Human CD8⁺ T-Cell Responses Against the 4 Dengue Virus Serotypes Are Associated With Distinct Patterns of Protein Targets

Abstract

Background. All 4 dengue virus (DENV) serotypes are now simultaneously circulating worldwide and responsible for up to 400 million human infections each year. Previous studies of CD8⁺ T-cell responses in HLA-transgenic mice and human vaccinees demonstrated that the hierarchy of immunodominance among structural versus nonstructural proteins differs as a function of the infecting serotype. This led to the hypothesis that there are intrinsic differences in the serotype-specific reactivity of CD8⁺ T-cell responses.

Methods. We tested this hypothesis by analyzing serotype-specific CD8⁺ T-cell reactivity in naturally infected human donors from Sri Lanka and Nicaragua, using ex vivo interferon γ–specific enzyme-linked immunosorbent spot assays.

Results. Remarkably similar and clear serotype-specific patterns of immunodominance in both cohorts were identified. Pooling of epitopes that accounted for 90% of the interferon γ response in both cohorts resulted in a global epitope pool. Its reactivity was confirmed in naturally infected donors from Brazil, demonstrating its global applicability.

Conclusions. This study provides new insight into differential serotype-specific immunogenicity of DENV proteins. It further provides a potentially valuable tool for future investigations of CD8⁺ T-cell responses in the typically small sample volumes available from patients with acute fever and children without requiring prior knowledge of either infecting DENV serotype or HLA type.

Four serotypes of dengue virus (DENV1–4) are responsible for 390 million infections and up to 100 million cases worldwide, making dengue a public health problem of increasing concern [1]. While primary infection with 1 serotype is thought to provide lifelong protection against the infecting serotype, secondary infection with another serotype can lead to more-severe forms of the disease, including dengue hemorrhagic fever and dengue shock syndrome. It is hypothesized that the higher frequency of severe disease during secondary infections may be caused by weakly neutralizing serotype–cross-reactive antibodies, leading to antibody-dependent enhancement [2], and/or by an aberrant CD8⁺ T-cell response, resulting in a cytokine storm [3]. A third, nonexclusive alternative is that different serotypes and the sequence of infection might be associated with intrinsic differences in pathogenicity.

Our understanding of the role of CD8⁺ T cells in response to infection with the various DENV serotypes remains incomplete and therefore requires further investigation. Previous studies in murine models demonstrated a protective role for CD8⁺ T cells and defined epitopes recognized in various murine strains, including HLA-transgenic mice [4–6]. Effector-memory CD8⁺ T cells are markedly activated in response to DENV viremia in humans [7]. Additional studies investigating human CD8⁺ T-cell responses in the dengue-hyperendemic country of Sri Lanka highlighted and demonstrated the association between the strength of responses restricted by HLA class I alleles and susceptibility to severe disease, suggesting a protective role for CD8⁺ T cells [8].

Previous studies from other groups and our own murine studies, the analysis of human volunteers immunized with dengue live attenuated vaccines, and the study of natural immunity observed in the dengue-endemic location in Sri Lanka all highlighted the general immunodominance of the DENV nonstructural proteins NS3, NS4B, and NS5 [4, 5, 8–11]. However, in the murine studies, it was noted that while DENV2 elicited a response almost exclusively directed to nonstructural proteins (NS1, NS2a, NS2b, NS3, NS4a, NS4b, and NS5), in the case of DENV3, a sizeable fraction of the response was directed to structural proteins (capsid [C], membrane [M], and envelope [E]). Since mice are not a natural host for DENV, the possibility that these results were reflective of a peculiarity of the mouse-adapted DENV strain and/or the interferon α/β receptor–deficient murine model could not be excluded.

The concern related to the murine models can be addressed by considering a recent study where the serotype-specific CD8⁺ T-cell epitopes elicited by vaccination with monovalent dengue live attenuated vaccine were investigated [9]. About 40% of all responses in the case of DENV3 were directed against structural proteins. However, in the case of DENV1, DENV2 and DENV4 responses against structural proteins comprised only 10% to 30% of the total [9].

Here, we sought to test whether these differences would hold in the case of human populations naturally infected with DENV. This was performed by first analyzing the data from the previous Sri Lanka study for differential patterns of serotype immunogenicity. Next, we generated a new data set to study the patterns of responses observed in the Nicaragua region, as an independent cohort.

DENV has been endemic in Nicaragua since 1985, and all 4 DENV serotypes are found in circulation, generally with 1 serotype dominating each season [12, 13]. In Managua, the capital of Nicaragua, by the age of 14 years, ≥90% of children have been exposed to DENV, suggesting that the vast majority of adults have experienced multiple DENV infections [12, 14, 15]. Because of the high incidence of heterotypic infections with any of the 4 serotypes in Nicaragua, we expected to be able to examine whether differences existed in terms of the immunodominance pattern in T-cell responses directed against each DENV serotype.

We used a previously described epitope identification approach based on testing predicted epitopes for a panel of the most common HLA-A and HLA-B alleles [8]. HLA-matched peptides were used to screen peripheral blood mononuclear cells (PBMCs) from individual blood bank donors who had been repeatedly infected with DENV. This approach allowed us to define responses for the vast majority of alleles expressed in the general Nicaraguan population and, thus, to characterize immunodominance at the level of specific HLA alleles. Determining the DENV-specific CD8⁺ T-cell response in multiple areas of endemicity further enabled creation and validation of a global epitope set, which will be useful for investigating responses in small sample volumes typically available from patients with acute fever and children.

MATERIALS AND METHODS

Ethics Statement

Samples were obtained from healthy adult blood donors from the Nicaragua National Blood Center and Ministry of Health, Colombo, Sri Lanka, in an anonymous fashion. Thus, it was not necessary to obtain written consent. According to local standards the University of Colombo, Ethics Review Committee reviewed and approved the project. Samples obtained were discarded buffy coats from routine blood donations and thus exempt from human subject review. Brazilian PBMC samples were also processed using a similar protocol, as previously described [7], and all participants signed the institutional review board (IRB)–approved informed consent document (IRB document 0652/09, CAPPesq, Hospital das Clínicas, Universidade de São Paulo). San Diego donors gave informed consent according to the guidelines of the IRB of the La Jolla Institute for Allergy and Immunology.

Human Blood Samples

A total of 130 samples were obtained from the National Blood Center in Managua. Donors were of both sexes, from the general population between 17 and 65 years old. Plasma samples were tested for total anti-DENV antibodies by inhibition enzyme-linked immunosorbent assay (ELISA), as previously described [16, 17]. Plasma specimens from a subset of 14 randomly selected Nicaraguan donors were tested using a serotype-specific reporter viral particle–based neutralization assay [15, 18]. All samples had a broad neutralization profile, suggesting that the donors had experienced at least 2 DENV infections prior to blood donation. Blood samples were collected from Sri Lankan participants in an identical fashion, as previously described [8].

A cohort of dengue cases was established in São Paulo, Brazil, by enrolling patients with confirmed DENV infection. Longitudinal samples were collected during the 2014 outbreak, with acute and convalescent specimens stored in the sample repository. The infecting DENV serotype was determined in all patients. Samples obtained at least 1 month after the onset of symptoms of acute dengue were made available for the present study.

Seronegative donors were recruited in San Diego, and seronegativity was confirmed by dengue virus immunoglobulin G–specific ELISA [19].

PBMC Isolation

PBMCs were purified by density gradient centrifugation (Ficoll-Paque Premium, GE Healthcare Biosciences, Kowloon, Hong Kong), resuspended in fetal bovine serum (Gemini Bio-products, Sacramento, California; Gibco Life Technologies) containing 10% dimethyl sulfoxide, and cryopreserved in liquid nitrogen [8, 20]. Brazilian PBMC samples were also processed using a similar protocol, as previously described [7].

HLA Typing

Genomic DNA for HLA typing was isolated from donor PBMCs, using QIAmp DNA isolation kits (Qiagen, Valencia, California). High-resolution Luminex-based typing of HLA class I genes was conducted as per the manufacturer's instructions (sequence-specific oligonucleotide typing; One Lambda, Canoga Park, California). Alternatively, high-resolution typing was performed by polymerase chain reaction–based methods (sequence-specific primer typing; One Lambda) as described elsewhere [8].

Peptide Prediction and Synthesis

DENV sequences were selected as previously described [8]. Additionally, we have added recently reported DENV sequences added to GenBank after 2009. Sequences available from DENV strains circulating in Nicaragua in particular, and in Central America in general, were also added. Every isolate with 1 amino acid difference was termed “unique,” and the number of unique isolates used was limited to 10 isolates per serotype per geographical location. Sequences of isolates were broken down into 9-mers and 10-mers and assessed for binding predictions.

Major histocompatibility complex (MHC) class I binding predictions and peptide selections were conducted using the Immune Epitope Database (IEDB) analysis resource (available at: http://tools.immuneepitope.org/) as previously described [8, 21]. The 9-mer and 10-mer peptides selected for synthesis were within the top 1% of predicted binders and were conserved in at least 30% of the strains within a serotype. Peptides from ≥2 serotypes that were predicted to bind to the same allele were classified as “conserved.” A total of 8894 peptides were synthesized and screened as described below. The peptides synthesized were predicted to bind at least 1 of 27 common HLA class I MHC molecules (Supplementary Table 1), which collectively account for approximately 97% of the HLA-A and HLA-B allelic variation [22].

Ex Vivo Interferon γ (IFN-γ)–Specific Enzyme-Linked Immunospot (ELISPOT) Assay

PBMCs were tested with sets of peptides predicted to bind the exact HLA alleles expressed in the particular donor. If an exact peptide set match at the 4-digit allele level was not available for a particular allele, the HLA allele was matched to the 2-digit allele level or (rarely) to the closest HLA allele within the same HLA supertype [23]. PBMCs (2 × 10⁵ cells/well) were incubated with 2 μg/mL of HLA-matched peptides in plates coated with the anti-human IFN-γ antibody 1-D1K (Mabtech, 5 μg/mL) at 37°C for 20–24 hours and developed as previously described [8]. PBMCs were first screened with peptide pools containing 10 peptides per pool and subsequently deconvoluted to identify the individual epitopes.

Flow Cytometry

In this study, we used anti-CD3 Alexa Flour 700 (UCHT1), anti-IFN-γ fluorescein isothiocyanate (4S.B3), anti-CD19 V500 (HIB19), and anti-CD14 V500 (M5E2), from BD Biosciences (San Diego, California); anti-CD8a BV650 (RPA-T8; Biolegend, San Diego, California); and fixable viability dye ef506 (ebioscience, San Diego, California). For intracellular cytokine staining, PBMCs were cultured in the presence of HLA-matched peptide pools (10 μg/mL) and GolgiPlug containing brefeldin A (BD Biosciences, San Diego, California) for 6 hours and subsequently permeabilized, stained, and analyzed as previously described [9]. Flow cytometry experiments in Brazilian samples were performed identically in the Laboratório de Investigação Médica 60, School of Medicine, University of São Paulo.

RESULTS

Analysis of Differences in Serotype Specificity in the Dengue-Hyperendemic Region of Colombo

A previous study analyzed the pattern of responses observed in the dengue-hyperendemic region of Colombo [8]. PBMCs were obtained from discarded buffy coats from normal blood donors from the Colombo Blood Bank. Each PBMC donation was HLA typed and was tested with predicted binding peptides in IFN-γ–specific ELISPOT assays. The aggregate data were previously reported [8] but have not separately been examined for the presence of serotype-specific patterns of recognition.

The analysis showed that responses directed against epitopes conserved in multiple serotypes accounted for about 26% of the total response (Figure 1A). Serotype-specific responses directed against epitopes derived from DENV2 accounted for 39% of the total, while responses directed to epitopes specific for DENV1, DENV3, and DENV4 accounted for 7%, 14%, and 13%, respectively. More than 90% of the epitopes conserved across serotypes were derived from nonstructural proteins (Figure 1B).

Figure 1.

Serotype distribution and conserved responses in Sri Lanka. A, Distribution of CD8⁺ T-cell epitope locations across the 4 dengue virus (DENV) serotypes and conserved regions in Sri Lankan donors (n = 181). B, Conserved CD8⁺ T-cell interferon γ (IFN-γ) responses across the DENV proteome. IFN-γ responses are displayed as spot-forming cells (SFCs) per 10⁶ peripheral blood mononuclear cells. The percentages in panel B represent the proportion of responses targeting epitopes in structural and nonstructural proteins.

Further, we found that serotype-specific epitopes were mostly derived from nonstructural proteins, accounting for 75%, 81%, and 97% of the total for DENV1, DENV2, and DENV4, respectively (Figure 2A, 2B, and 2D). In contrast, only 48% of the DENV3-specific responses were directed against nonstructural proteins (Figure 2C).

Figure 2.

Serotype-specific CD8⁺ T-cell responses in Sri Lanka. Dengue virus 1 (DENV1)–specific (A), DENV2-specific (B), DENV3-specific (C), and DENV4-specific (D) CD8⁺ T-cell interferon γ (IFN-γ) responses across the DENV proteome. IFN-γ responses are displayed as spot-forming cells (SFCs) per 10⁶ peripheral blood mononuclear cells. The percentages in all panels represent the proportion of responses targeting epitopes in structural and nonstructural proteins.

Accordingly, we concluded that, in the context of natural infection in humans, the DENV1, DENV2, and DENV4 serotypes were associated with CD8⁺ T-cell responses predominantly focused on nonstructural proteins, while nearly half of the DENV3-specific response was directed against structural proteins.

Breadth and Magnitude of DENV-Specific Responses in the Dengue-Hyperendemic Region of Managua

Next, we examined whether these observations held true in an independent donor population in a disparate geographical location. To define the repertoire and magnitude of the DENV-specific CD8⁺ T-cell responses in healthy blood donors from the dengue-endemic area of Managua, we followed the approach previously used to define CD8⁺ T-cell responses in Sri Lanka.

Ex vivo IFN-γ responses were measured directly for 130 donors, using ELISPOT assays. Reactivity was detected ex vivo for >30% of the donors. Responses were broad, with an average of 7 epitopes recognized per donor and were associated with an average magnitude per donor of 212 spot-forming cells (SFCs)/10⁶ PBMCs (Table 1). A total of 520 different responses were detected and mapped back to 314 different individual epitopes (available at: http://www.iedb.org/submission/1000582). When these peptides were clustered according to 80% homology (using the IEDB analysis resource) to identify nested and/or homologous regions, 232 total distinct antigenic regions were identified. The frequency and magnitude of these responses were comparable to what has been previously observed in secondary DENV infections in the Sri Lanka dengue-endemic area (ie, 43% responding donors, 11 epitopes/donor, and 220 SFCs/10⁶ PBMCs).

Table 1.

Dengue Virus–Specific CD8⁺ T-Cell Responses in Nicaragua

Parameter	Value
Frequency of donors responding, %	34
Breadth of response/donor, no. of epitopes	7
Magnitude of response/epitope, SFCs/106 PBMCs	212
Donors tested, no.	130
Responses detected, no.	520
Epitopes identified, no.	314
Antigenic regions identified, no.	232

Abbreviations: PBMCs, peripheral blood mononuclear cells; SFCs, spot-forming cells.

Serotype-Specific Patterns of Immunodominance Are Also Detected in the Nicaraguan Population

All 4 DENV serotypes circulated in Nicaragua in recent years, and as a result, the adult population has generally been exposed to multiple serotypes. Accordingly, we expected that significant T-cell reactivity would be detected against all 4 serotypes.

When the data were analyzed following the same procedure described above for the data derived from the Sri Lankan cohort, we observed that CD8⁺ T-cell reactivity to DENV2 was dominant, with 46% of the total response directed toward this serotype (Figure 3A). However, appreciable levels of T-cell reactivities to the rest of the DENV serotypes were also detected (ie, 10.5% to DENV1, 10.5% to DENV3, and 14% to DENV4), thereby enabling analysis of serotype-specific patterns of immunodominance. Responses against conserved epitopes accounted for 19% and were also predominantly located in the more conserved nonstructural proteins (Figure 3B).

Figure 3.

Serotype distribution and conserved responses in Nicaragua. A, Distribution of CD8⁺ T-cell epitope location across the 4 serotypes and conserved regions. B, Conserved CD8⁺ T-cell interferon γ (IFN-γ) responses across the dengue virus (DENV) proteome. IFN-γ responses are displayed as spot-forming cells (SFCs) per 10⁶ peripheral blood mononuclear cells. The percentages labeled in panel B represent the proportion of responses targeting epitopes in structural and nonstructural proteins.

Further inspection of the data revealed that different immunodominance patterns for the CD8⁺ T-cell response are observed as a function of different serotypes. DENV3-specific responses predominantly targeted structural proteins (58%; Figure 4C), while DENV1-, DENV2-, and DENV4-specific responding epitopes were found to disproportionately target nonstructural proteins (Figure 4A, 4B, and 4D).

Figure 4.

Serotype-specific CD8⁺ T-cell responses in Nicaragua. Dengue virus 1 (DENV1)–specific (A), DENV2-specific (B), DENV3-specific (C), DENV4-specific (D) CD8⁺ T-cell interferon γ (IFN-γ) responses across the DENV proteome. IFN-γ responses are displayed as spot-forming cells (SFCs) per 10⁶ peripheral blood mononuclear cells. The percentages labeled in all panels represent the proportion of responses targeting epitopes in structural and nonstructural proteins.

Definition of a Pool of Epitopes to Globally Detect DENV-Specific Responses

We next studied the overlap between the repertoire of epitopes identified in the current Nicaragua study and that previously identified in the Sri Lanka study [8]. We found that 69% of the epitopes identified in the Nicaragua study had also already been identified in the previous study. Many of the novel epitopes identified were restricted by alleles frequent in Nicaragua but not Sri Lanka, such as HLA A*30:02, A*68:02, B*44:02, and B*08:01. On the basis of this observation, we reasoned that a pool containing all epitopes identified in either study should be a generally applicable tool to characterize DENV-specific responses in bulk without requiring prior knowledge of either infecting serotype or HLA type.

Accordingly, we generated a pool of 268 epitopes selected to account for 90% of the IFN-γ response in both sample sets. Next, the reactivity to this DENV mega-pool was tested in intracellular cytokine staining assays, using PBMCs from randomly selected DENV-seropositive donors from Sri Lanka and Nicaragua. A positive ex vivo IFN-γ response was detected in 8 of 11 Sri Lankan donors and 7 of 12 Nicaraguan donors (Figure 5A and 5B). It should be possible to look at responses separately for the different serotypes in patients with an unknown HLA type but a known infecting serotype. However, because of the low number of primary infections with any of the 4 serotypes within our cohorts, we were unable to test and validate serotype-specific pools.

Figure 5.

CD8⁺ T-cell reactivity of the global epitope set. A, Gating strategy to assess the CD8⁺ interferon γ (IFN-γ) response against the mega-pool and representative fluorescence-activated cell-sorting plots from 1 donor each from Sri Lanka, Nicaragua, Brazil, and San Diego. Donor specimens were stimulated with the mega-pool for 6 hours, and the IFN-γ response was measured by intracellular cytokine staining. B, Percentages of CD8⁺ T cells that produce IFN-γ upon stimulation with the mega-pool in specimens from donors previously exposed to DENV from Sri Lanka (n = 11), Nicaragua (n = 12), and Brazil (n = 12) are shown. Specimens from DENV-seronegative donors from the greater San Diego area (n = 10) were also tested for reactivity against the mega-pool. The dotted line at 0.05% represents the cutoff for positivity. C, The average response (+standard error of the mean) for all cohorts is shown. Statistical significance was determined using a 2-tailed Mann–Whitney test. Abbreviations: FSC, forward scatter; NS, nonsignificant; SSC, side scatter.

As mentioned above, only about 30% of the epitopes identified in screening the Nicaraguan cohort were novel, despite the large differences in demographic and epidemiologic characteristics between the Nicaraguan and Sri Lankan cohorts. We estimated that further screening in new localities would reveal decreasing numbers of new epitopes. If linearity is assumed, a screen in a new cohort would be expected to reveal that only about 10% were previously unidentified epitopes. Accordingly, we hypothesized that this present epitope set would allow characterizing responses in other geographical locations.

To test whether this same peptide mega-pool could be used to stimulate DENV-specific responses in an independent cohort, we used PBMCs from 12 randomly selected DENV-seropositive donors from Sao Paulo. Positive responses were detected in 8 of 12 Brazilian donors (Figure 5A and 5B). As a control, the mega-pool was also tested in 10 seronegative donors from the greater San Diego area. No response responses were detected in any of these donors (Figure 5A and 5B). Responses measured in all cohorts of seropositive donors were significantly higher than responses measured in the seronegative control group (P = .001, P = .02, and P < .001 for Sri Lanka, Nicaragua, and Brazil, respectively, by the 2-tailed Mann–Whitney test; Figure 5C). Significant differences between the cohorts in terms of response frequencies or magnitudes were not detected (Figure 5C). In conclusion, the results presented herein suggest that the DENV mega-pool has reactivity in diverse global cohorts, suggesting that this tool allows broad characterization of DENV-specific CD8⁺ T-cell responses.

DISCUSSION

Here we report that different DENV serotypes are associated with distinct immunodominance patterns. DENV1, DENV2, and DENV4 all elicit a CD8⁺ T-cell response predominantly focused on nonstructural proteins (mainly NS3, NS4b, and NS5). In contrast, DENV3 is unique in that it elicits CD8⁺ T-cell responses that target both structural proteins (C, M, and E) and nonstructural proteins. This work confirms in naturally infected human populations the results of previous studies performed in a humanized mouse model and in human vaccinees receiving either monovalent or tetravalent attenuated dengue vaccines [4, 9]. In contrast to these previous studies, which were controlled experiments performed in dengue-naive mice and volunteers, this study analyzes individuals living in dengue-endemic areas who have encountered multiple DENV infections. In both cohorts, the response against DENV3 structural proteins was >50%, compared with 37% in the volunteers vaccinated with the monovalent DENV3 vaccination. It is possible that the attenuation of the vaccine strain results in a slightly lower skewing toward DENV3 structural proteins than natural infection. Therefore, although the findings may be confounded by the presence of multiple DENV infections, they reflect a more realistic scenario in which DENV infections occur.

Cross-reactivity is difficult to assess in humans since the exact sequence of the infecting virus is mostly unknown. Thus we cannot exclude that we may be underestimating the number of conserved epitopes. We have addressed this question in our murine system, in which we tested predicted peptides with no, 1, 2, 3, or 4 amino acid substitutions, compared with the DENV strain used as the infecting agent in the study [4]. The majority of cross-reactivity was detected between peptides with >80% sequence homology, which translates into 1–2 amino acid substitutions.

The differential skewing of DENV T-cell immunity has not previously been appreciated, and it is interesting to speculate on possible causes and implications. In terms of potential molecular mechanisms, it has been reported that DENV serotypes differ in their relative capacity to activate innate immunity and type I IFN pathways [24]. This may translate into differences in the processing pathways and the type of antigen-presenting cells involved in the elicitation of T-cell responses. It has been shown that autophagy promotes presentation of peptides from intracellular source proteins [25]. One of the mechanisms DENV uses to hijack host cell machinery to facilitate viral replication involves inducing autophagy, which is necessary for virus maturation and production of infectious virions [26, 27]. Conversely, if DENV3 antigens are selectively processed through cross-priming mechanisms, this might explain the induction of CD8⁺ T-cell responses directed against structural proteins.

Regardless of the underlying mechanism(s), the serotype-specific differential recognition of DENV could have important implications for immunity and vaccination. First, it is possible that these differences could influence disease outcomes, such as by leading to a less cross-reactive CD8⁺ T-cell response in heterotypic infections involving DENV3 that could aggravate potential antibody-dependent enhancement effects. Secondary infections by any of the 4 DENV serotypes can evolve to dengue hemorrhagic fever and dengue shock syndrome, but not all sequences of infection exhibit the same risk for development of severe disease [2]. In this respect, it is interesting and potentially important that infections with DENV3 followed by DENV2 appear to be among the most virulent combinations of secondary DENV infections.

A recent efficacy trial of the most advanced dengue vaccine candidate, a tetravalent live-attenuated chimeric vaccine (CYD), in which all nonstructural proteins are derived from the yellow fever 17D vaccine, demonstrated only partial protection despite high neutralizing antibody titers against all 4 serotypes in most subjects. These results challenge the hypothesis that neutralizing antibodies are the only reliable correlate of protection [28]. Three efficacy studies found protection against disease caused by DENV2 to be inferior to protection against disease caused by DENV1, DENV3, and DENV4 [28–30]. This weak protection might be attributed to the relative absence of CD8⁺ T-cell responses to DENV2 nonstructural proteins. However, CYD vaccination presents DENV3 structural proteins to the host, possibly contributing to the greater vaccine efficacy observed against DENV3. It is possible that the protection against DENV2 is more T-cell dependent than the other serotypes. Furthermore, it has been reported that heterotypic protection against disease wanes with time. To more firmly establish the contribution of T-cell immunity to this phenomenon, it may be important to characterize T-cell responses in individuals who have experienced a single infection with each of the 4 serotypes and at different intervals following infection.

In conclusion, we show here that different serotypes are associated with distinctive immunodominance profiles with respect to the specific antigens recognized by CD8⁺ T-cell responses. Finally, by defining a globally reactive DENV-specific mega-pool of CD8⁺ epitopes, we provide a tool to investigate the CD8⁺ T-cell responses in small sample volumes without the need for knowledge of infecting serotype or HLA type.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.