A Bispecific Antibody That Simultaneously Recognizes the V2- and V3-Glycan Epitopes of the HIV-1 Envelope Glycoprotein Is Broader and More Potent than Its Parental Antibodies.

Broadly neutralizing antibodies (bNAbs) can prevent a new HIV-1 infection and can at least temporarily suppress an established infection. However, antibody-resistant viruses rapidly emerge in infected persons treated with any single bNAb. Several bispecific antibodies have been developed to increase the breadth of these antibodies, but typically only one arm of these bispecific constructs binds the HIV-1 envelope glycoprotein trimer (Env). Here, we develop and characterize bispecific constructs based on well-characterized V2-glycan and V3-glycan bNAbs and show that at least one member of this class is more potent than its parental antibodies, indicating that they can simultaneously bind both of these epitopes of a single Env trimer. These data show that bispecific antibody-like proteins can achieve greater neutralization potency than the bNAbs from which they were derived.

form of CAP256.VRC26.25 might better position the Fc domain of a bispecific antibody near the V3-glycan epitope than a complete Fab domain. We thus generated scFv-Fc and full-length antibody forms of CAP256.VRC26. 25 and compared them in combination with full-length 10-1074 antibody, as represented in Fig. 2A. The linker separating the scFv and Fc domains was also varied, with different lengths of tetraglycine-serine linker (G 4 S) repeats. Homodimeric forms of these CAP256.VRC26.25 scFv-Fc constructs (L5, L10, L20, and L23, with the number indicating linker length and full nomenclature shown in Table 1) neutralized each of two HIV-1 isolates to comparable levels, albeit modestly less efficiently than the original CAP256.VRC26.25 bNAb (Fig. 2B). Combinations of each of these scFv-Fc constructs were then generated through cotransfection with plasmids expressing 10-1074 heavy and light chain, generating cotransfection combinations (CTC) of hetero-and homodimeric forms. These combinations neutral- Broad and Potent Bispecific Anti-HIV Antibodies ® ized the clade C isolate CE1176 more efficiently than either 10-1074 or an scFv-Fc form of CAP256.VRC26.25 with a 10-amino-acid linker (Fig. 2C). They also neutralized the clade A isolate BG505 isolate more efficiently than the 10-1074 antibody did and similarly to the CAP256.VRC26.25scFv-Fc homodimer. In contrast to these combinations, this scFv-Fc homodimer did not fully neutralize BG505. We also observed that mixtures, including a 10-amino acid scFv-Fc linker neutralized more efficiently than analogous mixtures with longer or shorter linkers. Because cotransfection combinations yielded both hetero-and homo-dimeric forms, we introduced knob-in-hole (KIH) mutations into the Fc domains to generate two additional constructs (26). BISC-1A (bispecific scFv) is a bispecific construct comprised of the CAP256.VRC26.25 single chain and 10-1074, whereas BICM-1A (bispecific cross Mab) is a bispecific construct of full-length versions of both antibodies using a cross-monoclonal antibody (Cross-Mab) architecture, ensuring that each light chain is associated with its own heavy chain (27). BISC-1A neutralized both CE1176 and BG505 more efficiently than the analogous cotransfection mixture (CTC-1A-L10), which in turn neutralized more efficiently than BICM-1A, the Cross-Mab form of 10-1074 and CAP256.VRC26.25 (Fig. 2D). All three combinations neutralized both isolates more efficiently than homodimeric forms of the CAP256.VRC26.25 scFv-Fc construct or the bNAb 10-1074. Notably, homodimeric CAP256.VRC26.25 scFv-Fc failed to completely neutralize CE1176 and neutralized BG505 less efficiently at higher concentrations, likely because it poorly neutralized a subset of Env conformational or glycosylation variants (28). In contrast, all combinations of CAP256.VRC26.25 and 10-1074 including CTC-1A-L10, BISC-1A, and BICM-1A neutralized these viruses completely. Thus, combinations of CAP256.VRC26.25 and 10-1074, and especially BISC-1A, neutralize these HIV-1 isolates markedly more efficiently than homodimeric 10-1074 or CAP256.VRC26.25 scFv-Fc.
CAP256-VRC26.25 scFv-Fc synergizes with multiple V3 glycan-targeting antibodies. We also investigated whether the enhanced neutralization observed with BISC-1A (the bispecific construct with CAP256.VRC26.25 scFv-Fc and 10-1074 arms) relative to its parental constructs could be extended to additional V3 glycan-targeting antibodies. We generated bispecific molecules consisting of the same CAP256.VRC26.25 scFv-Fc combined with full-length V3-glycan bNAbs PGT121 (BISC-1B) and PGT128 (BISC-1C) (7). These constructs were compared with each of their parental components for their ability to neutralize a 12-isolate global panel isolate and an additional 3 isolates (PVO4, ZM651, and BG505). Neutralization curves are shown for two isolates, CE1176 and BG505 that are differentially resistant to CAP256.VRC26.25 (Fig. 3A). Again, all three bispecific constructs neutralized these isolates more efficiently than any parental antibody, even when largely resistant to CAP256.VRC26.25. Fifty percent, 80%, and 95% inhibitory concentration (IC 50 , IC 80 , and IC 95 ) values are plotted for all 15 isolates in Fig. 3B to D (numerical values are provided in Tables S1 to S3 in the supplemental material). Notably, unlike any of the parental antibodies, all bispecific constructs neutralized all 15 isolates with IC 50 values less than 10 g/ml ( Fig. 3B and Table 2). Similarly, BISC-1B and BISC-1C neutralized all isolates with IC 80 values less than 10 g/ ml, whereas the IC 80 of BISC-1A for one isolate (CNE8) was greater than this concentration (Fig. 3C). Remarkably, BISC-1C neutralized all 15 isolates with IC 95 values less than 10 g/ml (Fig. 3D). Also, with two or three exceptions in each case, all bispecific Broad and Potent Bispecific Anti-HIV Antibodies ® constructs neutralized each isolate more efficiently than their parental antibodies ( Fig. 3E to G). Bispecific constructs neutralized isolates that were sensitive (IC 80 Ͻ 20 g/ml) to both parental antibodies more efficiently than these parental antibodies (Table 3). Similarly, they neutralized isolates resistant to CAP256.VRC26.25 scFv-Fc and sensitive to the parental V3-glycan antibody more efficiently than their parental antibodies. However, CAP256.VRC26.25 scFv-Fc in general neutralized isolates that were resistant to V3-glycan bNAbs but sensitive to CAP256.VRC26.25, although these isolates were nonetheless efficiently neutralized in all cases. Notably, BISC-1C neutralized every isolate tested that was resistant (IC 80 Ͼ 20 g/ml) to both PGT128 and CAP256.VRC26.25. Thus, bispecific constructs combining CAP256.VRC26.25 and any of three V3-glycan antibodies neutralize the majority of a global panel of HIV-1 isolates more efficiently than any of their component antibodies.
Multiple mathematical models have been described to predict the neutralization potency of combinations of antibodies based on their observed IC 50 and IC 80 values (19,29). We compared the Bliss-Hill and additive models of combinations of the BISC-1A components and equal mixtures of the parental components (Table S4). For 8 of the 15 isolates tested (CE1176, Zm651, x2278, CH119, BJOX2000, PV04, Tro11, and CE0217), BISC-1A performed better than either model or the experimentally determined 50:50 mixture. For the remaining seven isolates, including four fully resistant to 10-1074, the potency of BISC-1A was roughly equivalent to the models and mixture. Thus, BISC-1A demonstrates synergistic neutralization of some isolates, but even when synergy does not occur, it remains as potent as a mixture of its parental components. Bispecific molecules targeting the V3-glycan epitope with an additional V2apex epitope target. To evaluate whether additional V2-glycan bNAbs could be combined with V3-glycan bNAbs, we performed a similar study using an scFv-Fc version of the V2-glycan bNAb PGT145 (7). Specifically, we generated constructs identical to BISC-1A, BISC-1B, and BISC-1C, except that the PGT145 scFv-Fc replaced that of CAP256.VRC26.25. Thus, BISC-2A, -2B, and -2C combined this scFv-Fc with 10-1074, PGT121, and PGT128, respectively. The CE1176 isolate is completely resistant to homodimeric PGT145 scFv-Fc, but combinations of the construct with 10-1074, PGT121, and PGT128 neutralized this isolate more efficiently than any V3-glyan bNAb (Fig. 4A). Similarly, BISC-2A and BISC-2B neutralized BG505 more efficiently than any of their parental antibodies, but BISC-2C neutralized with an efficiency similar to that of the potent neutralization of PGT128. However, these PGT145 bispecific constructs generally did not neutralize a nine-isolate subset of the panel characterized in Fig. 3 more efficiently than their parental antibodies, nor was their breadth greater than those of these antibodies (Fig. 4B to G, Tables 4 and 5, and Tables S5 to S7). We conclude that bispecific-based PGT145 scFv-Fc do not consistently neutralize more efficiently than their parental bNAbs and that distinct properties of CAP256.VRC26.25 contribute to the increased breadth and potency of BISC-1A, -1B, and -1C.
BISC-1A simultaneously engages V2-and V3-glycan epitopes of the Env trimer. The increased potency of BISC-1A, -1B, and -1C relative to their respective parental bNAbs suggested that both arms of these bispecific constructs could simultaneously engage a single Env trimer. To determine whether this were the case, we compared binding of BISC-1A and its parental components for their ability to associate with monomeric gp120 of the BG505 isolate, soluble trimeric SOSIP Env of the same isolate, and cells transfected to expressed BG505 Env truncated in its cytoplasmic domain to facilitate efficient expression on the cell surface. 10-1074 bound monomeric gp120 more efficiently than BISC-1A, consistent with the dependence of the CAP256.VRC26.25 on a quaternary epitope (Fig. 5A). In contrast, BISC-1A bound BG505 SOSIP trimers Broad and Potent Bispecific Anti-HIV Antibodies ® markedly more efficiently than either of its parental constructs (Fig. 5B), indicating that both arms of the bispecific construct were simultaneously engaged. The marked difference between BISC-1A and 10-1074 could also be observed with cell-expressed BG505 Env (Fig. 5C). Competition binding studies, both enzyme-linked immunosorbent assay (ELISA) studies of SOSIP trimers and surface staining of cell-expressed Env, were also consistent with BISC-1A engaging both of its Env epitopes. For the ELISAs, plates were coated with BG505 SOSIP trimer and then incubated with a fixed 1-g/ml concentration of CAP256.VRC26.25-scFv-Fc, 10-1074, or BISC-1A. The plates were washed and then incubated with serial dilutions of either CAP256.VRC26.25-scFv-Fc (Fig. 5D) or 10-1074 (Fig. 5E) with a mouse Fc domain. Binding of the mouse Fc proteins was determined. Compared to the baseline binding in the absence of a competing antibody (gray), CAP256.VRC26.25-scFv-mFc binding was inhibited by both CAP256. VRC26.25-scFv-huFc and BISC-1A. Similarly, 10-1074-mFc was inhibited by both 10-1074-huFc and BISC-1A. There was minimal cross-competition between the parental antibodies. Similar results were obtained with surface staining (Fig. 5F and G). In this case, cells expressing BG505 ΔCT (cytoplasmic tail) were preincubated with various concentrations of the human constructs and subsequently stained with a fixed amount of the mouse Fc (mFc) constructs. Binding as determined by flow cytometry was normalized to the baseline without preincubation. As observed with ELISA binding, the parental antibodies were inhibited by BISC-1A and their human Fc (hu-Fc) versions. Combined with enhanced neutralization potency of BISC-1A relative to the parental  Broad and Potent Bispecific Anti-HIV Antibodies ® bNAbs, the data of Fig. 5 indicate that BISC-1A, and likely BISC-1B and BISC-1C, simultaneously bind both the V2-glycan and the V3-glycan epitopes of a single Env. BISC variants are broader and more potent than their parental components. We determined a theoretical breadth of BISC-1A, -1B, and -1C by analyzing all isolates for which neutralizing data were available for CAP256.VRC26.25 and each of the V3-glycan antibodies used in these bispecific constructs (15,19,20,(22)(23)(24)(25)(29)(30)(31)(32)(33)(34)(35). As shown in Fig. 6A, 403 isolates have been studied with both CAP256.VRC26.25 and 10-1074. Of these isolates, 226 or 56.1% are neutralized by CAP256.VRC26.25 with IC 80 values of less than 20 g/ml. Similarly, 246 or 61.0% are neutralized by 10-1074, but 345 or 85.6% are neutralized by at least one of these bNAbs, suggesting that BISC-1A would neutralize approximately this proportion of isolates. Similarly, 336 of 403 or 83.4% would theoretically be neutralized by BISC-1B ,and 232 of 280 (82.9%) would be neutralized by BISC-1C. Thus, these constructs are likely to neutralize a significantly wider range of isolates than CAP256.VRC26.25, 10-1074, PGT121, or PGT128. We then generated a potency-breadth plot of the observed neutralization efficiency for BISC-1A,  Table S2 were plotted for the 15 isolates tested in this study, each BISC construct had both increased breadth and increased potency compared to the parental components. Thus, BISC constructs neutralize a larger fraction of isolates than their components, and they do so more efficiently.

DISCUSSION
In this study, we examined the neutralization potency of bispecific molecules targeting the V2 apex and the V3 glycan sites on HIV-1 Env. We found that targeting a combination of these epitopes led to increased neutralization of a global panel of viruses. Specifically, heterodimeric combinations of CAP256.VRC26.25-scFv-Fc with any of three V3-glycan antibodies neutralized most viruses tested with markedly greater efficiency than homodimeric forms of their components. These observations contrast with most other reported bi-and multispecific bNAb constructs (20)(21)(22)(23). For example, Asokan et al. reported greater breadth, but no increase in potency when a panel of bNAb combinations were characterized (20). Similarly, bispecific and trispecific constructs reported by Khan et al. (21) and Steinhardt et al. (22), respectively, were not more potent, but again much broader, reflecting the sum of isolates neutralized by their component antibodies. One exception is the study of Bournazos et al. who observed greater neutralization when bNAbs were combined and a 62-amino-acid IgG3-derived linker was used to connect the Fab regions to the Fc domains (23). That study also observed greatest synergy when 3BNC117 and a less well-characterized V3-glycan antibody, PGT135, were combined but did not observe enhanced neutral-  Table S2 in the supplemental material for the indicated constructs as analyzed with CombiNAber online tools.
Broad and Potent Bispecific Anti-HIV Antibodies ® ization when 3BNC117 was combined with either 10-1074 or PGT128, both of which are markedly broader than PGT135 (7). The greater enhancement, higher potency, and more compact structure of BISC-1A, -1B, and -1C is consistent with the greater proximity of the V2-and V3-glycan epitopes compared with the epitopes targeted in the study of Bournazos et al. In addition, we were uniquely able to demonstrate greater binding of our bispecific constructs to SOSIP trimers and to cell-expressed Env, consistent with their higher potency.
We also observed that combinations of 10-1074 and an scFv-Fc form of CAP256. VRC26.25 neutralized more efficiently than the same combination in CrossMab format (Fig. 2). CAP256.VRC26.25 has not been solved in complex with Env. However, a structure of PGT145 bound to the Env trimer, used in Fig. 1, indicates that the C terminus of its scFv would be positioned closer than the C1 domain of the full-length antibody to the C terminus of the 10-1074 C1 domain. Nonetheless, we did not observe synergy when a PGT145 scFv-Fc was combined with 10-1074 using an identical 10-amino-acid linker, implying that the CAP256.VRC26.25 scFv is more closely positioned to the C1 domains of V3-glycan antibodies. Perhaps a longer linker or a different arrangement of scFv components might also allow PGT145 scFv-Fc to synergize with V3-glycan antibodies.
One limitation of V2-glycan antibodies is their frequent inability to completely neutralize otherwise sensitive HIV-1 isolates, reflected in the fact that their IC 80 and especially IC 95 values are higher than their IC 50 values would indicate ( Fig. 3A and 4A) (28). Incomplete neutralization remains a concern with this class of antibodies because it could accelerate the emergence of resistance. Notably, we consistently observed complete neutralization with our BISC variants despite the fact that CAP256.VRC26.25 scFv significantly contributed to neutralization. Thus, the bispecific variants described here do not share this limitation of V2-glycan antibodies.
It has previously been reported that one V2-glycan antibody, PG9, can interfere with neutralization by V3-glycan antibodies in some circumstances (29). Similar interference may account for the inability of PGT145 to combine usefully with V3-glycan antibodies in the BISC2 constructs characterized in Fig. 4. It may also suggest that the Cross-Mab construct BICM-1A is less potent than BISC-1A because the Fab of the former construct interferes with binding of its 10-1074 arm. A structure of CAP256.VRC26.25 in complex with Env may clarify the basis of this interference and facilitate the development of more potent variants. It is also unclear whether the BISC variants have pharmacokinetic (PK) properties that are similar to those of their component molecules, and the half-lives of bispecific and multispecific antibodies can frequently be limited by their shorter-half-life arm. Further evaluation and optimization of these constructs in vivo will be necessary before they might be considered for clinical use. Such optimization may be warranted: BISC-1A, -1B, and -1C are broader and more potent than either CAP256.VRC26.25 or their respective 10-1074, PGT121, and PGT128 component bNAbs. The architecture of these constructs allows for straightforward production in cell culture. Perhaps most importantly, they demonstrate that significant synergy can be obtained when two proximal Env epitopes are simultaneously engaged.
The variable heavy and light chains of PGT145 were cloned into human IgG1 expression vectors as previously described (41). 10-1074 expression plasmids were provided by Michel Nussensweig. PGT121 and PGT128 were provided by Dennis Burton. Heavy and light chains for PGT145 were synthesized by Integrated DNA Technologies (IDT) (Newark, NJ) and cloned into IgG1 expression vectors. scFv-Fc versions of Cap256-VRC26.25 and PGT145 were constructed by cloning the variable heavy and variable light chains separated by a 10-residue G 4 S linker fused to an IgG1 Fc domain. Different linkers were tested to separate the scFv and Fc domains, including 5-, 10-, and 20-residue G 4 S linkers and a 23-residue linker designed to limit glycosylation sites (amino acid sequence, GGAGGEAGAGGAGGAGGEAGAGG). Knobsin-holes mutations were generated by mutation of S354C/T366W in the V3-glycan antibodies ("knobs") and Y349C/T366S/L368A/Y407V in the apex antibodies and scFv-Fc ("holes") (26). For competition staining, the 10-1074 variable heavy chain and CAP256.VRC26.25 scFv were cloned into murine IgG2a Fc expression vectors. Expression vectors for the BG505 gp160-Δcytoplasmic tail were provided by John Moore and P. J. Klasse.
Protein production and purification. Antibodies were produced in Expi293 cells (Life Technologies, Carlsbad, CA). Cells were grown to a density of 3 ϫ 10 6 cells/ml prior to transfection with FectoPRO according to the manufacturer's instructions (Polyplus, New York, NY). Total DNA (140 g) was transfected in 250 ml Expi293 cells. Antibodies were produced by transfection of two plasmids encoding heavy and light chain, respectively, at a 1:1 ratio. Constructs using CAP256-VRC26.25 and PGT145 plasmids were cotransfected at an 4:1 ratio with plasmid encoding human tyrosine protein sulfotransferase 2 (TPST2) to ensure proper sulfation of CDRH3 residues. Bispecific constructs were produced by transfection of a 4:2:3:1 ratio of scFv-Fc:Heavy:Light:TPST2 plasmids. Five days posttransfection, the medium was collected for protein purification. Debris was cleared by centrifugation for 10 min at 4,000 ϫ g and filtered using 0.45-m filter flasks (Thermo Scientific, Waltham, MA). Antibodies and Fc-fusion proteins were purified from supernatants using HiTrap MabSelect SuRe columns (GE Healthcare Life Sciences, Pittsburgh, PA). After protein binding, columns were washed extensively with phosphatebuffered saline (PBS) before elution with IgG elution buffer (Thermo Scientific, Waltham, MA). Eluate pH was immediately adjusted with Tris-HCl 1 M pH 9.0 neutralization buffer (G-Biosciences, St. Louis, MO). Buffer was exchanged with PBS, and protein was concentrated to 1 mg/ml by ultrafiltration (Amicon Ultra, Millipore Sigma, Billerica, MA) at 4,000 ϫ g. Protein concentration was measured using the IgG setting (molar extinction coefficient of 210,000 M Ϫ1 cm Ϫ1 ) on a Nanodrop spectrophotometer (Thermo Scientific, Waltham, MA).
BG505 SOSIP trimer was produced by transfection of Expi293 cells with BG505.SOSIP.664 gp140 containing a His tag and pFurin plasmid. Supernatants were harvested after 96 h and initially purified on HIS-trap columns (GE Healthcare Life Sciences, Pittsburgh, PA). Trimer was further purified using PGT145 coupled to CNBr-activated Sepharose 4B (GE Healthcare Life Sciences, Pittsburgh, PA) according to the manufacturer's instructions (42). Trimer was eluted with 3 M MgCl 2 , and immediately buffer was exchanged with PBS and protein was concentrated to 1 mg/ml by ultrafiltration (Amicon Ultra, Millipore Sigma, Billerica, MA) at 4,000 ϫ g.
Pseudotyped viruses. HEK293T cells were transiently transfected with 45 g of a plasmid expressing Env of the indicated HIV-1 isolate and 45 g of pNL4.3ΔEnv, an HIV-1 expression vector lacking a functional env gene. Medium was changed at 12 h posttransfection. At 48 h after the medium change, supernatants were collected and filtered through a 0.45-m syringe filter. Aliquoted pseudoviruses were stored at Ϫ80°C.
TZM-bl neutralization assay. TZM-bl neutralization assays were performed as previously described (41,43,44). Briefly, antibody titrations were incubated with pseudotyped viruses for 1 h at 37°C. TZM-bl cells were diluted in DMEM to 100,000 cells/ml and added to the virus/inhibitor mix. Cells were then incubated for 48 h at 37°C. Viral entry was determined by luciferase readout with BriteLite Plus (Perkin Elmer, Waltham, MA) and read on a Victor X3 plate reader (Perkin Elmer, Waltham, MA).
ELISA. Enzyme-linked immunosorbent assay (ELISA) plates (Costar) were coated with 5 g/ml HIV-1 gp120 (Immune Technology Corp.) or BG505.664.SOSIP trimer and left overnight at 4°C. The plates were washed with PBS plus 0.05% Tween 20 (PBS-T) twice and blocked with 5% bovine serum albumin (BSA) in PBS for 1 h at 37°C. Dilutions of antibodies blocked with 5% BSA in PBS were added to the plate and incubated for 1 h at 37°C. Samples were washed five times with PBS-T and labeled with a horseradish peroxidase-conjugated secondary antibody (Jackson Immuno Research) recognizing human IgG1. Plates were incubated for 1 h at 37°C and then washed 10 times with PBS-T. Tetramethylbenzidine (TMB) solution (Fisher) was added and left for 10 min at room temperature, and then the reaction was stopped with TMB stop solution (Southern Biotech). Absorbance was measured at 450 nm with a Victor X3 plate reader (Perkin Elmer).
For competition binding, ELISA plates were coated with BG505.664.SOSIP trimers and left overnight at 4°C. The plates were washed with PBS plus 0.05% Tween 20 (PBS-T) twice and blocked with 5% BSA in PBS for 1 h at 37°C. A saturating concentration (1 g/ml) of antibodies with human Fc domains, blocked with 5% BSA in PBS, was added to the wells of the plate and incubated for 1 h at 37°C. Samples were washed five times with PBS-T before the addition of serial dilutions of CAP256.VRC26.25 scFvmouse-Fc or 10-1074 with a mouse Fc and incubated for 1 h at 37°C. Samples were washed five times with PBS-T and labeled with a horseradish peroxidase-conjugated secondary antibody (Jackson Immuno Research) recognizing mouse IgG. Plates were incubated for 1 h at 37°C and then washed 10 times with PBS-T. Tetramethylbenzidine (TMB) solution (Fisher) was added and left for 10 min at room temperature, and then the reaction was stopped with TMB stop solution (Southern Biotech). Absorbance was measured at 450 nm with a Victor X3 plate reader (PerkinElmer).
HIV Env surface staining assay. HEK293T cells were transfected with plasmids expressing the HIV-1 BG505 envelope glycoprotein lacking cytoplasmic residues 732 to 876 (HXBc2 numbering), BG505 ΔCT (cytoplasmic tail). Cells were collected 48 h posttransfection with nonenzymatic dissociation buffer (Sigma-Aldrich, St. Louis, MO). Cells were washed with flow cytometry buffer (PBS with 2% goat serum, 0.01% sodium azide) before incubation with antibodies for 1 h on ice. Cells were washed twice with flow cytometry buffer. Antibody binding was determined with allophycocyanin (APC)-conjugated goat antihuman secondary antibodies (Jackson ImmunoResearch, West Grove, PA). After incubation with secondary antibody, cells were washed once with flow cytometry buffer, once with PBS, and then resuspended in 1% paraformaldehyde in PBS. Binding was analyzed with an Accuri C6 flow cytometer, and data were analyzed with the C6 Software (BD Biosciences, San Jose, CA).
Broad and Potent Bispecific Anti-HIV Antibodies ® January/February 2020 Volume 11 Issue 1 e03080-19 mbio.asm.org 13 For competition binding, HEK293T cells were transfected with plasmids expressing BG505 ΔCT. Cells were collected 48 h posttransfection with nonenzymatic dissociation buffer (Sigma-Aldrich, St. Louis, MO). Cells were washed with flow cytometry buffer (PBS with 2% goat serum, 0.01% sodium azide) before incubation with serial dilutions of human Fc domain containing scFv-Fc, antibody, or BISC-1A for 1 h on ice. Cells were washed twice with flow cytometry buffer, and incubated with 200 ng/ml of mouse Fc containing CAP256.VRC26.25 scFv-mouse-Fc or 10-1074 with a mouse Fc. Antibody binding was determined with APC-conjugated goat anti-human and fluorescein isothiocyanate (FITC)-conjugated goat anti-mouse secondary antibodies (Jackson ImmunoResearch, West Grove, PA). After incubation with secondary antibody, cells were washed once with flow cytometry buffer, once with PBS, and then resuspended in 1% paraformaldehyde in PBS. Binding was analyzed with an Accuri C6 flow cytometer, and data were analyzed with C6 software (BD Biosciences, San Jose, CA).
Modeling and predictions. Predicted potency of combinations of antibodies and potency-breadth curves were calculated using the Los Alamos National Labs CombiNAber tool (19,29).
Statistical analysis. Wilcoxon signed-rank test analysis was performed in GraphPad Prism.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.

ACKNOWLEDGMENTS
This work was supported by grants from the National Institutes of Health to M.F. (DP1-DA043912 and R37-AI091476).
M.F. and M.R.G. are cofounders and have a financial stake in Emmune, Inc., a company focused on the development and delivery of HIV-1 entry inhibitors.