Research ArticleIMMUNOLOGY

Flavivirus serocomplex cross-reactive immunity is protective by activating heterologous memory CD4 T cells

See allHide authors and affiliations

Science Advances  04 Jul 2018:
Vol. 4, no. 7, eaar4297
DOI: 10.1126/sciadv.aar4297


How previous immunity influences immune memory recall and protection against related flaviviruses is largely unknown, yet encounter with multiple flaviviruses in a lifetime is increasingly likely. Using sequential challenges with dengue virus (DENV), yellow fever virus (YFV), and Japanese encephalitis virus (JEV), we induced cross-reactive cellular and humoral immunity among flaviviruses from differing serocomplexes. Antibodies against JEV enhanced DENV replication; however, JEV immunity was protective in vivo during secondary DENV1 infection, promoting rapid gains in antibody avidity. Mechanistically, JEV immunity activated dendritic cells and effector memory T cells, which developed a T follicular helper cell phenotype in draining lymph nodes upon secondary DENV1 infection. We identified cross-reactive epitopes that promote recall from a pool of flavivirus serocomplex cross-reactive memory CD4 T cells and confirmed that a similar serocomplex cross-reactive immunity occurs in humans. These results show that sequential immunizations for flaviviruses sharing CD4 epitopes should promote protection during a subsequent heterologous infection.


Flaviviral pathogens are primarily transmitted to humans by arthropod bites (1). This group is composed of several pathogenic viruses, including Japanese encephalitis virus (JEV), yellow fever virus (YFV), West Nile virus (WNV), dengue virus (DENV), and newly emergent Zika virus (ZIKV) (2, 3). The widespread distribution of flaviviruses results in hundreds of millions of infections annually and exposes many more to the risk of disease (1, 4). Moreover, cocirculation of multiple flaviviruses in some geographical regions (5) and gains in vaccination coverage for YFV and JEV (6), combined with the modern surge in human geographic mobility, have increased the likelihood of exposure to multiple flaviviruses within a lifetime. This raises the important question of how preexisting immunity will influence the outcome of subsequent vaccination or infection with a heterologous flavivirus.

The genus Flavivirus is composed of nearly 70 known viruses, organized into serocomplexes (7). Human infection results in production of both virus species-specific and flavivirus cross-reactive antibodies (8). The influence of serocomplex cross-reactive immunity during a subsequent flavivirus infection has not yet been clearly defined; however, several studies suggest that cross-reactive antibodies can be protective. For example, primary infection with JEV, YFV, or DENV1 protected animals from a virulent secondary YFV challenge in a hamster model, which was interpreted as showing a protective role for cross-reactive antibodies (9). DENV-immune humans were also observed to experience milder YFV symptoms than DENV-naïve individuals (10), again suggesting a protective role for previous flavivirus exposure during a secondary heterotypic infection.

Although few studies have examined cross-reactive immune responses among flaviviruses from differing serocomplexes (11, 12), more is known regarding cross-protection by viruses within the same serotype or closely related virus strains (13, 14). For example, cross-protection for a new serotype or strain of DENV, when it occurs, is thought to be achieved primarily through neutralizing antibodies that mainly recognize epitopes present on the viral envelope glycoprotein surface (15, 16). In contrast, subneutralizing (due to either antibody quality or concentration) flavivirus cross-reactive antibodies are believed to increase the severity of a secondary infection due to antibody-dependent enhancement (ADE) of infection (17, 18), a phenomenon that is well supported in the context of repeated infections by DENV serocomplex viruses (19, 20), as well as JEV serocomplex viruses (21, 22). ADE occurs when Fcγ receptors promote internalization of virus-antibody immune complexes within target cells (23, 24). In vivo, ADE has been shown using an immunocompromised mouse model (25, 26), yet evidence in humans remains indirect for DENV (2729), although we have recently shown that antibodies against JEV can lead to ADE of the YFV vaccine in humans (30).

Flaviviruses also induce serocomplex and serotype cross-reactive T cell responses (31, 32). For example, a recent study showed that DENV immunity enhances the kinetics and magnitude of T cell responses to ZIKV during ex vivo stimulation (33). During heterologous DENV infections in mice, cross-reactive T cells increase, suggesting a possible role of T cells in heterologous DENV immunity (34). Serotype cross-reactive T cells have also been identified in human DENV cohorts (35, 36). Some reports suggest an association between cross-reactive T cells and severe disease during heterologous DENV infections due to “original antigenic sin” (36). Yet, more recent evidence has indicated that although there is skewing toward the cross-reactive CD8 epitopes for the primary infecting strain during heterologous DENV infection, this cross-reactivity does not impair immunity (37). Thus, flavivirus cross-reactive immunity has the potential to either help or harm the course of immunity to a subsequent infection, but many factors including the degree of genetic similarity, conformation of structural proteins, and existence of cross-reactive T cell epitopes all may influence how immunity to one flavivirus can modify the development of immunity and clinical outcome to another.

The mechanisms by which flavivirus cross-reactive antibodies or memory T (TMEM) cells contribute to immune protection or pathology remain unclear. Here, using an immunocompetent mouse model, we defined experimentally how cross-reactive immunity modulates the course of infection during various secondary heterotypic flaviviral challenges. Serocomplex cross-reactive responses were also identified in humans. Our results suggest novel strategies for promoting cross-protection for flavivirus immunization.


Serocomplex cross-reactive immunity elicited by flaviviruses

To assess the influence of preexisting cross-reactive immunity on the development of immunity to a subsequent flavivirus challenge, we chose three representative pathogens from independent serocomplexes: JEV (SA-14-14-2), YFV (YFV-17D), and DENV1 (EDEN1). Considering the overlapping geographical distributions of these viruses (Fig. 1, A to C) and their genome similarity (with each other and recently emerged pathogens, WNV and ZIKV; Fig. 1D), we expected that the viruses would elicit cross-reactive immunity, experimentally mimicking the serocomplex cross-reactive responses of humans (8). Immunocompetent mice were chosen because the primary objective was to study functional immune responses. The strains chosen for YFV and JEV are live-attenuated vaccine strains, while the DENV1 strain is a clinical isolate, also chosen because these are the challenges most likely to be experienced by humans currently, where YFV and JEV vaccinations are becoming more common than natural infections, but natural DENV exposure remains high. We infected groups of mice with each virus and quantified specific and serocomplex cross-reactive antibody titers generated (Fig. 2, A to C). DENV1-infected animals produced high-titer DENV1-specific antibody (Fig. 2, A to C) and low-titer cross-reactive antibodies against JEV but did not induce cross-reactive antibodies against YFV (Fig. 2, D to F). In general, immunization with YFV and JEV vaccine strains produced lower-concentration specific antibodies compared to an equivalent inoculation with the clinical isolate of DENV1 (Fig. 2, A to C). Both YFV and JEV immunizations were able to produce high-titer cross-reactive antibodies against DENV1 (Fig. 2, D to F), raising the potential that these antibodies could influence the course of secondary flavivirus infection.

Fig. 1 Overlapping geographic distribution and polyprotein similarity of flaviviruses from various serocomplexes.

Maps showing the current distributions of (A) DENV, (B) JEV, and (C) YFV serocomplexes reveal geographic regions where multiple flaviviruses cocirculate. (D) Phylogenic tree showing the genetic distances between the strains used for this study for DENV, JEV, and YFV, as well as representative strains of WNV, ZIKV, and Spondweni virus. The scale bar indicates the genetic distances in substitutions per amino acid.

Fig. 2 Cross-reactive and cross-protective flavivirus immunity.

End-point titers of specific versus cross-reactive serum immunoglobulin G (IgG) against (A) DENV1, (B) YFV, or (C) JEV at each time point were measured by enzyme-linked immunosorbent assay (ELISA) for mice infected with DENV1, YFV, or JEV. Blood was collected 7, 14, and 21 days after infection. A comparison of specific versus cross-reactive end-point titers for mice infected with (D) DENV1, (E) JEV, or (F) YFV is presented for serum obtained 21 days after challenge. The avidity of serum antibodies toward each virus after challenge with (G) DENV1, (H) JEV, or (I) YFV is presented. Neutralization of virus by serum antibodies from mice infected with DENV1, YFV, and JEV or injected with saline was measured by plaque reduction neutralization test (PRNT) against (J) JEV, (K) YFV, or (L) DENV1. Results are presented as the percentage of neutralization compared to untreated control (virus alone) starting at a dilution of 1:10, with fourfold serial dilutions. For all viruses, only specific sera generated in mice inoculated with the same virus and not sera obtained after challenge with a related flavivirus were capable of virus neutralization. (M) Splenocytes were obtained from mice 5 weeks after infection and were used in proliferation assays. The results are given as the percentage of proliferation over splenocytes from the saline control mice. For all viruses, splenocyte proliferation was observed in response to homologous antigen, and for some viruses, splenocytes also proliferated in response to antigens from a heterologous virus. For (A) to (M), infections were performed by intraperitoneally injecting 1 × 106 plaque-forming units (PFU). (N) Serum and T cells were purified 4 to 5 weeks after infection from naïve and DENV1, JEV, or YFV post-immune mice (pooled from five mice per group). These products were transferred to naïve recipient mice (n = 5) before challenging all mice with DENV1. Alternatively, mice (n = 5) were given a secondary infection with DENV1 28 days after the primary challenge with DENV1, JEV, YFV, or saline. All secondary challenges were performed by subcutaneous injection with 1 × 105 PFU of DENV1. DENV1 was quantified in draining LNs after 24 hours by real-time reverse transcription polymerase chain reaction (RT-PCR). Results are expressed as a percentage relative to the primary DENV1 infection control (saline; followed by DENV1 infection). Viral clearance was enhanced during a homologous secondary DENV1 challenge after serum transfer, secondary infection, or T cell transfer. DENV1 was significantly reduced in JEV post-immune mice, while transfer of JEV post-immune serum enhanced DENV1 infection in LNs. Previous YFV immunity did not influence DENV1 viral load. For all panels, n = 5, *P < 0.05, and **P < 0.01. Cross-reactive low-avidity antibodies and T cells are generated by flavivirus infection; however, JEV, but not YFV, cross-reactive immunity enhances protection during secondary heterologous DENV1 challenge. ns, not significant.

To test the quality of the antibodies elicited, we measured their avidity to the virus structural antigens for each homologous or heterologous virus combination. The DENV1 clinical isolate induced high-avidity specific but low-avidity cross-reactive antibodies against YFV and JEV (Fig. 2, G to I). However, for JEV and YFV vaccine strains, both specific and cross-reactive antibodies generated were low avidity (Fig. 2, G to I). We next tested the capacity of serum from mice challenged with DENV1, YFV, or JEV to neutralize each virus and found that they were neutralizing against the primary challenge strain but not against the other related flaviviruses (Fig. 2, J to L). Thus, our mouse model results are consistent with the classification of DENV, JEV, and YFV into the same discrete serocomplexes as is observed in humans (8).

To measure cross-reactive cellular responses, peptide antigens were derived from infected cells and thus contain both structural and nonstructural antigens. Splenocytes showed robust proliferation when reexposed to antigens from the same viruses used to prime mice (Fig. 2M). Splenocytes from YFV- and JEV-primed mice also proliferated upon exposure to both DENV1 and JEV antigens (Fig. 2M). However, splenocytes from DENV1-infected mice only proliferated in response to JEV, but not YFV, antigens (Fig. 2M). Similarly, YFV, but not DENV1, antigen induced significant proliferation of JEV-primed splenocytes (Fig. 2M). These results support the idea that both serocomplex-specific and cross-reactive cellular immunity occur. Cumulatively, our mouse model recapitulates the cross-reactive immunity that characterizes human flavivirus infection, allowing further interrogation of the mechanisms that regulate immune priming or pathology during a secondary heterologous challenge.

The influence of cross-reactive immunity on subsequent flavivirus infection

To test whether preexisting cross-reactive immunity could reduce the viral burden during a secondary homologous or heterologous challenge, mice were challenged with DENV1, YFV, or JEV or given a vehicle control injection. To dissect the contributions of each adaptive immune branch to infection clearance, serum and T cells were collected and passively transferred to naïve mice. Recipient mice were infected subcutaneously with DENV1, 24 hours after transfer, followed by quantification of viral load in the draining lymph nodes (LNs) after another 24 hours, as subcutaneously injected DENV progresses to systemic infection via LNs in primates and mice (38, 39). As expected, adoptive transfer of serum or T cells from DENV1-immune mice or true secondary infection all led to a reduced viral burden upon DENV1 challenge, compared to transfer of those products from naïve (mock-immunized) animals (Fig. 2N), consistent with the belief that homologous secondary infections are efficiently cleared (28). However, transfer of serum or T cells from YFV-immunized mice or T cells from JEV-immunized mice did not influence the viral load (Fig. 2N). Mice given serum from JEV-immunized animals showed significantly enhanced viral burden in LNs after secondary heterologous DENV1 challenge (Fig. 2N), potentially attributable to ADE of DENV1 by antibodies generated during JEV immunization. In contrast, the viral load was significantly reduced in JEV-immune mice after a true secondary DENV1 infection (Fig. 2N). YFV immunity did not affect DENV1 clearance (Fig. 2N). The protective influence of JEV, but not YFV, immunity on DENV infection that was observed early at 1 day after infection (Fig. 2N) also persisted to later time points, as shown in LNs at 3 days after secondary cutaneous infection (fig. S1). Together, our data suggest that both humoral and cell-mediated cross-reactive immunity developed during a primary infection have the potential to be protective during secondary homologous or heterologous flavivirus infection by promoting viral clearance.

Cross-reactive antibodies can prime for early neutralizing immunity

Having observed that heterologous priming is protective during a secondary flavivirus challenge, we proceeded to evaluate how cross-reactive antibodies influence the development of adaptive responses during a secondary heterologous infection. Specifically, we questioned whether previous flavivirus infection might influence the production of neutralizing antibodies during memory recall. Mice were challenged with DENV1, YFV, JEV, or saline, and after 70 days, serum was collected to test its DENV1-neutralizing capacity. Mice were then rechallenged with DENV1, and serum was harvested after another 5 days to measure DENV1 neutralization. Sera obtained before secondary infection showed similar serotype-specific neutralization (Fig. 3, A and B), as in Fig. 2. Neutralization of DENV1 was improved after a secondary homologous DENV1 challenge (Fig. 3, C and D). JEV-immune mice also strongly neutralized DENV1 after a secondary DENV1 challenge at higher levels than DENV1-challenged mice that were previously flavivirus-naïve (saline control; Fig. 3, D and E). However, sera from mice that experienced a primary YFV challenge and secondary DENV1 challenge did not show improved neutralization of DENV1 (Fig. 3, C and E). These results suggest that JEV-induced cross-reactive immunity augments neutralizing antibody responses during a secondary flaviviral challenge.

Fig. 3 Serocomplex cross-reactive immunity boosts early serum neutralization after infection.

The DENV1-neutralizing capacity of serum antibodies raised in mice infected with (A) DENV1 or YFV or (B) DENV1 or JEV (or saline) was measured 70 days after primary challenge. Mice that were injected with (C) DENV1, YFV, or saline or (D) DENV1, JEV, or saline by intraperitoneal injection were reinfected 70 days after with DENV1 via intraperitoneal injection. (C and D) Neutralizing antibodies (day 75) were boosted by a secondary challenge with DENV1 in DENV1- and JEV-immune animals but not in YFV-immune animals. (E) The percentage neutralization at the highest dilutions from (A) to (D) was compared by analysis of variance (ANOVA) among all groups, for serum isolated after both primary challenge with DENV1, YFV, JEV, or PBS and secondary challenge with DENV1 (75 days). (F) Avidity of antibodies against DENV1 was measured in serum isolated 38 days after primary infection (10 days after secondary infection). Antibody avidity was significantly increased in mice experiencing a homologous secondary challenge of DENV1- and JEV-immune mice experiencing a heterologous challenge with DENV1. n = 5 per group. (G) DENV1 infection levels were measured in LNs 5 days following secondary DENV1 challenge by RT-PCR. n = 4 per group. *P < 0.05, **P < 0.01. Cross-reactive preexisting immunity to JEV enhances the neutralization and avidity of anti-DENV1 antibodies and coincides with reduced viral burden in vivo.

Next, we measured the avidity of antibodies generated against DENV1 in each of the primary immune experimental groups (saline, DENV1, JEV, and YFV), which were also given a secondary DENV1 challenge. Consistent with the results observed with the PRNT results, antibodies generated after a true homologous secondary infection with DENV1 had high avidity against DENV1 antigen (Fig. 3F). Similarly, antibodies generated in JEV-immune mice after a secondary DENV1 challenge showed significant improvement in their avidity against DENV1 antigen (Fig. 3F), while primary infection with YFV did not lead to improved avidity compared to the control group (Fig. 3F). At the same time point of 5 days after infection when the functionality of antibodies has improved (Fig. 3, D to F), protection is observed in terms of reduced DENV1 infection in the spleens of JEV-immune mice (Fig. 3G). JEV vaccination is able to prime a certain level of protection against an infection with DENV1.

JEV immunity primes for DC and T cell activation during DENV infection

Having shown that JEV can prime for functional protection against a subsequent infection with DENV1, we sought to understand the mechanism behind this. We hypothesized that JEV immunity may improve the mounting immune response during secondary DENV1 challenge by enhancing activation of dendritic cells (DCs), because DCs, as Fcγ receptor–bearing cells, could potentially bind to immune complexes formed during the heterologous secondary DENV challenge. To test this, mice were immunized with YFV, JEV, or saline, followed by a secondary challenge with DENV1 in the footpad skin. At 24 hours after challenge, DC activation was quantitated in draining LNs. The group with primary immunity to YFV showed similar levels of DC activation after DENV1 infection, compared to the saline control group, but there was a marked increase in the number of activated DCs in draining LNs after secondary DENV1 challenge in JEV-immune mice (Fig. 4A and fig. S2, A and B).

Fig. 4 Priming of DC activation and TMEM cell recall by serocomplex cross-reactive immunity.

Mice were injected subcutaneously with JEV, YFV, or saline and then rechallenged with DENV1 on day 21. LNs were isolated 24 hours after infection (n = 6). (A) Activated DCs were enumerated by flow cytometry after staining for CD11c, CD80, and CD86, showing increased numbers of activated DCs in the LNs of JEV-immune, but not YFV-immune, mice compared to naïve controls. Representative histograms of costimulatory molecule expression are shown in fig. S2 (A and B). (B) Gating strategy to define T cell subsets and their activation after staining against CD3, CD4, CD8, CD44, CD69, and CD62L. (C) The graph represents the ratio of activated TN cells (CD62L+CD44CD69+) to activated TEM cells (CD62LloCD44+CD69+) for CD4+ and CD8+ T cells. Higher proportions of activated CD4+ and CD8+ TEM cells in JEV post-immune mice and CD8+ TEM cells in YFV post-immune mice were detected. Total CD4+ and CD8+ (D) Teff, (E) TCM, and (F) TEM cells in LNs are presented. Significantly increased LN TEM cells staining intracellularly for the cytokines (G) IFN-γ and (H) IL-2 were observed in JEV- and DENV1-immune animals. (I) Donor Th1.2+ T cells from DENV1, JEV, and YFV post-immune mice were transferred 5 weeks after primary challenge to recipient Thy1.1+ mice. Recipient mice were given a secondary DENV1 challenge, and after 5 days, draining LNs were isolated. Donor Tfh cells with a memory phenotype (Thy1.2+CD4+CD44+CD62−/loPD-1+CXCR5+BCL6+) were higher in the LNs of DENV1- and JEV-immune animals. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. When trending toward significance (P < 0.1), P values are shown on the graph. Gating strategies for (G) to (I) are presented in fig. S2 (D and E). Increased TEM activation and decreased TN and TCM activation are associated with early DENV1-immune responses in animals with JEV preexisting immunity, and donor cross-reactive TEM cells adopt a Tfh phenotype in LNs during a secondary DENV1 challenge.

To better understand the role of T cells during DENV infection of JEV-immune mice, we examined the activation of various T cell populations in the draining LN. A flow cytometry gating strategy was used, where CD4+ and CD8+ T cell populations were further defined as naïve (TN; CD44CD62L+), effector (Teff; CD44CD62L), effector memory (TEM; CD44+CD62L), and central memory (TCM; CD44+CD62L+) T cells (Fig. 4B). CD69 was used to assess activation. We observed that the ratio of activated CD8+ and CD4+ TN cells to TEM cells was significantly reduced in the JEV-immunized group during a secondary DENV1 challenge (Fig. 4C). Although no change was observed in the YFV-immunized group in terms of the relative numbers of activated CD4+ TEM to TN cells, activated CD8+ TEM cells were significantly increased relative to TN cells upon secondary DENV1 challenge (Fig. 4C), demonstrating that the memory recall response favored the activation of TEM over TN cells. Similarly, the total TN CD4+ and CD8+ TN cells remained unchanged for the YFV-immunized group compared to the saline control group during DENV1 infection (fig. S2C). Total CD8+ TN cells in the JEV-immunized group did not differ compared to the saline control group; however, CD4+ TN cells were significantly reduced upon secondary DENV1 challenge (fig. S2C). Both YFV and JEV increased the numbers of Teff cells in LNs (Fig. 4D), but TMEM cells were differentially influenced during secondary DENV1 challenge (Fig. 4, E and F). Specifically, activated CD4+ and CD8+ TCM cells were also significantly reduced in both YFV- and JEV-immunized groups compared to the control group (Fig. 4E). In contrast, total numbers of activated CD4+ and CD8+ TEM cells were only significantly increased in the JEV-immunized group but not in the YFV-immunized group (Fig. 4F). To verify functional activation after a secondary heterologous challenge, we intracellularly stained TEM cells for the cytokines interleukin-2 (IL-2) and interferon-γ (IFN-γ), both of which were expressed at higher levels by TEM cells of DENV1- or JEV-immune animals after a secondary DENV1 challenge (Fig. 4, G and H). Thus, T cell responses to DENV1 infection are modulated by previous immunity to JEV and YFV to differing extents. Previous JEV exposure can prime for heightened DC activation and also improve TEM cell function during a subsequent DENV1 challenge, while neither of these indicators of effective immune activation was influenced by previous YFV exposure.

Presumably, TMEM cells are recalled and enter LN follicles during challenge by a heterologous serocomplex. To address this hypothesis, we adoptively transferred T cells from Thy1.2+ donor mice, which were isolated 5 weeks after flavivirus challenge (from either DENV1-, JEV-, or YFV-immune mice) into Thy1.1+ recipient mice. The recipient mice were then challenged with DENV1 subcutaneously, and the draining LNs were harvested 6 days after infection. LN cells were stained for donor T follicular helper (Tfh) cells with a memory phenotype (Thy1.2+CD4+CD44+CD62−/loPD-1+CXCR5+BCL6+). As expected, Tfh cells with a memory phenotype were more abundant in LNs of DENV1- and JEV-immune mice compared to naïve mice (Fig. 4I), indicating a genuine recall of flavivirus-experienced TMEM cells to participate in the germinal center reaction as Tfh cells.

Identification of DENV-specific and flavivirus cross-reactive CD4 epitopes

A plausible explanation for the increased cross-protection of JEV compared to YFV toward DENV could be its higher degree of genome similarity (Fig. 1D). To explore this possibility and further define how the genetic relationships between flaviviruses could influence TMEM activation, we examined the predicted CD4 epitopes along the DENV full-length polyprotein. Bioinformatic approaches identified 17 regions of the DENV polyprotein significantly likely to be presented on mouse major histocompatibility complex class II (MHC-II) (I-Ab; Fig. 5A and table S1): one in capsid, one in NS1, two in NS2b, seven in NS3, two in NS4b, and four in NS5. For these DENV1 peptides, we investigated whether the same region of JEV or YFV was also significantly likely to be presented on MHC-II. This analysis showed that most of the peptides were exclusive to DENV, with probability scores falling below cutoff levels, and only one sequence was predicted by this method to bind to MHC-II for all three viruses (Fig. 5A). Because of its higher degree of homology, JEV shared more sequences that were predicted to bind MHC-II with DENV than YFV, approximately 41% of predicted peptides for JEV versus 12% for YFV (Fig. 5A). To validate the predicted DENV1 CD4 epitopes and to assess whether these peptides could induce proliferation of CD4+ TMEM cells (CD3+CD4+CD44+) from JEV- or YFV-immune animals, we performed a proliferation assay. DENV peptides were chosen from those regions that were also predicted to be presented during JEV infection exclusively (but not YFV infection; Fig. 5B), YFV infection exclusively (but not JEV infection; Fig. 5C), or both infections (Fig. 5D). Alignments of these DENV peptides with the corresponding JEV and YFV sequences reveal varying degrees of homology and high levels of homology with ZIKV (Fig. 5, B to D). The selected peptides were used to pulse antigen-presenting cells (APCs), which were subsequently exposed to splenocytes from DENV1-immune animals. After incubation, the cells were stained for flow cytometry analysis of the frequency of CD4+ TMEM cells to identify any stimulatory effect on this subset. Of the six DENV1 peptides chosen for validation, five induced expansion of TMEM cells from DENV1-immune animals (Fig. 5E). Surprisingly, the one peptide that was predicted to bind MHC-II for DENV and also for the homologous regions in JEV and YFV (Fig. 5D) did not cause TMEM cell expansion in splenocytes from animals immune to DENV1, JEV, or YFV, suggesting that it may not be a true epitope that leads to CD4+ memory formation (Fig. 5E). Of the peptides corresponding to regions of JEV that were also predicted to bind to MHC-II, 80% also induced proliferation of TMEM cells from JEV-immune animals, demonstrating cross-priming of JEV-specific memory cells by DENV peptides (Fig. 5E). One of these corresponded to a region of YFV that was not significantly predicted to bind MHC-II (although approaching significance), but it shared a higher than average degree of homology to DENV and the other flaviviruses. This peptide also induced enrichment of TMEM cells and appeared to be a region broadly capable of cross-priming (Fig. 5E). Surprisingly, a DENV1 peptide that, similarly, corresponded to a YFV region predicted to bind to MHC-II, where the analogous region of JEV did not (Fig. 5C), induced expansion of JEV, but not YFV, post-immune TMEM cells (Fig. 5E). Proliferation was confirmed by staining for Ki-67 (fig. S4, A and B). Intracellular staining for IFN-γ and an IFN-γ enzyme-linked immunospot (ELISPOT) assay confirmed functional activation of specific and cross-reactive T cells by these peptides (fig. S4, C to G). Therefore, we have identified epitopes that are cross-reactive and allow TMEM recall from a pool of flavivirus serocomplex cross-reactive memory cells.

Fig. 5 Identification and validation of flavivirus cross-reactive CD4 epitopes.

Of the 17 regions of the DENV1 full-length polypeptide that were predicted MHC-II binders, we identified the homologous sequences in the JEV and YFV polypeptides. (A) The chart depicts the proportions of the DENV1 peptides for which the homologous regions of JEV and YFV either were not predicted to bind to MHC-II (DENV only) or were also predicted binders for JEV, YFV, or both (JEV and YFV). DENV1 peptides were selected for validation from those where homologous regions were predicted MHC-II binders for (B) JEV, (C) YFV, or (D) both JEV and YFV. For (B) and (C), DENV1 peptides are shown aligned to the corresponding regions of YFV, JEV, and ZIKV. Numbers indicate the amino acid position in the polyprotein, and its protein location is labeled blue. (E) Heat map representation of TMEM expansion after stimulation with synthetic peptide-pulsed APCs, relative to the no peptide control. The percentage of CD3+CD4+CD44+ of total T cells (CD3+) was calculated after analysis by flow cytometry. Raw data and bar graph representation are provided in fig. S3, and proliferation and cytokine production were validated in fig. S4.

Serocomplex cross-reactive TMEM cells are present in humans

To validate whether flavivirus-immune humans also have serocomplex cross-reactive CD4 TMEM cells that can be reactivated by a new flavivirus, we recruited donors with a known history of JEV vaccination. PRNT assays were used to confirm immunity against DENV, JEV, YFV, or ZIKV (Fig. 6, A to D). ZIKV was added to the study because the donors were recruited in Singapore, which has a high prevalence of natural DENV immunity that could limit our ability to focus on DENV cross-reactive responses in JEV-immune donors. Two donors appeared to have neutralizing antibodies against both JEV and DENV (Fig. 6, B and C), while one had neutralizing antibodies only to JEV (Fig. 6D). Similar to mice (Fig. 2), cross-reactive antibodies were generated, because all JEV-immune donors showed cross-reactive immunity to ZIKV (Fig. 6, E and F). However, those antibodies did not detect more distantly related YFV (Fig. 6, E and F). Also, as observed in mice, the cross-reactive avidity was weaker (here, against ZIKV) compared to the specific avidity (against JEV; Fig. 6G). To validate that TMEM cells could be cross-activated by antigens from differing serocomplexes, donor peripheral blood mononuclear cells (PBMCs) were stimulated with viral antigens. By flow cytometry, we observed increased activation of the human CD4 TEM (CD3+CD4+CD45RO+CCR7CD62L+) population, indicated by up-regulation of the activation marker human leukocyte antigen (HLA)–DR after incubation with JEV antigens for cells from JEV-immune donors but not from the naïve control (Fig. 6, I to K). Furthermore, both YFV and ZIKV antigens induced cross-activation of TEM cells (Fig. 6, J to L). ELISPOT assays were used to validate activation by measuring IFN-γ production after exposure of T cells to antigens from heterologous challenges (Fig. 6, L to O). This confirmed a functional cross-activation because JEV-immune patients produced IFN-γ in response to heterologous challenges of ZIKV (Fig. 6, M to O), and for one patient, IFN-γ was also produced in response to YFV (Fig. 6M), to which they had not been exposed. Therefore, cross-activation of TMEM cells from flavivirus-immune humans can occur upon exposure to a virus from a differing serocomplex.

Fig. 6 Confirmation of serocomplex cross-reactive T cell responses in flavivirus-immune humans.

PRNT assays were used to determine immunity to DENV1, JEV, YFV, or ZIKV in (A) presumed naïve or (B to D) JEV-immunized individual donors. (E and F) Specific and cross-reactive IgG titers against each virus were measured by ELISA. Average titers for JEV-immune donors are shown in (E), and individual titers are shown in (F). N, not detected. (G) Avidity of antibodies for JEV was high in JEV-immune donors and lower for cross-reactive antibodies. PBMCs from (H) flavivirus-naïve or (I to K) JEV-immunized donors were incubated with either JEV, DENV1, ZIKV, YFV, or control antigens for 3 days, followed by flow cytometry to identify the CD4+ TEM population. Activation of TEM cells from JEV-immune donors to homologous and heterologous flaviviruses was observed based on up-regulated staining for HLA-DR. Percentages of HLA-DR+CD4+ TEM cells of total T cells are included on the histograms. The flow cytometry gating strategy is presented in fig. S5. The frequency of spot-forming cells (SFC) after stimulation with flavivirus-derived or control antigens was determined for (L) flavivirus-naïve and (M to O) JEV-immunized donors by IFN-γ ELISPOT.


Our studies demonstrate the key contributions of serocomplex cross-reactive antibodies and T cells toward priming for enhanced immunological memory during a secondary flavivirus challenge. In the context of the substantial number of flavivirus-immune individuals and the cautious approach to vaccination that is taken in response to the possibility of cross-reactive immune pathology (11), this work provides a detailed understanding of how immunity influences a subsequent immunization or infection by a new flavivirus. Studies examining functional responses during a secondary heterologous challenge frequently assume that antibodies are the most consequential immune determinant for mediating cross-reactive protection, when it occurs (40, 41). However, we observed antibody-mediated enhancement of virus in a secondary target organ, the LN, when serum alone was transferred to recipient mice before a secondary challenge with a flavivirus from a distinct serocomplex (here, JEV followed by DENV infection; Fig. 2N). This is consistent with the concept of ADE. We reported that sequential immunization of humans with inactivated JEV vaccine followed by live-attenuated YFV vaccine resulted in higher YFV viremia than controls with no previous JEV vaccination (30). That enhancement of DENV1 infection was observed when mice were transferred serum from JEV-immune, but not YFV-immune, animals (Fig. 2N) indicates that the degree of genome similarity (Fig. 1D) might play a role in the potential of ADE. In contrast, JEV-immune mice that experienced a true primary infection, followed by a secondary infection with DENV1, had improved viral clearance in LNs (Fig. 2N). This observation cautions against assuming that T cell responses are dispensable for cross-protection against a secondary flavivirus infection. Rather, we observed that optimal memory recall relies on a combination of antibodies and T cells.

Our data suggest that increased DC presentation due to cross-reactive antibodies, combined with effective cross-reactive responses in the CD4 memory effector compartment, enhances protection observed with secondary challenge of DENV1, although YFV immunization did not result in a significant increase in DC activation in the draining LNs during secondary DENV1 challenge compared to the control group (Fig. 4A). That there was more efficient TEM recall over TN activation for CD8+ T cells from YFV-immune animals suggests some degree of cross-reactive T cell immunity in vivo in mice (Fig. 4, C and F). However, this cross-reactive response was not sufficient to protect significantly during DENV1 challenge (Figs. 2N and 3G and fig. S1). In humans, a heterologous CD8 T cell response also characterizes reinfection within the DENV serocomplex and, in one study, did not appear to be associated with more severe disease (37). Our results are consistent with that observation because cross-reactive CD8 T cells were not harmful in our model. We also observed that splenocytes from YFV-immunized mice proliferated in response to DENV1 antigen (Fig. 2M), despite not observing a cross-reactive functional protection against DENV1 challenge (Figs. 2N and 3G and fig. S1). In contrast, JEV-immunized mice displayed significant gains in DC activation and improved efficiency of the memory recall response because the ratio of activated TN/TEM cells decreased in JEV-immune mice compared to control mice for both CD4+ and CD8+ T cells (Fig. 4C). Furthermore, adoptive transfer experiments showed that, following a DENV challenge, TMEM cells from JEV-immune donors adopt a Tfh cell phenotype similar to T cells from DENV-immune donors (Fig. 4I), supporting both memory recall and participation in the germinal center reaction.

The in vivo evidence of TEM recall in JEV-immune animals with a challenge by DENV1 led us to identify and validate several predicted CD4+ epitopes for DENV1. Predicted epitopes were identified using bioinformatics, followed by validation in proliferation assays, where flow cytometry was used to verify the specific expansion of CD4 TMEM cells. As predicted, JEV and DENV shared five functionally cross-reactive epitopes, while all three viruses (DENV1, JEV, and YFV) shared only one highly cross-reactive epitope (Fig. 5E). This likely explains why JEV-immune mice show enhanced clearance of DENV1 (Fig. 2N) and substantial gains in antibody avidity and neutralization against secondary DENV1 (Fig. 3, D to F). While it is also likely that differences in antigen persistence between JEV and YFV could influence the quality of cross-reactive responses, we note that each infection induced high-titer neutralizing responses to itself (Fig. 2, A to C) and significant activation of homologous T cells (Fig. 2M), indicating efficient memory induction. This supports the findings that the differences between YFV and JEV in terms of T cell epitopes shared with DENV1 were due to a genuine lack of cross-activation of YFV-immune cells by most DENV1 epitopes. Because CD4+ T cells are key for providing B cell help and refining antibody responses, the unique capacity of JEV to induce heightened numbers of activated CD4+ TEM cells (Fig. 4F) could be responsible for the enhanced DENV1-binding capacity of antibodies generated immediately following a secondary DENV1 infection. Further studies are also needed to examine how CD4 TEM cells influence CD8 T cell responses because T cell help could also be provided to CD8 cells. We validated that TEM cells from JEV- or DENV/JEV-immune donors are cross-activated by related flaviviruses ZIKV and YFV (Fig. 6), although because humans have highly varied MHC-II molecules, the cross-presented epitopes are likely to differ from donor to donor. In our mouse challenges, we also observed that the attenuated JEV and YFV vaccine strains induced more cross-reactive antibodies in mice than the virulent DENV1 clinical isolate. This could potentially be attributed to mutations that might have been acquired on the surface glycoproteins of YFV and JEV vaccine strains in the process of attenuation (42). To our knowledge, that vaccination induces weaker avidity antibodies than a virulent infection has not been previously suggested and should be further investigated in humans.

Our findings are also informative in the context of the aim to develop and adopt a broadly effective DENV vaccine. Although multiple strategies have been proposed, such as virus-like particles, attenuated strains, and chimeric viruses (43, 44), most have aimed to develop a tetravalent vaccine to generate neutralizing antibodies to all DENV serotypes concurrently. The approved Sanofi Pasteur vaccine is based on using the YFV backbone to display DENV structural proteins. The rationale for using YFV as the backbone for a chimeric DENV vaccine was the genome stability of YF-17D (45, 46). YF-17D, itself, also has nearly unmatched efficacy as a vaccine, being capable of conferring recipients with lifelong immunity to YFV (47). In part, this high degree of efficacy is attributed to the ability of YF-17D to generate a robust specific CD4+ T cell response (48). However, questions remain regarding whether the YFV-directed cellular response prompted by YF-17D is efficient in priming DENV immunity in humans, because efficacy of the chimeric vaccine was limited in clinical trials, with protection from DENV between 30 and ~65% in the results published to date (49, 50). Our data show that preexisting immunity to JEV is more effective than YFV immunity in priming recall of TEM cells (Fig. 4F). That this translates into an increased efficiency of priming for DENV-specific antibodies (Fig. 3) and protection against DENV challenge (Fig. 2N) raises the potential that a JEV vaccine strain might be a more effective backbone for a chimeric DENV vaccine than YFV. Alternatively, validated cross-reactive CD4 epitopes could be added to an otherwise successful vaccine backbone such as YF-17D to generate broad flavivirus cross-reactive responses. Although humans have differing and more diverse CD4 epitopes than in-bred mice, in both cases, the degree of genome similarity between DENV and JEV would make conservation of epitopes more likely between these two. Given the importance of nonstructural proteins to generating T cell responses, we expect that live-attenuated vaccines rather than inactivated or subunit vaccines would be more likely to generate functionally protective cross-reactive T cell responses. Furthermore, although our work suggests that the degree of virus similarity is important for cross-protection by increasing the likelihood that memory CD4 T cell epitopes would be similar between the primary and secondary challenge strains, we cannot fully predict how all combinations of unique virus strains might behave in vivo.

Here, we show that mice evoke similar immune responses to flaviviruses as humans in terms of generating immunity consistent with human serocomplexes, where cross-reactive antibodies are nonneutralizing and have low avidity. Although the use of IFN-deficient mouse models are generally favored by the flavivirus field due to the viruses reaching high serum titers and the mice experiencing severe disease, they may be less ideal for studying immunity. IFN-γ has a central role in T cell polarization (51). Type I IFNs are key for function of plasmacytoid DCs and generating antibody responses (52, 53). Both type I and II IFNs promote antibody class switching and have an adjuvant effect during infection (53, 54). Studies have emphasized the importance of antibody subclasses to DENV functional immune responses (5557), making this an important justification for using an IFN-sufficient system to study DENV immunity. On the basis of these considerations, we have used an immunocompetent mouse model to investigate cross-reactive immunity among flaviviruses and have identified TMEM responses as a key component of cross-protection. By validating that serocomplex cross-reactive T cells are present in humans (Fig. 6), we have an indication that the mouse and human are similar in terms of displaying a potential for flavivirus cross-protective CD4 T cell responses.

There are major obstacles to improvement, development, validation, safety testing, production, and compliant use of vaccines against the many established flaviviral pathogens that cluster to discrete serocomplexes, including DENV, YFV, JEV, and WNV, as well as newly emerging pathogens, such as ZIKV. Approaches to rational design of vaccines that provide cross-protection through introduction of cross-reactive CD4+ epitopes may lessen the risk of infection in naïve individuals, in unvaccinated individuals, or to emerging pathogens from this genus. This study identifies novel correlates of protection based on the combined responses of TMEM cells and antibodies during a secondary heterologous flavivirus challenge. The results emphasize that cross-reactive flavivirus immunity is likely to prime for effective secondary responses but that cross-reactive CD4+ T cell responses are key for optimal enhanced secondary immunity to a heterologous infection. These results have implications for rational vaccine design for existing and emerging flaviviral pathogens.


Mouse infections and immunizations

C57B/6NTac mice were obtained from InVivos. B6.PL-Thy1a/CyJ mice (Thy1.1+) were obtained from The Jackson Laboratory. For intraperitoneal infections, mice were inoculated with 1 × 106 PFU of DENV1, JEV, or YFV in 100 μl of phosphate-buffered saline (PBS). For subcutaneous infections, mice were injected with 1 × 105 PFU in a 20-μl volume of PBS. Animals were bred and housed in the Duke–National University of Singapore (NUS) vivarium. The SingHealth Institutional Animal Care and Use Committee approved all animal protocols.

Human JEV-immune donors

Human donors that either were suspected flavivirus-naïve or had a known JEV immunization history were recruited as blood donors according to protocols approved by the Centralized Institutional Review Board of Singapore General Hospital.

Virus production

DENV1 (D1/SG/05K2402DK1), JEV (SA14-14), and ZIKV (H/PF/2013) were produced in C6/36 mosquito cells in RPMI 1640 with 2% fetal bovine serum (FBS) and 1% penicillin and streptomycin at 28°C and harvested 5, 3, and 4 days after infection, respectively. YFV (YFV-17D) was produced in Vero cells in Dulbecco’s modified Eagle’s medium with 2% FBS and 1% penicillin and streptomycin at 37°C, with 5% CO2, and harvested 3 days after infection. Viral titers were quantified by a standard plaque assay using BHK21 cell monolayers, as described previously (39).

Evaluation of serum antibody responses

To quantify the virus-specific or cross-reactive antibodies in serum, DENV1, YFV, JEV, or ZIKV were purified for use as ELISA capture antigens. In brief, supernatants of infected cells were clarified by centrifugation at 2500 rpm for 10 min and filtered using a 45-μm filter unit. The virus supernatants were then concentrated by ultracentrifugation (20,000 rpm for 2.5 hours). The pellets were resuspended in PBS, left for 2 hours on ice, and then applied to a 30% sucrose cushion and centrifuged at 25,000 rpm for 4 hours. After removing the supernatants, virus pellets were dried a few minutes before overnight rehydration in PBS. The resuspended viruses were stored at −80°C. Concentrations of viral proteins were measured using a Bradford assay (Bio-Rad) before coating of ELISA plates at a concentration of 5 μg/ml. Serum end-point titers against the various viruses were measured with a standard ELISA, using alkaline phosphatase–conjugated anti-mouse or anti-human detection antibodies, based on published protocols (58, 59). The end-point titer was calculated at twofold over naïve serum. For avidity ELISAs, washing was performed with either normal wash buffer or wash buffer containing urea (Sigma) at a 5 M concentration. The percentage of antibody that remained bound after stringent washing was calculated for each sample compared to normal wash conditions.

For PRNT assays, 2 × 105 BHK21 cells per well were seeded in 24-well plates with RPMI 1640 supplemented with 10% FBS and 1% penicillin and streptomycin. Cells were incubated until confluent at 37°C with 5% CO2. Sera from infected mice were heat-inactivated for 30 min at 56°C and serially diluted from 1:10 to 1:2560. Each dilution or control was mixed with virus (750 PFU/ml) in a ratio of 1:1 and incubated for 1 hour. Solutions of virus and serum (200 μl) were then added to the plates in duplicate and incubated for 1.5 hours at 37°C with 5% CO2. The plates were washed with PBS once, and 500 μl of carboxymethylcellulose overlay was added to each well. After 5 days for DENV1 and ZIKV or 3 days for YFV and JEV, the plates were fixed in paraformaldehyde (PFA) before staining with crystal violet. The plaques were counted, and the percentage of neutralization was calculated relative to the untreated virus control.

Flow cytometry to assess DC and T cell activation

To quantify activated DCs and T cell subsets after homologous or heterologous secondary infection, mice were injected subcutaneously with 1 × 105 PFU of JEV, YFV, or saline control, and after 21 days, all groups were rechallenged with 1 × 105 PFU of DENV1 subcutaneously in footpads. Draining popliteal LNs were isolated after an additional 24 hours (n = 6 per group), minced, and incubated in medium containing collagenase (Sigma). Single-cell suspensions were prepared using 70-μm nylon cell strainers (Falcon). Cells were blocked with 1% bovine serum albumin (BSA) in PBS (Gibco). To quantify DCs, cells were stained with anti-CD11c–Pacific Blue (Thermo Scientific), anti-CD80–BV605 (BD Biosciences), and anti-CD86–BV510 (BD Biosciences). To quantify T cells, cells were stained with anti-CD3e–PerCP (peridinin chlorophyll protein) Cy5.5, anti-CD4–BV650, anti-CD8–AF700, anti-CD44–BV510, anti-CD62L–PE (phycoerythrin)–Cy7, and anti-CD69–FITC (fluorescein isothiocyanate) (all from BD Biosciences). After washing, samples were fixed with 3.7% PFA and analyzed using a Fortessa flow cytometer (BD Biosciences) and FlowJo software. For some experiments, intracellular staining was performed subsequent to fixation. After PFA treatment, the cells were washed once with PBS–1% BSA and resuspended in PBS containing 1% BSA and 0.1% saponin for 15 min, followed by staining with anti-mouse IFN-γ–BV711 antibody (BD Horizon, reference 564336) and/or anti–IL-2–BV421 (BD Biosciences) in PBS with 1% BSA and 0.1% saponin. After washing, data were acquired by flow cytometry, as above.

T cell adoptive transfers

C57Bl/6 mice (Thy1.2+) were challenged with DENV1, JEV, and YFV (five mice per group) by intraperitoneal injection with 1 × 106 PFU or injected with saline. Five weeks after infection, the spleens were collected and the total T cells were purified by magnetic column using the Pan T Cell Isolation Kit II (mouse; MACS Miltenyi Biotec). C57Bl/6 mice (Thy1.1+) received an adoptive transfer by tail vein injection of 1 × 106 total T cells per mouse, with either T cells purified from mice exposed to DENV1, YFV, JEV, or saline. The mice were challenged with DENV1 24 hours after transfer, and after 6 days, popliteal LNs were collected and prepared as single-cell suspensions for flow cytometry. The cells were stained using the following primary conjugated antibodies: anti-Thy1.2–PerCP Cy5.5 (BioLegend), anti-CD4–BV650, anti-CD8–AF700, anti-CD44–BV510, anti-CD62L–PE-Cy7, CD69-FITC (all from BD Biosciences), anti-Bcl6–AF647 (BioLegend), anti–PD-1–BV711 (BioLegend), and anti-CXCR5–BV421 (BioLegend).

Proliferation assays

Mice (four groups of n = 5) were intraperitoneally injected with 1 × 106 PFU of DENV1, YFV, JEV, or the same volume of PBS. Spleens were collected from mice 5 weeks after infection, injected with 500 μl of (10 mg/ml) collagenase in RPMI 1640, and incubated for 30 min at 37°C. Single-cell suspensions were then made using a 70-μm cell strainer and an excess volume of PBS. Red blood cell lysis buffer [0.15 M NH4Cl, 10 mM NaHCO3, and 0.1 mM EDTA (pH 7.3)] was then used to treat the isolated cells, followed by washing and adjustment to a concentration of 1 × 105 cells/ml. JAWSII cells were preseeded 24 hours before the experiment at 5 × 103 cells per well in α-minimum essential medium, with 20% FBS, 1% penicillin and streptomycin, and granulocyte-macrophage colony-stimulating factor (GM-CSF) (5 ng/ml; Sigma). Antigens for DENV1, JEV, or YFV were prepared according to a published protocol (60). In brief, Vero cells were infected with DENV1, YFV, or JEV and harvested with a cell scraper when showing 50% cytopathic effect. Uninfected Vero cells were used to generate control antigens. Cell pellets were extensively washed before and after fixing with 0.025% glutaraldehyde. The suspensions were sonicated and clarified, and the antigens were stored at −20°C until use. Antigens (1.5 μg/well) were preincubated with JAWSII cells for 48 hours and then treated with mitomycin C (25 μg/ml) for 25 min, followed by washing to remove excess mitomycin C. Alternatively, synthetic peptides (synthesized by SABio; 2.5 μg/well) were preincubated with JAWSII cells. Purified splenocytes (100 μl of 1 × 105 cells/ml) were added to each well without pooling of cells so that each replicate represented the results from an individual mouse. After 72 hours, proliferation was measured using CellTiter 96 AQueous One Solution (Promega), according to the manufacturer’s instructions. For peptide proliferation assays, cells were prepared for flow cytometry analysis of TMEM cells with a staining and analysis strategy as described above, with the addition of the antibody anti–Ki-67–BUV395 (BD Biosciences).

Mouse ELISPOT assays

Spleens were harvested from mice injected with saline or DENV1, YFV, or JEV 5 weeks after challenge and treated with collagenase to obtain single-cell suspensions. One million splenocytes per well were seeded in a 96-well plate and stimulated with synthetic peptides AAIFMTATPPGSVEA or ASAWTLYAVATTIIT (25 μg/ml) for 48 hours at 37°C with 5% CO2. The peptide-stimulated or control splenocytes were then transferred to 96-well IFN-γ ELISPOT plate (BD Biosciences, catalog no. 552569) and incubated for an additional 17 hours before detection following the manufacturer’s instructions. Spot-forming cells were enumerated by counting.

Quantification of DENV1 infection in LNs and spleen

Mice were infected with DENV1 28 days after a primary infection with DENV1, YFV, JEV, or mock infection, or 24 hours after adoptive transfer of either antibodies or T cells purified from DENV1, YFV, or JEV post-immune mice. For adoptive transfer experiments, groups of five mice were intraperitoneally injected with PBS control or 1 × 106 PFU of DENV1, YFV, or JEV. For serum isolation, blood was collected by intracardiac puncture 28 days after infection. Sera from mice in the same primary infection group were pooled. Recipient mice (n = 5 per group) received 100 μl of serum. Alternatively, spleens were collected 5 weeks after infection of groups, as described above, and the total T cells were extracted using the Pan T cell Isolation Kit and LS magnetic columns (both from Miltenyi), according to the manufacturer’s instructions. For each group (n = 5), 1 × 106 T cells in PBS were given by tail vein injection into naïve mice. For all groups of mice infected via the subcutaneous route, the infection challenge was established by footpad injection of 1 × 105 of DENV1. After 24 hours, the draining popliteal LNs were collected, and RNA was extracted using the RNeasy Mini Kit (Qiagen). The iScript cDNA Synthesis Kit (Bio-Rad) was used according to the manufacturer’s instructions with an added DENV1 primer (5′-CTGAGTGAATTCTCTCTACTGAACC-3′) to ensure conversion of viral RNA to complementary DNA (cDNA). RT-PCR was performed using SYBR Green reagent (Bio-Rad) and validated DENV1-specific primers (forward, 5′-CAAAAGGAAGTCGTGCAATA-3′; reverse, 5′-CTGAGTGAATTCTCTCTACTGAACC-3′) (61). Alternatively, groups of DENV1, YFV, or JEV-immune mice or controls were infected intraperitoneally with 1 × 106 PFU of DENV1. Spleens were harvested 5 days after infection for RNA isolation. RNA was extracted using TRIzol reagent (Invitrogen). One-step quantitative PCR (qPCR) was performed using the SuperScript III Platinum One-Step qRT-PCR Kit (Invitrogen), as recommend by the manufacturer. The primers and the probe from the SinglePlex and MultiPlex RT-PCR for Detection and Serotype Identification of Dengue Virus [Centers for Disease Control and Prevention (CDC)] were used. A plasmid containing the DENV1 region of interest was used to generate a standard curve to quantify genome copies, and cDNA from uninfected mouse tissue was used as a negative control.

Generation flavivirus distribution maps

Current distributions of flaviviral pathogens were obtained from multiple recent sources including current literature (6265), the World Health Organization, and CDC. The Google GeoChart tool and Adobe Illustrator software were used to generate the maps.

Phylogenetic analysis

For phylogenetic analysis, full-length polyprotein amino acid sequences from DENV1, DENV2, DENV3, DENV4, JEV, ZIKV, WNV, Spondweni virus, and YFV were aligned using Clustal Omega or Molecular Evolutionary Genetic Analysis (MEGA) software. The phylogenetic tree was constructed by using the maximum likelihood method based on the Jones-Taylor-Thornton (JTT) matrix-based model. The tree with the highest log likelihood is shown. Initial tree(s) for the heuristic search was obtained automatically by applying Neighbor-Join and BioNJ algorithms to a matrix of pairwise distances estimated using a JTT model and then selecting the topology with superior log likelihood value. The bootstrap test of phylogeny with 1000 repeats was performed, and the tree was drawn to scale, with branch lengths measured in the number of substitutions per site. Evolutionary analyses were conducted using MEGA7 software (

MHC-II binding predictions

The MHC-II binding predictions were made for the MHC-II I-Ab locus on 1/25/2016 using the Immune Epitope Database Analysis Resource Consensus tool (66, 67) using the following protein sequences: DENV1 (GenBank: ABW82066.1), YFV (GenBank: AGO04419.1), and JEV (GenBank: BAA14219.1). Homologous protein sequences were identified by alignment using ClustalX 2.1 software. From the prediction, we selected DENV1 peptides with a percentile rank lower than 5 as predicted MHC-II binding peptides. Six peptides of 15 amino acids length were selected for validation based on their characteristic of having significant likelihood of MHC-II presentation for one or more of the flaviviruses used in this study: ASAWTLYAVATTIIT, RSGVLWDTPSPPEVE, AAIFMTATPPGSVEA, YKTWAYHGSYEVKPS, VILAGPMPVTVASAA, and AFLRFLAIPPTAGIL.

Human PBMC stimulation

PBMCs were isolated from the blood of donors that were either JEV-vaccinated or JEV-naïve and were stored at −80°C until use. After thawing, PBMCs were seeded (1 million cells per well in a 96-well plate) and stimulated for 72 hours with DENV1, YFV, ZIKV, or control antigens (50 μg/ml). The antigens were derived from cell culture to ensure the presence of nonstructural epitopes, as described above and according to published protocols (60). Before extracellular staining for surface markers, cells were treated with Fixable Viability Stain 510 (BD Biosciences, catalog no. 564406) according to the manufacturer’s instructions to indicate Live/Dead cells. Extracellular and intracellular staining of the PBMCs was performed, as previously described for staining of mouse cells, using the following antibodies: mouse anti-human CD3 PerCP-Cy5.5 (BD Pharmingen, catalog no. 560835), mouse anti-human CD4 BUV395 (BD Biosciences, catalog no. 742738), mouse anti-human CD8 BV650 (BD Biosciences, catalog no. 740587), mouse anti-human CD27 BV786 (BD Biosciences, catalog no. 740972), mouse anti-human CD62L BV711 (BD Biosciences, catalog no. 740783), mouse anti-human CD38 BV605 (BD Biosciences, catalog no. 740401), mouse anti-human CD45RO PE-Cy7 (BD Biosciences, catalog no. 560608), mouse anti-human HLA-DR APC-H7 (BD Biosciences, catalog no. 561358), and rat anti-human CD197 (CCR7) BV421 (BD Biosciences, catalog no. 740052). Data were acquired using a Fortessa flow cytometer (BD Biosciences) and analyzed using FlowJo software.

Human ELISPOT assay

PBMCs from either JEV-vaccinated or JEV-naïve donors were seeded at 1 million per well in a 96-well plate and incubated for 48 hours at 37°C with 5% CO2. The PBMCs were stimulated with DENV1, YFV, ZIKV, or control antigen (50 μg/ml). The PBMCs were then transferred to an IFN-γ ELISPOT plate and incubated for 17 hours before detection following the manufacturer’s instructions (Abcam, catalog no. ab46569). After plate scanning, the number of spot-forming cells was enumerated by counting.

Statistical analysis

Statistical analyses were performed using GraphPad Prism or Microsoft Office Excel. For comparison between two groups, Student’s unpaired t test was used. For multiple comparisons, one-way ANOVA was used with Dunnett’s post-test. The differences were considered significant when P was <0.05. The strength of significance is presented as follows: *P < 0.05, **P < 0.01, and ***P < 0.001.


Supplementary material for this article is available at

Table S1. List of the 17 predicted peptides with DENV1 MHC-II binding capacity and their position on the DENV1 polyprotein.

Fig. S1. Enhanced DENV1 clearance in LNs of DENV1- and JEV-immune mice.

Fig. S2. Assessment of LN cellularity following secondary DENV1 challenge.

Fig. S3. Relative TMEM expansion after stimulation with synthetic peptide-pulsed APCs.

Fig. S4. Validation of TEM proliferation and IFN-γ production after stimulation with synthetic peptides.

Fig. S5. Gating strategy to identify activated human CD4 TEM cells by flow cytometry.

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We thank the European Virus Archive for providing us with the ZIKV H/PF/2013 strain. Funding: This work was funded by NMRC/CBRG/0084/2015, TCRP ACP0113689, R01 HL112921, and start-up funding from the Duke-NUS Medical School. Author contributions: Experiments were performed primarily by W.A.A.S. with assistance from L.T. A.P.S.R. performed the phylogenetic analysis. Most data were analyzed and interpreted by A.L.S. with contributions from W.A.A.S. and A.P.S.R. The manuscript was written by A.L.S. and A.P.S.R. The study was designed and interpreted by A.L.S. Conception and funding were contributed by A.L.S., E.E.O., and S.N.A. Human patient recruitment and ethical approvals were led by J.L. with contributions from A.L.S., W.A.A.S., and E.E.O. All authors contributed to discussions and reviewed/edited the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.

Stay Connected to Science Advances

Navigate This Article