Isolation of human monoclonal antibodies to DENV

In order to better characterize the human antibody response to DENV, we isolated B cell clones from several independent cultures and characterized 70 DENV-reactive monoclonal antibodies (mAbs) (68 IgG1, either κ or λ, 1 IgG3, λ, and 1 IgG4, λ). All DENV-reactive mAbs bound to one or more DENV serotypes, as shown by staining of Vero cells infected with DENV-1, DENV-2, DENV-3, or DENV-4 (see examples in Figure 1A). Specificity and cross-reactivity were assessed using binding assays (ELISA with recombinant E, NS1 and NS3 proteins, staining of yeast displaying DI/DII or DIII, or Western blot of infected cell lysates) (see examples in Figure 1B–D). Mabs were also characterized using functional assays (neutralization or enhancement of infection by DENV vaccine strains using flow cytometry assays with Vero and K562 cells, respectively (Kraus et al., 2007; Littaua et al., 1990)) (see examples in Figure 1E, F).

Open in a separate window

Figure 1

Specificity and function of human mAbs against DENV

(A) Vero cells, uninfected or infected by DENV-1, −2, −3, and −4 vaccine strains, were fixed, permeabilized, and stained with human mAbs, followed by fluorescently-labeled anti-human IgG polyclonal antibodies. Representative dot plots show staining by mAbs isolated from Donor 76 reactive against DENV-3 alone (DV79.3), DENV-1 and DENV-3 (DV63.1) or cross-reactive against the four DENV serotypes (DV82.11, DV69.6, DV59.3). Numbers in quadrants indicate the percentage of positive cells. The data shown are for five mAbs and are representative of data obtained with 70 DENV-reactive mAbs and of at least two independent experiments. (B–D) Western blot analysis of mAbs that were not reactive by ELISA. Concentrated supernatants of uninfected control (−) or DENV-3-infected Vero cells (+) were separated under non-reducing conditions. B) E- and capsid-specific mAbs. E protein was identified by DV55.1, a mAb specific for E protein in ELISA. Capsid migrate with apparent molecular weights of 16 kD. C) shows prM-specific mAbs. PrM protein was identified by the mouse anti-prM mAb 2H2. D) shows NS1-specific mAbs. NS1 protein was identified by a rabbit anti-NS1 serum. The data are representative of at least two independent experiments. (E) Examples of viral neutralization assay in which serial dilutions of mAbs (DV63.1 and DV5.1) were incubated with DENV-1-4 before addition to Vero cells. Shown is the percentage of infected cells after 3 days as a function of increasing mAb concentration. Nonlinear regression analysis was used to calculate the EC₅₀, values indicated in the panel. (F) Infection enhancement assay in which serial mAb dilutions were incubated with DENV-1-4 before addition to K562 cells. Shown is the percentage of infected K562 cells. The range of mAb concentrations where infection enhancement was observed is indicated in gray. Numbers in the gray areas represent logarithm power. The data shown are from one representative experiment out of three independent experiments performed. Data obtained with the panel of DENV-reactive mAbs are shown in Tables 2, ​,33 and ​and44.

Serotype-specific and cross-reactive mAbs to DIII of E-protein

Of the 70 DENV-reactive mAbs isolated, 13 mapped to DIII of E protein and of these, 5 were serotype-specific, whereas 8 were cross-reactive with two, three or even all four DENV serotypes (Table 2). Of the five serotype-specific mAbs, three (DV3.7, DV25.5, and DV470.12) potently neutralized infection of Vero cells by either DENV-1 or DENV-2 vaccine strains, with EC₅₀ values in the low ng/ml range, whereas two mAbs (DV1.6 and DV14.5) showed decreased potency against DENV-2, suggesting that they may represent low affinity antibodies or may target private epitopes within DENV-2. Of the 8 cross-reactive mAbs, two (DV63.1 and DV55.1) potently neutralized DENV-1 and DENV-3, whereas mAb DV87.1 potently neutralized DENV-1, DENV-2 and DENV-3. Strikingly, the cross-reactive mAbs DV63.1 and DV87.1 neutralized infection in the same range of concentrations as observed for serotype-specific mAbs (Table 2). Finally, four mAbs bound all serotypes by ELISA and of these, two neutralized, albeit with low potency, all serotypes (DV21.5 and DV257.13), whereas the remaining mAbs showed a more restricted pattern of neutralization.

Consistent with what is known about the stoichiometric relationship between neutralization and enhancement (Pierson et al., 2007), DIII-specific mAbs also enhanced infection of K562 cells over a given range of concentrations with the same serotype specificity as shown in the neutralization assay (Table 2).

These results indicate that it is possible to isolate DIII-reactive human mAbs that potently neutralize infection of two or more DENV serotypes.

DI/DII-specific mAbs from secondary DENV infection cross-react with all four DENV serotypes

Of the 70 DENV-reactive mAbs, 34 mapped to DI/DII of E protein (Table 3). Of these, 8 were serotype-specific, 1 cross-reacted with DENV-1 and DENV-2, and 25 cross-reacted with all four DENV serotypes. When compared to DIII-specific mAbs, the DI/DII-specific mAbs showed in general a 10-50-fold lower potency of neutralization and enhanced infection over a wide range of concentrations (Table 3). Of interest, whereas DI/II-reactive mAbs isolated from primary infections (Donors 7, 13 and 76) were mostly serotype-specific, all DI/DII-reactive mAbs isolated from secondary infection (Donors 12 and 92) neutralized and enhanced infection by all four DENV serotypes. These findings suggest that broadly cross-reactive DI/DII-specific memory B cells may be selected during the course of secondary DENV infection.

Some of the broadly reactive DI/DII-specific mAbs were analyzed for reactivity against West Nile virus (WNV). Seven mAbs stained yeast cells expressing DI/DII of WNV E protein (Table S1). Using site-directed mutants in DII that had been previously generated (Crill and Chang, 2004; Oliphant et al., 2006), we localized binding of three mAbs to the fusion loop peptide at the tip of DII (W101R, G106E) and of a fourth mAb to the W233F residue on the dimmer interface of DII. The binding of the other three cross-reactive mAbs was not affected by these mutations, indicating that they map to DI/DII but outside of the fusion loop region.

Finally, 4 mAbs (DV66.1, DV57.4, DV61.2, and DV77.5), which were selected for binding to DENV-infected cells, did not bind to recombinant E proteins in ELISA but bound to E protein in Western blot (Figure 1B). DV66.1 and DV57.4 neutralized one and three DENV serotypes, respectively, whereas DV61.2 and DV77.5 did not neutralize DENV infection at the highest concentration tested. However, all 4 mAbs showed ADE activity (Table 3).

Human monoclonal antibodies to prM and to non-structural proteins

We next characterized 19 mAbs that stained DENV-infected cells but did not recognize E protein. Six mAbs reacted with a band in Western blot corresponding to prM, as indicated by the mouse anti-prM mAb 2H2 used as control (Falconar, 1999) (Figure 1C). As independent verification, we also performed cross-competition experiments and found that the anti-prM mAbs DV64.3, DV27.2, and DV69.6 inhibited binding of 2H2, suggesting that they may recognize the same or overlapping epitopes, whereas DV52.1, DV75.9, and DV62.5 did not compete with 2H2 and may recognize distinct epitopes (data not shown). Surprisingly, five of these mAbs stained a band corresponding to E-prM protein and three of them (DV64.3, DV52.1 and DV62.5) also recognized a band corresponding to the E protein. The reactivity of mAbs with both prM and E suggests a quaternary epitope with shared sites on the heterodimeric prM-E protein; a similar finding has been reported for mouse antibodies (Falconar, 1999; Gould et al., 1985; Puttikhunt et al., 2008). With the single exception of mAb DV52.1, all the prM-specific mAbs were broadly cross-reactive as they bound to cells infected with all four DENV serotypes (Table 4). All prM-reactive mAbs enhanced DENV infection over a broad range of concentrations whereas only three neutralized infection with low potency (Table 4). Thus, consistent with a recent study (Dejnirattisai et al., 2010), the prM-specific mAbs were highly cross-reactive among the DENV serotypes and potently promoted ADE.

Eleven mAbs were mapped to NS1 or NS3 protein, as determined by ELISA with recombinant proteins and Western blot (Figure 1D and data not shown). Five NS1-specific mAbs cross-reacted with all DENV serotypes, whereas the remaining two mAbs reacted with DENV-1 and DENV-3, and the four NS3-specific mAbs were all DENV-3-specific (Table 4). Finally, only one mAb (DV86.2) recognized the capsid protein (Table 4 and Figure 1B). Another mAb (DV34.4) was not assigned to a specific protein, although it showed high neutralizing activity against DENV-2 and, to a lower extent, DENV-4 (Table 4). This mAb likely recognizes a conformationally-sensitive epitope on E (e.g., hinge or interface between domains of oligomers) as was observed recently with two human mAbs against WNV (Vogt et al., 2009).

A broadly neutralizing antibody cocktail lacking enhancing activity

Given the availability of human mAbs that potently neutralized DENV infection, we set out to develop a candidate antibody-based therapeutic for dengue consisting of a “cocktail” of neutralizing mAbs. Ideally, this cocktail should: i) neutralize all DENV serotypes, ii) target at least two non- overlapping epitopes on each virus in order to minimize selection of escape mutants, and, iii) fail to enhance infection. We therefore selected two mAbs with potent and complementary neutralizing activity: DV87.1 specific for DIII of DENV-1, −2 and −3 and DV22.3 specific for DI/DII of DENV-4. As a complement we selected a DI/DII-specific mAb (DV82.11) that neutralized with comparable efficiency all four DENV serotypes (Figure 2A). Cross-competition experiments were consistent with the notion that the target epitopes recognized by the three mAbs did not overlap (Figure S1). Thus, a cocktail of DV87.1, DV22.3 and DV82.11 is expected to neutralize all DENV serotypes by targeting two distinct sites on each virus.

Open in a separate window

Figure 2

Three mAbs neutralize all four DENV serotypes by binding two distinct epitopes on each virus

Three mAbs (all IgG1) were selected for their specificity and neutralizing activity: DV87.1 binds DIII of DENV-1, −2 and −3; DV22.3 binds DI/DII of DENV-4, and DV82.11 cross-reacts with DI/DII of all DENV serotypes. (A) Neutralization (○) and enhancement (●) of infection by recombinant unmodified mAbs. (B) Neutralization (○) and enhancement (●) of infection by engineered LALA variant mAbs. Nonlinear regression analysis was used to calculate the EC₅₀, indicated in the panel. Data are mean of duplicates. One representative experiment out of three performed. (C) A cocktail of LALA-mAbs (DV82.11= 2 µg/ml, DV87.1=DV22.3= 0.2 µg/ml) was used in combination with mAbs directed against prM (DV69.6) or DI/DII E protein (DV78.6) or serum from a primary DENV-infected donor at potent enhancing concentration (0.1 µg/ml) on K562 cells. Numbers in quadrants indicate the percentage of infected cells, as revealed by staining with an anti-E mouse mAb. The data shown are from one experiment representative of two performed. See also Figure S1 and Table S2.

Because the selected neutralizing mAbs also enhanced infection (Figure 2A), we generated variants of DV87.1, DV22.3 and DV82.11 in which leucine residues at positions 1.3 and 1.2 of CH2 domain (according to the IMGT unique numbering for C-domain) were substituted with alanine residues. This modification, also known as “LALA” mutation, abolishes antibody binding to FcγRI, FcγRII and FcγRIIIa (Hessell et al., 2007). The variant and unmodified recombinant mAbs were expressed in HEK293T cells, and the purified mAbs were compared for their capacity to neutralize and enhance infection by the four DENV serotypes. As shown in Figure 2B, LALA variants retained the same neutralizing activity as unmodified mAbs, but were completely devoid of enhancing activity. In addition, the LALA cocktail, even at very low concentrations (DV82.11 = 2 µg/ml; DV87.1 = DV22.3 = 0.2 µg/ml), blocked ADE induced by enhancing prM or DI/DII mAbs or immune serum (Figure 2C).

To determine whether the combination of LALA mAbs might achieve increased neutralization we used a variable ratio approach (Zwick et al., 2001). The concentration of one mAb was varied in the presence of a fixed amount of a second mAb (added at a weakly neutralizing concentration) and the titration curves of the first mAb alone or in combination were compared. The EC₉₀ values shown in Table S2 indicate a two- to fourfold increase in neutralization potency, consistent with a modest additive effect.

Taken together, the above results suggest that a cocktail of three LALA variant mAbs targeting different epitopes on each virus can efficiently neutralize DENV infection, while avoiding and actually preventing enhancement of DENV infection in vitro.

Intracellular staining

C6/36 or Vero cells were infected with the DENV-3 recovered from Donor 76, with the DENV-1 recovered from Donor 12, with DENV-2 WHO reference strain 516803, or with DENV-1-4 vaccine strains at MOI leading to 70–90% of infected cells. After 5 days, cells were harvested, fixed with 2% formaldehyde (Sigma), and permeabilized in PBS 1% FCS 0.5% saponin (Sigma). Cells were then incubated for 1h on ice with B cell culture supernatants, washed and stained with Cy5-conjugated goat anti-human IgG (Jackson Immunoresearch) and analyzed on a FACSArray (BD Biosciences).

ELISA

96-well ELISA plates were coated at 4°C with recombinant E protein from DENV-1-4 (Rey, 2003), with recombinant DIII from the DENV-3 of Donor 76 or with DIII of DENV-4 (Simonelli et al., 2010) (all at 1 µg/ml in 0.1M Na carbonate buffer, pH 9.3). After washing and blocking, mAbs were added for 1h at 37°C. After further wash, mAbs bound were revealed using goat anti-human IgG coupled to alkaline phosphatase (Jackson Immunoresearch). In some experiments ELISA plates were coated with the 4G2 (1 µg/ml) followed by a mixture of DENV-1-4 antigens.

Yeast expression and staining

Yeast expressing DI/DII and DIII of DENV-2 E protein were generated as previously described (see Supplementary Experimental Procedures) (Oliphant et al., 2006; Sukupolvi-Petty et al., 2007). Yeast cells expressing wild type or mutant DI/DII or DIII of DENV or WNV (strain New York 1999) E protein were stained with mAbs on ice for 30 min. The cells were then incubated with a goat anti-human IgG conjugated to Cy5 (Jackson Immunoresearch) for 30 min on ice and analyzed by flow cytometry.

Immunoblot analysis

Equal amounts of concentrated supernatant from DENV-3- or DENV-2-infected or uninfected Vero cells were separated under non-reducing conditions by SDS-PAGE and transferred onto nitrocellulose membranes. Filters were blocked with 10% dry skim milk and probed with mAbs. Blots were incubated with horseradish peroxidase–conjugated mouse anti-human IgG, sheep anti-mouse IgG or donkey anti-rabbit IgG and developed with ECL Plus Western Blotting Detection System (GE Healthcare). E protein was identified using DV55.1, prM protein was identified using the mouse anti-prM mAb 2H2 (Falconar, 1999) and NS1 protein was identified using a rabbit anti-NS1 serum (Young et al., 2000). The C protein was identified as a 16-kD band.

Production of recombinant IgG and LALA variants

RNA was isolated with RNeasy kit (Qiagen) from EBV-immortalized B cell clones DV82.11, DV87.1, and DV22.3 and cDNA synthesized with M-MLV reverse transcriptase (Invitrogen). Variable regions of heavy-chain and light-chain genes were sequenced and cloned by PCR into human Igγ1, Igκ and Igλ expression vectors (provided by M. Nussenzweig, Howard Hughes Medical Institute, The Rockefeller University, New York) using gene-specific primers and Pfu Turbo (Stratagene) (Tiller et al., 2008). The vectors contain a murine Ig gene signal peptide sequence and a multiple cloning site upstream of human Igγ1, Igκ and Igλ constant regions, with either wild-type or with leucine-to-alanine mutations at positions CH2 1.3 and 1.2 of Igγ1, according to the IMGT unique numbering for C-DOMAIN (LALA mutant) (Hessell et al., 2007) introduced by site-directed mutagenesis (GenScript). Equal amounts of heavy and light chain vectors were mixed with an equal amount of polyethylenimine (Sigma), incubated for 15 min and added to 80% confluent HEK293T cells (ATCC) in DMEM supplemented with 1% Nutridoma (Roche). Antibodies were recovered from supernatants harvested 3 days after transfection and purified using protein A affinity chromatography followed by Sephadex 200 size-exclusion chromatography.

Infection neutralization and enhancement

DENV neutralization was measured using a flow based assay. The day before the infection, 5,000 Vero cells were plated in 96-well flat-bottom plates. Different dilutions of mAb were mixed with DENV (all at MOI of 0.04) for 1h at 37°C and then added to Vero cells. After three days the cells were fixed with 2% formaldehyde, permeabilized in PBS 1% FCS 0.5% saponin, and stained with 4G2. The cells were incubated with a goat anti-mouse IgG conjugated to R-phycoerythrin (Southern Biotechnology) and analyzed by flow cytometry. EC₅₀ values were determined by nonlinear regression analysis. ADE was measured by a flow assay using K562 cells. mAbs and DENV (MOI of 0.04) were mixed for 1 hour at 37°C and added to 5000 K562/well. After three days, cells were fixed, permeabilized, and stained with 4G2 and the number of infected cells was determined by flow cytometry as above.

Effect of LALA antibodies on ADE

Potent enhancing antibodies (0.1 µg/ml) or serum from primary dengue-infected donors were mixed with an equal volume of LALA cocktail (DV82.11 =2 µg/ml; DV87.1=DV22.3= 0.2 µg/ml). Attenuated DENV-1-4 virus strains (MOI of 0.04) and the antibody were mixed, added to K562 cells, and incubated for three days. Cells were stained with 4G2 and analyzed by flow cytometry as describe above.

Infection of AG129 mice

129/Sv mice lacking the interferon α/β and γ receptors (AG129) (van den Broek et al., 1995) were bred in the University of California, Berkeley Animal Facility. All experimental procedures were pre-approved by the University of California Berkeley Animal Care and Use Committee (ACUC) and were performed according to the guidelines of the UC Berkeley ACUC. Mice were injected i.p. with mAb, PBS, or anti-DENV sera in a total volume of 200 µl, then infected 18–24 hours later with either 10⁵ or 10⁶ pfu of DENV strain D2S10 diluted in 100µl volume by intravenous (i.v.) injection into the tail vein. In some experiments, 50 µg of mAb were transferred to mice i.v. 24 hours post-infection to assess the therapeutic efficacy of the mAb.

We thank Michel Nussenzweig for providing vectors to clone and express Ig genes, David Jarrossay for cell sorting, Catriona McElnea for the NS1 screening, and Ruben Lachica for his excellent technical assistance with the in vivo experiments. This research was supported in part by the Gottfried and Julia Bangerter-Rhyner-Stiftung (to F.S.), the Swiss National Science Foundation Grant N. 126027 (to A.L.), the Swiss Vaccine Research Institute (to L.V.), the Wellcome Trust (to C.P.S.), the Pediatric Dengue Vaccine Initiative (CRA14 to E.H), NIH grant R01-AI077955 (to M.S.D.), AI65359 (to E.H.) and U19-AI 057266 (to A.L.) and by the Intramural Research Program of the US National Institute of Allergy and Infectious Diseases, NIH (to S.S.W.). The Institute for Research in Biomedicine is supported by the Helmut Horten Foundation. AL is the scientific founder of Humabs LLC and a member of its Board of Directors. A.L. and F.S. are shareholders of Humabs.