HSV-1 is an important human pathogen, but there are no vaccines in active clinical development. In this report, we have shown that the proteins encoded by HSV-1 genes UL39 and UL46 have coordinated CD8 and CD4 immunogenicity in most individuals and are therefore rational vaccine candidates. We have provided an estimate of the complexity of the response, documented clustering of responses based on HLA type, and defined more than 40 novel CD8 T cell epitopes. We have identified a new hierarchy of responses based on HLA locus, with HLA-A more frequently responsible for antigen presentation than HLA-B, and HLA C having a minor contribution. Parallel studies determined that infected humans recognize a mean of 17 and 23 HSV-1 ORFs as CD8 and CD4 antigens, respectively. With the exceptions noted in the Introduction and Results, to our knowledge, the vast majority of the antigenic reactivities and epitopes we have defined are novel. We also showed applicability to another large-genome virus and anticipate that these systems may be useful for many pathogens.
T cell responses to pathogens have been difficult to access for several reasons. Microbe-specific T cells can occur at low abundance in the blood, such that an unbiased pre-enrichment step is helpful for new antigen or epitope discovery. Sylwester et al.’s probe of the response to the CMV proteome using peptides and direct PBMC ICC was enabled by the high overall abundance of T cell responses to CMV (48
). Responses to EBV are also large, such that a direct PBMC approach to CD8 responses using a cloned partial ORFeome has yielded hits (49
). Indeed, the low magnitude of direct PBMC IFN-γ ELISPOT responses to single HSV-1 peptides in the current report contrast with the high-magnitude responses to single epitopes noted in CMV and EBV (48
We have shown that restimulation methods tailored for CD8+ or CD4+ T cells, based on the biology of antigen presentation, can be used without change for two distinct viruses. Once responder cells are enriched and expanded, the large microbial genomes still harbor myriad potential epitopes that are challenging to decode. In this report, we first assigned CD8 reactivity at the level of HSV-1 ORFs, using efficient plasmid sets. Limited use of peptide binding algorithms and peptide synthesis efficiently confirmed antigenicity to the epitope level. The new, validated peptide reagents were in turn positive in a proportion of additional HLA-appropriate subjects in direct ex vivo assays. Even in our limited study population, CD8+ T cells from different individuals frequently had reactivity, in our primary co-transfection screens, with the same HSV-1 ORFs when studied using shared HLA cDNA molecules. For example, each HLA-A*0101–bearing donor had CD8+ T cells recognizing the HSV-1 ORFs UL1 and UL48 (Figure ). Clustering studies to measure this tendency will be the subject of future analyses.
In addition to identifying UL39
as rational candidate subunit vaccines, our data have implications for the design of whole virus formats. In fact, the diversity of the CD8 response, with an average of 17 antigens per person, implies that a whole virus rather than a subunit approach to vaccination for HSV-1 is most likely to mimic the immune response to natural infection. Globally, we found that broad kinetic and structural and functional spectra of HSV-1 proteins were recognized by the human CD8 response. Therefore, either replication-incompetent or attenuated replication-competent vaccine formats have the potential to stimulate broad CD8 responses in most individuals. Among whole HSV vaccine candidates that replicate discontinuously in normal cells, those that allow expression to proceed relatively completely may be quite rational (25
). To explore the virological rules of HSV-1 T cell antigenicity, we scored each protein dichotomously for CD8 responses in the study population, and then categorized each protein studied as immediate early, early, late but not otherwise characterized, early late with synthesis prior to DNA replication, or true late with expression only after DNA replication (1
). We also recorded, for each HSV-1 protein, its absence, detection at less than 1%, or detection at greater than 1% of adjusted virion mass using mass spectroscopy data of purified virions (51
). These analyses disclosed a weighting of CD8 responses toward HSV-1 proteins expressed prior to HSV-1 DNA replication, and toward abundant virion polypeptides. Specifically, only 3 of 17 true late proteins (18%) were recognized by CD8+
cells. In contrast, 4 of 5 immediate early proteins (80%), 9 of 12 early proteins (75%), 12 of 19 early late proteins (63%), 12 of 20 late proteins not specified as early or late (60%), and 0 of 1 proteins with no specified expression kinetics (0%) were CD8 antigens (highlighted in Supplemental Figure 5 using Figure data and expression classifications from a standard text; ref. 1
). Among the more abundant virion proteins, 17 of 23 (74%) were positive for CD8 responses. In contrast, only 23 of 51 (45%) proteins either absent from virions or present at levels less than 1% were CD8 antigens. The population prevalence for CD4 responses did not segregate by HSV-1 kinetics or structural class. Our data suggest that whole-virus HSV-1 format vaccines that express most proteins normally made prior to DNA synthesis, or that can be dosed to provide a large mass of virion input proteins, should retain the potential to stimulate broad CD8 responses.
A vaccine covering HSV-1 and HSV-2 would be desirable. Half of the minimal HSV-1 CD8 epitopes defined in this report are sequence-identical in HSV-1 and HSV-2 and appropriate for candidate type–common vaccines. Indeed, the HSV-2 homologs of three epitopes — HLA-A*0101/HSV-1 UL39
512–520, HLA-A*0201/HSV-1 UL25
367–375, and HLA-A*0201/HSV-1 UL27
448–456 — were found in our prior HSV-2 work (52
). HSV CD8 epitopes can also tolerate amino acid substitutions, as exemplified by UL46
354–362 of HSV-1 and HSV-2, differing at amino acid 2, and by ICP0 HSV-1 698–706 and its homolog HSV-2 ICP0 742–750, differing at amino acids 1 and 3. It is certainly possible that cross-reactive T cells could be involved in cross-protection against some aspects of HSV-2 infection or severity observed in HSV-1–infected individuals (55
). Most of our subjects were dually infected with both HSV types. Future cross-sectional studies comparing immune responses to HSV-1 in the presence or absence of HSV-2 coinfection could clarify the extent to which each infection contributes to the cross-reactive repertoire.
With regard to effector function, we showed that bulk HSV-specific CD8+
T cells enriched using cross-presentation and CD137 have brisk virus-specific cytotoxicity. We plan to enrich peptide-specific CD8+
cells with peptides (56
) or tetramers (57
) and study recognition of HSV-1–infected skin cells to more closely mimic physiologic target cells. The Cos-7 system has not been characterized for use in CTL assays, so we will move to a viral infection system. In HSV-2 studies, T cells recognizing diverse antigens were able to lyse HSV-infected skin cells, but the specific conditions, such as the dose and time of infection and the requirement for de novo viral protein synthesis or for IFN-γ pretreatment (56
), differed between epitopes. With the larger panel of HSV-1 epitopes, we hope to establish general rules for CD8 recognition of physiologically relevant cells that could inform vaccinology. Future cross-sectional study of populations with defined levels of HSV-1 severity and longitudinal research during the ontogeny of primary immune responses or during reactivations in the chronic phase may also contribute correlates of severity and reactivation that could further influence vaccine design.
was a strong HSV-1 CD8 antigen in both humans and the one mouse MHC haplotype studied, H-2b
is a virulence factor involved in evading innate immunity and apoptosis (59
), such that immune targeting of UL39
may be advantageous to the host. The CD8 repertoire in infected C57BL/6 mice had a breadth of 19 HSV-1 epitopes. These were concentrated in only three ORFs, gB1 (gene UL27
, and ICP8 (gene UL29
). The mice did not recognize immediate early HSV-1 polypeptides, while responses to ICP0, ICP4, ICP22, and ICP47 were detected in humans. In addition to MHC class I peptide binding preferences, these differences may reflect species-specific effects of HSV-1 HLA class I immune evasion genes (61
) and the fact that the human exposure to HSV-1 antigen is chronic and intermittent, while HSV-1 typically does not recur in mice. We conclude that study of adaptive immunity in the natural host is a necessary counterpoint to the powerful manipulative experiments that are possible in experiment models of HSV-1 infection.
Several factors may have influenced our results. Our CD8 workflow used cross-presentation at the re-stimulation step, but then switched to direct present by aAPCs or peptide-loaded cells at the readout step. We observed that cross-presentation is efficient in presenting diverse HSV-1 proteins to CD8+
T cells. The HSV proteins ICP47 (gene US12
) and vhs (gene UL41
) inhibit direct presentation (30
). We have previously re-stimulated memory CD8+
cells using direct presentation by HSV-infected B lymphoblastoid cell lines (B-LCLs), but this yielded a paucity of HSV-reactive CD8 clones reactive only with membrane glycoproteins (62
). HSV-infected fibroblasts failed in this endeavor. Direct presentation by infected DCs might uncover epitopes specific for this pathway, although HSV infection harms various DCs and renders them defective for antigen presentation (63
). Future studies will compare direct and cross-presentation at the re-stimulation stage.
Expression of the HSV-1 proteome was not totally complete. Genes UL15.5
, and UL43.5
are under development, as is the C-terminal approximately 500 amino acids of the UL36
protein. Genes predicted to be in-frame subsets of longer polypeptides were not included, but this will not lead to loss of potential epitopes. A poorly studied variable is allelic heterogeneity in HLA class I assembly with Chlorocebus sp.
β2m in Cos-7 cells. We overexpressed HSV-1 ORFs in isolation in aAPCs, where intracellular trafficking and class I presentation could differ from the viral context. There were subtle differences in some of our HLA-C constructs from the HLA-A and -B vectors, but our method has achieved excellent HLA-C expression (66
). There are interactions between HSV-1 proteins such as proteolysis and phosphorylation (1
), and possibly species-specific host protein–HSV-1 protein interactions, that would differ between infected and transfected cells. In HSV-2 work using a genomic DNA library and Cos-7 transfection, we decoded the fine specificity of each CD8 clone studied (52
) and therefore believe such situations are rare for HSV. We focused on IFN-γ readouts of CD8+
T cell activation and proliferation to detect CD4+
T cell responses. With regard to effector cytokines, rare HSV-reactive T cells in PBMCs making TNF-α or IL-2 but not IFN-γ have been described (54
). In preliminary studies, substitution of TNF-α for IFN-γ ELISA did not uncover additional specificities (data not shown). We noted one HSV-seronegative individual with CD8 responses to HSV-1 (participant 13 in Table ). Further work will be required to determine whether reactivity can be confirmed at the epitope level, as has been done for HSV-2 (68
We have extended the use of CD137 as an activation marker to two complex microbes for both CD4+
T cells. CD137 mediates a strong co-stimulatory signal to T cells. Thus, use of anti-CD137 to detect and purify antigen-reactive cells may assist their downstream expansion. Our data are consistent with some level of bystander CD137 expression, as the level of reactivity with whole HSV-1 among expanded CD137hi
cells varied between 4% and 45%. Enrichment was better for CD4 cells. We cannot be sure that all memory HSV-reactive cells upregulated CD137. CD137 is similar in this regard to other molecules used for enrichment, such as CD134, CD154, and captured IFN-γ (69
In summary, the T cell response to a complex and serious pathogen with a large genome, HSV-1, has been decoded with a linked set of cellular and molecular tools to reveal novel candidate vaccine antigens. We have preliminarily identified the proteins encoded by genes UL39 and UL46 as having high population prevalence of coordinated CD8 and CD4 responses and plan to study larger, defined populations to refine these conclusions. Cross-presentation followed by CD137-based selection also effectively enriches rare CD8+ cells specific for vaccinia virus, an effective live virus vaccine. CD137 is also suitable for enrichment of CD4+ T cells reactive with whole microbe preparations, as demonstrated for both HSV-1 and vaccinia virus. Our flexible gene cloning format allows integrated, efficient study of both CD8 and CD4 responses after one round of PCR-based cloning of microbial ORFs. Use of appropriate initial re-stimulation conditions, CD137 as a flexible selection marker, and the genomes and complete genome-covering ORF sets now available for Mycobacterium tuberculosis, Plasmodium falciparum, and other agents should speed comprehensive definition of T cell responses and vaccine design.