IgG antibodies to dengue enhanced for FcγRIIIA binding determine disease severity

See allHide authors and affiliations

Science  27 Jan 2017:
Vol. 355, Issue 6323, pp. 395-398
DOI: 10.1126/science.aai8128

A rare ability to enhance dengue virus disease

In some cases, secondary infections of dengue virus can be extremely serious and result in plasma leakage, thrombocytopenia, and hemorrhagic disease. This phenomenon has been attributed to antibody-dependent enhancement. Wang et al. show that a specific subclass of antibody, IgG1, which lacks fucosyl residues on the Fc segment of the heavy chain of the immunoglobulin, is elevated in patients with severe secondary dengue disease. These non-neutralizing antibodies bind activating Fc receptors and appear to cross-react with platelet antigens to cause platelet depletion, contributing to thrombocytopenia.

Science, this issue p. 395


Dengue virus (DENV) infection in the presence of reactive, non-neutralizing immunoglobulin G (IgG) (RNNIg) is the greatest risk factor for dengue hemorrhagic fever (DHF) or dengue shock syndrome (DSS). Progression to DHF/DSS is attributed to antibody-dependent enhancement (ADE); however, because only a fraction of infections occurring in the presence of RNNIg advance to DHF/DSS, the presence of RNNIg alone cannot account for disease severity. We discovered that DHF/DSS patients respond to infection by producing IgGs with enhanced affinity for the activating Fc receptor FcγRIIIA due to afucosylated Fc glycans and IgG1 subclass. RNNIg enriched for afucosylated IgG1 triggered platelet reduction in vivo and was a significant risk factor for thrombocytopenia. Thus, therapeutics and vaccines restricting production of afucosylated, IgG1 RNNIg during infection may prevent ADE of DENV disease.

Antibody-dependent enhancement (ADE) has been shown to occur in a variety of in vitro and in vivo dengue virus (DENV) infection models, but ADE does not explain why fewer than 15% of human DENV infections that occur in the presence of reactive, non-neutralizing immunoglobulin G (IgG) (RNNIg) progress to dengue hemorrhagic fever (DHF) or dengue shock syndrome (DSS) (17). This gap in our knowledge of the determinants responsible for progression to severe disease impedes the rational design of flavivirus vaccines that avoid eliciting ADE-mediating antibodies and limits proactive care to patients who are at higher risk for developing DHF/DSS. We therefore determined whether a unique property of these RNNIg, seen in the small proportion of patients that progress to DHF/DSS, could account for ADE.

The currently accepted model of ADE in DENV infection posits that RNNIg opsonizes viral particles and mediates increased infectivity of Fc gamma receptor (FcR)–expressing cells (810). Studies have shown that immune complex (IC)/FcR-mediated internalization of DENV virus can result in more infected cells (11, 12), enhanced fusion (7, 13), and suppression of innate immune signaling (14, 15). Severe DENV disease has been associated with specific combinations of virus serotypes and preexisting serotype immunity (16), viral genetic factors (1720), and host factors (2124), including association of asymptomatic DENV infection with the lower binding R131 polymorphism of FcγRIIa (25, 26). How these and other factors may contribute to ADE of human dengue disease remains unknown.

Most FcRs exhibit low affinity for monomeric IgG and are only engaged upon formation of multivalent IgG ICs. FcR-mediated effector functions include positive mechanisms such as antibody-dependent cellular cytotoxicity (ADCC), phagocytosis, and proinflammatory cytokine production, as well as negative functions, such as inhibition of inflammatory immune responses. Whether Fc domains within an IC engage activating or inhibitory FcRs is determined by Fc structure, which, in turn, is determined by IgG subclass and the precise composition of an N-linked glycan present on the CH2 domain of each heavy chain (2729). Humans have four IgG subclasses (IgG1 to 4), with IgG1 and IgG2 in highest abundance in serum, together constituting 90 to 95% of antigen-specific IgG (30). Each subclass is distinct in its ratio of binding to activating/inhibitory type I FcγRs, with IgG1 having significantly higher affinities for activating type I FcγRs than IgG2 (table S1). The Fc glycan has a core heptasaccharide structure that can be modified by the addition of specific residues; these modifications are dynamic and regulate the biological activity of IgGs by modulating Fc structure. Glycans lacking a core fucose have enhanced affinity for the activating FcγRIIIA and can confer improved effector activity compared with fucosylated Fc glycans (31). Fucosylated, sialylated glycoforms (sFc) confer binding ability of Fc to type II FcγRs, which mediate a variety of modulatory activities (table S1) (30, 32). We have previously demonstrated substantial variation in determinants of Fc structure between individuals (30), suggesting that some people may be predisposed to ADE of disease by virtue of production of IgGs that engage activating FcRs with higher affinity.

To determine whether individuals with severe DENV disease have a specific IgG Fc domain structure that might contribute to ADE, we analyzed the distribution of IgG subclass and Fc-associated glycoforms from hospitalized patients who were positive for IgG antibodies to DENV (anti-DENV IgG) in the early phase of disease, indicating previous infection with DENV or another flavivirus that elicited a cross-reactive IgG response (33) (34). Here, we refer to these patients who were positive for anti-DENV IgG in the early phase of disease as RNNIg+. RNNIg+ patients in this study were infected with any of the four viral serotypes and had clinical diagnoses either of dengue fever (DF), the relatively mild disease form, or DHF/DSS (table S2). Samples were obtained at an early time point (day 4 to 8 of fever) and a convalescent time point (>30 days after hospital discharge). As a comparator group for early viral infection, we studied anti-hemagglutinin (HA) IgG from subjects enrolled in a controlled influenza A virus (IAV) infection study. The IAV cohort was composed of healthy adults who had serum taken before and 7 days after intranasal inoculation with a wild-type IAV strain (A/California/04/2009/H1N1) (35).

Analysis of Fc-associated glycan composition showed that RNNIg+ patients with DENV infection exhibited a specific elevation of afucosylated Fc glycoforms (afucFc) on IgGs reactive with the dengue virus envelope protein (ENV), relative to patients infected with IAV (Fig. 1A). There was no evidence for differences in other glycoforms (Fig. 1B). Increased afucFc anti-ENV IgG during the early phase of infection persisted during the convalescent phase (Fig. 1A). In addition to the afucFc, anti-ENV IgGs in early infection were more likely to be IgG1 than IgG2 subclass (Fig. 1C), with both modifications (afucFc and IgG1) enhancing capacity of the Fc to engage the activating FcγRIIIA.

Fig. 1 Afucosylated Fc glycoforms and IgG1 subclass are enriched in dengue infection.

(A) Anti-ENV IgGs in early DENV or convalescent DENV infection (conv) show an increased abundance of afucosylated Fc glycans (afucFc) compared with healthy adults (uninfected), IgGs on day 7 after IAV infection (IAV) or Octagam or Flebogam intravenous immunoglobulin (IVIG) preparations. (B) No difference among sialylated Fc glycoforms (sFc) was observed. (C) Anti-ENV IgGs were skewed in distribution toward the IgG1 subclass during early or convalescent DENV infection compared with anti-HA IgGs from uninfected healthy adults, patients on day 7 of IAV infection, or Octagam or Flebogam IVIG preparations. (D) DENV patient IgGs that were reactive with DENV NS1 protein, cross-reactive/reactive with Zika virus ENV, or with HA protein were also elevated in afucFc relative to anti-HA IgG from IAV patients. (E) Zika ENV-reactive IgGs in early DENV infection were skewed in distribution toward the IgG1 subclass.

To determine whether elevated afucFc was specific for anti-ENV IgGs, we tested early-infection IgGs that were reactive with DENV NS1 protein, cross-reactive/reactive with Zika virus ENV (36), or specific for IAV HA protein. These other IgG specificities, including anti-HA IgGs, were similarly elevated in afucFc compared with anti-HA IgG from IAV patients (Fig. 1D), indicating that a global shift in IgG Fc structure had occurred in early DENV infection. In addition, IgGs reactive with the ENV protein of the infecting DENV serotype, or reactive with ENV proteins from the noninfecting DENV serotypes, were equivalent in abundance of afucFc in a subgroup analysis (fig. S1). Zika ENV-reactive antibodies were also elevated in IgG1/IgG2 ratio, whereas IgGs reactive with DENV NS1 or IAV HA proteins were not (Fig. 1E), indicating that modulation of Fc fucosylation was a more general feature in DENV infection than subclass bias.

Stratification of patients by clinical diagnosis showed that anti-ENV and anti-HA IgGs from DHF/DSS patients were elevated in afucFc compared with IgGs from either DF or IAV patients (Fig. 2A). Stratification of patients based on presence of thrombocytopenia (TH+) during disease course, which is a requisite criterion for DHF diagnosis, showed that TH+ patients had similarly elevated afucFc anti-ENV and anti-HA IgGs, with afucFc ≥ 10% being a significant risk factor for TH+ (P = 0.0139; odds ratio 11.00; 95% confidence interval 1.635 to 74.00; relative risk 1.833) (Fig. 2B). TH+ also correlated with an increased IgG1/IgG2 ratio for anti-DENV ENV (Fig. 2C). Total IgG from TH+ patients had measurably higher affinity for FcγRIIIA by surface plasmon resonance (fig. S2).

Fig. 2 Activating Fc phenotype in dengue infection correlates with disease severity.

(A) Increased afucFc on anti-ENV or anti-HA IgG correlated with DHF/DSS (DHF). (B) Stronger correlations yet were observed when patients were stratified based on thrombocytopenia (TH+) (platelets <100,000/ul) during hospitalization. TH+ patients had elevated abundance of afucFc on anti-ENV or anti-HA IgG compared with TH patients. (C) TH+ patient anti-ENV IgGs were skewed in distribution toward the IgG1 subclass, whereas anti-HA IgGs from TH+ patients were not. (D and E) Patients with the lowest recorded platelet count during hospitalization had the greatest abundance of afucFc and the highest IgG1/IgG2 ratio of anti-ENV IgGs. (F) Abundance of afucFc on anti-ENV IgG correlated with IgG1/IgG2 distribution; elevated afucFc and IgG1/IgG2 correlated with the severity of thrombocytopenia. The single patient with DSS in the cohort had the greatest abundance of afucFc and the highest IgG1/IgG2 ratio (marked with *). (G and H) Elevated hematocrit (HCT), a marker of plasma leak that distinguishes DHF/DSS from DF, correlated with anti-ENV IgGs having the greatest abundance of afucFc and the highest IgG1/IgG2 ratio.

Elevated afucFc and IgG1/2 ratio not only correlated with being TH+ during hospitalization but also correlated with the lowest platelet count recorded for each patient during hospitalization (Fig. 2, D and E). Further, these two determinants of higher-affinity binding to FcγRIIIA correlated with each other, and patients with the greatest abundance of afucFc IgG1 were most likely to have severe thrombocytopenia (Fig. 2F). The single patient with DSS had the greatest abundance of afucFc IgG1 of all patients in the cohort (Fig. 2F). In addition to clinical diagnosis and severity of TH+, elevated hematocrit, an indication of plasma leak that distinguishes DHF/DSS from DF, also correlated with the abundance of afucFc and the elevated IgG1/IgG2 ratio of anti-ENV IgGs (Fig. 2, G and H).

The correlation between high afucFc IgG1 and severe disease indicated that this Fc structure may play a role in ADE during DENV infection. In particular, the correlation between the abundance of afucFc IgG1 and the degree of thrombocytopenia (Fig. 2, D to F) led us to hypothesize that anti-DENV IgGs that cross-react with platelet antigens might contribute to platelet loss during dengue infection and thus to ADE of dengue disease. Because anti-DENV NS1 IgG has been shown to cross-react with platelets (37), we tested whether transfer of IgG from severely TH+ patients could mediate platelet reduction, in vivo, to a greater extent than IgG from TH patients. IgGs from TH+ patients caused a drop in platelets in mice humanized for FcRs (hFcR), whereas, even at high doses, IgG from TH patients did not reduce platelet numbers (fig. S3). IgG from TH+ patients was treated to remove Fc glycans, producing an aglycosylated TH+ pool (TH+agly) that would no longer engage FcRs; this pool had less effect on platelet numbers, whereas mice lacking all FcRs (α−/−) were resistant to TH+ IgG-mediated thrombocytopenia (Fig. 3A) (38). This loss of platelets was dependent on two low-affinity human activating FcRs, FcγRIIA (CD32A) and FcγRIIIA (CD16A), because deletion of either receptor rendered the mice resistant to TH+ IgG-induced platelet reduction (Fig. 3B).

Fig. 3 DHF IgG induces FcR-dependent platelet drop in vivo.

(A) IgG from TH+ patients, but not IVIG, caused a drop in platelets that was dependent on presence of the Fc glycan (TH+agly) and FcRs (α−/−). (B) Platelet drop was dependent on expression of FcγRIIA (CD32A) and FcγRIIIA (CD16A). (C) TH+ or TH IgG or IVIG was incubated with human platelets. IgG at equivalent concentrations eluted from platelets, or the remaining supernatant (sup), reacted with DENV NS1 protein but not DENV ENV protein. Data are representative of experiment performed in duplicate. (D) NS1-specific IgG from TH+ patients had elevated afucFc glycoforms over anti-HA IgG from patients with day 7 IAV (HA).

Because purified IgG alone caused a loss of platelets, we next investigated whether IgG from TH+ patients might bind to platelets directly. After incubation with human or mouse platelets, TH+ IgG could be eluted from platelets that bound the DENV NS1 protein but not the ENV protein (Fig. 3C and fig. S4). As with anti-ENV IgG from TH+ patients, anti-NS1 IgG had elevated afucFc (Fig. 3D), but the IgG1/IgG2 ratio was not different (fig. S5). All together, this showed that the IgGs enhanced for FcγRIIIA binding from patients who became thrombocytopenic during DENV infection could mediate FcR-dependent platelet loss in vivo. At least three mechanisms could contribute to this platelet reduction: anti-NS1/platelet IgG may activate platelets directly through platelet FcγRIIA, and/or cause sequestration or uptake of platelets by monocytes that express both FcγRIIIA and FcγRIIA, and/or ADCC of platelets could occur via FcγRIIIA (38).

These experiments showed that anti-DENV IgGs with enhanced affinity for FcγRIIIA could mediate ADE of disease, which is distinct from ADE of infection. Serum pools of both TH+ and TH patients mediated ADE of DENV infection in the standard U937 cell assay (fig. S6). This result, while confirming that IgGs from RNNIg+ DENV patients can mediate ADE, is not informative in the context of our observations that enhanced dengue disease is associated with an FcRIIIA-activating IgG phenotype. This is because U937 cells are FcγRIIIA, FcγRIIA+ (39–(41). Thus, we distinguish between ADE of infection and ADE of disease in our study.

That the precise Fc structure of antibodies present during DENV infection may contribute to disease severity raised the question of whether this Fc structure was present before infection or was triggered by DENV infection itself. To address this, we compared the Fc of antibodies obtained from TH+ patients during the early and convalescent phases of disease. The convalescent phase was marked by a significant drop in both afucFc and IgG1/IgG2 ratio (Fig. 4, A and B), indicating that, in patients with severe disease, DENV infection itself triggered an elevation in IgGs with enhanced affinity for FcγRIIIA.

Fig. 4 DENV infection contributed to activating Fc phenotype in TH+ patients.

(A and B) The convalescent phase of DENV infection was marked by a significant drop in both (A) afucFc and (B) IgG1/IgG2 ratio in TH+ patients, demonstrating an aberrant IgG response during infection.

The present finding that some individuals respond to DENV infection by producing IgGs with higher affinity for FcγRIIIA indicates a host determinant of susceptibility to severe DENV disease. Further studies will determine how this patient selectivity may contribute to additional mechanisms underlying ADE of DENV disease.

Supplementary Materials

Materials and Methods

Figs. S1 to S4

Tables S1 and S2

References (4245)

References and Notes

  1. Materials and methods are available online as supplementary materials.
  2. See the methods online.
  3. Acknowledgments: We thank S. Zhang and R. Sherwood at the Cornell Proteomics and Mass Spectrometry Facility for helpful discussions and excellent technical support. The authors also thank the staff of the Stanford Clinical Virology Laboratory for their assistance with dengue virus serologic testing. T.T.W. thanks B. Coller, S. J. Schlesinger, and the Rockefeller University KL2 Clinical Scholars Program for training and support. T.T.W. was supported, in part, by grant UL1TR001866 from the National Center for Advancing Translational Sciences, NIH, and the Clinical and Translational Science Award program. J.S. thanks S. Wang and S. Gunisetty for helping with preparation of sera from patient samples. K.P. was funded by NIH IRO1AI099385. S.B. was supported by an American Foundation for AIDS Research Mathilde Krim Fellowship in Basic Biomedical Research (109519-60-RKVA). Research reported in this publication was supported by the National Institute of Allergy and Infectious Diseases of the NIH under award numbers U19AI111825 (J.V.R.), U19AI109946 (J.V.R.), U19AI057266 (R.A. and J.W.), and U01AI115651 (R.A.). The influenza A virus challenge study was supported by the Intramural Research Program of the National Institute of Allergy and Infectious Diseases (NIAID), as well as the NIAID Extramural Clinical Research Acceleration Program. Support and infrastructure were also provided by The Rockefeller University and by Stanford University School of Medicine. Analysis of clinical samples in this work was approved by the Institutional Review Board of Rockefeller University (protocol #TWA-0804) (T.T.W.). The content is solely the responsibility of the authors and does not necessarily represent the official views of NIH. The data presented in this manuscript are in the main paper and in the supplementary materials.
View Abstract

Navigate This Article