Immune Protection by a Cytomegalovirus Vaccine Vector Expressing a Single Low-Avidity Epitope

Numerous Ags are weak and bind poorly to TCRs on responding CD8 T cells (1, 2), which poses a substantial challenge for the development of vaccines targeting intracellular pathogens or tumors. Vaccination strategies based on CMV vaccine vectors have shown unparalleled T cell–based immune protection in animal experimental models. CMV vectors provided protection against Ebola (3), and simian AIDS in rhesus macaques (4, 5). Moreover, experiments in the mouse model showed that the introduction of a single MHC class I–restricted epitope into mouse CMV (MCMV) induces protective immune responses against viruses (6–8), intracellular bacteria (9), and tumors (10, 11), arguing that the primary mechanism of immune protection relies on the induction of CD8 T cell responses. It is important to note, however, that all of these studies focused on well-characterized epitopes, known to induce high-avidity immune responses. Therefore, it has remained unclear if CMV-based vaccine vectors encoding low-avidity epitopes which are poorly recognized by CD8 T cells may also provide immune protection.

The design of CMV-based vaccine vectors demands a detailed understanding of the CMV-induced immune response. Human CMV is a ubiquitous herpesvirus that latently persists in most adults worldwide (12). It induces a life-long effector memory T cell response that is stronger than any other reported in human medicine (13). A similar immune response has been observed in the mouse model of CMV infection and called “memory inflation” (14): CD8 T cells specific for the MCMV accumulate in blood (14) and tissues (15) over weeks and months postinfection; never contract from peak response levels; and retain for life the effector memory phenotype (TEM), characterized by low CD62L and high CD44 surface expression. It has been observed that CD8 T cell responses to MCMV vary between MCMV Ags (16–18). Some MCMV epitopes like the m139-encoded epitope TVYGFCLL or the ie3 epitope RALEYKNL induce inflationary TEM responses. The M45 epitope HGRNASFI, or the M57 epitope SCLEFWQRV, however, induce conventional response kinetics, peaking at day postinfection (dpi) 7, followed by a contraction phase during which the majority of cells assume a central memory phenotype (TCM) (18). We showed previously that this difference does not depend on the sequence of the antigenic epitope but on the context of its gene expression (7). MCMV recombinants expressing the immunodominant HSV-1 epitope SSIEFARL (SL) in the context of the M45 gene induced conventional responses, whereas the same epitope fused to the immediate-early 2 (ie2) gene induced an inflationary CD8 T cell response (7). Furthermore, essentially all other successful CMV-based vaccine vectors induced TEM CD8 T cell responses. Hence, it has been proposed that the TEM of responding T cells may be critical for immune protection against intracellular pathogens by CMV vectors (19), yet a formal proof for this assertion has remained elusive. It has been shown that inflationary immune responses could be redirected by inserting exogenous Ags into the MCMV genome (6).

We generated MCMV recombinants expressing the low-avidity antigenic epitope KCSRNRQYL (KNL) (20) in the exact same positions as the high-avidity SL in our previous publication (7). This allowed us to study the effect of Ag avidity on immune kinetics, phenotype, and protection. We show that a single low-avidity epitope is sufficient to induce protective immunity upon challenge with recombinant vaccinia virus (VACV) if the epitope is expressed within the ie2 gene of a recombinant MCMV vaccine vector, but not if it is expressed by the early M45 gene. In contrast, the high-avidity control peptide was protective regardless of its expression context. We explain this pattern of protectivity by showing that the protective MCMV vectors induced inflationary CD8 T cell responses, whereas the recombinant with the low-avidity epitope expressed within the M45 gene induced a conventional response, where KNL-specific cells assumed a TCM over time. Finally, we show that the inflation and phenotype of Ag-specific cells are a result of competition of Ags for T cell responses, where epitopes with higher avidity and stronger expression crowd out other epitopes, but the intrinsic antigenic weakness of an epitope may be compensated by a strong expression context.

All animal experiments were performed in compliance with the German Animal Welfare Act (TierSchG BGBl. I S. 1206, 1313; May 18, 2006) and Directive 2010/63/EU. The mice were handled in accordance with good animal practice as defined by the Federation for Laboratory Animal Science Associations and Gesellschaft für Versuchstierkunde/Society of Laboratory Animal Science. All animal experiments were approved by the responsible state office (Lower Saxony State Office of Consumer Protection and Food Safety) under permits number 33.9- 42502-04-11/0426 and -14/1711.

129S2/SvPas Crl (129S2/SvPas) mice were purchased from Charles River (Sulzfeld, Germany), C57BL/6JRj mice were purchased from Janvier Labs. Mice were housed at the animal facility of the Helmholtz Centre for Infection Research under special pathogen-free conditions.

M2-10B4 (CRL-1972) and NIH 3T3 (CRL-1658) were maintained in DMEM supplemented with 10% FCS, 1% glutamine, and 1% penicillin/streptomycin; RK13 (CRL-1414) were maintained in MEM-Earl with 10% FCS, 1% glutamine, and 1% penicillin/streptomycin (all from American Type Culture Collection). C57BL/6 murine embryonic fibroblasts (MEFs) were prepared and maintained as described previously (21).

Bacterial artificial chromosome (BAC)–derived, wild-type MCMV (MCMV-WT) was propagated as described previously (22). The generation of the recombinant MCMV viruses MCMV-ie2-SL and MCMV-M45-SL was described previously (7). MCMV^ie2KNL and MCMV^M45KNL were generated following the protocol used in Dekhtiarenko et al. (7) inserting the AAKCSRNRQYL sequence (5′-GCT GCC AAA TGT TCT CGT AAT CGT CAA TAT TTA-3′) at the 3′ end of the MCMV ie2 gene (nucleotide positions 187,296–187,297 of the MCMV-WT genome). For MCMV^M45KNL, we inserted the same AAKCSRNRQYL-encoding sequence at the 3′ end of M45 gene (between nucleotides 59,513 and 59,514). VACV^SL (rVV-ES-gB498-505) has been described previously (23). VACV^KNL was generated by inserting the KNL coding sequence with an alanine spacer fused to the adenovirus type 5 E3/19k endoplasmic reticulum insertion sequence under control of the VACV P7.5 promoter into the TK locus of the VACV genome (strain WR) using a pSC11-based transfer vector (originally provided by B. Moss, National Institutes of Health, Bethesda, MD). VACV^KNL was selected and amplified following standard methodology (24). The VACVs were further propagated on RK-13 cells.

The peptides M45 (H-2D^b–restricted ⁹⁸⁵HGIRNASFI⁹⁹³), m139 (H-2K^b–restricted ⁴¹⁹TVYGFCLL⁴²⁶) (18), the HSV-1 glycoprotein–derived epitope gB (H-2K^b–restricted ⁴⁹⁸SSIEFARL⁵⁰⁵) (25), and the H-Y peptide KNL (20) were synthesized and HPLC purified (65–95% purity) at the Helmholtz Centre for Infection Research peptide-synthesis platform.

Recombinant viruses were reconstituted from BACs by transfection of BAC DNA into MEFs using FuGENE transfection reagent (Promega) according to the manufacturer’s protocol. Viruses were propagated as described (26) and virus stocks were prepared from M2-10B4 lysates purified on a sucrose cushion as described previously (27). Virus titers were determined on MEFs by plaque assay.

Recombinant VACVs were propagated on RK-13 cells and purified on a sucrose cushion as described (28). Virus titers were determined by plaque assay on RK-13 cells.

Monolayers of NIH 3T3 cells were infected at a multiplicity of infection of 0.1 with MCMV recombinants or MCMV-WT. After 1 h, the inoculum was removed, cells were washed with PBS, supplied with fresh medium, and incubated for 6 d. At selected time points postinfection, supernatants were collected and stored at −70°C until titration.

Female 6- to 12-wk-old mice were infected with purified, tissue culture–derived virus and housed in specific pathogen-free conditions throughout the experiment. MCMV-infected mice showing very weak immune priming (<30% of CD11a⁺ CD44⁺ in the CD8⁺ population) at dpi 7 were regarded as suboptimally infected and were excluded from the study. For experiments involving the establishment of infectious virus titers, organs were harvested under sterile conditions and stored at −70°C until titration.

MCMV from organ homogenate or tissue culture supernatants was titrated on MEFs as described previously (29). For identification of VACV titer in organs, titration was performed on RK-13 cells following standard methodology (28).

Cells were stimulated with peptides (1 μg/ml) and anti-CD28 and anti-CD49d (1.25 μg/ml; BD Biosciences) in 100 μl RPMI 1640 (supplemented with 10% FCS) for 6 h at 37°C with Brefeldin A (Cell Signaling Technology) added at a 10 μg/ml concentration for the last 5 h of stimulation. Negative control samples were generated for all tested groups by incubating cells in the same conditions, but in the absence of any peptide. Cells were tested for cytokine responses by intracellular-cytokine staining using flow cytometry. Cytokine responses observed in unstimulated samples were considered background noise due to unspecific Ab binding and were subtracted from the values observed in test samples. For the avidity assay, cells were stimulated with increasing concentrations of peptide (10⁻⁸, 10⁻⁷, 10⁻⁶, 5 × 10⁻⁶, 10⁻⁵, 5 × 10⁻⁵, 10⁻⁴, 5 × 10⁻⁴, 10⁻³, 5 × 10⁻³, 0.01, 0.05, 0.1, 0.5, and 1 μg/ml).

Single-cell suspensions were obtained by passing organ homogenates through 100-μm cell strainers. Cells were incubated with allophycocyanin-conjugated KNL-D^b or SL-K^b tetramers (tet) (obtained from Ramon Arens) for 30 min at room temperature, and subsequently stained for 30 min at 4°C with the following surface Abs: anti–CD4-Pacific Blue (clone GK1.5; BioLegend) or anti–CD4-BV650 (clone GK1.5; BD Biosciences), anti–CD8a-PerCP/Cy5.5 (clone 53-6.7; BioLegend), anti–CD44-Alexa Fluor 700 (clone IM7; BioLegend), anti–CD11a-PE-Cy7 (clone 2D7; BD Biosciences), anti–CD3-allophycocyanin-eFluor 780 (clone 17A2; eBioscience), anti–CD127-PE or -PE-Cy7 (clone A7R34; BioLegend), and anti–KLRG-1-bio (2F1/KLRG-1; BioLegend) or anti–KLRG-1-BV510 (2F1/KLRG-1; BD Biosciences). Where applicable, cells were washed and stained for 15 min at 4°C with Streptavidin-BV510 (BioLegend). For intracellular-cytokine staining, cells were subsequently fixed for 5 min with 100 ml IC Fixation Buffer (eBioscience), followed by 3 min permeabilization with 100 ml Permeabilization Buffer (eBioscience), and overnight incubation with anti–IFN-γ-allophycocyanin (clone XMG1.2; BioLegend). For analysis of Ki-67 and Bcl2 expression, cells were fixed for 20 min at room temperature using FoxP3/Transcription Factor Buffer Set (eBioscience) according to the manufacturer’s instructions, followed by a 30 min incubation with anti–Ki-67-PE (16A8; BioLegend) and anti–Bcl2-AlexaFluor488 (BCL/10C4; BioLegend). Cells were washed and an LSR Fortessa Cytometer (BD Biosciences) was used for cell acquisition. Cytometric results were analyzed with FlowJo software (version 9.8.3 for Mac).

Statistical analysis was performed using the GraphPad Prism7 program. Multiple samples were compared using the Kruskal–Wallis test followed by Dunn post hoc analysis. The EC₅₀ was determined by normalizing the data to 100% at the highest peptide concentration, performing logarithmic transformation, followed by fitting a nonlinear curve (log[agonist] versus response [three parameters]).

To define the role of Ag avidity on the kinetics of conventional and inflationary CD8 T cell responses in MCMV infection, we generated recombinant viruses expressing a model epitope inducing low-avidity responses. We chose the D^b-restricted epitope KNL which is encoded within the Y-chromosome Ag H-Y and is suggested to induce low-avidity CD8 T cell responses (20). By means of site-directed mutagenesis, we inserted this epitope in frame at the C terminus of the ie2 gene, known to induce inflationary responses (14); or of the M45 gene, inducing conventional responses (17). To normalize the flanking residues and thus peptide processing (30), two alanine residues have been added before the KNL sequence (Fig. 1A). Thus we generated the recombinant MCMVs that we called MCMV^ie2KNL and MCMV^M45KNL, respectively. When MCMV^ie2KNL and MCMV^M45KNL were tested for overall fitness in vitro in a multistep growth kinetic assay in NIH 3T3 cells, both mutants replicated like MCMV-WT (Fig. 1B). Additionally, we analyzed viral growth in vivo on dpi 5 in spleen, lung, and liver, and on dpi 21 in salivary glands. Mutant replication in vivo was not impaired when compared with MCMV-WT (Fig. 1C). Hence, we excluded that any difference in Ag availability upon infection with the two recombinants was due to a difference in replicative fitness, allowing us to focus exclusively on the promoter activity as determinant of antigenicity.

Finally, we verified the functional avidity of responding CD8 T cells during acute infection or during latency. As controls, we used MCMV^ie2SL, a recombinant MCMV expressing the high-avidity, K^b-restricted epitope SL (25) at the exactly same location (7) as MCMV^ie2KNL. Splenocytes of mice that had been infected with MCMV^ie2SL or MCMV^ie2KNL for 7 d (acute) or 9 mo (latent) were stimulated with increasing concentrations of the corresponding peptide. As expected, CD8 T cells specific for SL showed a significantly higher EC₅₀ (acute: 0.55 ng/ml; latent: 0.88 ng/ml) than KNL-specific CD8 T cells (acute: 77.8 ng/ml; latent: 157.3 ng/ml). There was no significant difference in functional avidity of CD8 T cells of the same specificity during acute and latent MCMV infection (Fig. 1D), in line with published data showing no differences in functional avidity of CD8 T cells during the acute and the latent phase of infection (18).

Taken together, we generated MCMV mutants that induced a low-avidity response (KNL) compared with the established high-avidity response (SL). MCMV^ie2KNL and MCMV^M45KNL growth was comparable to MCMV-WT in vitro as well as in vivo, ensuring that any observed effects are exclusively due to the gene expression pattern of the inserted Ags.

To test the ability of the low-avidity epitope to provide immune protection against a viral challenge, we infected mice with MCMV^ie2KNL or MCMV^M45KNL and challenged them 8 mo later with recombinant VACVs expressing the same Ag. We infected female 129S2/SvPas mice, which share the H-2^b haplotype with C57BL/6 mice, but lack the Ly49H NKR and thus do not recognize the m157 MCMV protein (31, 32). As positive controls we used the previously published recombinants MCMV^ie2SL and MCMV^M45SL (7), which express the high-avidity epitope SL within exactly the same genomic sites as MCMV^ie2KNL and MCMV^M45KNL, respectively. These mice were challenged with a recombinant vaccinia expressing the SL epitope. Negative controls included MCMV-WT–infected mice challenged with either of the recombinant VACVs. Mice were i.p. infected with 10⁶ PFU VACV^SL (23) or a newly generated recombinant VACV^KNL (see 2Materials and Methods), and 7 d later sacrificed, after which the infectious vaccinia titers were determined in the ovaries.

MCMV^ie2KNL induced protective responses, but MCMV^M45KNL-infected mice showed the same viral load after challenge with VACV^KNL as MCMV-WT infected controls (Fig. 2A), demonstrating that the low-avidity epitope is protective when expressed within the immediate-early gene. In contrast, the responses against the high-avidity epitope were protective regardless of the genetic expression context (Fig. 2A). A similar pattern was observed at 4 d postchallenge (Fig. 2B), arguing that the T cell control of vaccinia growth may have acted very early upon infection. To validate this assumption, KNL- and SL-specific CD8 T cells were identified by tet staining in blood, spleen, and ovaries on day 4 postchallenge. We found tet⁺ cells in all vaccinated groups (data not shown). The highest percentage of Ag-specific cells in the CD8 population was observed in the ovaries (Fig. 2C), where MCMV^SL mutants induced up to 76% SL tet–specific CD8 T cells and the MCMV^ie2KNL-infected animals showed an average of 32% KNL-specific cells in the ovarian CD8 population; whereas only 9% of tet⁺ cells were found in the MCMV^M45KNL group, which matched the frequency of SL-specific cells in MCMV-WT control groups challenged by VACV^SL (data not shown).

These results indicate that protection was mirrored by the percentage of Ag-specific CD8 T cells induced in the ovaries after challenge. Therefore, low-avidity epitopes may provide immune protection when expressed within immediate-early genes of CMV-based vaccine vectors, but not later in the virus replicative cycle. Intriguingly, this limitation did not affect the high-avidity SL epitope, which was protective even when expressed within the M45 gene.

To understand the reasons for the poor immune protection and low percentage of Ag-specific CD8 T cells induced by MCMV^M45KNL, we considered that the context of epitope expression defines the kinetics and magnitude of CD8 responses (7). Therefore, we analyzed the kinetics of the immune response induced by the MCMV mutants expressing low- or high-avidity epitopes. SL- and KNL-specific CD8 T cell responses were analyzed in vitro by 6-h peptide stimulation of blood leukocytes with the corresponding peptides, followed by intracellular staining for IFN-γ and multiparametric flow cytometry (Fig. 3A).

The SL and the KNL epitope expressed within the ie2 gene induced inflationary immune kinetics, but the percentage of KNL-specific CD8 T cells was lower than that of SL-specific CD8 T cells at all time points. This difference was more pronounced on dpi 7 than on dpi 180, with 2% of KNL-specific CD8 T cells compared with 10% of SL-specific CD8 T cells during acute infection (factor 5) and 9% of KNL-specific CD8 T cells and 19% of SL-specific CD8 T cells during latent infection (factor 2.1) (Fig. 3B).

The kinetics of the immune responses induced by MCMV^M45SL or MCMV^M45KNL were also similar to each other. Both responses peaked by dpi 7, contracted by dpi 14, and remained low thereafter (Fig. 3C). The percentage of KNL-specific CD8 T cells was 6.7-fold lower on dpi 7, with only 3% of responding cells in MCMV^M45KNL but 20% of responding cells in MCMV^M45SL. By dpi 180 the immune response to KNL was down to 1%, whereas the SL-specific response remained at 5% (Fig. 3C), resulting in a fivefold difference in magnitude.

These results demonstrated that the overall kinetics of MCMV-induced immune responses depend on the genetic context of the epitope expression, whereas the avidity influences the size of the CD8 T cell response but not the timing of its peak and contraction. Even more importantly, the data showed that MCMV^M45KNL induced the weakest Ag-specific CD8 response among the tested vectors, yet even this virus induced clearly detectable memory-cell responses.

The CD8 T cell response to MCMV^M45KNL and the immune protection conferred by the same virus were clearly the weakest ones among the tested viruses, providing a plausible explanation for the lack of immune protection. In contrast, immune protection by MCMV^M45KNL was completely absent, but the CD8 T cell responses were merely lower, but clearly detectable. Therefore, we considered that the quantitative difference in immune responses might not exclusively explain the qualitative difference in immune protection. Hence, we characterized the immune phenotype of responding CD8 T cells. The kinetics of the Ag-specific, CD8 T cell response was determined via peptide restimulation and intracellular-cytokine staining of blood leukocytes. Because CD62L is rapidly shed from T cells upon in vitro peptide restimulation, phenotyping TCM and TEM cells on CD62L expression would have resulted in false-negative readouts. Therefore, we classified the cells according to the surface markers CD44 and KLRG-1, where naive T cells are negative for both markers, short-lived effector cells (SLEC) are CD44⁺KLRG-1⁺ whereas CD44⁺KLRG-1⁻ include the central-memory and a part of the effector-memory subset (T memory [TM]) (Fig. 4A).

First we analyzed the immune phenotype of SL- and KNL-specific CD8 T cells induced by epitopes expressed in the ie2 gene context. The m139-specific CD8 T cells from MCMV-WT–infected mice were included in the analysis as controls representing an endogenous inflationary CD8 T cell response. We found that almost all analyzed immune responses remained SLEC, as it would be expected for inflationary responses. The low avidity of the KNL-specific response did not affect the immune phenotype (Fig. 4B, 4D).

Next we took a closer look at the immune phenotype of SL- and KNL-specific CD8 T cells with the Ag expressed in the M45 gene context. The CD8 T cell response to the endogenous M45-derived epitope HGIRNASFI in MCMV-WT infection was used as point of comparison. As expected for immune responses following a conventional immune kinetic, the KNL-specific CD8 T cells and the M45-specific CD8 T cells showed a SLEC phenotype initially, but reverted toward a TM phenotype over time. Contrary to the phenotype associated with a conventional T cell response, the SL-specific CD8 T cells induced by MCMV^M45SL did not follow this pattern, but remained mostly SLEC (Fig. 4C, 4E).

To validate these results, additional mice were infected with the MCMV mutants for 3 mo and analyzed by tet staining and the traditional surface markers CD44 and CD62L for the immune phenotyping (Fig. 5A). We found a high percentage of T effector (Teff)/TEM cells (CD44⁺CD62L⁻) within the tet⁺ population of MCMV^ie2SL-, MCMV^ie2KNL-, and MCMV^M45SL-infected mice (87, 99, and 93%, respectively) and the lowest percentage of Teff/TEM cells in MCMV^M45KNL (76%). Conversely, MCMV^M45KNL-induced tet⁺ CD8 T cells exhibited the highest percentage of TCM (CD44⁺ CD62L⁺) cells (Fig. 5B). The difference in absolute numbers of tet⁺ Teff/TEM and tet⁺ TCM cells was even more pronounced. In MCMV^ie2SL- and MCMV^ie2KNL-infected mice we counted ∼140 cells/μl of tet⁺ Teff/TEM cells, in MCMV^M45SL roughly 70 cells/μl, but <3 cells/μl in MCMV^M45KNL-infected mice (Fig. 5C); a 25- to 50-fold difference. The difference between the counts of tet⁺ TCM cells was almost negligible in comparison. In this study we observed the highest counts in MCMV^ieSL and MCMV^M45SL and the lowest in MCMV^M45KNL, but with an obvious overlap in values and a median two- to fourfold difference (Fig. 5C).

Analyzing the tet⁺ CD8 T cells in blood allowed us to assess the increase of Ag-specific cells in the first 4 d postchallenge. Both high-avidity mutants showed an increase of tet⁺ CD8 T cells in blood by a factor of 10, whereas there was no or only little increase for the low-avidity mutants (Fig. 5D). To establish whether the low-avidity cells were responding at all, we stained the KNL tet⁺ cells with the cell-division marker Ki-67 and found a significant increase in the percentage of Ki-67^hi cells within the tet⁺ populations in blood, spleen, and ovaries as compared with Ki-67 values in the blood prior to challenge (Fig. 5E, Supplemental Fig. 1). Importantly, the fraction of Ki-67^hi cells in the tet⁺ subset was higher in mice infected with MCMV^M45KNL than in those infected with MCMV^ie2KNL, arguing for their maintained ability to respond to challenge.

To confirm that our results are not limited to a specific mouse strain, we repeated these experiments using C57BL/6 mice, where MCMV recognition by the NKR Ly49H represses MCMV replication and impairs early CD8 T cell responses (33). The results indicated that the described effects are not limited to the SV129 strain, but can also be observed in the MCMV-resistant mouse strain (Supplemental Fig. 2A–F).

Taken together, our data indicated that the lack of protection in MCMV^M45KNL cannot be explained exclusively by unresponsiveness in the face of challenge, but rather by a lack of Ag-specific Teff/TEM cells.

Whereas immune protection correlated with the Teff phenotype/TEM of responding CD8 T cells, it remained unclear why the KNL epitope expressed within the M45 gene was the only one failing to elicit substantial Teff/TEM responses (Fig. 5C). It was not defined by the gene expression context alone, because the high-avidity epitope expressed from the same locus induced Teff/TEM responses, and it was not merely a function of peptide avidity, because the same epitope was able to induce Teff/TEM responses and immune protection when expressed within the ie2 gene. Therefore, the Teff/TEM response was largely attenuated only when suboptimal promoter activity was combined with low peptide avidity. Because inflationary epitopes compete for immunodominance with each other (7, 34, 35), we hypothesized that Teff/TEM responses to an epitope may be repressed by the presence of peptides with higher avidity and gene expression, but only if both of these parameters were to result in a competitive disadvantage.

To test if this might occur, we tested the responses against the natural inflationary m139 peptide TVYGFCLL in the presence of the high-avidity SL epitope driven by the strong ie2 promoter, the low-avidity KNL peptide driven by the weaker M45 promoter, and all combinations in between (Fig. 6A). Responses were reduced on dpi 7 in MCMV^ie2SL- and MCMV^M45SL-infected animals, but not in the case of viruses expressing the low-avidity epitope (Fig. 6B, 6C). By dpi 180 the response remained suppressed only in the context of the infection with the MCMV^ie2SL virus (Fig. 6B) strongly expressing the high-avidity epitope. In contrast, endogenous CD8 T cell responses at later time points were neither outcompeted by the strong expression of the low-avidity Ag (Fig. 6B) nor by the high-avidity epitope when expressed by the weaker promoter (Fig. 6C).

Analysis of the immune phenotype of m139-specific immune responses showed comparable levels of TM (Fig. 6D, 6E) and SLEC (Fig. 6F, 6G) in mice infected with MCMV-WT and those infected with MCMV^ie2KNL, MCMV^M45KNL, or MCMV^M45SL. In mice infected with MCMV^ie2SL, however, the percentage of SLEC was significantly reduced (69% on average) compared with MCMV-WT (average 88%) (Fig. 6F). In parallel, TM levels increased to 30% compared with 11% in MCMV-WT on dpi 180 (Fig. 6D). Hence, we compared the m139-responding cells in MCMV-WT or MCMV^ie2SL-infected mice in more detail. We calculated the fraction of TM or SLEC responders within the CD8 pool at all time points. Although the percentage of TM responses to m139 were essentially identical (Fig. 6H), we observed a clear and significant reduction in SLEC responses at all time points (Fig. 6I). In contrast, the response against the conventional M45 peptide HGIRNASFI was significantly impaired on dpi 7 in both SL mutants but not the KNL mutants (Supplemental Fig. 3A–C), and showed no obvious difference of immune phenotype compared with wild type (Supplemental Fig. 3D–G), arguing that the effects of the peptide competition on the phenotype of responding cells are restricted to epitopes that naturally sustain SLEC responses.

Taken together these data suggested that peptide competition depends on the combination of peptide avidity and the gene expression context. Therefore, the low-avidity Ag KNL in the context of weak gene expression was unable to compete with endogenous epitopes, resulting in the TCM of the immune response and the absence of protection against the vaccinia challenge.

It has been shown previously that MCMV epitopes induce diverse immune kinetics and phenotypes (18) and that the kinetics of immune responses depends on the promoter regulating Ag expression (7). In this article, we show that epitope avidity plays an equally important role in defining the kinetics and the phenotype of responding cells. Direct comparison of high- and low-avidity epitopes expressed by MCMV recombinants from the same location revealed similar response dynamics, arguing that the overall quality of responses depends on the context of antigenic expression. In contrast, the magnitude of responses was significantly lower in the low-avidity mutants (Fig. 3B, 3C). This difference was more pronounced during acute infection than in the context of memory inflation. A possible explanation for this might be that long-term Ag availability during latency may have compensated for the weak avidity of T cell binding. In that case, the prolonged period of stimulation in conditions of memory inflation would offset differences in avidity, whereas acute exposure to the epitope would accentuate them, which is consistent with our data. It is important to note, however, that an alternative explanation might be offered by different modes of Ag presentation that prevail during acute and latent infection. CD8 T cell priming during acute infection is mainly due to cross-presentation of Ag by dendritic cells (36, 37), whereas the maintenance of memory inflation depends on direct presentation of viral Ags on nonhematopoietic cells (38, 39). Hence, it cannot be excluded that the APC type skews epitope recognition toward high-avidity epitopes in early infection and low-avidity epitopes during viral latency.

Notably, only MCMV^M45KNL failed to provide immune protection against a VACV challenge (Fig. 2A, 2B). It is important to note that both the KNL epitope expressed by the strong ie2 promoter and the high-avidity SL epitope expressed within the M45 protein abrogated VACV growth in almost all of the tested animals, but the combination of these adverse features resulted in absolute vaccine failure. This stark dichotomy is reminiscent of the all-or-nothing immune control that was observed upon vaccination by rhesus CMV vectors against SIV (4, 5), and thus might be determined by the same mechanism. The rhesus CMV vaccine vectors did not abrogate SIV replication upon challenge, but nevertheless provided long-term control and eradication of SIV replication in the monkeys (40). Therefore, their effects depended on continuous patrolling of tissues by T cells, as one would expect from TEM populations. One possible explanation for the failure in immune protection that we observed upon immunization with MCMV^M45KNL was that epitope-specific Teff/TEM cells were substantially less abundant than in any other MCMV vector used in this study. This idea would postulate that immune protection by CMV-based vaccines kicks in only once a threshold is achieved, and then becomes very strong very rapidly. Nevertheless, this question can only be conclusively answered by comprehensive future studies, where mice immunized with graded doses of MCMV vectors would be challenged with graded doses of recombinant VACV.

Regardless of the underlying mechanism of immune protection on the host side, our study has identified a novel principle in the design of the CMV-based vaccine vectors that allows low-avidity epitopes to induce protective T cell responses. To understand how protective responses might be generated, we considered the phenomenon of peptide competition (7, 34), which is an extension of the immune sensing hypothesis (41). Immune responses directed against a high-avidity inflationary Ag suppress CD8 T cell responses against weaker endogenous MCMV epitopes (7, 34), and the outcompeted responder cells assume a TCM (34). Therefore, it was conceivable that a low-avidity exogenous epitope could also be outcompeted by stronger endogenous epitopes. Surprisingly, we observed that this occurs only when the low-avidity epitope was expressed by a weaker promoter, but does not happen when it is expressed by the strong ie2 promoter. Hence, our data indicated that peptide competition is influenced by the combination of epitope avidity and its gene expression context. In that case, one would predict that endogenous epitopes may only be outcompeted by exogenous epitopes that are both strongly expressed and possess a high avidity of binding to the TCR. In line with this prediction, we showed that the high-avidity epitope expressed by the early M45 gene did not repress the inflationary responses against endogenous peptides. Conversely, the low-avidity epitope could not repress the immune responses to endogenous peptides, even when it was expressed by the strong promoter (Fig. 5B). Therefore, peptide competition repressed responses against endogenous epitopes only when the high avidity of the exogenous peptide was combined with its strong expression. This finally explains why the exogenous peptide was outcompeted by endogenous ones, only when both its avidity and expression were weak.

Therefore, our results show that Ags chosen in CMV-based vaccines do not necessarily need to be of high avidity to be protective, as long as they are expressed in an inflationary genetic context and induce robust TEM responses. Similarly, the expression of a high-avidity antigenic epitope from a promoter that typically does not induce inflationary responses induced robust and protective immunity. Taken together, our results demonstrate that even low-avidity antigenic epitopes may provide robust inflationary T cell responses and immune protection if they are expressed within CMV genes with robust expression during latency. This insight is of particular interest if one was to use CMV vectors as tumor vaccines, considering that most tumor Ags are weak and do not naturally induce robust immune responses (1, 2). Our work provides a critical step in the deployment of such vaccine vectors, defining the molecular prerequisites for their further successful development.

We thank Ramon Arens for providing KCSRNRQYL-D^b and SSIEFARL-K^b tets, and Inge Hollatz-Rangosch, Ayse Barut, Ilona Bretag, Jennifer Wolf, as well as Janine Schreiber for technical assistance.

The authors have no financial conflicts of interest.