Potent Dengue Virus Neutralization by a Therapeutic Antibody with Low Monovalent Affinity Requires Bivalent Engagement

We recently described our most potently neutralizing monoclonal antibody, E106, which protected against lethal Dengue virus type 1 (DENV-1) infection in mice. To further understand its functional properties, we determined the crystal structure of E106 Fab in complex with domain III (DIII) of DENV-1 envelope (E) protein to 2.45 Å resolution. Analysis of the complex revealed a small antibody-antigen interface with the epitope on DIII composed of nine residues along the lateral ridge and A-strand regions. Despite strong virus neutralizing activity of E106 IgG at picomolar concentrations, E106 Fab exhibited a ∼20,000-fold decrease in virus neutralization and bound isolated DIII, E, or viral particles with only a micromolar monovalent affinity. In comparison, E106 IgG bound DENV-1 virions with nanomolar avidity. The E106 epitope appears readily accessible on virions, as neutralization was largely temperature-independent. Collectively, our data suggest that E106 neutralizes DENV-1 infection through bivalent engagement of adjacent DIII subunits on a single virion. The isolation of anti-flavivirus antibodies that require bivalent binding to inhibit infection efficiently may be a rare event due to the unique icosahedral arrangement of envelope proteins on the virion surface.


Introduction
Dengue virus (DENV) infection in humans causes symptoms ranging from a mild febrile illness to a severe and sometimes fatal disease. Over 3.6 billion people globally are at risk for DENV infection, with an estimated 390 million infections annually and no currently approved vaccine or antiviral therapy [1]. DENV belongs to the Flaviviridae family of medically important positivestranded RNA viruses. Within the DENV serocomplex, there is significant diversity, including four serotypes (DENV-1, -2, -3, and 4) that differ at the amino acid level of the envelope (E) protein by ,25 to 40 percent and multiple genotypes within a serotype that vary by up to ,3 percent [2,3].
A humoral response against DENV infection is believed to contribute to lifelong immunity against challenge by the homologous serotype. In comparison, protection against a heterologous DENV serotype infection is more transient (,6 months to two years) [4,5], allowing re-infection and disease to occur with a heterologous serotype in hyper-endemic areas of the world. Estimates suggest that greater than 90% of severe cases occur during secondary infection with a heterologous DENV serotype, possibly because sub-neutralizing amounts of cross-reactive antibody facilitate viral entry into myeloid cells expressing Fc-c receptors, a phenomenon termed antibody-dependent enhancement of infection (ADE) [6]. Antibody-mediated protection against homologous DENV infection correlates with a neutralizing antibody response directed predominantly against the viral E protein [7]. The ectodomain of E is comprised of three domains: domain I (DI), a central nine-stranded b-barrel that connects domain II (DII), which contains the fusion peptide at its distal end, and an immunoglobulin-fold like domain III (DIII) [6][7][8][9]. Although neutralizing antibodies have been mapped to all three domains of the E protein, many potently inhibitory anti-DENV mouse MAbs map to DIII [10][11][12][13][14][15][16][17], specifically to the lateral ridge or A-strand epitopes, and block flavivirus infection at a postattachment stage, likely by preventing E protein dimer-to-trimer transitions that are required for viral fusion [11,15,18].
We recently described a highly therapeutic MAb, DENV-1-E106 (hereafter termed E106), which neutralized infection of strains corresponding to all five DENV-1 genotypes and protected against lethal DENV-1 infection when administered as a single dose even four days after virus inoculation [13]. To understand the basis for the potency (plaque reduction neutralization titer (PRNT 50 ) of 0.6 ng/ml) [13] and specificity of this MAb, we solved the crystal structure of E106 Fab in complex with DENV-1 E DIII. Our analysis revealed a small antibody-antigen interface with contact residues corresponding to two previously characterized DIII epitopes. Remarkably, a ,20,000-fold disparity in neutralization by intact IgG and Fab correlated with distinct abilities to bind intact virions. Our results are consistent with a model in which our most potently inhibitory and therapeutic DENV-1 MAb requires bivalent binding through dual and simultaneous engagement of two antigen binding sites on a single virion to neutralize infection. E106 is therefore one of only a few unique antibodies described to date where effective neutralization requires a bivalent binding mechanism, and is the first such characterized MAb directed against flaviviruses.

Results
The crystal structure of E106 in complex with DIII reveals a small footprint E106 is a sub-complex specific therapeutic MAb that binds to DENV-1 and DENV-4 infected cells and neutralizes infection of all five DENV-1 genotypes efficiently (EC50 ranging from 1 to 50 ng/ml), without neutralizing DENV-2, 3 and 4 serotypes or WNV [13]. The 2.45 Å crystal structure of E106 Fab bound to DIII revealed only nine contact residues, from the A-strand (K307, K310), the end of the B-strand (K325, Y326), and the connecting BC (E327, T329, D330) and DE (K361, E362) loops ( Fig. 1A and Table 1); these results are consistent with prior mapping data by yeast surface display, which implicated five of these residues as essential recognition determinants (K310, T329, D330, K361 and E362) [13]. Yeast surface display results also implicated G328, P332, and P364 in E106 binding, and mutation of any of these residues would likely result in an altered presentation of the direct contact residues. Charge reversals at either E384 or K385 in the FG-loop also diminished E106 recognition, and this loop is adjacent to the primary E106 epitope. Overall, the contact residues contributed 24 van der Waal interactions, 14 hydrogen bonds, and 10 electrostatic interactions to the interface ( Table  S1). The E106 structural footprint represents a unique composite of previously identified DIII-specific neutralizing epitopes on flaviviruses including the lateral ridge (N-terminal region, BC, DE and FG loops) [19,20] and A-strand epitopes [10,21]. DENV-1 DIII was engaged by 11 heavy chain residues from CDR1 (I30, G31, Y32 and Y33), CDR2 (N52, E50, and R53), and CDR3 (R95, I196, N97 and W98) (Fig. 1B, top panel) and four light chain residues from CDR1 (D30, D32), CDR2 (E50), and CDR3 (L94) ( Table S1).
A comparison of the DIII structure in complex with Fab versus unbound DIII revealed small conformational differences, with a root mean square displacement of 0.9 Å in the a-carbons over 98 residues. Of the DIII residues that directly interacted with the E106 Fab, the greatest differences in a-carbon positions involved residue T329 (1.2 Å ); this was significant because a recently identified E106 MAb neutralization escape mutant showed a T to A amino acid change at position 329 (Fig. 1B, bottom panel) and [22]).
All nine DIII contact residues were conserved in the five DENV-1 genotypes (K361 is replaced by the conservative substitution R361 in genotype 5 strains); this likely explains why E106 neutralized infection of all five DENV-1 genotypes efficiently ( Fig. 1D and 1E), and [13]). In comparison, only one of the nine contact residues (Y326) was conserved in DENV-2, DENV-3, DENV-4 or WNV, a finding that is consistent with virological data showing that DENV-1-E106 MAb neutralizes infection in a serotype-specific manner [13]. E106 binds to DENV-4 but not to DENV-2 or DENV-3 [13]. The number of conserved contact residues did not correlate with DENV serotype binding specificity (DENV-4 has four whereas DENV-2 and DENV-3 each have five, Fig. 1E), which may instead be accounted for by other factors, including differential maturation [27] or relative virion dynamics [28].

E106 Fab binds DIII with micromolar monovalent affinity
To determine the significance of the small buried interface of our E106-DIII complex, we investigated E106 binding to DIII by surface plasmon resonance (SPR). Increasing concentrations of purified DENV-1 DIII monomer were flowed over immobilized

Author Summary
Dengue virus (DENV) is a globally important mosquitotransmitted human pathogen for which there is no approved vaccine or antiviral therapy. In recent years, the number and severity of DENV human infections have increased due to the expanded geographic range of the virus. Neutralizing antibodies are a key component of a protective natural and vaccine-induced immune response against human DENV infections. One recently described monoclonal antibody (E106) protects mice against infection of DENV-1 when administered before or several days after virus infection. Because of these results, we investigated the mechanism of action of E106 using a combination of structural and functional approaches. E106 engaged an epitope on domain III of the viral envelope protein that is a composite of two previously described epitopes. Unexpectedly, and in contrast to the intact IgG, Fab fragments of E106 were ineffective at neutralizing virus; this was explained by their weak micromolar affinity for virus particles. Our results suggest that neutralization by E106, our most potently inhibitory and protective anti-DENV MAb, requires bivalent binding of adjacent DIII subunits on a single virion. Immunization strategies with intact virions that skew the selection of neutralizing antibodies to those with bivalently binding properties could augment the potency of antiviral humoral responses against DENV and other flaviviruses.
E106 Fab ( Fig. 2A). Equilibrium analysis surprisingly revealed a micromolar affinity (4.862.1 mM) for this interaction. A similar result was observed when increasing concentrations of DENV-1-DIII monomer was flowed over immobilized intact E106 MAb in the solid phase (3.260.8 mM, Fig. 2B); this experiment eliminates the possibility that papain cleavage and the removal of the E106 Fc region altered the monovalent binding parameters. Binding to the ectodomain of DENV-1 E (DI-DII-DIII) also appeared equivalently weak (1.160.1 mM). As an independent measurement of affinity, we performed isothermal titration calorimetry under similar experimental conditions as SPR by injecting DENV-1 DIII into a solution of E106 Fab. Using this method with completely unmodified proteins we again measured micromolar affinity for the E106 Fab/DIII interaction (K D = 0.762 mM) (Fig. 2C).

E106 Fab binds and neutralizes DENV-1 poorly
The micromolar monovalent affinity of the highly therapeutic E106 antibody was unanticipated in light of its picomolar inhibitory activity (462 pM or 0.660.3 ng/mL; our most potent neutralizing anti-flavivirus MAb isolated to date); as a comparison, our therapeutic DIII-specific anti-WNV MAb E16 (inhibitory activity 30 to 80 pM or 4-18 ng/mL), which has advanced to human clinical trials [29], has a monovalent affinity of 3.4 nM [30]. We hypothesized that while E106 MAb potently inhibited DENV-1 infection, Fabs should lack this activity. To test this, we compared the ability of Fab and intact IgG from E106 and E103, a lateral ridge DIII-specific DENV-1 neutralizing antibody [13], to inhibit infection. While monovalent E103 Fab showed a 114-fold decrease in neutralization potency compared to the intact bivalent IgG, the Fab of E106 Fab showed a remarkable 18,150-fold decrease in inhibitory activity when compared to intact IgG (Fig. 3A).
To investigate this observation further, we performed a virionbinding assay by ELISA. DENV-1 virions (strain 16007) were captured with humanized DIII A-strand-specific antibody (DENV-1-E50) [31] and then detected with Fab or intact IgG of E103, E106, or WNV E16 (negative control). Notably, the amount of virus detected with E106 IgG was indistinguishable from E103 IgG (Fig. 3B). In comparison, E106 Fab bound virions to a significantly lower level (P,0.0001) at all concentrations tested, than those derived from E103 (Fig. 3B). Thus, disparate neutralization of E106 MAb and Fab correlated with discordant binding to DENV-1 virions and was consistent with the biophysical measurements: monovalent binding by DENV-1-E106 is surprisingly inefficient given the potent inhibitory activity of intact antibody.

E106 MAb binds bivalently to DENV-1 virus particles
Based on these experiments, we hypothesized that efficient neutralization of DENV-1 infection by E106 required bivalent binding. Using bio-layer interferometry (BLI), we measured the affinity and kinetics of MAb and Fab binding to intact DENV-1 virus particles [32] (Fig. 4). E106 MAb (Fig. 4A) bound DENV-1 particles with an apparent affinity (avidity), K Dapp of 1362 nM. In The immunoglobulin-like DIII is shown in blue (epitope regions in magenta), with the E106 heavy chain in green and light chain in cyan. (B) Heavy chain residue W98 binds in a deep pocket contributed by the aliphatic groups of side chain DENV-1 E DIII residues K307, K325, and E327 and main chain of Y326, in stereo (top panel). The electron density map is contoured at 1.1s. Ball and stick representation of the molecular interactions involving T329 (the residue that escapes neutralization [13] in stereo (bottom panel). (C) Surface representation of DENV-1 E DIII (top) and E106 Fab (bottom) highlighting residues making direct contacts in the complex (see Table S1). DENV-1 E DIII residues previously identified by yeast surface display are displayed in magenta [13]. (D) The structural epitope on DENV-1 DIII is shown in ribbon representation. (E) Sequence of DIII of DENV-1 aligned with that of DENV-2, -3, -4 and WNV highlighting the E106 structural epitope, which is conserved in DENV-1 genotypes but not DENV serotypes or WNV. Identical residues are represented by a dot, deleted residues by a hash. For comparison, The DIII structural epitopes of WNV E16, DENV-2 1A1D-2, DENV cross-reactive 4E11 and 2H12 MAbs and DENV-1-E111 contact residues are labeled with green, purple, light purple, orange and cyan asterisks respectively [10,17,20,21,33]. doi:10.1371/journal.ppat.1004072.g001  Table 2. DENV-1 DIII-E106 Fab structural interface. contrast, E106 Fab had an affinity of K D .1 mM, with a rapid dissociation rate (t 1/2 ,2 sec) that was at least 800-fold faster than E106 MAb (t 1/2 .400 sec) (Fig. 4B). These results contrast with more comparable binding of E103 MAb (K Dapp of 0.860.1 nM) (Fig. 4C) and E103 Fab (K D of 761 nM) to DENV-1 particles (Fig. 4D). Importantly, the binding affinities of E106 Fab engaging isolated DIII measured by SPR and ITC is remarkably similar to that observed by BLI for the binding to DENV-1 virions, suggesting that our structurally defined DIII epitope corresponds to the entire virion surface recognized by a single Fab.

Mechanistic correlates of E106 MAb neutralization
We next investigated the time-and temperature-dependence of E106 neutralization, as this analysis can provide information as to the relative accessibility of epitopes [28,33]. Changes in the time or temperature of incubation did not appreciably affect E106 neutralization ( Fig. 5A and B). By performing pre-and postattachment neutralization assays, we found that, similar to several other potently neutralizing DIII-specific antibodies against flaviviruses [15,30,34], E106 can neutralize infection even after virus attaches to cells (Fig. 5C). Finally, we tested the neutralization of E106 MAb as a function of the maturation state of the virus. DENV virions are a heterogeneous mixture of immature, partially mature and fully mature virions, with immature virions being generally less or non-infectious. In this assay, E106 MAb neutralization proved to be independent of the maturation state of the virus (Fig. 5D). In comparison, neutralization by E60, a DII fusion-loop-specific MAb, was sensitive to virion maturation, as seen previously [27].

Modeling of E106 binding to virus
In contrast to non-enveloped viruses where a structural understanding of bivalent antibody binding has emerged [35][36][37], there currently is no such data for icosahedral enveloped viruses including flaviviruses. To address how E106 might recognize DENV-1 bivalently we docked our structure onto the cryo-EM derived model of the mature DENV virion (Fig. 6) [38]. While the E106 epitope is predominantly exposed on all 180 E protein monomers (Fig. 6A), unimpeded binding is readily apparent only on monomers in the 3-fold and 2-fold symmetry axes, similar to what we observed for E16 binding to WNV [20,39]. However, minor reorientation of DIII subunits on the inner 5-fold symmetry axis would allow for up to two E106 Fabs to bind there at the same time as three Fabs could bind to the outer 5-fold (2-fold) related epitopes. We measured the distance  separating the docked Fab CH1 domain C-termini with the expectation that distances greater than 50 Å would be unlikely spanned by 16 hinge residues [36,37]. This analysis indicated the possibility for limited bivalent bridging, with the primary candidates being adjacent outer 5-fold epitopes (49 Å CH1 separation) ( Fig. 6B and D) as well as adjacent inner and outer  5-fold epitopes (24 Å CH1 separation) ( Fig. 6C and D). These epitopes are 85 Å and 79 Å apart, respectively, within the expected reach of a single IgG molecule (117-134640 Å ) [40]. We also examined the E106 epitope on the recent cryo-EM reconstruction of DENV-2 at 37uC [41,42]; this 'bumpy' virion supports a similar model of bivalent binding to DIII on the 5-fold and 2-fold symmetry axes.

Discussion
Epitope mapping studies have enhanced our understanding of the mechanisms of virus neutralization and identified sites on the E protein of flaviviruses that are targeted by neutralizing antibodies [7]. These include the lateral ridge of DIII of WNV and JEV [19,20], the A-strand of DIII of DENV [10,21], the CC9 loop of DIII of DENV-1 [33], the fusion loop of DII of WNV and DENV [43], a DI epitope of DENV-4 [11], and a complex epitope centered at the hinge of DI and DII on WNV [44] and DENV [45,46]. Here, we describe a composite epitope, comprised of regions of the lateral ridge and A-strand of DIII that is targeted by the therapeutic MAb E106. DIII residues contacted by E106 were highly conserved among DENV-1 genotypes but variable in other DENV serotypes. Consistent with this, E106 potently neutralized all five DENV-1 genotypes, but not other DENV serotypes nor WNV [13].
The E106 Fab-DIII complex was characterized by a smallburied interface, which correlated with an unexpectedly weak micromolar affinity, as determined both by SPR and isothermal calorimetry. We found no evidence for E106 binding to residues outside of DIII as the binding affinity to recombinant DIII appears to be very similar to the binding of Fab to E ectodomain protein or virions; consistent with this, neutralization escape studies only identified residues in DIII [22]. Monovalent E106 Fab poorly neutralized DENV-1 compared to intact E106 IgG, and this finding correlated with poor binding of Fab to intact virus. Although our structural models suggest that E106 can readily bind the isolated pre-fusion dimer and post-fusion trimer and possibly prevent the ,70u transition of DIII that is associated with membrane fusion [47,48], the inability of E106 Fab fragments to neutralize virus efficiently argues against this model. Our data are more consistent with a mechanism of neutralization that requires bivalent binding of E106 IgG to single virions, and cross-linking of E protein monomers in adjacent symmetry groups to prevent requisite E protein rearrangements during infection.
The measurement of micromolar monovalent affinity was unexpected given that E106 is our most potently neutralizing and protective anti-DENV-1 MAb (EC50 of 0.6 ng/ml against DENV-1 strain 16007), which is at least 10-fold more potent than our well-characterized DIII-lateral ridge-specific therapeutic MAb (E16) against WNV [20,30]. Indeed, E106 had the lowest EC50 value of ,500 anti-flavivirus MAbs (DENV-1, DENV-2, DENV-3, DENV-4, and WNV) generated to date in our laboratory [13,15,30,49]. Is there a correlation between neutralization potency and E106 bivalent binding to single virions? The icosahedral arrangement of the E protein on the mature DENV virion displays 180 copies of the E protein. Our in silico modeling predicts that in one possible arrangement, up to 48 of these sites may be available for bivalent engagement by 24 intact E106 IgG. Since monovalent binding was insufficient for neutralization, bivalent binding to single virions could neutralize infection by inhibiting an essential stage of the virus lifecycle (attachment, entry, or fusion). Alternatively, bivalent binding across virions could neutralize DENV infection by aggregating virus. Our postattachment studies suggest that E106 MAb was capable of neutralizing infection even after virus attached to the cell surface. Aggregation also appears less likely because the neutralization curves did not show a characteristic triphasic dose-response curve that was reported in inhibition studies describing antibody-virus aggregation [50]. We favor a model in which bivalent binding of E106 stabilizes and/or cross-links one or more E protein monomers in different symmetry groups, analogous to monovalent binding of WNV CR4354 MAb [44], and thus prevents radial expansion and rearrangement that is requisite for fusion of viral and host endosomal membranes [48,51].
E106 is one of very few MAbs that have been shown to require bivalent binding for efficient virus neutralization [35,36,52], and the first one directed against a flavivirus. While antibodies can be multivalent, with the potential to bind to virus particles with high avidity, the relatively small number of bivalent binding MAbs described to date may be attributed to the following: (i) the limited number of epitopes displayed on a single particle for some viruses; (ii) the position and orientation of epitopes that are beyond the reach of a bivalent MAb, which is limited by torsional flexibility of its hinge; (iii) the radius of curvature of virions, which may restrict the accessibility of epitopes by MAb (iv) post-translational modification (e.g. N-linked glycans) of virions that may hinder bivalent MAb engagement; (v) immunization protocols that rely on isolated recombinant envelope proteins, rather than envelope proteins in the structurally unique form of the intact virion; and (vi) assays that do not screen for bivalent neutralizing MAbs. E106 was generated after priming and boosting with infectious DENV-1 [13]. The repetitive E antigens in the icosahedral orientation of the virion may have facilitated selection of low monovalent affinity yet high avidity antibodies. While some antibodies against HIV have been described as bivalent, they actually are bispecific, with each arm binding distinct epitopes [53]. This is likely due to the paucity of trimeric spikes on the surface of the virus (,14) and their irregular spacing [54]. Although several MAbs have been proposed to require bivalent binding for efficient virus neutralization [55,56], compelling evidence is presented only for the nonenveloped positive strand RNA viruses, specifically, human rhinovirus 2 [36] and 14 [35,52] and the rabbit hemorrhagic disease calicivirus [37]. It may be that the quasi-icosahedral arrangement of the flavivirus envelope creates a landscape that permits limited bivalent MAb engagement.
Bivalent engagement of the virion by antibodies could be an important concept for DENV vaccine development. Immunity against DENV may not be achieved optimally using a subunit vaccine approach, as isolated E protein monomers may not induce antibodies that require bivalent binding for strong binding and neutralization. Analogously, some human MAbs against WNV bind a complex epitope at two independent positions on adjacent E protein monomers in different symmetry groups, which is only present on an intact WNV particle [44]. Human MAbs isolated from DENV-infected individuals are believed to recognize similar quaternary epitopes in E that are present only in the context of the intact DENV virion [45,46]. Given that E106 MAb was our most potent and therapeutic anti-DENV-1 MAb, strategies that enhance the likelihood of generating and identifying neutralizing antibodies that function through bivalent binding mechanisms may improve the potency of inhibitory humoral responses against DENV and other flaviviruses. Regardless, an understanding of the structural and mechanistic basis for the neutralization activity of E106 MAb provides new insights into the humoral response against flaviviruses.

Materials and Methods
E106 Fab-DIII crystal structure determination DENV-1 DIII (residues 293 to 399 of the E protein of strain 16007) was expressed in bacteria and re-folded oxidatively from isolated inclusion bodies as described previously [13]. Fab fragments of E106 were prepared using immobilized papain resin according to the manufacturer's instructions (Pierce). MAb (5 to 10 mg) was digested for 18 hours at 37uC, and passed over a protein A agarose resin to remove Fc fragments and undigested MAb and purified on a S-75 size exclusion chromatography column equilibrated in 20 mM HEPES pH 7.4 and 150 mM NaCl. Antibody-antigen complexes were formed by mixing papain-generated, gel filtered E106 Fab fragments with DIII at a ratio of 1.2:1 and crystallized by the hanging drop vapor diffusion method at a total protein concentration of 14 mg/mL in a solution of 22% PEG 6,000, and 0.1 M MES pH 5.0 (final pH 5.7). Crystals (in 1 mL crystallization drops) were cryoprotected by the addition of 0.2 mL aliquots of cryobuffer (in 23.5% PEG 6,000, 0.1 M MES pH 5.0, final pH 5.7, and 20% glycerol), then transferred to a fresh drop of cryobuffer and rapidly cooled in liquid nitrogen. X-ray diffraction data were collected at ALS beamline 4.2.2 (Lawrence Berkeley Laboratories) at a wavelength of 0.976 Å at 100 K with a CCD detector, and indexed and scaled in HKL2000 [57]. The crystals diffracted to 2.45 Å resolution and belonged to the space group P2 1 2 1 2 1 with unit cell dimensions of a = 82.7 Å , b = 91.8 Å , c = 92.6 Å , with one E106 Fab-DIII complex per asymmetric unit. Crystallographic phasing was obtained by molecular replacement using the program Phaser [58] and the coordinates of DENV-1 DIII (Protein Data Bank (PDB) 3EGP) and the Fab fragment of CTM01 IgG (PDB 1AD9). Iterative model building and refinement was performed using Coot [59] and Refmac [60] and later Phenix [61]. The final structure was refined to R cryst = 18.9% and R free = 23.9%. The final model includes DENV-1 DIII amino acid residues 297 to 394, E106 heavy chain residues 1 to 214 (Chothia numbering), and light chain residues 1 to 213. The atomic coordinates and structure factors of E106 Fab bound to DENV-1 DIII (CSGID target number IDP00272) have been deposited in the Protein Data Bank (www.rcsb.org) under PDB accession number 4L5F. Structural figures were prepared using CCP4MG [62] and Pymol [63] (surface representation using 1.4 Å solvent probe) and where shown, spheres represent van der Waal radii, vdw * 1.1.

Surface plasmon resonance and isothermal titration calorimetry
Monovalent antibody affinity analysis was performed using SPR (BIAcore T100, GE Company) and ITC (VP-ITC instrument, Microcal) at 10uC in 50 mM HEPES, pH 7.5 and 100 mM NaCl. For SPR, E106 MAb or Fab was immobilized at low concentrations (,500 Response Units) to a CM5 chip (GE healthcare) using aminecoupling chemistry. Bacterially-expressed DIII (residues 293-399) of DENV-1 (strain 16007) was injected at a flow rate of 65 ml/min at concentrations ranging from 0.2 mM to 500 mM for three minutes to saturate binding and then allowed to dissociate for seven minutes. The half-life of the monovalent interaction was short and did not require additional regeneration of the chip surface in preparation for the next DIII injection. The observed binding curves were double referenced to a non-reactive antibody (WNV E16) as well as buffer in the absence of DIII. Curves were analyzed by a steadystate fit for a 1:1 interaction, and a nonlinear least squares fit was used to evaluate the fit of the curve to the observed data. Alternatively, 500 response units (RU) of DIII were immobilized onto a CM5 chip and E106 Fab fragments were injected to saturation. Affinity measurements of E106 for the DENV-1 E ectodomain (DI-DII-DIII) were performed such that insect-derived DENV-1 E glycoprotein (ProSpec-Tany TechnoGene Ltd.) was in the stationary phase and E106 Fab fragments were in the mobile phase, in order to conserve limited protein and avoid avidity affects. Additional regeneration was not necessary because of the short halflife for the interaction. WNV E ectodomain was used as a negative control for E106 binding. Analysis was performed as with the DIII described above. ITC experiments were performed such that 4 to 8 mL of DENV-1-DIII protein (90 to 110 mM) was injected into 1.4 mL of E106 Fab protein (6 to 7 mM) over a total of 36 injections. The titration data were integrated and normalized in Origin (Microcal) to determine the reaction stoichiometry, n, and equilibrium constant Ka ( = K 21 d ).
Neutralization assays and capture ELISA Plaque reduction neutralization tests (PRNT) and pre-and postattachment neutralization assays were performed with DENV-1 strain 16007 on Vero cells as previously described [13,64]. Binding of intact MAbs or Fabs (E103, E106, and a negative control WNV E16) to DENV-1 virions (strain 16007) was detected by capture ELISA [13,64]. Briefly, humanized DENV-1 E50 MAb (subcomplex DIII A-strand specific antibody) was coated at 2 mg/ml on MaxiSorp (Nunc) polystyrene 96-well microtiter plates in a sodium carbonate (pH 9.3) buffer. Plates were washed three times in wash buffer (PBS with 0.02% Tween 20) and blocked for one hour at 37uC with blocking buffer (DMEM with 10% FBS). DENV-1 virions (2.5610 5 PFU) diluted in DMEM with 10% heat-inactivated FBS were captured on plates for two hours at 37uC. Wells were washed thrice with blocking buffer and DENV-1 MAb or Fab was then added at 100 mg/ml and 4-fold serial dilutions to duplicate wells and incubated for two hours at 37uC. Plates were washed five times and then sequentially incubated with goat anti-mouse (whole molecule) IgG-HRP (Sigma, St Louis, MO) and tetramethylbenzidine substrate (Dako). The reaction was stopped with the addition of 2 N H 2 SO 4 to the medium, and emission (450 nm) was read using an iMark microplate reader (Bio-Rad).
Time and temperature-dependent neutralization of DENV-1 A plasmid expressing the C-prM-E genes of DENV-1 (strain 16007) was co-transfected into HEK-293T cells with a plasmid encoding a WNV replicon expressing GFP. Transfected cells were incubated at 30uC and RVP harvested at 72 and 96 hours posttransfection, filtered through a 0.2 mM filter, and stored aliquoted at 280uC. DENV-1 RVP were incubated with serial dilutions of MAb under conditions of antibody excess at 4uC, 37uC, or 40uC for one or five hours. Subsequently, MAb-RVP mixtures were added to Raji-DCSIGNR cells and incubated at 37uC for 48 hours. Infected cells were assayed for GFP expression using a BD FACSCalibur flow cytometer as described [28].

The kinetics of DENV-1 RVP binding to DENV-1 MAbs and Fabs
All bio-layer interferometry studies were performed in PBS buffer supplemented with 1 to 2 mg/ml BSA (PBS-B) at 25uC using an Octet Red biosensor system (ForteBio). DENV-1 reporter virus particles (RVPs) (Western Pacific 74 strain) were produced as previously described [65]. To purify virus particles, RVP production supernatant was harvested, clarified through a 0.22 mm filter (Corning), and PEG precipitated using 7.5% PEG 8000 (Sigma). The virus particles were further purified through two 20% sucrose cushions before resuspension in HBS. Samples were stored at 2 80uC and gently thawed prior to use. RVPs were loaded onto streptavidin (SA) biosensor tips using a human monoclonal antibody against DENV-1 (1H7.2, the anti-prM antibody, a gift from James Crowe), which was captured using a biotinylated goat anti-human polyclonal antibody (GAH Fc, Southern Biotech). Briefly, GAH Fc was diluted to 5 mg/ml in PBS-B and bound to SA sensor tips for five minutes. Following a brief rinse in PBS-B, 1H7.2 (5 mg/ml in PBS-B) was captured for five minutes. After another brief rinse, DENV-1 RVPs diluted to 10 mg/ml (or 50 mg/ml (E106 Fab)) were loaded for 45 minutes followed by a five-minute stabilization. Antibody association was measured for up to 10 minutes followed by dissociation for 20 minutes (E106) or 45 minutes (E103) in buffer. Non-specific binding was assessed using sensor tips without RVPs as well as using sensor tips loaded with retroviral pseudotypes (Lipoparticles) containing only endogenous cell surface receptors (no viral envelope protein). Data analysis was performed using Octet Data Analysis v6.4 (ForteBio). Binding kinetics were analyzed using a standard 1:1 binding model.