Unique molecular signatures of antiviral memory CD8+ T cells associated with asymptomatic recurrent ocular herpes

The nature of antiviral CD8+ T cells associated with protective and pathogenic herpes simplex virus type 1 (HSV-1) infections remains unclear. We compared the transcriptome, phenotype, and function of memory CD8+ T cells, sharing the same HSV-1 epitope-specificities, from infected HLA-A*0201 positive symptomatic (SYMP) vs. asymptomatic (ASYMP) individuals and HLA-A*0201 transgenic rabbits. Compared to higher frequencies of multifunctional effector memory CD8+ TEM cells in ASYMP individuals, the SYMP individuals presented dysfunctional CD8+ TEM cells, expressing major exhaustion pathways. Compared to protected ASYMP HLA transgenic rabbits, the trigeminal ganglia of non-protected SYMP HLA transgenic rabbits had higher frequencies of dysfunctional tissue-resident CD8+ TRM cells. Moreover, blockade of T cell exhaustion pathways restored the function of CD8+ T cells, reduced virus reactivation, and diminished recurrent disease in HLA transgenic rabbits. These findings reveal unique molecular signatures of protective CD8+ T cells and pave the way for T-cell-based immunotherapy to combat recurrent ocular herpes.

Crosstalk between HSV-1 and CD8 + T cells restrains virus reactivation within the latently-infected TG 2,3,5,6,27 . The adaptive immunity plays a role in establishing latency as a high number of activated CD8 + T cells expressing a late effector memory phenotype were found to reside in latently infected TG. This indicates that activated late effector memory CD8 + T cells may control HSV-1 latency 7 . Functional antiviral CD8 + T cells, specific to multiple HSV-1 epitopes, are selectively retained and patrol latently-infected TG 2,5,6 . In contrast, dysfunctional HSVspecific CD8 + T cells appeared unable to control virus reactivation from latently infected TG, thus contributing to frequent virus shedding in tears as well as to frequent and severe recurrent herpes in SYMP individuals 5,28 .

Figure 1.
Differential gene expression in HSV-specific CD8 + T cells from HSV-1 infected symptomatic vs. asymptomatic individuals. (a) Experimental design and validation of differentially expressed genes in CD8 + T cells sharing the same HSV-1 epitope-specificities, from SYMP and ASYMP individuals. CD8 + T cells specific to HLA-A*0201-restricted HSV-1 gB 561-567 and VP11/12 702-710 epitopes were sorted from HLA-A*0201positive SYMP and ASYMP individuals, using specific tetramers. Total RNA was extracted from each clone of epitope-specific CD8 + T cells, and whole transcriptome analysis was performed using bulk RNA sequencing to determine the levels of expression of 25,638 genes. (b) Frequencies of CD8 + T cells specific to HLA-A*0201restricted HSV-1 gB 561-567 and VP11/12 702-710 epitopes detected by FACS in SYMP vs. ASYMP individuals. (c) Heatmap is showing 772 differentially expressed genes among SYMP and ASYMP individuals. (d) Heatmap showing statistically significant pathways that are affected in HSV-specific CD8 + T cells from SYMP vs. SYMP individuals. Parametric Gene Set Enrichment Analysis (PSGEA) method was applied based on data curated in Gene Ontology and KEGG. Pathway significance cut-off with a false discovery date (FDR) ≥ 0.2 was applied. (e) Bulk RNA heatmap comparing differentially expressed CAM pathway associated T cell co-stimulatory and T cell exhaustion genes in HSV-specific CD8 + T cells from SYMP vs. SYMP individuals.
Scientific RepoRtS | (2020) 10:13843 | https://doi.org/10.1038/s41598-020-70673-z www.nature.com/scientificreports/ Total RNA samples were isolated from sorted HSV-specific CD8 + T cells and subsequently processed for bulk RNA sequencing to screen the whole human transcriptome comprising 25,638 genes. Of this pool of genes, 20,126 genes were found with a minimum count per million (CPM) value ≥ 0.5. Our gene expression analysis revealed 772 genes to be statistically significant (FDR ≥ 0.1 and fold change ≥ 2) and differentially expressed among CD8 + T cells that share the same HSV-1 epitope-specificities, from SYMP vs. ASYMP groups (Fig. 1c). Of these genes, 583 genes were up-regulated, and 189 genes were down-regulated in HSV-specific CD8 + T cells from ASYMP individuals compared to SYMP patients ( Supplementary Fig. S1a, upper panel). Correlation matrix and Pearson's correlation coefficient confirmed a high degree of relatedness in the pattern of gene expression between SYMP and ASYMP individuals ( Supplementary Fig. S1a, middle panel). The Volcano plot showed significant log2 fold changes and -log10 (FDR) of each of the differentially expressed genes in HSV-specific CD8 + T cells from ASYMP vs. SYMP patients ( Supplementary Fig. S1a, lower panel). Further pathway enrichment analysis of the 772 differentially expressed genes between SYMP vs. ASYMP groups revealed a significant up-regulation of the cell adhesion molecule (CAM) pathway comprising major T cell exhaustion genes in HSV-specific CD8 + T cells from the SYMP group (P = 3.2e−04) (Fig. 1d). A bulk RNA sequencing specific heatmap confirmed the significant upregulation of CAM pathway-specific gene expression (i.e., CEACAM8, NRCAM, LAMA1, SELE, and NLGN3) and T cell exhaustion genes (i.e., PD-1, LAG-3, PSGL-1, CTLA-4, TIM3, and TIGIT) in HSV-specific CD8 + T cells from the SYMP patients compared to HSV-specific CD8 + T cells from the ASYMP individuals (Fig. 1e). These results indicate that an upregulation of CAM pathway associated genes along with major T cell exhaustion molecules in HSV-specific CD8 + T cells from SYMP patients is associated with frequent and severe recurrent ocular herpetic disease.
Due to the ethical and practical limitations in obtaining human TG samples from HSV-1-infected, SYMP and ASYMP individuals, the remainder of this study utilized our established HLA-A*0201 transgenic rabbit (HLA Tg rabbits) model of ocular herpes, which develops spontaneous virus reactivation, virus shedding in tears, and symptomatic recurrent ocular herpetic disease, as occurs in humans.

Blockade of PD-1 and LAG-3 immune checkpoint reduces recurrent ocular herpes infection
and disease in latently infected symptomatic HLA transgenic rabbits. Since up-regulation of genes for the PD-1 and LAG-3 exhaustion pathways in HSV-specific CD8 + T cells is associated with symptomatic herpes in humans, it was of interest to determine whether the blockade of PD-1 and LAG-3 immune checkpoint pathways would reduce virus reactivation and ease symptomatic recurrent ocular herpes.
As illustrated in Fig. 2a, HLA Tg rabbits (n = 20) were infected with 2 × 10 5 pfu of HSV-1 (McKrae Strain). Half of the rabbits (Group-1; n = 10) were treated with a combination of blocking α-PD-1 and α-LAG-3 mAbs, injected intravenously (i.v.) on days -3, -5, -7 before infection and then on days 3, 5, and 7 post-infection (p.i.) at 200 μg/dose. The other half of the rabbits (Group-2; n = 10) received saline injections (untreated controls). Eye swabs were collected daily for 5 days p.i., and recurrent ocular herpetic disease was monitored 15 days p.i. Rabbits were subsequently segregated into (1) asymptomatic (ASYMP) rabbits, with no apparent recurrent ocular herpetic disease, and (2) symptomatic (SYMP) rabbits with higher rates and severe recurrent ocular herpetic disease. The detailed characteristics of the SYMP and ASYMP HLA Tg rabbits used are described in "Methods". The α-PD-1/α-LAG-3 mAbs-treated group had the most ASYMP animals, with significantly less virus titers in the eyes (P < 0.04, Fig. 2b) and no apparent recurrent corneal herpetic disease (Fig. 2c). In contrast, the untreated group had the most SYMP animals, which showed a significantly higher level of virus replication in eyes associated with severe recurrent ocular herpetic disease (P < 0.04, Fig. 2b,c).
It was next determined whether the observed reduction of virus replication in the eyes of ASYMP HLA Tg rabbits treated with PD-1 and LAG-3 immune checkpoint blockade was associated with a reduction in virus reactivation locally in the TG, the site of latent HSV-1 infection/reactivation cycles. The viral load during exvivo TG reactivation is often determined using RT-PCR as a more sensitive alternative to plaque assay 34,36,37 . The excised latently infected TG from the same rabbit was digested and equally distributed in control, and the mAb treated wells, so the possibility of variation in DNA copy numbers due to variation in the level of latency does not arise. The amount of reactivated virus detected ex vivo in treated and untreated HSV-1 infected TG explants was determined daily for 8 days post-treatment. As shown in Fig. 2d, the HSV-1 infected TG explants incubated with α-PD-1, or α-LAG-3 mAbs showed a significant decrease of the reactivated virus on day 8 following the blockade (P < 0.05). The blockade of the LAG-3 pathway (P = 0.04) appeared slightly better in reducing virus reactivation from TG compared to blockade of the PD-1 pathway (P < 0.05).
Altogether, these results demonstrate that blockade of PD-1 and/or LAG-3 immune checkpoints reduced recurrent ocular herpes infection and disease in vivo in latently infected HLA Tg rabbits. The observed reduction of virus replication and recurrent ocular herpetic disease in PD-1 and LAG-3 treated HLA Tg rabbits was associated with a significant reduction in virus reactivation locally in the HSV-1 infected TG.
Since the HLA Tg rabbit model develops human-like CD8 + T cell responses to HLA-A*0201 restricted epitopes 3,21,38 , it offers the possibility to determine the phenotype, function, and transcriptome of TG-resident HLA-A*0201 restricted HSV-1 epitopes-specific CD8 + T cells associated with symptomatic vs. asymptomatic recurrent ocular herpes.
Single-cell RNA sequencing revealed CD8 + T cells and monocytes as the most significant CD45 + leukocytes infiltrating trigeminal ganglia of "protected" asymptomatic HLA Tg rabbits. TGderived CD45 + leukocytes were sorted by fluorescence-activated cell sorting (FACS) from (1) SYMP HLA Tg rabbits (n = 4); and (2) ASYMP HLA Tg rabbits (n = 4), as illustrated in Figs. 2a and 3a and subsequently processed for single-cell RNA sequencing (scRNASeq) using the 10 × Genomics platform. Eight cell populations Scientific RepoRtS | (2020) 10:13843 | https://doi.org/10.1038/s41598-020-70673-z www.nature.com/scientificreports/ were observed among the sorted CD45 + leukocytes from TG of ASYMP and SYMP groups: CD4 + T cells, CD8 + T cells, NK cells, B cells, macrophages, monocytes, granulocytes, and dendritic cells (Fig. 3b). CD8 + T cells (defined by CD8A gene) and monocytes (defined by CD14 gene) represented the most frequent CD45 + leukocyte populations in the TG of "protected" ASYMP HLA Tg rabbits compared to TG of "nonprotected" SYMP HLA Tg rabbits. We detected a total of 198 (26.7%) CD8 + T cells per TG in ASYMP HLA Tg rabbits, while only 116 (15.6%) CD8 + T cells were found in the TG of SYMP HLA Tg rabbits (Fig. 3c). After CD8 + T cells, the next most significant cell population detected in the TG of ASYMP HLA Tg rabbits were the monocytes at a frequency of 20.4% of the total cell population in comparison to only 4.3% among the SYMP HLA Tg rabbits (Fig. 3c). A relatively higher mRNA copy number was also found in the TG of ASYMP HLA Tg rabbits (Fig. 3d).
These results indicate that: (1) anti-PD-1 and anti-LAG-3 blockade induced compartmental remodeling of TG-infiltrating immune cells of HSV-infected HLA Tg rabbits; and (2) expansion of CD8 + T cells and monocytes in the TG of HSV-1 infected asymptomatic HLA Tg rabbits is associated with reduced virus reactivation and less recurrent ocular herpetic disease.
The pathway enrichment analysis carried out on the basis of differentially expressed genes among scRNA-Seq, and bulk RNA-Seq demonstrated downregulation of prominent T cell exhaustion molecules present in the cell adhesion molecules (CAMs) pathway (P = 0.04) (Accession no: ocu04514) among the ASYMP group (Fig. 4d). The CAMs pathway is comprised of PD-1-, LAG-3-, TIM3-, TIGIT-, CTLA4-, PSGL-1 genes. In contrast, the T cell activation pathway (P = 0.03) (Accession no: GO: 0042110), the chemokine-chemokine receptor signaling pathway (P = 0.04) (Accession no: ocu04062), and the cytokine-cytokine receptor interaction (P = 0.04) (Accession no: ocu04060) were all up-regulated in TG-resident HSV-specific CD8 + T cells from ASYMP HLA Tg rabbits (Fig. 4d). These results indicate that similar to SYMP patients above, in SYMP HLA Tg rabbits, there was an up-regulation of T cell exhaustion pathways in HSV-specific CD8 + T cells associated with frequent and severe recurrent ocular herpetic disease.
High frequencies of HSV-specific memory CD8 + t eM and CD8 + t RM cell subsets, with upregulated t-cell activation pathways, detected in tG of asymptomatic HLA tg rabbits. We next compared the frequencies of the three major subsets of memory CD8 + T cells that share the same HSV-1 epitopespecificities and are present in the TG of ASYMP vs. SYMP HLA Tg rabbits. As illustrated in Fig. 2a above, HLA Tg rabbits (n = 8) were first infected with 2 × 10 5 pfu of HSV-1 (McKrae Strain).
The CAMs pathway-specific genes were significantly down-regulated in TG-resident CD103 + CD69 + CD8 + T RM cells from ASYMP HLA Tg rabbits compared to TG-resident CD103 + CD69 + CD8 + T RM cells, with same epitopesspecificities, from SYMP HLA Tg rabbits. Major T cell exhaustion molecules like PD-1 (P = 0.005, logFC = − 2.02), LAG-3 (P = 0.04, logFC = − 4.15), CTLA4 (P = 0.02, logFC = − 1.78), were found to be downregulated among . Differential gene expression in HSV-specific CD8 + T cells from trigeminal ganglia of HSV-1 infected symptomatic vs. asymptomatic HLA Tg rabbits. (a) Experimental design and validation of differentially expressed genes in CD8 + T cells sharing the same HSV-1 epitope-specificities, from SYMP and ASYMP HLA Tg rabbits. CD8 + T cells specific to HLA-A*0201-restricted HSV-1 gB 561-567 , VP11/12 702-710, and gD 53-61 epitopes were sorted from TG of HLA-A*0201-positive SYMP and ASYMP HLA Tg rabbits, using specific tetramers. Total RNA was extracted from each clone of epitope-specific CD8 + T cells, and whole transcriptome analysis was performed using bulk RNA sequencing to determine the levels of expression of 23,669 rabbit genes (OryCun2.0 (GCA_000003625.1). (b) Frequencies of CD8 + T cells specific to HLA-A*0201-restricted HSV-1 gB 561-567 , VP11/12 702-710, and gD 53-61 epitopes detected by FACS in TG of HLA-Tg rabbits. (c) The heatmap is showing the most significant 2,879 differentially expressed genes among SYMP and ASYMP HLA Tg rabbits. Genes with minimum count per million (CPM) ≥ 0.5 were used for obtaining the transformed counts data for clustering using regularized log (rlog). (d) Bulk RNA heatmap shows the pathways that are different among ASYMP and SYMP HLA Tg rabbits. Genes differentially expressed in both single-cell RNA sequencing and bulk RNA sequencing were considered for pathway analyses. Parametric gene set enrichment analysis (PSGEA) method based on data curated in Gene Ontology and KEGG was applied. Pathway significance cut-off with a false discovery date (FDR) ≥ 0.2 was applied.  Table S2). Whereas, Inducible T-cell Co-Stimulator (ICOS) gene (P = 0.04, logFC = 4.03), and PRDM1 (BLIMP1) (P = 0.01, logFC = 3.1) known to be associated with T cell activation and differentiation were observed with a higher expression in the TG resident T RM cells among ASYMP HLA Tg rabbits ( Fig. 5a; Supplementary Table S2). PD-1 and LAG-3 genes were expressed in 21.6% and 18.9% of TG-resident CD103 + CD69 + CD8 + T RM cells in the SYMP HLA Tg rabbits compared to only 1.16% and 2.3% of TG-resident CD103 + CD69 + CD8 + T RM cells in the ASYMP HLA Tg rabbits (Fig. 5b).
When differential gene expression analysis was carried out between ASYMP vs. SYMP groups using bulk RNA sequencing, TIGIT (P = 0.05, logFC = − 2.70), CTLA4 (P = 0.01, logFC = − 1.04), PD-1 (P = 1.01E−06, logFC = − 5.58) genes were found to be significantly down-regulated along with other T cell exhaustion molecules (Fig. 5c, right panel; Supplementary Table S2). The bulk RNA sequencing-based on differential gene expression (DGE) analysis performed on TG-resident memory CD103 + CD69 + CD8 + T RM cells, with the same HSV-1 epitope-specificities, confirmed a heightened expression level of several genes associated with the T cell activation pathway in "protected" ASYMP HLA Tg rabbits (Fig. 5c). In contrast, significantly higher expression levels of several genes associated with the T cell exhaustion were confirmed by bulk RNA sequencing in "nonprotected" SYMP HLA Tg rabbits (Fig. 5c). Importantly, the single-cell RNA sequencing-based transcriptome of CD45 + leukocytes clearly showed increased frequencies of memory CD8 + TRM and CD8 + TEM cell subsets in the TG of "protected" ASYMP HLA Tg rabbits.
Interestingly, Fig. 5d shows upregulation of CD62L protein expression at the surface of the HSV-specific CD8 + T-cells from SYMP compared to ASYMP individuals, whereas the single-cell and bulk RNAseq data panels on Fig. 5a-c show downregulation of the CD62L mRNA expression in SYMP subjects. While the protein expression for the other markers (CD69, CD44, CD103) follows their mRNA expression, this does not seem to be the case with CD62L. This may be due to post-transcriptional/translational modification/regulation. Thus, it appears that the expression of CD62L on the surface on the T-cell (at the protein level) is regulated not only at the transcriptional level but also at a post-transcriptional level. In previous reports, it has been shown that the loss of CD62L by the activated T-cells can be regulated by the cleavage of CD62L from the cell membrane at K283-S284 by a disintegrin and metalloprotease ADAM17 in a process called CD62L shedding 39-42 . Heightened expression of genes for t cell attracting chemokines/receptors and t cell maintaining cytokines/receptors detected in TG-resident CD8 + t RM cells from asymptomatic HLA tg rabbits. The higher frequencies of HSV-specific CD8 + T RM cell subset observed in the TG of "protected" ASYMP HLA Tg rabbits prompted us to further compare in CD8 + T cells TG from SYMP vs. ASYMP HLA Tg rabbits the differential gene expression profile of genes for cytokines, chemokines and their receptors that are involved in T cell homing and T cell maintaining.
Altogether, these results suggest that a heightened activation of T cell attracting chemokines/receptors and T cell keeping cytokines/receptors pathways may lead to increased infiltration/retention of protective antiviral CXCR3 + CD8 + T RM cells observed in the TG of "protected" ASYMP HLA Tg rabbits, as illustrated in Fig. 7.   Figure 7. TG-resident HSV-specific memory CD8 + T RM cells downregulate the T cell exhaustion associated pathway and confer protection from ocular herpes in HSV-1 infected asymptomatic humans and HLA transgenic rabbits. (1) Upon exposure to stressors, the HSV-1 enters into the cornea and travels through neurons to Trigeminal ganglia. (2) Following primary HSV-1 infection, the vast majority (up to 95%) of antiviral effector CD8 + T cells die, leaving behind only about 5% of CD8 + T cells destined to differentiate into a heterogeneous pool of memory CD8 + T cells. (3) The effector memory (T EM ) and tissue-resident memory (T RM ) CD8 + T-cell subsets are found mainly in the HSV-infected but "naturally protected" asymptomatic subjects, whereas the lymphoid organ-resident central memory (T CM ) CD8 + T cell subsets are mainly present in non-protected Symptomatic subjects. (4) Reduced viral reactivation was observed among asymptomatic subjects possessing a higher frequency of CD8 + T RM cells resulting in a less severe herpes disease. (5) The findings study suggests that by blocking immune checkpoints, there is a reduced expression of T cell exhaustion molecules (PD-1, LAG-3, PSGL-1, CTLA-4, TIM3, and TIGIT) and T cell exhaustion associated Cell Adhesion Molecule pathway and increased retention of CD8 + T RM cell population in asymptomatic subjects. This memory CD8 + T cell population mediates recall responses and halts attempts of virus reactivation in the infected TG, thus accelerating viral clearance. More-so, reduced expression of T cell exhaustion pathway also gives rise to higher expression of genes associated with T cell function (CD107, IFN-γ), T cell homing (CXCR3, CCR7), and T-cell keeping (IL7R, IL15R). This helps in reducing the ocular herpes infection and recurrent herpetic disease.

Discussion
We report significant differences in the phenotype, function, and molecular signatures of HSV-specific CD8 + T cells that are associated with symptomatic vs. asymptomatic recurrent ocular herpes infection in humans and HLA Tg rabbits. We were not able to detect tetramer-positive HSV-specific CD8 + T cells in the herpes seronegative control individuals. Due to the lack of HSV-specific CD8 + T cells, we were not able to include these cells in the RNA sequencing experiments as baseline control. Compared to multifunctional blood-derived CD8 + T EM cells in ASYMP individuals, SYMP individuals displayed high frequencies of dysfunctional HSV-specific blood-derived effector memory CD8 + T EM cells, expressing genes of major exhaustion pathways. Similarly, the trigeminal ganglia of "protected" ASYMP HLA Tg rabbits contain a high number of functional CD8 + T cells with phenotypic and functional features of tissue-resident memory T (T RM ) cells that expressed high levels of effector proteins. In contrast, TG of "non-protected" SYMP HLA Tg rabbits with increased virus reactivation from TG and increased severity of recurrent ocular herpes, contain a high number of dysfunctional HSV-specific CD8 + T EM cells and tissue-resident memory CD8 + T RM cells, with significant up-regulation of PD-1, LAG-3, TIGIT and TIM-3 genes. As exhaustion markers are found to be expressed on circulating CD8 + T cells in humans, it is likely that a reservoir of circulating memory CD8 + T cells immigrate into nonlymphoid tissues, including TG, which may trigger the observed exhaustion. Similar studies by Klonowski et al., using parabiosis experiments, have demonstrated that circulating memory CD8 + T cells, in a resting host, comprised a pool of cells, which were capable of relocating into nonlymphoid tissues 43 . Due to the ethical and practical limitations in obtaining human TG samples from HSV-1-infected SYMP and ASYMP individuals, it was not possible to study exhaustion of tissue-resident HSV-specific CD8 + T cells in humans. Instead, we used our established HLA Tg rabbits' model of ocular herpes to demonstrate increased expression of "functional exhaustion" markers on TG-resident CD8 + T cells in SYMP HLA Tg rabbits. The report suggests that antiviral CD8 + T EM and CD8 + T RM cells contribute to herpes immunosurveillance in latently infected TG and their number and function are key targets of a boost by a therapeutic vaccine combined with immune checkpoints blockade (Fig. 7). Beside the central antiviral CXCR3 + CD8 + T RM cells observed in the TG of asymptomatic individuals and rabbits, it is likely that peripheral corneal T cell immunity contribute to protection in asymptomatic individuals and rabbits. Although our human transcriptome profiling reveals unique molecular characteristics of blood-derived "protective" antiviral CD8 + T cells associated with asymptomatic recurrent ocular herpes, it is important to note that the molecular characteristics of blood-derived CD8 + T cells may not reflect those of tissue-resident HSV-specific CD8 + T cells. In this study, we used our established HLA-A*0201 transgenic rabbit model of recurrent ocular herpes, which develops spontaneous virus reactivation, virus shedding in tears, and symptomatic recurrent ocular herpetic disease, as occurs in humans. Besides, the HLA Tg rabbit model develops human-like CD8 + T cell responses to HLA-A*0201 restricted epitopes 3,21,38 , and as such offered the possibility to determine the phenotype, function, and transcriptome of TG-resident HLA-A*0201 restricted HSV-1 epitopes-specific CD8 + T cells that would be associated with symptomatic vs. asymptomatic recurrent ocular herpes. This constitutes the first report deciphering the differential gene expression pattern of CD8 + T cells in general, and of TG-resident CD8 + T cell subsets in particular, in the rabbit model during symptomatic and asymptomatic herpes infection. We observed increased frequencies of functional CD8 + T RM cells in TG tissues associated with protection from recurrent ocular herpes in ASYMP HLA Tg rabbits, and identified a conserved tissue signature and activation states for those CD8 + T RM cells conserved across many HSV-1 epitopes. Moreover, we demonstrated that high frequencies of functional CD8 + T cells in TG explants positively correlated with less virus reactivation ex vivo. Thus, while the antiviral CD8 + T cells play a crucial role in reducing HSV-1 reactivations from latently infected TG, the small number of functional CD8 + T RM cells present in latently infected TG of SYMPP rabbits (and presumably in TG of SYMP humans) may not be sufficient to prevent virus reactivation 23,44 . In such a scenario, to increase the number of functional TG-resident CD8 + T cells, we used the antagonist mAbs approach to block T cell exhaustion pathways. We demonstrate, in both ex vivo in TG explants and in vivo in HSV-1 infected HLA Tg rabbits, that blockade of PD-1 and/or LAG-3 immune checkpoints significantly boosted T-cell immunity associated with a significant reduction in virus reactivation from TG explants and with protection from recurrent ocular herpes infection and disease in HSV-1 latently infected HLA Tg rabbits. To the best of our knowledge, this is a first report using scRNASeq in the rabbit model, providing transcriptional signatures across TG-resident CD45 + leukocytes and showed eight transcriptionally distinct major leukocyte subpopulations, CD8 + T cells, CD4 + T cells, NK cells, B cells, dendritic cells, monocytes, granulocytes, and macrophages. Among these cell populations, we identified increased numbers of CD8 + T cells and monocytes as a cellular signature associated with a reduction in recurrent ocular herpes infection and disease in latently infected HLA Tg rabbits.
The profile of biological pathways and genes differentially expressed during herpes infection and disease will aid in the development of therapeutic herpes vaccine. In this study, we found the memory CD8 + T EM cells and CD8 + T RM cells from TG of "protected" ASYMP HLA Tg rabbits expressed high levels of T cell attracting chemokines/chemokine receptor genes including: CCL5, CCL14, CCR7, and CXCR3. The bulk-RNA sequencing also revealed an up-regulation of CCL2-, CCL4-, CCL5-, CXCL9-, CXCL10-, CXCL11-, CCR7-, CXCR3-, and VCAM-1-like genes associated with the chemokine signaling pathway. The HSV-specific CD8 + T EM cells and CD8 + T RM cells infiltrating the TG of protected HLA Tg rabbits heightened expression of inflammatory chemokines CCL2, CCL4, CCL5, CXCL9, and CXCL10 45,46 . Recent studies reporting high transcript levels of CCL4 in quiescent T RM cells associate the phenomenon with the production of such chemokines by T RM cells 47 . These chemokines may subsequently trigger the attraction of neutrophils and monocytes like innate myeloid cells to the site of infection, and further enhance the immune response 48,49 . Similarly, memory CD8 + T cells from TG of "protected" ASYMP HLA Tg rabbits expressed high levels of CXCR3, a receptor known to be associated with recruitment of T cells to infected tissues in response to CXCL9-and CXCL10 chemokines 50 . In contrast, a Scientific RepoRtS | (2020) 10:13843 | https://doi.org/10.1038/s41598-020-70673-z www.nature.com/scientificreports/ down-regulation of the CAM pathway, which is comprised of major T cell exhaustion molecules, in HSV-specific CD8 + T cells was associated with asymptomatic herpes in both humans and HLA Tg rabbits. We further observed an up-regulation of genes IL7R, IL15R, IL17R, IL6ST (CD130), and TNFRSF9 (CD137) cytokines and cytokine receptors in the CD8 + T RM cells from TG of ASYMP HLA Tg rabbits. In other systems, T RM cells associated to protection upregulate receptors for IL-7 and IL-15 51,52 . Moreover, IL7R and IL15R contribute to antigen-independent maintenance of T RM cells 51,52 and the continued presence of IL-15 aids in long-term maintenance of T RM cells in infected tissues 52 . TNFRSF9 (CD137) helps clonal expansion, survival and development of T cells. In both mice and humans, T RM cells across different tissues express higher levels of transcript for the pro-inflammatory cytokines compared to their circulating counterparts 53,54 . Once stimulated by the virus, such as after HSV-1 reactivation from TG, the local CD8 + T RM cells release pro-inflammatory cytokines, IFN-γ, TNF-α, and IL-2, which amplifies the present several other tissue-resident immune cell, such as the monocytes present with high frequencies in TG of protected HLA Tg rabbits.
We detected an increased frequency of functional HSV-specific CD103 + CD69 + CD8 + T RM cells with activation pathway-specific genes, such as CD69, CD62L (SELL), and CD44, in the TG of "protected" ASYMP HLA Tg rabbits compared to TG of "non-protected" SYMP HLA Tg rabbits. Our bulk RNA sequencing results also revealed up-regulation of the T cell activation pathway with increased expression of CD69, CD62L (SELL) and IFN-γ genes in ASYMP HLA Tg rabbits. This finding is in light of previous reports of increased frequencies of CD8 + T RM cells expressing CD69 and/or CD103 in protected tissues of many systems [54][55][56][57][58][59][60][61] . Interestingly, the single cell and bulk RNA-sequencing data showed increased expression of central memory marker CD62L (L-selectin) in TG-resident memory cells. It appears that the expression of CD62L on the surface on the T-cell (at the protein level) is regulated at both the transcriptional and post-transcriptional level. CD62L expressed on T cells plays a crucial role during recirculation of central memory T cells (T CM cells), especially into peripheral lymph nodes and are also involved in mediating entry of T CM cells into infected non-lymphoid tissues, such as TG 62 . T cells in non-lymphoid organs such as lung, intestine and liver have been shown to have more CD62L expression 63 . Thus, the presence of CD62L (+) CD8 + T CM cells in TG suggests a migration of HSV-specific T CM cells from periphery into infected TG. The HSV-specific CD69 + CD8 + T RM cells from TG of "protected" ASYMP rabbits with less virus reactivation appeared to be multifunctional, proliferating and producing high levels of CD107 and IFN-γ. In agreement with previous reports that rapid control of viral replication is associated to the abundant IFN-γ production by cytotoxic CD8 + T RM cells 64 . Local CD8 + T RM -derived IFN-γ is responsible for increasing the expression of the homing molecule vascular cell adhesion molecule 1 (VCAM-1) on endothelial cells, and is also known to enhance the recruitment of other memory T cells, such CD8 + T RM cells from circulation 45,46 .
We found RNA sequencing of CD8 + T cells specific to several HSV-1 epitopes provide a high-resolution map that defined T cell states associated with protective vs. pathogenic recurrent ocular herpes infection 31 . Using RNA profiling in humans and rabbits, we identified a module of co-inhibitory receptors that includes genes for several known co-inhibitory receptors (PD-1, TIM-3, LAG-3 and TIGIT). Sustained overexpression of co-inhibitory receptors, such as PD-1 and LAG-3, on antiviral CD8 + T cells promote T cell dysfunction or exhaustion, leading to impaired ability to clear virus reactivation, as reported in other systems [65][66][67][68][69][70][71][72] . Such T cell dysfunction may contribute to suboptimal herpes immunity and subsequently to a symptomatic virus shedding seen in SYMP patients and SYMP HLA Tg rabbits 31,44 . Thus, even in the presence of a higher frequency of CD8 + T RM cells in the TG, as an immune evasion strategy the HSV-1 can still manage to reactivate by inducing exhaustion of these TG-resident CD8 + T RM cells 3,6,34 . We phenotypically and functionally validated two co-inhibitory receptors PD-1 and LAG-3 in HSV-1 infected HLA Tg rabbits, where the TG-resident CD8 + T cells exhibited transcriptional features of exhausted CD8 + T cells. We observed an increased frequency of LAG-3 + CD8 + T cells and PD-1 + CD8 + T cells in "non-protected" SYMP HLA Tg rabbits in comparison to increased frequency of functional LAG-3 -CD8 + T cells and PD-1 -CD8 + T cells in TG of "protected" ASYMP HLA Tg rabbits. It is likely that similar to "non-protected" SYMP HLA Tg rabbits, exhausted CD8 + T RM cells in HSV-1 latently infected TG occurs in TG of SYMPP patients. Exhaustion of HSV-specific CD8 + T cells may occur during multiple productive replication attempts and/or repetitive virus reactivations, events that might be underestimated with actual molecular detections 73 . Moreover, we found that LAG-3 and PD-1 immune checkpoints blockade in HSV-1 latently infected HLA Tg rabbits restored T cell functions associated with less recurrent ocular herpes infection and disease, confirming our previous studies 18,28,31,35 . The findings in this report prompt to question whether HSV-1 infections can be categorized as a latent or chronic infection.
Our transcriptome profiling data identified the transcription factors ICOS and PRDM1 (BLIMP-1) as cooperative regulators of the co-inhibitory (exhaustion) module identified in antiviral CD8 + T cells associated with asymptomatic herpes infection. The inducible T cell co-stimulator (ICOS) has been reportedly associated with induction of Th1, Th2, Th17 immunity and regulation of effector T cell differentiation and with a role in adaptive immunity 74 . Remarkably, an up-regulation of ICOS has been reported in human and mice CD8 + T RM cells 54,75 . Similarly, Mackay et al. observed that BLIMP-1 expression by memory T cells was highest eight days after herpes infection and later was maintained at a lower level in memory T cells in the skin 53 . They further suggested that BLIMP-1 expression is required to form short-lived effector T cells and BLIMP-1 along with HOBIT are necessary for the development of T RM cells but not of circulating memory T cells, suggesting that BLIMP-1 and HOBIT cooperate to promote T RM cell development 53 . Our results provide insights into the transcriptional regulation that influence memory formation and CD8 + T cell protective immunity that contribute to asymptomatic herpes. Among the limitations are that the select tissues from SYMP and ASYMP herpes patients profiled in this study may not include the full diversity of T cell transcriptional programs throughout the body, and that quantification of various memory T cell subsets may be subject to dissociation biases between the individual tissue compartments.
In summary, our study demonstrates that HSV-1 infected ASYMP humans and "protected" ASYMP HLA Tg rabbit model have unique transcriptomic, phenotypic and functional characteristics in their antiviral CD8 + T Scientific RepoRtS | (2020) 10:13843 | https://doi.org/10.1038/s41598-020-70673-z www.nature.com/scientificreports/ cells. Bulk and scRNASeq profiling of HSV-specific memory CD8 + T cells from blood and TG, the site of latent infection, demonstrate that high-resolution map of antiviral T cells in symptomatic and asymptomatic herpes serve as a new baseline for defining the coding genes and the immunological pathways that are differentially expressed in CD8 + T RM cells during HSV-1 symptomatic vs. asymptomatic infections. Future studies will address the biological mechanisms that regulate "protective" and "pathogenic" pathways in CD8 + T RM cells during HSV-1 symptomatic vs. asymptomatic infections. Importantly, this work strengthens the current strategy of developing T-cell based immunotherapeutic approaches to protect from recurrent herpes.

Methods
Human study population. All clinical investigations in this study were conducted according to the Declaration of Helsinki. All subjects were enrolled at the University of California, Irvine under approved Institutional Review Board protocols (IRB#2003-3111 and IRB#2009-6963). Written informed consent was received from all participants prior to inclusion in this study. During the last 17 years (i.e. January 2003-January 2020), we screened 955 individuals for HSV-1 and HSV-2 seropositivity. Since HSV-1 is the main cause of ocular herpes, individuals who were HSV-1 seropositive were enrolled in this study. The HSV-1-seropositive individuals were divided into two groups: (1) HLA-A*02:01positive ASYMP individuals who, despite being infected, never had any clinically detectable herpes disease; and (2) HLA-A*02:01-positive SYMP individuals with a history of numerous episodes of clinically documented recurrent ocular herpes diseases, such as herpetic lid lesions, herpetic conjunctivitis, dendritic or geographic keratitis, stromal keratitis, and iritis consistent with rHSK, with one or more episodes per year for the past 5 years. Only SYMP patients who were not on Acyclovir or other antiviral or anti-inflammatory drug treatments at the time of blood sample collections were enrolled 16,22,26,76 . SYMP and ASYMP groups were matched for age, gender, serological status, and race. Among the SYMP and ASYMP individuals, 4 HLA-A*02:01-positive patients (ASYMP: n = 2, and SYMP: n = 2) were enrolled in this study for transcriptome profiling through bulk RNA sequencing. More importantly, as present study was aimed to study the differentially expressed genes in HSVspecific CD8 + T cells in ASYMP and SYMP individuals. We have observed none or very few tetramer-positive cells in the seronegative control group, which was not enough to run the RNA sequencing experiment further. Therefore, we did not include seronegative individuals as baseline control in this study.
HLA transgenic rabbits. An HLA transgenic rabbit colony was bred at UC Irvine 19,49 and the New Zealand White (NZW) rabbits, purchased from Western Oregon Rabbit Co., were used for all the experiments. All rabbits were housed and treated in accordance with ARVO (Association for Research in Vision and Ophthalmology), AAALAC (Association for Assessment and Accreditation of Laboratory Animal Care), and NIH (National Institutes of Health) guidelines. A colony of human leukocyte antigens (HLA) transgenic (Tg) rabbits were maintained at UC Irvine. The HLA Tg rabbits retain their endogenous rabbit major histocompatibility complex (MHC) locus and express human HLA-A*02:01 under the control of its normal promoter 44,78 . Prior to this study, the expression of HLA-A*02:01 molecules on the PBMC of each HLA-Tg rabbit was confirmed by fluorescenceactivated cell sorting (FACS) as described previously 10 . Herpes simplex virus production and ocular herpes infection. The HSV-1 (lab strain McKrae) was used in this study. The virus was triple plaque purified and prepared as previously described 49,55 . Two groups (Group-1: n = 10; Group-2: n = 10) of HLA Tg rabbits (8-10 weeks) received an ocular HSV-1 challenge (2 × 10 5 pfu, lab McKrae strain) without scarification. Following ocular infection, rabbits from both groups were monitored for ocular herpes, virus infection, and disease 44,78 . The Group-1 HSV-1 infected rabbits were observed with no recurrent corneal herpes and categorized as asymptomatic (ASYMP). In contrast, the Group-2 HSV-1 infected rabbits were found with severe corneal herpetic disease and classified as symptomatic (SYMP). On the Scientific RepoRtS | (2020) 10:13843 | https://doi.org/10.1038/s41598-020-70673-z www.nature.com/scientificreports/ severity of disease scoring scale ranging between 0 and 4; the ASYMP rabbits scored 0-1, whereas the SYMP rabbits scored 2-4. immune checkpoints blockade. Cross-reactive anti-PD-1 mAb (RMPI-14) and anti-LAG-3 mAb (C9B7W) were purchased from BIOXCELL (West Lebanon, NH). NZW and HLA Tg rabbits were ocularly infected with 2 × 10 5 pfu of lab strain McKrae and received an intravenous injection of 200 μg of anti-PD-1 mAb and/or anti-LAG-3 mAb on scheduled days, as illustrated in Fig. 2a.
Detection of rabbit ocular infectious virus. Tears were collected from both eyes using a Dacron swab (type 1) (SPECTRUM LABORATORIES, CA, USA) following the commencement of blockade post-challenge. Individual swabs were transferred to a 2 mL sterile cryogenic vial containing 1 mL culture medium and stored at − 80 °C until use. The HSV-1 titers in tear samples were determined by standard real-time PCR based on previously described reaction conditions 79 . Bulk RnA-sequencing. HSV-specific CD8 + T cells were FACS sorted from PBMC of ASYMP and SYMP human subjects using tetramers specific to immunodominant epitopes selected from both glycoprotein and tegument proteins: (1) glycoprotein gB (gB 561-569 ); and (2) tegument protein VP11/12 (VP11/12 702-710 ). Post-sorting acquisition was performed for the purpose of post-sort purity check ( Supplementary Fig. S2). Similarly, TG of blockade-treated and mock-treated HLA Tg rabbits were treated using tetramers specific to immunodominant epitopes selected from both glycoproteins and tegument proteins: (1)  Droplet-based scRnASeq. For droplet-based scRNASeq, libraries were prepared from CD45 + T cells sorted from the TG of ASYMP (HLA Tg rabbits treated with α-PD-1 and α-LAG-3 mAbs) and SYMP (HLA Tg rabbits not subjected to PD-1 and LAG-3 blockade treatment) HLA Tg rabbits using the Chromium Single Cell 3′ Reagent Kits v.3 according to the manufacturer's protocol (10 × GENOMICS). The generated scRNASeq libraries were sequenced using Hiseq 4000 (150 cycles).

Ex-vivo culture to detect virus reactivation (HSV-1 DNA copies) in Trigeminal Ganglia.
Analysis of droplet-based scRnASeq. Gene counts were obtained by aligning reads to the rabbit genome assembly OryCun2.0 (GCA_000003625.1) using Cell Ranger software (v.3.0.2) (10 × Genomics). A gene-count matrix was generated after quantification of mRNA using the Cell Ranger count pipeline. The filtered gene count matrix files were used for downstream analysis using Seurat 3.0. Then, a cut-off value of 200 unique molecular identifiers (UMIs) was used to select single cells for further analysis. This resulted in an initial dataset that was subsequently examined to exclude low-quality libraries 80 .
Quality control for cell inclusion. The initial dataset contained 906 cells for ASYMP group and 886 cells for SYMP group of HLA Tg rabbits. t-distributed stochastic neighbor embedding (t-SNE) was used to project the entire dataset onto the two-dimensional space based on the top 10 principal components. For each cell, we evaluated; (1) the number of genes for which at least one read was mapped (which is indicative of library complexity), (2) the total number of counts, (3) the percentage of counts mapping to the top 50 genes, and (4) the percentage of reads mapped to mitochondrial genes. Cells showing a relatively lower proportion of endogenous RNAs might suggested that either the cells were stressed or dead and resulted in RNA degradation. Outlier cells in these quality metrics were found to cluster together in the t-SNE two-dimensional space. Further, k-means clustering algorithm was applied and cells with an abnormally high ratio of counts mapped to mitochondrial genes were removed and gene counts associated with protein-coding genes were included for downstream analysis. At the end of quality control, 902 cells among ASYMP and 745 cells among SYMP groups remained.
Scientific RepoRtS | (2020) 10:13843 | https://doi.org/10.1038/s41598-020-70673-z www.nature.com/scientificreports/ Cell clustering and marker identification. Data normalization and clustering was performed on quality control passed ASYMP (n = 902) and SYMP (n = 745) cells with the Seurat 3.0. Counts for all were scaled by the total library size multiplied by 10,000 and transformed to log space. A total of 3,078 and 2,964 highly variable genes for ASYMP and SYMP groups were identified based on mean and dispersion, and the values were rescaled. Principal component analysis (PCA) was performed on the variable genes, and t-SNE was run on the top 10 principal components (PCs) and the Louvain graph-clustering algorithm was applied to identify cell clusters.
Differential gene expression analysis. The differentially expressed genes (DEGs) were analyzed by using integrated Differential Expression and Pathway analysis (iDEP) tool 81 . iDEP is a comprehensive analysis toll comprising of 63 R/Bioconductor packages, two web services, and well-annotated pathway databases for multiple animal species including that of humans and rabbits. The gene count data obtained after applying quality control parameters were subject to DEG analysis. The genes were converted to Ensemble gene IDs and filtered based on gene count. Subsequently, the pre-processed data was used for k-means and hierarchical clustering. The pairwise comparison of human and rabbit (ASYMP vs. SYMP) was performed using the DESeq2 package with a threshold of false discovery rate (FDR) < 0.5. and fold-change > 1.5. Although DESeq, SAMseq, EBSeq methods can be used to identify differentially expressed genes (DEG) in RNASeq data, in this study, we preferred DESeq over the other approaches because it allows obtaining better results with small samples (e.g., two samples per condition as used in this study), as reported previously 82,83 . Unlike the DESeq approach used in this study, the results obtained by other approaches, such as SAMseq, can be influenced by the small sample size and may produce many false positives 84 . Similarly, in scRNA Seq, the statistical power comes from the analysis of multiple cells rather than biological samples, which justifies the use of two samples per group, as previously reported 85 . Moreover, a hierarchical clustering tree and network of enriched GO/KEGG terms were constructed to visualize the potential relationship. Gene set enrichment analysis (GSEA) method was performed to investigate the related signal pathways activated among ASYMP and SYMP groups. Parametric gene set enrichment analysis (PSGEA) method was applied based on data curated in Gene Ontology and KEGG. Pathway significance cutoff with a false discovery date (FDR) ≥ 0.2 was applied.
flow cytometry assays. Trigeminal ganglia were analyzed by flow cytometry, as we previously reported 86 .
The following antibodies were used: mouse anti-rabbit CD8 (clone MCA1576F, SEROTEC), mouse anti-human CD103 (clone H4A3) FITC, CD69 (clone H4B4) APC/Cy7 (BIOLEGEND), and rat anti-IFN-γ (clone XMG1.2) (BD Biosciences). For surface staining, mAbs against various cell markers were added to a total of 1 × 10 6 cells in phosphate-buffered saline containing 1% FBS and 0.1% sodium azide (FACS buffer) and left for 45 min at 4 °C. For intracellular staining, mAbs were added to the cells and incubated for 45 min on ice and in the dark. A total of 50,000 events were acquired by LSRII (BECTON DICKINSON, Mountain View, CA, USA), followed by analysis using FlowJo software (TREESTAR, Ashland, OR, USA).
Statistical analyses. Data for each differentially expressed gene among blockade-treated, and mocktreated groups of HLA Tg rabbits were compared by analysis of variance (ANOVA) and Student's t test using GraphPad Prism version 5 (La Jolla, CA, USA). ANOVA and multiple comparison procedures were followed to calculate the statistical differences between the study groups. Data are expressed as the mean ± SD. Results were considered statistically significant at P ≤ 0.05.