Poor Antigen Processing of Poxvirus Particles Limits CD4⁺ T Cell Recognition and Impacts Immunogenicity of the Inactivated Vaccine

CD4⁺ T cells have a diverse set of functions, which make them a crucial immune cell type for protection against a diverse array of infectious diseases. Key roles include coordination of B cell and CD8⁺ T cell responses, production of inflammatory cytokines, and, in some cases, direct killing of infected cells. For these reasons CD4⁺ T cell engagement is a critical consideration in rational vaccine design.

CD4⁺ T cell activation is initiated by interactions at the cell surface with MHC class II (MHC-II) molecules in complex with antigenic peptides (epitopes). A large body of work with stable globular proteins has suggested that the majority of MHC-II–restricted epitopes are derived from exogenous Ags digested in the endosomal compartment. However, these model Ags do not predict the processing of more complex structures, such as viral particles. Our work with influenza has shown that a more complicated network of MHC-II Ag processing and presentation is at play, in many cases using as processing substrates proteins synthesized within the APC (1, 2). This requirement derives both from the ability of nascent viral proteins to be engaged by a diverse network of cellular components capable of producing a wide array of peptides and from the poor processability of whole virions (1). However, it is unclear whether these observations reflect general principles in viral immunity or are specific to influenza.

Like influenza, orthopoxviruses continue to be a public health concern, and the current vaccination strategy, although effective, can cause severe adverse reactions. Greater mechanistic understanding of how poxviruses engage the adaptive immune system could be of considerable benefit. Various orthopoxviruses can cause severe disease in people, most notably smallpox, one of the most lethal diseases in human history (3) and an ongoing threat as a bioterrorism agent. More recently, zoonotic poxviruses such as monkeypox have emerged as pathogenic in humans, and evidence with monkeypox suggests that the virus is rapidly evolving to more efficiently counteract the human immune system, resulting in more severe disease (4). Despite promising results with subunit vaccines (5–7), the gold standard of immunization against poxviruses remains replication-competent vaccinia (VACV), which has been used for centuries as a live vaccine against smallpox (8). Although VACV immunization is efficacious, cases of severe pathogenesis, secondary infection, and even death have been reported, with symptoms ranging from mild, such as fatigue and headaches, to severe, such as progressive VACV, eczema vaccinatum, and, rarely, fatal encephalitis (9–16). During the era of smallpox eradication there were attempts at an inactivated whole-virus vaccine, and although there was limited success under boosting conditions, no efficacy in producing neutralizing Ab titers in previously naive individuals was observed (17–19). Thus, efforts in this regard have largely ceased.

The production of neutralizing Abs is considered a requirement for protection against secondary poxvirus infections, and many of the vaccination studies have focused primarily on the Ab response. Thus far, the CD4⁺ T cell response generated in response to inactivated versus live poxvirus agents has not been characterized despite CD4⁺ T cell help being crucial for a strong Ab response as well as direct CD4⁺ T cell–mediated killing of pox-infected cells (20, 21). Although we have previously characterized the poor Ag processing of whole virions as a key determinant in the poor CD4⁺ T cell response to inactivated influenza, the influenza virion is relatively fragile, surviving only minutes in solution (22). In contrast, orthopoxvirus virions are remarkably durable, owing to large size with an extremely dense protein content (23, 24). Remarkably, these resilient particles can be freeze-dried and subsequently reconstituted with complete viability (25). Because of these properties, we speculated that orthopoxvirus-specific CD4⁺ T cell responses are dependent upon infectivity and the production of more processable forms of Ag, even more so than what we observed for influenza-specific responses.

To characterize the CD4⁺ T cell response to inactivated poxvirions, we turned to a murine model of smallpox, ectromelia (ECTV), colloquially known as mousepox and a close relative to VACV as well as smallpox (26). ECTV naturally infects mice, displays a restricted host range to that species, and closely mimics the disease progression of smallpox and monkeypox in humans (27, 28). Compared with VACV, ECTV is better able to infect immune cells, produce a systemic infection, and cause fatal disease at low infectious doses in certain inbred strains (29, 30). Furthermore, immunization of mice with live VACV can protect even susceptible mouse strains from a secondary infection with ECTV (31), thereby mimicking the current human immunization protocols.

Studies focusing on CD4⁺ T cell immunogenicity have been limited because of the difficulties in predicting MHC-II–binding epitopes as well as the focus on B and CD8⁺ T cell responses. Thus, by using a panel of algorithm-selected synthetic peptides and CD4⁺ from ECTV-primed C57BL/6 mice, we first identified a large number of novel ECTV-derived I-A^b–restricted epitopes. Assaying for responses to these epitopes under various conditions, we observed that poxvirion particles are exceptionally poor processing substrates for the MHC-II Ag-processing machinery even when an adjuvant was included, providing an important guiding principle in rational poxvirus vaccine design.

Bone marrow dendritic cells (BMDCs) were derived from the bone marrow of pooled female C57BL/6 mice (The Jackson Laboratory, 00064) and cultured for 7 d in RPMI 1640 (10% FBS, antibiotics, l-glutamine, and 2-ME) with GM-CSF (Gemini Bio-Products).

Female C57BL/6 mice 6–8 wk of age were inoculated with appropriate virus (three mice per group). ELISpot plates were coated with anti–IFN-γ at 1:200 in PBS and incubated overnight at 4°C. Ten days postinfection, spleens from each group were harvested. CD4⁺ T cells were purified from bulk splenocytes using negative bead isolation (Invitrogen Dynabeads Untouched Mouse CD4⁺ Isolation Kit) and incubated with BMDCs and 15-mer peptides, virions, or subviral material. After an overnight incubation, the plates were developed (BD Biosciences ELISpot IFN-γ Ab pair and AEC Substrate Kit), and IFN-γ spots produced by activated T cells were counted (ImmunoSpot reader).

All protein sequences for the ECTV genome (NC_004105.1) were collected from GenBank (173 open reading frames [ORFs]) and broken down in silico into 15-mer peptides, overlapping by 10 residues, starting at position 1, and including a peptide covering the C terminus. For example, for a 27-residue protein, peptides 1–15, 6–20, 11–25, and 13–27 would be selected (total = 10,984 15-mers). After removing duplicate peptides (because of sequence homology between the inverted terminal repeat regions), binding predictions for the remaining 10,721 15-mers were performed for the MHC allele I-A^b using the consensus method available at the Immune Epitope Database and Analysis Resource (IEDB) MHC-II binding prediction tool (32). The 1000 peptides predicted for highest binding to I-A^b were synthesized (Pepscan Systems). Female C57BL/6 mice were immunized with 3 × 10³ PFU ECTV (Moscow strain) in 30 μl of saline via the footpad route. After 10 d, CD4⁺ T cells were purified from pooled spleens, and peptides were assessed in an ELISpot assay for CD4⁺ T cell recognition. The 1000 peptides were screened individually via a matrix approach, and then the top responders were assayed in triplicate. Spots represent IFN-γ–producing cells per 100,000 purified CD4 cells. Four independent experiments were performed, and for each experiment the percent of the total response was calculated for all individual peptides. The mean and SEM of the percent of the responses across all four independent experiments was calculated for each peptide that appeared in three or more independent experiments and is shown in Fig. 1.

TK⁻ cells infected with ECTV (Moscow) and VACV (Western Reserve) for 72 or 48 h, respectively, were lysed by repeated freeze, thaw, sonication cycles. The cell lysate was then purified through a 36% sucrose cushion and resuspended in 1 mM Tris-HCL for use in animal and in vitro experiments.

Female C57BL/6 mice were immunized with either live or inactivated ECTV. Live ECTV was inoculated in the footpad with 3 × 10³ PFU of virus in 30 μl of saline, a commonly used infection route and dose for this virus; alternately, 3 × 10⁶ PFU of virus was inactivated with UV radiation/psoralen and injected i.p. Ten days postinjection, CD4⁺ T cells from pooled spleens were tested in an ELISpot assay for their ability to recognize the 42 consistent peptide hits. Spots represent IFN-γ–producing cells per 100,000 purified CD4 cells.

Female C57BL/6 mice were inoculated with ECTV in the footpad with 3 × 10³ PFU of virus in 30 μl of saline; alternatively, mice were inoculated i.p. with 3 × 10⁵ PFU of VACV (WR strain). Ten days postinjection, CD4⁺ T cells from the pooled spleens from each set of mice were tested in an ELISpot assay for their ability to recognize the 42 consistent peptide hits. Spots represent IFN-γ–producing cells per 100,000 purified CD4 cells. To normalize the two conditions, each peptide is shown as the percentage of the total response (total spot count).

Female C57BL/6 mice were immunized with live ECTV, inactivated ECTV, or inactivated ECTV plus adjuvant. Adjuvant consisted of 10% aluminum hydroxide gel (Rehydragel LV, Chemtrade Chemicals), 1 mg/ml saponin (Sigma-Aldrich), as previously published by (33). Live ECTV was inoculated in the footpad with 3 × 10³ PFU of virus in 30 μl of saline; alternately, 3 × 10⁶ PFU of virus was inactivated with UV radiation/psoralen and injected s.c. in 500 μl of either saline or adjuvant. Ten days postinjection, CD4⁺ T cells from individual mice were tested in an ELISpot assay for their ability to recognize five consistently strong peptide hits. Spots represent IFN-γ–producing cells per 100,000 purified CD4 cells.

Female C57BL/6 mice were infected with either 3 × 10³ PFU live VACV, 3 × 10⁶ PFU VACV inactivated with UV radiation/psoralen, or a saline control injected i.p. Four weeks later, serum was collected via cheek bleed prior to inoculation with 3 × 10³ PFU live ECTV via footpad scarification. Five days postchallenge, mice were sacrificed, the dorso-plantar thickness of the infected foot was measured for swelling by digital calipers, and spleens and livers were collected for organ titering. Harvested organs were processed using a gentleMACS dissociator, and live virus was titrated by plaque assay on BSC-1 cells from homogenates.

Serum was collected from the heart of primed mice immediately following death and analyzed for IgG titer by ELISA. Briefly, serum was serially diluted in PBS supplemented with 1% low-IgG BSA (Gemini Bio-Products) ranging from 1:500 to 1:16,000 and incubated in high-binding EIA/RIA plates (Corning) precoated with 6.25 × 10⁴ PFU purified ECTV. Plates were then washed with PBS + 0.01% Tween and incubated with peroxidase-labeled anti-mouse IgG (H+L) (catalog no. PI-2000; Vector Laboratories) at 1:1500 dilution in PBS/BSA (1%). Plates were developed using ABTS Peroxidase Substrate (KPL) and read at detection wavelength of 405 nm.

Serum from infected mice was subjected to a previously published protocol for assessing neutralizing Abs to poxviruses (34). Briefly, serum was incubated for 1 h with a β-galactosidase expressing VACV prior to dispensing on a monolayer of HeLa cells for overnight infection (18–20 h). β-galactosidase activity was measured using a plate reader at 405 nm and compared with a condition prepared without serum (100% infectivity).

TK⁻ cells were infected with ECTV for 3 d until significant cytopathic effect was observed. Cells were harvested and subjected to three consecutive cycles of freezing, thawing, and vortexing. Whole virions were separated from free proteins via a previously published procedure (35), namely ultracentrifugation, at 36,000 × g for 30 min at 4°C. The pellet (virions) was resuspended in PBS, and both pellet and supernatant were assayed for live virus by plaque assay on BSC-1 cells.

Samples were boiled in nonreducing conditions and loaded onto a precast NuPage 4–12% Bis-Tris gel (Thermo Fisher Scientific). Following semidry transfer onto a nitrocellulose membrane and subsequent blocking (Licor blocking solution), presence of A27L (Santa Cruz Biotechnology sc-69950) and A33R (BEI Resources NR-49231) structural proteins were probed and detected using Licor suitable secondary reagents.

Our previous work has illustrated on numerous occasions the heterogeneous MHC-II–processing properties of individual epitopes within a complex pathogen (1, 36–38) and the inaccurate generalizations that can result from studying a limited number of epitopes or bulk responses. We therefore set out to obtain a large panel of individual epitopes that we could use to probe CD4⁺ T cell responses to ECTV. This was accomplished by measuring reactivity of CD4⁺ T cells from ECTV-infected mice to a panel of synthetic peptides derived from the ECTV proteome. Orthopoxviruses have extremely large genomes, encoding on the order of 200 distinct proteins. Rather than creating a comprehensive overlapping peptide screen covering the entire proteome, we used in silico epitope prediction software to focus our screening efforts. All protein sequences for the ECTV genome (NC_004105.1) were collected from GenBank (173 ORFs) and broken down in silico into 10,984 15-mer peptides overlapping by 10 residues, starting at position 1, and including a peptide covering the C terminus. For example, for a 27-residue protein, peptides 1–15, 6–20, 11–25 and 13–27 would be selected. After removing duplicate peptides (because of sequence homology between the inverted terminal repeat regions), binding predictions for the remaining 10,721 15-mers were performed for the MHC allele I-A^b using the consensus method available at IEDB MHC-II binding prediction tool (39). Based on the IEDB consensus percentile score (peptides with lower percentile score being better binders), the 1000 15-mer peptides with strongest predicted binding affinity to the I-A^b MHC-II molecule expressed by C57BL/6 mice, excluding those already identified in previous screens (40, 41), were synthesized. An additional stipulation was that each of the 173 ECTV ORFs be represented by at least two peptides, resulting in a range of 2–36 peptides per ORF.

The peptides were tested individually in ELISpot assays using purified CD4⁺ T cells from ECTV-infected mice, and the numbers of IFN-γ–producing cells were recorded. Over four independent experiments, we identified a group of 42 novel peptides that reproducibly induced responses above background (present in three or more experiments) and 17 novel peptides that appeared sporadically (Supplemental Tables I, II). The set displays a reproducible hierarchy of activation, which is depicted in Fig. 1A. Relevant to the heterologous prime/challenge experiments detailed below, immunization with VACV displayed a broadly similar repertoire of epitopes, as expected from the high degree of homology between ECTV and VACV, with only two peptides not showing responses against VACV (both of which had amino acid changes between ECTV and VACV) (Supplemental Fig. 1).

MHC-II binding predictions have historically been more challenging than for MHC-I because MHC-II molecules have less stringent and predictable binding requirements and can therefore accept a greater range of potential peptides (39). In general, it has been found that binding affinity is a necessary but not sufficient condition for T cell immunogenicity (42, 43). In this light, we assessed the relationship of predicted binding strength to observed CD4⁺ T cell stimulation. Although we saw only a modest correlation between predicted rank and actual activation of T cells (Fig. 1B), seven out of the top 10 peptides were predicted to bind I-A^b quite strongly, with predicted binding strengths of >0.2. This suggests effective discrimination in the predictive algorithm’s performance and confirms that prediction of MHC binding can be used to identify candidates for immunogenicity screen, but MHC binding alone is an incomplete predictor of T cell immunogenicity, presumably reflective of other factors, such as the available T cell repertoire and the varying efficiency of the Ag-processing machinery for the production of individual epitopes (43, 44).

Because our screen was designed for broad representation of the ECTV genome, we were able to determine whether there are trends among the parent proteins of the novel epitopes that were identified. Poxviruses have well-defined promoter sequences that segregate protein expression into early, intermediate, and late phases during an infectious cycle; a minority of promoters allow for expression at both early and late timepoints, and some genes appear to have immediate-early kinetics (44, 45). Work with a previous set of CD4 T cell–activating VACV-derived peptides revealed a bias toward late-stage genes (40), which was attributed to the greater expression levels of the proteins in this category. The data from our larger screen suggest that proteins from all phases of infection can be presented by MHC-II in response to both ECTV and VACV (Fig. 1C, Supplemental Fig. 1, Supplemental Table I). In this study, the expression profile of our group of activating peptides more closely mirrors the overall distribution of poxvirus genes (40). Importantly, our data set also show that proteins with structural, regulatory, and virulence roles can also be presented by MHC-II (Supplemental Table I).

Having identified a large panel of MHC-II–restricted ECTV-derived epitopes, we were positioned to analyze the CD4⁺ T cell response to immunization with inactivated virus. We infected C57BL/6 mice with live ECTV via footpad injection, the standard infection route for this virus, or immunized with UV-inactivated ECTV at 100× the input dose to account for the inability of this virus to proliferate. In addition, for inactivated virus, we immunized i.p. rather than via footpad to allow immune cells more direct exposure to the inactivated virus particles. We then used our large panel of epitopes to probe the CD4⁺ T cells resulting from these challenges using IFN-γ ELISpots. In line with our prediction that whole virions would be poor processing substrates for MHC-II, the inactivated virus induced exceedingly weak CD4⁺ T cell responses to a very small number of peptides despite eliciting a detectable Ab response (Fig. 2A, 2B). Indeed, only a single ECTV peptide (741) demonstrated a consistent response above background, and the magnitude of the response to this epitope was substantially lower than that elicited by live virus (Fig. 2A). Three other peptides were sporadically detectable (one out of three independent experiments), with all of these decidedly low in magnitude when at all detectable. As observed with influenza (1), screens of the entire 1000-peptide library did not uncover any novel hits with inactivated virus (data not shown), arguing against immunodominance effects. Notably, there was a low level of virus-specific IgG Abs in the serum of mice infected with UV-inactivated ECTV, suggesting some class switching in the absence of detectable CD4-mediated help. In comparison with our previous work with influenza, where several epitopes of a much smaller peptide panel displayed relatively robust responses against inactivated virus (1), CD4⁺ T cell responses against inactivated ECTV are markedly worse. What is more, whereas boosting with inactivated influenza produced an appreciable and expanded secondary response (1), boosting with UV-inactivated ECTV did not detectably enhance the response (Fig. 2C).

To determine whether an inflammatory milieu could help boost CD4⁺ T cell responses to inactivated poxvirus, we employed a well-established adjuvant, namely aluminum hydroxide + saponin (33). Importantly, this adjuvant mixture has recently been shown to increase Ab responses to inactivated ECTV in a prime/boost vaccination setting via s.c. immunization (33). After confirming that immunizing with inactivated virus via i.p. versus s.c. routes did not alter the CD4⁺ T cell response (data not shown), we infected C57BL/6 mice with live ECTV via footpad or inactivated ECTV s.c. with or without adjuvant, using the same dosages as previously discussed. When the serum IgG titers were analyzed, in contrast to the i.p. route, we did not observe any titers in mice immunized with inactivated virus without adjuvant (Fig. 3B). However, we did observe a low level of serum IgG titers following immunization with inactivated virus in the presence of adjuvant, although this was several orders of magnitude lower than when mice were infected with live virus (Fig. 3B). Irrespective of any boost to Ab responses when adjuvant was included, when we assessed CD4⁺ T cell responses in individual mice for a representative subset of our 42 peptides (Fig. 3A), we did not observe any enhancement of CD4⁺ T cell responses to inactivated virus. These data suggest that the lack of inflammatory environment was not a causative explanation for the lack of CD4⁺ T cell responses to inactivated poxvirus.

We predicted that the low level of CD4⁺ T cell activation we observed following inactivated ECTV immunization, and the resultant poor Ab response, would not be protective against a secondary challenge, correlating with the empirical observations in vaccine trials with inactivated smallpox (17–19). We used VACV as the immunization agent, as it is the smallpox vaccine virus and displays limited pathology in mice while still generating an immune response. Importantly, UV-inactivated VACV showed a similar dearth of epitope-specific CD4⁺ T cell activation as inactivated ECTV (data not shown). We immunized C57BL/6 mice via the i.p. route with live VACV, inactivated VACV, or a saline control and 4 wk later assessed protection by challenging with ECTV via footpad scarification. As C57BL/6 mice will invariably survive this challenge (31), we used footpad inflammation as well as viral titers in the spleen and liver as correlates of protection. As expected, the mice immunized with live VACV were completely protected from secondary challenge, whereas the mice immunized with the saline control displayed significant footpad inflammation as well as ECTV titers in both the spleen and the liver (Fig. 4A–C). In line with our hypothesis, we observed that the cohort of mice immunized with inactivated VACV displayed footpad inflammation as well as organ titers similar to the saline control mice. In addition, we observed low levels of serum IgG in response to the inactivated virus condition at four wk postvaccination, consistent with our previous observations at 10 d (Fig. 4D). Using a well-established viral neutralization assay, it was apparent that this low level of virus-specific Abs did not display neutralizing capabilities (Fig. 4E).

Thus far, using a large panel of epitopes, we observed that CD4⁺ T cells are poorly primed in response to immunization with inactivated poxvirions and that this low level of CD4⁺ T cell activation is not protective. We hypothesized that poor processing of whole virions by the cellular machinery provides an explanation for these results. In contrast, live virus infection creates a pool of viral proteins that are not incorporated into virions and could be more efficiently converted to epitopes by the MHC-II–processing machinery. To test this notion, we infected mouse fibroblasts with ECTV and separated the lysate into whole virions and free viral proteins using a previously published centrifugation technique (35), validating the procedure by observing the presence of infectious virus in the virus but not protein fraction (data not shown). We then incubated BMDCs with dilutions of the virus fraction or the equivalent volumes of the protein fraction and probed the ability of these preparations to reactivate ECTV-experienced CD4⁺ T cells via ELISpot (Fig. 5A). As we had incubated BMDCs with the same volumes of the protein and virus fractions, we were able to normalize the number of reactivated CD4⁺ T cells based on the total protein in each fraction as assessed by a standard protein quantification assay. As shown in Fig. 5A, the free protein fraction stimulated a substantially more robust CD4⁺ T cell response than the whole-virus fraction. Furthermore, we confirmed via Western blot the greater individual protein content in the whole-virus fraction, reflecting the substantially higher stimulatory capacity of poxvirus proteins unencumbered by the dense structure (Fig. 5B).

CD4⁺ T cells are central to coordinating many aspects of the adaptive immune response, including the production of protective neutralizing Ab responses, the elaboration of proinflammatory cytokines, and, in some cases, direct cytolytic killing of infected cells. For all these reasons, the optimization of CD4⁺ T cell function and activation is a key consideration in rational vaccine design. However, the factors necessary for the development of effective CD4⁺ T cell responses remain poorly understood. Although epitope–MHC-II complex recognition by a cognate TCR is established as the critical initiating step of CD4⁺ T cell activation, the steps preceding complex formation, processing of viral proteins into MHC-II–binding epitopes, is increasingly understood to entail a complex network of host proteins functioning in many cellular compartments. In addition, specific aspects of the viral replication cycle are expected to play a large role in which cellular factors are required for processing and presenting antigenic material. Therefore, a careful analysis of epitope-specific CD4⁺ T cell responses against a given virus could greatly inform how vaccination strategies could be modified to enhance protection.

Vaccination has provided a particularly effective means of protecting against poxvirus infection, as neutralizing Abs can cross-react with many pathogenic poxviruses, including both smallpox and emerging zoonotic poxviruses such as monkeypox (46). Rational vaccine design against poxviruses, both smallpox and emerging zoonotic poxviruses, remains a high priority, as the only licensed vaccine is live VACV, which can cause severe pathogenesis in vaccinees, even death in some cases (9–16). However, the basis for the superior immunogenicity of live virus, motivating continued use of a vaccine that has significant risks, is poorly understood. In this study, we carried out an extensive analysis of epitope-specific CD4⁺ T cell responses to a pathogenic poxvirus, providing an opportunity to both analyze CD4⁺ T cell responses to inform vaccination strategies as well as add new information to our growing understanding of the determinants necessary to drive a strong CD4⁺ T cell response.

To carry out these studies, the first step was to establish a panel of I-A^b–restricted ECTV epitopes that could be leveraged in subsequent studies. It was critical that we examine CD4⁺ T cell responses to many epitopes because not all epitopes have the same properties. For this reason, bulk responses to the whole virus would also be of limited utility; detection of a polyclonal CD4⁺ T cell response to virus does not indicate whether the response is targeted largely to a single epitope or is distributed across many epitopes. Prior to our studies, there had been a dearth of published MHC-II–restricted epitopes derived from the ECTV proteome. A previous study performed a large-scale epitope analysis of the vaccine strain VACV–Western Reserve and identified 14 I-A^b–restricted epitopes, the majority of which (12/14) bound with intermediate/high affinity to I-A^b (40). In designing our screen, we consciously excluded these previously identified epitopes to focus on identification of novel epitopes. Our peptide screen, selected from a virtual overlapping peptide library, combined with predictive binding affinity yielded 42 novel I-A^b–restricted epitopes that consistently and, in many cases, robustly reactivated CD4⁺ T cells from ECTV-infected mice. As the binding algorithms continue to undergo refinement (47), a screen of the entire ECTV peptidome would likely have yielded additional epitopes, both previously published and novel.

CD4⁺ T cell epitopes are potentially skewed toward certain types of source proteins, such as abundantly produced late-phase proteins. The parental proteins that produced our epitopes are from every phase of the poxvirus lifecycle, as was previously observed with a large-scale epitope mapping study of VACV (40). In contrast to this previous study where the epitope-bearing early gene products were not virulence factors, several of our epitopes deriving from early proteins are predicted to be encoded by virulence genes. This most likely reflects the larger set of epitopes identified in our study, as, indeed, we too found a higher number of epitopes deriving from early-expressed proteins involved in gene regulation as compared with virulence.

Our first objective was to use this panel of novel ECTV epitopes to determine the potency of the CD4⁺ T cell response to inactivated poxvirus. The adverse reactions to immunization with live VACV are considerable (9–16), but the live vaccine has remained the standard because early attempts to immunize with inactivated virus failed to elicit strong Ab responses and protection (17–19). Based on our previous work in other viral systems, we hypothesized that one reason for the failure of inactivated poxviruses to provide protection is the absence of a strong CD4⁺ T cell response. The basis for CD4⁺ T cell activation and recognition of peptide/MHC-II complexes has been deduced mainly through use of stable globular proteins. This has perhaps led to the assumption that CD4⁺ T cell responses to virus infections entails internalization of whole virions followed by Ag processing and MHC-II loading via the classical, endosomal pathway. However, whole virions are far more complex structures than monomeric globular proteins that may not be so readily converted to MHC-II–binding peptides. Furthermore, viruses that infect APCs such as influenza and ECTV can interact with host processing machinery beyond the endocytic compartments. Indeed, our previous studies with influenza demonstrated that, in fact, inactivated virions were poor drivers of a CD4⁺ T cell response compared with live virus (1). By extension, if processability of whole virions is an important factor in the strength of the CD4⁺ T cell response, the more stable and refractory to processing the virion in question, the poorer the CD4⁺ T cell response should be. Although influenza is a fragile RNA virus surviving only minutes outside the host (22), poxvirus virions are far more durable structures that can even be lyophilized without loss of infectivity, a property that was instrumental in the eradication of natural smallpox (25).

When we assessed the recall response of CD4⁺ T cells isolated from mice immunized with inactivated ECTV, no consistent specificities were detected against any of our 42 novel epitopes; one epitope sporadically induced a vastly reduced response compared with what is elicited by live virus. In addition, unlike our previous work with influenza, the poor epitope-specific CD4⁺ T cell response was not amplified in a prime/boost scenario. The results further support our prediction that poxvirions are poor MHC-II Ag-processing substrates. Furthermore, we did not observe responses to the epitopes derived from late structural proteins present in the mature virion, suggesting limited efficacy of endosomal proteases in the digestion of whole virions.

Despite the absence of a detectable CD4⁺ T cell response, we did observe a low level of serum IgG against inactivated ECTV. Considering the lack of a strong response to any peptide we tested, it seems unlikely that class switching was driven by CD4⁺ T cells whose specificities were not analyzed. Rather, we favor the notion of T-independent class switching, as has been observed in other viral infections (48–55). An alternative is the presence of APC populations in the peritoneum that are distinct from those in the skin because s.c. immunization did not produce detectable serum anti-ECTV IgG. Regardless of the mechanism at play, this class- switched Ab response was ∼8-fold lower than that induced by a much lower input dose of live ECTV and was not sufficient to protect mice from a secondary challenge.

There are several factors, in addition to poor virion processability, that could contribute to the poor CD4⁺ T cell responses generated against inactivated virus. One is reduced viral load due to the absence of replication. We aimed to address this factor by immunizing with a much higher dose of inactivated virus compared with live virus and also by altering the immunization route to one that would result in more direct exposure of viral particles to immune cells. It is also possible that lack of inflammatory signals raised against inactivated virus could contribute to poor CD4⁺ T cell activation. This was a possibility that we ruled out in our previous influenza studies; coimmunization with an infectious non–cross-reactive strain of influenza that provided inflammatory signals had no impact on response to inactivated virus (1). Recently published work corroborates our finding that inactivated VACV does not prime a protective Ab response; however, the inclusion of particular adjuvants were reported to confer protection (33, 56). Therefore, we used the same adjuvant as was previously published preparation in conjunction with our inactivated virus. Although we did observe induction of serum anti-ECTV IgG at low levels with inclusion of adjuvant, we did not observe enhancement of CD4⁺ T cell responses. This suggests that lack of inflammatory signals against inactivated poxvirus is not a causative explanation for our results. It should be noted, however, that more dramatic Ab production might have been observed had adjuvantation been implemented in a prime/boost immunization as in (33).

Having determined that CD4⁺ T cell responses are compromised following inactivated poxvirus immunization, we sought to probe directly whether poor processability of whole virions was in fact a mechanistic explanation. Therefore, after growing ECTV in cell culture, we separated whole virions from a fraction containing poxviral proteins produced during replication but not incorporated into virions. Despite the presence of more poxvirus-derived proteins in the virus fraction than in the protein fraction, we observed that the protein fraction facilitated considerably more robust CD4⁺ T cell activation than the virus fraction. Thus, virion-free protein is far more efficiently processed, resulting in stronger CD4⁺ T cell responses, perhaps providing a basis for the promising results observed with subunit vaccines (5–7). These data are consistent with published literature suggesting that individual proteins rather than whole virions drive CD4⁺ T cell–B cell collaboration during the response to VACV infection (57).

In this paper, we have demonstrated that inactivated poxvirus elicits poor CD4⁺ T cell responses and minimal protection to a secondary challenge. Further, we provided a mechanistic explanation in that whole virions are poor substrates for the MHC-II–processing machinery, which leads to poor CD4⁺ T cell activation. This work adds to the growing evidence in the field that MHC-II processing and presentation is much more complex than generally envisioned. In addition, it points to principles that may be critical considerations in the rational design of vaccines intended to provoke strong CD4⁺ T cell responses.

We thank Stuart N. Isaacs for critical reading of the manuscript.

The authors have no financial conflicts of interest.