Cutting Edge: Protection by Antiviral Memory CD8 T Cells Requires Rapidly Produced Antigen in Large Amounts

CD8 T cells scan the surface of professional APC in secondary lymphoid organs in search of an antigenic peptide bound to MHC class I (MHC I) molecules. In all cells, MHC I molecules bind peptides derived from the normal degradation of the cellular proteome and display them at the cell surface (1). These self-peptides normally do not induce a CD8 T cell response. However, when cells become infected by a virus, some of the peptides presented by MHC I derive from the degradation of viral proteins. When the cells presenting viral peptides are professional APC, specific CD8 T cells become activated, increase their numbers, distribute throughout the body, and kill any cell that presents their cognate peptide on MHC I. After the infection subsides, an expanded pool of virus-specific CD8 T cells remains. These cells, known as memory CD8 T cells, can potentially protect from subsequent infections by viruses carrying the cognate peptide (2). Thus, memory CD8 T cells induced by vaccines can theoretically protect from virulent viruses. Yet, although CD8 T cell vaccines have been effective in some experimental setups, they have failed to fulfill their promise (3, 4). A reason for this failure might be that we do not know what makes an Ag an effective target of protective memory CD8 T cells.

Orthopoxviruses (OPV) are a genus of DNA viruses that include the agent of human smallpox variola virus, vaccinia virus (VACV), which was used as the vaccine that eradicated smallpox, and ectromelia virus (ECTV), a pathogen of the laboratory mouse. Following footpad infection, ECTV rapidly disseminates through the lympho-hematogenous route, causing a lethal disease known as mousepox in susceptible mice but not in resistant strains of mice. The main target organs of ECTV are the liver and the spleen, where it causes massive necrosis in mousepox-susceptible strains but not in mousepox-resistant strains. Indeed, death in susceptible strains is thought to be due to the liver necrosis.

As with many other viruses, the anti-OPV CD8 T cell responses are directed to multiple peptides; the one eliciting the strongest CD8 T cell response is called “immunodominant,” and those that induce lower responses are called “subdominant.” The immunodominant peptide of ECTV and VACV is TSYKFESV (amino acid single letter code) (5). It is derived from the degradation of the high-abundance protein B8, an IFN-γ decoy receptor encoded by the early/late genes B8R in VACV and EVM158 in ECTV (6). TSYKFESV binds with high affinity to the mouse MHC I molecule K^b, which is present in both mousepox-resistant C57BL/6 (B6) and mousepox-susceptible B6.D2-(D6Mit149-D6Mit15)/LusJ (B6.D2-D6) mice, a B6 congenic recombinant inbred strain.

After footpad infection with ECTV, B6 mice show few signs of disease and mount strong antiviral CD8 T cell responses to multiple peptides but dominated by anti-TSYKFESV cells (5, 7). Conversely, male B6.D2-D6 mice die of mousepox with almost absent CD8 T cell responses and extreme lymphopenia in secondary lymphoid organs (7, 8). Yet, the inadequate B6.D2-D6 responses to ECTV are not due to an intrinsic defect in CD8 T cell immunity because they mount very strong CD8 T cell responses to VACV, including to TSYKFESV. Notably, B6.D2-D6 mice can be protected from mousepox by memory CD8 T cells specific for TSYKFESV and to some but not all subdominant peptides (7). Thus, not all immunogenic peptides are obligate targets of protective memory CD8 T cells, suggesting a possible reason for the failure of many CD8 T cell vaccines.

Stocks were produced and titers were determined as previously described (9). ECTV recombinants were generated as previously described (10). The scheme for the construction of the viruses in this paper is depicted in Supplemental Fig. 1A and 1B.

L929 cells (American Type Culture Collection) in 24-well plates were infected with 1 PFU per cell, and RNA was extracted at the indicated times. When indicated, AraC (Sigma-Aldrich) was added at a concentration of 40 μg/ml. Quantitative PCR was performed as previously described (11).

We used standard procedures. Rabbit polyclonal anti-B8 (12) was from Dr. R. Mark Buller (Saint Louis University, Saint Louis, MO). Rabbit polyclonal anti-EVM166 has been described (13). Anti-β actin Ab was from Abcam.

B6 mice were from Taconic, and B6.SJL (CD45.1) mice were from NCI. B6.D2-D6 mice were bred at Fox Chase Cancer Center. Vaccination with dendritic cells (DC) pulsed with TSYKFESV (DC-TSYFESV) was as before. This results in ∼4% of the CD8 T cells specific for TSYKFESV 1 mo after boosting (7). Unless indicated, mice were infected with ECTV in the left footpad with 30 μl PBS containing 3 × 10³ PFU and were monitored as before (14). These studies were approved by the Institutional Animal Care and Use Committee of Fox Chase Cancer Center.

Histopathology was as previously described (13).

Flow cytometry was as previously described (14). Abs to CD4, CD8a, CD45.2, and CD69 were from BioLegend. Anti-human/mouse Granzyme B (GzmB) was from Caltag. H-2K^b:Ig recombinant fusion protein (Dimer-X; BD) incubated with synthetic peptides was used as recommended by the manufacturer.

Unless indicated, data correspond to one representative experiment of at least two similar experiments with groups of three to six mice. Statistical analysis was performed using GraphPad Prism software. One-way ANOVA, unpaired two-tailed t, Mann–Whitney, or log-rank (Mantel-Cox) tests were used as applicable.

We made ECTV lacking EVM158 but producing VACV B8 (from an introduced VACV B8R gene) driven by the promoters of the VACV genes B8R, C3L, and D2L (hereafter, ECTV pB8R, pC3L, and pD2L, respectively). Because these viruses carried the Escherichia coli gene gpt for selection, we also made an EVM158-null virus that carried gpt but not B8R (in this article, ECTV gpt) (Supplemental Fig. 1A, 1B). We made the viruses producing VACV B8 and not ECTV B8 because VACV B8 fully conserves TSYKFESV but does not bind mouse IFN-γ (6). We chose the promoters of B8R, C3L, and D2L given their expected levels of expression, according to Assarsson et al. (15).

During OPV infection, early genes are expressed before and late genes are expressed after DNA replication (16). In cells infected with ECTV pB8R and pD2L, expression of B8R was detectable by reverse transcription quantitative PCR at 2 h postinfection (hpi) and continuously increased up to 24 hpi (the last time point tested), suggesting B8R is an early/late gene in ECTV pB8R and pD2L. Yet, expression was always higher in cells infected with ECTV pB8R than with pD2L. In cells infected with ECTV pC3L, B8R expression was not detectable at 2 hpi, was low at 3 hpi, and steadily increased up to 24 hpi (Supplemental Fig. 1C), suggesting B8R is a late gene in ECTV pC3L. Infection for 8 h in the presence of cytosine arabinoside (araC), which prevents the expression of late but not early viral genes (17), significantly decreased the expression of B8R in cells infected with ECTV pC3L but not with ECTV pB8R or ECTV pD2L (Supplemental Fig. 1D), further suggesting that B8R is early/late in ECTV pB8R and pD2L, and late in ECTV pC3L. Western blot (WB) at 6 hpi showed that cells infected with ECTV pB8R produced more B8 than cells infected with ECTV pD2L, and cells infected with ECTV pC3L did not produce B8 (Fig. 1A). Control WB for actin and the viral protein EVM166 demonstrated similar loading and infection efficiency, respectively. Serial dilution indicated that the difference in B8 content between cells infected with pB8R and pD2L was ∼3-fold (Fig. 1B). WB at 24 hpi showed that cells infected with ECTV pB8R and p3CL produced high amounts of B8, whereas cells infected with p2DL produced less (Fig. 1C). Serial dilution showed that cells infected with ECTV p3CL and pB8R contained similar amounts of B8 (Fig. 1D), whereas those infected with pD2L contained ∼30-fold less (Fig. 1E). Thus, ECTV pB8R produces relatively high and ECTV pD2L produces low amounts of B8 protein during early and late infection (early/late high and early/late low, respectively), and ECTV pC3L produces high amounts of B8 protein but only late (late/high). Importantly, all the viruses remained similarly fully virulent because 3000 PFU killed B6.D2-D6 mice with indistinguishable kinetics. Control mousepox-resistant B6 mice infected with ECTV gpt survived the infection (Fig. 1E).

We have shown that memory CD8 T cells induced by vaccination with DC-TSYKFESV protect B6.D2-D6 mice from death and pathology caused by wild type (WT) ECTV (7). Hence, we vaccinated B6.D2-D6 mice with DC-TSYKFESV or DC-SIINFEKL as control. As before (7), this resulted in ∼4% TSYKFESV⁺ CD8⁺ T cells (data not shown). Next, we challenged the vaccinated mice with the different viruses to determine survival, virus loads, histopathology, and memory CD8 T cell responses (experimental scheme in Supplemental Fig. 2A). DC-TSYKFESV–vaccinated mice were fully and significantly protected from death caused by ECTV pB8R (early/late high) or ECTV pC3L (late high) but not from ECTV pD2L (early/late low) or control ECTV gpt. Control mice immunized with DC-SIINFEKL succumbed to ECTV pB8R (Fig. 2A). Thus, DC-TSYKFESV vaccination fully protects from lethality when B8 is produced in high amounts, whether early/late (ECTV pB8R) or only late (ECTV pC3L) but does not protect when B8 is produced early/late in low amounts (ECTV pD2L). This was not due to diverging virulence of the viruses because pB8R (full protection) and pD2L (no protection) were both fully lethal not only at 3000 PFU (Fig. 1E) but also at 300 PFU and partially lethal at 30 PFU (data not shown).

Although protection from death is essential, vaccines should also protect from pathology. Notably, mice immunized with DC-TSYKFESV controlled the replication ECTV pB8R (early/late high) in the spleen (Fig. 2B) and liver (Fig. 2C) significantly better than the replication of all the other viruses, including ECTV pC3L (late high). Consistently, the livers of mice infected with ECTV pB8R had much less staining with an Ab to the viral protein EVM135 than the livers of mice infected with the other viruses (Supplemental Fig. 2B). Thus, DC-TSYKFESV immunization protected from lethal mousepox caused by ECTV pB8R (early/late high) and pC3L (late high), but only those infected with ECTV pB8R were protected from high virus loads and organ pathology. These findings indicated that the timing of Ag expression is critical for protection against pathology and disease.

Next, we analyzed the memory CD8 T cell recall responses of DC-TSYKFESV–immunized B6.D2-D6 mice at 7 d after challenge with the different viruses. Only the mice infected with ECTV pB8R had significantly more TSYKFESV-specific CD8 T cells in their livers and spleens than mice challenged with ECTV gpt (Fig. 2D, 2E, Supplemental Fig. 2C). The ability of ECTV pB8R to induce strong memory CD8 T cell responses in susceptible mice was likely associated with increased Ag presentation because when we adoptively transferred CFSE-labeled splenocytes from DC-TSYKFESV–immunized B6 mice into B6-CD45.1 mice (experimental scheme in Supplemental Fig. 2D), donor anti-TSYKFESV memory CD8 T cells proliferated and expanded to a significantly larger extent in mice challenged with ECTV pB8R than in mice challenged with ECTV pD2L (Fig. 2F, Supplemental Fig. 2E), and more of them produced GzmB (Fig. 2G, Supplemental Fig. 2E). Thus, only a virus that rapidly produced high amounts of B8 significantly recalled and expanded the anti-TSYKFESV memory CD8 T cells in mousepox-susceptible B6.D2-D6 mice and induced better proliferation and effector differentiation in mousepox-resistant B6-CD45.1 mice.

We next asked whether reactivating the memory CD8 T cells could improve protection from ECTV pD2L. We vaccinated B6.D2-D6 mice with DC-TSYKFESV or DC-SIINFEKL (as control) as in the experiments above, but some mice were boosted 3 d before analysis and/or virus challenge with DC-TSYKFESV or DC-SIINFEKL (experimental scheme in Supplemental Fig. 2F). Boosting with DC-TSYKFESV but not with DC-SIINFEKL resulted in the expression of GzmB and CD69 in anti-TSYKFESV CD8 T cells, indicating that they had become effectors (Fig. 3A). Mice vaccinated and boosted with DC-TSYKFESV survived infection with ECTV pD2L (Fig. 3B). However, the protection induced by boosting was incomplete, because similarly treated mice had high virus loads in the liver (Fig. 3C) and spleen (Fig. 3D) 7 d after challenge, had a significant reduction in the cellularity of their spleens, which is a typical sign of mousepox (Supplemental Fig. 2G), and had wide dispersal of the virus in their livers as determined by immunohistochemistry (Supplemental Fig. 2H). Thus, although effector CD8 T cells can improve survival, they do not protect from disease when the virus produces the Ag in low quantities.

It was possible that anti-TSYKFESV memory CD8 T cells failed to protect from a virus producing low B8 because the amount of B8 was insufficient for the display of K^b-TSYKFESV complexes at the surface of infected cells in vivo. Alternatively, K^b-TSYKFESV complexes could be displayed but at densities that were insufficient to recall a protective memory CD8 T cell response in susceptible mice. Thus, we next tested whether the different viruses could elicit anti-TSYKFESV CD8 T cell responses in mousepox-resistant B6 mice. We found that B6 mice infected with ECTV pB8R, pC3L, or pD2L but not ECTV gpt had significantly more TSYKEFSV-specific cells than uninfected mice at 7 d postinfection (dpi). However, the anti-TSYKFESV responses were highest in mice infected with ECTV pB8R (early/late high), intermediate in mice infected with ECTV pC3L (late high), and lowest in mice infected with ECTV pD2L (early/late low) (Fig. 4A, 4B), whereas the overall anti-ECTV CD8 T cell responses (the sum of all the specificities), as determined by gzmB staining of total CD8 T cells, were similar for all the viruses (Fig. 4A, 4C). Although these results confirm that the amount of Ag affects the magnitude of the primary CD8 T cell response (18), when compared with other known ECTV epitopes (5, 7), the anti-TSYKFESV response was immunodominant in mice infected with ECTV pB8R and codominant with ECTV pC3L or pD2L (Fig. 4D). Thus, in vivo, K^b-TSYFESV is displayed at the surface of cells in sufficient amounts to elicit dominant or codominant CD8 T cell response for all viruses in mousepox-resistant mice.

We show that memory CD8 T cells are efficiently recalled and fully protect a susceptible host from disease and/or death only when their cognate Ag is produced in relatively large quantities and with rapid kinetics. Notably, ECTV pC3L (late high) and pD2L (early/late low) were not good targets of protective anti-TSYKFESV memory CD8 T cells despite that the amounts of B8 they produced resulted in presentation of K^b-TSYKFESV complexes sufficient to elicit codominant antiviral CD8 T cell response in resistant mice. Hence, epitopes that induce potent primary CD8 T cell responses may not induce protective recall responses, and the amount of Ag produced by the virus might be more important for the recall of memory CD8 T cells in susceptible hosts than for the elicitation of a primary response in resistant hosts. Preactivating the memory cells before challenge with pD2L (early/late low) significantly improved survival but did not prevent pathology, suggesting that low-expressed Ags are also inadequate targets for secondary effector CD8 T cells. In addition, our data also demonstrate that the amount of Ag produced by a virus affects the strength but is not critical for the induction of a strong primary CD8 T cell response in genetically resistant hosts.

Our work seems to contradict the dogmatic view that the amount of peptide-MHC at the surface of target cells may not be important for memory CD8 T cell protection. This idea likely arose from the finding that in vitro, one or a few peptide-MHC complexes are sufficient for CD8 T cell–mediated killing (19). Although for technical reasons we were unable to directly quantify K^b-TSYKFESV complexes or processed TSYKFESV, from the B8 expression data and the experiments with adoptively transferred memory cells, we propose that in vivo, a relatively high density of peptide-MHC at the surface of infected cells is necessary for full antiviral protection by secondary effectors. A possible reason for our finding may be that protection requires not only cytotoxicity but also rapid CD8 T cell expansion, which requires high levels of peptide-MHC at the surface of APC (20). This is supported by our adoptive transfer experiments.

Our results show that an insufficient quantity of Ag produced by the pathogen may explain why many naturally immunogenic peptides fail to recall protective memory CD8 T cell responses. This finding contrasts with the prevalent view that the obstacle for the production of protective CD8 T cell vaccines is in the induction of a sufficient number of high-quality memory cells (21, 22). We have previously shown that TSYKFESV-specific memory CD8 T cells deficient in IFN-γ (which by definition are “poor quality”) fully protect B6.D2-D6 mice from ECTV WT lethality and pathology (23). Thus, for protective immunity, Ag amount might be more important than memory CD8 T cell quality. In summary, our results suggest that the vaccines more likely to succeed are those that induce memory CD8 T cells directed to Ags produced in large amounts and possibly at early stages of the virus life cycle and are likely to fail when the antigenic protein is produced in low amounts. Our results are in principle applicable to a poxvirus and are likely useful to develop novel vaccines to large DNA viruses. Whether it also applies to smaller RNA viruses needs to be tested, but it may be complicated to do because of the difficulty at unlinking genetic manipulation and virulence in this type of viruses. It would also be of interest to determine if this also applies to cancer vaccines.

We thank Fox Chase Cancer Center Laboratory Animal, Biostatistics, Flow Cytometry, Histopathology, and Tissue Culture Facilities, Thomas Jefferson University laboratory animal facility, Felicia Roscoe for technical work, Dr. Andres Klein-Szanto for histopathology, Dr. Samuel Litwin for statistical analyses, and Jennifer Wilson for editing the manuscript.

The authors have no financial conflicts of interest.