cytokines, can improve the efficacy of oncolytic virotherapy (10–14).

Another promising approach to control tumor progression is engineering immune cells with a chimeric antigen receptor (CAR), which is an artificially modified fusion protein, including an extracellular antigen recognition domain fused to an intracellular signaling domain and designed to enhance the specificity of T cells, NK cells, or other immune cells (15). CAR-modified T cells are currently showing promising efficacy to treat solid tumors, including glioblastoma (GBM; refs. 16, 17). Recently, CAR-modified NK cells have also shown early success in lymphoid malignancies (18, 19). However, CAR T or CAR NK cells in combination with OV have not yet been widely explored.

Here, we hypothesized that the combination of OV-IL15C with off-the-shelf EGFR-CAR NK cells could improve antitumor efficacy in GBM. The frozen and readily available off-the-shelf EGFR-CAR NK cells were produced from the peripheral blood of individual donors, which are able to target both EGFR and EGFR variant III (EGFRvIII). Both OV-IL15C and EGFR-CAR NK cells demonstrated better cytotoxic activity compared with control OV or empty vector (EV)–transduced NK cells in vitro, respectively. In vivo, the combination of OV-IL15C and EGFR-CAR NK cells produced a synergistic antitumor effect and a significant prolongation of survival over either therapy alone, correlating with increased intracranial infiltration and activation of NK and CD8⁺ T cells.

Vero cells (derived from monkey kidney epithelial cells) were used for viral production and plaque assay-based viral titration. Human GBM cell lines (U251, LN229, U87vIII) and the mouse GBM cell line CT2A, as well as Vero and human embryonic kidney 293T cells, were cultured in DMEM media (Gibco) supplemented with 10% FBS, penicillin (100 U/mL), and streptomycin (100 μg/mL). GBM30 spheroid cells derived from a patient with GBM and modified to express luciferase (named GBM30-luc) were maintained with neurobasal media (DMEM/F12) supplemented with 2% B27 (Gibco), human epidermal growth factor (StemCell), basic fibroblast growth factor (StemCell), heparin (StemCell), and Glutamax (Gibco) in low-attachment cell culture flasks. Primary human cells used for tropism testing were purchased from ScienCell and cultured following the manufacturer's instructions. U251 cells and GBM30 cells were authenticated by the University of Arizona Genetics Core via STR profiling in January 2015 and March 2018, respectively. Vero cells, 293T, CT2A, U87vIII, and LN229 cells were not authenticated after receipt. These cell lines were either purchased from the ATCC or obtained from Dr. E. Antonio Chiocca's laboratory at Harvard University (Cambridge, MA). For all experiments, cells were frozen down in low passage cultures and used within 2–3 passages when thawed. All cell lines were routinely tested for the absence of Mycoplasma using the MycoAlert Plus Mycoplasma Detection Kit from Lonza.

Generation of OV-IL15C was performed using the fHsvQuik-1 system as previously described (20). The full-length human IL15 and IL15Rα sushi domain DNA sequences were cloned into the pT-oriSIE4/5 plasmid and driven by the herpes simplex virus (HSV) pIE4/5 promoter. For evaluating viral production and infectivity, plaque-forming assays were undertaken with the GBM cell lines U251 and LN229 infected with OV-Q1 or OV-IL15C at an indicated multiplicity of infection (MOI). The plaques in each group were imaged by using Zeiss fluorescent microscopy at different time points after infection.

An immunoblotting assay was performed to detect the IL15/IL15Rα complex. Cell lysates (from virus-infected 293T cells) were prepared with RIPA Lysis and Extraction Buffer (ThermoFisher) containing protease/phosphatase inhibitor cocktail (Halt). Protein concentration was assessed using a Rapid Gold BCA Protein Assay Kit (Pierce). Anti-HA-tag antibody (Cell Signaling Technology; clone: C29F4) used as a primary antibody was added for overnight incubation at 4°C, followed by incubation with an anti-rabbit secondary antibody (Cell Signaling Technology) for one hour at room temperature. β-Actin was used as a loading control.

An ELISA was performed to quantify the IL15/IL15Rα complex secreted into the supernatants from various virus-infected GBM cell lines, using a DuoSet ELISA kit (R&D Systems, Catalog No: DY6924) according to the manufacturer's instructions.

Peripheral blood cones were collected from healthy donors in the City of Hope National Medical Center Donor Apheresis Center (DAC) with written informed consent approved by the Institutional Review Board. The studies were conducted in accordance to the Declaration of Helsinki. Primary human NK cells isolated from peripheral blood mononuclear cells (PBMC) were preincubated with 10× concentrated supernatants from OV-Q1- or OV-IL15C–infected U251 cells for 18 hours at 37°C. Primary human CD8⁺ T cells were isolated and preincubated similarly, but the incubation time is 48 instead of 18 hours. A standard ⁵¹Cr-release assay was used to measure cytotoxicity levels of the above preincubated NK cells and a flow cytometry–based assay using a Fortessa X-20 flow cytometer for the preincubated CD8⁺ T cells. Before these assays measuring cytotoxicity levels, the preincubated NK cells were cultured with GBM30 or K562 target cells for 4 hours whereas the preincubated CD8⁺ T cells for 12 hours. The assays were performed in at least three technical replicates with human NK and CD8⁺ T cells from different donors. The flow cytometry data were analyzed by using Flowjo V10 software (Tree Star).

Primary human NK and CD8⁺ T cells were cultured in supernatants from OV-Q1- or OV-IL15C–infected U251 cells in the absence of recombinant human IL2 (rhIL-2) at 37°C. From days 0 to 4, the number of NK and CD8⁺ T cells in each group was counted daily by microscopy after staining with Trypan Blue Solution (ThermoFisher). The assays were performed with human NK and CD8⁺ T cells from different donors in at least three technical replicates for each donor.

Immunoblotting of phospho-STAT5 was used to determine NK and CD8⁺ T-cell activation after 30 minute- and 12-hour culture, respectively, in the supernatants from OV-Q1- or OV-IL15C–infected U251 cells. Cell lysates were prepared similarly as for immunoblotting of the IL15/IL15Rα complex, as described above. The anti–phospho-STAT5 (Tyr694; D47E7) XP Rabbit mAb (Cell Signaling Technology) was used as a primary antibody, and anti-rabbit mAb (Cell Signaling Technology) was used as a secondary antibody.

Primary human NK cells were isolated from PBMCs, and the purity was assessed by flow cytometry using anti-human CD3 (Miltenyi Biotec) and anti-CD56 mAbs (BD Biosciences). Highly purified NK cells were seeded at 1 × 10⁵ cells/mL and cultured with irradiated (25 Gy) autologous PBMCs (1 × 10⁶ cells/mL) for expansion in culture medium, containing 5% human AB serum, and supplemented with hrIL-2 (NIH; 1,000 IU/mL) and anti-CD3 (OKT3; Invitrogen; 10 ng/mL). After 5 days, the media were completely changed, and cells were resuspended in fresh culture media containing rhIL-2 (1,000 IU/mL) without anti-CD3 mAb. Thereafter, half of the media were replaced by fresh media with cytokines every 2 days. When the cell density was high, cells were harvested and transferred to larger flasks.

To produce retrovirus, GP2–293 cells with a confluency of 70%–80% were transfected with a pCIR retrovirus vector expressing an anti-EGFR CAR cassette or with a corresponding empty control plasmid, using Lipofectamine 3000 Reagent (ThermoFisher), as previously described with some modifications (21). The EGFR-CAR cassette sequentially includes a signal peptide, the light chain of an anti-EGFR antibody, a linker, the heavy chain of an anti-EGFR antibody, a hinge, the CD28 costimulatory domain, and CD3ζ. Following the manufacturer's protocol, RetroNectin (Takara Bio) coating plates were used for 2-hour retroviral infection of the expanded NK cells described in the Expansion of NK cells section with 2,000 × g centrifugation. The infected cells were washed and cultured with rhIL-2 (1,000 IU/mL) for 48 hours before the assessment of infection efficiency by flow cytometry. The infected cells were cultured for a certain period before being frozen.

For measuring CAR activity, EGFR-CAR NK cells or empty vector–transduced NK cells or untransduced NK cells were incubated with tumor cells at an E/T ratio of 5:1 for 4 hours at 37°C in the presence of anti-CD107a antibody and 1 mg/mL of GolgiPlug (BD Biosciences). The cells were harvested, followed by being permeabilized, fixed, and stained with an anti-IFNγ antibody and an anti-TNFα antibody before subjected to a flow cytometric analysis. For testing whether the IL15/IL15-Rα complex secreted by OV-IL15C–infected GBM cells could improve anti-GBM activity of EGFR-CAR NK cells, these CAR NK cells were preincubated with 10× concentrated supernatants from OV-Q1- or OV-IL15C–infected U251 cells for 18 hours. The preincubated EGFR-CAR NK cells were then washed, counted, and co-cultured with GBM target cells (LN229 or U87vIII) at an E/T ratio of 5:1 for an additional 4 hours in the presence of anti-CD107a antibody and 1 mg/mL of GolgiPlug. The cocultured cells were harvested to assess CD107a degranulation and IFNγ as well as TNFα production, gated on CD56⁺ cells.

All animal experiments were approved by the City of Hope Institutional Animal Care and Use Committee. Mice were sacrificed when they became moribund with neurologic impairments or obvious weight loss (up to 20%). NOD/SCID/IL-2rg (NSG) and C57BL/6J mice were purchased from The Jackson Laboratory. Details are included in Supplementary Materials.

6–8-week-old male C57BL/6 mice were implanted with the murine CT2A cell line–expressing human EGFR (CT2A-hEGFR). Five days after the CT2A-hEGFR cells were implanted, mice were treated with 2 × 10⁵ pfu OV-IL15C alone, 1 × 10⁶ frozen and unsorted EGFR-CAR NK cells alone, the combination of the two agents, or saline alone. Three days after the treatment, mice were sacrificed to harvest brain tissues to isolate mononuclear cells, as previously described (20, 22). The mononuclear cells were subjected to assess NK and T-cell infiltration or cultured with PMA (BioLegend) and 1 mg/mL of GolgiPlug for 4 hours before assessing the activation capacity of these cytolytic cells by flow cytometry.

The CT2A-hEGFR mouse model was established and treated as described above. Mice were sacrificed to collect sera for cytokine release array assay using a Quantibody Mouse Inflammation Array 1 Kit (RayBiotech Life, Inc.).

Statistical analyses were performed using GraphPad software Prism v.8.0, R.3.3.1, or SAS 9.4. Values are presented as mean ± SEM or SD. Continuous endpoints were compared between two or more groups by a two-sample t test or one-way ANOVA model. The linear mixed model was used to account for the covariance structure due to repeated measures from the same donor. Survival functions were estimated by the Kaplan–Meier method and compared by the log-rank test. P values were adjusted for multiple comparisons by the Holm's procedure. A P value of <0.05 was considered statistically significant.

We successfully engineered an oncolytic herpes simplex virus (oHSV), referred to as OV-IL15C, which expresses a fusion protein containing a complex of human IL15 and an IL15Rα sushi domain. The genetic maps of wild-type human HSV-1, parental oHSV control OV-Q1, and OV-IL15C are shown in Fig. 1A. Immunoblotting analysis showed the IL15/IL15Rα complex tagged with HA was detected with an expected size of 32 kDa (Fig. 1B). ELISA results indicated that secretion of the IL15/IL15Rα complex occurred in a time-dependent manner with approximately 600, 300, 800, 1,000, and 1,200 pg/mL secreted by U251, LN229, GBM30, CT2A, and 293T cells, respectively, 96 hours after OV-IL15C infection, whereas the IL15/IL15Rα complex was undetectable in uninfected and OV-Q1–infected cells (Fig. 1C). To investigate the capacity of viral production by OV-IL15C, the GBM cell lines were infected with the OVs at different MOIs. The OV-IL15C–infected U251 and LN229 cells produced a similar amount of virus compared with the same cells infected with the control OV-Q1 under either unsaturated or saturated conditions (Fig. 1D), suggesting that OV-IL15C engineering does not affect the capacity for viral production.

Both NK and CD8⁺ T cells are cytolytic lymphocytes, and their activation by cytokines is critical for their antitumor activities. NK cells against GBM30 cells were found significantly increased in the presence of 10× concentrated supernatants from OV-IL15C–infected GBM cells compared with 10× concentrated supernatants from OV-Q1–infected GBM cells (Fig. 2A). Similarly, the killing ability of CD8⁺ T cells against GBM30 cells also significantly increased in the presence of enriched supernatants from OV-IL15C–infected GBM cells compared with enriched supernatants from OV-Q1–infected GBM cells (Fig. 2B). Furthermore, the enriched supernatants from OV-IL15C-infected GBM cells significantly prolonged the survival of NK and CD8⁺ T cells compared with enriched supernatants from OV-Q1–infected GBM cells, correlating to increased levels of phosphorylated STAT5 in both NK and CD8⁺ T cells (Fig. 2C–F; Supplementary Fig. S1A and S1B).

To evaluate the efficacy of OV-IL15C on GBM therapy in vivo, we established an orthotopic xenograft GBM mouse model by intracranially injecting luciferase-expressing GBM30 cells (GBM30-luc) into NOD/SCID/IL-2rg (NSG) mice on day 0. On days 5 and 12, mice received an intratumoral co-injection of OV-Q1 plus human CD8⁺ T cells or OV-IL15C plus human CD8⁺ T cells or saline on each day (Fig. 3A). The co-injected human CD8⁺ T were pre-activated by anti-CD3/anti-CD28 beads. The tumor growth of the mice was monitored by serial in vivo bioluminescence imaging (BLI). Mice treated with OV-IL15C plus human CD8⁺ T cells showed a significant reduction in tumor burden 8 (Fig. 3B and C) and 15 days (Fig. 3D and E) after tumor implantation and survived significantly longer (Fig. 3F), compared with mice treated with OV-Q1 plus human CD8⁺ T cells or saline. To test whether OV-IL15C also enhances GBM therapy in the presence of human NK cells, we re-established the GBM30-luc xenograft GBM model with an intratumoral co-injection of OV-Q1 plus human NK cells or OV-IL15C plus human NK cells, similarly as described above for co-injection of OV and human CD8⁺ T cells (Supplementary Fig. S2A). However, no significant difference was observed for tumor burden and mouse survival between the OV-Q1 plus human NK cells and the OV-IL15C plus human NK cells group (Supplementary Fig. S2B–S2D).

To improve NK-cell antitumor activity in immunotherapy, we generated CAR NK cells targeting wild-type EGFR and EGFRvIII GBM cells. For this, human peripheral blood NK cells were transduced with an EGFR-CAR retroviral vector (Fig. 4A), followed by further expansion. We were able to generate over 1 × 10¹¹ EGFR-CAR NK cells from the peripheral blood of an individual donor. Cells were frozen and ready to use, so called off-the-shelf, for both in vitro and in vivo experiments after thawing. We assessed cytotoxicity of EGFR-CAR NK cells against the LN229 (EGFR-positive cell line; ref. 23) or the EOL-1 (no EGFR or EGFRvIII expression; ref. 24) at different effector/target (E/T) ratios. Across all E/T ratios, EGFR-CAR NK cells exerted superior killing of LN229 cells compared with EV-transduced NK cells (Fig. 4B, left). The EGFR-CAR NK cells were equally efficient as EV-transduced NK cells in killing EOL-1 target cells (Fig. 4B, right). These data indicate that the enhanced killing of LN229 targets by the transduced cells is mediated by the CAR. Furthermore, we measured CD107a degranulation and IFNγ as well as TNFα secretion in response to LN229, U87vIII (EGFRvIII positive cell line; ref. 23), and EOL-1 target cell lines, and all these effector functions of EGFR-CAR NK cells were significantly increased against LN229 and U87vIII target cells compared with the EV-transduced NK cells, whereas the EGFR-CAR NK cells and EV-transduced NK cells showed similar effector function against the EOL-1 cells (Fig. 4C). Next, the luciferase-expressing U87vIII orthotopic GBM model was established as described previously for the GBM30-luc model, followed by treatment with EGFR-CAR NK cells or EV-transduced NK cells as depicted in Fig. 4D. Results showed that a single intracranial infusion of EGFR-CAR NK cells possessed significantly better anti-GBM activity and significantly improved survival of mice bearing GBM tumor, compared with EV-transduced NK or control groups (Fig. 4E–G).

The above results showed EGFR-CAR NK cells are superior to EV-transduced NK cells. We next tested whether the IL15/IL15-Rα complex secreted by OV-IL15C–infected GBM cells can improve EGFR-CAR NK-cell activity. For this purpose, after confirming that EGFR-CAR expression and activity remained after being frozen (Supplementary Fig. S3A and S3B), we thawed frozen EGFR-CAR NK cells and preincubated them with 10× concentrated supernatants from OV-Q1- or OV-IL15C–infected U251 cells, followed by co-culture with GBM target cells (LN229 or U87vIII). Results showed that the 10× concentrated supernatants containing the IL15/IL15Rα complex secreted from OV-IL15C–infected-GBM cells significantly increased or showed a trend of an increase of the secretion of IFNγ and TNFα, compared with similarly 10× concentrated supernatants from OV-Q1–infected GBM cells (Fig. 5A). To test whether the in vitro effects can be translated to in vivo, that is, whether the combination of EGFR-CAR NK cells with OV-IL15C is superior to its combination with OV-Q1, we used the orthotopic GBM30-luc xenograft model (Fig. 5B). Indeed, the combination of OV-IL15C and EGFR-CAR NK cells resulted in a more significant reduction in tumor burden (Fig. 5C and D) and prolonged mouse survival (Fig. 5E), compared with the combination of OV-Q1 and EGFR-CAR NK cells. These in vivo data provide a proof of concept for superior combination immunotherapy of OV-IL15C with CAR NK cells. Next, we investigated whether the combination of OV-IL15C and EGFR-CAR shows a synergistic effect. For this purpose, we re-established the orthotopic GBM30-luc xenograft model and treated it with or without the combination of OV-IL15C and EGFR-CAR NK cells of their monotherapies, as shown in Fig. 5F (4 groups in Fig. 5F–I vs. 3 groups in Fig. 5B–E). Luciferase-based in vivo BLI showed the mice that received either OV-IL15C alone, EGFR-CAR NK cells alone, or the combination of both had significantly inhibited tumor growth, compared with injection with saline control (Fig. 5G; Supplementary Fig. S3C). Importantly, the suppression of tumor growth showed a synergistic effect after treatment with the combination of OV-IL15C and EGFR-CAR NK cells compared with each monotherapy (Fig. 5H). The combination of the two agents also rendered mice to survive significantly longer than those in all other groups (Fig. 5I).

To investigate the effects of OV-IL15C on endogenous immune cells, we established an immunocompetent GBM model with an intracranial injection of the murine GBM cell line CT2A. Five days after tumor implantation, mice received an intracranial injection of OV-Q1, OV-IL15C, or saline. Three days after treatment, mice were sacrificed to check the infiltration of the immune cells in both the brain and spleen. Mice treated intracranially with OV-IL15C showed a significant infiltration of endogenous NK and T cells into the brain compared with those treated with OV-Q1 or saline (Fig. 6A); however, no significant difference between OV-IL15C and OV-Q1 was observed in the spleen (Fig. 6B). In both organs, both OVs significantly recruited more or showed a trend of increase of NK and T cells (Fig. 6A and B). Also, OV-IL15C treatment significantly prolonged mouse survival when compared with OV-Q1 or saline treatment (Fig. 6C). To evaluate the benefit of combining OV-IL15C with EGFR-CAR NK cells on immune responses, we modified the immunocompetent GBM model using the murine GBM cell line CT2A expressing human EGFR (CT2A-hEGFR). Five days after tumor implantation, mice were intratumorally treated with OV-IL15C, EGFR-CAR NK cells, the combination of the two agents, or saline. Three days after treatment, mice were sacrificed to check the immune cell infiltration and activation. Blood sera were collected to measure cytokine levels, as detailed below. We found that significantly more endogenous NK and CD8⁺ T cells infiltrated into the brain when mice were treated with the combination of OV-IL15C and EGFR-CAR NK cells versus their monotherapies, except for the combination versus OV-IL15C in T cells, which only showed a trend of an increase (Fig. 6D). The combination also resulted in significantly more TNFα secretion than EGFR-CAR NK cells rather than OV-IL15C in CD8⁺ T cells, whereas similar benefit was not observed in NK cells and both types of cytolytic cells did not show any benefit of the combination regarding IFNγ production (Fig. 6D). The combination of OV-IL15C and EGFR-CAR NK cells also significantly prolonged mice survival, compared with the other groups (Fig. 6E). To examine whether OV-IL15C can make EGFR-CAR NK cells persist longer without increasing exhaustion, we re-established the orthotopic CT2A-hEGFR immunocompetent model. Mice were treated with EGFR-CAR NK cells alone or in combination with OV-IL15C 5 days after tumor implantation. Twenty-four, 48, and 72 hours after treatment, we observed that EGFR-CAR NK cells persisted longer (72 vs. 48 hours) in the presence of OV-IL15C, compared with EGFR-CAR NK cells alone. On each of 3 days, more EGFR-CAR NK cells in terms of percentages and/or absolute cell numbers were detected in the brain in the OV-IL15C plus EGFR-CAR group versus the EGFR-CAR alone group (Fig. 6F). Also, compared with EGFR-CAR NK cells alone, mice treated with the combination of OV-IL15C plus EGFR-CAR NK cells did not show increased CAR NK cell exhaustion, marked by the expression of PD-1 and Tim-3 (Fig. 6G).

For evaluating the safety profile of OV-IL15C, an in vitro tropism change assay and an in vivo safety assay were performed. In vitro, no discernible change in tropism was observed following infection with either OV-Q1 or OV-IL15C in primary human oral fibroblasts (HOrF), primary human hepatic sinusoidal endothelial cells (HHSEC), or primary human pulmonary microvascular endothelial cells (HPMEC; Fig. 7A, top). We also quantified the effects of OV-Q1 and OV-IL15C on infection of these primary human cell lines, and no significant difference was observed. For HPMEC, it is hard to quantify the plaque number because these cells are highly susceptible to oHSV infection, and the formed plaques did not separate very well (Fig. 7A, bottom). To assess in vivo safety of OV-IL15C, BALB/c mice were intracranially injected with wild-type HSV-1 (as a negative control for safety), OV-Q1, or OV-IL15C at a high dose of 1 × 10⁶ pfu. Mice injected with wild-type HSV-1 died rapidly (less than 12 days), whereas all mice treated with OV-Q1 or OV-IL15C survived for several weeks (Fig. 7B). Because cytokine release syndrome (CRS) is common and major toxicity in CAR T cells, using a Quantibody Mouse Inflammation array, we measured levels of 37 cytokines in the sera of the mice treated with saline (control), OV-IL15C, EGFR-CAR NK cells, or the combination of OV-IL15C plus EGFR-CAR NK cells. Results showed that the levels of these cytokines, including IL6, IL10, IFNγ, and GMCSF, had no significant difference among the 4 treatment groups (Fig. 7C), suggesting that unlikely OV-IL15C, EGFR-CAR NK cells, and their combination will induce CRS in vivo.

GBM is the most common and aggressive primary malignant brain tumor in humans without curative therapy at its advanced stages (25). OV is a good approach to treat GBM as its local administration can induce immune infiltration and activation as well as direct lysis of GBM cells (26). OV can also serve as a cargo for local delivery of anti-GBM arsenals. We, therefore, engineered an OV-expressing IL15/IL15Rα (OV-IL15C) to further improve anti-GBM immune responses. As the IL15/IL15Rα complex is released locally due to locoregional administration of OV-IL15C, the virus may not increase too much systemic inflammation but locally produces strong antitumor immunity (27). Among OVs, oHSV has been approved by the FDA for cancer treatment and is the furthest along in the clinic (28). The OV-IL15C engineered for this study is based on HSV-1, and we demonstrated that it secretes a human IL15/IL15Rα complex and is safe in vivo. Accordingly, the complex secreted from OV-IL15C–infected GBM cells could prolong survival and activate both NK and CD8⁺ T cells in vitro. Consistent with this, the combination therapy of OV-IL15C with CD8⁺ T cells or EGFR-CAR NK cells significantly improves the therapeutic outcomes in the xenograft and/or immunocompetent GBM mouse models.

IL15 is an essential cytokine for NK and CD8⁺ T-cell development and function (29). Our group originally discovered that this cytokine is critical in regulating NK-cell survival (30). IL15 complexed with its high-affinity receptor alpha (IL15Rα) shows promising advantages over IL15 alone (31). The complex can significantly enhance the half-life and bioavailability of IL15 (5, 32, 33). The IL15/IL15Rα complex has a prolonged half-life time in serum (∼20 hours) compared with IL15 (∼1 hour; refs. 32, 34, 35). This might be more critical in the brain, as historically the brain was recognized as an immune-privileged site (36). In some in vivo mouse models, the IL15/IL15Rα complex could cause a rapid and significant regression of GBM and melanoma; however, IL15 alone could not (33, 37). On the basis of such advantages, engineered cells expressing IL15/IL15Rα complex for preclinical and clinical studies have been explored (9). Tumor cells ectopically expressing IL15/IL15Rα improve NK-cell– and CD8⁺ T-cell–mediated in vivo tumor lysis (38). Intratumoral vaccination of an adenovirus vector–expressing IL15/IL15Rα inhibits murine breast and prostate cancers (8). Furthermore, CD8⁺ T cells or dendritic cells transfected with IL15/IL15Rα enhance the antitumor response in vivo (39, 40). A previous study showed that engineered an oncolytic vaccinia virus–expressing the murine IL15/IL15Rα complex elicits potent antitumor immunity (41). Thus, OV-IL15C alone or its combination with another effective agent may maximize the activity of IL15 and improve oncolytic virotherapy for GBM. However, using an oHSV to express human IL15/IL15Rα and the study of how it alone or its combination with another effective therapy modulates the immune response in the tumor microenvironment and the subsequent therapeutic efficacy have not been explored.

CD8⁺ T cells play an important role in regulating the antitumor immune response. Our results showed that the IL15/IL15Rα complex secreted by OV-IL15C–infected GBM cells could significantly enhance cytotoxicity and survival of CD8⁺ T cells in vitro. We believe that this enhanced cytotoxicity of CD8⁺ T cells should be antigen non-specific. A previous study demonstrated that IL15–treated healthy donor T cells kill over 50% of target cells (the P815 cell line), whereas untreated T cells only kill about 5% of the same target cells (42). Our in vivo xenograft GBM model showed a better therapeutic outcome of OV-IL15C plus CD8⁺ T cells compared with OV-Q1 plus CD8⁺ T cells. T-cell responses belong to the adaptive immune response, which need time to be ready for fighting tumor cells. In contrast, NK cells belong to the innate immune response and can quickly respond to target cells, including tumor cells, without prior activation. In vitro, we observed that the IL15/IL15Rα complex secreted by OV-IL15C–infected GBM cells enhances anti-GBM activity of both CD8⁺ T cells and NK cells, whereas in vivo we only observed improved survival by OV-IL15C when combined with CD8⁺ T cells instead of NK cells. One reason to explain these results may be due to the fact that primary human NK cells have a limited survival period and/or effector function in the brain, especially when they are not armed with a CAR targeting a tumor-associated antigen.

Engineering NK cells with a CAR to effectively treat cancer is necessary, as shown in several preclinical cancer models, including GBM (43–45). However, using NK cells to treat patients with solid tumors remains very limited, in large part due to the inability of NK cells to traffic into tumor tissue as well as the immunosuppressive microenvironment of the tumor. Our previous study demonstrated the efficacy and safety of an intracranial injection of EGFR-CAR–modified NK-92 cells in a GBM orthotopic xenograft model (46). However, NK-92 cells were derived from a lymphoma cell line, it should be irradiated before their infusion into patients, and the irradiation may cause the loss of some antitumor activity (47). In the current study, we used primary human NK cells obtained from PBMCs to engineer EGFR-CAR NK cells. We show that EGFR-CAR NK cells derived from peripheral blood and manufactured with expansion by an autologous PBMC condition successfully recognize the EGFR antigen on GBM cells and lead to enhanced anti-GBM activity both in vitro and in vivo. We were able to generate up to 1 × 10¹¹ frozen and ready to use, so called “off-the-shelf” EGFR-CAR products under good laboratory practices-like conditions for all in vitro and in vivo studies, with maintained EGFR-CAR expression as well as good effector function over 6 months after being frozen.

A recent study suggested that IL15 preserves the CAR T-cell Tscm (T-stem cell memory phenotype) and improves their metabolic fitness (48). However, there are at least two advantages for combining OV-IL15C with CAR NK cells instead of CAR T cells: (i) CAR NK cells can be allogeneic, off-the-shelf, and thus the cost for manufacturing will be lower compared with autologous CAR T cells; (ii) one of the major concerns with CAR T-cell therapy is near-lethal or lethal toxicity, such as CRS and inflammatory encephalomyelitis. Without data from the clinic, we are unsure whether the IL15/IL15Rα complex produced by our OV will worsen CRS of CAR T cells, considering that OV itself may potentially further activate immune cells above the level of CRS. The addition of IL15 may further exacerbate inflammatory or toxic conditions for patients. In contrast, NK cells do not have “memory” in the traditional sense of T cells. Therefore, the likelihood of massive clonal expansion upon exposure or re-exposure to tumor-associated antigens is extremely small, likely making CAR NK-cell therapy much less toxic and better tolerated than CAR T-cell therapy. The recent clinical trial showed that the infusion of anti-CD19 CAR NK cells expressing IL15 did not induce an increase of the levels of IL6 or CRS in 11 patients with lymphoid cancer (18). Our in vivo safety data demonstrated that there was no significant difference in levels of 37 cytokines, including IL6, among combination therapy of OV-IL15C with EGFR-CAR NK cells or their monotherapies.

GBM is a very heterogeneous cancer. Even in the same individual patient, some tumor cells express wild-type EGFR, some express EGFRvIII, some express both genes, and some do not express either (49). EGFR-CAR NK cells directly target EGFR and EGFRvIII-positive tumor cells, whereas OV has selectivity for tumor cells versus normal cells and can, therefore, kill tumor cells lacking EGFR and EGFRvIII expressions. EGFR-CAR NK cells and OV both can launch endogenous immune responses to tumor cells (20, 21, 50–52). OV may also enhance intracranial infiltration of EGFR-CAR NK cells that are systemically intravenous administered, as we previously showed for endogenous NK cells in our animal models (53). Furthermore, EGFR-CAR NK cells persist longer in the presence of OV-IL15C compared with EGFR-CAR NK cells alone without showing exhaustion in mice, resulting in an enhanced anti-GBM activity. All of these effects converge to transform a “cold” TME with few immune effector cells into a “hot” TME with increased immune cells and provide a strong rationale to combine OV, or OV-IL15C, with EGFR-CAR NK cells to target heterogeneous GBM.

In summary, we developed an innovative and promising method by intracranial co-injection of OV-IL15C with off-the-shelf EGFR-CAR–modified primary human NK cells to target GBM. Our current results provide strong experimental proof applicable for future clinical application.

M.A. Caligiuri reports grants from NIH during the conduct of the study; as well as personal fees from Cytovia, OncoC4, Imugene, and CBMG outside the submitted work; and a patent for EGFR-CAR issued and licensed to CytoImmune Therapeutics and a patent for OV-IL15C pending. J. Yu reports grants from NIH during the conduct of the study; as well as other support from CytoImmune, Inc. outside the submitted work; and a patent for EGFR-CAR issued and licensed to CytoImmune, Inc., and a patent for OV-IL15C pending. No disclosures were reported by the other authors.

R. Ma: Investigation, methodology, writing–original draft. T. Lu: Methodology. Z. Li: Methodology. K.Y. Teng: Methodology. A.G. Mansour: Methodology. M. Yu: Methodology. L. Tian: Methodology. B. Xu: Methodology. S. Ma: Methodology. J. Zhang: Data curation. T. Barr: Writing–review and editing. Y. Peng: Supervision. M.A. Caligiuri: Supervision, writing–review and editing. J. Yu: Supervision, writing–review and editing.

This work was supported by grants from the NIH (NS106170, AI129582, CA247550, and CA223400 to J. Yu; CA210087, CA068458, and CA163205 to M.A. Caligiuri), the Leukemia and Lymphoma Society (1364-19), and The California Institute for Regenerative Medicine (DISC2COVID19-11947 to J. Yu).

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.