The diploid gene γ34.5 (RL1) encodes a multifunctional tegument protein that is the most common gene mutated or deleted in oHSVs targeting brain tumors, including all oHSVs in GBM clinical trials ( Table 1 ; Figure 2 ). This is because γ34.5 is the major HSV gene driving neuropathogenicity (Peters and Rabkin, 2015). Among the other activities of γ34.5 are: (i) overcoming PKR-induced block to host protein shutoff, whereby its GADD34 homology domain promotes dephosphorylation of p-eIF2α and restoration of translation; (ii) blocking autophagy by binding beclin 1; and (iii) inhibition of RIG-I, TBK1, and cGAS/STING signaling that blocks IRF3 activation and type 1 interferon (IFN) signaling (Dogrammatzis et al., 2020; Kangas et al., 2021). Most tumors suppress antiviral innate immune responses (translation control and IFN responses) due to their anti-proliferative, pro-apoptotic, and immune activities (Du et al., 2017; Castiello et al., 2018; Boehmer et al., 2021; Musella et al., 2021; Solomon et al., 2022). Thus, γ34.5 deletions provide cancer selectivity, as well as significant safety in the brain. This was demonstrated with first-generation oHSVs, containing deletions of both copies of γ34.5 in different HSV parental strains, HSV1716 in strain 17 and R3616 in strain F, which were replication competent in human glioma and other cancer cell lines, and xenografts in immunodeficient mice (Jahan et al., 2021).

HSV1716 (Seprehvir; Sorrento Therapeutics) was the first oHSV to enter clinical trial in Europe, with 3 trials in GBM patients ( Table 1 ). The first study enrolled 9 patients with recurrent high-grade gliomas (HGG) who received escalating intratumoral doses of 10³-10⁵ pfu in 1 ml. Four of nine patients survived longer than 14 months, and no evidence of virus shedding or reactivation was detected, demonstrating the safety and feasibility of HSV1716 (Rampling et al., 2000). In the second study, HGG patients had their tumors resected 4-9 days after intratumoral treatment to evaluate virus replication. Infectious virus, more than input, was recovered at the injection site in two of twelve patients, both being seronegative and seroconverted, while virus DNA was detected in 10 patients with 4 being at distal sites (Papanastassiou et al., 2002). Finally, in a third study, virus (10⁵ pfu in 1 ml) was injected into the rim of the resection cavity after maximal surgical resection, with 3 of 12 patients alive and stable after 15-22 months (Harrow et al., 2004). None of the patients in any of the 3 clinical trials experienced adverse events attributable to the virus (Harrow et al., 2004). There was concern about the safety of oHSVs with only a single genetic alteration, even deletion of γ34.5, which is a reason why the HSV1716 clinical trials dose escalated only up to 10⁵ pfu. This prompted the development of multimutated second-generation oHSVs, of which G207 is the exemplar.

HSV encodes a number of proteins involved in nucleotide metabolism in order to replicate in non-cycling post-mitotic cells. Mutations in these genes confer specificity for dividing cells (ie., cancer cells), which express cellular nucleotide metabolism enzymes, and often also attenuate pathogenicity, making them important mutations for oHSV. In addition to TK, mutations in uracil DNA glycosylase (UNG or UL2) and ICP6 (UL39) have been used to construct oHSVs (Pyles et al., 1997; Peters and Rabkin, 2015). ICP6 is the large subunit of the viral ribonucleotide reductase (RR), an enzyme that converts ribonucleotides (ribonucleoside diphosphate, NDPs) into deoxyribonucleotides (dNDPs). RR activity is encoded in the C-terminus and is required for viral DNA replication (Peters and Rabkin, 2015). The N-terminal RHIM domain binds RIP3 to trigger necroptosis (Wang et al., 2014). In contrast to TK mutants, ICP6 and UNG mutants are hypersensitive to nucleoside analog drugs (Mineta et al., 1994; Pyles et al., 1997). ICP6 mutants are attenuated in pathogenicity, including in the brain. An ICP6 mutated oHSV, hrR3, with an E. coli LacZ insertion inactivating ICP6, and forming a fusion protein, was efficacious in inhibiting the growth of orthotopic glioma xenografts and spreading in the tumors (staining for LacZ) (Mineta et al., 1994). LacZ provides a reporter gene, to easily evaluate virus spread, and as a unique identifier for oHSV, enabling discrimination with patient clinical isolates during clinical trials. Because of these properties, ICP6 mutations have been combined with other mutations to enhance oHSV safety, especially in the brain, and glioma specificity; including, 1716-6 (γ34.5Δ ICP6^-) from HSV1716 (γ34.5Δ), MG18L (Us3^- ICP6^-) from R7041 (Us3^-), and Δ68H-6 (γ34.5-BBDΔ ICP6^-) from Δ68H (γ34.5-BBDΔ) (Peters and Rabkin, 2015).

G207 ( Table 1 ; Figure 2 ) was constructed with the goal of clinical translation to patients with GBM. Thus, safety was of paramount importance. It was derived from R3616 (γ34.5 deleted (Δ)) with an inactivating insertion of LacZ into the ICP6 gene, as in hrR3 (Mineta et al., 1995). From a safety point of view, it has a number of important elements: (i) two pathogenicity genes mutated/deleted that are broadly spaced in the genome making recombination/reversion highly unlikely; (ii) hypersensitivity to anti-viral nucleoside analog drugs; (iii) temperature-sensitivity to compromise replication under fever conditions; (iv) selectively grows in and kills glioma cells and not astrocytes or neurons; and (v) is safe after intracerebral injection in mice and non-human primates, as opposed to wild-type HSV at a 10⁴ lower dose (Mineta et al., 1995). In addition, preclinical studies in immunocompetent mouse models provided the first demonstration of OV-induced in situ cancer vaccination (Toda et al., 1999), including in mice with both subcutaneous and intracerebral tumors, where G207 treatment of the subcutaneous tumor inhibited the growth of the intracerebral tumor and provided protection to tumor rechallenge in the brain (Todo et al., 1999; Jahan et al., 2021).

G207 was the first oHSV to enter clinical trial in the US in a dose-escalation (10⁶ - 3 x 10⁹ pfu) phase I study for recurrent HGG (Markert et al., 2000). G207 was administered stereotactically to a single site in a contrast-enhancing region, except for the final dose. There were no dose-limiting toxicities attributable to G207, and responses were not dose-dependent and seen at low doses. For example, the longest surviving GBM patient (>17 months from treatment) received a dose of 3 x 10⁷ pfu (Markert et al., 2000). In a follow-on phase Ib trial, G207 was inoculated via intratumoral catheter (1.5 x 10⁸ pfu) followed by tumor resection 2-5 days later and virus injection (1 x 10⁹ pfu) adjacent to the resection cavity (Markert et al., 2009). Four of six patients displayed increased CD3+ cells (IHC) post-G207, and viral DNA was detected in all patients’ tumors by PCR. Two patients had HSV-1 detected in their saliva at 7 days post-resection, however, these were demonstrated to be wild-type virus based on negative PCR for LacZ, while no spread of G207 was detected in serum, conjunctival swabs, or saliva (Markert et al., 2009). The final clinical trial with G207 for adult GBM was in combination with radiation (5 Gy) 24 hr post-G207 (Markert et al., 2014). In this trial, 5 of 8 patients were seronegative, much lower than in the 2 previous trials at 32%, and only 1 seroconverted (Markert et al., 2014). Like the two prior trials, treatment was well tolerated, and no patients developed encephalitis or serious adverse events attributed to the virus, G207 was not detected in the saliva or serum, and some patients had significant responses (Markert et al., 2014).

A recently completed phase I clinical trial of G207 in pediatric HGG has reported promising results, including a median OS of 12.2 versus 5.6 months in historical controls (Friedman et al., 2021). All 12 patients were IDH1 wild-type without favorable H3.3 mutations. The trial included 4 groups, a single administration of 10⁷ or 10⁸ pfu via slow infusion from 1-4 catheters with or without 5 Gy of radiation. G207 was not associated with any grade 2-4 adverse events or evidence of peripheral shedding (saliva). One patient with a saliva-positive HSV PCR result was LacZ negative demonstrating reactivation of a clinical strain (Friedman et al., 2021). Seroconversion occurred in 3 of 9 seronegative patients, all at the 10⁸ pfu dose. Intriguingly, the median OS was 5.1 months for 3 seropositive patients and 18.3 months for those that seroconverted. Significant increases in tumor-infiltrating lymphocytes, both CD4+ and CD8+ cells, were detected in patient biopsies 2-9 months post-treatment, while no HSV-1 staining was observed (Friedman et al., 2021).

While the γ34.5 deletion has a large impact on attenuating pathogenicity and replication in normal cells, it also attenuates virus replication and cell killing in cancer cell lines (Todo et al., 2001; Shah et al., 2007; Kanai et al., 2012b). Once human GSCs were identified and isolated, it was possible to ask whether they were permissive to γ34.5Δ oHSV replication. Unfortunately, γ34.5Δ oHSVs (1716, R3616, G207) replicate poorly, if at all, in patient-derived GSCs in vitro that have been tested (Wakimoto et al., 2009; Kanai et al., 2012b; Peters et al., 2018). However, they do replicate well in patient-matched serum-cultured GBM cells (ScGCs), similar to glioma cell lines, which were used in the preclinical studies (Peters et al., 2018). This is due to a block in true late viral gene translation in GSCs (Peters et al., 2018). Virus yield of γ34.5Δ C101 (parent of C134 ( Figure 2 )) in 3 human patient-derived xenograft (PDX)-derived GSC xenolines was 10² - 10⁴-fold lower than with wild-type HSV, and somewhat decreased or not in hypoxia where CD133+ cells were greatly increased (Friedman et al., 2012). In a comparison of stem-like cells from patient-derived pediatric high-grade brain tumors with adult GBM, pediatric brain tumor xenograft stem-like cells were ≥ 6-fold more sensitive to G207 cytotoxicity than cells from adult tumors, having a mean sensitivity at MOI~3 after 3 days of only 46% (Friedman et al., 2018). Pediatric tumor GSCs expressed significantly higher levels of HSV entry mediator nectin-1 (CD111) than adult GSCs, which inversely correlated with IC50 (Friedman et al., 2018). In one pediatric embryonal supratentorial primitive neuroectodermal tumor (sPNET) stem-like cell-derived intracranial xenograft model, G207 significantly extended survival, whereas, in 3 different GSC-derived intracranial tumor models, irrespective of nectin-1 expression, G207 extended survival, but not significantly (Friedman et al., 2018).

This limitation of γ34.5Δ oHSVs led to strategies to overcome the attenuated virus replication without significantly increasing pathogenicity, including second-site suppressors (G47Δ), targeted mutations in γ34.5 (Δ68H-6), and transcriptionally-targeted γ34.5 expression (rQNestin34.5). Initially, an in vitro screen for second-site suppressor mutations of γ34.5Δ virus replication in non-permissive cancer cells was performed. This identified deletions of ICP47 (Us12, α47) and the Us11 late promoter (Mohr and Gluzman, 1996). This deletion placed the late Us11 gene under control of the Us12 IE promoter, allowing US11 to be expressed as an early instead of late protein (Mulvey et al., 1999). Us11 binds to double-stranded RNA, blocking PKR activation and eIF2α phosphorylation, thus complementing the lack of γ34.5 (Peters and Rabkin, 2015). ICP47 binds to transporter associated with antigen processing (TAP) to prevent MHC I expression, shielding virus-infected cells from CD8+ T cells (Todo et al., 2001; Dogrammatzis et al., 2020). Unfortunately, ICP47 activity is species-specific and minimally active in rodents (Ahn et al., 1996), so in mice, HSV-1 doesn’t inhibit TAP, and ICP47Δ oHSV behaves like ICP47⁺ in humans making it difficult to study the effects of ICP47Δ. This could be somewhat circumvented by using ‘humanized’ mice, immunodeficient mice with human immune cells.

G47Δ (Delytact®, Teserpaturev; Daiichi Sankyo) ( Table 1 and Figure 2 ) is a third-generation oHSV engineered from G207 by deletion of ICP47 and the Us11 promoter (Todo et al., 2001). It replicates in and kills glioma cells, and importantly GSCs, and inhibits tumor growth more effectively than G207 (Wakimoto et al., 2009). Intracerebral injection of G47Δ in A/J mice was as safe as with G207 (Todo et al., 2001). There have been two clinical trials of G47Δ for recurrent HGG in Japan. A phase I/II clinical trial of 13 patients with progressive or recurrent GBM at the University of Tokyo where 3x10⁸ pfu (3 patients) or 1x10⁹ pfu was injected two times (5-14 days apart) at the same coordinates (UMIN000002661) (Todo et al., 2022a). The median overall survival was 30.5 months from initial diagnosis and the 1-year survival rate from the last G47Δ administration was 38.5%, with 3 patients surviving more than 46 months; none were IDH1 mutants (Todo et al., 2022a). On MRI, all patients typically showed clearing at the injection site and enhancing lesion enlargement (immunoprogression) after the first G47Δ injection. HSV staining was negative at autopsy, while an influx of CD4+ and CD8+ cells were seen in biopsies at the second virus injection and persisted at autopsy (Todo et al., 2022a). Seroconversion occurred one week after G47Δ in all four seronegative patients. The most common G47Δ-related adverse events were headache, fever, or vomiting in 12 of the patients, with no evidence of virus shedding (blood, urine, saliva) within two weeks of treatment (Todo et al., 2022a). This first study demonstrated the safety and potential therapeutic activity of G47Δ for GBM.

In the registration single-arm phase II trial with 19 patients with recurrent GBM, G47Δ (1x10⁹ pfu in 1 ml) was stereotactically injected into 1-3 sites up to 6 times, at intervals of 5-14 days for doses 1 and 2, and 2-6 weeks for the remainder (Todo et al., 2022b). The primary endpoint of 1-year survival was met by 84% of patients, leading to a conditional approval of G47Δ for the treatment of malignant glioma in Japan, the first OV approved in Japan (Daiichi-Sankyo, 2021). Median overall survival was 28.8 months from the first surgery/diagnosis, with 3 patients alive over 3 years from the last viral dose. Interestingly, the median OS was virtually the same in the 6 IDH1 mutant patients as in the IDH1 wildtype patients (Todo et al., 2022b). These results were better than the 20.9 months seen in the randomized phase III clinical trial of tumor-treating fields and TMZ maintenance in patients who had completed standard therapy (Stupp et al., 2017). The magnetic resonance imaging (MRI) observations were informative and consistent with other oHSV trials for glioma; contrast-enhancement clearing at the injection site, some enlargement of target lesions right after virus injection (immunoprogression), and reductions in tumor size that often took over 9 months (Todo et al., 2022b). An important feature of this trial was the acquisition of tumor biopsies before each virus injection, which were used to evaluate histology and T cell infiltration. CD4+ T cells tended to infiltrate at the 2^nd injection, a little sooner than CD8+, but both increased with later injections and persisted after treatment ceased for up to 50 months, while the number of Foxp3+ T regulatory cells (Tregs) was very low throughout (Todo et al., 2022b). Despite this, tumor shrinkage took 4 months or more to detect. The safety profile was good, with only 2 grade 3 events (fever and vomiting) attributable to the virus and frequent low-grade fever and headache. A survival benefit without remarkable tumor responses (ORR of only 5.3%), delayed time to treatment response by MRI after an initial inflammatory response, and an early influx of T cells may be characteristics of oHSV therapy in GBM.

Δ68H-6 ( Figure 2 ; Table 2 ) has a deletion in the γ34.5 beclin 1 binding domain (BBD; 68 to 87 aa) and LacZ inactivated ICP6 (Kanai et al., 2012b). Amongst other activities, γ34.5 inhibits autophagy by binding to beclin 1 through the BBD, which modestly reduces neurovirulence (Talloczy et al., 2006; Orvedahl et al., 2007). Δ68H-6 was not as effective at inhibiting p-eIF2α as its BBD rescued virus (Δ68HR-6), but better than γ34.5Δ oHSV (Kanai et al., 2012b). However, it replicated in and killed GSCs as well as γ34.5+ oHSVs, and inhibited the growth of GSC-derived orthotopic tumors, while causing only transient and minimal neurologic symptoms (Kanai et al., 2012b).

Alternate strategies to overcome γ34.5Δ attenuation include expressing a mammalian or viral orthologue of γ34.5. C134 ( Figure 2 ; Table 1 ) has the HCMV IRS1 gene, which prevents PKR activation, under control of the HCMV IE promoter inserted into the UL3-UL4 intragenic region of γ34.5Δ R3616 (Cassady, 2005). C134 infection of glioma cell lines restores late viral protein synthesis and replication close to levels of wild-type HSV-1 (Shah et al., 2007), such that C134 replicated much better in GSCs than C101 (γ34.5Δ) in vitro (Friedman et al., 2015). At a low dose, C134 was significantly better at extending the survival of mice with intracranial glioma cell line xenografts than C101, which was not significantly better than saline (Shah et al., 2007). C134 was found to be safe after intracerebral injections of HSV-1 susceptible mice and non-human primates (1 x 10⁷ pfu) (Cassady et al., 2017), and a phase I clinical trial for recurrent GBM patients is now recruiting (NCT03657576).

rQnestin34.5v2 ( Figure 2 ; Table 1 ) has a copy of γ34.5 under the control of the nestin promoter inserted into the ICP6 locus of a γ34.5Δ virus (Kambara et al., 2005). Nestin is upregulated in neural progenitor cells and glioma and is a stem cell marker for GSCs (Ludwig and Kornblum, 2017; Xie et al., 2021). rQnestin34.5v2 replicated better and was more cytotoxic than parental rHsvQ1 (γ34.5Δ) in glioma cell lines and GSCs while retaining non-permissivity in normal cells (Kambara et al., 2005; Chiocca et al., 2020). In vivo, rQnestin34.5v2 treatment doubled the survival of athymic mice bearing U87 glioma tumors compared to the control virus (rHsvQ1) (Kambara et al., 2005). Moreover, rQnestin34.5v2 was not toxic, with no adverse effects detected in mice (Chiocca et al., 2020). A phase I clinical trial was approved in 2017 to test the safety of the virus in humans and to define the correct dose of virus to be administrated (NCT03152318). NG34 is similar to rQnestin34.5, except the nestin promoter drives expression of human GADD34, which is homologous to the carboxy-terminal PP1 domain of γ34.5 (Nakashima et al., 2018) ( Table 2 ). It is as cytotoxic as rQnestin34.5 in vitro to GSCs and in vivo at inhibiting GSC-derived tumor growth, and more cytotoxic than parental rHsvQ1 (Nakashima et al., 2018). Interestingly, it is less pathogenic than rQnestin34.5 after intracerebral injection in BALB/c and athymic mice, suggesting γ34.5 has additional pathogenicity activities (Nakashima et al., 2018).

Another way to overcome γ34.5Δ oHSV attenuated replication in cancer cells and especially its inability in GSCs, is to introduce mutations in other genes endowing oncolytic activity in the presence of γ34.5. MG18L ( Figure 2 ; Table 2 ) contains a deletion of the US3 gene, as well as the LacZ inactivation of ICP6 (Kanai et al., 2011). U_S3 is a non-essential gene encoding a serine-threonine kinase that blocks apoptosis and Akt activation (Leopardi et al., 1997). Apoptosis-related pathways are altered/dysfunctional in 80% of GBMs (Parsons et al., 2008). Thus, MG18L provides tumor specificity by enhancing apoptosis in normal cells, blocking virus replication and spread, and also to some extent in GSCs (Kanai et al., 2011). The ED50 of MG18L is about 4-fold lower in GSCs, similar to G47Δ, than in astrocytes. In addition, MG18L induces PI3K/AKT signaling, so that it synergizes with PI3K inhibitors in most GSCs, an interaction not induced by G47Δ (Kanai et al., 2011). Intracerebral injection of 4 x 10⁶ pfu of MG18L caused only minor and transient neurologic deficits in 25% of mice, illustrating its safety, similar to G47Δ (Kanai et al., 2011). In a GSC-derived xenograft model, a single intratumoral injection of MG18L significantly extended survival by 25% and increased apoptosis in the tumor, while the addition of a PI3K/AKT inhibitor resulted in 50% long-term survivors (Kanai et al., 2011). MG18L also exhibits either synergy or additivity with TGFβ receptor inhibitor and PARP inhibitors in GSCs in vitro and in vivo (Esaki et al., 2017; Ning et al., 2017).

An alternate strategy to endow HSV with oncolytic activity and patient safety is to retarget oHSV to only infect specific cancer cell types by detargeting its natural receptors and targeting cancer-specific cell surface molecules (Campadelli-Fiume et al., 2016; Goins et al., 2016). If this targeting is specific enough and limited to cancer cells, there should be no need to mutate virus genes for cancer selectivity and safety. HSV entry occurs in a multiple step process in which essential glycoproteins gB, gD and gH/gL are required, with gD interacting with HSV receptors HVEM and Nectin-1 (Goins et al., 2016). Therefore, most oHSV retargeting strategies are based on modifying gD, however, ligand insertions into gB and gH have also been reported (Campadelli-Fiume et al., 2016; Goins et al., 2016). Nectin-1 expression (CD111; HSV receptor) ranged from 4-76% of adult GBMs and inversely correlated with the IC50s of G207 in adult GBM, including PDX-derived GSCs, and pediatric brain tumors (Friedman et al., 2018). This provides a rationale for targeting other non-HSV receptor cell surface molecules in GBM. Conversely, it also illustrates a major downside to targeting specific cell surface molecules; the inherent heterogeneity of their expression.

Epidermal growth factor receptor (EGFR) is overexpressed in 60% of primary GBMs (Hersh et al., 2022). KNE ( Table 2 ) was retargeted to human EGFR- and EGFRvIII-overexpressing cells by inserting an scFv against EGFR into gD, which was also mutated for the HVEM and nectin-1 binding sites, and with gB mutations that enhance entry (Uchida et al., 2013). KNE infection was selective for EGFR-overexpressing cells in vitro and inhibited GSC-derived tumor growth in immunodeficient mice (Uchida et al., 2013). R-LM113 ( Figure 2 ; Table 2 ) targets human HER2 (EGFR2, erbB2) (Menotti et al., 2008), which is expressed in a majority of GBMs, as well as GSCs (Ahmed et al., 2010). It efficiently infected hGSCs and was efficacious in orthotopic glioma implant models with hHER2+ murine glioma cells in immunodeficient and immunocompetent mouse models (Gambini et al., 2012; Alessandrini et al., 2019). R-613, similar to R-LM113 except with an scFv to EGFRvIII inserted into gD, effectively infected hGSCs expressing EGFRvIII but not those without (Appolloni et al., 2021) ( Table 2 ). CXCR4+ cells were targeted using a nanobody sequence against human CXCR4 inserted into gD (oHSV/Nb-gD), as in KNE (Sanchez Gil et al., 2022). oHSV/Nb-gD was constructed on a G47Δ-like parent (γ34.5Δ, ICP6^-, and ICP47Δ) ( Table 2 ). CXCR4 is overexpressed in GBM and GSCs, and correlates with tumor size, progression, and recurrence, while the CXCL12/CXCR4 pathway is associated with GBM cell migration. oHSV/Nb-gD demonstrated efficacy in orthotopic xenografts of GSCs and U87MG.CXCR4+ cells (Sanchez Gil et al., 2022). Unfortunately, scFv’s and nanobodies against human cell surface molecules are usually species-specific, so proper toxicity testing of retargeted viruses for off-cancer target effects on normal cells cannot be evaluated in mice.

Another means to decrease pathogenicity and replication selectively in normal cells is the use of miR target sequences to repress virus gene expression in normal cells that express that miR. KG4:T124 is KG with an insertion of target sequences to miR-124, which is expressed in healthy neurons but not in GBM cells (Gaur et al., 2007), into the 3’UTR of ICP4 to block virus replication in the brain, but not in GSCs, and is also deleted for the IR/joint region and ICP47, and contains mutant gB (Mazzacurati et al., 2015). Other miRs that are selective for the brain include miR-128, -137, -219a, and -204, which have been incorporated into ONCR-159 (Kennedy et al., 2020). KGE4:T124 is KG4:T124 retargeted to human EGFR, as in KNE, so that it only infects human EGFR-overexpressing cells, and its replication is restricted to cells that do not express miR-124 (Mazzacurati et al., 2015) ( Table 2 ). It replicated as well as parental KG in hGSCs in vitro and inhibited GSC-derived tumor growth in vivo. In contrast to KG, KG4:T124 was non-toxic after intracranial injection in mice (Mazzacurati et al., 2015).

IL-12, a heterodimeric master regulator of cell-mediated immunity and an angiogenesis inhibitor, is the most potent anti-tumor cytokine expressed from oHSV (Nguyen et al., 2020). IL-12 is too toxic when delivered systemically, or in transduced tumor-infiltrating lymphocytes (TILs) (Conlon et al., 2019), thus, local expression, as with oHSV, is critical. G47Δ-mIL12 expresses murine IL-12 from the ICP6 locus of G47Δ (Cheema et al., 2013) ( Table 2 ). The anti-angiogenic activity of IL-12 was examined in vitro, where conditioned media from G47Δ-mIL12 infected glioma cells or GSCs inhibited endothelial tube formation, and in vivo where it extended survival of hGSC-derived tumor-bearing mice, coincident with decreased neovasculature and VEGF expression (Zhang et al., 2013). There is a dearth of non-immunogenic mouse glioma cell lines for interrogation of immunotherapy in mouse orthotopic implant models that are representative of human GBM. Mouse 005 cells were the first murine GSC model developed (Marumoto et al., 2009). 005 mGSCs form lethal non-immunogenic orthotopic tumors that resemble human GBM in C57BL/6 mice, which present with intratumoral heterogeneity, invasiveness, neovascularity, and immunosuppressive TME (Cheema et al., 2013). This rigorous GBM model was used to evaluate the immunovirotherapeutic efficacy of G47Δ-mIL12. Intratumoral injection of G47Δ-mIL12 in 005-derived brain tumors induced a significant survival increase compared to unarmed G47Δ. This increased survival was associated with a significant reduction of tumor cells, Treg’s, and vascularity, and was dependent on T cells but not NK cells (Cheema et al., 2013).

M002 and M032 ( Figure 2 ; Table 1 ) express mouse or human IL-12 respectively, from the Egr-1 promoter in a γ34.5Δ HSV (Parker et al., 2000). In vivo, M002 was more efficacious than its parent non-expressing R3659 in the murine syngeneic 4C8 glioma model and safe after intracerebral injection into the brains of Aotus nancymae primates (Markert et al., 2012). Human pediatric medulloblastoma cancer stem cells (CSCs) express high levels of nectin-1 and CD133 (Friedman et al., 2016). M002 was cytotoxic to medulloblastoma CSCs (LD50 ~ 0.5 MOI) in vitro and extended survival in mice with intracranial tumors to a similar extent as G207 (Friedman et al., 2016). Based on the promising M002 results, M032 was constructed for clinical translation (Roth et al., 2014). An extensive toxicology study was performed in Aotus nancymae primates, where it was found to be safe at 1x10⁶ pfu and 1 of 4 animals becoming moribund at the highest dose (1x10⁸ pfu) (Roth et al., 2014). Elevated white blood cells and neutrophils were observed on day 3 and then declined to baseline. Viral DNA was present in the brain, highest at the injection site, and decreased over time but still present 91 days after injection (Roth et al., 2014). Interestingly, a phase I clinical trial in dogs with glioma was performed before the human clinical trial (Omar et al., 2021). While HSV-1 is not pathogenic in dogs, canine glioma cells are similarly susceptible to killing as mouse cells. M032 treatment induced transcriptional signatures of immune modulation, both inflammatory and immunosuppressive (Chambers et al., 2021). For dogs with high-grade gliomas, the mean survival time after treatment was 108 days, and no significant adverse events were attributable to M032 (Omar et al., 2021).

R-115 expresses mouse IL-12 from the retargeted oHSV, R-LM113 ( Figure 2 ; Table 2 ) (Alessandrini et al., 2019). To evaluate immune-mediated effects, a new mouse syngeneic GSC model was developed; PDGF-B transduced BALB/c neural progenitor cells were implanted orthotopically, gliomas were harvested, and isolated GSCs were transduced with human HER-2 (mHGG^pdgf-HER2) (Reisoli et al., 2012). Infection of human and mGSCs was dependent on HER-2 expression. While R-115 did not extend the median survival of mice with mGSC-derived tumors compared to parental R-LM113, it did result in about 30% long-term survivors with none occurring in the R-LM113 group, and these were protected from rechallenge with mHGG^pdgf GSCs with and without HER-2 (Alessandrini et al., 2019).

OV-IL15C encodes the IL-15 complex of IL-15 and partial IL15Rα in oHSV (γ34.5Δ, ICP6^-) ( Table 2 ). IL-15 has pleiotropic roles in NK and T cell survival and activation, while the IL-15 complex with IL15Rα acts as a super-agonist. Conditioned media from infected glioma cells increases human NK or CD8+ T cell survival and killing of hGSCs (Ma et al., 2021). Treatment of orthotopic hGSC-derived tumors with hCD8+ T cells administered with OV-IL15C was significantly better than with parental OV-Q1 (Ma et al., 2021).

OV-Cmab-CCL5 expresses a secreted bispecific cancer cell-targeted (anti-EGFR)-chemokine (human or mCCL5) Fc IgG1 fusion protein inserted in oHSV fHsvQuik1 (γ34.5Δ, ICP6^-) (Tian et al., 2022) ( Table 2 ). Fc IgG1 induces Fc receptor-mediated NK cytotoxicity and macrophage phagocytosis and anti-EGFR tethers CCL5-Fc to tumor cells. In a humanized mouse model with implanted EGFRvIII⁺ hGSC tumors and adoptive transfer of PBMCs and activated T cells, OV-Cmab-hCCL5 was significantly better than parental OV-Q1. In an immunocompetent GBM model, CT2A-hEGFR cells were implanted and treated with OV-Cmab-mCCL5, which induced immune cell infiltration and activation, and ‘cured’ about a third of mice when injected early after implantation. Efficacy was abolished by depleting T cells and reduced by depleting NK cells or macrophages (Tian et al., 2022). These studies illustrate how a multimodel therapy with 6 distinct activities (oHSV inflammation and cytotoxicity, chemokine activity, Fc-mediated antibody-dependent cellular cytotoxicity (ADCC) and antibody-dependent cellular phagocytosis (ADCP), and inhibition of EGFR) that target multiple cell types can ‘cure’ rigorous GBM models in mice (Sanchez Gil and Rabkin, 2022).

oHSV-TRAIL expresses secretable TRAIL, a potent death receptor-dependent apoptosis inducer, from G47Δ (Tamura et al., 2013) ( Table 2 ). GBM cells exhibit variable TRAIL resistance, whereas most cells are sensitive to oHSV, providing a rationale for oHSV-TRAIL (Tamura et al., 2013). In vitro, oHSV-TRAIL was more efficacious in killing temozolomide-resistant human primary and recurrent GSCs (7 of 8 lines) than oHSV (G47Δ) (Jahan et al., 2017). Intratumoral injection of oHSV-TRAIL potently inhibited the growth of human chemoresistant primary and recurrent GSC-derived intracerebral tumors through extensive induction of apoptotic cell death (Jahan et al., 2017). Thus, local expression of TRAIL from oHSV avoids systemic toxicity and overcame TMZ resistance.

G47Δ-mAngio expresses murine angiostatin, which inhibited HUVEC tube formation after hGSC infection (Zhang et al., 2013) ( Table 2 ). In vivo, G47Δ-mAngio significantly though modestly inhibited the growth of hGSC-derived brain tumors compared to G47Δ-empty, which was further improved in combination with G47Δ-mIL12, which also has anti-angiogenic activity (Zhang et al., 2013). This improved efficacy was associated with a decrease in tumor vascularity (CD31+ vessels) and VEGF expression and increased LacZ (oHSV) expression (Zhang et al., 2013). VAE is armed with an endostatin-angiostatin fusion gene inserted into a γ34.5Δ, ICP6- oHSV (Zhang et al., 2014) ( Table 2 ). In an intracranial GSC-derived tumor model, VAE was significantly better than parental oHSV or recombinant endostatin at extending survival. This improvement was associated with a significant decrease in microvessel density (Zhang et al., 2014). RAMBO, a γ34.5Δ, ICP6^- oHSV expressing vasculostatin (extracellular fragment of BAI-1), modestly extended median survival in mice bearing intracerebral hGSC-derived tumors, which was further extended about 10% in combination with bevacizumab (anti-VEGFR) (Tomita et al., 2019) ( Table 2 ).

Another strategy employing Fc-mediated cytotoxicity was recently reported with OV-αCD47-G1, constructed on oHSV fHsvQuik1 (γ34.5Δ, ICP6^-) and armed with anti-CD47, an immune checkpoint that blocks phagocytosis by binding to SIRPα on myeloid cells (Xu et al., 2021) ( Table 2 ). In vitro, conditioned media from OV-αCD47-G1 infected glioma cells significantly increased mouse myeloid cell phagocytosis, human NK cytotoxicity on co-cultured primary GBM cells from PDX tumors, and human macrophage-induced cytokine gene expression (NOS2, IL12A, IL1B) (Xu et al., 2021). Intratumoral injection of OV-αCD47-G1 in orthotopic hGSC xenografts significantly extended survival compared to parental OV-Q1, with the majority being long-term ‘cures’ (Xu et al., 2021). To test this strategy in a fully-syngeneic model, OV-A4-IgG2b was constructed with anti-mCD47 on mouse IgG2b. Intratumoral injection of OV-A4-IgG2b in CT2A tumors significantly extended survival and ‘cures’ that were dependent on macrophages (Xu et al., 2021). The results with OV-αCD47-G1 and OV-Cmab-CCL5 illustrate the power of secreted bispecific molecules interacting with Fc receptors to target both the TME and tumor cells.

oHSV efficacy can be improved by increasing virus spread in the tumor and reducing NK cell killing of infected tumor cells. OV-CDH1 was constructed on fHsvQuik1 (γ34.5Δ, ICP6^-) and armed with human E-cadherin (CDH1) under the control of pIE4/5 promoter (Xu et al., 2018) ( Table 2 ). E-cadherin binds to KLRG1 on NK cells to block NK cytolysis and cooperates with nectin-1 at cell-cell adherens junctions to facilitate oHSV infection. oHSV infection of GSCs increase their killing by NK cells, which was modestly reduced with OV-CDH1 (Xu et al., 2018). In vivo, OV-CDH1 injection of hGSC-derived orthotopic tumors significantly extended survival compared to parental OV-Q1, which was associated with increased virus spread and infectious virus yield, and infiltrating NK cells (Xu et al., 2018). OV-CDH1 was safe after intracerebral or intravenous injection.

The therapeutic index for chemotherapeutic drugs is quite narrow, with severe dose-limiting toxicities, while the therapeutic index for oHSV is large with limited toxicities or resistance, and distinct cell death mechanisms. This makes the combination potentially synergistic (Kanai and Rabkin, 2013). TMZ is an oral alkylating agent that induces DNA breaks and is immune depleting. The major source of initial drug resistance in patient tumors and GSCs is the expression of MGMT, which removes the methyl guanine adducts, and is often characterized by the methylation status of the gene (Blough et al., 2010; Kanai et al., 2012a; Jahan et al., 2017). Like patients, some GSCs are resistant to TMZ (MGMT-positive) and some are sensitive. The combination of TMZ with G47Δ was synergistic in killing TMZ-sensitive hGSCs, but for TMZ-resistant GSCs was only synergistic when combined with an MGMT inhibitor O⁶-benzylguanine, while in normal human astrocytes was antagonistic (Kanai et al., 2012a). Synergy was due to oHSV-induced double-strand DNA breaks, manipulation of the DNA damage response (DDR), and sequestration of ATM (Kanai et al., 2012a). Combination treatment of TMZ-sensitive hGSC-derived orthotopic tumors was also synergistic, with about 50% long-term survivors, while a combination effect with TMZ-resistant GSCs was dependent on O⁶-benzylguanine (Kanai et al., 2012a). In contrast, when the combination of TMZ and G47Δ-mIL12 was evaluated in an immunocompetent orthotopic 005 mGSC-derived GBM model, TMZ abrogated the efficacy seen with G47Δ-mIL12 alone, even with O⁶-benzylguanine, likely due to its cytotoxic effects on tumor-infiltrating T cells and macrophages (Saha et al., 2020). This illustrates how a synergistic effect in immuno-deficient models can be reversed in an immunocompetent model and that all combinations are not beneficial.

Etoposide (VP-16), a topoisomerase II inhibitor, has been in clinical trials for GBM and pediatric brain tumors; however, high doses are very toxic. The combination of G47Δ with low doses of etoposide was moderately synergistic in killing hGSCs due to increased apoptosis (Cheema et al., 2011). In an intracerebral etoposide-resistant hGSC tumor model, the combination of low-dose etoposide with G47Δ resulted in a significant prolongation of survival compared to monotherapy that was associated with a large increase in apoptosis, as occurs in vitro (Cheema et al., 2011).

Poly(ADP-ribose) polymerase (PARP) plays a key role in the DDR; required for base excision repair and single-strand break repair. PARP inhibition leads to double-strand breaks (DSBs) and was found to be synthetic lethal with homologous recombination deficiency. This is the basis for the approval of PARP inhibitors (PARPis) for the treatment of BRCA1/2 mutated ovarian and breast cancer (Lord and Ashworth, 2017). Like with other therapeutics, some hGSCs are sensitive to PARPis and some are resistant, likely reflecting the tumor from which they were isolated, although all had PARP activity that was inhibited by PARPis (Ning et al., 2017). oHSVs G47Δ and MG18L synergized in vitro with PARPis, such as olaparib, in both sensitive and resistant hGSCs due to proteasomal degradation of Rad51 and Chk1 (Ning et al., 2017). The combination of MG18L with olaparib significantly enhanced the survival of mice bearing PARPi-resistant and -sensitive hGSC-derived tumors compared to a single agent alone, and this was associated with increased apoptosis and DSBs, and decreased Rad51 and Chk1, as in vitro (Ning et al., 2017).

Epigenetic alterations regulate expression of GSC stemness and DDR (Hersh et al., 2022). oHSV infection induces innate immunity, which inhibits virus replication and spread, while histone deacetylase (HDAC) inhibitors can upregulate virus gene expression and downregulate IFN-stimulated genes (Nakashima et al., 2015b). Pretreatment of hGSCs with pan-HDAC inhibitor VPA increased rQNestin34.5 replication and cytotoxicity, and reduced associated type I IFN-responsive gene expression (Otsuki et al., 2008; Nakashima et al., 2015a). Class IIb HDAC6 deacetylates tubulin and other cytoplasmic proteins, which are involved in endocytic uptake and lysosome fusion. Virion uptake in GSCs is mostly through endocytosis. HDAC6 inhibitors tubacin and tubastatin A significantly increased virus replication in 2 of 6 hGSCs, however, there was no correlation between HDAC6 expression and GSC sensitivity (Nakashima et al., 2015a). In vivo analysis of rQNestin34.5 treatment of hGSC-derived orthotopic tumors found that tubastatin A significantly increased infectious virus in the tumors and extended survival, but not significantly (Nakashima et al., 2015a). Demethylating agents such as 5-azacytidine (5-Aza), also increased the replication of rQNestin34.5 in GSCs and synergized in killing GSCs, due in part to hypermethylation of the viral nestin promoter in hGSCs (Okemoto et al., 2013). The combination of rQNestin34.5 with 5-Aza or decitabine significantly extended the survival of mice with hGSC-derived tumors compared to either agent alone, including over 30% long-term survivors (Okemoto et al., 2013).

Transforming growth factor-beta (TGFβ) plays a key role in normal development, the maintenance of GSC stemness, and reducing innate immune responses (Hersh et al., 2022). Based on the immunosuppressive properties of TGFβ, its effects on oHSV were examined. TGFβ1 treatment of NK cells rendered them less cytolytic against rQNestin34.5-infected GSCs in co-culture (Han et al., 2015). In both an hGSC-derived orthotopic xenograft and 4C8 syngeneic mouse model, pretreatment with TGFβ1 or NK cell depletion followed by rQNestin34.5 significantly extended survival, while TGFβ neutralizing antibody 1D11 abrogated oHSV efficacy (Han et al., 2015). In the immunocompetent 4C8 model, TGFβ pretreatment decreased NK and macrophage tumor infiltration and activation (Han et al., 2015). This indicated that administration of TGFβ before oHSV transiently inhibited innate immune cells, enhancing therapeutic outcomes. On the other hand, the combination of TGFβ receptor kinase inhibitor (TGFβRi) with MG18L and G47Δ synergized in killing primary and recurrent hGSCs, including TMZ-resistant (Esaki et al., 2017). Recurrent GSC TGFβR signaling, viability, and sphere formation were inhibited by TGFβRis. TGFβRi increased virus replication in recurrent hGSCs and inhibited JNK signaling, as did JNK inhibitor SP600125 (Esaki et al., 2017). Pretreatment of recurrent hGSC-derived brain tumors with TGFβR1i galunisertib followed by MG18L significantly inhibited tumor growth compared to monotherapy and resulted in about 60% long-term survivors (Esaki et al., 2017). The discordant results between the two studies could be due to the use of athymic versus SCID mice, blocking TGFβ versus TGFβR, differences between primary and recurrent hGSCs, and different oHSVs.

Integrin β1 is expressed on glioma cells, GSCs, and macrophages, and a humanized integrin β1 blocking antibody, OS2966, has shown antitumor activity. It boosted the therapeutic efficacy of rHSVQ (γ34.5Δ, ICP6^-) through inhibition of interferon signaling and proinflammatory cytokine expression, which increased oHSV replication and cytotoxicity in hGSCs (Lee et al., 2019). In a hGSC-derived orthotopic xenograft model, OS2966 in combination with a single injection of rHSVQ significantly but modestly increased mouse survival compared to single treatments (Lee et al., 2019).

MEK kinase inhibitor trametinib is FDA-approved for BRAF-mutant melanoma. Although BRAF mutants are rare in GBM, but more frequent in pediatric gliomas, the MEK/ERK signaling pathway is often activated due to NF1 or upstream receptor kinase mutations. The combination of BRAF-mutant and MEK inhibitors has shown some clinical efficacy in BRAF-mutant gliomas (Toll et al., 2019). Trametinib treatment of macrophages co-cultured with hGSCs in vitro resulted in a significant reduction in tumor necrosis factor-alpha (TNF-α) secretion and an increase in rHSVQ-Luc replication in hGSCs (Yoo et al., 2019). When GSCs alone were treated, trametinib increased oHSV cytotoxicity and reduced virus replication. In vivo, the combination extended median survival of mice with hGSC-derived tumors modestly, in contrast curing 50% of immunodeficient mice bearing glioma cell line xenografts (Yoo et al., 2019).

Notch signaling is activated in hGSCs by oHSV infection of adjacent GSCs (inside-out signaling), which can be blocked with γ-secretase inhibitors (Otani et al., 2020). Combination treatment with oHSV 34.5ENVE of mice bearing hGSC-derived brain tumors significantly extended survival that was associated with reduced Ki67+ and increased cleaved caspase-3+ cells (Otani et al., 2020).

VEGFR tyrosine kinase inhibitor axitinib, an anti-angiogenic agent, is FDA-approved. It extended survival of immunodeficient and immunocompetent mice bearing hGSC- and mGSC-derived intracerebral tumors significantly, but modestly, which was associated with decreased vascularity (Lu et al., 2015). Mouse brain microvascular endothelial cells (MBMEC) were sensitive to axitinib, but not G47Δ-mIL12 cytotoxicity, whereas hGSCs and mGSCs (005) were sensitive to both (Saha et al., 2018). Axitinib also inhibited endothelial transdifferentiation-tube formation (Soda et al., 2011) of mGSCs but not hGSCs, while the combination with G47Δ-mIL12 reduced mGSC sphere formation, an indicator of stemness, to a greater extent than either single treatment (Saha et al., 2018). Combination treatment of recurrent hGSC xenografts with high-dose axitinib, which blocked neovascularization, and G47Δ-mIL12 significantly extended survival compared to efficacious single treatments, and this was associated with increased macrophage infiltration, tumor necrosis, and inhibition of PDGFR/ERK pathway activation (Saha et al., 2018). With the mGSC model, combination therapy was only effective in immunocompetent but not immunodeficient mice, indicating dependence on T cells, and was not enhanced with immune checkpoint inhibitors (Saha et al., 2018).

Immune checkpoint inhibitors (ICI) have been exceptionally effective in many solid tumors, but not GBM due to excessive immunosuppression, low tumor mutation burden (number of nonsynonymous somatic mutations (potential neoantigens) in the tumor, and ‘cold’ TME (Himes et al., 2021). Because oHSV infection induces an inflammatory TME, it is reasonable to examine whether this will overcome the resistance to ICI in GBM. In a representative orthotopic mGSC model (005), single treatments with G47Δ-mIL12, anti-PD-1, or anti-CTLA-4 had a significant but modest effect on survival, which was improved by dual combination therapy (Saha et al., 2017). Triple combination therapy acted synergistically and resulted in most mice being ‘cured’, which was associated with a decrease in CD4+ Tregs and increases in the CD8+ T/CD4+ Treg ratio, and macrophage infiltration and M1-like polarization (Saha et al., 2017). Immune cell depletion experiments revealed a complicated dependency, with CD4+ depletion completely abrogating efficacy, and CD8+ or macrophage depletion eliminating all ‘cures’ (Saha et al., 2017). Therefore, four components were necessary for curative therapy; oHSV, local IL-12 expression, and two distinct ICIs, acting with a complex interconnectedness that reflects on the challenges for GBM immunotherapy.