Effect of γ34.5 Deletions on Oncolytic Herpes Simplex Virus Activity in Brain Tumors

Abstract

The ICP34.5 protein of herpes simplex virus (HSV) is involved in many aspects of viral pathogenesis; promoting neurovirulence, inhibiting interferon-induced shutoff of protein synthesis, interacting with PCNA and TBK1, inhibiting dendritic cell (DC) maturation, and binding to Beclin 1 to interfere with autophagy. Because of its key role in neuropathogenicity, the γ34.5 gene is deleted in all oncolytic HSVs (oHSVs) currently in clinical trial for treating malignant gliomas. Unfortunately, deletion of γ34.5 attenuates virus replication in cancer cells, especially human glioblastoma stem cells (GSCs). To develop new oHSVs for use in the brain and that replicate in GSCs, we explored the effect of deleting the γ34.5 Beclin 1 binding domain (BBD). To ensure cancer selectivity and safety, we inactivated the ICP6 gene (UL39, large subunit of ribonucleotide reductase), constructing ICP6 mutants with different γ34.5 genotypes: Δ68HR-6, intact γ34.5; Δ68H-6, γ34.5 BBD deleted; and 1716-6, γ34.5 deleted. Multimutated Δ68H-6 exhibited minimal neuropathogenicity in HSV-1-susceptible mice, as opposed to Δ68H and Δ68HR-6. It replicated well in human glioma cell lines and GSCs, effectively killing cells in vitro and prolonging survival of mice bearing orthotopic brain tumors. In contrast, 1716 and 1716-6 barely replicated in GSCs. Infection of glioma cells with Δ68H-6 and 1716-6 induced autophagy and increased phosphorylation of eIF2α, while inhibition of autophagy, by Beclin 1 short hairpin RNA (shRNA) knockdown or pharmacological inhibition, had no effect on virus replication or phosphorylated eIF2α (p-eIF2α) levels. Thus, Δ68H-6 represents a new oHSV vector that is safe and effective against a variety of brain tumor models.

INTRODUCTION

Oncolytic viruses are a new class of cancer therapeutic with a unique mechanism of action: selective virus replication and associated killing of cancer cells but not normal tissue (11). Since the first reported genetically engineered oncolytic herpes simplex virus (oHSV) (40), at least 8 different HSV-1 genes, including TK (UL23), ICP6 (UL39), γ34.5, and Us3, have been deleted/mutated to generate oHSV (49). A number of oHSVs (G207, 1716, OncoVex^GMCSF, NV1020, HF10, G47Δ) have entered clinical trials, including phase III, for a range of cancers (27, 49).

Glioblastoma (GBM), the most common primary brain tumor in adults, is invariably fatal despite optimal multimodal therapy, with a median survival of 12 to 16 months that has not substantially improved over the past 30 years (5, 67). Recently, a subpopulation of cells, cancer stem cells or tumor-initiating cells, have been isolated from a variety of tumors, including GBM, that have properties of adult stem cells, such as self-renewal and differentiation into more mature multiple lineages, as well as being highly tumorigenic in immune-deficient mice (63). Glioblastoma stem cells (GSCs) are thought to be important in GBM initiation, progression, heterogeneity, recurrence and resistance to therapy (13). Importantly, GSCs form tumors in mice that closely resemble, histopathologically and genotypically, the patient's tumor from which they were isolated, thus providing a highly representative disease model (30, 53, 65), which is in contrast to established GBM cell lines (33).

Because HSV is a human pathogen, a critical issue in designing oHSVs is endowing tumor selectivity with safety (61). Especially for use in the brain, where HSV encephalitis can be lethal, safety is paramount. Therefore, γ34.5, the major HSV-1 neurovirulence gene (14), has been deleted in all oHSV vectors clinically evaluated in the brain (1716 and G207) (39, 50, 71). ICP34.5 is a multifaceted protein with a variety of diverse activities: (i) it counteracts PKR-mediated innate immune responses and translation shutoff through its interaction with protein phosphatase 1α (PP1α) and dephosphorylation of phosphorylated eIF2α (p-eIF2α), which is mediated by the carboxyl-terminal GADD34 homology domain containing PP1α (amino acids [aa] 193 to 195) and eIF2α binding domains (aa 233 to 248) (15, 18, 34); (ii) it binds to PCNA, including a region overlapping a region of the GADD34 homology domain (8, 17); (iii) it interacts with TBK1 to disrupt IRF3 activity and beta interferon (IFN-β) expression through an internal amino-terminal domain (aa 72 to 106) (62); (iv) it is important in efficient virus egress/release in mouse cells (aa 30 to 106) (7, 22); and (v) it binds to Beclin 1 (Atg6) (aa 68 to 87) and inhibits autophagy, which contributes to neurovirulence in a PKR-dependent fashion (1, 47).

Autophagy is a major degradative pathway by which cytoplasmic constituents, including viruses, are delivered to lysosomes in response to a variety of cellular stresses (29). The role of autophagy in cancer is complex; allelic loss of Beclin 1 occurs in cancer, defective autophagy can promote tumorigenesis, and autophagy is often protective to therapy (68). HSV-1 induces autophagy, likely through a stress response and PKR activation, as it is dependent upon eIF2α phosphorylation (19, 54). ICP34.5 is the major viral protein inhibiting autophagy (55), indirectly through inhibition of eIF2α and directly by inhibiting Beclin 1 through the Beclin 1 binding domain (BBD; aa 68 to 87), which is distinct from the carboxyl-terminal PP1α binding/GADD34 homology domain but overlaps the TBK1 binding domain (47). BBD deletion mutants are able to dephosphorylate eIF2α and overcome host protein shutoff (47). γ34.5-deficient HSV-1 is unable to replicate in mouse embryo fibroblasts (MEFs) or human SK-N-SH neuroblastoma cells, even in autophagy-deficient MEFs (Atg5^−/−), whereas BBD deletion mutants replicated similarly to their wild-type revertants (1, 47). This would suggest that autophagy and its inhibition are not critical to HSV replication in vitro in the cell lines that have been examined so far. In vivo, BBD deletion mutants (Δ68H and Δ68-87) were neuroattenuated, indicating that the BBD contributes to γ34.5-mediated virulence in mice (31, 47).

Unfortunately, deletion of γ34.5 attenuates virus replication even in cancer cells, and especially in GSCs (57, 64). Therefore, it is important to develop new oHSVs that maintain the safety of γ34.5-deficient mutants yet replicate efficiently in GSCs. To this end, we examined the effect of the γ34.5 BBD deletion (Δ68H) on safety in the brain and efficacy toward GSCs. To engender selectivity for cancer cells and improve safety, we inactivated the ICP6 gene (UL39), encoding the large subunit of ribonucleotide reductase, to generate Δ68H-6. For isogenic controls in strain 17⁺ backgrounds, we constructed mutants possessing the identical ICP6 inactivation, Δ68HR-6 from Δ68HR with intact γ34.5 and 1716-6 from 1716 lacking γ34.5. Δ68H-6 was very efficacious in killing GSCs in vitro and prolonging survival of mice bearing GSC-derived brain tumors. As opposed to Δ68HR-6, it caused minimal morbidity after intracerebral injection. Pharmacological inhibition of autophagy or short hairpin RNA (shRNA)-mediated knockdown of Beclin 1 did not affect the replication or cytotoxicity of any of the viruses, irrespective of γ34.5 status. This is the first study to demonstrate the benefit of γ34.5 BBD deletion for oHSV virotherapy.

MATERIALS AND METHODS

Cells and reagents.

U87, U373, T98 human glioma, and Vero (African green monkey kidney) cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA). The N18 cell line was derived from an A/J mouse and obtained from K. Ikeda (Tokyo Institute of Psychiatry) (58). These cell lines were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum (FCS). The isolation of human GSCs was previously described (64). The MGG number refers to a passaged GSC culture isolated from an individual patient; MGG numbers were previously referred to as the GBM numbers (25, 64). MGG genotypes and tumor histopathology are described in reference 65. BT74 was obtained from S. Kesari (University of California at San Diego, CA) and originally isolated by D. James (University of California at San Francisco, CA) as GBM6 (48). GSCs were cultured in EF20 medium composed of Neurobasal medium (Invitrogen, Carlsbad, CA) supplemented with 3 mM l-glutamine (Mediatech, Manassas, VA), 1× B27 supplement (Invitrogen), 0.5× N2 supplement (Invitrogen), 2 μg/ml heparin (Sigma), 20 ng/ml recombinant human epidermal growth factor (EGF) (R&D Systems, Minneapolis, MN), 20 ng/ml recombinant human fibroblast growth factor 2 (FGF-2) (Peprotech, Rocky Hill, NJ), and 0.5× penicillin G-streptomycin sulfate-amphotericin B complex (Mediatech). Spheres were dissociated using a NeuroCult chemical dissociation kit (StemCell Technologies, Vancouver, BC, Canada). All cells were maintained at 37°C and 5% CO₂. 3-Methyladenine (3-MA) (Sigma, St. Louis, MO) was dissolved in phosphate-buffered saline (PBS). All passaged cells were confirmed to be mycoplasma free (LookOut mycoplasma kit; Sigma).

Viruses.

All viruses were constructed on a HSV-1 strain 17syn⁺ background. Δ68H, containing a deletion of γ34.5 BBD (aa 68 to 87), and Δ68HR, a marker-rescue virus of Δ68H, were kindly provided by D. Leib (Dartmouth-Hitchcock Medical Center, NH) (1). 1716 has a 759-bp deletion located within each copy of the BamHI fragment of the long repeat region of the genome, which removes most of γ34.5, and was kindly provided by N. Fraser (University of Pennsylvania, PA) (38). Viruses were grown on Vero cells and purified, and their titers were determined on Vero cells (42).

Construction and characterization of ICP6 mutants.

Construction and characterization of the ICP6 (UL39) mutants (Δ68H-6, Δ68HR-6, 1716-6) was as previously described (42). Briefly, the 5.3-kb fragment of pKX2-βG3 (a gift from S. K. Weller, University of Connecticut Health Center), containing the E. coli lacZ sequence inserted in-frame in UL39, was cotransfected with parental viral DNAs into Vero cells using Lipofectamine (Invitrogen). Recombinant viruses isolated by limiting dilution and identified as plaques staining blue after 5-bromo-4-chloro-3-indolyl-β-d-galactoside (X-Gal) histochemistry were plaque purified three times in Vero cells. The genomic structures of ICP6 mutants were confirmed by restriction endonuclease digestion and Southern blot analysis. The presence or absence of BBD in Δ68HR-6 and Δ68H-6 was confirmed by PCR followed by HinfI digestion, as described previously (1).

Virus replication assay.

Cells were seeded into 24-well plates at 2 × 10⁴ cells/well in 500 μl of media and infected with HSV at a multiplicity of infection (MOI) of 0.1. In some experiments, infection was done in the presence of 3-MA at 1 mM. Cells were harvested with supernatants at indicated times in triplicate. After three freeze/thaw cycles and sonication, the titers of infectious progeny virus were determined by plaque assay on Vero cells.

Cell susceptibility assay.

Cells were seeded into 96-well plates at 5,000 cells in 100 μl of media and infected or mock infected with virus at indicated MOI. Three and a half days after infection, cell viability was assessed with a 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium (MTS) assay (Promega, Madison, WI) for cells in suspension and a 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide (MTT) assay (Sigma) for adherent cells according to the manufacturers' instructions. Cell viabilities of treatment groups were compared to those of the mock-infected controls.

Immunoblots.

Cells were mock treated or treated with HSV at an MOI of 2. At indicated times after infection, cells were collected and lysed in radioimmunoprecipitation buffer (Boston Bioproducts, Worcester, MA) with a cocktail of protease inhibitors (Roche, Indianapolis, IN). Fifteen micrograms of protein was separated by 10 to 15% SDS-PAGE and transferred to polyvinylidene difluoride membranes by electroblotting. After blocking with 5% nonfat dry milk in TBS-Tween 20 (TBST; 20 mM Tris [pH 7.5], 150 mM NaCl, 0.1% Tween 20), membranes were incubated with antibodies to p-eIF2α (Ser51) (rabbit monoclonal; Abcam, Cambridge, MA), eIF2α (mouse monoclonal; Cell Signaling Technology, Beverly, MA), LC-3 (rabbit polyclonal; Novus Biologicals, Littleton, CO), Beclin 1 (rabbit polyclonal; Novus Biologicals), or actin (rabbit polyclonal; Sigma). The membranes were washed in TBST and incubated with appropriate peroxidase-conjugated secondary antibodies. Protein-antibody complexes were visualized using the enhanced chemiluminescence kit (Amersham Bioscience, Piscataway, NJ).

shRNA-mediated knockdown of human or mouse Beclin 1.

Plasmid constructs containing shRNA sequences against human Beclin 1 (TRCN0000033550, TRCN0000033551) and mouse Beclin 1 (TRCN0000087288, TRCN000087290) were obtained from Open Biosystems (Huntsville, AL). Nontargeting shRNA (SHC002; should not target human or mouse genes but will engage RISC) was obtained from Sigma. Generation of lentiviral constructs for shRNA-mediated knockdown were as described in the protocol listed at the RNAi Consortium/Broad Institute website (http://www.broadinstitute.org/rnai/public/static/protocols/TRC_Protocols_Section_II_Viral_Production_100809.pdf). Cells were infected with lentiviral supernatants, which were removed after overnight incubation and grown in EF20 (MGG4, MGG8) or in DMEM-FCS (U87, N18) for 36 h before selection with puromycin (Sigma). Protein levels of target genes were assessed by Western blot analysis.

Animal experiments.

All in vivo procedures were approved by the Subcommittee on Research Animal Care at Massachusetts General Hospital.

Safety evaluation study in mice.

Female A/J mice at 7 weeks of age were obtained from the National Cancer Institute, Frederick, MD. Viruses at indicated doses in 3 μl of virus buffer (150 mM NaCl, 20 mM Tris [pH 7.5]) were stereotactically inoculated into the right striatum as described previously (64). Mice were then checked daily for 28 days to follow their general appearance, spontaneous activity, reaction to external stimuli, or neurological deficits, if any. Mice were euthanized with a lethal dose of pentobarbital (intraperitoneal [i.p.] injection) when they became moribund, lethargic, dehydrated, or distressed.

Efficacy study.

Female athymic mice at 6 to 8 weeks of age were obtained from the National Cancer Institute. To generate orthotopic xenograft models, 2 × 10⁵ MGG4 or 3 × 10⁵ U87 in 3 μl were stereotactically implanted into the brains of athymic mice, as described previously (64). After implantation, mice were randomly placed into three (MGG4) or four (U87) groups. On day 11 (MGG4) or 10 (U87), mice were treated by intratumoral injection of 2 × 10⁶ PFU (MGG4) or 5 × 10⁵ PFU (U87) of virus in 3 μl of virus buffer or virus buffer alone (mock). Mice were then observed until they became moribund, at which point they were sacrificed, and the presence of intracranial tumors was confirmed. For the syngeneic brain tumor model, 1 × 10⁵ N18 cells were implanted into the brains of 6- to 8-week-old A/J mice, which were treated with 1 × 10⁶ PFU virus on day 5. For histological studies using MGG4, mice (35 days after tumor implantation) were mock treated or treated with Δ68H-6 or 1716. Upon sacrifice, brains were removed and fixed in 4% paraformaldehyde, and frozen sections subjected to X-Gal and hematoxylin staining or immunocytochemistry with rabbit anti-HSV1 antibody (Dako, Carpinteria, CA) followed by incubation with Cy3-conjugated secondary antibody (Jackson ImmunoResearch, West Grove, PA).

Statistics.

Comparisons of data in viral replication and cytotoxicity assays were performed using a two-tailed Student t test (unpaired) or two-way analysis of variance (ANOVA) with Bonferroni multiple comparisons. Survival analysis was conducted by Kaplan-Meier curves, and comparisons were determined by log rank test. P values less than 0.05 were considered statistically significant.

RESULTS

Construction of ICP6 mutants from Δ68H, Δ68HR, and 1716.

In order to develop new oncolytic HSV vectors that target GSCs and are safe in the brain, we examined deletion of γ34.5 BBD, which was reported to attenuate HSV pathogenicity but not virus replication in vitro (1, 31, 47). While deletion of γ34.5 BBD reduces neuropathogenicity (31, 47), further attenuation is necessary for therapeutic use as an oHSV in the brain. Therefore, we introduced an ICP6 (UL39) mutation (inactivating LacZ insertion) into Δ68H (γ34.5 BBD deletion) to generate Δ68H-6, as well as into Δ68HR (marker-rescued Δ68H, γ34.5⁺) to generate Δ68HR-6, and into 1716 (γ34.5 deletion) to generate 1717-6 (Fig. 1A). All 6 viruses are on a strain 17⁺ background. The genomic structures of Δ68HR-6, Δ68H-6, and 1716-6 were confirmed by restriction endonuclease digestion and Southern blot analysis or PCR (Fig. 1B and D). All three ICP6 mutants produce similar-sized plaques on Vero cells, which were similar to the parental viruses, except they stained with X-Gal (Fig. 2A).

Fig 1.

Virus constructs. (A) Schematic of recombinant viruses. Δ68HR-6, Δ68H-6, and 1716-6 contain E. coli lacZ coding sequences inserted into the UL39 gene of Δ68HR, Δ68H, and 1716, respectively. Δ68HR retains both copies of the intact γ34.5 gene and is a marker rescue revertant of Δ68H, Δ68H contains a deletion of the BBD in both copies of the γ34.5 gene, and 1716 has a 759-bp deletion of the BamHI fragment of the long repeat region of the genome, removing most of the γ34.5 gene (C). Boxes represent inverted repeat sequences flanking the long (U_L) and short (U_S) unique sequences (B, BamHI; G, BglII; X, XhoI). (B) DNA structure of plaque-purified recombinant viruses Δ68HR-6, Δ68H-6, and 1716-6 was confirmed by restriction endonuclease mapping and Southern blot hybridization. The lacZ insertion in the ICP6 gene was confirmed by hybridization of NdeI-digested viral DNAs with a probe containing lacZ sequence flanked by ICP6 gene sequence. Hybridization of NdeI-digested viral DNAs detects the 13-kb fragments of Δ68HR, Δ68H, and 1716 and the expected 5.7-kb and 10.3-kb fragments of Δ68HR-6, Δ68H-6, and 1716-6 (bottom). (D) The presence or absence of BBD in Δ68HR-6 and Δ68H-6 was confirmed by PCR, followed by HinfI digestion as described previously (1). HR, Δ68HR; H, Δ68H; S, 1716; HR6, Δ68HR-6; H6, Δ68H-6; S6, 1716-6; M, DNA fragment ladder. Lanes marked 1 and 2 contain independent clones of each of the recombinants.

Fig 2.

Virus plaques and neurovirulence. (A) Representative images of virus plaques on Vero cells. Vero cells were infected with parental viruses (left) or their ICP6 mutants expressing LacZ (right), and 42 h later, cells were fixed, stained with X-Gal, and counterstained with neutral red. Scale bars = 500 μm. (B to D) Morbidity caused by intracerebral inoculation of BBD mutants evaluated with the following scoring scale: general appearance, spontaneous activity, reaction to external stimuli or neurologic deficits, scoring each from 1 (severely impaired) to 4 (normal) for a total score of 3 to 12. Time course of morbidities caused by Δ68H (100 PFU) (B) with 1 of 4 survivors, Δ68HR-6 (1 × 10⁷ PFU) (C) with 4 of 8 survivors, and Δ68H-6 (1 × 10⁷ PFU) (D) with 8 of 8 survivors is indicated for each mouse (symbols).

Δ68H-6 is highly neuroattenuated in vivo.

For neuropathogenicity studies we used A/J mice, which are highly susceptible to HSV1 (57). One hundred PFU of Δ68H was lethal in 3 out of 4 mice, whereas all mice survived the highest obtainable dose of Δ68H-6 (1 × 10⁷ PFU) and only half the mice survived the same dose of Δ68HR-6 (Fig. 2B to D). It was previously reported that the LD₅₀ for Δ68H after intracerebral inoculation in C57BL/6 mice, more HSV resistant than A/J mice (37), was about 100 PFU (31). While the ICP6 mutation (Δ68HR-6) greatly attenuates neuropathogenicity, the combination with γ34.5 BBD deletion (Δ68H-6) provides increased safety and no detectable mortality (Fig. 2D). Thus, Δ68H-6 was deemed sufficiently safe for treatment in the brain.

Replication of ICP6 mutants and parental HSVs in vitro.

One of the key determinants of oHSV efficacy is efficient replication in tumor cells. We examined the replication of the HSV mutants using a panel of human glioma cell lines and GSCs and Vero cells (Fig. 3). Both Δ68H and Δ68HR replicated similarly in glioma cell lines and Vero cells, as reported previously (1), as did Δ68H-6 and Δ68HR-6 (Fig. 3A to D). However, the ICP6 mutation significantly decreased virus yield in glioma cells, except for Δ68H-6 and Δ68HR-6 in U87 at 72 h. 1716, lacking γ34.5, had significantly reduced replication in glioma cell lines compared to Δ68H or Δ68HR (i.e., from 9- to 63-fold at 72 h in U87 and T98, respectively, Δ68H versus 1716) but just slightly less (∼3-fold) in Vero cells (Fig. 3A to D). Δ68H replicated well in all GSCs and somewhat less than Δ68HR (2- to 10-fold), except in MGG4. Addition of the ICP6 mutation slightly reduced Δ68HR-6 replication in all GSCs (compared to Δ68HR at 72 h). Δ68H and Δ68H-6 were only significantly different in MGG13 and BT74 at 72 h postinfection (Fig. 3E to H). Overall, the mean virus yield of Δ68H-6 in the glioma cell lines was about 8-fold greater than in the GSCs. 1716 and 1716-6 barely replicated at all in any of the four GSCs (Fig. 3E to H), as seen previously with γ34.5-deleted G207 and R3616 (64). All viruses replicated poorly in the mouse N18 neuroblastoma cells, with a further ICP6-dependent decrease (∼8-fold) with Δ68H-6 and Δ68HR-6 (Fig. 4A). Mouse tumor cells tend to be less permissive to oHSV replication, especially for γ34.5-deleted viruses (6).

Fig 3.

Replication of HSV mutants and phosphorylated state of eIF2α in infected cells in vitro. (A to H, upper panels) Cells as indicated were infected at an MOI of 0.1, and virus titers were determined at indicated times after infection. *, P < 0.05, and **, P < 0.0001, between indicated pairs at 72 h (two-way ANOVA, Bonferroni posttest comparisons). In panel A, Δ68HR and Δ68HR-6 are significantly different at 24 and 48 h (P < 0.01), and Δ68H and Δ68HR are significantly different from 1716 at 24 and 48 h (P < 0.01). In panel B, Δ68H-6 and Δ68HR-6 are significantly different from 1716 at 72 h (P < 0.05). In panels A and F, Δ68H is significantly different from Δ68H-6 at 24 and 48 h (P < 0.01). In panel G, Δ68H and Δ68HR are significantly different at 24 and 48 h (P < 0.01). In panel H, Δ68H and Δ68HR are significantly different only at 72 h. Dashed line indicates the input virus titer. Error bars are standard deviations (SD). (A to H, lower panels) Twenty hours after infection (MOI = 2), cells were collected and processed for Western blotting. p-eIF2α, phosphorylated eIF2α (Ser51); M, mock; HR, Δ68HR; H, Δ68H; S, 1716; HR6, Δ68HR-6; H6, Δ68H-6; S6, 1716-6.

Fig 4.

Virus infection of mouse N18 neuroblastoma cells. (A) N18 cells were infected with indicated viruses at an MOI of 0.1, and virus titers were determined at 24 and 48 h after infection. Dashed line indicates input virus titer. *, P < 0.05, and **, P < 0.0001, between indicated pairs (two-way ANOVA, Bonferroni posttest comparisons). (B) Cell viability was assessed by MTT assay (in triplicate) at 3.5 days after infection at indicated MOIs. Relative cell viability was normalized to that of mock-infected controls. There was no statistical difference between BECN1 and NT shRNA transduced cells (two-way ANOVA, Bonferroni posttest comparisons). (C) Transduced N18 cells (BECN1, Beclin 1 shRNA; NT, nontargeting shRNA) were processed for Western blotting with anti-Beclin 1 and actin antibodies. (D) Transduced cells were infected with indicated viruses (MOI = 2) and, 20 h later, collected and processed for Western blotting. p-eIF2α, phosphorylated eIF2α (Ser51); M, mock; HR, Δ68HR; H, Δ68H; S, 1716; HR6, Δ68HR-6; H6, Δ68H-6; S6, 1716-6.

Phosphorylation state of eIF2α after infection with BBD mutants.

One of the well-characterized biological functions of γ34.5 is to dephosphorylate eIF2α via binding to and activating PP1α (18). In Vero cells, infection with either Δ68H or Δ68HR leads to dephosphorylation of eIF2α, with no detectable p-eIF2α, while infection with 1716 (lacking γ34.5) induced p-eIF2α compared to mock infection (Fig. 3A, lower panels), similar to that reported previously for SK-N-SH cells (47). In contrast, in the human glioma cell lines and N18, p-eIF2α levels were higher after Δ68H infection than with Δ68HR, although usually less than in mock infection (Fig. 3B to D, lower panels, and 4D). In the GSCs, Δ68H infection actually increased p-eIF2α levels compared to mock infection, indicating virus induction of p-eIF2α and an inability of γ34.5 lacking BBD to fully counteract this. Interestingly, in Δ68HR-infected MGG4 and BT74 cells, p-eIF2α levels were not reduced compared to mock infection, as occurs in Vero and glioma cells (Fig. 3E and H, lower panels), indicating that even intact γ34.5 was not sufficient. Infection with 1716 consistently produced the highest levels of p-eIF2α in all cells. In all cells tested, the addition of the ICP6 mutation did not alter p-eIF2α levels compared to the corresponding parental viruses (Δ68HR, Δ68H, 1716) (Fig. 3, lower panels).

Cytotoxicity of ICP6 mutants and parental HSVs against human glioma cells.

We next examined the oncolytic activity of the viruses in the glioma cell lines and GSCs. Similar to what was seen with virus replication, Δ68H was very efficacious in inhibiting the growth of all glioma cells in vitro at a low MOI (Fig. 5). Mutation of ICP6 in Δ68H-6 significantly reduced cytotoxicity in all GSCs except MGG13 (Fig. 5D-G). Δ68H-6 was also less cytotoxic than Δ68H in glioma cell lines U87 and U373 (Fig. 5A and B). The mean viability of the glioma cell lines after Δ68H-6 infection was about 70% that of the GSCs. In all GSCs, except MGG13, Δ68HR-6 was significantly more cytotoxic than Δ68H-6, whereas in the glioma cell lines there was no difference (Fig. 5). 1716 and 1716-6 were much less effective in GSCs (Fig. 5D to G), which corresponds with their limited replication (Fig. 3). Only in U87 was there no difference between 1716-6 and Δ68H-6 or Δ68HR-6, and U87 supported the highest replication of 1716-6 of any of the tested cells. Decreases in cell viability were dose dependent for all the glioma cells (Fig. 6). At an MOI of 1, the differences between viruses were not significant in the glioma cells, except between Δ68H-6 and 1716-6 in T98, while differences remained in GSCs (Fig. 6).

Fig 5.

Cytotoxic effects mediated by HSV mutants. Cell viability was assessed by MTT (A to C) or MTS (D to G) assay (in triplicate) at 3.5 days after infection at an MOI of 0.3. Relative cell viability was normalized to that of mock-infected controls. Error bars are SD. Statistical comparisons (unpaired t test) were between indicated viruses (brackets), where * and ** represent P ≤ 0.05 and P ≤ 0.005, respectively.

Fig 6.

Cytotoxicity of HSV mutants in vitro at MOIs of 0.1 and 1. Cell viability was assessed by MTT (upper panels) or MTS (lower panels) assay (in triplicate) at 3.5 days after infection at indicated MOIs (same experiment as Fig. 5). Relative cell viability was normalized to that of mock-infected controls. Dashed lines indicate 50% viability. Error bars are SD. In the GSCs, Δ68H-6 is significantly different from 1716, 1716-6, and Δ68HR-6 at MOIs = 0.1 and 1, except Δ68H-6 versus 1716 in MGG4 at an MOI of 0.1, and Δ68H-6 is significantly different from Δ68H in MGG4 (MOI = 1) and MGG8 (MOI = 0.1) (P < 0.05, unpaired t test). In glioma cell lines (U87, U373, T98), Δ68H-6 is significantly different from 1716-6 at an MOI of 0.1 and in T98 at an MOI of 1.

Pharmacological inhibition of autophagy with 3-MA does not affect replication or cytotoxicity of oHSV in vitro.

It is becoming recognized that many viruses, including HSV-1, interact with or exploit cellular autophagic machinery during their viral life cycle (35). We thus examined the significance of autophagy and its inhibition on the efficacy and replication of oHSV in glioma cells. We used lipidation of microtubule-associated protein 1 light chain 3 (LC3), conversion of LC3 I to LC3 II, as a marker for autophagy (28, 43), with rapamycin being a positive control for autophagy induction (Fig. 7A, U87 [Rapa]). Autophagy is induced and elevated by HSV lacking BBD (Δ68H-6 and 1716-6), as has been reported previously in normal cells (1, 47), to a greater extent than with rapamycin in U87 (Fig. 7A). While p-eIF2α levels were higher in 1716-6- than Δ68H-6-infected cells, the induction of autophagy (LC3 II) was similar with both viruses (Fig. 7A, S6 versus H6). Inhibition of autophagy with 3-MA blocked induction of LC3 II but not p-eIF2α after virus infection (Fig. 7A). 3-MA inhibition had no effect on virus replication (Fig. 7B) or cytotoxicity at multiple MOIs (Fig. 7C), indicating that autophagy is not contributing to the attenuated replication of γ34.5 mutant HSV in glioma cells.

Fig 7.

Inhibition of autophagy by 3-MA does not affect oHSV. (A) Autophagy and eIF2α phosphorylation was examined in the presence or absence of 3-MA (1 mM). Twenty hours after infection (MOI = 2), cells were collected and processed for Western blotting with antibodies to indicated proteins. p-eIF2α, phosphorylated eIF2α (Ser51); LC3 I, soluble form; LC3 II, membrane-bound form; M, mock; HR, Δ68HR; H, Δ68H; S, 1716; HR6, Δ68HR-6; H6, Δ68H-6; S6, 1716-6; Rapa, rapamycin (μM). (B) Replication of oHSV was examined in the presence or absence of 3-MA (1 mM). Cells were infected at an MOI of 0.1, and virus titers were determined at indicated times after infection. Dashed line indicates the input virus titer. There was no significant difference with 3-MA for any virus or cell. (C) Cell killing mediated by indicated oHSV was examined in the presence or absence of 3-MA (relative cell viability; normalized to mock infected). Cell viability was assessed by MTT (U87) or MTS (MGG4, MGG8) assay (in triplicate) at 3.5 days after infection at indicated MOIs. Dashed lines indicate 50% viability. There was no significant difference with 3-MA for any virus or cell.

shRNA-mediated knockdown of Beclin 1 does not affect replication or cytotoxicity of oHSV in vitro.

Because ICP34.5 interacts with Beclin 1, which is one of the essential components for autophagy, we examined the effect of Beclin 1 shRNA knockdown using lentivirus-transduced cells. Beclin 1 shRNA reduced the steady-state levels of Beclin 1 compared with nontargeting shRNA (Fig. 4C and 8B, inset). Similar to the results obtained using 3-MA, genetic inhibition of autophagy by means of knocking down Beclin 1 inhibited virus-induced autophagy but did not affect p-eIF2α levels, virus replication, or cytotoxicity after oHSV infection of human glioma cells (Fig. 8) and mouse N18 cells (Fig. 4B and D).

Fig 8.

shRNA-mediated knockdown of Beclin 1 does not affect oHSV. Cells were transduced with lentivirus vectors expressing Beclin 1 (BECN1) or nontargeting (NT) shRNA. (A) Transduced cells were infected with virus (M, mock; HR, Δ68HR; H, Δ68H; S, 1716; HR6, Δ68HR-6; H6, Δ68H-6; S6, 1716-6) at an MOI of 2 and, 20 h later, collected and processed for Western blotting. p-eIF2α, phosphorylated eIF2α (Ser51); LC3 I, soluble form; LC3 II, membrane-bound form. (B) Cells were infected (Δ68HR-6, Δ68H-6, 1716-7) at an MOI of 0.1, and virus titers were determined at 24 and 48 h after infection. Dashed lines indicate the input virus titer. There was no statistical difference between BECN1 and NT shRNA transduced cells. (Inset) Cells (BECN1, Beclin 1 shRNA; NT, nontargeting shRNA) were processed for Western blotting with anti-Beclin 1 and actin antibodies. (C) Cell viability was assessed by MTT (U87) or MTS (MGG4, MGG8) assay (in triplicate) at 3.5 days after infection at indicated MOIs. Relative cell viability was normalized to that of mock-infected controls. Dashed lines indicate 50% viability. There was no statistical difference between BECN1 and NT shRNA transduced cells.

In vivo efficacy of Δ68H-6 in intracerebral brain tumor models.

We evaluated the in vivo efficacy of a single intratumoral injection of Δ68H-6 using 3 different intracerebral tumor models (Fig. 9). In the U87 glioma model in athymic mice, Δ68H-6 was significantly more efficacious than 1716-6 at prolonging survival of tumor-bearing mice, while Δ68HR-6 and Δ68H-6 were similarly effective (Fig. 9A). In light of the report showing that Δ68H induced a more robust adaptive immune response (31), we examined the efficacy of Δ68H-6 in an immunocompetent tumor model: N18 neuroblastoma tumors in the brains of syngeneic A/J mice. In N18, Δ68HR-6 and Δ68H-6 were also similarly effective, although only Δ68H-6 was significantly more effective than 1716-6 (Fig. 9B). In this experiment, 1716-6 was not significantly different from mock treatment or 1716 (data not shown). Considering the importance of GSCs in gliomagenesis and therapy, we also assessed the efficacy in mice bearing GSC (MGG4)-derived intracerebral tumors. MGG4 formed a relatively well-circumscribed mass with intratumoral hemorrhage (Fig. 9C), as previously described (64, 65). Δ68H-6 efficiently infected and spread within the tumor, as illustrated by the extensive X-Gal staining (Δ68H-6-infected cells) at 2.5 days postinfection, which was more extensive than at 18 h (Fig. 9D). Immunostaining with anti-HSV antibody (Δ68H-6-infected and 1716-infected cells) confirmed Δ68H-6 spread within the tumor and lack of 1716 spread (Fig. 9D), as expected from the in vitro studies (Fig. 3). Intratumoral injection of Δ68H-6 significantly prolonged survival of mice (median survival of 61 and 47 days for Δ68H-6 and mock treatment, respectively), while 1716 had no effect (Fig. 9E).

Fig 9.

Therapeutic efficacy of Δ68H-6 in vivo. (A) Kaplan-Meier survival curves of athymic mice bearing intracerebral U87 tumors after treatment with intratumoral oHSV (5 × 10⁵ PFU) or mock treatment on day 10 postimplantation (mock, n = 7; Δ68HR-6, n = 6; Δ68H-6, n = 6; 1716-6, n = 6). P < 0.05 (mock treatment versus 1716-6, Δ68HR-6 versus 1716-6, Δ68H-6 versus 1716-6) (log-rank test). (B) Kaplan-Meier survival curves of A/J mice bearing intracerebral N18 tumors after treatment with intratumoral oHSV (1 × 10⁶ PFU) or mock treatment on day 5 postimplantation (mock, n = 8; Δ68HR-6, n = 7; Δ68H-6, n = 7; 1716, n = 6; 1716-6, n = 6). P < 0.05 (mock treatment versus 1716, mock treatment versus Δ68HR-6, mock treatment versus Δ68H-6, Δ68H-6 versus 1716-6) (log-rank test). (C) Athymic mice intracerebrally implanted with MGG4 were sacrificed on day 35. Shown is a hematoxylin-and-eosin-stained representative section of the brain with tumor (surrounded by arrowheads). Intratumoral hemorrhage was frequently observed (arrows). (D) Athymic mice with MGG4 tumors were treated with Δ68H-6 or 1716 and sacrificed 18 h (upper panels) and 2.5 days (lower panels) after virus injection. Brains were sectioned and tumors stained with X-Gal (left panels) or immunostained (immunofluorescence [IF]; middle and right panels) with antibodies to HSV (red, Δ68H-6 and 1716) (scale bars = 500 μm). (E) Kaplan-Meier survival curves of mice bearing intracerebral MGG4 tumors after treatment with intratumoral Δ68H-6, 1716 (2 × 10⁶ PFU), or mock treatment on day 11 postimplantation (mock, n = 8; Δ68H-6, n = 7; 1716, n = 7). P < 0.05 (mock treatment versus Δ68H-6 and Δ68H-6 versus 1716) (log-rank test).

DISCUSSION

The γ34.5 gene of HSV-1, which is not essential for virus replication in human nonneural cells in vitro, is the major viral neurovirulence factor, since loss of this gene renders HSV-1 avirulent in mice after intracerebral inoculation (14). The principal role of ICP34.5 in neurovirulence is thought to be in counteracting IFN-α/β-induced PKR-mediated protein synthesis shutoff, as γ34.5-deleted HSV-1 regains wild-type neurovirulence and replication in the nervous system in either PKR or IFN-αβR knockout mice (32). A second-site extragenic mutation that complements the γ34.5-deleted defect in blocking protein shutoff was found to delete ICP47 and the Us11 promoter, resulting in immediate-early expression of Us11, which inhibits PKR activation (44, 46). Importantly, this suppressor mutation significantly improves virus yields of γ34.5-deleted HSV-1 in a variety of cancer cells, without increasing neurovirulence (45, 57), which suggests that activities of γ34.5 beyond dephosphorylation of p-eIF2α are important in neurovirulence.

Recently, a region in the amino-terminal domain of γ34.5 (aa 68 to 87) that interacts with Beclin 1 and inhibits autophagy has been identified (47). Deletion of the γ34.5 BBD resulted in a virus (34.5Δ68-87) that induced autophagy in neurons and was attenuated for neuropathogenicity after intracerebral inoculation. We found that Δ68H was only minimally attenuated in vivo, with 1 in 4 survivors at 10² PFU, which is similar to what Leib et al. reported; an LD₅₀ of approximately 10² PFU in less sensitive C57BL/6 mice and all mice succumbing to marker-rescued virus (Δ68HR) at this dose (31). While we found no effect of autophagy on virus replication in vitro, the effects in vivo likely involve host immune responses. Neurovirulence of the BBD deletion mutant was restored in PKR^−/− mice, as PKR is also involved in inducing autophagy (55), but not IRF3^−/− or MyD88^−/− mice, consistent with this innate immune pathway or TBK1 binding not contributing to BBD-dependent neurovirulence (31, 47). The rescue of neurovirulence in Rag1^−/− mice further suggests that the primary function of BBD is to preclude autophagy-mediated adaptive immune responses. Virus-induced autophagy enhances major histocompatibility complex (MHC) class I and II antigen presentation and induces a stronger CD4⁺ T cell response (16, 31, 59).

While the BBD deletion reduced HSV neurovirulence, the effect was relatively small, and this necessitated combination with an ICP6 mutant for optimal safety in the brain. ICP6 is another determinant of HSV-1 neuropathogenicity (9, 69), as we showed here with Δ68HR-6 and previously for FΔ6 (25). In addition to its ribonucleotide reductase activity in complex with UL40, ICP6 protein also binds to eIF4G to promote assembly of the eIF4F complex and protein synthesis in quiescent cells (66). The latter function is an additional translational control perturbation by HSV and is likely important in attenuating ICP6⁻ HSV-1 neuropathogenicity.

Because of the central role of γ34.5 in HSV neurovirulence, it has often been deleted in engineering oHSVs (61), despite it causing reduced replication in cancer cells. Four of 6 oHSV vectors translated to the clinic, and all in clinical trial for glioblastoma, are deleted for γ34.5 (49). The inability of γ34.5-deleted oHSV to replicate in or kill GSCs may limit its use in patients with GBM. To address the reduced replication and efficacy of γ34.5 deletion mutants, several strategies have been employed: (i) deletion of the α47 gene and late Us11 promoter causing immediate early Us11 expression, which increases the replication and antitumor efficacy of γ34.5 deletion mutants (36, 56, 57). In addition, loss of ICP47 enhances presentation of MHC class I antigens on infected cells (57). (ii) Expression of inhibitors of PKR-mediated protein shutoff from other viruses, such as HCMV TRS1 or IRS1 (52), or the γ34.5 homologous domain from cellular MyD116 (GADD34) (4). (iii) Targeted expression of γ34.5 in cancer cells; transcriptionally through the use of tumor specific promoters, such as glioma selective nestin and musashi1 (23, 24), or translationally through the use of a picornavirus IRES (10). Here, we have chosen to examine a deletion of the γ34.5 BBD that was reported not to affect phosphatase activity (47).

As expected from previous studies in nonpermissive SK-N-SH neuroblastoma and MEF cells (1, 47), Δ68H replicated similarly to Δ68HR in Vero cells, glioma cell lines, and mouse N18 cells. In GSCs, Δ68H replicated efficiently but virus yields were somewhat reduced compared to Δ68HR (3- to 15-fold). Unexpectedly, this was associated with increased levels of p-eIF2α compared to mock treatment but still less than with 1716 and not altered by the ICP6 mutation. In general, there were higher levels of p-eIF2α even in wild-type HSV-infected GSCs than in glioma cell lines or Vero cells. This suggests that PP1-mediated phosphatase activity is insufficient and/or eIF2α kinases other than PKR are elevated in GSCs. Recombinant γ34.5 protein with a deletion of the amino-terminal region (aa 1 to 83), overlapping BBD and TBK1 binding domains, had decreased eIF2α phosphatase activity compared to a deletion of aa 1 to 52 or the wild type, suggesting that this region may impact the phosphatase function of the carboxy-terminal GADD34 homology domain (12). The decreased replication of Δ68H in GSCs compared to Δ68HR might indicate that the BBD affects other cellular activities or Beclin 1 interactions, possibly through apoptosis (26) or elevated p-eIF2α levels.

In a survey of human glioma cell lines, 1716 produced about 10-fold less virus than wild-type-HSV (41). We found a 12- to 79-fold difference between 1716 and Δ68HR in glioma cell lines after low-MOI infection. R3616 an F strain deleted for γ34.5, was similarly or more attenuated for replication in glioma cell lines compared to wild-type F (3, 57). In GSCs, 1716 was highly attenuated for replication, from 200- to 6,300-fold less than Δ68HR. However, there seemed to be some replication occurring, especially in BT74, in contrast to a lack of replication with R3616 and G207 (64).

γ34.5 blocks autophagy, both the number of autophagosomes and viruses in autophagosomes (xenophagy), in a PKR-dependent fashion (1, 47, 55). We found that infection of GSCs with either Δ68H-6 or 1716-6 induced autophagy, as detected by LC3-II accumulation, which was blocked by 3-MA or Beclin 1 shRNA. While initial studies suggested that γ34.5-deleted HSV was degraded by xenophagy (55), there is no evidence that autophagy inhibits HSV replication. Thus, Δ68H replication in MEFs lacking Atg5, involved in autophagosome assembly, is no different from that of Δ68HR or that in Atg5^+/+ MEFs (1). We found no detectable role for autophagy in HSV-1 replication or glioma cell cytotoxicity, either by inhibiting autophagy with 3-MA or by shRNA knockdown of Beclin 1, and whether cells were infected with HSV lacking γ34.5 or BBD or with intact γ34.5. Conversely, 1716-6 and Δ68H-6 induce autophagy similarly in glioma cells, yet there is a large difference in their replication. This is in contrast to oncolytic adenovirus, where virus induction of autophagy in glioma cells increased cytotoxicity that was sensitive to 3-MA (2, 20, 60, 70). There are contradictory reports as to whether inhibition of autophagy reduces adenovirus replication or not, which may be related to infected cancer cell type (21, 51). This indicates that autophagy induction by different viruses will have different outcomes on infected cancer cells.

The overall goal of these studies is to develop new oHSVs that would be safe for use in the brain and efficacious against GSCs. Therefore, it was necessary to test the BBD-deleted HSV in brain tumor models in vivo. In the U87 glioma model, Δ68H-6 was as effective as Δ68HR-6 in prolonging survival and significantly better than 1716-6. The U87 tumor model is not very representative of human disease, so we also tested a GSC-derived intracerebral tumor model (MGG4). In this model, 1716 was ineffective, as expected from in vitro studies, while Δ68H-6 extended survival by ∼30%. In light of enhanced CD4⁺ T cell responses in BBD-deleted virus infected mice compared to wild-type HSV-1 (31), we examined a syngeneic tumor model, mouse N18 neuroblastoma in A/J mice. Again, Δ68H-6 injection extended survival but to the same extent as Δ68HR-6. Unfortunately, because of the rapidity of tumor growth, there was likely insufficient time for the effects of an adaptive antitumor immune response to be detected (58).

In summary, we show for the first time that deletion of γ34.5 BBD in combination with ICP6 mutation generates an oHSV that is safe for intracerebral administration and efficacious against human GSCs in vitro and a variety of brain tumors in vivo. Inhibition of autophagy does not affect HSV replication or cytotoxicity in glioma cells nor HSV-1-regulated p-eIF2α levels. This strategy of mutating domains in γ34.5 to improve virus replication with attenuated neurovirulence is likely to be a productive approach to generating additional oHSV vectors and improving our understanding of γ34.5 activities in infected cancer cells. The new oHSV construct, Δ68H-6, should be useful in a wide variety of solid tumors, in addition to brain tumors.