Research ArticleDengue

The Structural Basis for Serotype-Specific Neutralization of Dengue Virus by a Human Antibody

See allHide authors and affiliations

Science Translational Medicine  20 Jun 2012:
Vol. 4, Issue 139, pp. 139ra83
DOI: 10.1126/scitranslmed.3003888


Dengue virus (DENV) is a mosquito-borne flavivirus that affects 2.5 billion people worldwide. There are four dengue serotypes (DENV1 to DENV4), and infection with one elicits lifelong immunity to that serotype but offers only transient protection against the other serotypes. Identification of the protective determinants of the human antibody response to DENV is a vital requirement for the design and evaluation of future preventative therapies and treatments. Here, we describe the isolation of a neutralizing antibody from a DENV1-infected patient. The human antibody 14c10 (HM14c10) binds specifically to DENV1. HM14c10 neutralizes the virus principally by blocking virus attachment; at higher concentrations, a post-attachment step can also be inhibited. In vivo studies show that the HM14c10 antibody has antiviral activity at picomolar concentrations. A 7 Å resolution cryoelectron microscopy map of Fab fragments of HM14c10 in a complex with DENV1 shows targeting of a discontinuous epitope that spans the adjacent surface of envelope protein dimers. As found previously, a human antibody specific for the related West Nile virus binds to a similar quaternary structure, suggesting that this could be an immunodominant epitope. These findings provide a structural and molecular context for durable, serotype-specific immunity to DENV infection.


Dengue virus (DENV), a member of the Flaviviridae family, is a positive-sense RNA virus infecting about 100 million people each year. Other members of the same family include West Nile virus (WNV), yellow fever, and Japanese encephalitis virus, all of which are important human pathogens. There are four DENV serotypes. In human populations, sequential infection with different DENV serotypes is common, and this is linked to an increased risk of developing severe disease, namely, dengue hemorrhagic fever and dengue shock syndrome. This is mediated by cross-serotype binding (heterotypic) antibodies with poor protective activity engendered by the original infection—a process termed antibody-dependent enhancement (ADE) (1). Durable immunity to DENV is associated with the appearance of neutralizing antibodies targeting the infecting serotype (2), but little is known about their fine specificity or mode of action.

The DENV particle includes a nucleocapsid core, which consists of an 11-kb single-stranded RNA wrapped by core proteins. The nucleocapsid core is encapsulated by a lipid bilayer membrane that has 180 copies of membrane proteins and 180 copies of envelope proteins anchored on it. The DENV envelope protein is the dominant antigen, and it forms dimeric structures that lie parallel to the viral lipid membrane. The virus surface envelope and membrane proteins are arranged in a pseudo T = 3 symmetry (3). The triangulation number (T) reflects the number of structural units in each of the 60 asymmetric units of an icosahedral virus. The ectodomain of the envelope protein consists of three distinct domains: E-DI, E-DII, and E-DIII. E-DIII mediates receptor binding, and the fusion loop at the tip of E-DII facilitates fusion with endosomes (4, 5). Studies on mouse antibodies to DENV have implicated the lateral ridge of E-DIII as the principal determinant of virus neutralization (6, 7, 8, 9). However, a comparison of the neutralization activities of DENV patient sera before and after depletion of E-DIII–specific antibodies showed no reduction in potency. This, combined with studies on the antibody repertoire engendered by a natural DENV infection, suggested poor concordance with the mouse (10, 11). To better understand the nature of human serotype-specific antibody responses to DENV, we generated a panel of Epstein-Barr virus (EBV)–immortalized B cell lines from a DENV1-infected patient (12). These formed the source of the anti-DENV1 monoclonal antibody (mAb) characterized in this study.


A neutralizing DENV1-specific antibody 14c10 from a DENV1-infected patient

A group of B lymphocyte cell lines secreting antibodies with serotype-specific binding and neutralizing activity for DENV1 were identified, subcloned, and expanded. One of these cell lines, BCL-14c10, produced immunoglobulin G (IgG) with significantly stronger binding and neutralizing activity than the others (fig. S1A). This cell line was used as a source of Ig gene templates for polymerase chain reaction amplification and expression of recombinant human IgG1 antibodies (fig. S1Ba). One recombinant human antibody 14c10 (HM14c10) had comparable binding activity for DENV1 with the parental antibody BCL-14c10 (fig. S1Bb). HM14c10 antibody bound and neutralized the DENV1 serotype but not DENV serotype 2, 3, or 4 (Fig. 1A and fig. S1C).

Fig. 1

HM14c10 is a human antibody specific for DENV1. (A) The HM14c10 antibody exhibits neutralization activity specific for DENV1 as assessed by the PRNT assay. (B) HM14c10 induces homotypic ADE for DENV1 at subneutralizing concentrations. The control HM4G2 antibody induces heterotypic ADE for DENV1 at subneutralizing concentrations. (C) The Fab fragment of HM14c10 or mutation (N297Q) of the IgG1 Fc region of HM14c10 reduced homotypic ADE. (D) Different subclasses of human IgG (HM14c10) mediate differential levels of homotypic ADE. (E) HM14c10 is neutralizing for multiple DENV1 genotypes. The genotypes are indicated in brackets beside the virus designation. Error bars represent SDs of triplicate samples, and all experiments were conducted at least three times.

The ADE activity linked to the development of dengue hemorrhagic fever and dengue shock syndrome has been proposed to occur when subneutralizing concentrations of antibodies and DENV form complexes that bind to Fc receptor–bearing cells. This leads to an increase in virus uptake and secretion of proinflammatory cytokines and chemokines (13). We compared the activity of the HM14c10 antibody with a humanized anti-flavivirus mAb HM4G2 (14) by means of an established in vitro assay using the myelomonocytic cell line K562 that expresses Fcγ receptors (15). HM4G2 has cross-serotype binding activity and targets a conserved fusion loop of E-DII on the four DENV serotypes (16). As expected, we observed that HM14c10 exhibits some homotypic enhancement of DENV1 infection at subneutralizing concentrations but no enhancing activity for DENV2, 3, or 4, to which it does not bind. In contrast, HM4G2 mediates enhancement of all four serotypes at subneutralizing concentrations (Fig. 1B and fig. S2A). To investigate the contribution of the Fcγ receptors expressed by K562 cells to the observed homotypic ADE activity of HM14c10, we expressed the antibody as a Fab fragment or we reduced Fcγ receptor binding by removing the glycosylation site on human IgG1 through substitution of the asparagine residue (N) at position 297 for glutamine (Q) (17). Both the Fab fragment of HM14c10 and the N297Q mutant exhibited a reduction in their homotypic ADE activity compared with whole IgG1 control antibodies (Fig. 1C). We next compared the influence of the IgG subclass on ADE activity and observed a partial correlation with the reported binding activities for Fcγ receptor IIA expressed by K562 cells (18). The ADE activity can be ranked as follows: IgG3 > IgG1 > IgG2 > IgG4, with IgG3 being highest and IgG4 the lowest (Fig. 1D). Thus, the ADE activity of this neutralizing anti-DENV antibody appears dependent on Fcγ receptor binding, although it should be noted that the influence of the high-affinity Fcγ receptor 1 and complement components on virus neutralization were not addressed in these experiments (19).

An additional complexity of DENV is the presence of multiple genotypes within a single serotype. DENV1 genotypes can vary by up to 3% in their amino acid composition, and previous reports of mouse anti-DENV antibodies have suggested that protective activity can vary between genotypes (6, 7, 20, 21). The binding activity of HM14c10 antibody to the five DENV1 genotypes was compared with HM4G2. Both HM14c10 and HM4G2 exhibited binding activity for the genotypes tested, with HM4G2 displaying better binding characteristics in most cases (fig. S2B). HM14c10 exhibited neutralization activity for all five genotypes tested with moderately better activity against genotypes 1, 3, and 5 compared with genotypes 2 and 4 (Fig. 1E). The plaque reduction neutralization test (PRNT50) values range from 5 ng/ml (3146 SL, genotype V) to 1.505 μg/ml (16007 DV1, genotype 2) (fig. S2C).

HM14c10 binds to a quaternary structure–dependent epitope

A cryoelectron microscopy (cryoEM) structure of a Fab HM14c10-DENV1 complex was solved to 7 Å resolution (Fig. 2A). At full occupancy, 120 copies of Fab HM14c10 bind to all of the available 180 copies of envelope proteins on the virus surface. To identify the footprint of HM14c10 on the envelope protein, we fitted the crystal structure of DENV1 envelope protein (22) into the cryoEM density map (fig. S3 and table S1). The 7 Å resolution cryoEM map showed clear density connections between the HM14c10 Fabs and the envelope proteins, allowing the identification of envelope protein residues at the interacting interface (Fig. 2B and fig. S4). The epitope recognized by HM14c10 is dependent on the quaternary structure of the virus. Two Fabs of HM14c10 bind to three envelope proteins in the virus asymmetric unit (Fig. 2, C and D). Each antibody binds across two adjacent envelope proteins with half of the epitope on E-DIII and the other half on E-DI and the E-DI–E-DII hinge of a neighboring envelope protein.

Fig. 2

HM14c10 binds a virus quaternary structure–dependent epitope. (A) CryoEM map of a Fab HM14c10-DENV1 complex showing 120 Fabs (blue) binding to 180 envelope proteins on the virus surface (cyan). Black triangle represents an asymmetric unit. (B) View of connecting densities of a Fab HM14c10 (I) to envelope protein epitope (purple spheres). Envelope proteins E-DI, E-DII, and E-DIII are colored in red, yellow, and blue, respectively. (C) Densities of Fab molecules on envelope protein Cα chains in two asymmetric units. Fab HM14c10 (I) and HM14c10 (II) are the two independent molecules in an asymmetric unit. (D) Epitopes of Fab HM14c10 (I) (purple spheres) and HM14c10 (II) (cyan spheres) on the three envelope proteins (shaded in gray) in an asymmetric unit.

To understand the Fab interaction with the envelope protein, we created a homology model of the variable region of HM14c10 on the basis of a reference human antibody structure [Protein Data Bank (PDB) code 2GHW] using the Modeller program (23). The variable region of the light and heavy chain of the homology model was then fitted into the cryoEM density map. Although the structures of both chains are similar, there is a distinctive fit that gives a better correlation to the density (fig. S5, A and B). Analysis of the Fab–envelope protein interface suggests that all complementarity-determining regions of the heavy and light chains are involved in the interaction (fig. S5C).

The binding footprints of the two HM14c10 Fabs in an asymmetric unit are not identical (Fig. 2D), with 12 amino acids common to both interfaces but 4 that are unique (Table 1). Sequence comparison of the epitope residues between DENV1 genotypes I and V indicate that all residues are conserved (fig. S6A). However, HM14c10 showed different levels of neutralization activity against these genotypes (Fig. 1E). This suggests a differential level of exposure of the target residues on the virus surface between genotypes.

Table 1

Fab HM14c10 epitope on DENV1 envelope protein. Envelope protein residues in the epitope were identified when connecting densities to Fab molecules were observed at 2.5σ contour level.

View this table:

Comparison of the epitope residues between DENV serotypes and WNV (fig. S6B) showed that the epitope is not conserved, which is consistent with the specific binding activity of HM14c10 to DENV1 (fig. S1Cb).

Time-lapse confocal microscopy reveals the neutralization mechanism of HM14c10

Antibodies can neutralize viral infections by diverse mechanisms including inhibition of virus attachment or fusion to endosomal membranes, or through blocking conformational changes of the virus surface glycoproteins (24, 25). To understand the mechanism of HM14c10 neutralization of DENV1, we used time-lapse confocal microscopy to track the infection of cells by infectious, fluorescently tagged DENV (26) (Fig. 3, A to C). When baby hamster kidney (BHK) cells were incubated with DENV1 and isotype control mAbs (non-DENV binding), the virus coalesced in multiple, predominantly perinuclear, intracellular compartments (Fig. 3Aa and movie S1). Neutralizing concentrations of HM4G2 induced the formation of viral aggregates in the extracellular space, but these were also successfully internalized, confirming that HM4G2 does not inhibit virus attachment or internalization (Fig. 3, A and B, and movie S2). In contrast, HM14c10 induced the formation of smaller aggregates, but efficiently blocked attachment, with most of the small viral particles remaining in the extracellular space after 1 hour (Fig. 3, A and C, and movie S3). HM4G2 delayed the accumulation of intracellular viruses compared to the isotype control (Fig. 3B, upper and middle panels). HM14c10-DENV1 complexes failed to enter cells but could be seen deflecting from their surface (Fig. 3B, lower panel). The degree of fluorescent DENV1 internalized under all three conditions was quantified (Fig. 3C). In addition, a pre- and post-attachment neutralization assay indicates that HM14c10 can also inhibit DENV1 infection after the virus has bound to its host cell, although the concentration of antibody required for this effect was significantly higher than that required for pre-attachment inhibition (Fig. 3D). These data suggest that the principal mode of inhibition of DENV1 by HM14c10 is by blocking virus attachment to host cells. Because the virus-antibody complex is unable to attach to cells, the post-attachment inhibition activity of HM14c10 may not play an important role in the inhibition of virus infection. However, this observation is consistent with the structural data showing that HM14c10 would lock the virus in its quaternary structure, disallowing any envelope protein structural changes that could lead to the fusion of virus to endosomal membranes.

Fig. 3

HM14c10 neutralizes DENV1 by inhibiting attachment and fusion of DENV1 and exhibits potent antiviral activity in vivo. (A) Time-lapse confocal microscopy demonstrating DENV1 infection of BHK host cells in the presence of (a) isotype control mAb, (b) HM4G2, and (c) HM14c10. Left panels of (a) to (c): DENV1 and mAbs were labeled with Alexa Fluor 647 (red) and Alexa Fluor 488 (green), respectively. Right panels of (a) to (c): cell boundaries (white dotted lines) and the distribution of DENV1 in cells. (B) Close-up of live infection events. DENV1 virions were observed inside BHK cells from 18 min in the isotype controls and from 28 min with HM4G2 antibody. HM14c10-DENV1 complexes were unable to attach to BHK cells. (C) Virus (red fluorescence intensity) quantified in 120 randomly selected cells after 1 hour of infection. **P < 0.0001, Student’s t test. (D) Pre- and post-attachment neutralization of DENV1 by HM14c10. HM14c10 inhibits virus attachment at 1 μg/ml; at higher concentrations, it can also inhibit a post-attachment step of infection. Two independent experiments were performed with duplicates. Error bars represent SDs. (E) HM14c10 was tested for antiviral activity in vivo; antibody is administered to DENV1-infected AG129 mice at days 0 and 2 after infection, respectively. HM14c10 had antiviral effects whether the virus was injected (a) subcutaneously or (b) intraperitoneally. n = 5 in both cases. *P < 0.05; **P < 0.0001, compared with PBS controls, Student’s t test.

HM14c10 exhibits potent antiviral activity in vivo

Although DENV is not a natural pathogen in immunocompetent rodents, it is possible to induce a dose-dependent viremia in AG129 mice deficient in receptors for type I/II interferon (IFN). We injected these mice with unmodified DENV1 subcutaneously (model I, fig. S7Aa) or intraperitoneally (model II, fig. S7Ab) and then quantified viremia 3 and 4 days later, respectively (27). Two DENV1 clinical isolates, representing disparate genotypes (EHI-D1 genotype I versus Westpac genotype IV), were used to determine the in vivo efficacy of HM14c10. In both models, HM14c10 lowers viremia when given to mice 24 hours before DENV1 infection, or when given 48 hours after infection (Fig. 3E). The lowest concentration of HM14c10 where a significant reduction in viremia was observed is 0.6 μg per mouse (or 160 pM), indicating strong in vivo potency.


Recent reports on the humoral responses engendered by DENV infection suggest that there is a dominance of antibodies that are mostly DENV serotype cross-reactive with weak neutralization activities. Although scarce in the human serum repertoire, E-DIII antibodies are suggested to protect against DENV infection (28, 29), and this is consistent with studies on the murine antibody response to DENV (9). The human antibodies characterized have principally been specific for DI and DII of the virus envelope protein. A small number of characterized antibodies were observed to bind to the whole virus but not to recombinant E protein, suggesting specificity for quaternary structure–dependent epitopes (28). Here, we have isolated and thoroughly characterized a neutralizing antibody against DENV serotype 1. This antibody neutralized DENV1 in both in vitro and in vivo systems. Because it binds only to DENV1, it does not cause enhanced infection of myelomonocytic K562 cells by other DENV serotypes.

The 7 Å resolution cryoEM map of the Fab HM14c10-DENV1 complex allows direct mapping of the epitope because densities connecting the Fab and envelope protein can be observed. The footprint of HM14c10 spans across E-DIII and E-DI–E-DII from a neighboring envelope protein (Fig. 2D). A report on a human antibody CR4354 specific for WNV has also implicated this region as a target for immunity (30). Although the cryoEM structure of WNV complexed with Fab CR4354 is solved to lower resolution (14 Å resolution) (Fig. 4A), the fitting of the Fab CR4354 crystal structure generates a pseudo-atomic resolution structure. This allowed identification of interacting residues. Comparison of the CR4354 epitope on WNV and the HM14c10 epitope on DENV1 (Fig. 4B) showed that CR4354 has a bigger proportion of the footprint on E-DIII, whereas HM14c10 has most of the interacting residues on E-DI. Sequence comparison of the epitope residues showed that only about 20% of the CR4354 epitope overlaps with HM14c10, and the overlapping residues are mostly nonconserved (Fig. 4C).

Fig. 4

Comparison of epitopes bound by a WNV antibody CR4354 and DENV1-specific HM14c10 antibody. (A) Fit of human antibody CR4354 and the homology model of HM14c10 to envelope proteins on WNV (left) (30) and DENV (right), respectively. CryoEM density is displayed at 2.8σ (CR4354-WNV) or 2.5σ (HM14c10-DENV1) contour level. (B) An asymmetric unit of WNV (left) and DENV1 (right) with antibody CR4354 or HM14c10 footprints shown in spheres. Epitopes at the two independent binding sites in an asymmetric unit are colored in purple and cyan. The three envelope proteins in an asymmetric unit are shaded in gray. An asymmetric unit is shown as a black triangle. (C) Comparison of amino acid residues in the two independent epitopes (a and b) between CR4354 (on WNV) and HM14c10 (on DENV). Residues in the two independent epitopes are colored as in (B).

CR4354 and HM14c10 epitopes exist in similar regions on the virus surface, and thus, we may expect the mechanism of inhibition of infection to be similar. However, this is not the case. CR4354 was shown to preferentially inhibit WNV fusion (31), whereas HM14c10 primarily inhibits the attachment of DENV to host cells (although at much higher concentrations, it can also inhibit the post-attachment step of infection).

The surface proteins of DENV have been suggested to undergo constant changes under physiological   conditions—termed “breathing” (25). It is possible that breathing may play a role in facilitating attachment of the virus to cells. Because HM14c10 cross-links surface envelope proteins, it may then inhibit attachment by preventing the surface proteins from undergoing breathing. Alternatively, because E-DIII has been shown to be important for host cell attachment, the binding of HM14c10 to E-DIII may sterically hinder this process.

The envelope proteins of most flaviviruses have similar quaternary structures based on a high degree of similarity between the cryoEM structures of WNV (32) and DENV (33). Therefore, all flavivirus surface envelope proteins may undergo similar structural rearrangements during their infection cycle. Antibodies that target a region similar to HM14c10 or CR4354 in other flaviviruses may therefore be protective. Because HM14c10 and CR4354 antibodies are the only two antibodies characterized with this binding activity and both are derived from human sources, this indicates that this type of epitope is probably a determinant for generalized flavivirus immunity.

Although human E-DIII–specific antibodies elicit strong protective responses, they represent a minority in the human antibody repertoire, which consists of mostly E-DI–E-DII antibodies (28, 29). This suggests that E-DIII is not immunodominant and may therefore be unable to stimulate strong antibody responses in the general human population. On the other hand, large numbers of quaternary structure–dependent antibodies are frequently detected in human serum (28). This highlights the technical challenges inherent in DENV vaccine design because preserving these complex tertiary structures may be vital to induce the right type of antibody response. This points to the importance of using whole virion approaches such as attenuated, inactivated, or subviral particles as candidate vaccine platforms.

Finally, given that HM14c10 is neutralizing against all DENV1 genotypes and has potent antiviral effects in vivo, this antibody represents a good therapeutic candidate for the treatment of DENV1-infected patients.

Materials and Methods

Ethics statement

Informed consent was obtained and all procedures were carried out under an approved protocol from the National University Institutional Review Board (NUS-IRB number 06-196).

Cells and viruses

C6/36 and BHK-21 cells were cultured as described previously (34). Dengue EHI strains D1 and DC164D01 were obtained from the Environmental Health Institute, Singapore (EHI) and PVP159 (DENV1/SG/07K3640DK1/2008) from the EDEN patient cohort (35). Dengue strains 16007 DV1 and 3146 SL were obtained from the Washington University School of Medicine.

Cloning of B cells

Isolation and immortalization of B cells were carried out as described previously (12). After 15 days of culture, supernatants were screened for DENV-specific antibodies by enzyme-linked immunosorbent assay (ELISA) and PRNT.

ELISA binding assays

Ninety-six–well flat-bottom plates (Maxisorp plates, Nunc) were coated with mouse 4G2 antibody at 5 μg/ml overnight. Plates were washed three times with phosphate-buffered saline (PBS)/0.01% Tween 20. Different DENV stains were added at 1 × 105 plaque-forming units (PFU) in 50 μl per well and further incubated for 2 hours. Plates were washed three times with PBS/0.01% Tween 20. HM14c10 was added to the plates and incubated for a further 1 hour. Plates were washed three times with PBS/0.01% Tween 20. Anti-human IgG-conjugated horseradish peroxidase (Pierce) was added and incubated for 1 hour. Tetramethylbenzidine substrate (GE Healthcare) was added, and 0.1 M sulfuric acid was used to stop the reaction.

Plaque reduction neutralization test

PRNT assays were conducted as described previously (36). Percentage neutralization was determined by comparing the number of plaques in various antibody dilutions to that when no antibody is present. PRNT50 is defined as the concentration of antibody where there is a 50% reduction in the number of plaques, which is determined by nonlinear regression in GraphPad Prism (version 5.0a).

Production of recombinant HM14c10

RNA from B cells was extracted with an RNA extraction kit (Qiagen). The cloning and expression of recombinant antibodies were conducted as previously described (37).

ADE assay

DENV (5 × 102 PFU/ml) was preincubated with medium, individual mAbs (HM4G2, HM14c10, or HM14c10 N297Q), subclasses of HM14c10 mAbs (IgG1, IgG2, IgG3, or IgG4), or HM14c10 Fab fragments and then added to 105 K562 cells. After an hour, cells were washed extensively with PBS to remove unbound virus and mAb. After an additional 48 hours, supernatants were harvested, and viral titers were determined by plaque assay on BHK-21 cells.

In vivo mouse experiments

AG129 mice are deficient in IFN-α/β and IFN-γ receptors (38). The mice were handled in accordance with the Institutional Animal Care and Use Committee recommendations (protocol no. 018/11). A schematic diagram detailing the prophylactic and therapeutic applications of HM14c10 versus PBS-treated controls is provided in fig. S7A. Mice were killed and viremia was quantified by an established plaque assay (39). The data are presented as averages, with error bars representing the SEM of the PFU/ml obtained from the five serum samples from each group.

Time-lapse confocal live-cell imaging

All time-lapse live-cell microscopy was performed on an inverted A1Rsi confocal microscope (Nikon) with Plan-Apochromat 100× 1.4–numerical aperture lens. Live-cell imaging was performed with living, unfixed BHK cells grown on 25-mm glass coverslips (Marienfeld GmbH) mounted onto a chamber holder (Nikon). Cells were seeded at a density of 4 × 104 per well 1 day before the experiment and cultured in RPMI 1640 supplemented with 10% fetal calf serum (FCS). For the simultaneous detection of Alexa Fluor 488–labeled antibodies (HM14c10 and HM4G2 were used at 10 μg/ml) and Alexa Fluor 647–labeled DENV1 (fig. S7B), the 488-nm line of an argon ion laser and the light of a 633-nm helium neon laser were directed over an HFT UV/488/633 beam splitter, and fluorescence was detected with an NFT 545 beam splitter in combination with a 505 to 530 bandpass filter for Alexa Fluor 488 detection and a 650 longpass filter for Alexa Fluor 647 detection. Images were captured at 30-s intervals at one frame per second for 30 to 60 min. All live-cell imaging experiments were performed with cells incubated at 37°C in a 5% CO2 microscope cage incubator system (Okolab). The images were analyzed and processed by Nikon Imaging Software (NIS) elements C software (64 bit, version 3, SP7/build 547) (Nikon).

Quantification of intracellular fluorescence

The effect of antibody on the endocytosis of DENV1 was evaluated by measuring the relative level of fluorescence within the living cells. After treatment with the respective antibody, images of at least 100 cells were randomly acquired with an A1Rsi confocal microscope from three independent experiments. The intracellular region of the cells was then individually demarcated manually with the “region of interest” (ROI) function of NIS Elements software (Nikon), and the relative fluorescence level of Alexa Fluor 488 within each cell was measured with the ROI statistics function of the software. The average, SD, and Student’s t test were calculated for each cell population with Microsoft Excel. The fluorescence from untreated cell populations infected with DEN1 was normalized to 100% and used as a comparison to antibody-treated infected cells.

Pre- and post-attachment neutralization assay

To determine whether HM14c10 inhibited at a pre- or post-attachment step, we performed a modified PRNT assay. In the post-attachment assay, BHK-21 cells were prechilled to 4°C, 100 PFU of DENV1 were added to cells, and viral adsorption was allowed for 1 hour at 4°C. Cells were washed three times with cold medium to remove unbound virus, and HM14c10 was added at the specified concentrations. Virus-antibody complexes were allowed to form for 1 hour at 4°C and then were washed three times with cold medium followed by an overlay with 0.8% methylcellulose (Aquacide II, Calbiochem) in RPMI 1640 medium and 2% FCS (without antibody or compound). For the pre-attachment assay, a PRNT assay with all cells and solutions at 4°C, in which virus-antibody was mixed for 1 hour at 4°C before addition to cells, was performed. The virus-antibody mixture was incubated with cells at 37°C for 1 hour, after which the cells were washed with medium and overlaid with 0.8% methylcellulose as described above. For both pre- and post-attachment assays, 5 μM compound 6, which is a dengue entry inhibitor blocking viral fusion with a median effective concentration (EC50) of 0.11 μM (40), and 10 μM NITD008, which is a dengue RNA synthesis inhibitor with an EC50 of 0.82 μM (41), were used as positive and negative controls, respectively (fig. S7C). The infected cells with methylcellulose were incubated at 37°C for 4 days, and plaques were visualized by fixing with 3.7% formaldehyde and staining with 1% (w/v) crystal violet. Plaques were counted and then normalized to the average of data from the medium control wells.

Cryoelectron microscopy

DENV (strain PVP159) was prepared as described previously (3). Virus was mixed with Fab HM14c10 in a molar ratio of one Fab molecule to every E protein and incubated at 37°C for 30 min and then 4°C for 2 hours. The complex was then flash-frozen in liquid ethane on lacey carbon grids, which were coated with a thin layer of continuous carbon. Virus particles were imaged with a 300-kV FEI Titan Krios in the following conditions: electron dose of 16 e2, magnification of 47,000, and defocus range of 1 to 3 μm. The images were recorded on a 4K by 4K Gatan charge-coupled device camera, resulting in a pixel size of 1.9 Å per pixel. A total of 5566 particles were boxed, and contrast transfer function parameters were determined with the programs boxer and ctfit, respectively, in the EMAN (42) program suite. Orientation of the particles was determined with multipath simulated annealing protocol (43). WNV was used as an initial model (32). The three-dimensional map was generated with the program make3d in EMAN. The resolution of the final map was found to be 7 Å resolution as determined by the Fourier shell coefficient cutoff of 0.5. The DENV1 post-fusion E protein crystal structure (22) does not fit well into the cryoEM density map as a rigid body; the domains in the E protein were thus broken up and then fitted separately. The fit of the molecules into the cryoEM map (set at 4σ contour level) was then optimized with the “fit-in-map” function of Chimera (44). The epitope was mapped by observing the connecting densities to antibody at 2.5σ contour level. To fit the Fab molecule into the cryoEM map, we first created a homology model of HM14c10 variable region on the basis of a Fab crystal structure (PDB code 2GHW), which has the highest sequence identity to HM14c10, with the program Modeller (23). The heavy and light chains of the homology model were fitted separately into the cryoEM map (set at 3σ contour level) in the two possible orientations of the Fab (fig. S5).

Supplementary Materials

Fig. S1. Identification and recombinant expression of a fully human antibody with neutralizing activity for DENV.

Fig. S2. HM14c10 exhibits binding and neutralizing activity for all DENV1 genotypes (I to V) and does not cause ADE to DENV serotypes 2, 3, and 4.

Fig. S3. Fit of the post-fusion crystal structure of DENV1 E proteins into the cryoEM map of Fab HM14c10 complexed with DENV1.

Fig. S4. Stereo diagram of the Fab HM14c10 and E protein binding interface.

Fig. S5. Fitting of the homology model of HM14c10 variable region into HM14c10-DENV1 cryoEM density map.

Fig. S6. HM14c10 epitope on DENV1 (genotype PVP159) and comparison of the epitope with (A) other DENV1 genotypes and (B) dengue serotypes and West Nile virus (WNV).

Fig. S7. Mechanism of inhibition of DENV1 by HM14c10 in vitro and pre- and post-infection treatment efficacy of HM14c10 in vivo.

Table S1. Fitting of DENV 1 E protein domains into HM14c10-DENV1 cryoEM density.

Movie S1. Live infection of BHK cells by Alexa Fluor 647–labeled DENV1 in the presence of an Alexa Fluor 488–labeled isotype control antibody (human IgG1).

Movie S2. Live infection of BHK cells by Alexa Fluor 647–labeled DENV1 in the presence of an Alexa Fluor 488–labeled humanized mouse monoclonal HM4G2 (human IgG1).

Movie S3. Neutralization of Alexa Fluor 647–labeled DENV1 by Alexa Fluor 488–labeled HM14c10.


References and Notes

  1. Acknowledgments: We thank the Nikon Imaging Centre (Biopolis, Singapore) for providing the confocal equipment and C. Khaw (Nikon Imaging Centre) for technical advice. We thank W. Jin and T. S. Leng at OPAT, National University Hospital System, for recruitment of dengue-infected patients and Y.-Q. Bai from FEI (Singapore) in assisting in the microscopy work. We thank M. Diamond (Departments of Medicine, Molecular Microbiology, Pathology and Immunology, Washington University School of Medicine) for providing the DENV1 genotype II and V viruses and L. C. Ng (National Environment Agency, Singapore) for providing the genotype III virus. J. D. Brien (Department of Medicine, Washington University School of Medicine) and S. K. Austin (Department of Pathology and Immunology, Washington University School of Medicine) are thanked for providing envelope protein sequences of DENV1 genotypes I and V. Funding: Supported by grants from the National Medical Research Council (STOP-Dengue TCR, NMRC-182-003-220-275, and NCIS C-000-999-002-001) and the National Research Foundation (NRF-182-000-218-281 and NRF2-182-005-172-281) to P.A.M. and National Research Foundation fellowship (NRF-913-301-015-281) to S.-M.L. K.G.C.S. was supported by the Wellcome Trust and National Institute for Health Research Cambridge Biomedical Research Centre. B.J.H. is funded by grants from the Defense Research and Technoogy Office, Singapore. Author contributions: E.P.T., E.W.T., and S.H.C. made and characterized the EBV-transformed B cell lines. A.P.C.L. and B.J.H. cloned recombinant HM14c10. A.Y., W.S., P.-Y.S., G.K.T., and S.A. conducted in vivo experiments. M.L.N., E.E.O., and P.K. purified DENV. P.K. prepared the cryoEM samples, fitted the protein structures, and mapped the Fab HM14c10 epitope. S.-M.L., P.K., and V.A.K. solved the cryoEM structure of the complex. M.A. took the cryoEM images. S.-M.L. supervised the structural studies. E.P.T., E.W.T., and T.T.T. conducted live infection confocal microscopy. D.F. and Y.S.L. coordinated the clinical aspects of study. G.Z. conducted the pre- and post-attachment inhibition assays. S.-M.L., P.K., D.M.K., K.G.C.S., B.J.H., and P.A.M. contributed to the writing of the paper. A.H.Y.C. conducted the flow cytometry based neutralization assays. J.K.W.N. conducted mouse studies (model 1). Competing interests: HM14c10 is the subject of U.S. Provisional patent application #61/423,085 to P.A.M., B.J.H., S.-M.L., P.K., and E.P.T. “A recombinant human monoclonal antibody with specificity for dengue virus serotype 1 E protein and uses thereof.” The other authors declare that they have no competing interests. Data and materials availability: The cryoEM density map of the HM14c10-DENV1 complex has been deposited in Electron Microscopy Data Bank under accession number EMD-5268. The atomic coordinates of fitted DENV1 E protein of one asymmetric unit have been deposited with the PDB under accession code 3J05.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article