Varicella Zoster Virus Induces Differential Cell-Type Specific Responses in Human Corneal Epithelial Cells and Keratocytes

Purpose While VZV DNA and antigen have been detected in acute and chronic VZV keratitis, it is unclear whether productive infection of corneal cells is ongoing or whether residual, noninfectious VZV antigens elicit inflammation. Herein, we examined VZV-infected primary human corneal epithelial cells (HCECs) and keratocytes (HKs) to elucidate the pathogenesis of VZV keratitis. Methods HCECs and HKs were mock- or VZV infected. Seven days later, cells were examined for morphology, proinflammatory cytokine and matrix metalloproteinase (MMP) release, ability to recruit peripheral blood mononuclear cells (PBMCs) and neutrophils, and MMP substrate cleavage. Results Both cell types synthesized infectious virus. VZV-infected HCECs proliferated, whereas VZV-infected HKs died. Compared to mock-infected cells, VZV-infected HCECs secreted significantly more IL-6, IL-8, IL-10, and IL-12p70 that were confirmed at the transcript level, and MMP-1 and MMP-9; conditioned supernatant attracted PBMCs and neutrophils and cleaved MMP substrates. In contrast, VZV-infected HKs suppressed cytokine secretion except for IL-8, which attracted neutrophils, and suppressed MMP release and substrate cleavage. Conclusions Overall, VZV-infected HCECs recapitulate findings of VZV keratitis with respect to epithelial cell proliferation, pseudodendrite formation and creation of a proinflammatory environment, providing an in vitro model for VZV infection of corneal epithelial cells. Furthermore, the proliferation and persistence of VZV-infected HCECs suggest that these cells may serve as viral reservoirs if immune clearance is incomplete. Finally, the finding that VZV-infected HKs die and suppress most proinflammatory cytokines and MMPs may explain the widespread death of these cells with unchecked viral spread due to ineffective recruitment of PBMCs.

V aricella zoster virus (VZV) produces varicella then establishes latency in ganglionic neurons, including cranial nerve ganglia. 1 With aging or immunosuppression, VZV reactivates and typically presents as zoster with 20% occurring in the ophthalmic distribution of the trigeminal nerve (herpes zoster ophthalmicus, HZO), 2,3 of which 23% have eye involvement. 4 Epithelial and stromal keratitis are corneal complications of VZV and can lead to vision loss after acute disease and during recurrence due to long-term corneal scarring and haze. 5 While multiple studies have found viral DNA and/or antigen in early and late VZV keratitis, it is unclear whether these represent ongoing infection or persistent viral DNA and/or antigens that elicit a cell-mediated immune response. For example, punctate epithelial keratitis is transient early in disease, but may progress into pseudodendrites composed of swollen, heaped up, poorly adherent epithelial cells. VZV has been cultured from these abnormal epithelial cells, suggesting that active viral replication plays a role. 6 Chronic epithelial pseudoden-drites or mucus plaques can occur up to 2 years after HZO [6][7][8][9][10][11] and contain VZV DNA. 10,12 In contrast, VZV DNA has been detected in the cornea up to one month after acute epithelial keratitis and both VZV DNA and antigen have been detected in corneas up to 10 years after zoster without clinical symptoms and signs. 13,14 Based on its rapid resolution upon corticosteroid treatment, stromal keratitis was thought to be an immune response to viral antigens within the corneal stroma, comprised of stromal fibroblasts underlying the epithelial cell layer. 15,16 However, Matoba and colleagues 17 found VZV DNA in the cornea of a patient with stromal keratitis, and a recent case of VZV keratitis demonstrated herpesvirus capsids in degenerative-appearing keratocytes, suggesting active VZV infection. 18 Supporting evidence for active VZV infection in stromal keratitis is that in patients who develop recurrent keratitis, the persistence of VZV antigens seems unlikely because one would expect persistent inflammation; rather, the recurrent keratitis is more likely due to recurrent deposition of VZV to stroma with reactive inflammation. 19 Thus, many questions remain concerning the role of active VZV infection in morphological changes seen in keratitis; the ability of corneal cells to support active VZV replication and transmit infectious virus to adjacent cells; the ability of VZV to persist in cornea, and; the role of VZV-induced inflammation in disease pathogenesis.
Since there is no animal model for VZV keratitis, we infected primary human corneal epithelial cells (HCECs) and stromal fibroblasts (keratocytes; HKs) with VZV and analyzed the cytopathology and inflammatory responses. Understanding the role of ongoing virus replication and inflammation in VZV keratitis can guide antiviral and corticosteroid therapy, mitigating ocular damage and vision loss produced by VZV.

Virus and Cells
The VZV Gilden strain (GenBank #MH379685) and primary HCECs and HKs from adult human cornea (ScienCell, Carlsbad, CA, USA) were used. Cells were confirmed by immunofluorescence antibody assay (IFA) and quantified by Fiji image processing software (https://fiji.sc/).
HCECs were seeded in corneal epithelial cell medium containing HCEC growth supplement/1% penicillin-streptomycin (20,000 cells/cm 2 ; ScienCell). Cells were co-cultured with lysates from uninfected or VZV-infected human fetal lung fibroblast (HFL; ATCC, Manassas, VA) for 24 hours (200 plaque-forming units [PFU]/cm 2 ), 20 medium was then replaced to eliminate the possibility that ongoing infection was due to residual lysate. To confirm that no HFLs contaminated the corneal cultures, HFL lysates were plated and did not grow out cells. HKs were seeded in basal fibroblast medium containing 2% fetal bovine serum (FBS)/1% fibroblast growth serum/1% penicillin-streptomycin (ScienCell). After 24 hours, medium was changed to quiescent basal medium supplemented with 0.1% FBS/1% penicillin-streptomycin and replenished every 72 hours for 7 days. Quiescent HKs were mock-or VZV-infected with HFL lysates. Cells were counted at 1, 3, 5, and 7 days postinfection (DPI) at the height of cytopathic effect (CPE).

Transmission of VZV from Corneal Cells to HFLs
VZV-infected HCECs were harvested at 3, 5, and 7 DPI; each sample was serially diluted onto 70% confluent monolayers of uninfected HFLs in DMEM F12 medium with 10% FBS (ATCC) to demonstrate virus transmission from HCECs to HFLs. Similarly, VZV-infected HKs at 3, 5, and 7 DPI were diluted onto HFLs. Three days later, HFLs were stained with 0.1% crystal violet and PFUs/mL were calculated.

Immune Cell Migration Assay
At 7 DPI, mock-and VZV-infected supernatants were assayed for their ability to attract PBMCs and neutrophils. Supernatants were loaded in a 96-well chemotaxis plate (Neuroprobe, Gaithersburg, MD, USA), 5-lm pore size filters placed on top, immune cells pipetted onto filters (50,000 cells/well), and plate incubated at 378C for 4 hours. Filters were removed; migrating immune cells were visualized by microscopy and quantified using ImageJ 3D Objects Counter (http://imagej. nih.gov/ij/; provided in the public domain by the National Institutes of Health, Bethesda, MD, USA). Positive control for PBMC and neutrophil migration was chemokine (C-C motif) ligand 2 (CCL2; Biolegend, San Diego, CA, USA) diluted in either HCEC or HK medium (20 pg/mL) and 50 ng/mL IL-8 (Biolegend), respectively. Negative controls for PBMC and migration assays were HCEC and HKs medium only. To determine specificity for IL8-induced neutrophil migration in infected supernatant, mouse monoclonal IL-8 blocking antibody was added at 1 lg/mL (aIL8, R&D Systems, Minneapolis, MN). Fold differences of VZV-infected cells and positive and negative controls were relative to mockinfected cells.

Statistical Analysis
Statistical analysis was performed using graphing software (GraphPad Prism; GraphPad, San Diego, CA, USA). Cytokine and MMP differences among mock-and VZV-infected cells were determined by multiple unpaired t-tests with a False Discovery Rate (q-value 0.05) using the two-stage step-up method of Benjamini et al. 23 Significant differences in cell count, viral titer, PBMC/neutrophil migration and MMP activity were determined using a 1-way ANOVA with a Tukey's multiple comparisons test. Alpha was set at 0.05 (*P < 0.05, **P < 0.01, ***P < 0.001).
HCECs and HKs exposed to uninfected HFL lysates had no CPE or VZV gB (red; Figs. 1A3, 1A4 and 1A7, 1A8, respectively). HCECs exposed to VZV-infected HFL lysates had a CPE and contained regions of cells expressing VZV gB (Figs. 1A5, 1A6) that accumulated and spread, indicating that HCECs can harbor replicating VZV that spreads to adjacent cells. VZV-infected HKs demonstrated an expanding CPE with VZV gB expression (Figs. 1A9, 1A10).
Aside from VZV-infected corneal cells having the capacity to spread VZV to adjacent cells of the same type, VZV-infected HCECs and HKs were tested for their ability to transmit VZV to another cell type by cell-to-cell spread. VZV-infected HCECs and HKs at 3, 5, and 7 DPI were cocultured with uninfected HFLs; PFUs were counted at 3 DPI. The PFUs observed were due to VZV-infected HFLs since input VZV-infected HCECs perish in DMEM F12 medium, and input infected HKs were present in low amounts and are morphologically distinguishable. VZV-infected HKs die and form cell clearings, whereas VZV-infected HFLs swell and form syncytia. Input VZV-infected HCECs at 3, 5, and 7 DPI significantly increased the PFU/mL in HFLs at 367 6 219, 2300 6 82, and 5250 6 204, respectively (

VZV-Infection Induced Cell Proliferation in HCECs and Cell Death in HKs
Phase-contrast imaging indicated that VZV-infected HCECs proliferate; VZV-infected HKs sloughed off.  HCECs and HKs were mock-or VZV-infected and analyzed at 7 days postinfection by phase microscopy and IFA using mouse anti-VZV glycoprotein E (gB) antibody. In mock-infected HCECs, phase images showed a cell monolayer without a CPE (A3) and no VZV gB (A4), whereas VZV-infected HCECs showed a CPE with areas of cell accumulation on phase-contrast (A5) that contained VZV gB by IFA (A6, red). In mock-infected HKs, phase images showed a monolayer of cells without CPE (A7) and no VZV gB (A8), whereas VZV-infected HKs showed a CPE on phase-contrast (A9) that corresponded to cells expressing VZV gB (A10, red). Blue color indicates cell nuclei. Mag 400X, A1 and A2; 100X, A3-A10. At 3, 5, and 7 days postinfection, infectious virus transmission from VZV-infected HCECs and HKs was measured by serially diluting cells onto uninfected HFLs.

Proinflammatory Cytokines and Infiltration of Immune Cells by VZV-Infected HCECs and HKs
At 7 DPI, compared to mock-, VZV-infected HCEC supernatant significantly increased IL-2, IL-6, IL-8, IL-10, IL-12p70, and IFNc; no significant differences in IL-1b or IL-13 was seen and IL-4 and TNF-a were not detected (Table 1: mock-and VZV-HCEC mean concentrations, cytokine detection range and fold differences in columns 2 through 5, respectively; fold differences in Fig. 3A, black bars; MFD 6 SEM; n ¼ 5). To verify cytokine expression, transcripts were analyzed by RT-qPCR and fold differences between VZV-to mock-infected HCECs determined (Table 1, column 6). Compared to mock-, the following transcript changes corresponded to the secreted cytokines measured in VZV-infected HCECs: no significant increase in IL-1b; significant increases in IL-6, IL-8, IL-12p70, and; significant increase in IL-10 compared to no detection of IL-10 transcript in mock. The following fold differences in transcripts did not correspond to that detected for secreted proteins: IL-2, IL-4, IL-13, IFN-c, and TNF-a.
At 7 DPI, compared to mock-, VZV-infected HK supernatant had significantly decreased concentrations of IL-1b, IL-2, IL-4, IL-6, IL-10, IL-12p70, IL-13, and IFN-c; significantly increased levels of IL-8 and; no detection of TNF-a ( Table 2: mock-and VZV-infected HK mean concentrations, cytokine detection range and fold differences in columns 2 through 5, respectively; fold differences in Fig. 3A, gray bars; MFD 6 SEM; n ¼ 5). To verify cytokine expression, transcripts from mock-and VZVinfected HKs were analyzed by RT-qPCR and fold differences between VZV-to mock-infected HKs determined (Table 2, column 6). Compared to mock-, IL-8 transcript was significantly increased and TNF-a transcript was not detected corresponding to their secreted cytokine levels during VZV infection. The following fold differences in transcripts did not correspond to that detected for secreted proteins: IL-1b, IL-2, IL-4, IL-6, IL-12p70, IL-10, IL-13 and IFN-c.

DISCUSSION
Herein, we show both HCECs and HKs can harbor productive VZV infection yet differ with respect to cytopathology, proinflammatory cytokine production and matrix metalloproteinase release, which may contribute to the clinical differences in VZV epithelial and stromal keratitis.
The question of whether HKs are able to synthesize infectious virions was recently raised following a case report of a patient who developed VZV stromal keratitis after zoster vaccination. 18 VZV particles appeared inside keratocytes; however, capsids were empty and it was unclear whether VZV was able to complete all the replication steps and release complete virions to infect adjacent keratocytes. 19 To our knowledge, only one previous study examined the effects of VZV infection on human corneal stromal fibroblasts in vitro, 24 which showed that VZV can infect stromal fibroblasts and that VZV deficient in functional ORF66 protein kinase expression is severely growth-impaired in these cells. Our study showed that both VZV-infected HCECs and HKs can synthesize infectious virus particles following exposure to VZV-infected HFL lysates, as demonstrated by the proliferation and ''piling up'' of HCECs expressing VZV glycoproteins, reminiscent of epithelial pseudodendrites, and by an expanding cytopathic effect in VZV-infected HKs that express VZV glycoproteins at the periphery of plaques, reminiscent of degenerating stromal cells seen by Jastrzebski and colleagues. 18 While conditioned supernatant from VZV-infected HCECs and HKs were unable to infect HFLs, indicating that cell-free virus was not produced, infected HCECs and HKs were able to transmit infectious virus particles to HFLs by cell-to-cell spread, demonstrating that HCECs and HKs are permissive to VZV infection.
Compared to mock-, VZV-infected HCECs secreted significantly higher levels of IL-2, IL-6, IL-8, IL-10, IL-12p70 and IFN-c in the conditioned medium; no significant differences in IL-1b or IL-13 were measured. IL-4 and TNF-a were not detected. However, only IL-1b, IL-6, IL-8, IL-10, and IL-12p70 were confirmed at the transcript level. The reason(s) for the disparities in transcript and secreted cytokine levels remain to be determined. Consistent with the increased proinflammatory cytokine levels, VZV-infected HCEC supernatant recruited PBMCs and neutrophils in a migration assay, consistent with inflammatory infiltrates observed in VZV-infected corneal epithelium of patients. The number of proinflammatory cytokines secreted by VZV-infected HCECs is greater than that previously reported in VZV-infected human brain vascular adventitial fibroblasts (HBVAFs), vascular smooth muscle cells (HBVSMCs), perineurial cells (HPNCs) and HFLs. 25 During VZV infection, IL-8 was increased in all four of these primary human cell cultures; IL-6 was increased in all but HBVSMCs, while IL-2 was increased only in HPNCs. However, there were no increases in IL-4, IL-10, IL-12p70 or IFN-c, possibly due to the different VZV strains used (VZV Ellen strain in the previous report and the Gilden strain herein) or differential cell type specific responses to infection. The cytokines secreted from VZV-infected HCECs differ from those secreted from herpes simplex-1 (HSV-1; KOS strain)-infected HCECs. While both viruses led to an increase in IL-6 and IL-8, HSV-1-infected HCECs also showed an increase in TNF-a and IFN-b, 26 which was not detected and not measured in VZV-infected HCECs, respectively. In a study of HSV-1-infected human cornea organotypic cultures, IL-6 and IL-1b were not significantly increased, but TNF-a was modestly increased; IL-17 was not detected. 27 These differential cytokine responses in HCECs against two different viruses may contribute to differences in the level of inflammation and morphological changes (i.e., ulcerations and perforation, seen in VZV and HSV-1 epithelial keratitis).
In VZV-infected HCECs, levels of secreted MMP-1 and MMP-9 were significantly increased, unlike the HSV-1-induced increases in MMP-2 and MMP-9 in HCECs. 28 MMP-1 is an enzyme that cleaves collagen that is present in the extracellular matrix of the cornea. 29 Similarly, MMP-9 degrades denatured collagen. 30 Both MMPs can mediate epithelial-stromal interactions, immune cell infiltration, corneal damage and corneal ulceration. 31 Since the ability of MMPs to cleave extracellular matrix depends on the balance of activated MMPs and tissue inhibitors of MMPs, an MMP activity assay was used that showed increased extracellular matrix substrate cleavage. This increase in MMP activity associated with VZV infection of HCECs could lead to the loss of epithelial layer integrity and promote an environment that is more permissive to immune cell infiltration and epithelial cell migration and proliferation.
Compared to mock-infected HKs, VZV-infected HKs only secreted significantly higher levels of IL-8 that was confirmed at the transcript level, similar to HSV-1-induced IL-8 secretion in keratocytes; 32 conditioned supernatant recruited neutrophils in the migration assay. Specificity for increased IL-8 in HK supernatant acting as a neutrophil attractant is seen by decreased infiltration when IL-8 was neutralized with an anti-IL-8 antibody. In addition, MMP-1, -2, -3, and -9 were suppressed in VZV-infected HKs. The inability of VZV-infected HKs to secrete proinflammatory cytokines, recruit PBMCs and cleave extracellular matrix, in contrast to the ability of VZVinfected HCECs to create a proinflammatory environment, may be due to differences in developmental origins: HCECs are embryonic ectoderm-derived, whereas HKs are neural crestderived. 33 Overall, our data shows that VZV-infected HCECs recapitulate the clinical findings seen in patients with VZV keratitis, including epithelial cell proliferation, pseudodendrite formation and increased proinflammatory cytokines and MMPs, providing an in vitro model for VZV infection of HCECs. The proliferation and persistence of VZV-infected HCECs suggest that these cells may serve as viral reservoirs if immune clearance is incomplete. Finally, the VZV-induced death of HKs and suppression of most proinflammatory cytokines and MMPs may explain the widespread death of these cells with unchecked viral spread due to ineffective recruitment of PBMCs.