Dengue virus immune protection is specific to the serotype encountered and is thought to persist throughout one’s lifetime. Many serotype cross-reactive memory B cells isolated from humans with previous dengue infection are specific for the nonstructural and the prM structural viral proteins, and they can enhance infection in vitro. However, plasmablasts circulating in enormous numbers during acute secondary infection have not been studied. In this study, we analyzed single plasmablasts from two patients by sorting the cells for Ig sequence analysis and for recombinant expression of Abs. In contrast to memory B cells, most plasmablast-derived Abs bound to the structural E protein of dengue, and protection experiments in mice revealed that virus serotypes encountered during past infections were neutralized more efficiently than were the serotypes of the current infection. Together with genetic analyses, we show evidence that plasmablasts in dengue patients are a polyclonal pool of activated E protein–specific memory B cells and that their specificity is not representative of the serum Abs secreted by long-lived plasma cells in the memory phase. These results contribute to the understanding of the phenomenon of original antigenic sin in dengue.
The symptoms of dengue disease are primarily immune mediated, with the most severe manifestations occurring only after patient viremia has decreased from its peak (1–3). Typical dengue symptoms include high fever, vascular leakage, rashes, headache, and bone and muscle pain. Dengue patients exhibit hallmarks of severe inflammation, including increased serum cytokine concentrations (4, 5), high numbers of activated T cells, B cells, enormous numbers of plasmablasts (PBs) (6–8), and high titers of circulating dengue-specific Abs. Sera from infected patients contain Abs against dengue virus (DENV) structural proteins (including prM, the E glycoprotein, and capsid protein), as well as against nonstructural proteins (primarily against NS1, which is secreted by infected cells) (9). DENV has four serotypes, and it is possible for a single individual to be infected four separate times by these alternative strains of virus. Intriguingly, successive rounds of infection are associated with increased disease severity, suggesting that pre-existing immunity can have a detrimental effect on the course of disease. Neutralizing Ab titers decline rapidly postinfection (10), and reactivation of specific memory B cells (MBCs) most likely contributes to protection after homologous infection. Moreover, cross-reactive MBCs might also be beneficial during heterologous infection, because they can lead to the production of large amounts of dengue-binding Abs, which, although not providing sterile protection, could positively influence the outcome of the disease by decreasing the virus load earlier in secondary cases compared with primary cases (11). In fact, serum from both primary and secondary patients is serotype cross-reactive in ELISA during the first weeks postinfection (6), and patients seem to be protected against all serotypes for a limited time (12), suggesting a potential protective role for the polyclonal Ab response generated during acute disease.
Previous reports described the specificity of MBCs isolated several weeks after acute infection with a known dengue serotype, followed by clone immortalization using EBV, screening of the secreted Abs, and cloning of Ab genes (13–16). Most Abs isolated in this way were found to react with conserved epitopes in the prM and E domain of the virus or with nonstructural proteins and, therefore, were largely serotype cross-reactive. Very few Abs exhibited a high neutralizing capacity in vitro or were sufficient to mediate protection in mouse-infection experiments.
DENV activates B cells polyclonally. During secondary infection, large numbers of PBs are observed 6–7 d after the onset of fever (6, 17); yet, the specificity of these PBs and their origin from either naive or MBCs has not been studied. PBs (CD20−CD19dimCD138+/− cells) are transiently present in the blood of patients during secondary dengue infection, coinciding with a dramatic increase in neutralizing Ab titers against all four serotypes (6, 10). CD138 is a marker to identify plasma cells in tissues such as the bone marrow. CD138+ PBs in the blood might represent a differentiation stage between PB and tissue-resident plasma cells; however, there are no studies proving this.
We isolated PBs from two patients during acute-phase secondary infections to study the specificity and protective capacity of PB-derived Abs at the single-cell level. This analysis revealed that E protein–binding Abs dominated the response and showed a high neutralizing capacity against multiple serotypes in vitro. Instead, the protective capacity of the Abs in mice was higher against the previously encountered serotype of infection, which is consistent with the phenomenon of original antigenic sin (8, 18, 19). Together with a detailed sequence analysis, we characterize the phenomenon of original antigenic sin in dengue on a cellular level in two patients.
Materials and Methods
The research involving patients enrolled in the Early Dengue Infection and Outcome study was approved by the Institutional Review Board of Singapore National Healthcare Group Ethical Domain (DSRB B/05/013), and patients gave written informed consent. As part of the collaborative program “STOP Dengue” (http://www.stopdengue.sg) (20), adult patients (age >21 y) presenting at community primary care clinics with acute-onset fever (>38.5°C for <72 h) without rhinitis or clinically obvious alternative diagnoses were included in the study. Whole-blood samples were collected into EDTA-vacutainer tubes (Becton Dickinson) at recruitment (acute phase), at 4–7 d (defervescence), and at 3–4 wk after fever onset (convalescence). Patients were diagnosed by DENV-specific RT-PCR. DENV-specific IgM and IgG Abs were detected by ELISA using the commercially available Panbio kit (Inverness Medical, Queensland, Australia). Both patients described in this study had DENV-specific serum IgG Abs at the time of fever onset and, therefore, were classified as having secondary infections.
Cell lines and virus strains
All viruses used were produced in C6/36 mosquito cells (American Type Culture Collection). The following patient-isolate strains were used for PB-screening ELISA and neutralization assays: DENV1-05K2916 [EU081234 (21)], DENV2-TSV01 (22), DENV3-VN32/96 (EU482459), and DENV4-My04 31580. DENV3 and DENV4 were gifts from Dr. Cameron Simmons (Oxford University Clinical Research Unit, Ho Chi Minh City, Vietnam) and Prof. Shamala Devi (University of Malaya, Kuala Lumpur, Malaysia), respectively. DENV1-Westpac 74 (U88535), DENV1-05K2392 (20), DENV2-TSV01, and DENV2-08K3115 were used for mouse experiments.
Cell sorting and sequence analysis
The mRNA of human IgG H and κ L chains were amplified from single B cells by RT-PCR, according to the protocol published by Smith et al. (23). The sequence quality was checked with Lasergene or CodonCode Aligner software. Sequences with high background noise or overlapping peaks were excluded from the analysis. vBase2 (http://www.vBase2.org) was used to determine the V and J gene family (Fig. 4). To calculate CDR3 length and mutations, sequences were analyzed with JOINSOLVER (24). Germline sequences were from the ImmunoGenetics database (http://www.imgt.org) (25).
Cloning, expression, and purification of Abs
The mRNA of human IgG H and κ L chains were amplified from single B cells by RT-PCR using the protocol described by Smith et al. (23). For cloning into expression vectors, RT-PCR products of selected B cells were used for nested PCR with modified primers: SalI and NheI sites were added at the 5′ and 3′ ends of the H chain, and SalI and BbvCI sites were added at the 5′ and 3′ ends of the κ L chain. The PCR products were cloned into the pTT5 mammalian expression vector (26) (licensed from the National Research Council Biotechnology Research Institute, Montreal, QC, Canada) using Infusion (Clontech), following the manufacturer’s protocol. Plasmids containing productive sequences were subsequently used for rAb expression in HEK293-6E cells (26). The control human IgG1 mAb HA4 (kindly provided by DSO National Laboratories, Singapore) is specific for H5N1 influenza virus, and human IgG1 mAb 5F-RC (27) is specific for chikungunya virus.
For DENV-specific ELISA, MaxiSorp plates (Nunc) were coated with PEG-precipitated DENV serotypes 1–4. Plates were blocked with PBS, 0.05% Tween 20, and 3% skim milk. Supernatants from Ab-expressing HEK cells were incubated on virus-coated plates for 1 h at room temperature before washing with PBS 0.05% Tween 20 and detection of virus-binding Abs with a secondary anti-human IgG-HRP (Sigma). For determination of absolute concentrations of IgG, plates were coated with anti-Ig Ab (Caltag), and an IgG standard at different concentrations was included to generate a standard curve. EDIII-specific Abs were measured on plates coated with EDIII of all four serotypes (ProSpec-Tany TechnoGene) at a concentration of 150 ng/well. Purified Abs were used for screening. E protein–specific Abs were measured on plates coated with 300 ng/well E protein of DENV1 or DENV3 (ProSpec-Tany TechnoGene) or 150 ng/well E protein of DENV2-TSV01, which was produced in S2 cells, as described (28). Supernatants from Ab-expressing 293HEK cells or purified Abs were used for screening. Pooled serum from several dengue-immune healthy donors was used as a positive control. 3,3,5,5-tetramethylbenzidine HRP substrate solution (Sigma) was used as substrate for all ELISAs. An OD value 2-fold higher than the background was defined as a positive signal.
A flow cytometry–based neutralization assay was used with modifications (29). BHK21 cell monolayers were grown in 96-well plates. Heat-inactivated plasma samples or protein G–purified mAbs diluted in RPMI 1640 medium without FCS were incubated with DENV1 (05K2916), DENV2 (TSV01), DENV3 (VN32/96), or DENV4 (My04 31580) at a multiplicity of infection ∼1 for 1 h at 37°C. Plasma–virus mixtures were then transferred onto the BHK21 monolayers and incubated for 2 h at 37°C before adding RPMI 1640, 5% FCS. After an incubation of 2–3 d, cells were stained intracellularly with Abs against NS1 and E protein and then acquired on an LSRII flow cytometer (Becton Dickinson). Data were analyzed using FlowJo software (TreeStar). The proportions of infected cells were plotted against the dilution factor, and EC50 was calculated with Prism5 (GraphPad Software), applying a three-parameter nonlinear curve fit. Values with a curve fit of R2 > 0.95 were included in the analysis.
AG129 mice (B & K Universal, Grimston, U.K.) were housed under specific pathogen-free conditions at the Biological Resource Center, Singapore. A total of 100 μg purified mAb in PBS was injected i.p., followed by virus infection via the same route 5–24 h later (3–6 × 106 PFU/mouse). The animal experiments were conducted according to the rules and guidelines of the Agri-Food and Veterinary Authority and the National Advisory Committee for Laboratory Animal Research, Singapore. The experiments were reviewed and approved by the Institutional Review Board of the Biological Resource Center, Singapore (Institutional Animal Care and Use Committee; protocol #090461).
Data were analyzed with Prism Software (version 5 for Mac), and the statistical tests used are indicated in the figure legends. A p value < 0.05 was considered statistically significant.
PBs isolated during secondary infection are dengue specific but serotype cross-reactive in vitro
We isolated total PBs from two dengue patients (10/63 and 10/50; Table I) to characterize PB-derived Abs in infected individuals. We first established the DENV immune status of these subjects using a flow cytometry–based infection-neutralization assay (29) (Table I). These data indicated that patient 10/63 had undergone a previous DENV1 infection, whereas patient 10/50 had experienced a previous DENV2 infection and possibly a DENV4 infection. RT-PCR revealed that the viruses of the ongoing infection were serotypes DENV2 (ID10/63) and DENV3 (ID10/50). PBs were identified and sorted as CD20−CD19+CD27high lymphocytes for patient 10/63 and as CD20−CD19+CD27highCD138+ or CD138− lymphocytes for patient 10/50. Ab secretion by B cell subsets sorted from dengue patients according to this strategy was verified by ELISPOT (Supplemental Fig. 1). PBs were sorted into single wells of 96-well plates for nested RT-PCR to amplify the variable regions of the Ab H (VH) and L (VL) chains and for subsequent cloning into IgG1-expression vectors (23, 30) (Fig. 1A). CD138+ PBs were chosen from patient 10/50 for the expression of Abs. IgG1 Abs were tested in ELISA for their capacity to bind to viral particles (Fig. 1B, Supplemental Table I), which indicated that mAbs from patient 10/63 preferentially bound to DENV1 and DENV3, whereas mAbs from patient 10/50 preferentially bound to DENV2. Ab binding to DENV1-, DENV2-, DENV3-, or DENV4-infected BHK21 cells in fluorescent-histology assays exhibited equal binding to all four virus serotypes for most clones (Fig. 1C, Supplemental Table I).
In summary, 77% of the PBs in patient 10/63 and 60% of the PBs in patient 10/50 demonstrated binding activity to DENV by ELISA and/or in histology (Fig. 1D).
The original serotype of infection is neutralized more efficiently in vivo
Patient plasma obtained during the time of PB isolation (day 4 or 6, respectively) already showed efficient neutralization of the new serotype of infection in vitro (Fig. 2A). To study the neutralizing capacity of individual PB-derived Abs compared with polyclonal plasma, the clones from each donor that showed the strongest binding in ELISA (Fig. 1B) were tested in a flow cytometry–based neutralization assay. The 50% neutralizing titer (NT50) for most clones occurred at concentrations between 0.01 and 1 μg/ml, and most clones neutralized all four serotypes (Fig. 2B, Supplemental Table I).
In vitro–neutralization assays offer only limited predictive value for dengue protection, because in vitro culture of the highly mobile DENV seems to expose poorly accessible viral epitopes, leading to increased Ab binding and enhanced virus neutralization (31). To address the in vivo efficacy, we next tested the protective capacity of selected mAbs in a mouse model of dengue infection (32). The mAbs selected were chosen based on their high in vitro–neutralizing capacity against previous and/or current serotypes of infection (Fig. 2C). A total of 100 μg of each mAb was injected i.p. 5–24 h before infection with DENV1, DENV2, or DENV3 i.p. (Fig. 2D). In contrast to in vitro–neutralization data (Fig. 2C), but in line with pre-existing titers in the plasma (Table I), mAbs derived from patient 10/63 were more protective against DENV1, and mAbs derived from patient 10/50 were more protective against DENV2, as shown by the reduction in viremia compared with the control mAb HA4 or 5F-RC (Fig. 2D). 10/50 Abs also reduced DENV1 viremia 4–10-fold (data not shown); however, there was no protection against the current serotype of infection. In fact, patient 10/50–derived Abs tended to increase DENV3 infection (Fig. 2D).
In summary, these data are consistent with the reactivation of cross-reactive MBCs with a higher protecting capacity against the previous serotypes of infection.
PB Abs bind to the virus surface E protein
The virus particle ELISA used for the first Ab screen (Fig. 1A) showed that most Abs bound to structural proteins that make up the coat of virus particles. The virus coat of mature particles consists of the E protein and the M protein; M is a transmembrane protein and is not exposed on the surface (33). However, the uncleaved form, called prM, is maximally exposed on immature virus particles, and virus preparations normally contain small amounts of immature particles. To confirm E protein binding, we coated ELISA plates with recombinant E protein and found that 86–100% of the mAbs bound to the recombinant E protein serotype of either the current or previous infection (Fig. 3A, Table II). The mAbs tested in mice were further assessed using Western blot. mAbs bound to E protein monomers (∼48 kD) and dimers (∼100 kD) (Fig. 3B), and no binding to prM was observed, with the exception of the pooled human serum that was used as a positive control. All mAb clones were also tested for binding to EDIII, the E protein domain that contains the most effective neutralizing epitopes (34, 35). However, none of the Abs tested bound to EDIII, at least not to the soluble monomeric form used in these assays (Table II).
In contrast to the memory B response, which is reportedly dominated by prM- and nonstructural protein–specific cells (6, 13–15), PBs preferentially bound to the E protein. This, in turn, implies that mature virus particles are the activating Ag that triggers B cells early during infection.
The PB response is polyclonal and generated from affinity-matured and selected MBCs based on sequence analysis
PBs isolated in the current study were class switched (IgG) and, thus, were likely to have undergone selection in a germinal center. Given that DENV is an acute infection, we hypothesized that the B cell clones studied in the current study may have undergone limited affinity maturation and, therefore, may have maintained the capacity to partially neutralize several serotypes. To address this question, we compared the affinity maturation of PBs relative to that of nonspecific “general” MBCs based on numbers of CDR mutations. The MBCs isolated from patient 10/50 were CD19+CD27+ and nondengue specific and were selected based on their inability to bind to fluorescently labeled DENV2 virus, whereas cells from patient 10/63 were collected from the total pool of CD19+CD27+ MBCs and, thus, might contain dengue-specific memory cells in addition to the larger pool of nonspecific memory cells. PCR products from single cells were sequenced for this analysis (Fig. 1A). In the two patients analyzed in this study, the number of CDR mutations was similar in PBs and MBCs (Fig. 4A). We next tested whether few high-affinity clones were expanded and dominated the PB response, as observed after influenza infection (30). Fig. 4B illustrates that, although clonal expansion was observed particularly for the PB population, the response remained polyclonal. Interestingly, amino acid mutations at the junction and CDR3 region were observed for many clones. Among the PB sequences chosen for rAb expression, five mAbs from patient 10/50 used the same H and L chain V(D)J elements and, thus, appeared to originate from the same progenitor cell. However, the introduction of mutations led to the loss of dengue specificity in three of the five PBs (Fig. 4C).
In summary, PBs were generated from a very diverse, affinity-matured and selected pool of MBCs that did not proliferate extensively before differentiating into PBs. However, limited proliferation might allow for the introduction of mutations for further repertoire diversification, which can also lead to the loss of dengue specificity.
Dengue-specific PBs use unique VDJ combinations compared with MBCs
To address a potential dengue-specific genetic pattern, we next determined the V and J gene family usage of both H and L chains in dengue-specific and nonspecific B cell populations of the multiserotype-specific PBs isolated (Fig. 4D). Sequences from all PBs were compared with sequences from confirmed DENV-binding PBs and with sequences from nonspecific MBCs that were isolated as described in the previous paragraph. For both patients 10/63 (Supplemental Fig. 2) and 10/50 (Fig. 4D), the VDJ gene family usage of PBs was restricted compared with the broader VDJ gene family usage in nonspecific “general” MBCs, and the unique VDJ usage in confirmed dengue-binding PBs (expression of recombinant mAbs; Fig. 1A) was even more pronounced than in total PBs. The repertoire of CD138+ and CD138− PBs was similar for patient 10/50 (data not shown).
Interestingly, dengue-binding PBs from both patients showed a preference for VH1 family gene usage, whereas VH3 gene usage was dominant in MBCs. VH3 normally dominates total peripheral B cells in healthy individuals (36, 37). These data suggest that DENV selectively binds to B cells using rare V family genes, which are efficiently activated and differentiate into PBs during acute disease.
During acute disease, pre-existing Abs in the circulation produced by long-lived plasma cells are present alongside Abs produced from newly activated PBs. Because the PB response in dengue patients is fulminant and likely has an impact on the disease, the aim of the current study was to address the Ag specificity of the total PB response and to identify the specificity and protective capacity of individual cells. Moreover, because many symptomatic dengue infections are secondary cases, and with the prospect of a dengue vaccine becoming available in the near future, we asked whether the PB response in secondary patients originated from specific memory and/or naive B cells. Therefore, we sorted patient-derived PBs during acute-phase disease and analyzed the mAbs that they expressed. PBs were found to be predominantly DENV specific in the patients studied (60 and 77% specificity, respectively). Thus, PBs comprising mostly reactivated MBCs dominate the secondary acute response, accounting for the original antigenic sin phenomenon (18, 19). At the same time, a primary response to the new serotype of infection is initiated, eventually leading to immunity against past and current serotypes of infection. The percentages of specific cells are consistent with the PB responses in influenza-immune patients infected with a new serotype of influenza (30). Moreover, Lee et al. (38) recently reported that the PB response to respiratory syncytial virus and influenza virus infection measured by ELISPOT was remarkably Ag specific and showed little bystander activation.
On the individual cell level, we found that E protein–specific cells dominated the acute PB response. This is in contrast to the previously reported MBC repertoire in dengue patients in which ≥50% mAbs bind to NS1, NS3, prM, or capsid (13–15). All Abs that bound to infected BHK-21 cells in histology also bound to virus particles in ELISA (Supplemental Table I). The virus particle preparations used for ELISA did not contain NS1 (the plasma control containing anti-NS1 Abs did not show a band typical for NS1 in the Western blot in Fig. 3 for which the same virus particles were used); therefore, we excluded binding to NS1. However, we did not specifically test for binding to NS and prM proteins, and not all Abs binding in the virus particle ELISA were available for the recombinant E protein ELISA (Table II), which might have introduced a bias toward E specificity in our analysis.
The E protein is exposed on the virus coat and contains neutralizing Ab epitopes. Therefore, E protein specificity of our Abs can explain the serotype-specific partial protection observed in animal experiments (Fig. 2C). However, the Ab binding sites seemed to include highly conserved amino acids, because most Abs bound to all serotypes in vitro. In fact, recently reported DENV-neutralizing Ab epitopes were shown to consist of quaternary structures, spanning over two adjacent E protein dimers and involving E domains I and II or III (39, 40). These findings showed that Ab binding may involve both serotype-specific and conserved epitopes. Despite binding activity of the PB mAbs in Western blot, which indicated that our Abs recognize linear epitopes, we did not observe binding to peptides (15mers with a 5-aa overlap). The exact epitope(s) of PB-derived Abs isolated remains to be established.
In addition to the specificity of Abs, a key question is the biological origin of the B cells secreting them. Cross-reactive Abs prevail after both a primary and secondary infection; therefore, it is not clear which B cells are activated during a secondary heterologous infection. Knowledge about the nature of activated or reactivated cells is of particular relevance in the context of vaccination.
PB Ig sequences identified after influenza infection were highly mutated, exceeding the mutation rate of MBCs (30). We found that dengue infection generated PBs that showed a mutation rate similar to memory cells in the same patients. Moreover, PBs had undergone selection based on replacement versus silent mutations in framework and CDR regions (data not shown), further demonstrating their MBC origin. Preferential activation of B cells with the highest affinities was demonstrated after influenza vaccination in humans (30) and in mouse models (41, 42), suggesting that high affinity of a memory cell is a general prerequisite for the initiation of PB differentiation during acute disease. At least for the two patients analyzed, PB responses were relatively polyclonal compared with the pauci-clonal response in influenza vaccinees with pre-existing immunity (30).
Strikingly, dengue-specific B cells used VH and VL gene families with a bias toward VH1 and Vκ1 (Fig. 4D). Likewise, HIV-neutralizing Abs from unrelated long-term survivors are surprisingly similar in their epitope recognition and show a bias toward VH1 gene usage (43, 44), suggesting that a VH1 gene bias is representative of a B cell population activated in the context of viral infections. Although it is not possible to generalize from two patients, BCRs using VH1 gene families may be particularly suited to bind to virus coat glycoproteins and might reside in a lymphatic environment where the virus is accumulated or filtered.
In conclusion, secondary infections with a heterologous virus serotype engage pre-existing dengue-specific MBCs from previous dengue infections and activate nonspecific MBCs to a lesser extent. The E protein specificity and diversity of PB-derived Abs may help to clear the mutating virus faster after secondary infection than after primary infection (11). It seems unlikely that these Abs have an impact on severity by enhancing infection, because the PB response peaks when viremia is already declining. It will be important to determine the extent to which PB clones are selected into the memory pool and whether they change their specificity or affinity in the memory phase, because this could help to determine correlates of protection.
The authors have no financial conflicts of interest
We thank colleagues of the “STOP Dengue” program for helping to collect dengue patient samples and providing clinical parameters and diagnosis of the virus strains. We also thank Anis Larbi and the team for help with cell sorting, Michael Poidinger for help with statistical analysis, and Neil McCarthy of Insight Editing London for manuscript editing. DENV3 and DENV4 were kind gifts from Dr. Cameron Simmons and Prof. Shamala Devi, respectively. The control human IgG1 mAb HA4 was generously provided by DSO National Laboratories, Singapore, and 9F12 was a kind gift from Dr. Subhash Vasudevan, Duke-National University of Singapore.
This study was funded by the Agency for Science, Technology and Research, Singapore.
The online version of this article contains supplemental material.
Abbreviations used in this article:
- dengue virus
- memory B cell
- 50% neutralizing titer
- Received June 19, 2012.
- Accepted October 5, 2012.
- Copyright © 2012 by The American Association of Immunologists, Inc.