Genotype-Specific Neutralization and Protection by Antibodies against Dengue Virus Type 3

Abstract

Dengue viruses (DENV) comprise a family of related positive-strand RNA viruses that infect up to 100 million people annually. Currently, there is no approved vaccine or therapy to prevent infection or diminish disease severity. Protection against DENV is associated with the development of neutralizing antibodies that recognize the viral envelope (E) protein. Here, with the goal of identifying monoclonal antibodies (MAbs) that can function as postexposure therapy, we generated a panel of 82 new MAbs against DENV-3, including 24 highly neutralizing MAbs. Using yeast surface display, we localized the epitopes of the most strongly neutralizing MAbs to the lateral ridge of domain III (DIII) of the DENV type 3 (DENV-3) E protein. While several MAbs functioned prophylactically to prevent DENV-3-induced lethality in a stringent intracranial-challenge model of mice, only three MAbs exhibited therapeutic activity against a homologous strain when administered 2 days after infection. Remarkably, no MAb in our panel protected prophylactically against challenge by a strain from a heterologous DENV-3 genotype. Consistent with this, no single MAb neutralized efficiently the nine different DENV-3 strains used in this study, likely because of the sequence variation in DIII within and between genotypes. Our studies suggest that strain diversity may limit the efficacy of MAb therapy or tetravalent vaccines against DENV, as neutralization potency generally correlated with a narrowed genotype specificity.

Dengue viruses (DENV) cause the most common arthropod-borne viral infection in humans worldwide, with ∼50 million to 100 million people infected annually and ∼2.5 billion people at risk (13, 61). Infection by four closely related but serologically distinct viruses of the Flavivirus genus (DENV serotypes 1, 2, 3, and 4 [DENV-1 to -4, respectively]) cause dengue fever (DF), an acute, self-limiting, yet severe, febrile illness, or dengue hemorrhagic fever and dengue shock syndrome (DHF/DSS), a potentially fatal syndrome characterized by vascular leakage and a bleeding diathesis. Specific treatment or prevention of dengue disease is supportive, as there is no approved antiviral therapy or vaccine available.

DENV has an ∼11-kb, single-stranded, positive-sense RNA genome that is translated into a polyprotein and is cleaved posttranslationally into three structural (envelope [E], pre/membrane [prM], and capsid [C]) and seven nonstructural (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) proteins. The three structural proteins encapsidate a single infectious RNA of the DENV genome, whereas the nonstructural proteins have key enzymatic or regulatory functions that promote replication. Additionally, several DENV proteins are multifunctional and modulate cell-intrinsic and cell-extrinsic host immune responses (10).

Most flavivirus-neutralizing antibodies recognize the structural E protein (reviewed in reference 40). Based on X-ray crystallographic analysis (32, 33), the DENV E protein is divided into three domains: domain I (DI), which is an 8-stranded β-barrel, domain II (DII), which consists of 12 β-strands, and domain III (DIII), which adopts an immunoglobulin-like fold. Mature DENV virions are covered by 90 antiparallel E protein homodimers, arranged flat along the surface of the virus with quasi-icosahedral symmetry (25). Studies with mouse monoclonal antibodies (MAbs) against DENV-1 and DENV-2 have shown that highly neutralizing anti-DENV antibodies are serotype specific and recognize primarily the lateral-ridge epitope on DIII (15, 49, 53). Additionally, subcomplex-specific MAbs, which recognize some but not all DENV serotypes, recognize a distinct, adjacent epitope on the A β-strand of DIII and also may be inhibitory (16, 28, 42, 53, 56). Complex-specific or flavivirus cross-reactive MAbs recognize epitopes in both DII and DIII and are generally less strongly neutralizing (8, 53).

Beyond having genetic complexity (the E proteins of the four distinct serotypes are 72 to 80% identical at the amino acid level), viruses of each serotype can be further divided into closely related genotypes (43, 44, 57). DENV-3 is divided into 4 or 5 distinct genotypes (depending on the study), with up to 4% amino acid variation between genotypes and up to 2% amino acid variation within a genotype (26, 58, 62). The individual genotypes of DENV-3 are separated temporally and geographically (1), with genotype I (gI) strains located in Indonesia, gII strains in Thailand, and gIII strains in Sri Lanka and the Americas. Few examples of strains of gIV and gV exist from samples isolated after 1980 (26, 62). Infection with one DENV serotype is believed to confer long-term durable immunity against strains of the homologous but not heterologous DENV serotypes due to the specificity of neutralizing antibodies and protective CD8⁺ T cells (45). Indeed, epidemiological studies suggest that a preexisting cross-reactive antibody (7, 24) and/or T cells (34, 35, 64) can enhance the risk of DHF/DSS during challenge with a distinct DENV serotype. Nonetheless, few reports have examined how intergenotypic or even strain variation within a serotype affects the protective efficacy of neutralizing antibodies. This concept is important because the development of tetravalent DENV vaccines with attenuated prototype strains assumes that neutralizing antibody responses, which are lower during vaccination than during natural infection, will protect completely against all genotypes within a given serotype (60). However, a recent study showed markedly disparate neutralizing activities and levels of protection of individual anti-DENV-1 MAbs against different DENV-1 genotypes (49).

Herein, we developed a panel of 82 new DENV-3 MAbs and examined their cross-reactivities, epitope specificities, neutralization potential at the genotype level in cell culture, and protective capacities in vivo. The majority of strongly neutralizing MAbs in this panel mapped to specific sites in DIII of the E protein. Remarkably, because of the scale of the sequence variation of DENV-3 strains, most of the protective antibodies showed significant strain specificity in their functional profiles.

MATERIALS AND METHODS

Viruses and cells.

The following DENV-3 strains were used in this study: 16652 (gI, from R. Kinney, Centers for Disease Control and Prevention), H87 (gI, from R. Tesh, University of Texas Medical Branch), UNC3043 and UNC3044 (gI, from A. de Silva, University of North Carolina), UNC3046 and UNC3049 (gII, from A. de Silva), UNC3006 and UNC3018 (gIII, from A. de Silva), and UNC3050 (gIV, from A. de Silva). Strains representing other DENV serotypes also were tested: 16007 (DENV-1), 16681 (DENV-2), and H241 (DENV-4) (53). All DENV isolates were propagated in C6/36 Aedes albopictus cells as described previously (11). Vero T144 cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 5% fetal bovine serum albumin (FBS), nonessential amino acids (NEAA), sodium pyruvate, and 100 IU/ml of penicillin and streptomycin. Raji-DC-SIGN cells were generated by electroporating Raji cells with a plasmid (pUNO; Invivogen) that encodes human DC-SIGN-1A and a blasticidin selection marker. Single-cell clones were isolated after blasticidin (5 μg/ml) selection, and cell surface expression was confirmed by flow cytometry with an anti-DC-SIGN MAb (DCN46; BD Pharmingen). Raji-DC-SIGN cells were maintained in RPMI 1640 supplemented with 10% FBS, NEAA, sodium pyruvate, penicillin, and streptomycin.

Cloning and expression of DENV-3 E protein and domains.

cDNA encoding the ectodomain (amino acid residues 1 to 407) and DIII (amino acid residues 291 to 407) of the E protein of DENV-3 strain 16652 were amplified from viral RNA isolated from infected C6/36 cells using Superscript III and Platinum HiFi Taq polymerase according to the manufacturer's instructions (Invitrogen). The PCR product was cloned into the pET21a bacterial expression plasmid (EMD Biosciences) using flanking NdeI and XhoI restriction sites, sequenced, and then expressed in BL21 Codon Plus (Stratagene) Escherichia coli by autoinduction (51). Two independent QuikChange (Stratagene) reactions were completed to generate the D328G or P330G mutation within DIII from DENV-3 strain 16652. Inclusion bodies containing insoluble aggregates were denatured in the presence of 6 M guanidine hydrochloride and 20 mM β-mercaptoethanol and refolded in the presence of 400 mM l-arginine, 100 mM Tris base (pH 8.0), 2 mM EDTA, 0.2 mM phenylmethylsulfonyl fluoride, and 5 mM reduced and 0.5 mM oxidized glutathione. Refolded protein was separated from aggregates on a Superdex 200 size exclusion column using fast-protein liquid chromatography (GE Healthcare). The DENV-3 E protein ectodomain (amino acid residues 1 to 395, strain 16652) was cloned downstream of a honeybee melittin signal peptide (amino acid residues 1 to 21) into a baculovirus expression vector (pFastBac; Invitrogen) using BamHI and XhoI cloning sites. Recombinant baculoviruses were propagated in SF9 insect cells in Grace's medium under serum-free conditions.

Generation and purification of anti-DENV-3 MAbs.

Anti-DENV-3 MAbs were generated as part of six independent splenocyte-myeloma fusions as described previously (37). For the first three fusions, which produced MAbs DENV-3-E1 to DENV-3-E43, C57BL/6 mice were immunized and boosted iteratively with 25 μg of recombinant DENV-3 E protein and DIII. For the remaining fusions, DENV-3-E49 to DENV-3-E98 MAbs were generated after infection of alpha/beta interferon receptor-negative (IFN-α/βR^−/−) C57BL/6 mice with 10⁴ PFU of DENV-3 (strain 16652, gI) by an intraperitoneal route and rechallenge 3 weeks later with the same strain, route, and dose. Three weeks after secondary challenge, mice were boosted subcutaneously with 20 μg DENV-3 DIII (16652) mixed with alum (100 μl) and 10 μg CpG1826 (IDT) in a final volume of 200 μl per mouse. Mice were bled, and those with sera having plaque reduction neutralization titers (PRNT) of 1/1,000 or greater were selected. Mice received a final boost with 20 μg of purified DIII in phosphate-buffered saline (PBS) intravenously. Three days later, splenocytes were fused to P3X63Ag8.53 myeloma cells (19). MAbs were subcloned by limiting dilution, isotyped (Southern Biotech), and purified using protein A or G affinity chromatography (Invitrogen).

In vitro neutralization assay.

Because several of the DENV-3 strains do not form visible plaques on BHK21-15 or Vero cells, focus-forming assays (47) were used to measure the neutralizing titers of MAbs against DENV-3 isolates. Fourfold serial dilutions of MAb (5 μg/ml to 2.5 ng/ml) were mixed with ∼100 focus-forming units (FFU) of virus, incubated at 37°C for 1 h, and added to Vero T144 monolayers in 24-well plates for 1 h at 37°C to allow virus adsorption. Medium was removed, and cells were overlaid with 1% methylcellulose mixed with DMEM containing 5% FBS and incubated for 4 days. Monolayers were washed thrice with PBS to remove methylcellulose, fixed with 1% paraformaldehyde in PBS for 10 min at room temperature, rinsed, and permeabilized in Perm Wash (PBS, 0.1% saponin, and 0.1% BSA). Infected cell foci were stained by incubating cells with the flavivirus-cross-reactive, chimeric, human West Nile virus (WNV) MAb E18 (1 μg/ml) (39) for 1 h at 37°C and then washed three times with Perm Wash. Foci were detected after the cells were incubated with a 1:2,000 dilution of horseradish peroxidase-conjugated goat anti-human IgG (Sigma) for 1 h. After three washes with Perm Wash, staining was visualized by addition of TrueBlue detection reagent (KPL) and terminated by rinsing monolayers in water. Infected foci were enumerated by counting cells in wells using a Nikon dissecting microscope with a 10× objective lens.

Flow cytometry analysis of DENV-infected cells.

To assess MAb reactivity with heterologous DENV serotypes, Raji-DC-SIGN or C6/36 cells (depending on the strain) were infected with individual strains at a multiplicity of infection (MOI) of 1 and harvested 3 or 8 days later, respectively. Cells were washed in PBS, fixed with 1% paraformaldehyde, permeabilized, and incubated sequentially with 20 μg/ml MAb and 4 μg/ml Alexa-647-conjugated goat anti-mouse IgG (Invitrogen) for 30 min on ice. After a final wash, cells were processed on a FACSArray flow cytometer (Becton Dickinson) and analyzed using FlowJo software v8 (Treestar).

Domain mapping by yeast surface display.

A cDNA fragment encoding DIII (amino acid residues 293 to 409) was amplified from the DENV-3 strains 16652 (gI), UNC3043 (gI), UNC3049 (gII), UNC3006 (gIII), and UNC3050 (gIV) by reverse transcription) (RT-PCR using Superscript III reverse transcriptase and Platinum HiFi Taq DNA polymerase with BamHI and XhoI sites added at the 5′ and 3′ ends, respectively. Cloning, transformation, and expression were performed as described previously (37, 53), with one exception: yeast synthetic dropout medium (Sigma) was used in place of Casamino Acids. Saccharomyces cerevisiae cells were stained after sequential incubation with 50 μl of purified (20 μg/ml) MAb and Alexa-647 goat anti-mouse IgG secondary antibody (4 μg/ml), each for 30 min. Yeast cells were processed by flow cytometry without permeabilization and analyzed as described above.

Generation of a yeast expressing DIII variants.

Site-specific mutations were engineered into a DENV-3 strain 16652 DIII yeast expression plasmid by a reverse genetics approach using the QuikChange II mutagenesis kit (Stratagene). Point mutations were generated at amino acid positions 301, 302, 303, 328, and 329 to define potential variation in MAb binding to the different genotypes within the DENV-3 serotype. Mutations at amino acid residues 305, 306, 308, 309, 325, 330, 340, 384, 386, and 389 in the A strand, BC and FG loops, and G strand were generated based on previously described epitopes for WNV, DENV-1, and DENV-2 (15, 16, 28, 37, 42, 49, 53). Surface expression of individual DENV-3 DIII variants was confirmed by flow cytometry after they were stained with two nonneutralizing MAbs, DENV-3-E1 and DENV-3-E8.

Mouse protection experiments.

All mouse studies were approved and performed according to the guidelines of the Washington University School of Medicine Animal Safety Committee. IFN-α/βR^−/− × IFN-γR^−/− mice on the 129 Sv background (AG129 mice) were a gift from H. Virgin (Washington University School of Medicine) and bred in a pathogen-free barrier facility. For protection experiments, mouse-adapted strains of DENV-3 16652 (gI) and UNC3050 (gIV) were used. Adapted strains were generated by passaging virus between AG129 mice (after intracranial [i.c.] infection) and C6/36 insect cells, with DENV-3 16652 passaged twice and DENV-3 UNC3050 thrice in this manner. In prophylaxis experiments, AG129 mice were administered a single dose of individual MAbs (500 μg via the intraperitoneal [i.p.] route) 1 day before infection. Subsequently, mice were challenged with DENV-3 strain 16652 (gI, 5 × 10³ PFU) or UNC3050 (gIV, 1 × 10⁵ PFU) by the i.c. route, and mortality was monitored for 50 days. In postexposure therapeutic experiments, a single dose (500 μg) of MAb was administered by i.p. injection 2 days after i.c. infection with 5 × 10³ PFU of strain 16652.

Mapping of mutations onto the DENV-3 DIII structure.

Figures were prepared using the atomic coordinates of DENV-3 DIII (RCSB protein data bank accession number 1UZG) with the program PyMOL (www.pymol.org). The alignment of DENV-3 DIII from different genotypes was created with the program ALSCRIPT (3).

Bioinformatic analysis.

Nine hundred ninety-five full-length DENV-3 E protein amino acid sequences were downloaded from the Virus Variation database (NCBI) and aligned using MUSCLE (http://www.drive5.com/muscle). Sequences were manually trimmed to include only DIII sequences (amino acid residues 295 to 408). Duplicate sequences were removed by DuplicatesFinder (bioinfotutlets.blogspot.com), resulting in 139 unique DENV-3 DIII sequences. The locations and percentages of amino acid variation were completed by calculating percent identity at each amino acid position. The solvent-accessible surface area was determined using GETAREA (12).

Statistical analysis.

Data were analyzed for statistical significance using Prism software (GraphPad Software). For survival analysis, Kaplan-Meier survival curves were analyzed by the log rank test. For neutralization assays, an unpaired Student t test was used. Focus reduction neutralization titers (FRNT) that resulted in 50% inhibition (50% effective concentration [EC₅₀]) were determined using nonlinear-regression analysis.

RESULTS

Generation of MAbs.

Studies from several decades ago showed that polyclonal sera directed against the E protein can protect against lethal flavivirus challenge in animals (6, 41). More recent experiments with WNV, DENV-1, and DENV-2 (4, 15, 16, 37, 46, 49, 53) have suggested that DIII of the E protein is a key target for strongly neutralizing MAbs, although the human B cell repertoire may be directed away from this region (38, 55, 59). To gain insight into the structural specificities of protective MAbs that recognize the DENV-3 E protein, we performed six independent myeloma-B cell fusions and generated 82 new mouse MAbs after screening 4,300 primary wells. Three fusions were performed after BALB/c mice were immunized with recombinant DENV-3 E protein (strain 16652) generated in insect cells, resulting in 33 MAbs (see Table S1 in the supplemental material). Because these MAbs had poorly neutralizing activities, we modified the immunization protocol and instead infected IFN-α/βR^−/− C57BL/6 mice with strain 16652; the immunodeficient mice supported increased DENV-3 replication compared to that in wild-type mice, which are relatively nonpermissive for infection (48). Infection was followed by virus rechallenge and boosting with DIII of the homologous strain; this augmented neutralizing polyclonal antibody titers substantially (data not shown). By this approach, 49 additional anti-DENV-3 MAbs were generated (see Table S1 in the supplemental material).

All MAbs were initially characterized for neutralizing activity, isotype, epitope recognition, and cross-reactivity. MAbs were initially tested semiquantitatively for neutralization of the homologous DENV-3 strain by a single-endpoint plaque reduction assay using Vero cells and neat hybridoma supernatant (∼10 μg/ml). Of the MAbs generated, 32 showed no inhibitory activity (<15% neutralization), 28 had modest inhibitory activity (15 to 90% neutralization), and 22 were strongly inhibitory (>90% neutralizing) (see Table S1 in the supplemental material); one of the MAbs (DENV-3-E83) was not tested because we were unable to isolate an individual subclone. Subsequently, MAbs were screened for E protein domain recognition using yeast cells expressing DENV-3 DI-DII or DIII on their surfaces (see Table S1 in the supplemental material). MAbs in the sequences of DENV-3-E2 to DENV-3-E37, with the exception of DENV-3-E8 and DENV-3-E22, bound yeast expressing DENV-3 DI-DII (data not shown). Mapping studies using DIII of the homologous strain 16652 identified 28 DIII-specific MAbs. Notably, 16 MAbs (DENV-3-E52, DENV-3-E58, DENV-3-E70, DENV-3-E72, DENV-3-E75, DENV-3-E76, DENV-3-E81, DENV-3-E82, DENV-3-E85, DENV-3-E87, DENV-3-E88, DENV-3-E89, DENV-3-E92, DENV-3-E93, DENV-3-E94, and DENV-3-E96) failed to bind yeast cells expressing DI-DII or DIII on their surfaces yet recognized fixed and permeabilized SF9 insect cells infected with a baculovirus that expressed the ectodomain of DENV-3 E protein (data not shown) and, thus, were defined as E protein specific. The remaining six MAbs were either IgM and not analyzed further or were tested by Western blotting of DENV-3-infected cell lysates under nonreducing conditions and failed to produce a signal, leaving their antigen specificity undetermined.

All MAbs were tested for cross-reactivity using cells infected with WNV or different serotypes of DENV (Fig. 1 and see Table S1 in the supplemental material). Twenty-three of the 54 DI-DII-specific MAbs and 4 of the 25 DIII-specific MAbs were subcomplex or complex specific and bound to some or all of the other serotypes of DENV. Only 3 (DENV-3-E31, DENV-3-E32, and DENV-3-E33) of the 82 MAbs in our panel were completely cross-reactive and bound all DENV serotypes and WNV. DENV-3-E11 and DENV-3-E15 displayed an unusual reactivity pattern and recognized DENV-1, DENV-3, and WNV but not DENV-2 or DENV-4.

FIG. 1.

Defining the serotype cross-reactivities of anti-DENV-3 MAbs by flow cytometry. MAbs DENV-3-E77, DENV-3-E61, DENV-3-E3, DENV-3-E55, and DENV-3-E47 were added to fixed, permeabilized, uninfected Raji-DC-SIGN cells (filled histograms) and cells infected (dashed-line histograms) with strains corresponding to different DENV serotypes (DENV-1 [strain 16007], DENV-2 [strain 16681], DENV-3 [strain 16652], and DENV-4 [strain H241]). Results are representative of 4 independent experiments. MAb reactivity was determined by the loss of binding to infected cells, as judged by a decrease (shift to the left) in the mean fluorescent intensity of staining.

Characterization of strongly neutralizing MAbs against DENV-3.

Twenty-one of the 23 MAbs that neutralized infection at a level of greater than 50% in the single-endpoint plaque reduction neutralizing tests were purified and used for all remaining functional and epitope mapping experiments. MAb DENV-3-E66 was isotyped as an IgM antibody and not examined further in this study. MAb DENV-3-E55 produced poor antibody yields and, thus, was used only for some in vitro experiments. Of the 21 strongly neutralizing MAbs, 3, 15, 1, 1, and 1 were isotyped as IgG1, IgG2a, IgG2b, IgG2c, and IgG3, respectively (see Table S1 in the supplemental material). The variation in IgG2 isotypes (IgG2a and IgG2c) is due to allotypic differences between BALB/c and C57BL/6 mice (29). Twenty of the 23 neutralizing MAbs mapped to DENV-3 DIII, as determined by recognition of yeast cells that expressed DIII of strain 16652 on their surfaces.

Genotype-specific reactivities of anti-DENV-3 MAbs.

To further evaluate the specificity of these neutralizing MAbs against a range of genetically diverse DENV-3 isolates, we infected C6/36 cells with nine different DENV-3 virus strains that represented four DENV-3 genotypes (Fig. 2 A and B and data not shown). This allowed us to determine whether epitope recognition by these MAbs was strain or genotype specific. Four gI strains (16652, H87, UNC3043, UNC3044), two gII strains (UNC3046, UNC3049), two gIII strains (UNC3006, UNC3018), and one gIV strain (UNC3050) were used for this analysis. Remarkably, among the neutralizing MAbs, only DENV-3-E61 recognized all nine DENV-3 strains equivalently (Table 1). Nine MAbs bound to cells infected only with gI and gII strains, whereas two MAbs recognize cells infected with gI, gII, and gIII strains. DENV-3-E62 was unique in its ability to recognize cells infected with gI, gII, and gIV viruses. In comparison, a nonneutralizing MAb (DENV-3-E8) that mapped to DIII reacted with cells infected with all DENV-3 strains regardless of genotype.

FIG. 2.

Genetic variation among DENV-3 genotypes within DIII. (A) Sequence alignment of nine different DENV-3 strains. The sequence of DIII from DENV-3 strains of different genotypes (gI, strains H87, 16652, UNC3043, and UNC3044; gII, strains UNC3046 and UNC3049; gIII, strains UNC3006 and UNC3018; and gIV, strain UNC3050) were aligned. The secondary structures of DENV-3 E DIII residues 291 to 401 from the strains that have not been crystallized were predicted by DSSP (22) using the H87 strain coordinates. Black blocks highlight residues of genotypic variation. Colored boxes correspond to specific neutralizing antibody and structural recognition determinants according to epitopes mapped by yeast surface display (Table 5). The results of the yeast surface display epitope mapping are denoted underneath in red to indicate the number of neutralizing MAbs in our panel that lose binding when a specific amino acid is changed. Green or cyan circles denote solvent-accessible or inaccessible amino acids, respectively, in the mature dengue virion at sites of variation between any DENV-3 strains. (B) C6/36 cells infected with the indicated DENV-3 strains were fixed, permeabilized, and incubated with the indicated MAbs (20 μg/ml). Uninfected (filled histogram) and infected (dashed-line histograms) cells after exposure to DENV-3 16652 (gI), DENV-3 UNC3044 (gI), DENV-3 UNC3046 (gII), DENV-3 UNC3018 (gIII), and DENV-3 UNC3050 (gIV) are shown. Results are representative of four independent experiments.

TABLE 1.

MAb binding to C6/36 insect cells infected with strains representing different DENV-3 genotypes^a

MAb	Specificity	Genotype recognition	% of control value for indicated strain
			16652 (gI)	H87 (gI)	UNC3043 (gI)	UNC3044 (gI)	UNC3046 (gII)	UNC3049 (gII)	UNC3006 (gIII)	UNC3018 (gIII)	UNC3050 (gIV)
Neutralizing
DENV-3-E47	Type	Varies	113	36	36	1	24	52	35	0	166
DENV-3-E48	Type	I, II	113	117	100	104	104	79	25	0	8
DENV-3-E49	Type	I, II	107	118	90	103	107	98	3	0	1
DENV-3-E50	Type	I, II	113	118	95	104	107	101	2	0	0
DENV-3-E51	Type	I, II, III, IV	112	119	100	104	106	100	69	11	111
DENV-3-E52	Type	Varies	11	7	0	0	1	0	0	0	0
DENV-3-E53	Type	I, II, III	114	77	68	86	94	78	73	36	32
DENV-3-E54	Type	I, II	101	118	82	99	105	81	1	0	4
DENV-3-E56	Type	Varies	104	83	2	5	70	54	6	0	1
DENV-3-E57	Type	I, II	99	118	87	103	107	93	9	0	2
DENV-3-E58	Type	Varies	103	5	1	0	11	26	5	0	5
DENV-3-E59	Type	I, II	100	119	76	104	108	84	7	0	4
DENV-3-E60	Type	I, II	95	118	73	102	106	78	1	0	1
DENV-3-E62	Type	I, II, IV	88	118	80	102	107	80	33	0	118
DENV-3-E63	Type	I, II	115	119	100	102	106	98	0	0	ND
DENV-3-E64	Type	Varies	100	20	4	0	0	9	1	0	8
DENV-3-E67	Type	I, II	97	118	69	101	105	69	1	0	1
DENV-3-E55	Subcomplex	I, II, III	102	117	91	103	106	86	87	115	1
DENV-3-E61	Subcomplex	I, II, III, IV	97	119	72	104	107	70	66	116	64
DENV-3-E77	Complex	I, II, III	109	117	99	103	104	82	109	114	ND
Poorly neutralizing
DENV-3-E97	Type	I, III	98	112	91	96	103	26	89	103	ND
DENV-3-E8	Type	I, II, III, IV	101	100	100	100	100	100	100	100	100
DENV-3-E84	Type	I, II, III	102	112	93	97	103	24	88	106	ND
DENV-3-E98	Complex	I, III	88	84	68	84	23	10	39	71	ND

^aC6/36 cells were infected with the indicated DENV-3 strains and stained with MAbs (20 μg/ml). Data are the averages of results from three to four independent experiments and are normalized to values for percentage-positive cells based on staining with the control MAb DENV-3-E8. Boldface indicates that the level of binding of a given MAb was from 0 to 25% of the control value, and underlining indicates that the level of binding was from 26 to 50% of the control value. Note that values that exceed 100% represent MAbs that likely have a higher avidity for a given strain than were indicated by detection by DENV-3-E8. ND, not done.

Neutralizing potential of MAbs against different DENV-3 genotypes.

Given the variation in binding of our panel of MAbs to insect cells infected with different DENV-3 genotypes, we assessed their ability to neutralize infection more quantitatively. All DENV-3 MAbs that had strong neutralizing activity in the single-endpoint PRNT assay (see Table S1 in the supplemental material) were tested over a range of concentrations to establish an EC₅₀ using one DENV-3 strain from each genotype. In these experiments, a focus-forming reduction assay (to determine FRNT) was substituted for a classical PRNT assay, as many of the low-passage-number DENV-3 strains failed to consistently cause plaques in BHK21 or Vero cell monolayers. Notably, no single MAb efficiently neutralized strains corresponding to all four DENV-3 genotypes (Fig. 3 A and Table 2). One MAb, DENV-3-E51, strongly neutralized (EC₅₀ ≤ 514 ng/ml) viruses from three of the DENV-3 genotypes (gI, gII, and gIII). Seven MAbs efficiently neutralized two of the four genotypes, whereas, and somewhat surprisingly, 10 MAbs neutralized only the strain that was used for immunization to generate the panel of MAbs (16652, gI). When the levels of cell binding and inhibitory activities of all of the neutralizing MAbs were compared across genotypes, we observed a largely direct relationship between preservation of binding to a given genotype and neutralizing activity (see Table S2 in the supplemental material).

FIG. 3.

Neutralization of DENV-3 strains. (A) MAb neutralization of different genotypes of DENV-3. Neutralization of different DENV-3 genotypes (16652, gI; UNC3046, gII; UNC3018, gIII; and UNC3050, gIV) by MAbs DENV-3-E47 (left), DENV-3-E51 (middle), and DENV-3-E60 (right). Increasing concentrations of purified MAbs were mixed with 10² PFU of the indicated DENV-3 strains, corresponding to all genotypes, and inhibition was assessed by an FRNT assay in Vero cells. Graphs were generated after regression analysis using Prism statistical software. The data are representative of at least four independent experiments. (B) MAb neutralization of different gI strains. Neutralization of different gI DENV-3 strains (16652, H87, UNC3043, and UNC3044) by MAbs DENV-3-E47 (left), DENV-3-E51 (middle), and DENV-3-E49 (right). Increasing concentrations of purified MAbs were mixed with 10² PFU of the indicated gI DENV-3 strains, and neutralization was assessed by an FRNT assay in Vero cells. Graphs were generated after regression analysis using Prism software. The data are representative of at least four independent experiments.

TABLE 2.

MAb neutralization of different DENV-3 genotypes^a

MAb	Specificity	Domain location	EC50 for:
			gI 16652	gII UNC3046	gIII UNC3018	gIV UNC3050
DENV-3-E47	Type	DIII	110	>5,000	>5,000	4,600
DENV-3-E48	Type	DIII	500	>5,000	>5,000	>5,000
DENV-3-E49	Type	DIII	8	4	>20,000	>20,000
DENV-3-E50	Type	DIII	24	>5,000	>5,000	>20,000
DENV-3-E51	Type	DIII	20	25	514	>5,000
DENV-3-E52	Type	E	3,739	ND	ND	ND
DENV-3-E53	Type	DIII	38	457	>5,000	>20,000
DENV-3-E54	Type	DIII	9	>5,000	1,982	>20,000
DENV-3-E56	Type	DIII	2	>20,000	4,700	>20,000
DENV-3-E57	Type	DIII	6	<1.0	>5,000	>20,000
DENV-3-E58	Type	DIII	20	>20,000	>20,000	>20,000
DENV-3-E59	Type	DIII	42	3	>5,000	>20,000
DENV-3-E60	Type	DIII	90	67	>20,000	3,586
DENV-3-E62	Type	DIII	9	162	>5,000	>5,000
DENV-3-E63	Type	DIII	33	ND	>20,000	>20,000
DENV-3-E64	Type	DIII	4	>20,000	>20,000	>20,000
DENV-3-E67	Type	DIII	5	383	>20,000	>20,000
DENV-3-E55	Subcomplex	DIII	>5,000	>5,000	1,980	>5,000
DENV-3-E61	Subcomplex	DIII	281	>5,000	99	3,084

^aThe assay to determine focus reduction neutralization titers (FRNT) was performed on Vero cells after incubating 10² PFU of strains of the indicated DENV-3 genotype with increasing concentrations of purified MAbs. All EC₅₀ values are indicated in ng/ml. Values are the averages of results from three to five independent experiments performed in duplicate. EC₅₀ values were calculated by nonlinear-regression analysis. ND, not done.

Neutralizing potential of MAbs against different gI DENV-3 strains.

Because of the unanticipated extent of genotype-specific neutralization, we evaluated whether the MAbs efficiently inhibited three additional DENV-3 strains (H87, UNC3043, and UNC3044) of gI that vary within DIII by 1.5% at the amino acid level (Fig. 3B and Table 3). These gI viruses vary at three amino acid positions in DIII: 302, 303, and 391 (Fig. 2A). DENV-3-E47 neutralized only strain 16652, whereas DENV-3-E60 and DENV-3-E62 neutralized two of the four gI viruses. Four MAbs (DENV-3-E49, DENV-3-E50, DENV-3-E54, and DENV-3-E57) inhibited the four gI viruses, albeit to various degrees, whereas three MAbs (DENV-3-E51, DENV-3-E59, and DENV-3-E61) strongly neutralized all four gI strains.

TABLE 3.

MAb neutralization of DENV-3 genotype I viruses^a

MAb	EC50 for:
	gI 16652	gI H87	gI UNC3044	gI UNC3043
DENV-3-E47	109	>20,000	>20,000	>20,000
DENV-3-E49	4	113	2,478	1,229
DENV-3-E50	24	16	1,136	522
DENV-3-E51	20	41	218	242
DENV-3-E54	9	49	3,121	2,570
DENV-3-E57	6	19	1,013	1,079
DENV-3-E59	42	71	112	3
DENV-3-E60	2	160	>20,000	>20,000
DENV-3-E61	281	853	54	95
DENV-3-E62	9	43	>20,000	>20,000

^aFRNT analysis was performed on Vero cells with increasing concentrations of purified MAbs and 10² PFU of the indicated DENV-3 gI strain. All EC₅₀ values are indicated in ng/ml. Values are the averages of results from three to five independent experiments performed in duplicate.

Epitope mapping of DIII-neutralizing MAbs.

Given the differences in genotype-specific inhibition by several DIII-specific MAbs, we mapped binding determinants using yeast surface display to gain structural insight into the basis for differential neutralization. Initial studies were performed with 23 of our DIII-specific MAbs and four DIII variants (UNC3043, UNC3049, UNC3006, UNC3050) corresponding to the different DENV-3 genotypes (Table 4). There are 11 amino acid positions within DENV-3 DIII that vary between genotypes: positions 301, 302, 303, 322, 329, 340, 368, 380, 383, 386, and 391, most of which are surface exposed (underlined) (Fig. 2A). The greatest diversity overall between genotypes occurs within the N-terminal linker of DIII at positions 301, 302 and 303. Of the five poorly neutralizing MAbs, two (DENV-3-E8 and DENV-3-E98) bound to DIII from all four genotypes, whereas three (DENV-3-E1, DENV-3-E84, DENV-3-E97) recognized DIII from gI, gII, and gIII viruses. Unlike with staining of infected, fixed, and permeabilized C6/36 cells, two neutralizing MAbs (DENV-3-E77 and DENV-3-E61) recognized DIII from all four genotypes on yeast. Two neutralizing MAbs (DENV-3-E51 and DENV-3-E57) recognized DIII from three gI, gII, and gIV strains, whereas eight neutralizing MAbs (DENV-3-E49, DENV-3-E50, DENV-3-E54, DENV-3-E59, DENV-3-E60, DENV-3-E62, DENV-3-E63, and DENV-3-E67) recognized DIII of gI and gII strains only. Six neutralizing MAbs (DENV-3-E47, DENV-3-E48, DENV-3-E53, DENV-3-E56, DENV-3-E58, DENV-3-E64), however, bound to yeast expressing DIII from only the 16652 strain. Although binding to DIII on yeast generated from heterologous DENV-3 strains did not always correlate with data from infected cells, it more directly predicted neutralizing activity. For example, DENV-3-E48, which failed to neutralize gII viruses, bound insect cells infected with gII viruses but not yeast cells expressing gII DIII. Consistent with this, DENV-3-E47, DENV-3-E48, DENV-3-E51, DENV-3-E53, and DENV-3-E62 bound insect cells but failed to recognize yeast-displayed DIII or neutralize the infection of a given individual genotype (Table 4).

TABLE 4.

Binding of DENV-3 MAbs to yeast cells expressing DIIIs of different DENV genotypes^a

MAb	% of control value for indicated strain
	16652 (gI)	UNC3043 (gI)	UNC3049 (gII)	UNC3006 (gIII)	UNC3050 (gIV)
Neutralizing
DENV-3-E47	137	13	40	18	12
DENV-3-E48	131	67	7	3	11
DENV-3-E49	138	54	75	6	17
DENV-3-E50	136	66	77	12	11
DENV-3-E51	125	60	62	2	57
DENV-3-E53	123	13	20	13	1
DENV-3-E54	131	36	67	3	1
DENV-3-E56	124	1	37	1	1
DENV-3-E57	133	61	66	30	74
DENV-3-E58	119	1	2	1	1
DENV-3-E59	123	67	57	2	65
DENV-3-E60	132	47	57	2	29
DENV-3-E62	124	32	58	2	27
DENV-3-E63	120	29	51	2	36
DENV-3-E64	110	1	2	1	1
DENV-3-E67	131	37	88	42	1
DENV-3-E61	117	62	53	82	69
DENV-3-E77	130	106	94	133	94
Poorly neutralizing
DENV-3-E1	115	105	110	114	1
DENV-3-E8	100	93	92	95	86
DENV-3-E84	106	79	58	99	0.3
DENV-3-E97	114	86	65	107	6
DENV-3-E98	105	78	65	95	90
WNV E111	56	57	45	83	85

^aStaining of yeast cells expressing DENV-3 DIIIs from different genotypes after incubation with 20 μg/ml of MAb. Data are the averages of results from four independent experiments and are normalized to the percentage of cells that were positive after staining with the poorly neutralizing MAb DENV-3-E8. Boldface indicates that the level of binding of a given MAb was from 0 to 25% of the control value, and underlining indicates that the level of binding was from 26 to 50% of the control value.

On the basis of studies that mapped DIII-specific MAbs against WNV, DENV-1, and DENV-2 (15, 16, 28, 37, 49, 53), we engineered mutations (L301T, S302N, S303A, V305G, L306G, K308E, E309K, K325E, D328G, D328N, A329G, A329T, A329V, P330G, G340N, A384Q, L385A, K386E, K386N, I387A, W389G, and W391A) on residues of the N-terminal linker, BC and FG loops, and the A and G β-strands of DIII from strain 16652 (gI) and displayed these variants on the surfaces of yeast cells. DIII-specific MAbs were screened for loss of binding to the mutants to identify potentially critical recognition residues (Fig. 4 A and Table 5). Using this strategy, we localized the epitopes of the different classes of neutralizing MAbs against DENV-3. For example, the complex-specific MAb DENV-3-E77 showed loss of binding when residues in the A strand (L306), BC loop (A329 and P330), and G strand (L385) were altered. For subcomplex-specific MAbs (e.g., DENV-3-E61), changes in the A strand (L306, K308, E309) and G strand (L385 and W389) diminished or abolished binding. For the strongly inhibitory type-specific MAbs (e.g., DENV-3-E54), substitutions in the BC loop (D328, A329, and P330) abrogated binding to DIII on yeast (Fig. 5 A). Due to the large number of MAbs that lost binding to the D328G or P330G DIII mutant when DIII was displayed on yeast cells, we independently confirmed this phenotype by performing binding assays with wild-type DIII or a mutant (D328G or P330G), recombinant, bacterially expressed DIII (Fig. 4B). With the exception of DENV-3-E77, all MAbs showed binding patterns similar to those of mutants containing the D328G and P330G mutations when DIII was displayed on the surfaces of yeast cells or as a recombinant protein in an enzyme-linked immunosorbent assay.

FIG. 4.

Epitope localization of anti-DENV-3 MAbs to DIII of the E protein. Yeast expressing wild-type E protein and DIII point mutation (S302N, L306G, E309K, D328G, W389A) strains of DENV-3 were incubated with the indicated DENV-3 (DV3) MAbs and analyzed by flow cytometry. Histograms are shown for the MAbs WNV E16 (negative control), DENV-3-E8, DENV-3-E61, DENV-3-E77, and DENV-3-E47, with yeast variants. (B) ELISA with plate-bound wild-type, D328G, or P330G DENV-3 DIII protein expressed in bacteria after detection by DENV-3 MAbs. Values are the averages of results from two independent experiments performed in duplicate. O.D. 450 nm, optical density at 450 nm; wE111, a cross-reactive MAb against WNV that binds all DENV serotypes and was used as a control.

TABLE 5.

Summary of MAb binding to DENV-3 DIII variants expressed on the surfaces of yeast cells^a

MAb	% of control value for variant with:
	L301T	S302N	S303A	V305G	L306G	K308E	E309K	K325E	D328G	D328N	A329G	A329T	A329V	P330G	G340N	A384Q	K386E	K386N	L385A	W389G	W391A
DENV-3-E77	87.0	102.6	94.5	98.4	19.6	123.9	101.1	93.8	89.9	93.3	2.4	95.9	94.4	0.4	91.9	67.0	95.0	60.6	23.6	50.7	78.5
DENV-3-E61	111.2	104.5	94.5	62.9	1.4	15.8	16.0	49.1	1.1	1.6	2.4	87.6	116.7	1.3	105.8	72.3	75.6	77.2	2.3	8.3	27.8
DENV-3-E48	80.1	80.1	55.2	64.5	4.2	95.0	70.7	61.6	3.6	1.3	6.3	84.7	61.7	2.3	54.3	106.9	0.9	86.8	4.9	71.2	52.7
DENV-3-E47	123.0	8.5	90.2	93.9	10.4	107.5	88.4	71.7	1.5	1.9	17.9	95.9	103.1	9.6	101.7	100.0	60.7	91.2	32.6	88.6	84.9
DENV-3-E50	112.4	102.6	84.7	97.1	51.3	120.0	91.7	67.9	3.3	2.5	13.7	83.5	100.6	4.8	87.9	102.0	5.6	99.1	53.9	93.2	67.7
DENV-3-E51	80.7	98.1	66.1	93.1	18.6	112.5	86.2	67.9	3.3	3.5	6.2	80.6	58.0	3.0	52.6	88.1	1.7	92.1	27.4	87.7	72.0
DENV-3-E49	98.1	96.8	85.8	98.8	28.3	111.3	99.4	113.2	16.4	27.3	56.2	84.7	91.4	6.1	90.2	103.0	2.7	97.4	44.7	97.7	78.5
DENV-3-E54	118.3	133.7	106.6	93.9	41.6	103.8	91.7	65.4	6.8	19.1	4.8	116.5	102.8	3.9	111.8	104.0	2.8	98.2	51.1	90.4	71.0
DENV-3-E53	93.2	3.3	106.6	76.7	4.6	92.5	84.0	54.1	4.2	2.7	33.8	115.6	100.0	1.9	105.8	79.2	43.8	83.3	13.6	71.2	58.1
DENV-3-E57	120.5	97.4	86.9	80.0	11.5	97.5	80.7	65.4	16.4	23.7	4.8	75.3	106.8	2.2	80.9	96.0	1.3	89.5	32.9	84.9	73.1
DENV-3-E56	123.0	57.7	95.1	84.1	2.5	93.8	79.6	54.1	0.2	0.3	3.1	81.8	110.5	1.3	97.1	80.2	57.7	91.2	17.1	76.7	66.7
DENV-3-E58	113.0	0.8	88.0	79.2	2.6	88.8	79.6	57.9	0.6	0.2	2.6	78.8	109.9	2.0	84.4	83.2	53.7	89.5	15.4	74.0	64.5
DENV-3-E59	101.9	96.8	86.9	70.2	1.7	85.0	69.6	49.1	1.7	1.4	3.4	75.9	95.7	1.3	94.2	76.2	0.7	79.8	5.1	67.6	60.2
DENV-3-E60	100.0	92.3	76.0	92.2	17.6	103.8	82.9	55.3	15.1	18.6	10.3	74.1	86.4	1.7	78.6	86.1	2.6	85.1	36.5	86.8	73.1
DENV-3-E62	118.0	92.9	95.1	75.1	22.1	91.3	81.8	60.4	18.7	23.7	4.8	78.8	96.3	3.0	85.5	87.1	2.3	88.6	31.1	84.0	66.7
DENV-3-E63	105.0	94.2	90.2	84.1	10.5	86.3	65.2	49.1	19.3	25.3	4.9	72.9	110.5	1.3	91.9	93.1	1.1	83.3	22.8	76.7	55.9
DENV-3-E64	114.3	0.9	86.9	50.6	1.2	80.0	50.8	32.7	3.7	0.5	2.4	70.6	74.7	1.0	69.4	75.2	1.0	82.5	3.7	59.4	61.3
DENV-3-E67	99.4	84.0	84.2	97.1	6.0	95.0	87.3	52.8	1.2	3.7	2.6	74.7	92.6	1.4	78.6	91.1	0.4	88.6	35.6	79.5	67.7
DENV-3-E1	100	115	111	104	104	102	106	115	115	94	102	69	105	120	0.5	106	118	93	111	98	111
DENV-3-E84	89.4	93.6	94.5	24.8	0.1	83.4	39.6	56.8	0.1	0.1	0.2	90.6	80.2	0.1	0.1	57.1	75.2	41.6	0.1	0.1	10.1
DENV-3-E97	92.5	103.8	101.6	35.5	0.4	96.9	50.5	77.8	0.2	0.3	0.4	99.4	95.1	0.4	0.2	66.0	85.1	59.7	0.2	5.0	52.7
DENV-3-E98	105.6	99.3	107.6	58.1	0.4	79.8	71.4	67.9	7.6	0.4	10.3	ND	ND	13.4	102.4	85.7	94.1	77.9	0.4	37.0	78.5
WNV E111	94	69	26	2	60	15	35	2	2	4	4	7	92	50	68.3	58	82	2	4	9	10

^aStaining of yeast cells that express DENV-3 16652 DIII with single point mutations. Yeast cells were stained with 20 μg/ml of MAb. Data represent the averages of results from four to six independent experiments. Data are normalized to the percentage of cells that stained positive for DENV-3-E8. Boldface indicates that the level of binding of a given MAb was from 0 to 25% of the control value, and underlining indicates that the level of binding was from 26 to 50% of the control value.

FIG. 5.

Structural analysis of the effects of sequence diversity in DIII on the neutralization of DENV-3. (A) Localization of neutralizing epitopes on DENV-3 DIII as determined by yeast surface display. Ribbon diagrams of the DENV-3 DIII structure (Protein Database identifier 1UZG) (33) are shown with amino acid residues that affect the binding of neutralizing MAbs colored as follows: blue, residues involved in neutralization by complex-specific neutralizing MAbs; magenta and blue, residues involved in neutralization by subcomplex-specific MAbs; and orange and blue, residues that alter binding of type-specific neutralizing MAbs. K386 mutants sometimes reduced the binding of type-specific MAbs and, thus, are grouped with the subcomplex MAbs in this figure. The disulfide bond between positions 300 and 331 is highlighted in yellow. (B) Ribbon diagram of DENV-3 DIII showing amino acid sequences that vary among the different DENV-3 genotypes, with solvent-inaccessible and -accessible residues depicted in cyan and green, respectively. (C) Ribbon diagram of DENV-3 DIII showing amino acid variation of DIII from all 139 available unique DENV-3 strain sequences, with solvent-inaccessible and -accessible residues depicted in cyan and green, respectively.

Structural analysis of genotypic variation.

Based on the DENV-3 strains used in this study, there are 11 sites of genotypic variation in DENV-3 DIII, four of which are conservative and seven are nonconservative (Fig. 2A and 5B). Four sites of variation in DIII are not predicted to be solvent accessible in the mature virion structure; they are located in the B strand (V/I322), C-C′ loop (G/V340), E strand (E/D368), and C-terminal linker to the stem anchor region (R/K391). In contrast, the N-terminal linker contains three sites of variation, all of which are nonconservative and solvent exposed (L/T/S301, N/S/G302, T/A/S303). The BC loop and FG loop contain three sites of nonconservative, solvent-accessible variation (A/V329 and I/T380 [BC loop] and K/N383 [FG loop]). The final site of variation is solvent accessible and conservative in the G strand (K/R386). Of note, intragenotypic variation is observed at positions 302, 303, and 391 within the gI strains, position 368 within the gII strains, and position 329 within the gIII strains.

Because of the variation in the intragenotypic neutralization pattern (Table 3), we evaluated the overall amino acid variation among DENV-3 strains independently of genotype. To determine the conservation of amino acid residues within DIII, we aligned all available DENV-3 envelope sequences in the NCBI database. Sequences were trimmed and duplicates eliminated, and the resulting 139 unique DENV-3 DIII sequences were analyzed for conservation at each amino acid position. Using this approach, we independently confirmed variation at positions 301, 302, 303, 329, 380, 383, 386, and 391 (Fig. 5C and Table S3 in the supplemental material). Three additional sites (345, 360, and 377) of amino acid variation were identified and mapped on the DENV-3 DIII structure. Residue 360 (DE loop) is the sole solvent-accessible amino acid, as residues 345 (C-C′ loop) and 377 (F strand) are not accessible on the mature virion.

In vivo protection against homologous and heterologous DENV-3 challenge.

Preexposure passive transfer of neutralizing MAbs against DENV-1, DENV-2, and DENV-4 protects against infection in mice (2, 21, 49, 54). To confirm that neutralizing anti-DENV-3 MAbs protect in vivo, we developed a lethal infection model for the homologous DENV-3 16652 strain. This was not straightforward, as DENV-3 strains replicate poorly in wild-type and immunodeficient mice. To our knowledge, only one mouse-passaged strain (H87, gI) has been reported to cause subtotal lethality in mice after intracranial inoculation (9). To generate a more virulent mouse-adapted strain, DENV-3 16652 was passaged twice between IFN-α/βR^−/− × IFN-γR^−/− AG129 mice after intracranial infection with C6/36 insect cells. The resultant isolate caused 100% lethality in AG129 mice after i.c. infection (Fig. 6). Although this DENV-3 challenge model does not recapitulate human DENV disease, it serves as a stringent assay to define highly protective antibodies because lower concentrations of MAb cross the blood-brain barrier and accumulate in the central nervous system (20).

FIG. 6.

Prophylactic and therapeutic efficacy of strongly neutralizing antibodies in mice after DENV-3 infection. (A) Mice were administered a single 500-μg dose of the indicated MAbs via the i.p. route 1 day prior to infection and then i.c. infected with 5 × 10³ PFU of the mouse-adapted DENV-3 strain 16652 and monitored over 60 days for survival. The left and right panels show MAbs with different levels of protection. (B) The therapeutic efficacy of the strongly protective neutralizing antibodies was tested by administering a single 500-μg dose of MAb i.p. 2 days after infection with 5 × 10³ PFU i.c. (C) Mice were administered a single 500-μg dose of the indicated MAb via an i.p. route 1 day prior to i.c. infection with 10⁵ PFU of the mouse-adapted DENV-3 UNC3050 strain and monitored over 30 days for survival. The survival analysis represents the results from two to six independent experiments with at least 4 mice per group.

A single 500-μg dose of a control anti-WNV MAb (E16, IgG2b) 1 day prior to infection had no effect, as expected, with 6% survival (1 of 18 mice) and a mean time to death (MTD) of 14 days (Table 6 and Fig. 6A). Three MAbs (DENV-3-E47, DENV-3-E48, DENV-3-E52) showed little, if any protection in this model. For DENV-3-E47, this was quite surprising as the neutralization potency was strong, with EC₅₀ values of ∼110 ng/ml against the parent strain (Table 2). Due to the disparity between the EC₅₀ values and protection, the neutralization potential of DENV-3-E47, DENV-3-E48, and DENV-3-E52 was confirmed using the mouse-adapted 16652 strain; notably, no difference in EC₅₀ value from that for the parental virus was observed (data not shown). In comparison, 13 MAbs protected significantly, and these were separated into two (moderate- and high-level) groups. DENV-3-E49, DENV-3-E51, DENV-3-E53, DENV-3-E54, DENV-3-E58, DENV-3-E61, DENV-3-E62, and DENV-3-E64 protected between 20 and 40% (P < 0.05) of AG129 mice, whereas DENV-3-E50, DENV-3-E57, DENV-3-E59, DENV-3-E60, and DENV-3-E63 protected between 60 and 83% (P < 0.005) of animals from lethal infection. Surprisingly, and in contrast to that seen with WNV, DENV-1, and DENV-2 (37, 39, 49, 52), we observed no clear relationship between neutralizing potency in vitro and protective activity in vivo.

TABLE 6.

Preexposure prophylaxis with neutralizing MAbs against challenge of AG129 mice by DENV-3 16652 (gI)^c

MAb	No. of mice that survived	No. of mice that died	Total no. of mice	% Survival	MTDa	±SD	P valueb
DENV-3-E47	0	7	7	0	15	4.4	>0.1
DENV-3-E48	4	11	21	33	18	8.6	<0.0005
DENV-3-E49	2	3	5	40	24	5.1	<0.001
DENV-3-E50	9	6	15	60	23	6.6	<0.0001
DENV-3-E51	1	4	5	20	23	3.2	<0.0005
DENV-3-E52	0	5	5	0	12	4.7	>0.5
DENV-3-E53	2	3	5	40	22	5.9	<0.001
DENV-3-E54	9	6	15	46	13	9.6	>0.1
DENV-3-E56	2	2	4	50	18	1.4	<0.005
DENV-3-E57	5	3	8	63	23	5.0	<0.0001
DENV-3-E58	1	3	4	25	18	9.2	<0.05
DENV-3-E59	10	2	12	83	14	4.9	<0.0001
DENV-3-E60	3	2	5	60	13	8.5	<0.01
DENV-3-E61	8	6	17	47	17	6.8	<0.0001
DENV-3-E62	2	3	5	40	19	8.2	<0.01
DENV-3-E63	7	2	9	78	10	1.4	<0.005
DENV-3-E64	2	3	5	40	23	2.1	<0.0005
DENV-3-E67	4	2	6	67	21	2.8	<0.005
WNV E16	1	17	18	6	14	3.3

^aMTD, mean time to death.

^bP values were calculated using the log rank test by comparing values for the negative-control MAb (WNV E16) and antibody-treated mice.

^cFour- to 5-week-old AG129 mice were passively administered 500 μg of the indicated MAbs 1 day before infection with 5 × 10³ PFU of strain 16652 by an i.c. route. Mice were monitored for survival for 60 days after infection.

Based on the prophylaxis studies, the four most protective MAbs (DENV-3-E48, DENV-3-E57, DENV-3-E59, and DENV-3-E63) were selected for evaluation in a postexposure therapeutic model. A single 500-μg dose of MAb was transferred passively 2 days after i.c. infection with DENV-3 16652, and survival was monitored. Two MAbs (DENV-3-E57 and DENV-3-E63) showed an ∼75% survival (P < 0.001) rate in this model (Fig. 6B and Table 7). To determine the possible utility of our protective MAbs as a therapeutic against genetically diverse DENV-3 strains, we developed a heterologous DENV-3 challenge model by generating a mouse-adapted UNC3050 (gIV) strain after passaging it between AG129 mice and C6/36 cells. AG129 mice were passively transferred a single 500-μg dose of DENV-3-E47, DENV-3-E60, and DENV-3-E61, the three MAbs that showed modest cross-neutralization of UNC3050 in cell culture, prior to i.c. challenge with 10⁵ PFU (Fig. 6C). However, none of the three MAbs protected against the adapted heterologous UNC3050 virus in this stringent challenge model.

TABLE 7.

Postexposure therapy with neutralizing MAbs after challenge of AG129 mice with DENV-3 16652 (gI)^c

MAb	No. of mice that survived	No. of mice that died	Total no. of mice	% Survival	MTDa	±SD	P valueb
WNV E16	0	12	12	0	14	3.4
DENV-3-E48	0	7	7	0	19	9.7	>0.1
DENV-3-E57	8	3	11	73	14	4.0	<0.001
DENV-3-E59	3	7	10	30	14	5.6	>0.1
DENV-3-E63	10	3	13	77	20	1.5	<0.0001

^aMTD, mean time to death.

^bP values were calculated using the log rank test by comparing values for the negative-control MAb and antibody-treated mice.

^cFour -to 5-week-old AG129 mice were administered 500 μg of the indicated MAbs 2 days after infection with 5 × 10³ PFU of strain 16652 by the i.c. route. Mice were monitored for survival for 60 days after infection.

DISCUSSION

In two recent studies, we evaluated large panels of MAbs against DENV-1 and DENV-2 for their neutralizing activity, epitope-specific binding patterns, and in vivo protective capacity (49, 52). Several strongly neutralizing MAbs failed to inhibit infection of DENV-1 strains of a heterologous genotype, suggesting that variable antibody neutralization may be due to the limited amino acid dissimilarity that occurs within a DENV serotype. For studies with DENV-2, we identified a similar trend, as 4 of 18 strongly neutralizing MAbs failed to inhibit in vitro infection of a heterologous genotype. These data suggested that it may be important to assess whether antibody responses generated against an individual strain of a given serotype neutralize infection by heterologous genotypes effectively. As this concept had implications for DENV vaccine development and evaluation, here, we assessed the cross-genotype neutralizing and protective activities of an extensive panel of newly generated anti-DENV-3 MAbs. Twenty-four of the 82 MAbs strongly neutralized the homologous virus strain (DENV-3 16652), and 20 of these mapped to epitopes in DIII of the E protein. Remarkably, only two of the neutralizing MAbs efficiently recognized cells infected with all four DENV-3 genotypes, with many showing reduced binding to multiple genotypes. None of the strongly inhibitory MAbs efficiently neutralized all DENV-3 isolates used in this study, and the majority inhibited only strains classified as gI and gII. Although we identified MAbs with similar functional characteristics within the panels of MAbs against DENV-1 and DENV-2 (49, 52), the extent of the genotype-specific neutralization was greatest for the DENV-3 panel. Consistent with this, several MAbs were highly protective in mice as prophylaxis or postexposure therapy in an i.c.-challenge model of homologous (gI) DENV-3 infection, yet none of the tested MAbs protected against infection by a heterologous (gIV) DENV-3 strain.

Although there is significant genetic variation throughout the genomes of strains of the DENV-3 serotype, the diversity is greatest within the E protein (5, 23, 26, 57), the target of the majority of neutralizing MAbs. The DENV-3 serotype is segregated into four different genotypes, with gI, gII, and gIII causing the majority of contemporary human infections (1). Within the DENV-3 serotype, one area of genetic variation among natural viral isolates is in the E protein along the lateral-ridge epitope of DIII (58). This region, which includes amino acids in the N-terminal linker and BC loop, has been identified as a key recognition site for MAbs that potently neutralize the infection of other flaviviruses, including WNV (4, 37, 46), DENV-1 (49), DENV-2 (15, 53), tick-borne encephalitis virus (50), and Japanese encephalitis virus (14, 63). The lateral-ridge epitope on DENV-3 DIII was the dominant site of recognition for the majority of the strongly neutralizing mouse MAbs in our study and also for two mouse MAbs reported by others (58). With analysis of 82 new MAbs, our experiments confirm and extend the results of recent studies which characterized five anti-DENV-3 MAbs, including two type-specific neutralizing antibodies and one subcomplex-specific neutralizing antibody (58). Consistent with their results and the structural variation of DIII among strains at positions along the lateral-ridge epitope, none of our strongly type-specific MAbs efficiently recognized and inhibited DENV-3 from all four genotypes. In comparison, only one subcomplex-specific MAb (14A4) from the prior study, which recognized a residue (K308) in the A-strand epitope, neutralized all four genotypes, albeit with rather modest EC₅₀ values ranging from 0.25 to 2.2 μg/ml, values that were 100- to 1,000-fold less potent than many of the MAbs described in our study.

Two of the subcomplex-specific neutralizing MAbs (DENV-3-E61 and DENV-3-E77) described here localized to additional amino acids, 306 and 385, on the A and G β-strands of DIII. These regions were described as recognition sites for other subcomplex- or complex-specific neutralizing MAbs against DENV (16, 27, 30, 31, 42, 53, 56). It is interesting to note that these two MAbs, despite binding multiple DENV serotypes, still failed to neutralize efficiently all of the genotypes within the DENV-3 serotype. Thus, future evaluation of therapeutic anti-DENV MAbs should consider efficacy testing against all genotypes within a DENV serotype, even if the MAb retains the ability to bind and neutralize heterologous DENV serotypes.

To evaluate the protective activity of the neutralizing MAbs, we developed two new lethal-challenge mouse models using homologous gI and heterologous gIV strains. This model, which does not reflect DENV-3 pathogenesis in humans, nonetheless is particularly stringent for protection, as immunocompromised AG129 mice are challenged by an i.c. route and only a small fraction (∼0.1%) of IgG crosses the blood-brain barrier in rodents during flavivirus infection (36). This challenge route was necessary, as virtually all DENV-3 isolates replicated poorly in mice after peripheral (subcutaneous, intraperitoneal, or intravenous) inoculation. Our challenge model is similar to one that was used recently to monitor the effect of protective polyclonal antibodies against DENV-3 after DNA plasmid vaccination (9). Against the homologous DENV-3 genotype, two of our neutralizing MAbs significantly protected when they were used as pre- or postexposure therapy, the latter even 2 days after i.c. challenge. However, neither of the most protective MAbs (DENV-3-E57 or DENV-3-E63) had activity against the gIV strain, even as prophylaxis. These results are analogous to what was observed with several strongly neutralizing anti-DENV-1 MAbs generated against a genotype 2 isolate, which failed to protect against challenge by a heterologous genotype 4 strain (49).

Our functional studies with anti-DENV-3 MAbs potentially have implications for evaluating candidate tetravalent DENV vaccines. It is widely believed that a successful DENV vaccine requires a robust and durable neutralizing antibody response against all four serotypes (60). This is essential because subneutralizing antibody responses against any single serotype could, in theory, facilitate antibody-dependent enhancement and enhanced disease severity upon infection with a natural isolate (17). Our data with MAbs against DENV-3, along with recent and earlier studies on DENV-1 (49) and DENV-2 (18), demonstrate variable neutralization among strains of a given serotype, likely because of the sequence diversity of the E protein between genotypes. A key question is whether the genetic diversity across DENV strains in nature will affect the neutralization potential observed during vaccine trials in the context of a polyclonal response, especially with the less robust or durable neutralizing antibody responses that occur after immunization with attenuated strains or recombinant proteins. As neutralizing activity in polyclonal antibody responses in humans appears to be serotype specific (59), there is at least a possibility of incomplete neutralization of specific genotypes within a serotype, especially over time as immunity wanes. Thus, it may be important to evaluate candidate vaccines for neutralizing antibody responses against an array of viruses within a serotype encompassing genotypic variation to confirm the efficacy and breadth of a protective antibody response.

Our results suggest that the lateral-ridge epitope of DIII of the flavivirus E protein is a target for strongly neutralizing antibodies produced, but sequence variation within that epitope can limit the breadth of activity of antibodies that target that region. Moreover, the potency of antibody neutralization generally correlated with narrowed genotype specificity. The identification of key target determinants for antibody protection against all DENV serotypes and genotypes ultimately may allow the generation of epitope-based diagnostic reagents that can better predict the qualitative, quantitative, and functional profiles of polyclonal antibody responses within humans during natural infection or after vaccination.