Abstract
CD8+ T cells that recognize virus-derived peptides presented on MHC class I are vital antiviral effectors. Such peptides presented by any given virus vary greatly in immunogenicity, allowing them to be ranked in an immunodominance hierarchy. However, the full range of parameters that determine immunodominance and the underlying mechanisms remain unknown. In this study, we show across a range of vaccinia virus strains, including the current clonal smallpox vaccine, that the ability of a strain to spread systemically correlated with reduced immunodominance. Reduction in immunodominance was observed both in the lymphoid system and at the primary site of infection. Mechanistically, reduced immunodominance was associated with more robust priming and especially priming in the spleen. Finally, we show this is not just a property of vaccine and laboratory strains of virus, because an association between virulence and immunodominance was also observed in isolates from an outbreak of zoonotic vaccinia virus that occurred in Brazil.
Introduction
CD8+ T cells are vital effector cells in antiviral immunity that recognize infected cells displaying antigenic fragments of virus proteins (epitopes) in association with MHC class I through their TCR (1, 2). Virus infection can result in many epitopes being presented to CD8+ T cells, but their immunogenicity varies over orders of magnitude (3). Epitopes that elicit a strong CD8+ T cell response can be considered immunodominant, whereas others that induce a smaller but still detectable response are subdominant. The two main factors that intersect to determine immunodominance are the abundance of a given epitope that is presented and the number of T cells in the preimmune repertoire with a cognate receptor (4, 5). However, these factors do not completely explain dominance hierarchies. Other factors, such as recruitment levels of naive CD8+ T cells and immunodomination, also play roles (6–8). Immunodomination is the ability of T cells responding to dominant epitopes to suppress other responses (9). Understanding these phenomena is important, as they reflect the basic function of the immune system as well as underpin attempts to explain and then perhaps manipulate the breadth of immunity to pathogens via vaccination (2).
Vaccinia virus (VACV) belongs to the orthopoxvirus family and was the live vaccine used to eradicate smallpox. In the absence of smallpox, VACV is being stockpiled in several countries against a potential nefarious release of variola virus, the causative agent of this now historic disease. Furthermore, smallpox vaccination provided protection against a range of orthopoxviruses, some of which continue to emerge from animal reservoirs and cause human disease. The most concerning of these is monkeypox virus, but human cowpox infections are being reported regularly in Europe and even VACV itself is a cause of human infection in South America (10–12). Within a species of orthopoxvirus, wide ranges of virulence have been observed, for example between variola major and minor (or alastrim) and the two main clades of monkeypox virus (13, 14). Likewise, the reactogenicity of smallpox vaccines was known to vary and this has been modeled in mice (15, 16). However, the immunological consequences of differing virulence have not been explored. Furthermore, there remains relatively few descriptions of the full virulence range of VACV strains, which is relevant for policy with regard to vaccination of laboratory workers and also understanding the risk associated with emerging VACV, for example in Brazil.
VACV in mice provides an attractive model for understanding CD8+ T cell responses with well-characterized infections that are acute and an extensive list of mapped epitopes that have been ranked in a predictable immunodominance hierarchy (3, 17, 18). This model has been useful for examining the role of previously primed T cells, regulatory T cells, TCR diversity, and route of infection in immunodominance (8, 19–21). In the present study, we set out to determine whether genetic diversity of the virus is also a determinant of immunodominance using first a set of vaccine and laboratory strains, and then confirming the results using two isolates from a recent outbreak of zoonotic VACV in Brazil.
Materials and Methods
Viruses and cell lines
VACV strains were grown and titrated in BHK-21 and BS-C-1 cells, respectively, both of which were maintained in DMEM (Invitrogen) with glutamine and 10% FBS (D10). VACV strains were Western Reserve (WR, National Institutes of Health tissue culture adapted [NIH-TC], GenBank accession no. AY243312, http://www.ncbi.nlm.nih.gov/nuccore/AY243312) and modified vaccinia Ankara (MVA) (22), B. Moss, National Institutes of Health; ACAM2000 (23), R. Weltzin, Acambis; Copenhagen, Lister, and Tian Tan (24), G.L. Smith, University of Cambridge; VACV WR NP-S-GFP (25), referred to in this study as WR TK−, J. Yewdell and J. Bennink, National Institutes of Health; and Guarani P1 virus (GP1V) and Guarani P2 virus (GP2V) (10), E.G. Kroon, Universidade Federal de Minas Gerais (Belo Horizonte, Brazil).
Mice and infections
Specific pathogen-free female C57BL/6 mice 8–20 wk of age were obtained from the Animal Resource Centre (Perth, WA, Australia) and from the Australian National University Bioscience Research Facility. Within experiments mice were age matched and derived from one source. Mice were housed and experiments were done according to the relevant ethical requirements and under an approval from the Australian National University Animal Ethics and Experimentation Committee (approvals F-BMB-38.8, A2013.037, and A2011.01). For most experiments, mice were infected i.p. with 1 × 106 PFU VACV strains (see Table I) in 200 μl PBS. To determine replication in skin, ear pinnae (intradermal [i.d.]) infections were done with 1 × 103 PFU VACV in 10 μl PBS (26, 27).
Namea . | Sequence . | H-2 . | Changes in Promoter and Protein across VACV Strains Relative to WR . | ||
---|---|---|---|---|---|
Strainb . | Promoterc . | Protein . | |||
B820 | TSYKFESV | Kb | A2K | Nil | K205E |
Cop | Nil | K157E, P160T, V179A, K205E, S213N | |||
Lister | Nil | K242T | |||
MVA | Nil | Del95–102, del115–121, trnc241 | |||
A8189 | ITYRFYLI | Kb | A2K | Nil | K247E |
Cop | Nil | K247E | |||
Lister | Nil | Nil | |||
MVA | Nil | Nil | |||
A3270 | KSYNYMLL | Kb | A2K | C-85T | Q51K |
Cop | Nil | Nil | |||
Lister | C-85T | Nil | |||
MVA | G-44A, C-85T | Q51K | |||
A23297 | IGMFNLTFI | Db | A2K | Nil | P58H |
Cop | A-41G | P58H | |||
Lister | Nil | Nil | |||
MVA | Nil | P58H | |||
K36 | YSLPNAGDVI | Db | A2K | Nil | K22N |
Cop | Nil | F36S | |||
Lister | Ins-41GACAATAA | K22N, F36S | |||
MVA | Nil | K22N, F36S | |||
A47138 | AAFEFINSL | Kb | A2K | Del-36 | G68E, K71E, M236I, H240Y, E244V, trnc245 |
Cop | Del-36 | K71E, M236I, H240Y, E244V, trnc245 | |||
Lister | Nil | K71E, H240Y | |||
MVA | Del-36 | Del29–33, K71E, del153–161, M236I | |||
L253 | VIYIFTVRL | Kb | A2K | Nil | T5A, del71-72, A74V |
Cop | Nil | T5A | |||
Lister | Nil | T5A | |||
MVA | Nil | T5A | |||
J3289 | SIFRFLNI | Kb | A2K | Nil | S227L, I313V, N317D, S318F |
Cop | Nil | V87M, A229V, S331G | |||
Lister | Nil | S227L | |||
MVA | A-77G, A-73G | Nil | |||
A870 | IHYLFRCV | Kb | See above for A8189 | ||
A4288 | YAPVSPIVI | Db | A2K | InsA-18, A-62G | Nil |
Cop | InsA-18 | Nil | |||
Lister | Nil | Nil | |||
MVA | InsA-18 | Del45–49 | |||
G834 | LMYIFAAL | Kb | A2K | Nil | E94K |
Cop | Nil | Nil | |||
Lister | Nil | Nil | |||
MVA | Nil | A257T | |||
A1947 | VSLDYINTM | Kb | A2K | T-37C | Nil |
Cop | T-37C | Nil | |||
Lister | Nil | Nil | |||
MVA | T-37C | Nil |
Namea . | Sequence . | H-2 . | Changes in Promoter and Protein across VACV Strains Relative to WR . | ||
---|---|---|---|---|---|
Strainb . | Promoterc . | Protein . | |||
B820 | TSYKFESV | Kb | A2K | Nil | K205E |
Cop | Nil | K157E, P160T, V179A, K205E, S213N | |||
Lister | Nil | K242T | |||
MVA | Nil | Del95–102, del115–121, trnc241 | |||
A8189 | ITYRFYLI | Kb | A2K | Nil | K247E |
Cop | Nil | K247E | |||
Lister | Nil | Nil | |||
MVA | Nil | Nil | |||
A3270 | KSYNYMLL | Kb | A2K | C-85T | Q51K |
Cop | Nil | Nil | |||
Lister | C-85T | Nil | |||
MVA | G-44A, C-85T | Q51K | |||
A23297 | IGMFNLTFI | Db | A2K | Nil | P58H |
Cop | A-41G | P58H | |||
Lister | Nil | Nil | |||
MVA | Nil | P58H | |||
K36 | YSLPNAGDVI | Db | A2K | Nil | K22N |
Cop | Nil | F36S | |||
Lister | Ins-41GACAATAA | K22N, F36S | |||
MVA | Nil | K22N, F36S | |||
A47138 | AAFEFINSL | Kb | A2K | Del-36 | G68E, K71E, M236I, H240Y, E244V, trnc245 |
Cop | Del-36 | K71E, M236I, H240Y, E244V, trnc245 | |||
Lister | Nil | K71E, H240Y | |||
MVA | Del-36 | Del29–33, K71E, del153–161, M236I | |||
L253 | VIYIFTVRL | Kb | A2K | Nil | T5A, del71-72, A74V |
Cop | Nil | T5A | |||
Lister | Nil | T5A | |||
MVA | Nil | T5A | |||
J3289 | SIFRFLNI | Kb | A2K | Nil | S227L, I313V, N317D, S318F |
Cop | Nil | V87M, A229V, S331G | |||
Lister | Nil | S227L | |||
MVA | A-77G, A-73G | Nil | |||
A870 | IHYLFRCV | Kb | See above for A8189 | ||
A4288 | YAPVSPIVI | Db | A2K | InsA-18, A-62G | Nil |
Cop | InsA-18 | Nil | |||
Lister | Nil | Nil | |||
MVA | InsA-18 | Del45–49 | |||
G834 | LMYIFAAL | Kb | A2K | Nil | E94K |
Cop | Nil | Nil | |||
Lister | Nil | Nil | |||
MVA | Nil | A257T | |||
A1947 | VSLDYINTM | Kb | A2K | T-37C | Nil |
Cop | T-37C | Nil | |||
Lister | Nil | Nil | |||
MVA | T-37C | Nil |
Copenhagen nomenclature; subscript denotes position of the first amino acid in the epitope.
A2K, ACAM2000; Cop, Copenhagen.
Changes in the 85 bp upstream of the ATG of the ORF (bp denoted numbering down from A).
Stimulations and intracellular staining of IFN-γ
Mice were euthanized 7 d postinfection and peritoneal wash, spleens, and mediastinal lymph nodes were taken for analysis of CD8+ T cell responses by intracellular cytokine staining (ICS) as described (28, 29). Briefly, cells were plated at 1–2 × 106 cells/well in D10 into round-bottom 96-well plates. Synthetic peptides were added to a final concentration of 10−7 M and plates were incubated at 37°C and 5% CO2. After 1 h, 5 μg/ml brefeldin A (Sigma-Aldrich) was added and plates were incubated for another 3 h. Plates were spun at 4°C, medium was removed, and cells were resuspended in 50 μl 1:150 diluted anti–CD8-PE (clone 53-6.7; BioLegend). After 30 min incubation on ice, cells were washed, resuspended in 50 μl 1% paraformaldehyde, and incubated at room temperature for 20 min before another two washes and staining with 50 μl 1:200 diluted anti–IFN-γ-allophycocyanin (clone XMG1.2; BioLegend) overnight in PBS with 2% FBS and 0.5% saponin (Sigma-Aldrich) at 4°C. Cells were washed three times before acquisition using a FACS LSR II (BD Biosciences). Analysis was done using FlowJo software (Tree Star). Events were gated for live lymphocytes on forward scatter by side scatter followed by CD8+ T cells using CD8 by side scatter and displayed as CD8 by IFN-γ. Data were recorded as IFN-γ+, CD8+ cells as a percentage of total CD8+ cells. Backgrounds as determined using irrelevant peptides were usually on the order of 0.1% and were subtracted from the values presented for test samples.
Measurement of infectious virus in mouse tissues
Organs were removed after infection, including ovaries, spleen, kidneys, liver, lungs, heart, brain, and mediastinal lymph nodes for i.p. infected mice and ear pinnea after i.d. infection. Organs were briefly rinsed with 80% ethanol and then ground in 1 ml DMEM with 2% FBS in small tissue glass grinders (no. 358103; Wheaton, Milville, NJ) and then subjected to three cycles of freezing and thawing. Samples were sonicated before virus was titrated by plaque assay on BS-C-1 cells (27).
In vivo cytotoxicity assay
As published (29), splenocytes from uninfected C57BL/6 mice were labeled with 5 μM Vybrant DiD cell labeling solution (Molecular Probes) in DMEM for 1 h at 37°C, washed, and split into two populations. One population was pulsed with 10−7 M B820 peptide for 1 h at 37°C and labeled with a high concentration (5 μM) of CFSE for 8 min at 37°C (CFSEhigh cells). The second population was left without peptide and was labeled with a low concentration (0.5 μM) of CFSE (CFSElow cells). Cells were mixed in equal proportions and a total of 4 × 107 cells were injected i.v. into infected (test) or uninfected (control) C57BL/6 mice. The test mice were infected 2 or 4 d prior to the injection of CFSE-labeled targets. In the latter case, mice were injected daily with 1 μg/g 2-amino-[2-(4-octylphenyl])-1,3-propanediol hydrochloride (FTY720; Sigma-Aldrich) starting from the day before infection (30, 31). Mice were euthanized 4 h after injection of targets, and spleens and lymph nodes were examined for the relative proportions of CFSEhigh and CFSElow populations amoungst DiD+ events by flow cytometry. To calculate specific lysis, the following formula was used: ratio = percentage CFSElow/percentage CFSEhigh. The percentage of target cell killing was: [1 – (ratio uninfected [control] recipients/ratio infected [test] recipients)] × 100.
Statistical analyses
Statistical comparisons were done using an unpaired t test with a Welch correction for unequal variance, or ANOVAs with Tukey pairwise posttests (GraphPad Prism, GraphPad Software, La Jolla, CA).
Results
VACV strains fit into two distinct patterns of immunodominance
To determine how virus strain affected CD8+ T cell responses, we tested a wide range of VACV strains, including WR, the most common laboratory strain; ACAM2000, the current United States smallpox vaccine, which is a clone derived from Dryvax (23, 32, 33); Lister, the most widely used smallpox vaccine; Copenhagen and Tian Tan, which were used in Europe and China (34); and MVA, a replication-deficient VACV strain used as a vaccine vector in many clinical trials (35–39). For each virus, mice were infected with 1 × 106 PFU by i.p. injection and splenic CD8+ T cell responses to a panel of 12 peptides were measured by a standard assay that detects IFN-γ–secreting cells after a brief in vitro stimulation with peptides in the presence of brefeldin A (referred to as IFN-γ–ICS). The peptides used correspond to epitopes conserved across all strains (Table I), and, as shown recently, this assay allows highly accurate enumeration of epitope-specific CD8+ T cells in this model (28). For all strains, the overall hierarchy of the 12 epitopes was similar, with B820 being the immunodominant epitope (IDE) and the other 11 peptides being subdominant epitopes (SDE) (Fig. 1A). However, the total size of the responses (when summed across all epitopes) varied. WR, Tian Tan, and Copenhagen elicited responses of >4 × 106 CD8+ T cells per spleen, whereas ACAM2000, Lister, and MVA induced around half that number. Furthermore, the difference in responses across the strains was due to significantly lower responses to the SDE in the cases of ACAM2000, Lister, and MVA compared with WR and Tian Tan and for ACAM2000 and MVA compared with Copenhagen (p < 0.05, ANOVA with Tukey multiple comparisons test; Fig. 1B). We have recently shown that infection of mice with WR by different routes alters the ratio of IDE/SDE responses (8), and this can be most easily seen when showing B820-specific responses as a fraction of the sum of responses to all epitopes. Shown this way, the six VACV strains fall into two groups with respect to the ratio of IDE/SDE responses (Fig. 1C): the IDE either accounts for ∼42% (WR, Tian Tan, Copenhagen) or 61% (ACAM2000, Lister, MVA) of the total measured responses to VACV. This was supported statistically by pairwise comparisons. For all pairs across the groups (e.g., WR and ACAM2000), the difference in the IDE/SDE ratio was significant (p < 0.001), but within each group of viruses (e.g., WR and Copenhagen or ACAM2000 and Lister), no pair had a significant difference. When data are viewed for the IDE and sum of SDEs as a percentage of CD8+ T cells, the same pattern in immunodominance is seen, but the differences in the total size of responses are less apparent (Fig. 1D). In summary, significant variations in both the size of the total CD8+ T cell response and the extent to which these are dominated by B820 are seen across these strains of VACV, but interestingly the level of domination by B820 appeared to be bimodal, rather than being graded across strains.
CD8+ T cell responses to VACV epitopes after immunization of mice with different strains of VACV. Groups of C57BL/6 mice were infected i.p. with 106 PFU VACV strains WR, Copenhagen (Cop), Tian Tan (TT), Lister (Lis), MVA, or ACAM2000 (A2K). Seven days later, numbers of splenic CD8+ T cells that produce IFN-γ in ex vivo stimulations with the indicated individual peptides were measured by ICS. (A) Results for each of the peptides and strains, showing the total numbers of epitope-specific CD8+, IFN-γ+ cells. (B) The same as in (A), but simplified by showing only the response to the IDE (B820) and the sum of responses to the SDE based on numbers of CD8+, IFN-γ+ cells. Virus strain was found to be a statistically significant source of variation in this experiment (p = 0.002; ANOVA), and pairwise significant differences were found for SDE responses between ACAM2000, Lister, and MVA compared with WR and Tian Tan and for ACAM2000 and MVA compared with Copenhagen (p < 0.05, Tukey multiple comparisons test). (C) Relative proportion of the total measured CD8+ T cell responses that are accounted for by the IDE and sum of SDE responses, based on (A) and (B). Comparisons between WR, TT, and Cop or between ACAM2000, Lis, and MVA are not significant, but all comparisons across these sets of three viruses (e.g., WR and ACAM2000) are significant (p < 0.001). (D) Response to the IDE and the sum of responses to the SDE based on the percentage of CD8+ cells that are IFN-γ+. Data shown are means and SEM of groups of 6–12 mice.
CD8+ T cell responses to VACV epitopes after immunization of mice with different strains of VACV. Groups of C57BL/6 mice were infected i.p. with 106 PFU VACV strains WR, Copenhagen (Cop), Tian Tan (TT), Lister (Lis), MVA, or ACAM2000 (A2K). Seven days later, numbers of splenic CD8+ T cells that produce IFN-γ in ex vivo stimulations with the indicated individual peptides were measured by ICS. (A) Results for each of the peptides and strains, showing the total numbers of epitope-specific CD8+, IFN-γ+ cells. (B) The same as in (A), but simplified by showing only the response to the IDE (B820) and the sum of responses to the SDE based on numbers of CD8+, IFN-γ+ cells. Virus strain was found to be a statistically significant source of variation in this experiment (p = 0.002; ANOVA), and pairwise significant differences were found for SDE responses between ACAM2000, Lister, and MVA compared with WR and Tian Tan and for ACAM2000 and MVA compared with Copenhagen (p < 0.05, Tukey multiple comparisons test). (C) Relative proportion of the total measured CD8+ T cell responses that are accounted for by the IDE and sum of SDE responses, based on (A) and (B). Comparisons between WR, TT, and Cop or between ACAM2000, Lis, and MVA are not significant, but all comparisons across these sets of three viruses (e.g., WR and ACAM2000) are significant (p < 0.001). (D) Response to the IDE and the sum of responses to the SDE based on the percentage of CD8+ cells that are IFN-γ+. Data shown are means and SEM of groups of 6–12 mice.
Sequence differences in Ags and promoters do not correlate with immunodominance
The epitopes themselves were all conserved across the VACV strains, but it was possible that changes in promoter sequences or the wider protein altered expression or processing of epitopes, respectively. If a pattern of change was consistent for the strains grouped by immunodominance, this would offer an easy explanation for the CD8+ T cell data above. High-quality full-genome sequences were available for WR, Copenhagen, ACAM2000, Lister, and MVA, so assuming that our stocks of these strains were similar, we looked for changes in the predicted amino acid sequences of Ags and in the 85 bp immediately upstream of each of their open reading frames. These are shown in Table I using the WR sequence as the reference; whereas at least one change was found in at least one region for all 12 Ags across the strains, the pattern of these never corresponded with the immunodominance data in Fig. 1.
Virus replication across VACV strains
We noticed that the VACV strains where immunodomination was more marked were those found previously to be less virulent in an i.d. model of infection in mice (16). A feature of this i.d. model is very restricted spread of infection (8, 26, 40), so measurement of virus loads in ears provides a simple assay for virus growth in vivo. We used this model to determine whether virus growth across the strains correlated with the immunodominance observations. Virus loads were measured in the ears of mice 5 d postinfection with 1 × 103 PFU VACV strains (Fig. 2). Levels of infectious WR, Copenhagen, and Tian Tan were significantly higher than Lister and ACAM2000. However, whereas the first three strains had very similar levels of virus, Lister and ACAM2000 appeared to differ from each other. Furthermore, MVA (not tested here) does not replicate in mice at all (41). If a dose effect were in play, ACAM2000 and MVA would be expected to show even sharper dominance than Lister, but instead we found two discrete IDE dominance patterns. This suggested that simple differences in levels of replication might not fully explain the immunodominance phenomenon.
Virus growth in ear pinnae after i.d. infection with different VACV strains. Groups of C57BL/6 mice were i.d. infected into the ear pinnae with 1 × 103 PFU VACV strains as shown. Five days later, ears were ground and virus was titrated. Data shown are individual titers for each mouse, and the line denotes the mean for each group.
Virus growth in ear pinnae after i.d. infection with different VACV strains. Groups of C57BL/6 mice were i.d. infected into the ear pinnae with 1 × 103 PFU VACV strains as shown. Five days later, ears were ground and virus was titrated. Data shown are individual titers for each mouse, and the line denotes the mean for each group.
Immunodominance correlates with virus spread
Our immunodominance profiles across strains were obtained from mice infected by the i.p. route, which unlike i.d. injection allows VACV spread. Previously we have shown that the lack of spread after inoculation by i.d. injection and other peripheral sites is associated with increased immunodominance (8). Therefore, we wondered whether differences in dissemination might help explain the two immunodominance profiles across VACV strains. Mice were infected i.p. with the VACV strains, and virus loads were measured in spleen, mediastinal lymph node, heart, serum, brain, lungs, kidneys, liver, and ovaries at days 2, 5, and 8 postinfection (Fig. 3). WR showed surprisingly broad spread with very substantial levels of virus across all organs except serum and brain. Tian Tan and Copenhagen were also able to spread to multiple sites, but levels of virus were very much lower than for WR, and detection at most sites was sporadic across the groups. In contrast, Lister and ACAM2000 did not demonstrate convincing spread. These experiments revealed very large differences in virus loads and spread across the strains. Furthermore, they demonstrate that the ability of VACV to spread to multiple organs after i.p. injection correlates well with immunodominance profile.
Virus growth and spread in organs of mice infected with different VACV strains. Groups of C57BL/6 mice were i.p. infected with 106 PFU VACV strains as shown on each chart. At days 2, 5, and 8 postinfection, virus was titrated from spleen, mediastinal lymph nodes (LN), heart, serum, brain, lungs, kidneys, liver, and ovaries. Data shown are individual titers for each mouse per organ, and the line denotes the mean for each group; the dashed line shows the limit of detection (10 PFU/organ).
Virus growth and spread in organs of mice infected with different VACV strains. Groups of C57BL/6 mice were i.p. infected with 106 PFU VACV strains as shown on each chart. At days 2, 5, and 8 postinfection, virus was titrated from spleen, mediastinal lymph nodes (LN), heart, serum, brain, lungs, kidneys, liver, and ovaries. Data shown are individual titers for each mouse per organ, and the line denotes the mean for each group; the dashed line shows the limit of detection (10 PFU/organ).
To draw out distinctions between ability to replicate and virus spread further, we took advantage of a recombinant WR virus that is less able to replicate in vivo owing to loss of thymidine kinase (TK) function, but is otherwise identical to WR. Inactivation of TK increases the lethal dose of WR virus by several orders of magnitude after intracranial infection of mice and limits virus loads after high-dose i.p. infection (42). Using the same experimental methods as above, we compared CD8+ T cell immunodominance (Fig. 4A–C). This showed that WR TK− had a similar immunodominance profile as the parental WR virus, with the IDE accounting for ∼40% of the total measure CD8+ T cell response. Next we compared WR TK− growth in ear pinnae with that of Copenhagen (reduced immunodominance) and Lister (increased immunodominance), but in this experiment there was no significant difference between these three strains, again confirming that growth at a site that restricts further spread does not predict immunodominance (Fig. 4D). After i.p. infection, WR TK− was able to disseminate to multiple organs when examined at multiple times, but this spread was sporadic (Fig. 4E). To be certain that WR TK− could spread as well as another strain that had demonstrated reduced immunodominance, the experiment was repeated, but this time Copenhagen and Lister were included (Fig. 4F). This confirmed that strains Copenhagen and WR TK−, but not Lister, spread to multiple organs. Taken together, these data suggest that it is virus spread, rather than simple replication differences, that determine CD8+ T cell immunodominance patterns. Finally, these data place the attenuation of TK− WR viruses into the context of other VACV strains and demonstrate that even without TK, this virus remains more virulent than traditional vaccine strains such as Lister, when judged by ability to spread. We also noted that similar to Lister and not TK− WR, ACAM2000, the current clonal smallpox vaccine used in the United States, was never seen to spread (Fig. 3).
Immunodominance profile is more closely linked to virus spread than replication alone. (A–C) Groups of C57BL/6 mice were infected i.p. with 1 × 106 PFU WR or WR TK− (TK−) and the number and percentages of splenic CD8+ T cells that produce IFN-γ in ex vivo stimulations with the indicated peptides were measured by ICS. (A) Results for each of the peptides, showing the total numbers of epitope-specific CD8+, IFN-γ+ cells. (B) Relative proportions of the total measured CD8+ T cell responses that are accounted for by the IDE and sum of SDE responses, based on (A). (C) The response to the IDE and the sum of responses to the SDE shown as the percentage of CD8+ cells that are IFN-γ+. (D) Virus titers in ear pinnae 5 d after i.d. infection with 1 × 103 PFU VACV strains Copenhagen (Cop), WR TK−, or Lister. (E) Mice were injected with 1 × 106 PFU WR TK− i.p. and amounts of virus in the organs shown were measured 2, 5, and 8 d later. (F) Mice were injected with 1 × 106 PFU WR Copenhagen, WR TK−, or Lister i.p. and amounts of virus in the organs shown were measured 2 d later. In (A)–(C), data are means and SEM of three and eight mice for WR and WR TK−, respectively, over two experiments; in (D)–(F) results for individual mice are shown with lines denoting means and the dashed line showing the limit of detection (10 PFU/organ).
Immunodominance profile is more closely linked to virus spread than replication alone. (A–C) Groups of C57BL/6 mice were infected i.p. with 1 × 106 PFU WR or WR TK− (TK−) and the number and percentages of splenic CD8+ T cells that produce IFN-γ in ex vivo stimulations with the indicated peptides were measured by ICS. (A) Results for each of the peptides, showing the total numbers of epitope-specific CD8+, IFN-γ+ cells. (B) Relative proportions of the total measured CD8+ T cell responses that are accounted for by the IDE and sum of SDE responses, based on (A). (C) The response to the IDE and the sum of responses to the SDE shown as the percentage of CD8+ cells that are IFN-γ+. (D) Virus titers in ear pinnae 5 d after i.d. infection with 1 × 103 PFU VACV strains Copenhagen (Cop), WR TK−, or Lister. (E) Mice were injected with 1 × 106 PFU WR TK− i.p. and amounts of virus in the organs shown were measured 2, 5, and 8 d later. (F) Mice were injected with 1 × 106 PFU WR Copenhagen, WR TK−, or Lister i.p. and amounts of virus in the organs shown were measured 2 d later. In (A)–(C), data are means and SEM of three and eight mice for WR and WR TK−, respectively, over two experiments; in (D)–(F) results for individual mice are shown with lines denoting means and the dashed line showing the limit of detection (10 PFU/organ).
Virus spread correlates with priming sites of VACV-specific CD8+ T cells
A potential link between virus spread and CD8+ T cell immunodominance is the number or nature of lymphoid organs where immune responses can be primed. Furthermore, a precedent for this has been shown for VACV WR when administered by different routes (8). For this reason we wanted to examine priming sites for a representative of each of the two immunodominance groups of VACV strains, that is, WR (less dominance) and ACAM200 (more dominance). To do this we used an in vivo cytotoxicity assay 2 d postinfection. This approach has been shown to reflect sites of priming after HSV (43) and VACV (8) infection of mice and is based on the observation that primed CD8+ T cells rapidly acquire cytotoxic function but remain at their original priming site for 3–4 d before entering circulation (see supplemental figure 1 in Ref. 8). Therefore, mice were infected i.p. with 1 × 106 PFU WR or ACAM2000, and B820-specific in vivo cytotoxicity assays were done 2 d later looking at results obtained across a variety of lymphoid organs (Fig. 5A). B820-specific killing was highest in mediastinal lymph nodes, which drain the peritoneal cavity, but it was also well above background for all sites in WR-infected mice. After ACAM2000 infection, cytotoxicity was also seen in multiple sites, but it was significantly less than for WR in spleen, mediastinal, and cervical lymph nodes. Furthermore, the greatest difference between the strains was for the spleen, where the average killing for ACAM2000 was marginal as shown by some mice having more B820-labeled than unlabeled cells remaining (generating a specific killing value of <0). To address concerns that there might be some recirculation of primed CD8+ T cells, we repeated this experiment in mice given FTY720, doing the in vivo killing assay 4 d postinfection to increase the numbers of primed CD8+ T cells that could be detected (Fig. 5B). As expected, the average B820-specific killing was higher in all sites in these mice, but it was always higher for WR-infected compared with ACAM2000-infected mice, and this difference was again statistically significant in the spleen. Taken together, these data suggest that priming occurs across multiple secondary lymphoid organs for both viruses, with the second experiment allowing more time for the primed cells to build up and be detected. However, this priming was clearly more limiting for ACAM2000 than WR, and priming in the spleen was always significantly lower for this strain of VACV.
Early and FTY720-blocked B820-specific cytotoxicity shows different levels of CD8+ T cell priming in spleens and lymph nodes of WR- and ACAM2000-infected mice. (A) Spleen cells from naive mice were pulsed with B820 peptide and labeled with CFSE to give a high level of fluorescence intensity (CFSEhigh). Unpulsed spleen cells were labeled with CFSE to give low fluorescence intensity (CFSElow). A 1:1 mixture of 1 × 107 cells of each cell population was injected i.v. into naive mice or in mice that had been infected i.p. with 1 × 106 PFU VACV-WR or ACAM2000 2 d previously. After 4 h, mice were sacrificed and lymph nodes and spleens were analyzed for CFSEhigh and CFSElow target cells. To quantify in vivo cytotoxicity, the elimination of B820-pulsed CFSEhigh-labeled cells relative to unpulsed CFSElow-labeled cells was monitored. (B) As for (A), but mice were administered FTY720 daily starting 24 h prior to infection and in vivo cytotoxicity measured 4 d postinfection. Results of individual mice and means of groups are shown. *p < 0.05. LN, lymph node.
Early and FTY720-blocked B820-specific cytotoxicity shows different levels of CD8+ T cell priming in spleens and lymph nodes of WR- and ACAM2000-infected mice. (A) Spleen cells from naive mice were pulsed with B820 peptide and labeled with CFSE to give a high level of fluorescence intensity (CFSEhigh). Unpulsed spleen cells were labeled with CFSE to give low fluorescence intensity (CFSElow). A 1:1 mixture of 1 × 107 cells of each cell population was injected i.v. into naive mice or in mice that had been infected i.p. with 1 × 106 PFU VACV-WR or ACAM2000 2 d previously. After 4 h, mice were sacrificed and lymph nodes and spleens were analyzed for CFSEhigh and CFSElow target cells. To quantify in vivo cytotoxicity, the elimination of B820-pulsed CFSEhigh-labeled cells relative to unpulsed CFSElow-labeled cells was monitored. (B) As for (A), but mice were administered FTY720 daily starting 24 h prior to infection and in vivo cytotoxicity measured 4 d postinfection. Results of individual mice and means of groups are shown. *p < 0.05. LN, lymph node.
Immunodominance is sharper at the site of infection than in the spleen
To determine whether the immunodominance differences in the spleen might have an effect on the repertoire of T cells that are able to combat infection, we examined CD8+ T cell responses to our panel of peptides in cells from the peritoneal cavity of i.p. infected mice (Fig. 6). Approximately 10-fold more CD8+ T cells were detected in the peritoneal cavity of WR-infected compared with ACAM2000-infected mice (5 × 107 versus 5 × 106 CD8+ T cells, respectively; naive mice have 1 × 106 CD8+ T cells). Additionally, responses were more heavily skewed toward the IDE in the peritoneal cavity than the spleen for both strains of VACV (compare Figs. 1 and 6). However, there remained a difference between the strains with the IDE accounting for 53 and 73% of the measured response for WR and ACAM2000, respectively (Fig. 6B). When viewed as a percentage of CD8+ T cells, responses to ACAM2000 appear to be stronger than those to WR, emphasizing the importance for taking total numbers into account in these experiments (Fig. 6C). Therefore, all CD8+ T cell responses were more skewed toward the IDE at the site of infection, but differences in immunodominance between VACV strains were maintained.
Splenic CD8+ T cell immunodominance profiles are maintained at the site of infection. Groups of C57BL/6 mice were infected i.p. with 106 PFU WR or ACAM2000 and 7 d later CD8+ T cells responses were determined in peritoneal exudate cells. (A) Results for each of the peptides, showing the total numbers of epitope-specific CD8+, IFN-γ+ cells. (B) Relative proportion of the total measured CD8+ T cell response that are accounted for by the IDE and sum of SDE responses, based on (A). (C) The response to the IDE and the sum of responses to the SDE shown as the percentage of CD8+ cells that are IFN-γ+. Data shown are means and SEM of groups of six to nine mice from three experiments.
Splenic CD8+ T cell immunodominance profiles are maintained at the site of infection. Groups of C57BL/6 mice were infected i.p. with 106 PFU WR or ACAM2000 and 7 d later CD8+ T cells responses were determined in peritoneal exudate cells. (A) Results for each of the peptides, showing the total numbers of epitope-specific CD8+, IFN-γ+ cells. (B) Relative proportion of the total measured CD8+ T cell response that are accounted for by the IDE and sum of SDE responses, based on (A). (C) The response to the IDE and the sum of responses to the SDE shown as the percentage of CD8+ cells that are IFN-γ+. Data shown are means and SEM of groups of six to nine mice from three experiments.
Differences in immunodominance are a feature of VACV isolates from outbreaks
One of the concerns with any studies using laboratory or vaccine strains of VACV is the extent to which their artificial maintenance on animals and in culture over long periods have endowed phenotypes that are not relevant to survival in nature. Isolates of VACV from zoonotic outbreaks in Brazil provide an opportunity to examine strains that have the ability to persist in the environment and cause human disease. In one such outbreak, two strains of VACV were isolated, namely GP1V and GP2V, and further investigation of these found that the former strain had much greater virulence, spread, and replication than did the latter strain in mice (10, 44). We used these isolates to test whether the replication and virulence differences between recent isolates of VACV would determine their CD8+ T cell immunodominance profiles as was seen in the laboratory and vaccine strains. Mice were infected i.p. and the response to the set of 12 peptides was determined (Fig. 7). Results for these strains fell into the same two groups based on size of total response and especially immunodominance, with the IDE accounting for 45 and 63% of the measured response for virulent GP1V and avirulent GP2V, respectively.
Zoonotic isolates of VACV fit into two patterns of immunodominance. Groups of C57BL/6 mice were infected i.p. with 106 PFU GP1V or GP2V and 7 d later CD8+ T cells responses were determined in the spleen. (A) Results for each of the peptides, showing the total numbers of epitope-specific CD8+, IFN-γ+ cells. (B) Relative proportion of the total measured CD8+ T cell response that are accounted for by the IDE and sum of SDE responses, based on (A). (C) The response to the IDE and the sum of responses to the SDE shown as the percentage of CD8+ cells that are IFN-γ+. Data shown are means and SEM of groups of eight mice from two experiments.
Zoonotic isolates of VACV fit into two patterns of immunodominance. Groups of C57BL/6 mice were infected i.p. with 106 PFU GP1V or GP2V and 7 d later CD8+ T cells responses were determined in the spleen. (A) Results for each of the peptides, showing the total numbers of epitope-specific CD8+, IFN-γ+ cells. (B) Relative proportion of the total measured CD8+ T cell response that are accounted for by the IDE and sum of SDE responses, based on (A). (C) The response to the IDE and the sum of responses to the SDE shown as the percentage of CD8+ cells that are IFN-γ+. Data shown are means and SEM of groups of eight mice from two experiments.
Discussion
In this study, we explored the apparent difference in CD8+ T cell responses to epitopes conserved across VACV strains that were noted when the first minimal determinants were identified (3). We confirmed that for MVA, responses to SDE are more compromised than those recognizing the IDE and extend this to a wider set of less virulent strains. This group of viruses that elicits responses with a greater focus on the IDE is notable in that it includes strains considered safer for use in humans as smallpox or recombinant vaccines. We have not tested whether the shift in immunodominance and thus narrowing of specificity for these might compromise protection against subsequent challenge, but this is worth further consideration. On the one hand, poorer responses elicited by the less virulent strains have been investigated previously and shown to correlate with reduced engagement of costimulators such as OX40 and CD27 (45). On the other hand, requirement of OX40 was equal both for IDE and SDE, suggesting that this does not underlie the immunodominance changes we report (46). Also, in considering the impact of changes in immunodominance on protection, it is noteworthy that immunization with single CD8+ T cell epitopes can protect mice against lethal VACV and ectromelia virus challenge (3, 47, 48). We found in the present study that B820-specific (IDE) responses differ less across the strains, and thus if adequately primed, the response to this epitope alone would provide significant protection irrespective of the vaccine. Finally, Ab responses will remain, and we have not addressed any possible differences in the specificity of these here. Alternatively, the comparison between virulent VACV strains causing systemic infection and those that remain limited might be a model of the difference between smallpox vaccination (with VACV) and disease. It has been argued strongly that smallpox survivors had superior immunity to those vaccinated against the disease (49). Although there are many possible reasons for this (including those noted above), it is possible that in the face of smallpox, all aspects of immunity need to be optimal to provide protection, and that includes breadth of cellular responses.Although it was beyond the scope of this investigation, it would be of interest to know whether breadth of humoral immunity was also greater in mice that experience systemic, compared with local, VACV infections.
Among the VACV strains we examined, several are used in people or are in clinical trials, and these are all from the less virulent group and responses were characterized by increased immunodominance. This could be of potential importance for recombinant vaccines where immunodominance of responses to the vector are thought to be a problem (50). Of greatest direct importance, it impacts the design of preclinical work and suggests clearly that all such work should be done with a strain suitable for using in humans, rather than a virulent laboratory strain such as WR. This is needed to avoid complications that might be caused by changes in immunodominance rather than overall immunogenicity of vectors or immunization strategies (8, 51).
Our data examining priming sites should be interpreted with a recent publication examining immunodominance changes after infection with WR by different routes of infection (8). Taken together, these suggest that differences in systemic spread of virus drive the changes in immunodominance. After i.d. immunization compared with i.p. or i.v. immunization, spread of WR was limited, resulting in reduced priming of CD8+ T cells in the spleen and increased domination by the IDE. Additionally, greatly reducing the dose of WR given by the i.p. route such that priming was restricted to the mediastinal lymph node similarly increased immunodomination (8). The large difference in ability of the various strains of VACV to spread was reflected in the level of priming seen in different lymphoid organs and especially the spleen. Therefore, a similar mechanism probably underpins both observations. In the case of the different routes we found that competition for costimulation was a driver for increased immunodomination where priming was confined to local draining lymph nodes (8). This is likely to be occurring in this case where less virulent viruses are unable to spread beyond a local draining lymph node.
Studies with VACV are sometimes regarded as irrelevant, owing to the unknown origin and incomplete passage history of the laboratory strains. For this reason we included viruses that were derived from a VACV zoonotic outbreak in Brazil (10). The two VACVs have previously been reported to have higher (GP1V) and lower (GP2V) virulence. We found that this virulence difference influenced the size of immune responses and the immunodominance profile in the same way as it did across the laboratory and vaccine strains. This suggests that our findings are relevant to viruses that are maintained in an ecological niche, which we note for similar isolates of VACV from Brazil has been shown to include rodents (52). Although we did not test the virulence of these strains again, our study is the most comprehensive examination of VACV spread and virus loads across laboratory and vaccine strains. This work demonstrated a very striking difference in the ability of VACV strains to replicate and spread from a peritoneal site of infection, with the virulence of strain WR being outstanding. At least by this model, the vaccine strains of VACV are highly attenuated. Taking into account their lack of spread as well as reduced replication in the skin, strains such as ACAM2000 and Lister are more attenuated than a TK− recombinant of WR. This finding is important in that it places the extent of attenuation that is achieved by inactivating the TK gene from WR (often considered to greatly increase safety) in the context of VACV vaccines and supports vaccination with ACAM2000 to protect against laboratory exposure with TK− WR viruses. Indeed, the replication and spread in vivo of TK− WR was similar to that of Copenhagen, which was one of the more reactogenic smallpox vaccine strains. It also demonstrates the very wide range of virulence that can be found across VACV strains even though all replicate equally in culture. Finally, the ability of TK− WR to spread detectably to multiple organs, despite reduced replication in the skin, suggests that replication at the primary site of infection and dissemination may be independently controlled phenotypes.
The reason for the increased spread of a subset of VACV strains (WR, Copenhagen, and Tian Tan) is not clear, but it might be related to a particular arsenal of immune-modulating genes. We have examined a set of 30 known immune modulators of VACV to determine whether the possession of any of these might correlate with ability to spread and reduced immunodominance (Table II). The only candidate that fit this criteria across the fully sequenced VACV strains was the gene encoding the soluble receptor for type I IFN (B18R or WR200 in WR/B19R in Copenhagen) (53). This important immune evasion gene was found in both Brazillian strains used in this study (10), which weakens the correlation, but full sequences have not been reported and there is a P75L substitution in the less virulent GP2V not found in GP1V or other VACV strains, so it remains a molecule of interest. The two caveats to this discussion are that we have not sequenced the particular isolates we used to ensure that none has changes in these genes, but just as importantly, there remain uncharacterized genes in the VACV genome. Finally, many single gene deletion mutants of VACV are attenuated in vivo, and previous studies have found that it is not straightforward to predict the virulence of poxviruses based on knowing the complement of virulence genes in a given strain (16, 54).
Protein/Function . | WR . | Cop . | A2K . | Lister . | MVA . |
---|---|---|---|---|---|
Secreted type I IFN-binding proteina | Y | Y | N | N | N |
Secreted CC chemokine-binding protein (35K) | N | N | N | Y | N |
Secreted type II IFN-binding protein | Y | Y | Y | Y | N |
Secreted IL-1β–binding protein | Y | N | Y | Y | Y |
Secreted IL-18–binding protein | Y | N | Y | Y | Y |
Secreted TNF-binding protein CrmC | N | N | N | Y | N |
Secreted TNF-binding protein CrmE | N | N | N | Y | N |
A41 secreted chemokine-binding protein | Y | Y | Y | Y | Y |
SPI-1 serpin | Y | Y | Y | Y | N |
SPI-2/CrmA serpin | Y | N | N | N | N |
SPI-3 serpin | Y | Y | Y | Y | Y |
A39 semaphorin | N | Y | N | Y | N |
A44 3-β-hydroxysteroid dehydrogenase | Y | Y | Y | Y | Y |
A52 intracellular TLR and IL-1 signal inhibitor | Y | Y | Y | Y | N |
A46 Toll–L1 receptor-like protein | Y | N | Y | Y | Y |
B14 NF-κB inhibitor | Y | Y | Y | Y | N |
A49 NF-κB inhibitor | Y | Y | Y | Y | Y |
K7 immune modulator | Y | Y | Y | Y | Y |
F1 antiapoptotic | Y | Y | Y | Y | Yb |
M2 NF-κB inhibitor | Y | Y | Y | Y | N |
A55 Kelch BTB | Y | Y | Y | Y | N |
C2 Kelch BTB | Y | Y | Y | Y | N |
F3 Kelch BTB | Y | Y | Y | Y | Y |
A35 MHC class II inhibitor | Y | Y | Y | Y | Y |
K3 IFN resistance, eIF-2a homolog | Y | Y | Y | Y | Y |
E3 IFN resistance, dsRNA binding proteina | Y | Y | Y | Y | Y |
N1 virulence factor | Y | Y | Y | Y | Yb |
O1 ERK1/2 pathway inhibitor | Y | Y | Y | Y | N |
C2 complement control protein (VCP) | Y | Y | Y | Y | N |
C7 host range gene | Y | Y | Y | Y | Y |
Protein/Function . | WR . | Cop . | A2K . | Lister . | MVA . |
---|---|---|---|---|---|
Secreted type I IFN-binding proteina | Y | Y | N | N | N |
Secreted CC chemokine-binding protein (35K) | N | N | N | Y | N |
Secreted type II IFN-binding protein | Y | Y | Y | Y | N |
Secreted IL-1β–binding protein | Y | N | Y | Y | Y |
Secreted IL-18–binding protein | Y | N | Y | Y | Y |
Secreted TNF-binding protein CrmC | N | N | N | Y | N |
Secreted TNF-binding protein CrmE | N | N | N | Y | N |
A41 secreted chemokine-binding protein | Y | Y | Y | Y | Y |
SPI-1 serpin | Y | Y | Y | Y | N |
SPI-2/CrmA serpin | Y | N | N | N | N |
SPI-3 serpin | Y | Y | Y | Y | Y |
A39 semaphorin | N | Y | N | Y | N |
A44 3-β-hydroxysteroid dehydrogenase | Y | Y | Y | Y | Y |
A52 intracellular TLR and IL-1 signal inhibitor | Y | Y | Y | Y | N |
A46 Toll–L1 receptor-like protein | Y | N | Y | Y | Y |
B14 NF-κB inhibitor | Y | Y | Y | Y | N |
A49 NF-κB inhibitor | Y | Y | Y | Y | Y |
K7 immune modulator | Y | Y | Y | Y | Y |
F1 antiapoptotic | Y | Y | Y | Y | Yb |
M2 NF-κB inhibitor | Y | Y | Y | Y | N |
A55 Kelch BTB | Y | Y | Y | Y | N |
C2 Kelch BTB | Y | Y | Y | Y | N |
F3 Kelch BTB | Y | Y | Y | Y | Y |
A35 MHC class II inhibitor | Y | Y | Y | Y | Y |
K3 IFN resistance, eIF-2a homolog | Y | Y | Y | Y | Y |
E3 IFN resistance, dsRNA binding proteina | Y | Y | Y | Y | Y |
N1 virulence factor | Y | Y | Y | Y | Yb |
O1 ERK1/2 pathway inhibitor | Y | Y | Y | Y | N |
C2 complement control protein (VCP) | Y | Y | Y | Y | N |
C7 host range gene | Y | Y | Y | Y | Y |
Present in the Brazillian VACV strains GP1V and GP2V but with at least one coding change.
Present but with small inframe deletions, function unknown.
N, no; Y, yes.
In conclusion, we found that less virulent viruses elicited far weaker responses that were more focused on the IDE. Sharpened dominance was strictly associated with a lack of systemic spread of virus, rather than just replicative ability, and determined the specificity of responses at the site of infection as well as the spleen. Finally, by finding the same two patterns of CD8+ T cell immunodominance in two isolates from a Brazilian outbreak of epizootic VACV, we show that this phenomenon is not an artifact of viruses maintained for many years under laboratory conditions.
Acknowledgements
We thank Drs. J.W. Yewdell, J. Bennink, and B. Moss (National Institute of Allergy and Infectious Diseases, National Institutes of Health), Prof. G.L. Smith (University of Cambridge), Dr. R. Weltzin (Acambis), and Prof. E.G. Kroon (Universidade Federal de Minas Gerais) for provision of viruses. We also thank Stewart Smith for general laboratory assistance and Research School of Biology animal services for husbandry of mice.
Footnotes
This work was supported by National Institutes of Health Grants R01 AI067401 and U19 AI100627, National Health and Medical Research Council (Australia) Grant APP1023141, and by Australian Research Council Future Fellowship FT110100310.
Abbreviations used in this article:
References
Disclosures
The authors have no financial conflicts of interest.