Immunodominant T-cell responses to dengue virus NS3 are associated with DHF

Abstract

Dengue infections are increasing at an alarming rate in many tropical and subtropical countries, where epidemics can put health care systems under extreme pressure. The more severe infections lead to dengue hemorrhagic fever (DHF), which can be life threatening. A variety of viral and host factors have been associated with the severity of dengue infections. Because secondary dengue infection is more commonly associated with DHF than primary infections, the acquired immune response to dengue, both B cells and T cells have been implicated. In this study, we set out to study T-cell responses across the entire dengue virus proteome and to see whether these were related to disease severity in a cohort of dengue-infected children from Thailand. Robust responses were observed in most infected individuals against most viral proteins. Responses to NS3 were the most frequent, and there was a very strong association between the magnitude of the response and disease severity. Furthermore, in DHF, cytokine-high CD107a-negative cells predominated.

Dengue virus is transmitted to man via the bite from an infected mosquito, and infections are a growing threat to public health in a number of tropical and subtropical countries. There are estimated to be up to 50 million dengue infections annually, and epidemic activities have the potential to almost overwhelm healthcare systems (1). Dengue viruses display substantial sequence diversity, allowing them to be classified into four distinct serotypes that differ in primary amino acid sequence by 30–35%. This sequence divergence is nearly as great as the difference between dengue and other members of the family flaviviridae. Importantly, the sequence divergence means that infection with one serotype of dengue does not provide immunity to infection with other serotypes (2). In many parts of the world, all four serotypes of dengue cocirculate, meaning that sequential or secondary dengue infections are common (3–5).

After the bite from the infected mosquito, there is a period of fever that occurs during a time of high viraemia. In most cases, referred to as dengue fever (DF), the virus is controlled, the fever remits, and the patient makes an unremarkable recovery. In 1–5% of cases, the disease can become more severe with extensive plasma leak and sometimes hemorrhage (6). This syndrome is known as dengue hemorrhagic fever (DHF) and can be life threatening with a 20% fatality rate that can be reduced to well below 1% with expert management that mainly relies primarily on careful fluid management (1).

Forty years ago, it was recognized that the majority of cases of DHF occur in patients suffering secondary dengue infections (6–8), although there is also a peak of severe dengue in children experiencing primary infections during the first year of life (9, 10). The association with secondary infections is perhaps best exemplified by careful epidemiological studies of dengue infection in island populations such as Cuba (7).

The pathogenesis of DHF is the subject of some debate in the field. The etiology is likely to be multifactorial with contributions from both the virus and host. However, there are three key observations that must be borne in mind. First, as mentioned above, DHF is more common after a secondary infection (6–8). Second, a high peak virus load is associated with a higher risk of DHF (11, 12). Finally, the severe symptoms of DHF tend to occur at the time of virus control when virus loads are falling steeply (11, 12).

In 1977, Halstead put forward the hypothesis of antibody-dependent enhancement (ADE) to explain the link between severity and secondary infection (13). However, although patients with high dengue viraemia are unwell with high fever and constitutional symptoms, they do not appear to develop DHF until around the time they control the virus (11, 12). The association of DHF with virus control and a peak in inflammatory cytokine production has led some to suggest that the disease is a form of immunopathology (14, 15), which shares some features with experimental models of T cell-mediated immunopathology in mice (16–19).

In this study, we have performed a global scan for T-cell responses by using overlapping peptides covering the entire dengue proteome. Robust T-cell responses can be seen after infection, and although the whole of the virus can be recognized, responses are particularly focused on nonstructural protein 3 (NS3). A clinical study of children suffering dengue infection showed that the amplitude of response was highly associated with the severity of the disease with high cytokine and low CD107a expressing T cells predominating the response in DHF.

Results

Detection of Dengue-Specific T Cells in PBMC from Dengue-Infected Patients.

Dengue is a positive-strand RNA virus belonging to the family flaviviridae. The nucleic acid is ≈11 kb in length and is transcribed as a polyprotein that is cleaved by both viral and host-encoded proteases into 10 polypeptides (20). The virus has three structural proteins—capsid, premembrane (prM), and envelope—together with seven nonstructural (NS) proteins—NS1, NS2a, NS2b, NS3, NS4a, NS4b, and NS5. A number of T-cell epitopes for dengue have been reported (21–23), but to our knowledge, no previous studies have looked for responses to the whole dengue virus proteome.

Overlapping peptides from dengue serotype 2 strain 16681 (Den2) were aliquoted into seven pools corresponding to individual dengue serotype 2 proteins (C-prM), E, NS1, (NS2a-NS2b), NS3, (NS4a-NS4b), and NS5. Staphylococcal enterotoxin B (SEB; 1 μg/mL) and peptide diluent (0.8% DMSO in PBS) were used as positive and negative controls, respectively.

PBMC were collected, after informed consent, and cryopreserved from children seen in hospital with dengue infection. Dengue infection was determined serologically, and virus serotype where possible was determined by serotype specific RT-PCR (24). Overlapping peptides were used to stimulate 2-wk convalescent PBMC from 40 dengue-infected patients—18 secondary DF, 22 secondary DHF; the patient details are shown in Table S1. Responses of CD4⁺ and CD8⁺ T cells were measured by using intracellular cytokine staining (ICS) for IFN-γ and TNF-α, and the ICS strategy is outlined in Fig. 1A.

Fig. 1.

T-cell responses to the dengue proteome. Shown are representative plots for CD4⁺ and CD8⁺ T-cell subsets from PBMC stimulated with peptide pools covering the dengue proteome. (A) Gating strategy. Representative sets of responses for CD4 (B) and CD8 (C) T cells.

Significant responses were seen in most patients with low levels of background reactivity. A representative set of responses is shown in Fig. 1 B and C, where robust responses can be seen in both the CD4⁺ and CD8⁺ T-cell subsets to a variety of viral proteins, in particular to NS3.

NS3 Is the Immunodominant T-Cell Antigen in Dengue Infection.

Next, we analyzed the data from all 40 individuals selected at random to compare the level of responses between the different dengue peptide pools, the details of these patients are shown in Table S1. Positive responses were called when they were greater than the percentage of cytokine-producing cells plus 3 SDs from five dengue nonimmune volunteers. Most patients showed responses to several epitopes, particularly patients in the DHF group where all of the patients responded to NS3 protein (Fig. 2A). Responses to E, NS1, and NS5 proteins were also common in DHF, with >70% of DHF patients having T-cell response to these proteins. Similar response patterns were found in the DF group where NS3 elicited T-cell responses in the majority of patients (78%). Responses to E, NS1, NS2, and NS5 were found in most DF patients; 30% for E, NS1, and NS2 and 44% for NS5. Similar patterns of reactivity were seen when the CD4⁺ and CD8⁺ T-cell subsets were analyzed separately (Fig. S1 A and B).

Fig. 2.

Breadth and magnitude of dengue specific T-cell responses. (A) T-cell responses against dengue virus were broad; the numbers above the bars represent the number of responding patients. (B) Magnitude of the responses to the peptide pools in the DF and DHF patient groups. (C) Sum of responses to all peptide pools for each patient. Differences between the two groups were tested by using the nonparametric two tailed Mann–Whitney test. Asterisks indicated P values: *P < 0.05; **P < 0.005; ***P < 0.0005.

Importantly, we found a highly significant association between the magnitude of the T-cell response to dengue and disease severity (Fig. 2B). The sum of all of the responses across the entire dengue proteome in the DHF group was 5.22 ± 3.85% versus 2.02 ± 1.59% in the DF patients, P = 0.0008 (Fig. 2C).

Comparison of T-Cell Responses to NS3 in Primary and Secondary Responses.

The analyses presented above were performed by using peptides derived from the dengue serotype 2 strain 16681. The patients recruited into the study were suffering infection with a variety of dengue serotypes, meaning that in many cases, the test antigen did not match the serotype of the current dengue virus infection. To correct for this mismatch, we synthesized additional overlapping peptides covering the immunodominant antigen NS3 from prototype strains of Den1, Den3, and Den4.

In this second analysis, four groups of patients were selected, 10 primary DF cases, 5 primary DHF cases, 15 secondary DF cases, and 15 secondary DHF cases, and details of these patients are shown in Table S2. PBMC were incubated with an overlapping pool of NS3 peptides corresponding to the serotype of the current infection as determined by RT-PCR. After incubation, cells were stained for CD4 and CD8 and intracellular staining performed for TNF-α and IFN-γ, and in addition, cells were stained with anti-CD107a, which is used as a surrogate measure of degranulation (25).

Studies of T-cell responses in dengue have to date concentrated on individuals suffering secondary infection in large part because of the relative scarcity of primary cases, and in particular, primary DHF. T-cell responses could be readily detected in primary cases by stimulation with the NS3 overlapping peptides. Responses in primary DHF patients were significantly higher than in DF patients (TNF-α, P = 0.0080, and IFN-γ, P = 0.0013) (Fig. 3A). This highly significant association between the magnitude of the T-cell response and severity was also found in secondary dengue infections, TNF-α (P < 0.0001) and IFN-γ (P = 0.0001). These findings were similar in CD4⁺ and CD8⁺ T cells (Fig. S2).

Fig. 3.

NS3-specific responses in primary and secondary infections. (A) PBMC were stimulated with NS3 from the serotype of their current infection. Total responding cells (CD107a or IFN-γ or TNF-α) or individual responses to CD107a, IFN-γ, and TNF-α are shown. Differences between the two groups were tested by using the nonparametric two-tailed Mann–Whitney test. (B) Responding T cells from Fig. 3A were divided into three groups: degranulation (CD107a) only, degranulation and cytokine production, and cytokine production only. Pie charts show the percentages of the functional cells to total responding T cells from patients in the DF and DHF groups. The average percentages were calculated by summing the percentages from individuals and dividing by number of cases.

More Cytokine Production and Less Degranulation in DHF.

There has been much recent interest in the development of T-cell vaccines and measures of their efficacy. It is becoming clear that it is not just the magnitude of the T-cell response but also the quality of the induced T cells that are important for function, with the suggestion that polyfunctional T cells are the most desirable.

Analysis of the T cells induced in dengue illustrated two interesting features. First, the majority of T cells are monofunctional with respect to IFN-γ, TNF-α secretion, and degranulation (CD107a) (Fig. S3). Second, there were clear differences in the type of response mobilized in DF when compared with DHF.

In Fig. 3B, we have broken these data down into three groups: CD107a only, cytokine (TNF-α or IFN-γ) only, and double positive for cytokine and CD107a. CD107a-expressing cells were relatively distinct from cells producing TNF-α and IFN-γ, with the majority expressing no associated cytokine. In addition, CD107a-producing cells were rarer in cases of DHF where cytokine-producing cells were the dominant population. In primary cases, 71% of cells were CD107a-positive in DF compared with 13% in DHF (P = 0.0007), and similarly in secondary infections, CD107a cells accounted for 45% of responding cells in DF and 18% in DHF (P = 0.0055). This difference was mirrored by a commensurate rise of cytokine only producing cells in both primary and secondary DHF. Both CD4⁺ and CD8⁺ T cells gave similar patterns (Fig. S4).

In a final series of experiments, we used MHC tetramer staining to focus on an immunodominant HLA-A11–restricted epitope from NS3 (GTS) that we have described (22). PBMC from 10 HLA-A*11–positive individuals (5 DF, 5 DHF, all secondary infections; details shown in Table S3) were stained with HLA-A*11 tetramer refolded with the peptide corresponding to the serotype of their secondary infection. Cells were subsequently stimulated with the same dengue serotype specific peptide and stained for CD107a, IFN-γ, and TNF-α.

The results of these experiments are shown in Fig. 4A, where it can be seen that the production of IFN-γ and TNF-α were higher in the DHF group compared with DF; this reached statistical significance for IFN-γ. Furthermore, in the DF patients, a higher proportion of responding cells showed CD107a staining compared with the DHF group DF 60% vs. DHF 31% (P = 0.0079) (Fig. 4B).

Fig. 4.

MHC tetramer analysis of the HLA-A11-restricted NS3-GTS response. (A) PBMC from 10 secondary dengue-infected patients, 5 DF and 5 DHF (Table S3), were stained with the HLA-A*11-NS3-GTS tetramer corresponding to the serotype of their secondary infection. After stimulation with the same peptide used for tetramer staining and in the presence of anti-CD107a Ab, cells were stained for CD8, TNF-α, and IFN-γ. (B) Pie charts showing the details of the tetramer-specific responses by following the format of Fig. 3B.

Discussion

To our knowledge, this is the first study that has comprehensively examined T-cell responses in dengue-infected patients across the entire dengue proteome. These data from 40 cases show that responses to the virus are broad and can recognize most of viral proteins, particularly in cases of DHF (Fig. 2 A and B). When the responses to individual proteins are compared, responses to NS3, a viral protease, NTPase, and helicase, are dominant. Significant responses were also seen against envelope and NS5. These results concur with the reports of a number of T-cell epitopes in dengue infection where NS3 responses are the most frequently seen, although this analysis shows the response to be considerably broader than reported (21–23, 26). In a recent report on T-cell responses in human vaccines, receiving the yellow fever 17D vaccine, the NS3 response was subdominant to NS5 in most individuals (27). Interestingly, in this study, there was a major dominant epitope in NS4b found in HLA-A*02-expressing individuals, with the similar result was seen in dengue vaccines (28).

The determinants of immunodominance are not fully understood but are presumed to result from a variety of events occurring in the generation of the T-cell response: the amount of antigen produced, the timing of protein production, and host-specific factors such as antigen processing, MHC, and TCR repertoires (29). The reason for the dominance of the NS3 response in dengue is not clear, unlike other viruses in which programs of antigen production lead to the emergence of different antigens early and late during the virus lifecycle dengue is transcribed as a single polyprotein that is cleaved into its 10 constituent polypeptides by host- and virus-specific proteases that include NS3 (20). This processing presumably means that the polypeptides are produced in equimolar concentrations. However, possible differences in intracellular targeting and stability may lead to preferential processing and presentation of NS3 peptides.

Extensive sequence diversity is a feature of many RNA viruses because of poor fidelity of the RNA polymerase enzyme. Most sequence changes are detrimental to viral fitness, reducing the efficiency of virus replication. However, many sequence changes are tolerated, and the emergence of new viral strains is believed in part to be driven by immune responses, leading to the emergence of immune escape variants. This phenomenon is perhaps best exemplified for the antibody response by influenza and for the T-cell response by HIV (30–32). Sequence diversity in dengue viruses is ≈30–35%, meaning that viruses differ from each other by nearly as much as the dengue viruses themselves differ from other members of the flavivirus family such as Japanese encephalitis virus (50%). The sequence variation in dengue is such that the virus can be divided into four distinct serotypes where by definition immunity to one serotype will not provide long-lasting immunity to the others.

In dengue-endemic areas, multiple viral serotypes cocirculate, although the relative frequencies of the four viruses vary from year to year, perhaps driven by the heard immunity of susceptible human hosts (3–5). Secondary dengue infections are common, and the severe disease manifestations of DHF and shock are more commonly seen in these secondarily infected individuals (6–8). DHF pathogenesis is likely to be multifactorial with contributions from the virus and the host. High peak virus loads are associated with a higher risk of DHF (11, 12), and one explanation for the higher virus loads is antibody-dependent enhancement. ADE can be readily demonstrated in vitro and has also been demonstrated in vivo in both murine and primate models (33–36). ADE may also explain the peak in DHF seen in children during the first year of life where passively transferred maternal antibody may enhance infection (10, 37).

However, although there is a correlation between virus/antigen load and severity, the severe sequaelae of dengue infection occur when the virus load is falling steeply and are coincidental with sharp rises in inflammatory cytokine production. Many of these cytokines can be T cell-derived, and the coincidence of severe disease with viral clearance has led a number of investigators to suggest the DHF may result from T cell-mediated immunopathology (15, 21, 38, 39). Immunopathology is now being recognized to be a feature of a number of viral diseases where the protective antiviral response, although capable of controlling virus, may cause considerable inflammation and tissue damage that can, in some instances, be fatal. In humans, immunopathology has been invoked in severe influenza infections, RSV after ineffective vaccination, severe hepatitis in hepatitis B virus infections, and the immune reconstitution syndrome seen in individuals coinfected with tuberculosis and HIV upon the commencement of antiretroviral therapy (40–43).

We had described pronounced original antigenic sin in the T-cell response to dengue virus whereby the T-cell response in secondary infection is shaped in part by the response to the primary encounter with dengue. This phenomenon leads to highly cross-reactive responses that in some cases are more avid to the primary encountered virus than to the secondary infection (22, 23). In this report, we went on to examine the T-cell response to NS3 in four groups of patients: primary and secondary infections on both DF and DHF groups; in each case, we examined the response to NS3 from the serotype of virus responsible for their current infection.

There were highly significant differences in both IFN-γ and TNF-α production in the DHF groups when compared with DF groups regardless of primary or secondary infection. This finding is in keeping with reports of correlations between the serum levels of inflammatory cytokines such as IFN-γ, TNF-α, and interleukins and disease severity (6). Polyfunctional analysis of T-cell responses has recently been of great interest in the vaccine field where it is now recognized that the ability of T cells to control virus replication does not correlate well with the overall magnitude of the response. Results from analysis of effective vaccines such as bacillus Calmette–Guérin, vaccinia, and yellow fever 17D virus or effective HIV controllers versus rapid progressors suggest that T cells with multiple functions including IFN-γ, TNF-α, CD107a, and IL-2 are the most effective at controlling virus (27, 44–47).

In this study, we found that the majority of patients made monofunctional or oligofunctional responses and cells producing cytokine together with CD107a were rare. Of great interest in this study was the finding that in both primary and secondary DHF, <20% of T cells produce CD107a, whereas >80% produce cytokine only. This situation is reversed in primary DF where 71% produce CD107a and 29% cytokine only, and secondary DF 45% produce CD107a. CD107a expression was found on CD4⁺ T cells in addition to CD8⁺ in DF, which fits with reports of cytotoxicity of antidengue CD4⁺ T-cell lines, and it is possible that this population plays an important cytotoxic role in vivo (48, 49).

These results suggest that the functional phenotype of dengue responsive T cells, together with the magnitude of the T-cell response, may play a role in the development of severe disease. We propose that the limited cytotoxic potential of T cells in a subset of patients may not allow early and effective viral control. Instead, the cytokine-only producing T cells are stimulated, which in the context of high virus/antigen loads will lead to excessive production of inflammatory cytokines triggering the tissue damage and leak characteristic of DHF.

Materials and Methods

Samples.

Blood samples were taken from normal healthy subjects and from patients admitted to the Pediatric Departments of Khon Kaen and Songkhla hospitals after informed consent. The ethical committees of Khon Kaen, Songkhla, and Siriraj hospitals, Thailand, as well as the Riverside ethical committee in the United Kingdom approved the studies. Acute dengue infection was identified by RT-PCR-based dengue gene identification or dengue-specific IgM capture ELISA (24, 50, 51). Secondary dengue infection (an acute infection in a patient who had previously encountered dengue on one or more occasions) was defined as a dengue-specific IgM/IgG ratio < 1.8, by IgM and IgG capture ELISA by using paired acute and convalescent sera.

Disease severity was classified according to the World Health Organization criteria (52). PBMC were isolated from whole blood by Ficoll-Hypaque density gradient centrifugation and cryopreserved for subsequent studies. Patient details were made anonymous and stored on a password protected database, patients in both DF and DHF groups were selected at random. In both the global scan of the whole virus proteome and the targeted study of NS3 responses, the ages and male:female ratio (in parentheses) of children were as follows: for the global scan, DF = 9.4 ± 3.6 y (7:11) and DHF = 10.3 ± 2.8 y (10:12) and for NS3 responses, DF = 10.3 ± 2.2 y (6:4) and DHF = 10.8 ± 2.9 y (3:2) for primary infection, and DF = 10.4 ± 2.9 y (8:7) and DHF = 11.1 ± 0.3 y (8:7) for secondary infection. The details of these patients are shown in Tables S1–S3.

Dengue nonimmune controls were a group of five individuals who had never traveled to a dengue endemic area and who had not received either yellow fever or Japanese encephalitis vaccination. These were students from the laboratory who donated blood after informed consent and under a protocol approved by the Riverside Ethics Committee UK.

Peptides.

15-mer peptides overlapping by 5 amino acids corresponding to sequences of dengue serotype 2 strain 16681 (CprM, E, NS1, NS2, NS3, NS4, and NS5) and to the NS3 sequences of dengue serotype 1 strain Hawaii, dengue serotype 3 strain H87, and dengue serotype 4 strain H241 were synthesized to >80% purity (Sigma Aldrich). Lyophilized peptides were resuspended in dimethyl sulfoxide (DMSO; Sigma) at 80 mg/mL for peptide mixtures. The overlapping peptides were grouped together in pools corresponding to individual dengue proteins (dengue 2: CprM-26, E-97, NS1-71, NS2-70, NS3-123, NS4-70, NS5-95, dengue 1 NS3-124, dengue 3 NS3-124, and dengue 4 NS3-124). The final concentration of any single peptide was 2 μg/mL, and the final concentration of DMSO was <1% in all experiments.

Cell Stimulation for Multifunctional Assessment.

PBMC were thawed and rested in complete RPMI media (RPMI 1640 supplemented with 10% heat inactivated FCS, L-glutamine, and antibiotics) for at least 2 h at 37 °C in a 5% CO₂ incubator; viability was then examined by using Trypan blue dye exclusion. Cell viability was >85% for all cases. For stimulation, briefly, PBMC were incubated in the presence of peptide mixes (final concentration of each peptide was 2 μg/mL) and the titrated amount of CD107a FITC (BD Pharmingen) for 1 h at 37 °C in a 5% CO₂ incubator, followed by an additional 5 h in the presence of the protein transport inhibitors monensin (BD Biosciences) and brefeldin A (BFA) (Sigma). A negative control, PBMC incubated in the peptide diluents (0.8% DMSO in PBS), and a positive control, PBMC incubated with 1 μg/mL of staphylococcal enterotoxin B (SEB), were included in all experiments.

Tetramer Staining and Cell Stimulation.

HLA-A11 was refolded around a 10-mer epitope GTS present between positions 133 and 142 of the NS3 protein. Three separate variants of this peptide were used for refolding GTSGSPIVNR for serotype 1, GTSGSPIVDR from serotype 2, and GTSGSPIINR, which is found in both dengue serotypes 3 and 4. PBMC were stained with the phycoerythrin (PE)-conjugated GTS-A*11 tetramer, which corresponded to the serotype of their current secondary dengue infection, at 37 °C for 20 min. Thereafter, cells were incubated with 10 μg/mL of the same GTS peptide that corresponded to the serotype of their current infection. Cells were stained for CD107a, CD8, and cytokine as described above and below. A negative control, cells incubated without peptide, was included to control for spontaneous production of cytokine and/or expression of CD107a. A positive control, cells incubated with 1 μg/mL of SEB, was also included in all assays. A gate was set around the CD8⁺ tetramer⁺ cell population, and the function of these cells with respect to CD107a, TNF-α, and IFN-γ secretion were calculated by using gates set on the negative control-unstimulated tetramer positive cells.

Immunofluorescent Staining.

Immediately after stimulation, PBMC were washed once and fixed/permeabilized by using FACS permeabilization buffer II (BD Pharmingen). For four-color analysis, the cells were washed and stained with the titrated amount of IFN-γ APC, TNF-α FITC, CD4 PE, and CD8 PerCP. For five-color analysis, the cells were stained with the titrated amount of IFN-γ APC, TNF-α PE-Cy7, CD4 PE, and CD8 APC-Cy7. The cells were then washed and fixed with 1% formaldehyde in PBS. All antibodies were purchased from BD Pharmingen.

Flow Cytometric Analysis.

Two hundred thousand to 500,000 events were collected per sample. Analysis was performed by using FlowJo software (version 7; TreeStar). Background responses detected in negative control tubes were subtracted from those detected in stimulated samples for every specific functional combination.

Statistical Analysis.

Differences between the DF and DHF groups were all evaluated by the nonparametric two-tailed Mann–Whitney test by using the Prism program (Graph Pad Software).