Role of the DNA Sensor STING in Protection from Lethal Infection following Corneal and Intracerebral Challenge with Herpes Simplex Virus 1

ABSTRACT

STING is a protein in the cytosolic DNA and cyclic dinucleotide sensor pathway that is critical for the initiation of innate responses to infection by various pathogens. Consistent with this, herpes simplex virus 1 (HSV-1) causes invariable and rapid lethality in STING-deficient (STING^−/−) mice following intravenous (i.v.) infection. In this study, using real-time bioluminescence imaging and virological assays, as expected, we demonstrated that STING^−/− mice support greater replication and spread in ocular tissues and the nervous system. In contrast, they did not succumb to challenge via the corneal route even with high titers of a virus that was routinely lethal to STING^−/− mice by the i.v. route. Corneally infected STING^−/− mice also showed increased periocular disease and increased corneal and trigeminal ganglia titers, although there was no difference in brain titers. They also showed elevated expression of tumor necrosis factor alpha (TNF-α) and CXCL9 relative to control mice but surprisingly modest changes in type I interferon expression. Finally, we also showed that HSV strains lacking the ability to counter autophagy and the PKR-driven antiviral state had near-wild-type virulence following intracerebral infection of STING^−/− mice. Together, these data show that while STING is an important component of host resistance to HSV in the cornea, its previously shown immutable role in mediating host survival by the i.v. route was not recapitulated following a mucosal infection route. Furthermore, our data are consistent with the idea that HSV counters STING-mediated induction of the antiviral state and autophagy response, both of which are critical factors for survival following direct infection of the nervous system.

IMPORTANCE HSV infections represent an incurable source of morbidity and mortality in humans and are especially severe in neonatal and immunocompromised populations. A key step in the development of an immune response is the recognition of microbial components within infected cells. The host protein STING is important in this regard for the recognition of HSV DNA and the subsequent triggering of innate responses. STING was previously shown to be essential for protection against lethal challenge from intravenous HSV-1 infection. In this study, we show that the requirement for STING depends on the infection route. In addition, STING is important for appropriate regulation of the inflammatory response in the cornea, and our data are consistent with the idea that HSV modulates STING activity through inhibition of autophagy. Our results elucidate the importance of STING in host protection from HSV-1 and demonstrate the redundancy of host protective mechanisms, especially following mucosal infection.

INTRODUCTION

Herpes simplex virus 1 (HSV-1) is a member of the Alphaherpesvirus subfamily with high seroprevalence in the human population (1). Infection at mucosal surfaces such as the mouth, eyes, and genitalia leads initially to lytic replication in mucosal epithelial cells, followed by infection of the innervating sensory neurons. HSV-1 then travels in a retrograde direction to the neuronal cell body, where it establishes latency. It is this ability to establish latency that renders HSV-1 refractory to clearance by the immune system, allowing persistence for the lifetime of the host. During periods of reactivation from latency, HSV-1 can travel in the anterograde direction to mucosal tissues, causing diseases ranging in severity from the common cold sore to herpetic stromal keratitis (HSK), the most common cause of infectious blindness in the developed world (2). HSV-1 can also gain entry into the central nervous system (CNS) to cause herpes simplex encephalitis (HSE) (3). HSE is a leading cause of viral encephalitis, further underscoring the significant morbidity and mortality associated with HSV-1.

In order to effectively respond to infection, host cells have evolved a broad spectrum of sensors of evolutionarily conserved pathogen-associated molecular patterns (PAMPS) (for reviews, see references 4 and 5). To establish a directed and metered antiviral state, the cellular responses to viral nucleic acids are particularly important. While responses to endosomal nucleic acids by Toll-like receptor 3 (TLR3), TLR7, and TLR9 have been well studied, our understanding of responses to cytosolic nucleic acids is less developed (4, 6). Such cytoplasmic sensing pathways are candidates for efficient sensing of HSV because genomic HSV-1 DNA is found free in the cytoplasm following proteosomal degradation of the HSV-1 virion (7). STING (also known as MITA, ERIS, and TMEM173) is an adaptor protein activated by cytosolic double-stranded DNA (dsDNA) or cyclic dinucleotides (8–12). Once activated, STING translocates to the endoplasmic reticulum, activating the TBK1/IRF-3 pathway which upregulates interferon beta (IFN-β) (9). In addition, IFI16, a ligand of STING, is a DNA sensor that presents nuclear and cytoplasmic HSV DNA to STING, thereby potentiating the STING-dependent sensing of infection (13, 14). Consistent with this, STING is critical for survival of mice following high-titer intravenous (i.v.) HSV-1 infection (15). An additional study showed that STING-deficient mice have increased HSV-1 titers in corneas relative to control mice, but no further parameters of pathogenesis were measured (16). These studies confirm a role for STING in HSV infection and collectively suggest that STING deficiency renders the host highly susceptible to HSV infection, equivalent perhaps to deficiencies in STAT1, type I IFN receptors, or IRF-3/7 (17–20). These studies notwithstanding, how STING-driven responses shape HSV pathogenesis following peripheral challenge remains unknown. The efficacy of STING-driven responses to HSV varies by cell type in vitro, which further complicates predictions of the role of STING in vivo (9, 15, 21). Furthermore, recent work has demonstrated cross talk between the STING and autophagy pathways and shown it to be important for activation of IRF-3 (22, 23). These host defenses notwithstanding, HSV contains a variety of genes that serve to counter IFN-driven innate responses and autophagy (24–27). The γ34.5 gene of HSV-1 is an especially potent neurovirulence factor in mice and humans that serves to counter the IFN- and protein kinase PKR-driven antiviral state, as well as strongly modulating autophagy/xenophagy through a specific interaction with the essential autophagy protein Beclin 1 (24, 28, 29). This provokes the idea that through γ34.5, HSV may thereby inhibit the STING-driven response through its ability to modulate the autophagy pathway (30).

In this study, we examined central (intravenous [i.v.] and intracerebral [i.c.]) and peripheral (ocular) routes of HSV-1 infection in STING-deficient and control mice. In agreement with previous studies, we found STING was essential for control of HSV-1 following central challenge. In contrast to previous studies, however, STING was dispensable for survival following peripheral challenge, even with high titers of virus that were routinely lethal to STING^−/− mice at low i.v. doses. We also demonstrate a role for tumor necrosis factor alpha (TNF-α) and CXCL9 in the increased pathology observed in STING-deficient mice and show that γ34.5 counters both STING-driven antiviral and autophagy responses in the infected host.

MATERIALS AND METHODS

Viruses, cells, and mice.

Vero cells were used to propagate and determine the titers of viruses as described previously (31). The wild-type (WT) HSV-1 strains used were strain KOS and strain 17 syn⁺ (32, 33). KOS/Dlux/OriL, the luciferase-expressing HSV strain used for bioluminescent imaging (BLI), was previously described (34). The γ34.5-null virus, Δγ34.5, the γ34.5 Beclin-binding domain-deleted virus Δ68H (termed herein ΔBBD), and the thymidine kinase null virus 17/tBTK⁻ (termed herein ΔTK) were all on the strain 17 syn⁺ background and described previously (35–37). Heterozygous STING^-/+ mice in a mixed C57BL/6 and 129SvEv background were generously provided by Glen Barber (University of Miami) and were described previously (9). This study was carried out in accordance with guidelines set forth by the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. Our protocols were approved by the Dartmouth IACUC (05/06/12, approval no. leib.da.1).

Animal infection, organ harvest, and periocular disease scoring.

Heterozygous STING^+/− mice were bred to generate STING^−/− and wild-type littermate control mice and genotyped according to methods and primers described previously (9, 38). Male and female mice, aged 6 to 14 weeks, were anesthetized with intraperitoneal (i.p.) injection of ketamine (87 mg/kg body weight) and xylazine (13 mg/kg body weight). Corneas were scarified in a 10 × 10 crosshatch pattern, and mice were either inoculated with 2 × 10⁶ PFU virus (unless otherwise noted) in 5 μl of inoculation medium (Dulbecco's modified Eagle's medium [DMEM; HyClone] with 2% fetal bovine serum, 60 U/ml penicillin [HyClone], and a 60-μg/ml final concentration of streptomycin [HyClone]) or mock infected with 5 μl of inoculum medium (31). For intracranial infections, mice were anesthetized as described above and then injected with 1 × 10⁴ PFU of the indicated virus in 10 μl inoculum medium using a Hamilton syringe and a 26G needle. For intravenous infections, 1 × 10⁷ PFU of strain KOS in a volume of 150 μl was injected directly into tail veins.

Mice were sacrificed at the specified times postinfection or once they met endpoint criteria as defined by our IACUC protocol. Eye swabs were collected at indicated time points as previously described (39). Blood was harvested and serum was separated by centrifugation at 2,000 relative centrifugal force (RCF) for 5 min and then stored at −80°C. Eye swabs, spleens, livers, brains, brain stems, and trigeminal ganglia were frozen in the appropriate volume of inoculation medium at −80°C. Tissues were prepared for titer determination by homogenization/disruption with glass beads and sonication as previously described (31).

Mice were scored for disease and weighed at the specified times postinfection. Periocular disease scoring was performed as previously described and summarized here: 0, no pathology; 0.5, minor eyelid swelling; 1.0, minor eyelid and nasal swelling; 1.5, moderate eyelid and nasal swelling; 2.0, severe eyelid swelling with minor periocular hair loss and skin damage; 3.0, neurological symptoms (20).

IFN-β ELISA.

Mice were infected as indicated previously, and organs were harvested, weighed, and placed in extraction reagent I (Invitrogen, Carlsbad, CA). Organs were homogenized using an electric (Omni International, Kennesaw, GA) homogenizer prior to enzyme-linked immunosorbent assay (ELISA). Samples were then centrifuged, and supernatants were analyzed using an IFN-β ELISA-HS kit according to the manufacturer's instructions (PBL Interferon Source, Piscataway, NJ).

Quantitative real-time PCR.

Organs were harvested into tissue extraction reagent I (Invitrogen, CA) and then homogenized (Omni International, GA) at the time points indicated. RNA was extracted using TRIzol (Life Technologies, NY) and further purified using an RNeasy kit (Qiagen). cDNA was synthesized using SuperScript III (Life Technologies, NY) and random hexamer primers (Promega, WI). This cDNA was used for SYBR green real-time PCR (Life Technologies, NY). IFIT1 and TNF-α were measured relative to GAPDH (glyceraldehyde-3-phosphate dehydrogenase) using primers as previously described (40, 41) and analyzed using the 2^−ΔΔCT threshold cycle method (42).

Cytokine quantification.

Organs were harvested into tissue extraction reagent (Invitrogen, Carlsbad, CA) and homogenized (Omni International, Kennesaw, GA) at the time points indicated. Cytokines were quantified using a mouse 32-plex Luminex assay (MPXMCYTO70KPMX32; EMD Millipore, Darmstadt, Germany). CXCL9 levels were further analyzed using the bead-based cytokine quantification assay Super X-Plex (Antigenix America, Melville, NY) according to the manufacturer's instructions with samples treated as per the Luminex assay.

BLI.

Mice infected corneally with KOS/Dlux/OriL and at the appropriate times postinfection were injected i.p. with filter-sterilized d-luciferin potassium salt (Gold Biotech, Olivette, MO) in phosphate-buffered saline (PBS) at 150 μg/g body weight. Mice were anesthetized with 2.5% isoflurane and imaged using a cooled charge-coupled device (CCD) camera-equipped instrument (IVIS 100; Caliper LifeSciences, MA) as previously described (43). Mice were imaged for 30 s with f-stop 1, binning 8, and a field of view of 6.6. To quantify luminescent signals, identical regions of interest (ROI) were placed over images encompassing the mouse's head from the dorsal view. ROI results were recorded using total photon flux in photons per second per ROI. All images were analyzed using Igor Pro Living Image software (version 2.60) (PerkinElmer, Akron, OH).

RESULTS

STING is dispensable for protection from lethality following peripheral HSV-1 infection.

We wished to address the hypothesis that the impact of STING on HSV pathogenesis is dependent on the route of infection. Previous results demonstrated that infection of STING^−/− mice was uniformly lethal following i.v. infection with 1 × 10⁷ PFU HSV-1 strain KOS, while ≥75% of control mice survived (15). We wanted to determine whether this mortality pattern would occur following infection at peripheral sites such as the cornea, a natural route of infection in humans. To match the lethal input inoculum of the previously published i.v. study, we used 1 × 10⁷ PFU KOS. This dose is significantly higher than that required for IFN-αβγR^−/−, STAT1^−/−, or IRF-7^−/− mice to succumb to corneal infection (20, 44). We therefore infected STING^−/− and littermate control mice either via the cornea, the tail vein (i.v.), or the cerebrum (i.c.) with HSV-1 KOS and observed the mice for 21 days or until endpoint health criteria were reached (Fig. 1A to C). Surprisingly, STING^−/− and control mice corneally infected with 1 × 10⁷ PFU KOS were comparable in terms of survival, with 1/9 of STING^−/− mice and 0/11 control mice reaching endpoint criteria within 21 days (Fig. 1A). In contrast, following i.v. infection with 1 × 10⁷ PFU KOS, 7/8 STING^−/− mice reached endpoint criteria within 7 days relative to only 3/10 control mice (Fig. 1B). To assess whether STING plays an important role during infection within tissues of the CNS, mice were infected i.c. with 1 × 10⁴ PFU HSV-1 KOS. We observed that 13/14 i.c.-infected STING^−/− mice reached endpoint criteria within 5 days postinfection (dpi) (Fig. 1C), in stark contrast to only 1/14 control mice reaching endpoint criteria within the 21-day experimental period. Consistent with this increased mortality following i.c. infection, we also observed significantly increased viral titers in the brains of STING^−/− mice relative to control mice when infected i.c. (Fig. 1D). Taken together, these data confirm the requirement for STING in protecting the host from HSV-1 in the CNS. In addition, reduced mortality following peripheral challenge suggests a STING-independent bottleneck to lethal infection via the cornea when infected with HSV-1 strain KOS.

FIG 1.

Survival of STING^−/− and littermate control mice following corneal, i.v., or i.c. infection. STING^−/− and control mice were infected with 1 × 10⁷ PFU HSV-1 KOS total via the cornea (A) or i.v. (B) or with 1 × 10⁴ PFU HSV-1 KOS i.c. (C). Survival was recorded until endpoint criteria were met. (D) Brain titers of STING^−/− and control mice infected with 1 × 10⁴ PFU HSV-1 (KOS) i.c. and sacrificed on day 3 postinfection. **, P < 0.005. Statistical significance was determined by a Mantel-Cox test, with two independent experiments performed with ≥9 mice.

STING protects the cornea from acute HSV infection but is not required for viral clearance.

To gain a rapid overview of viral replication and spread we used real-time bioluminescent imaging (BLI) in conjunction with KOS/Dlux/OriL (KOSDlux), a virus that expresses firefly luciferase (32, 34). Following corneal infection with 2 × 10⁶ PFU/eye of KOSDlux, STING^−/− and control mice were imaged daily on days 2 through 9 postinfection (dpi). Significant luciferase activity above baseline was observed in the eyes and the skin of the snout of both strains of mice on days 4 through 7 (Fig. 2A) (data not shown). Quantification of the BLI signals revealed a significantly higher photon flux in the STING^−/− mice relative to littermate controls over the 9-day time course (Fig. 2B). Despite this increase, photon flux in both the STING^−/− and wild-type mice returned to baseline on day 9, indicating both strains of mice were able to clear the infection in the cornea (Fig. 2B). Consistent with this, the visceral dissemination and mortality previously observed with corneal KOSDLux infection of IFN-αβγR ^−/− and STAT1^−/− mice was not observed in STING^−/− mice (44). These results demonstrate that while STING plays a role in the control of corneal HSV-1 infection, it is dispensable for the clearance of virus from the cornea and from tissues subsequently infected via this route.

FIG 2.

Bioluminescent imaging (BLI) and quantification of HSV-1 (KOSDlux) in STING^−/− and littermate control mice following corneal infection. (A) BLI of STING^−/− and control mice following corneal infection with 2 × 10⁶ PFU/eye KOSDLux. All images are formatted to a log scale with a minimum 5 × 10⁴ photons (p)/s/cm²/sr (purple) and maximum of 1 × 10⁶ p/s/cm²/sr (red). (B) Quantification of total photon flux performed on heads of all animals imaged using region of interest (ROI) analysis in Living Image software (Xenogen). Data show significant differences between STING^−/− and control mice using area under the curve analysis and an unpaired two-tailed t test (P = 0.0171). Results are from three independent experiments with ≥9 mice.

To confirm and extend our BLI results to tissues that cannot be easily imaged, STING^−/− mice and littermate controls were corneally infected with 2 × 10⁶ PFU/eye of HSV-1 KOS, and viral titers were quantified in corneas, brain, brain stem, liver, spleen, lymph nodes, and serum. STING^−/− mice had significantly increased viral titers compared to control mice in both eyes and trigeminal ganglia (Fig. 3A and B). Viral loads within the brain stem (Fig. 3C) and brain tissues (Fig. 3D) did not significantly differ between control and STING^−/− mice, confirming that there is no additional dissemination through the tissues of the CNS following peripheral infection. This was in contrast to mice that were infected i.c., in which there were significantly increased titers in the brains of STING^−/− mice (Fig. 1D). Virus was not detected in the serum, lymph nodes, spleens, or livers of STING^−/− or control mice on days 3 and 5 postinfection (data not shown), confirming the lack of visceral dissemination following peripheral challenge.

FIG 3.

Viral titers of STING^−/− and littermate control mice following corneal infection with 2 × 10⁶ PFU/eye of HSV-1 (KOS). Mice sacrificed on days 3 and 5 were used to collect titer data for eye swabs (A), trigeminal ganglia (B), brain stems (C), and brains (D). Periocular disease (E) and weight change (F) of STING^−/− and control mice were measured by a masked observer (20). **, P < 0.005; ***, P < 0.0005. Statistical significance was tested using an unpaired two-tailed t test for each day. Disease scores and weights postinfection were measured during two independent experiments, using a total of 14 to 23 mice per group.

Given the increased titers in corneas and trigeminal ganglia of STING^−/− mice, we wished to assess whether these mice would also exhibit altered periocular disease or weight loss (Fig. 3E and F). Periocular disease scores were significantly higher for STING^−/− mice than for control mice on 6 of 10 days monitored postinfection, and at all time points, the mean disease scores were higher for the STING^−/− mice. The weight loss of STING^−/− mice was overall greater than that in control mice, although it failed to reach statistical significance except for 8 days postinfection. Taken together, these data suggest that there is STING-dependent control of viral replication in the cornea but that STING is dispensable for viral clearance. Moreover, STING deficiency is associated with significantly increased periocular disease.

Interferon responses of STING^−/− mice following corneal infection.

STING acts as an adaptor that bridges cytosolic DNA sensing with upregulation of IFN-β—a critical cytokine in host defense to viral infection (45–47). We observed in this study that the survival of STING^−/− mice was dependent upon the route of infection. We therefore hypothesized that the survival differences observed in STING^−/− mice might result from anatomically distinct patterns of IFN-β signaling. To test this hypothesis, we ocularly infected STING^−/− and control mice with HSV-1 KOS and measured IFN-β in the cornea by ELISA (Fig. 4A). At 3 days post-corneal infection, IFN-β levels were significantly elevated in infected relative to mock-infected corneas. Surprisingly, IFN-β levels were slightly higher in the STING^−/− corneas relative to those of the littermate controls (Fig. 4A). We also measured IFN-β levels in brain tissue of mice following intracranial infection with HSV-1 KOS. Again, at 3 days post-i.c. infection, IFN-β levels were significantly elevated in infected relative to mock-infected brains (Fig. 4B). Consistent with previous studies, slightly lower levels of IFN-β were observed in STING^−/− mice compared to littermate controls (15). To further measure IFN synthesis following corneal and i.c. infections at earlier time points, we also measured IFN at 2, 4, 6, and 8 h postinfection, but all samples were at or below the level of detection of the ELISA (data not shown).

FIG 4.

Interferon responses following corneal or intracranial HSV-1 KOS infection. (A) IFN-β ELISA of STING^−/− and control corneas 3 days following infection with 2 × 10⁶ PFU/eye HSV-1 KOS. (B) IFN-β ELISAs on STING^−/− and littermate control brains 3 days following i.c. infection with 1 × 10⁴ PFU HSV-1 KOS. (C) Real-time PCR of IFIT1 mRNA expression relative to that in control mock-infected mice normalized to GAPDH in the corneas of STING^−/− and control mice 3 days post-corneal infection with 2 × 10⁶ PFU/eye HSV-1 KOS. (D) Real-time PCR of IFIT1 mRNA expression relative to that in control mock-infected mice normalized to GAPDH in the brains of STING^−/− and littermate control mice 3 days post-i.c. infection with 1 × 10⁴ PFU/eye HSV-1 KOS. *, P < 0.05. Statistical significance was determined using an unpaired two-tailed t test for IFN-β ELISAs and IFIT1 mRNA expression results. IFN-β ELISAs were carried out in two independent experiments on a total number of mice as follows: corneal infection, WT, n = 7, and STING^−/−, n = 7; i.c. infection, WT, n = 12, and STING^−/−, n = 14. IFIT1 real-time PCR was carried out in two independent experiments on a total of 7 mice per group.

Although these differences in IFN-β levels in the corneas and brains of STING^−/− relative to control mice were statistically significant (P < 0.05), the magnitude of these changes seemed insufficient to explain the differences in viral pathogenesis observed. Nonetheless, it was possible that small changes in IFN-β levels could have a disproportionate effect on downstream antiviral IFN-stimulated genes (ISGs). Furthermore, the sensitivity of ELISA for detecting IFN-β could be limiting since the values obtained were close to the limits of detection. To examine this further, induction of the ISG IFIT1 was measured using real-time PCR. Consistent with the patterns seen with IFN-β, there was strong induction of IFIT1 expression in corneas and brains of all infected mice (both STING^−/− and WT), relative to mock-infected mice (Fig. 4C and D). Furthermore, levels of IFIT1 expression were statistically indistinguishable in STING^−/− and control mice. Together, these data are consistent with the idea that the increased pathology and lethality caused by HSV infection in STING^−/− relative to control mice are largely independent of the expression and downstream effects of IFN-β.

Other cytokine responses in STING^−/− mice.

Given that the different survival phenotypes arising in STING^−/− and control mice following central or peripheral HSV-1 infection appeared to be independent of IFN synthesis, we hypothesized that a dysregulated immune response may be causing the pathology observed. A key early regulator that mediates corneal damage during HSV-1 infection is tumor necrosis factor alpha (TNF-α) (48, 49). Using real-time PCR, we measured TNF-α expression in corneas dissected from STING^−/− and control mice infected with HSV-1 KOS on day 3 postinfection (Fig. 5A). TNF-α expression was significantly elevated in corneas of STING^−/− mice relative to those of control or mock-infected mice. Following i.c. infection, however, the levels of TNF-α expression were comparable in STING^−/− and control mice (Fig. 5B). To further probe potential chemokines and cytokines that may contribute to the outcomes of corneal and i.c. infection of STING^−/− and control mice, we performed a 32-plex Luminex screen (data not shown). The corneas of corneally infected STING^−/− mice showed a particularly strong increase in the chemokine CXCL9 (Fig. 5C), consistent with the observation that that infected STING^−/− corneas exhibit increased inflammation. In contrast, wild-type and STING^−/− mice infected i.c. showed comparable levels of brain CXCL9 expression, although both groups had elevated CXCL9 compared to mock-infected mice (Fig. 5D). Corneas of STING^−/− mice therefore exhibit elevated inflammatory cytokine responses to HSV-1 infection. In contrast, there were no demonstrable differences in inflammatory cytokines between STING^−/− and control mice following i.c. infection.

FIG 5.

Cytokine RNA expression in infected STING^−/− and littermate control mice 3 days postinfection. (A and B) Real-time PCR of TNF-α mRNA expression relative to that in control mock-infected mice performed on the excised corneas of STING^−/− and control mice 3 days postinfection with 2 × 10⁶ PFU/eye HSV-1 KOS (A) or excised brains of mice 3 days post-intracranial infection with 1 × 10⁴ PFU HSV-1 KOS (B). (C and D) Luminex analysis of CXCL9 in corneas of mice 3 days post-corneal infection with 2 × 10⁶ PFU/eye HSV-1 KOS (C) or brains of mice 3 days post-intracranial infection with 1 × 10⁴ PFU HSV-1 KOS (D). *, P < 0.05. Statistical significance was determined using an unpaired two-tailed t test for TNF-α expression and CXCL9 concentration results. TNF-α real-time PCR and CXCL9 cytokine quantification were carried out in two independent experiments on a total of 5 mice per group.

Virulence of γ34.5-deficient viruses is restored in STING^−/− mice following i.c. infection.

The STING-driven antiviral response in infected cells activates and is modulated by autophagy (22, 23, 30, 50). Given the autophagy-modulating role of HSV-1 γ34.5, it was of interest to examine its role in modulating the effects of STING (24). We used two different recombinant viruses (Δ34.5 and ΔBBD) with mutations in the γ34.5 gene. Δ34.5 lacks the entire γ34.5 open reading frame (ORF) and is thereby unable to block the IFN- and PKR-driven antiviral response and cannot counter autophagy (35, 36, 51). ΔBBD is fully capable of blocking the IFN- and PKR-driven antiviral response but cannot block autophagy because the Beclin-binding domain of γ34.5 has been deleted (34). As a control, we also used 17ΔTK (17/tBTK⁻), which lacks the viral thymidine kinase gene (37). 17ΔTK is highly attenuated in vivo through a pathway unrelated to IFN responses or autophagy and therefore served to test the caveat that STING deficiency may nonspecifically increase or restore the virulence of any attenuated virus. All viruses used in these experiments were in the strain 17 background. While strain 17 has higher virulence than KOS, the mutant viruses are profoundly attenuated. To increase the probability of determining a clear phenotype with these strain 17 mutants, we therefore used a higher inoculum (2 × 10⁴ PFU) in these i.c. experiments than in those with KOS shown in Fig. 1C and D. For consistency within this experiment, we also used this higher inoculum with WT strain 17 (Fig. 6). Control mice showed a significant (P < 0.02) survival advantage compared to STING^−/− mice, although all mice reached endpoint criteria by 6 days postinfection (Fig. 6A). When lower doses of strain 17 (50 PFU) were administered, 82% of STING^−/− and 91% of control mice succumbed to infection, although a significant survival advantage for control mice was again observed (P < 0.02) (data not shown), consistent with the data for KOS (Fig. 1C). To test whether STING-dependent antiviral responses against HSV were dependent on autophagy and whether HSV can counter this response, we infected STING^−/− and control mice i.c. with ΔBBD (Fig. 6B). As previously shown, ΔBBD was significantly attenuated in control animals relative to strain 17, with only 50% of mice reaching endpoint criteria within 21 days (28). In contrast, in STING^−/− mice, the virulence of ΔBBD was significantly increased compared to that of littermate controls, with all STING^−/− mice succumbing to infection within 8 days. We next infected mice with the γ34.5 null mutant (Δγ34.5), which cannot modulate autophagy and is unable to dephosphorylate the α subunit of eukaryotic initiation factor 2 (eIF2α) to prevent host-imposed translational arrest (29, 52). Upon infection of control mice with Δγ34.5, there was profound attenuation, with all control mice surviving infection to 21 days (Fig. 6C). In contrast, all STING^−/− mice succumbed to infection with Δγ34.5 within 9 days (Fig. 6C), with kinetics that were statistically indistinguishable from those following infection with ΔBBD (Fig. 6B). To confirm that these changes in virulence in STING^−/− mice were specific to the IFN and autophagy pathways, we infected STING^−/− mice and controls with ΔTK (Fig. 6D). We observed complete survival of both wild-type and STING^−/− mice over the 21-day infection period, supporting the idea that the attenuation of the γ34.5 mutants is largely due to their inability to counteract the STING-dependent autophagy pathway.

FIG 6.

Survival analysis of STING^−/− and littermate control mice following intracranial infection with 2 × 10⁴ PFU of HSV-1 strain 17 (A), ΔBBD (B), Δγ34.5 (C), and ΔTK (D). Mortality was recorded upon endpoint criteria being met. Statistical significance was determined by a Mantel-Cox test. For each type of infection, two independent experiments were performed. The numbers of mice used for each survival experiment are as follows: strain 17, WT, n = 6, and STING^−/−, n = 10; ΔBBD, WT, n = 6, and STING^−/−, n = 7; Δγ34.5, WT, n = 9, and STING^−/−, n = 10; and ΔTK, WT, n = 11, and STING^−/−, n = 8.

To further understand the role of STING in corneal infection, STING^−/− and littermate control mice were infected corneally with the strain 17, Δγ34.5, ΔBBD, and ΔTK viruses. The inoculum used for all experiments was 1 × 10⁵ PFU/eye, as strain 17 is more neurovirulent than strain KOS. Following HSV-1 strain 17 corneal infection, all STING^−/− mice and 57% of control mice succumbed to infection (Fig. 7A). To test whether HSV-1 counters this STING-dependent autophagic response through the BBD domain of γ34.5, we infected mice corneally with ΔBBD (Fig. 7B). All STING^−/− mice succumbed to this infection by day 11 postinfection, while all control mice survived the challenge. The comparable endpoint of STING^−/− mice infected with strain 17 and ΔBBD, along with the increased survival of control mice, demonstrates that the HSV-1 γ34.5 BBD is most likely countering the STING-dependent autophagy response following corneal infection. STING^−/− and control mice infected with Δγ34.5 showed no statistically significant difference in survival (Fig. 7C). To confirm these results were specific to the autophagy and IFN pathways, we corneally infected STING^−/− and control mice with ΔTK (Fig. 7D). We found that all mice survived the ΔTK corneal challenge. Together with Fig. 1, these data suggest that neurovirulent strains of HSV-1 can overcome the corneal STING-independent barrier to cause significant mortality and that protection is mediated at least partially by STING.

FIG 7.

Survival analysis of STING^−/− and littermate control mice following corneal infection with 1 × 10⁵ PFU/eye of HSV-1 strain 17 (A), ΔBBD (B), Δγ34.5 (C), and ΔTK (D). Mortality was recorded upon endpoint criteria being met. Statistical significance was determined by a Mantel-Cox test. For each infection, two independent experiments were performed. The numbers of mice used for each survival experiment are as follows: strain 17, WT, n = 7, and STING^−/−, n = 7; ΔBBD, WT, n = 6, and STING^−/−, n = 6; Δγ34.5, WT, n = 5, and STING^−/−, n = 6; and ΔTK, WT, n = 4, and STING^−/−, n = 5.

DISCUSSION

STING is pivotal in the innate response to a variety of pathogens (15, 50, 53, 54). This is largely through its ability to recognize cytosolic dsDNA and subsequently interact with TBK, facilitating phosphorylation of IRF-3 and induction of IFN-dependent antiviral responses. Consistent with these activities, STING is indispensable for the protection of mice from HSV-1-induced mortality following intravenous challenge with the low-virulence HSV-1 KOS strain (15). The studies described herein are consistent with these findings, but our use of a mucosal route of infection via the cornea has revealed that STING is dispensable for survival following a peripheral challenge with KOS. These findings contrast with studies of other mediators of the IFN response, such as STAT1 and IFN receptors, which are essential for prevention of generalized infection and mortality following peripheral challenge with KOS. Our findings also provide an interesting contrast with mice deficient in IRF-3 which were originally shown to be completely resistant to challenge by intravenous HSV-1 but subsequently shown to be susceptible to lethal infection via the corneal and i.c. routes (55, 56). These data therefore underscore the importance of testing immune-deficient mice using a variety of infection routes. Furthermore, they show that even when the immune deficiencies are in the same antiviral pathways, different and unexpected resistance patterns can emerge.

IFN responses are a key determinant of the outcome of corneal infection by HSV-1. In this study, STING^−/− mice exhibited increased ocular titers relative to controls, consistent with previous studies (16). Using IVIS imaging, we were able to extend these observations and visualize viral tropism over a longer time course, allowing us to observe a more robust ocular infection in STING^−/− mice but with eventual clearance of the virus. This pattern was predicted based on previous studies, but it was unexpected that the altered titers were largely independent of IFN production in the tissues examined. One possible explanation is that the STING pathway induces autophagy, which in turn controls virus replication in the cornea (22, 23). STING^−/− mice may therefore have a reduced ability to control HSV through autophagy/xenophagy in the cornea, and as discussed further below, this may also apply to infection of CNS tissues. It was also notable that STING^−/− mice had greater periocular disease scores than control mice. This was consistent with increased TNF-α expression and CXCL9 concentration in the corneas of STING^−/− mice. Given that STING upregulates the expression of TNF-α, it seems likely that the increased TNF-α expression in STING^−/− mice results indirectly from increased viral titers rather than being a direct consequence of STING deficiency (57). Histology performed on corneas of infected mice revealed increased immune cell infiltration in the corneas of STING^−/− mice compared to controls (data not shown), consistent with the increases in observed TNF-α and CXCL9. While STING is clearly dispensable for mediating survival following corneal infection with strain KOS, it is important for limiting HSV replication in the cornea and thereby necessary to avoid the induction of immunopathological cytokine expression and subsequent periocular disease. These findings are consistent with previous data showing increased and protracted genital inflammation in STING^−/− mice following vaginal infection, although no survival data were presented (15). Furthermore, it has been observed during skin infections with HSV-1 that while STING^−/− mice have higher viral titers in the skin, virus is not found in the brain and STING^−/− mice do not succumb to infection (S. Bedoui, personal communication).

The patterns of HSV replication and virulence in the brains of STING^−/− mice in this study appear to be determined largely by the route of infection. Following corneal infection with HSV-1 KOS, despite significantly increased titers in the cornea and trigeminal ganglia of STING^−/− mice, there was no difference in viral replication in the brain or statistically discernible changes in survival. Furthermore, analysis of brain cytokines by BioPlex and ELISA on days 3 and 5 post-corneal infection showed no significant changes (data not shown). In contrast, following i.c. infection of STING^−/− mice with HSV-1 KOS, we observed significant changes in viral replication in the brain, proinflammatory cytokine production, and survival. Both i.c. infections with KOS and strain 17 demonstrated a significant survival advantage for control relative to STING^−/− mice. This survival advantage is greater for KOS, as expected given its reduced neurovirulence relative to strain 17, but both survival phenotypes are representative of the importance of STING when natural barriers to infection are breached. When the neurovirulent strain 17 was used in corneal infections, >50% of control mice and 100% of STING^−/− mice succumbed to the infection, demonstrating that neurovirulence of the virus can overcome the corneal replication bottleneck. It is possible that once the more neurovirulent virus is able to reach the brain, a STING-dependent response is necessary to control the virus: therefore, STING^−/− mice readily succumb to infection, although with the high neurovirulence of strain 17, even half of control mice succumb to infection. These data suggest a CNS-specific requirement for STING to promote host survival. The requirement for STING in the CNS, however, appeared largely independent of IFN since there were only modest changes in IFN-β and no differences in ISG (IFIT1) expression between STING^−/− and control mice after i.c. infection with KOS. These findings are consistent with the data and hypothesis that the nervous system largely utilizes nondestructive innate responses to infection, such as autophagy/xenophagy (58, 59). This mode of pathogen clearance preserves the viability of neurons, which are largely a nonreplicating irreplaceable population of cells. IFN-driven antiviral responses are therefore necessarily muted or ineffective in the nervous system, and xenophagy is a dominant antiviral defense in both the peripheral nervous system (PNS) and CNS. It is also of note that there were no discernible differences in the ability of HSV-1 to reactivate from latently infected trigeminal ganglia explanted from STING^−/− or control mice (data not shown). The relationship between STING and autophagy is complex and an area of a great deal of recent research. There is clear evidence, however, that in addition to stimulating the IFN-driven antiviral response, STING-dependent responses also stimulate the xenophagy pathway that promotes the clearance of intracellular pathogens (60). Indeed, STING is critical for the HSV-induced autophagy response, at least in bone marrow-derived dendritic cells (BMDCs) (61). Notably, STING transcription is low in cornea and brain tissues but high in antigen-presenting cells, cells that undergo significant autophagic flux (62, 63). The data described herein strongly implicate that HSV γ34.5 counters the antiviral effects of STING largely through its ability to modulate autophagy through binding to Beclin 1 rather than altering IFN responses through its ability to bind TBK, eIF2α, or PP1α (22, 24, 28, 29, 52, 64). Further support for this idea comes from studies in our laboratory showing that levels of IRF-3 phosphorylation, largely mediated by TBK, are comparable between ΔBBD and wild-type viruses (R. Manivanh and D. A. Leib, unpublished data). Data from this study are consistent with the idea that while the STING-driven IFN-dependent response is important in vivo in some instances, the STING-dependent activation in autophagy and xenophagy is also a critical antiviral mechanism, especially in the CNS. Both of these responses are powerfully countered by γ34.5 and likely by other viral factors, such as ICP0 and US11, which act in concert to regulate the intricate life cycle of HSV-1 (26, 27, 65).