Direct detection and quantification of HIV particles and their envelope trimers bound by combinations of broadly neutralizing antibodies

Summary

Broadly HIV-neutralizing antibodies (bnAbs) targeting viral envelope (Env) antigens are promising for HIV/AIDS treatment, functional cure, and prevention. However, individual bnAb epitope classes show distinct gaps in coverage of Env diversity, supporting the use of antibody combinations to improve breadth and potency. We examined how frequently multiple bnAb classes bind simultaneously to individual virions using quantitative single-particle fluorescence correlation spectroscopy (FCS) and FRET approaches. These methods quantified unmodified virion subpopulations from multiple strains bound by one or more fluorescently labeled bnAbs. Analysis of N49P7, PGT121, and PGDM1400, targeting the CD4-binding site, V3-glycan region, and trimer apex, respectively, revealed subpopulations bound by one, two, or three bnAbs. Across strains, the proportion of virions bound by all three antibodies correlated significantly with neutralization activity. These findings define the distribution of combinatorial bnAb binding across heterogeneous virion populations and support quantitative single-particle fluorescence approaches for evaluating antibody combinations against HIV.

Graphical abstract

Highlights

• •Combination of bnAbs reveal distinct Env epitope targeting on single HIV-1 virions
• •FCS and burst analysis quantify single virions bound by one, two or three bnAbs
• •FRET maps simultaneous multi-bnAb engagement on the same HIV-1 Env trimer within ≤10 nm
• •Single-virion mapping links combinatorial bnAb occupancy to neutralization across strains

Applied sciences

Introduction

The development of broadly neutralizing antibodies (bnAbs) against HIV with remarkable breadth and potency offers a promising approach for the treatment, prevention and potential functional cure of HIV/AIDS.^1,2,3,4 These bnAbs categorized by their target epitope clusters on the HIV envelope glycoprotein (Env) have shown considerable advantages in preclinical studies compared to traditional antiretroviral drugs (ARVs). These benefits include less frequent dosing, reduced risk of side effects and genotoxicity, the ability to target latent HIV reservoirs, enhancement of host antiviral immune responses and resistance to conventional ARV resistance mechanisms.^5,6,7 BnAbs can inactivate Env, opsonize single virions and engage Fc-dependent effector functions. BnAbs are also amenable to rational engineering for greater functionality, and for extended half-lives to sustain efficacy.⁸ BnAbs targeting trimeric HIV Env spikes are considered a promising approach to prevent HIV transmission through PrEP, suppress viral replication therapeutically, or achieve a functional cure.^9,10 In contrast conventional combination antiretroviral therapy (cART) does not elicit these immune modulating effects.¹¹ Encouraged by these attributes and supportive preclinical findings, several bnAbs have advanced to clinical trials, with some already completed.^12,13,14 On HIV particles, the envelope spike is a heavily glycosylated trimer composed of three heterodimers, each containing gp120 surface subunits and gp41 transmembrane proteins. The gp120 subunit binds to the host cell receptor CD4 and a co-receptor triggering gp41 to facilitate membrane fusion and viral entry. Env is highly variable in sequence and structure, enabling the virus to evade immune responses. Simultaneously, Env presents highly conserved epitopes, such as those within glycan domains, the CD4 binding sites or on gp41.^15,16,17 These conserved regions are promising targets for vaccine development as bnAbs against them can provide robust neutralization and potent sterilizing protection, as demonstrated in macaque infection models using SHIV challenges.^18,19,20,21

Advances in the discovery of highly potent bnAbs, sourced from rare HIV infected individuals with high plasma neutralizing activity²² have enabled clinical testing of anti Env antibodies. Current leading bnAbs categorized by the specific vulnerable epitope clusters (“supersites”) target on the trimeric Env spike.²³ These supersites include the apex region comprising gp120 V1/V2 amino acids and glycans, the gp120 V3 loop domains associated with the N332-glycan, high mannose regions of gp120,²⁴ the CD4 binding site (CD4bs) of gp120, the membrane proximal external region of gp41 and combined sequences from proximal gp120 and gp41 regions.²⁵ Traditionally bnAb development has been guided by the assumption that clinical success depends on their ability to neutralize a broad spectrum of targets with high potency.²⁶ A persistent concern is that all known “native” bnAbs exhibit incomplete coverage of HIV strains.²⁷ Such limits are apparent in standardized assays using cell lines and genetically diverse pseudovirus panels.^28,29 There are numerous bnAb classes with varying coverage and newly engineered variants continue to emerge.²⁷ Extensive in vitro testing with panels comprising highly diverse HIV strains and subtypes reveal that each bnAb class is associated with a small but distinct group of neutralization resistant strains^30,31,32 which becomes an impediment to successful clinical trials based on bnAb monotherapy.^33,34 Clinical proof of concept has shown that dual administration of VRC01LS, a bnAb targeting the CD4 binding site of gp120 and 10-1074, a V3-glycan-directed bnAb targeting the N332 glycan supersite can maintain HIV-1 suppression in a subset of early treated children following analytical treatment interruption. In one such trial, sustained virologic control was associated with a smaller reservoir size and viral susceptibility to 10-1074, highlighting both the safety and therapeutic promise of bnAb combinations as ART sparing strategies.³⁵ However, the effectiveness of bnAbs is limited by gaps in their coverage of global HIV variants, which could impact their overall efficacy. Achieving success in this area requires an understanding of the epitope patterns displayed by virions and virus populations. Each bnAb class resistance group shares signature Env escape sequences, with limited apparent genotypic and phenotypic overlap. Such variants evolve during spreading infection due to HIV genetic diversity and rapidly emerge under single bnAb class selective pressure in vivo. Collectively, these features point to combining bnAb epitope classes to achieve a polyclonal, more escape resistant approach to provide broad clinical efficacy.^36,37

Evidence of protective and therapeutic efficacy in animal models has spurred the advancement of several bnAbs to clinical trials, with some studies already completed.^38,39 Clinical trials based on bnAb monotherapy have demonstrated limited efficacy largely due to the presence or emergence of resistant viral variants.^27,40,41 However, recent clinical trials evaluating double or triple bnAb combinations consistently report favorable safety profiles and importantly, therapeutic efficacy.^7,37,42 Triple bnAb treatment trials have reported that, in a fraction of participants, virological control persisted for up to one year without additional bnAb or conventional ARV treatment, even after circulating bnAb titers had substantially declined.^7,10 The lack of success in achieving this outcome in other patients most likely reflects failure of the bnAb treatment to adequately cover the extant variant population.

Previous structural and ensemble-based approaches, including electron microscopy and biochemical analyses, have provided important insights into Env-antibody interactions but largely yield static or population-averaged information that cannot resolve heterogeneous combinatorial antibody occupancy within intact virion populations. Likewise, earlier FCS-based studies, including our own, primarily examined the binding of individual antibodies or ligands to virions. In the present study, we selected representative bnAbs targeting three distinct and clinically relevant Env vulnerability sites: the CD4 binding site antibody N49P7, the V3-glycan supersite antibody PGT121 and the trimer apex-directed antibody PGDM1400.^26,43,44,45 These antibodies were chosen based on their breadth, potency, and complementary epitope specificities, thereby enabling evaluation of how distinct bnAb classes concurrently engage heterogeneous HIV-1 populations.^46,47,48 N49P7 was selected as the representative CD4bs bnAb because of its broad and potent neutralizing activity, recognition of highly conserved CD4bs-associated gp120 elements and previously demonstrated robust binding to intact virions in single-particle FCS studies, making it well suited for quantitative combinatorial occupancy analysis in the present FCS/FRET framework.⁴⁶ Using quantitative cross-correlation FCS, three-color coincidence analysis, burst counting and FRET-FCS, the present work establishes a single-particle framework that directly measures the fraction of unmodified virions simultaneously occupied by one, two, or three distinct bnAb classes in solution (Figure 1). Importantly, this approach enables quantitative comparison of combinatorial occupancy patterns across genetically distinct HIV-1 strains and relates these occupancy subclasses to functional neutralization activity. By integrating single-virion occupancy measurements with neutralization analyses, the study provides important insight into how the extent and pattern of multi-bnAb engagement across virion populations influence antiviral potency and may contribute to reduced viral escape.

Figure 1.

Multicolor single-virion bnAb-binding scenarios

Potential bnAb binding situations distinguishable by multicolor burst counting, cross-correlation FCS, and three-color FCS-FRET.

Results

Broadly neutralizing antibodies targeting distinct HIV-1 Env epitopes

For this study, broad and potent representatives of bnAbs classes that have been evaluated in preclinical and clinical studies were used; including the anti-CD4 binding site-N49P7; anti-V3 glycan-PGT121; and anti-V2 apex-PGDM1400. Each bnAb was labeled with different Alexa fluorophores and tested against CCR5-tropic tier 1 BaL pseudovirus, tier 2 clade B YU2 infectious molecular clones, clade C 1086c transmitted-founder viruses and CXCR4-tropic NL4-3 infectious molecular clones (see STAR Methods). Previously, we utilized FCS and fluorescently labeled proteins to study the binding of individual anti envelope monoclonal antibodies (mAbs) or soluble CD4 (sCD4) to HIV particles from various strains, with all components in solution.^46,47,48,49 These experiments demonstrated that Alexa Fluor labeled broadly neutralizing anti gp120 antibodies 2G12, PG9, b12 and N49P7 as well as the non-neutralizing anti-gp41 mAb F240, bound efficiently and reliably to virions.^{47,48,50,51,52} Unlike previous studies that only examined mAb binding to virions, this work directly counts individual virions bound by one, two or three types of bnAbs, offering quantitative data on both bound and unbound particles using an approach having a translational significance. Using cross-correlation measurements and FRET-FCS, the approach quantitatively resolves virion subpopulations simultaneously occupied by up to three bnAb classes across neutralization-sensitive and resistant HIV-1 strains.^47,48

Quantitative analysis of different bnAbs binding to single HIV-1 BaL virions by FCS

Prior to performing FCS and FRET analyses, we evaluated whether fluorophore conjugation altered bnAb antigen recognition. Binding of unlabeled and fluorophore-labeled PGDM1400, PGT121 and N49P7 was compared by ELISA. Overall, labeled antibodies retained binding activity comparable to their unlabeled counterparts, indicating that fluorophore conjugation did not substantially impair antigen recognition. These findings support the use of the labeled bnAbs in subsequent single-particle FCS and FRET experiments. In order to quantify the dynamics of interactions between multiple bnAbs and HIV-1 particles three colored FCS experiments were undertaken. Two complementary fluorescence strategies were used to distinguish overall combinatorial bnAb occupancy from proximity-dependent co-engagement on individual virions. Simultaneous excitation at 470 nm, 560 nm and 635 nm directly excited PGDM1400-A488, PGT121-A594, and N49P7-A647 respectively enabling quantitative analysis of virions bound by one, two, or three bnAbs through burst coincidence and cross-correlation measurements. In contrast, single-wavelength excitation at 470 nm selectively excited the donor fluorophore PGDM1400-A488, such that acceptor-channel autocorrelation signals arose predominantly through FRET, reporting closely spaced multi-bnAb co-occupancy events on a single trimer on the virion surface. Three-color FCS experiments were performed with HIV-1 BaL pseudoviruses (1.6×10⁹/mL) and triple bnAb combinations evaluating the bnAb-virion binding signatures in solution. Reactions used dilutions of virus that place ≤1 virion in the femtoliter focal volume at any one time. In such experiments, it is necessary to distinguish bound versus unbound bnAb fluorescence burst signals. To do so, we considered that unbound bnAbs in solution have faster diffusion rates versus those bound to virions. Thus, the burst intensities of free bnAbs will, on average, be lower than those of bound virions because they exhibit a shorter dwell time (or faster diffusion time) in the detection window. Bursts from labeled bnAb with the pseudovirus were recorded over a 120 s period (Figures 2A–2C). The threshold was conservatively set at twice the average burst intensity detected for the antibody only (see STAR Methods). Burst intensities above this threshold in the reactions with BaL were considered to be bnAb bound virions. On excitation at individual excitation wavelengths of 635 nm, 560 nm, and 470 nm for the virion bound bnAbs N49P7-A647, PGT121-A594, and PGDM1400-A488, the number of bursts were obtained (Figures 2A–2C). These numbers translated into a count of 9×10⁸, 7.5×10⁸ and 6.5×10⁸ virions/ml bound to N49P7-A647, PGT121-A594, and PGDM1400-A488 respectively (see STAR Methods). Coincident bursts from three separate channels are depicted in Figure 2D. Subsequently, the number of virions bound to all three antibodies were found to be 5×10⁸/mL. These numbers are reflected in the percentage virion binding to individual and all three antibodies at the same time (Figure 2E). The cross-correlated signals using labeled bnAb combinations, PGDM1400 + PGT121 + N49P7 was fitted to a single component 3D-diffusion model with a diffusion coefficient of 6 μm²/s, consistent with a 100 nm retroviral particle (Figures 2F–2H). At this point it is important to mention that the cross-correlation signals between any two channels of detection arises from the coincident bursts only. Since the cross-correlation functions for all three combinations yield the same diffusion coefficient of 6 μm²/s, it is reasonable to conclude that coincident binding of PGDM1400 + PGT121 + N49P7 to the BaL pseudovirion has occurred. Furthermore, the auto-correlation curve (Figure 2I), corresponding to only the coincident bursts (Figure 2D) could similarly be fitted to a diffusion coefficient of 6 μm²/s. This quantifies the fractional virion population which is bound to all three antibodies.

Figure 2.

Quantification of multiple bnAbs binding to BaL virions by single-particle burst analysis and three-color FCS

Intensity-time traces for the interaction of bnAbs (A) N49P7-A647, (B) PGT121-A594, and (C) PGDM1400-A488 with BaL pseudovirus in three separate channels on excitation at individual wavelengths of 635 nm, 560 nm, and 470 nm corresponding to the three labeled antibodies. The threshold for detecting fluorescently labeled bnAb bound virion bursts is set at two times the intensity from the fluorescent bnAbs alone.

(D) Coincident bursts corresponding to the population of the virions bound to all three antibodies at the same time.

(E) The fraction of BaL pseudovirus adopting a lower diffusion coefficient by binding to single or all three bnAbs at the same time. The binding fraction is calculated from the bursts shown in (A–D). Data are presented as mean ± SD from triplicate measurements. Cross-correlation curves of (F) N49P7-A647 and PGT121-A594; (G) PGDM1400-A488 and PGT121-A594; (H) N49P7-A647 and PGDM1400-A488 with BaL pseudovirus on excitation at wavelengths of 635 nm, 560 nm, and 470 nm simultaneously.

(I) Correlation curve generated from the coincident bursts signifying the fractional population of the virions bound to all three bnAbs.

Quantitative FRET-based analysis on combination of bnAbs bound to HIV-1 BaL virions

In three-color FRET experiments the sample was excited at a single wavelength of 470 nm corresponding to the excitation of A488 labeled PGDM1400 and bursts were counted in all three channels (Figures 3A–3C). To ensure a single-donor system for three-color FRET, A488 labeled PGDM1400 was selected as the donor because only a single PGDM1400 molecule binds to an Env trimer on the virion surface.¹⁹ The appearance of bursts in the acceptor channels corresponding to PGT121-A594 and N49P7-A647 upon donor excitation indicates FRET between fluorophore-labeled bnAbs in close spatial proximity. When interpreted together with the coincident burst detection and virion-scale diffusion coefficient (∼6 μm²/s) obtained from auto and/or cross-correlation analyses, these data strongly support simultaneous association of the three bnAbs on the same intact virion and are most consistent with engagement of the same Env trimer. Accordingly, engagement of three different bnAbs to the same Env trimer is interpreted as an integrated biophysical inference from the combined measurements rather than as a directly resolved structural observation. Following the similar burst counting approach as described in the previous section, the percentage virion bound population to each of the labeled antibody were calculated (Figure 3I). A striking feature in this case was the relatively less virion bound population to N49P7-A647 and PGT121-A594 (Figure 3I) as compared to the former (Figure 2D). This can be explained by taking into account that in the former case the sample was illuminated simultaneously at 470 nm, 560 nm, and 635 nm corresponding to the excitation wavelengths in all three channels. It has to be taken into account that the percentages derived from the FRET experiments do not represent total virion binding by PGT121-A594 or N49P7-A647, but rather the subset of virions generating detectable acceptor signals under donor excitation; therefore, these values are influenced by FRET efficiency and fluorophore geometry in addition to antibody occupancy. Hence the virion bound PGT121-A594 and N49P7-A647 emanated signals (bursts) in the latter case is due to FRET that can only occur when all three bnAbs are on a single envelope on a virion. Importantly, the interpretation of simultaneous multi-bnAb engagement does not rely on FRET measurements alone. While FRET establishes donor-acceptor proximity within the ∼2–10 nm range, the coincident fluorescence bursts and correlation analyses additionally demonstrate that these fluorophore signals originate from particles exhibiting a diffusion coefficient of ∼6 μm²/s, consistent with intact ∼100 nm retroviral particles. Species such as free antibodies, dissociated Env fragments, or membrane-associated debris would be expected to exhibit substantially faster diffusion kinetics. Thus, the combined FRET and diffusion-based measurements strongly support the conclusion that the fluorophore-labeled bnAbs are associated with the same intact virion and given the known dimensions of the HIV-1 Env trimer, these observations are most consistent with simultaneous engagement on the same Env trimer.⁵³

Figure 3.

FRET detection of concurrent bnAbs binding to individual Env trimers on the surface of HIV-1 BaL virions

Burst counts from intensity-time traces for the interaction of bnAbs (A) N49P7-A647, (B) PGT121-A594, and (C) PGDM1400-A488 with BaL pseudovirus in three separate channels on excitation at a single wavelength of 470 nm. The threshold for detecting fluorescently labeled bnAb bound virion bursts is set at two times the intensity from the fluorescent bnAbs alone. Autocorrelation curves of (D) acceptor N49P7-A647 (A2); (E) acceptor PGT121-A594 (A1); (F) donor PGDM1400-A488 (D) with BaL pseudovirus. FRET histograms of triple bnAb combinations reacted with virions. FRET efficiencies between (G) PGDM1400-A488 (D) to PGT121-A594 (A1) (one-step FRET).

(H) Two-step FRET process from PGDM1400-A488 (D) to PGT121-A594 (A1) to N49P7-A647 (A2). The solid lines in (G and H) are fitted with a Gaussian distribution to the experimental FRET histogram data.

(I) The binding fraction is calculated from the bursts shown in (A–C). Data are presented as mean ± SD from triplicate measurements.

The autocorrelation curves in the donor channel (PGDM1400-A488) corresponding to excitation at 470 nm was fitted to a sum of two components in a 3D diffusion model (Figure 3F).⁴⁷ The slow component of 6 μm²/s corresponds to the bnAb bound virion particle and the faster one is because of unbound labeled antibodies. However, in both the acceptor channels (PGT121-A594 and N49P7-A647) the decay of the autocorrelation could be fitted to a single component of 6 μm²/s indicating the coincident binding of two bnAbs or all three bnAbs to the same Env on a virion surface (Figures 3D and 3E). In the studies involving three-color FCS-FRET, BaL pseudovirus was reacted with a cocktail of the labeled antibodies: donor (A488 “D”) PGDM1400, acceptor-1 (A594 “A1”) PGT121 and acceptor-2 (A647 “A2”) N49P7. These fluorophores were selected such that D does not efficiently activate A2. The one step FRET (D-A1) followed a Gaussian distribution with an average efficiency of approximately 40% (Figure 3G), while the average three-color FRET efficiency (D-A1-A2) was predictably lower at around 20% (Figure 3H). The FRET method relies on standard dipole-dipole interactions between paired donor and acceptor fluorophores, which occurs within the distance range of 2–10 nm. Thus, paired fluorescent bnAbs will not create FRET signals unless they bind to tightly localized epitopes. This gives a quantitative estimate of the fraction of antibodies bound to the same envelope on a virion surface.

Quantitative analysis of bnAbs binding to different clades of HIV-1 virions by FCS and FRET

To verify the proposition the same set of experiments were performed with YU2 (1.8×10¹⁰/mL). HIV-1 YU2 is a Tier 2, clade B, CCR5-tropic infectious molecular clone which is moderately resistant to neutralization.⁵⁴ Upon burst analysis following excitation at individual excitation wavelengths of 635 nm, 560 nm, and 470 nm for the YU2 bound bnAbs, the number of virion bound bnAbs were found to be 8×10⁹, 4.9×10⁹, and 5.2×10⁹/mL for N49P7-A647, PGT121-A594, and PGDM1400-A488 respectively. The number of virions bound to all three bnAbs was found to be 3.2×10⁹/mL (Figure 4D). These numbers are reflected in the percentage virion binding to individual and all three bnAbs at the same time (Figure 4E).

Figure 4.

Three-color FCS analysis of multi-bnAb interactions with YU2 virions

Single-particle measurements characterizing the binding dynamics between a combination of bnAbs and individual YU2 HIV-1 particles. Intensity-time traces for the interaction of bnAbs (A) N49P7-A647, (B) PGT121-A594, and (C) PGDM1400-A488 with YU2 viruses in three separate channels on excitation at individual wavelengths of 635 nm, 560 nm, and 470 nm corresponding to the three labeled antibodies.

(D) Coincident bursts corresponding to the population of the virions bound to all three antibodies at the same time.

(E) The fraction of YU2 virus adopting a lower diffusion coefficient by binding to single or all three bnAbs at the same time. The binding fraction is calculated from the bursts shown in (A–D). Data are presented as mean ± SD from triplicate measurements. Cross-correlation curves of (F) N49P7-A647 and PGT121-A594; (G) PGDM1400-A488 and PGT121-A594; (H) N49P7-A647 and PGDM1400-A488 with YU2 on excitation at wavelengths of 635 nm, 560 nm, and 470 nm simultaneously.

On similar lines, three-color FRET experiments were performed with YU2 viruses. These virions were incubated with donor (A488 “D”) PGDM1400, acceptor 1 (A594 “A1”) PGT121 and acceptor 2 (A647 “A2”) N49P7, with excitation at a single wavelength of 470 nm. Coincident bursts detected in all three channels support simultaneous association of all three bnAbs with the same Env structure on the virion surface. Autocorrelation analysis showed a slow component (∼6 μm²/s) for bound virions in both the acceptor channels, indicating coincident binding (Figures 5D and 5E). The autocorrelation trace of the donor channel was fitted to a sum of two components in a 3D diffusion model where the faster component (∼55 μm²/s) corresponds to the unbound PGDM1400-A488 and the slow component (∼6 μm²/s) corresponding to PGDM1400-A488 bound virions. Compared to direct multi-wavelength excitation, fewer virions were detected with PGT121-A594 and N49P7-A647, as their signals here occurred only via FRET (Figure 5I). The one step FRET efficiency (D-A1) averaged ∼46%, while the three-color FRET efficiency (D-A1-A2) averaged ∼27%, reflecting close spatial proximity of all three antibodies on the same viral envelope.

Figure 5.

Quantitative three-color FRET-FCS analysis revealing coordinated binding dynamics of bnAb combinations on HIV-1 YU2 virions

Burst counts from intensity-time traces for the interaction of bnAbs (A) N49P7-A647, (B) PGT121-A594, and (C) PGDM1400-A488 with YU2 viruses in three separate channels on excitation at a single wavelength of 470 nm. Autocorrelation curves of (D) acceptor N49P7-A647 (A2); (E) acceptor PGT121-A594 (A1); (F) donor PGDM1400-A488 (D) with YU2 virus. FRET histograms of triple bnAb combinations reacted with virions. FRET efficiencies between (G) PGDM1400-A488 (D) to PGT121-A594 (A1) (one step FRET).

(H) Two-step FRET process from PGDM1400-A488 (D) to PGT121-A594 (A1) to N49P7-A647 (A2). The solid lines in (G and H) are fitted with a Gaussian distribution to the experimental FRET histogram data.

(I) The binding fraction is calculated from the bursts shown in (A–C). Data are presented as mean ± SD from triplicate measurements.

The clade C transmitted founder (T/F) HIV-1 strain 1086c is a key model in HIV research due to its relevance in early infection and vaccine development.^55,56 As a representative of viruses capable of crossing the mucosal transmission bottleneck, it provides critical insight into the traits enabling successful establishment of infection. 1086c serves as a benchmark for testing immunogen designs helping access antigenic properties, antibody binding profiles and the balance between eliciting broad neutralization and targeting specific protective epitopes. To validate the proposition the same experimental workflow was applied to the 1086c strain which had 2.4 × 10⁹ virions/ml as determined by RT-qPCR. Burst analysis was performed following excitation at 635 nm, 560 nm and 470 nm for 1086c bound fluorescently labeled bnAbs N49P7-A647, PGT121-A594, and PGDM1400-A488 respectively. The resulting burst counts of virion bound labeled antibodies were 1.2×10⁹ for N49P7-A647, 1.1×10⁹ for PGT121-A594 and 8×10⁸/mL for PGDM1400-A488. Notably, the number of virions simultaneously bound to all three antibodies was measured at 4.5×10⁸/mL. These absolute counts were further translated into the corresponding percentages of virion binding for each individual antibody, as well as for the triple bound population, as depicted in Figure 6E.

Figure 6.

Three-color FCS resolves multi-bnAb engagement on HIV-1 1086c virions at the single-particle level

Intensity-time traces for the interaction of bnAbs (A) N49P7-A647 (B) PGT121-A594 and (C) PGDM1400-A488 with 1086c viruses in three separate channels on excitation at individual wavelengths of 635 nm, 560 nm and 470 nm corresponding to the three labeled antibodies.

(D) Coincident bursts corresponding to the population of the virions bound to all three antibodies at the same time.

(E) The fraction of 1086c virus adopting a lower diffusion coefficient by binding to single or all three bnAbs at the same time. The binding fraction is calculated from the bursts shown in (A–D). Data are presented as mean ± SD from triplicate measurements. Cross-correlation curves of (F) N49P7-A647 and PGT121-A594; (G) PGDM1400-A488 and PGT121-A594; (H) N49P7-A647 and PGDM1400-A488 with 1086c on excitation at wavelengths of 635 nm, 560 nm, and 470 nm simultaneously.

In three-color FRET experiments with 1086c virions, excitation at 470 nm and occurrence of concurrent bursts in all channels confirmed binding of PGDM1400-A488, PGT121-A594 and N49P7-A647 to the same Env on a virion. Autocorrelation showed a slow diffusion (∼6 μm²/s) in acceptor channels for bound virions (Figures 7D and 7E), while the correlation curve in the donor channel fit a two-component model distinguishing bound (∼6 μm²/s) from unbound (∼55 μm²/s) antibodies (Figure 7F). Compared to multi wavelength excitation, fewer virions were detected for PGT121-A594 and N49P7-A647 since signals in acceptor channels occurred solely from FRET. One step (D-A1) and three-color (D-A1-A2) FRET efficiencies averaged ∼60% and ∼20%, respectively, indicating close proximity of all three antibodies on the same viral envelope (Figures 7G and 7H). Our proposition was further bolstered from experiments with HIV-1 NL4-3 virions (1.1×10¹⁰/mL by RT-qPCR), burst analysis detected 3.4×10⁹/mL N49P7-A647, 1.55×10⁹/mL PGT121-A594 and 1.05×10⁹/mL PGDM1400-A488 bound virions, including 4.0×10⁸/mL simultaneously bound by all three bnAbs, with the low but detectable occupancy likely reflecting heterogeneous Env presentation and limited antibody engagement of partially accessible or nonfunctional Env despite poor neutralization sensitivity (Figure S1). Three-color FRET analysis of NL4-3 virions confirmed simultaneous binding of PGDM1400-A488, PGT121-A594, and N49P7-A647 with autocorrelation showing a slow diffusion component (∼6 μm²/s) for virion bound antibodies and an additional fast component (∼55 μm²/s) for unbound donor species (Figure S2). Additionally, the three-color fluorescence data were analyzed using pairwise two-color coincidence burst analysis, allowing direct identification of individual virions simultaneously bound by two distinct bnAbs (Figure S3). Quantification of coincident burst events provides a particle resolved and fully quantitative measure of dual antibody occupancy, revealing strain dependent differences in multivalent binding that are not accessible from ensemble averaged measurements.

Figure 7.

Multiparametric three-color FRET-FCS unraveling the binding behavior of bnAb combinations on HIV-1 1086c particles

Burst counts from intensity-time traces for the interaction of bnAbs (A) N49P7-A647 (B) PGT121-A594 and (C) PGDM1400-A488 with 1086c viruses in three separate channels on excitation at a single wavelength of 470 nm. Autocorrelation curves of (D) acceptor N49P7-A647 (A2); (E) acceptor PGT121-A594 (A1); (F) donor PGDM1400-A488 (D) with 1086c virus. FRET histograms of triple bnAb combinations reacted with virions. FRET efficiencies between (G) PGDM1400-A488 (D) to PGT121-A594 (A1) (one-step FRET).

(H) Two-step FRET process from PGDM1400-A488 (D) to PGT121-A594 (A1) to N49P7-A647 (A2). The solid lines in (G and H) are fitted with a Gaussian distribution to the experimental FRET histogram data.

(I) The binding fraction is calculated from the bursts shown in (A–C). Data are presented as mean ± SD from triplicate measurements.

Comparative bnAb binding and neutralization across diverse HIV-1 strains

BnAbs differ not only in their capacity to engage HIV Env but also in their ability to translate binding into functional neutralization. The present dataset highlights these differences by examining the performance of three bnAbs viz. PGT121, N49P7, and PGDM1400. These were tested individually and in combination against BaL pseudoviruses, NL4-3 and tier-2 clade B YU2 infectious molecular clones and finally with clade C transmitted founder 1086c viruses. Alongside traditional neutralization assays, burst analysis from FCS measurements provided quantitative estimates of the percentage of virions bound by each antibody, allowing a direct comparison between binding and functional inhibition. The neutralization curves (Figures S4A–S4D) reveal that PGT121 and N49P7 consistently achieved high levels of neutralization across BaL, YU2 and 1086c whereas PGDM1400 displayed modest neutralization.²¹ For NL4-3, N49P7 and PGDM1400 mediated partial inhibition, but PGT121 was largely ineffective.^57,58 Interestingly the equimolar mixture of all three bnAbs generally retained strong potency across strains, underscoring the benefit of cocktail approaches in broadening coverage (Figure S4 and Table S1). Synagis used as a negative control showed no measurable neutralization, confirming specificity of the bnAb responses.

Discussion

BnAbs that target the HIV Env glycoprotein represent a promising strategy for HIV treatment, functional cure, and prevention. Nevertheless, each bnAb epitope class shows incomplete coverage of the extensive Env diversity present among people living with HIV. As a result, clinical trials based on bnAb monotherapy have demonstrated limited efficacy largely due to the presence or emergence of resistant viral variants. These limitations support the clinical development of bnAb combinations, which can generate polyspecific anti-Env activity to enhance potency and reduce viral escape. In this context, a recent clinical trial evaluating a triple bnAb combination- PGT121, PGDM1400, and VRC07-523LS reported a favorable safety profile and prolonged viral suppression in most participants following ART interruption, with some individuals maintaining viral control for nearly one year, highlighting its potential as an alternative or adjunct to ART.⁷ However, the basis for the improved performance of bnAb combinations remains incompletely understood. Simultaneous engagement of individual virions by multiple bnAb classes may contribute to broader coverage and reduced opportunities for viral escape. Although structural approaches such as electron microscopy and crystallography have been used to address this issue, they provide static, ensemble averaged snapshots that often focus on antigen-Fab interactions and fail to capture the heterogeneity of bnAb virion binding patterns in unmodified virus populations.^59,60 The present approach addresses this limitation by resolving combinatorial antibody occupancy at the level of individual virions in solution, thereby enabling direct quantification of heterogeneous binding subclasses within viral swarms. Importantly, these measurements can be related to neutralization phenotypes across strains, providing a functional framework that is not accessible through ensemble-averaged structural analyses alone. Single particle burst analysis based FCS reveals the fraction of virions bound by each bnAb types, reflecting intrinsic heterogeneity in Env epitope presentation across the virus populations. For BaL pseudovirus, the CD4 binding site antibody N49P7 exhibits the highest binding fraction (∼60%), followed by the V3-glycan antibody PGT121 (∼50%) and the V2-apex antibody PGDM1400 (∼35%–40%), while the fraction of virions simultaneously bound by all three antibodies remains lower (∼30%), consistent with steric and epitope availability constraints at the level of individual particles (Figure 2E). For YU2 isolates, PGDM1400 displays binding of ∼30%, approaching that of PGT121 (∼25%–30%), while N49P7 continues to show the highest overall binding fraction ∼45% (Figure 4E). This behavior is consistent with the strong dependence of PGDM1400 on Env trimer conformation and intact V2-apex quaternary epitope, which appears to be uniformly or stably presented on a subset of YU2 virions (Figure 4E).²¹ Because V2-apex antibodies are highly sensitive to trimer architecture and glycan microheterogeneity, even modest strain dependent differences in V1/V2 loop packing or apex glycan occupancy can translate into substantial changes in the fraction of antibody-bindable particles.⁶¹ By contrast, CD4 binding site antibodies such as N49P7 target a comparatively conserved and broadly accessible epitope, explaining their consistently higher binding fractions across all three viruses examined, including 1086c (∼50%) (Figure 2E, 4E, and 6E). The reduced binding of PGT121 to YU2 relative to BaL and 1086c is also consistent with the known sensitivity of V3-glycan directed antibodies to N332-region glycan composition and loop context.⁶² Finally, the consistently lower fraction of triple antibody bound virions across all strains (∼15%–30%) reflects the combined effects of finite functional Env trimer numbers per virion, epitope competition, conformational heterogeneity, and the stringent requirement that all three antibodies engage the same particle.

The neutralization dose-response curves for BaL, YU2, 1086c, and NL4-3 quantitatively highlight the strain dependent limitations of individual bnAbs that have been extensively investigated in prior studies.^21,63,64,65 V3 glycan targeting antibodies such as PGT121 typically exhibit IC50 values in the range of 0.01–0.1 μg/mL against relatively sensitive tier 1 BaL and some tier 2 viruses including YU2 (Figures S4A and S4B). However they show reduced potency and incomplete neutralization plateaus against NL4-3 and more resistant strains such as 1086c which is consistent with large scale neutralization datasets and cross clade analyses (Figures S4C and S4D).⁶⁶ Similarly V2 apex antibodies such as PGDM1400 have been shown to potently neutralize select strains like YU2 with sub 0.1 μg/mL IC50 values, while exhibiting weak or delayed neutralization against viruses such as BaL and clade C transmitted founder strains due to differences in trimer apex architecture and glycan organization.²¹ An important point to note here is that that apex directed antibodies such as PGDM1400 are particularly sensitive to Env quaternary structure, glycan occupancy, producer-cell-dependent glycosylation and virion preparation conditions, all of which can influence measured neutralization potency across assay platforms.^60,65,67,68 These quantitative disparities are reflected in the right shifted or shallow neutralization curves observed for antibody monotherapy.^69,70 In contrast, the equimolar antibody mixture consistently achieves higher maximal neutralization exceeding 90 percent across all strains, in agreement with previous preclinical and clinical studies demonstrating that bnAb combinations reduce effective IC₅₀ or IC₈₀ values by two to 10-fold and increase the fraction of fully neutralized virions compared with single antibody treatment.^33,69 This was even more evident from the significantly lesser IC₈₀ values obtained when an equimolar mixture of all three bnAbs were used as compared to individual bnAbs (Figures S4J–S4L) for BaL, YU2 and 1086c strains. Though a reversal in trend could be seen for NL4-3 yet it was not statistically significant (p value = 0.16) (Figure S4M). This can be rationalized from the very low binding efficiencies of NL4-3 with either of the three bnAbs (Figure S1E). Importantly, the correlation shown in panel (Figure S4E) reveals a statistically significant (p < 0.04) positive relationship between percentage neutralization and the fraction of virions bound by at least one bnAb, with viruses achieving approximately 80%–90% neutralization corresponding to ≥80% bnAb combination engagement, whereas NL4-3 displays both lower bnAb binding around 40% concomitant with reduced neutralization. The linear correlation between percentage virion binding and percentage neutralization is consistent with a relationship between antibody occupancy and antiviral activity. These findings support the hypothesis that the extent of antibody engagement may contribute to neutralization efficiency across the virus panel. This quantitative correlation is consistent with earlier mechanistic models and experimental observations suggesting that viral escape is minimized when multiple antibody classes simultaneously occupy the same virions rather than distributing across distinct particles.⁷¹ The single-particle FCS, burst coincidence and FRET analyses directly quantify bnAb-associated virion populations, coincident fluorescence events and virion-scale diffusion behavior. From these experimentally observed properties, we infer that multiple bnAbs can simultaneously associate with the same Env epitopes on intact virions. Similarly, the statistically significant relationship between antibody occupancy and neutralization is interpreted as supportive evidence consistent with the proposed model of combinatorial bnAb activity. Although the observed occupancy-neutralization relationships are consistent with enhanced cooperative bnAb activity, the current experiments were not designed to independently manipulate virion occupancy and therefore do not establish causality directly.

Notably, understanding the interactions of bnAbs with HIV Env on replicating virion populations is critical for the development of many immune based HIV/AIDS countermeasures. However, HIV-1 population contains several replication incompetent particles which can present abundant non-functional Env alongside functional spikes yet display immunoreactive epitopes. Further, the size of this subpopulation can differ among strains. Defective particles generally present sparse functional and/or abundant non-functional envelope structures; however, the latter may still express certain immunoreactive neutralizing epitopes. For example, our previous studies using FCS and FCS-FRET to study bnAb binding at the single virion level revealed concurrent binding of non-neutralizing anti-Env mAbs (marking nonfunctional Env) and bnAbs to the same Env structures on several virus strains.^47,48 Conversely, replicating viruses may express nonfunctional Env structures along with functional ones.^72,73 Collectively, these features could substantially “decoy” bnAb pools, thus lowering their efficacy. Theoretically, the most efficacious bnAb or bnAb combination would be one that avoids nonreplicating viruses, focuses on functional envelopes and strongly favors replicating viruses. However, signals marking concurrent bnAb epitope-binding (e.g., by trivalent combinations), when immunochemically permissible, could best reflect the size of the replicating virus subpopulation. Although the antibodies target distinct Env epitopes, the measured coincidence and FRET efficiencies reflect the combined effects of epitope accessibility, antibody occupancy and any potential steric constraints under the experimental conditions. Together, these results show that single-virion bnAb binding profiles are associated with neutralization potency and provide a quantitative framework for investigating the activity of bnAb combinations (Figure 8 and Table S1).

Figure 8.

Schematic overview of combinatorial bnAb binding coverage across virions in virus swarms, quantified using multicolor single-particle burst analysis

Schematic representation of quantified coverage of the combinations of bnAbs binding to virions in virus swarms determined by multi-color single particle burst analyses approaches. Percentage of coverage are noted under each subset (Table S1). (A) Each bnAb individually and at least one bnAb in an equimolar mixture of three; (B) fraction of virions occupied by all three bnAbs at the same time; (C) FRET detection reveals virions on which three bnAbs bind concurrently to single Env structures on virion surfaces.

In conclusion, bnAbs continue to emerge as one of the most promising classes of agents for HIV treatment, prevention and cure strategies offering distinct advantages. Their ability to persist longer in circulation, exert Fc-mediated effector functions and potentially target viral reservoirs positions them as a transformative tool in the ongoing battle against HIV/AIDS. However, the formidable challenge of incomplete strain coverage and the risk of viral escape necessitate rational design of bnAb cocktails that maximize breadth and potency while minimizing redundancy. The development of single particle analytical approaches, including FCS, cross-correlation analyses, three-color coincidence analysis, FRET and quantitative burst counting represents a methodological advancement in this regard. These techniques provide a capacity to determine whether multiple bnAbs colocalize and bind simultaneously to the same virion and/or envelope spike and overlapping binding at the single virion level. The burst analysis approach developed and described in this study extract relevant signals from fluorescence fluctuations, offering a readout for identifying virions bound concomitantly by three distinct fluorophore labeled bnAbs. Likewise, FRET efficiencies arising from pairs of bnAbs provide strong biophysical evidence of close spatial proximity on the viral envelope, supporting overlapping structural engagement on intact virions that is critical for neutralization. Application of these approaches to diverse HIV-1 strains including BaL, NL4-3, YU2, and the transmitted founder strain 1086c demonstrates not only that triple antibody occupancy can be achieved but also binding efficiencies vary across viral subtypes, reflecting the inherent diversity. Importantly, the correlation between the fractions of virion populations bound by one or more fluorescent labeled bnAbs and neutralization levels reveals a quantitative association between antibody-virion occupancy and the inhibition of HIV-1 infectivity. These findings support the importance of quantitative evaluation of bnAb cocktails, providing a framework that may assist future selection and evaluation of bnAb combinations against diverse viral variants. Ultimately, integrating such single particle biophysical approaches with advances in antibody engineering and clinical development will be crucial for realizing the full therapeutic and preventive potential of bnAbs. By bridging structural immunology, quantitative virology and translational medicine this framework sets the stage for the next generation of bnAb based interventions that could complement current antiretroviral therapy in durability, breadth and efficacy.

Limitations of the study

Several limitations of the present study should be acknowledged. First, the analyses were restricted to three bnAbs representing selected Env epitope classes and therefore do not capture the full diversity of antibody specificities relevant to HIV-1 infection. Second, although multiple viral strains were examined, the study evaluated a limited set of isolates and primarily utilized clonal viral populations rather than the heterogeneous viral quasispecies typically present in vivo. Third, the effects of potentially competing or interfering antibodies, including non-neutralizing antibodies and antibodies capable of inducing Env conformational changes, were not investigated. Consequently, the observed patterns of combinatorial occupancy may not fully reflect the complexity of antibody-virus interactions occurring during natural infection or therapeutic administration. Future studies applying this platform to larger antibody panels, including non-neutralizing antibodies and bnAbs known to induce conformational changes alongside patient-derived viral populations and heterogeneous HIV-1 quasispecies, may provide additional insight into antibody cooperativity, competition and Env conformations at the single-virion level.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Krishanu Ray (kray@som.umaryland.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

All data supporting findings of this study are provided within the article and its supplemental information section. This study does not report any original code. Any additional information regarding the data reported in this study is available from the lead contact upon reasonable request.

Acknowledgments

The research reported in this study was supported by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health under Award Number R01-AI172487, R61/R33-AI176561 (K.R.) and P01-AI162242. The authors thank Dr. Shih-Chu Liao from ISS in developing the customized VistaVision analyses software module. The content is solely the authors’ responsibility and does not necessarily represent the official views of the National Institutes of Health. The authors thank the NIH bioart repository for some of the vector diagrams that have been used in the graphical representations.

Author contributions

S.D. performed experiments, analyzed data, and wrote the manuscript. A.T.S. and S.K. contributed to experiments and analyzed data. G.K.L. and A.L.D. analyzed data and edited the manuscript. K.R. designed and performed research, analyzed data, and wrote the manuscript.

Declaration of interests

The authors declare no competing interests. K.R. and A.L.D. are the inventors on US non-provisional patent application no. 18/443,665.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
IgG1	Calbiochem	Catalog #401114
Synagis (anti-RSV mAb)	MedImmune	N/A
mAb PGDM1400	DHVI	https://doi.org/10.1073/pnas.1415789111
mAb PGT121	NIH AIDS Reagent Program - ARP-12343	https://doi.org/10.1038/nature10373
mAb N49P7	In house	https://doi.org/10.1016/j.cell.2018.03.061
Virus strains
HIV-1BaL	AIDS Research and Reference Reagent Program	https://journals.asm.org/doi/10.1128/jvi.03048-13
HIV-1YU2	AIDS Research and Reference Reagent Program	https://journals.asm.org/doi/10.1128/jvi.66.11.6587-6600.1992
HIV-1NL4-3	AIDS Research and Reference Reagent Program	https://journals.asm.org/doi/10.1128/jvi.03048-13
HIV-11086c	The Duke Human Vaccine Institute (DHVI)	https://doi.org/10.1016/j.str.2025.04.016
Chemicals, peptides and apparatus
Trolox	Sigma-Aldrich	Catalog #238813
Critical commercial assays
Bright-Glo™ Luciferase Assay System	Promega	Catalog #E2610
NuPAGE™ 4–12% Bis-Tris Protein Gels	Thermo Fisher Scientific	Catalog # NP0322PK2
Whatman® Optitran® reinforced nitrocellulose membranes	MilliporeSigma	Catalog # 10493396
DMEM	GibcoR	Catalog #21-063-029
DPBS (1×)	GibcoR	Catalog #14190144
FuGENE 6 transfection reagent	FuGENE	Catalog #F6-1000
Alexa Fluor™ 647 Antibody Labeling Kit	Invitrogen™	Catalog # A20186
Alexa Fluor™ 488 NHS Ester	Invitrogen™	Catalog #A20100
Alexa Fluor™ 594 NHS Ester	Invitrogen™	Catalog #A37572
Sodium bicarbonate	Sigma-Aldrich	Catalog #S5761
Press-To-Seal silicone isolators	Grace Bio-Labs	Catalog#665208-25 EA
Zeba™ Spin Desalting Columns, 7K	ThermoFisher SCIENTIFIC	Catalog #89882
Bovine Serum Albumin (BSA) Microbiological Grade Powder	fisherscientific	Catalog #BP9700100
Dynabeads™ CD8	ThermoFisher SCIENTIFIC	Catalog #11148D
Corning™ Costar™ 96-Well	fisherscientific	Catalog #07–200-90
High-binding 96-well plates	Thermo Scientific IMMULON®	Catalog #3455
HRP-Streptavidin	Revvity	Catalog #405210
One Component HRP Microwell TMB Substrate	BioFX	Catalog #TMBW-1000–01
Stop Solution	BioFX	Catalog #NSTP-0100–01
LookOut® One-Step Mycoplasma Detection Kit	Sigma-Aldrich	Catalog MP0050-25TST
Experimental models: Cell lines
HEK293T/17 Cells	ATCC	Catalog # CRL-11268
TZM-bl Cells	NIH AIDS Reagent Program	Catalog # 8129
Software and algorithms
VistaVision	ISS	N/A
Origin	OriginLab	N/A
Prism	Graphpad	N/A

Experimental model and study participant details

The experimental models comprised TZM-bl cells obtained from the NIH AIDS Reagent Program and HEK293T/17 cells obtained from ATCC. Cell lines were authenticated by the respective suppliers prior to distribution and no additional authentication was performed by the authors. Cells were routinely tested for mycoplasma contamination using the LookOut® One-Step Mycoplasma Detection Kit and were confirmed to be mycoplasma-negative prior to use. Suppliers, catalog numbers for all resources used in this study are provided in the key resources table. No human study participants were included in this study.

Method details

Labeling of antibodies

PGDM1400 was labeled using Alexa 488. In a solution containing 73 μL of 1× PBS, 25 μL of PGDM1400 (1 mg/ml) was added followed by addition of 1 μL Sodium bicarbonate (NaHCO₃, 10 M) to maintain a pH of ∼8.3. Finally, 1.6 μL of 5 mM Alexa Fluor 488 NHS ester (Invitrogen, Catalog No. A20000) solution was added to the solution containing the antibody. The reaction mixture was kept in dark at room temperature for 90 min. PGT121 and N49P7 were labeled using similar protocols and concentrations using Alexa Fluor 594 NHS ester (Invitrogen, Catalog No. A20004) and Alexa Fluor 647 NHS ester (Invitrogen, Catalog No. A37573). The labeled antibodies were isolated from unreacted dye using centrifugation through a spin column at 1,200×g for 5 min. Dye to protein ratios were assessed by measuring absorbance at 280 nm for protein and 488, 594 or 650 nm for different dyes, with ratios ranging from 2 to 3 depending on the antibody. Quantification of the labeled antibodies was performed using a UV visible spectrometer (Nanodrop 2000; Thermo-Scientific, Wilmington, DE). Care was taken to maintain a low labeling density, as the experiments were performed under single-molecule fluorescence conditions and minimal perturbation of bnAb functionality was essential.⁴⁸

ELISA for assessment of antigen binding by labeled and unlabeled antibodies

The antigen-binding activity of labeled and unlabeled antibodies was assessed by enzyme-linked immunosorbent assay (ELISA). High-binding 96-well plates were coated overnight at 4°C with BG505 SOSIP at 0.3 μg/mL. Plates were subsequently washed four times with PBS containing 0.05% Tween 20 (PBS-T) and blocked with 5% non-fat dry milk in Tris-buffered saline (TBS). Labeled and unlabeled forms of PGT121, N49P7 and PGDM1400 were diluted to 1 μg/mL in blocking buffer and added to the plates in duplicate wells. Following incubation for 1 h at 37°C, plates were washed four times with PBS-T and incubated for 1 h at 37°C with biotinylated anti-human IgG antibody diluted 1:3000 in blocking buffer. Plates were then washed four times with PBS-T and incubated for 1 h at 37°C with streptavidin-HRP diluted 1:3000 in blocking buffer. Following a final wash step, plates were developed using BioFX One Component HRP Microwell TMB Substrate. The reaction was stopped using BioFX Stop Solution and absorbance was measured at 450 nm using a microplate reader.

Preparation of pseudoviruses and molecular clones

HIV-1 BaL pseudoviruses were produced by co-transfecting HEK293T cells with an Env-deficient HIV-1 backbone plasmid, pNL4-3-ΔE-EGFP and Env expression plasmids.^74,75 Specifically, pHIV-1-BaL 0.1 (sourced from the AIDS Research and Reference Reagent Program, Division of AIDS, NIAID) were used. The transfections were performed using FuGENE 6 (Roche, Indianapolis, IN) as the transfection reagent, maintaining a 3:1 reagent to DNA ratio. To generate the HIV-1 NL4’-3 infectious molecular clone same HEK293T cells were transfected with a full length pNL4-3 HIV genome expression plasmid, which was acquired from the AIDS Research and Reference Reagent Program, Division of AIDS, NIAID using FuGENE at a reagent to DNA ratio of 3:1.¹⁵ Replication competent HIV-1 YU2 was generated by cloning the YU2 envelope gene into the NL4-3 molecular backbone followed by transient transfection of HEK293T cells and harvesting of cell free viral supernatants.⁷⁶ Transmitted founder virus 1086c was prepared by infecting CD8 depleted PBMCs (using CD8 Dynabeads-Invitrogen) with 1086c (TCID₅₀ 3.2×10⁶).^77,78 Viral RNA copy numbers of all the viruses were quantified using reverse transcription quantitative PCR (RT-qPCR) following extraction of viral RNA from the samples. RNA copies for the HIV BaL pseudovirus preparation were 3.2×10⁹/mL by RT-qPCR. Assuming on average two viral RNA molecules/virion, this translates to a virion concentration of 1.6×10⁹/mL.

Sample preparation: FCS and FRET experiments

In this method, HIV-1 BaL pseudovirions, NL4-3 and YU2 molecular clones and transmitted founder virus 1086c were incubated with 100 μg/mL of non-specific IgG1 (1.5 μL of a 7 mg/mL stock) for 90 min at 37°C to block non-specific binding. Subsequently, 2 μL of Alexa-conjugated test mAbs (20 nM each) was added to a reaction volume of 20 μL and incubated for another 90 min at 37°C so as to maintain the final concentration of the labeled antibody at 2 nM. For spectroscopic analysis, 11 μL of the reaction mixture was loaded onto an FCS slide reservoir, sealed and placed on an ISS Q2 confocal microscope equipped with a high numerical aperture water objective (Olympus 60× magnification, NA 1.2) and a Fianium SC-400 super-continuum laser for excitation.^46,49,79,80

Three-color fluorescence correlation spectroscopy

Fluorescence Correlation Spectroscopy (FCS) is a technique used for real time detection of multiple protein-protein or antibody-virion interactions in solution by analysing the diffusion and reaction kinetics of fluorescently-labeled biomolecules.^46,48,49,79 Three-color FCS has been employed to study the binding of three distinct bnAbs to a single HIV-1 virion. For example, the interaction of Alexa 488, 594 or 647 labeled bnAbs (PGDM1400, PGT121, N49P7) with HIV-1 virions was tracked by observing the diffusion of their fluorescent labels in three distinct detection channels.

The setup employed an acousto-optical tunable filter (AOTF) to select excitation wavelengths, which were further refined using narrow bandpass clean-up filters. The samples were simultaneously excited at 470, 560 and 635 nm followed by collection of fluorescence signals from Alexa 488, Alexa 594 and Alexa 647-labeled bnAbs in three separate detection channels with high quality emission band pass filters (500–550 nm (D), 575–625 nm (A1) and 650–720 nm(A2)) over a 120 s measurement period within a fixed detection volume (∼1 fL). ISS Vista Vision software was used to compute the autocorrelation and cross-correlation functions of fluorescence fluctuations, which correlates the intensity of fluorescence at a given time with the intensity at a delayed time interval (τ). This function, calculated over time intervals ranging from 10^–2 ms–10² ms, provided insights into the binding dynamics of the bnAbs. The autocorrelation function G(τ) of only the labeled antibodies fitted to a single-component 3D diffusion model yields a diffusion coefficient value of 55 μm²/s, which is in line with our previous reports.^47,80 In experiments involving bnAbs and virions, the autocorrelation data were analyzed using a two-component 3-D diffusion model with one component corresponding to the diffusion coefficient of 55 μm²/s (unbound bnAb) and another one of 6 μm²/s, consistent with the expected behavior of fluorescent bnAbs bound to a 100 nm retroviral particle. Cross-correlation metrics were used to assess whether the signal intensities in the three separate detection channels were correlated, indicating synchronized or independent fluctuations. Positive correlation amplitude was observed only for pairs of coincident photon counts from any of the two distinct channels (Ch1-Ch2, Ch2-Ch3, Ch1-Ch3). Three-color cross-correlation data showing positive correlation amplitude, specifically for two different antibodies in the presence of an HIV-1 virion, provided evidence that the antibodies were bound to the same virion particle. In the absence of test viruses, no cross-correlation signals or coincident binding events were observed when PGDM1400-A488, PGT121-A594 and N49P7-A647 were mixed and analyzed via three-color FCS. Furthermore, Alexa Fluor 647 labeled nonspecific human IgG1 served as a negative control exhibiting no detectable binding to any of the virions tested.^47,48 Error bars denote the standard deviation calculated from a minimum of three independent experimental replicates reflecting the variability and reproducibility of the measurements.

Burst counting analysis

The fundamental principles of FCS to investigate bnAb-virion interactions involve labeling bnAbs with distinct fluorophores and introducing them to virus samples. As individual virions pass through the ∼1 fL focal volume, the detector channels capture signal “bursts” from the bound bnAbs. To ensure precise measurements, the virion densities are set so that typically only one particle occupies the focal volume at any time. The system was calibrated and cross-validated for burst analyses using commercially available fluorescent beads (TetraSpeck Microspheres, ThermoFisher) of 100 nm diameter with a translational diffusion coefficient of ∼5 μm²/s. Each virion generates a characteristic fluorescent burst. Previously we analyzed these bursts using autocorrelation and cross-correlation plots, providing insights into the binding efficiencies of individual anti-Env Abs to single virions and the simultaneous binding of two bnAbs or Abs to the same virion.^47,48,49 To partition particle subsets based on bnAb binding signals, the approach required identifying individual bursts (representing virions), assigning specific fluorescence characteristics (indicating bound bnAbs) to each and quantifying the number of bursts sharing the identical fluorescence signatures. To achieve this, a “burst analysis” method was developed. This approach extracts the necessary signals from fluorescence fluctuation data using the analysis module and the maximum likelihood estimator in ISS Vista-vision software.^81,82 Fluorescence intensity thresholds, which determine when a burst qualifies as a virion were used operationally to distinguish virion-associated bursts from rapidly diffusing free fluorophore-conjugated bnAbs in solution and are established from fluorescence fluctuation data for fluorophore-conjugated bnAbs tested at the final concentration in the absence of the virus. The number of bursts detected during the experimental time frame is analyzed using the translational diffusion time (4 msec) of the virion-bnAb complex in a confocal volume of 1 fL to estimate the number of virions per unit volume in a sample displaying a specific fluorescent bnAb binding pattern. Thresholding alone was not used to assign virion populations occupied by single, double or triple antibody combinations. Rather, dual and triple associated events were identified through coincident burst detection across independent fluorescence channels and further validated by auto- and/or cross- correlation analyses yielding diffusion coefficients consistent with intact virion particles (∼6 μm²/s) (Figure 2I). Under limiting conditions, each burst is counted and interpreted as a single virion. The fluorescence burst signal characterizes the types of bnAbs bound to the virion and bursts with a specific signal pattern can be quantified and compared to the total virion count in the sample. In practice, burst detection was performed under both single-color and multi-color excitation conditions. For single-color excitation at 470 nm (PGDM1400-A488 donor), bursts appearing in the acceptor channels at 560 nm (PGT121-A594) and 635 nm (N49P7-A647) were observed exclusively through FRET from the donor channel enabling indirect identification of acceptor labeled bnAb binding. In contrast, during multi-color excitation (470, 560 and 635 nm), bursts were independently detected in each channel after applying the predefined fluorescence thresholds. Thresholds were conservatively established from fluorophore-conjugated bnAbs in the absence of virus and set at approximately 2-fold above the average free-antibody burst intensity distribution to eliminate the inclusion of rapidly diffusing unbound species. Coincident bursts were then defined as fluorescence events that appeared simultaneously in all three detection channels signifying virions simultaneously bound by PGDM1400, PGT121 and N49P7. The coincident burst dataset was subsequently analyzed to obtain the autocorrelation curve which allowed for the extraction of diffusion properties of labeled multiple bnAb bound virion complexes. Meanwhile, cross-correlation between channels provided quantitative evidence for the co-binding of multiple bnAbs to the same virion. To assess robustness of the burst-classification procedure, analyses were repeated across a range of threshold values surrounding the selected operational cutoff. While the absolute occupancy fractions varied modestly with threshold definition, the overall trends in single, dual and triple-coincident virion populations, as well as the associated virion-scale diffusion coefficients, remained qualitatively preserved (Figure S5). This analysis therefore supports the robustness of the main qualitative trends across the tested threshold range. To eliminate potential sources of false-positive coincidence events viz. random temporal overlap of bursts, detector noise and spectral cross-talk between fluorescence channels, different aspects of analysis and acquisition have been put into place to minimize these effects. First, experiments were performed under limiting dilution conditions corresponding to ≤1 virion within the femtoliter observation volume at a given time, reducing random coincidence probability. Second, fluorophore combinations were selected to minimize spectral bleed-through. Third, coincident events were independently verified by cross-correlation analyses yielding diffusion coefficients of ∼6 μm²/s, consistent with intact retroviral particles rather than free antibodies or fragmented material.

FRET analysis

For three-color FRET measurements, PGDM1400 labeled with Alexa Fluor 488 (donor, D), PGT121 with Alexa Fluor 594 (acceptor, A1) and N49P7 with Alexa Fluor 647 (acceptor, A2) were used. Complexes were formed with BaL, NL4-3, YU2, transmitted founder 1086c prior to FRET measurements. Experiments were carried out on a confocal microscope (ISS Q2) equipped with a supercontinuum laser and AOTF for selective excitation of fluorophores as they diffused through the 1 fL confocal volume. Data acquisition and histogram analyses were performed using ISS Vistavision software.

Fluorescence signals from the donor and both acceptors were separated with dichroic beam splitters and detected on avalanche photodiodes (APDs) using time-correlated single photon counting in TTTR mode on a Becker & Hickl SPC-150 module. High quality bandpass filters (Chroma) were used for laser cleanup and detection channels. Single-photon data were binned in 4 ms intervals, corresponding to the diffusion time of a 100-nm virion, yielding intensity-time traces of 120 s per measurement. Burst thresholds were applied independently to each channel to discriminate single-particle fluorescence from background. Coincident bursts were recorded across donor and acceptor channels and FRET efficiencies were calculated using E = I_A/(I_A+γI_D), where I_D and I_A represent donor and acceptor photon counts respectively with corrections for detection efficiency (γ) between channels. The resulting data were compiled into FRET efficiency histograms representing the fraction and distribution of FRET events per sampling bin over the acquisition period. Error bars represent the standard deviations from a minimum of three independent experimental replicates reflecting the variability and reproducibility of the measurements. In the absence of test viruses, no FRET signals were detected when PGDM1400-A488, PGT121-A594 and N49P7-A647 were added to the reaction mixture and analyzed by the three color FRET-FCS method.

Neutralization assay

This assay measures neutralization in TZM-bl cells by assessing the reduction in Tat-induced luciferase (Luc) reporter gene expression following a single round of virus infection. Neutralizing activity was assessed using HIV-1 viruses and TZM-bl reporter cells as previously described.⁸³ Briefly, 50 μL of different HIV-1 viruses were incubated with 50 μL of serially diluted antibodies (starting at 0.003–9 μg/mL) for 1 h at 37 °C in a CO₂ incubator. Next, TZM-bl cells (10,000 cells/well) were added to the virus-bnAb solution and incubated at 37 °C in 6% CO₂ incubator. Virus control wells (cells plus virus) and background controls (cells only) were included in this assay. After 72 h, Bright-Glo reagent (Promega, Madison, WI, USA) was added and RLU was measured using SpectraMax ID3 (Molecular Devices, CA, USA) and the percentage of neutralization was calculated. Error bars represent standard deviations of at least three replicates. The normalized neutralization curves (Figures S4F–S4I) for the additive (blue) were constructed using a sum of the neutralization values of each individual bnAb at 1/3^rd the concentration of the mixture (0.33 μg/mL) at which the percentage neutralization was determined experimentally. A two tailed Welch’s t test was performed on the IC₈₀ values obtained from the fitted neutralization curves to ascertain the statistical significance in this variation.

Quantification and statistical analyses

Statistical analyses were performed using GraphPad Prism version 10.1.2. Pairwise comparisons between pseudovirion groups were conducted using two-tailed Welch’s t-tests, which account for potential differences in variance between groups. Linear relationships between variables were assessed using simple linear regression and the statistical significance of each regression was determined by testing whether the fitted slope differed significantly from zero. The strength of each linear relationship was expressed as the coefficient of determination (R²). Exact p values, R² values and the number of independent replicates (n) are provided in the corresponding figure legends.