Acquisition of Quaternary Trimer Interaction as a Key Step in the Lineage Maturation of a Broad and Potent HIV-1 Neutralizing Antibody

SUMMARY

Although most broadly neutralizing antibodies (bNAbs) specific for the CD4-binding site (CD4-BS) of HIV-1 interact with a single gp120 protomer, a few mimic the quaternary binding mode of CD4, making contact with a second protomer through elongated heavy chain framework 3 (FRH3) or complementarity-determining region 1 (CDRH1) loops. Here, we show that a CDRH3-dominated anti-CD4-BS bNAb, CH103, establishes quaternary interaction despite regular-length FRH3 and CDRH1. This quaternary interaction is critical for neutralization and is primarily mediated by two FRH3 acidic residues that were sequentially acquired and subjected to strong positive selection during CH103 maturation. Cryo-electron microscopy structures confirmed the role of the two FRH3 acidic residues in mediating quaternary contact and demonstrated that CH103 reaches the adjacent gp120 protomer by virtue of its unique angle of approach. Thus, the acquisition of quaternary interaction may constitute a key step in the lineage maturation of a broad and potent HIV-1 neutralizing antibody.

eTOC

Liu et al. identify a neutralizing antibody against HIV-1, designated CH103, that binds simultaneously to two HIV-1 envelope subunits and they illustrate the structural basis for such quaternary interaction. The authors show that this unusual feature was acquired during the antibody maturation path, progressively increasing the potency of the precursor forms.

Graphical Abstract

INTRODUCTION

The development of a protective HIV-1 vaccine remains a high priority for ending the HIV/AIDS pandemic on a global scale. Although broadly neutralizing antibodies (bNAbs) directed against different sites of vulnerability ^1,2 in the envelope (Env) trimer have been isolated from chronically HIV-1-infected individuals and shown to confer sterilizing immunity in pre-clinical models ^3–8, their induction by a vaccine has proven very difficult ^9–11. This drawback is mainly due to the protected configuration of the conserved epitopes targeted by bNAbs, which are effectively concealed by an extensive glycan shield that covers most of the exposed surface of Env, in addition to steric occlusion and conformational masking ^1,12. To overcome these hurdles, several bNAbs are endowed with unusual structural and biological features such as a long third complementarity-determining region of the heavy chain (CDRH3), high levels of somatic mutations and polyreactivity with non-HIV-1 antigens, all of which suggest substantial barriers to their elicitation ^2,13. An additional obstacle is the relative rarity of the germline precursors for bNAbs, also defined as unmutated common ancestors (UCAs), in the human B-cell repertoire ^2,9. Once such precursors are engaged and expanded, a painstaking process of affinity maturation begins in order to reach the final maturation stage ^14–16.

One of the primary targets of bNAbs in the HIV-1 Env is the binding site for the main viral receptor, CD4 ^15,17. Several potent bNAbs against the CD4-binding site (CD4-BS) have been isolated from chronically infected individuals, and their lineage evolution and structural mode of interaction have been extensively characterized ^18–25. Most anti-CD4-BS bNAbs have high levels of somatic mutation, suggesting complex and prolonged maturation pathways ²⁶. We previously reported that the CD4-BS has a quaternary configuration as it involves not only a primary CD4-interactive region in the gp120 outer domain, but also a secondary binding site, defined CD4-BS2, that is located in the inner domain of the neighboring protomer and includes four highly conserved charged residues, E62, E64, H66, and K207 ²⁷. Although the majority of anti-CD4-BS bNAbs interact with a single gp120 protomer, a few VH gene-restricted CD4-BS bNAbs mimic CD4 in its quaternary trimer interaction, establishing contact with two neighboring gp120 protomers ²⁷. To achieve the secondary contact, these antibodies have developed unusual structural features, namely, elongated and acidic heavy chain framework 3 (FRH3) loops, as in the case of VRC03 and VRC06, or complementarity-determining region 1 (CDRH1) loops, as in the case of VRC-CH31 and 1–18 ^24,28. Mutagenesis studies have demonstrated that the secondary contact plays a major functional role in increasing the binding affinity and the neutralization potency of these uncommon bNAbs ^28,29.

Longitudinal studies have shown that HIV-1 bNAbs evolve through different molecular paths ^20,23,30. A precise characterization of these paths and the different mechanisms and modalities of bNAb maturation may provide invaluable information to guide vaccine design. We previously reported the isolation of a unique anti-CD4-BS bNAb, CH103, from an African individual (CH505) who was followed from the acute phase of HIV-1 infection to the time of bNAb development, more than 3 years later ²². The CH103 bNAb lineage was first detected as early as 14 weeks after acute infection and found to be less mutated than most other anti-CD4-BS bNAbs. CH103 uses a CDRH3-dominated mode of epitope recognition, which is distinct from the CDRH2-mediated interactions with the CD4-binding loop adopted by VH gene-restricted anti-CD4-BS antibodies derived from VH1–2 and VH1–46 ²¹. Sequential sampling led to the identification of multiple intermediate forms during the lineage evolution of CH103, providing a unique window into the molecular mechanisms for the progressive acquisition of high-affinity binding that underlies the high potency of this bNAb.

In the present study, we provide structural and functional evidence that CH103 establishes quaternary interaction with the HIV-1 Env trimer despite belonging to the CDRH3-dominated class of anti-CD4-BS bNAbs and lacking the typical extensions in the FRH3 or CDRH1 loops that characterize quaternary-interactive VH gene-restricted CD4-BS bNAbs. Furthermore, we found that two acidic residues in the FRH3 region that are critical for establishing the quaternary interaction were not present in the antibody UCA, but were progressively acquired during the antibody lineage maturation. These findings identify the acquisition of quaternary interaction with the HIV-1 Env trimer as a critical mechanism for increasing the neutralization potency of a broad and potent HIV-1 neutralizing antibody during the process of affinity maturation.

RESULTS

Structural modeling predicts a quaternary interaction of bNAb CH103 with the CD4-BS

While most bNAbs directed against the CD4-BS of the HIV-1 Env trimer interact with a single gp120 protomer, a few were shown to engage two neighboring gp120 protomers, thus establishing a quaternary interaction that plays an important functional role in trimer binding and virus neutralization ^24,28. These quaternary-binding antibodies, all of which belong to the VH gene-restricted group of anti-CD4-BS bNAbs ²¹, display unusual structural features that enable them to reach a second gp120 protomer, such as elongated acidic heavy chain FRH3 loops, e.g., VRC03, or CDRH1 loops, e.g., 1–18 (Figure 1A). The present study is focused on CH103, an anti-CD4-BS bNAb derived from an African donor, whose lineage evolution has been thoroughly characterized ²⁰. Although CH103 is a CDRH3-dominated anti-CD4-BS bNAb ²³ and encompasses FRH3 and CDRH1 loops of regular length (Figure 1B), by structure alignment and and visualization in the context of the Env trimer, we predicted that this bNAb can establish a quaternary interaction with the HIV-1 Env trimer, with its acidic FRH3 loop making contact with a second gp120 protomer (Figure 1C). Our structural model predicted the quaternary contact to be mediated, on the gp120 side, by residues H66 and K207 of CD4-BS2 and a basic patch at the base of the V3 loop formed by residues R304, R308 and Y318, and on the antibody side, by two acidic residues at the apex of the FRH3 loop of CH103, i.e., E75 and D76 (Figure 1C).

Figure 1. Structural modeling of antibody CH103 interaction with a quaternary epitope in the HIV-1 Env trimer.

(A) Experimentally-determined or modelled structures of the HIV-1 Env trimer (PDB ID: 4TVP) bound to anti-CD4-BS monoclonal antibodies VRC03 (PDB ID: 6MPG), 1–18 (PDB ID: 6UDJ) or N6 (modeling created by aligning the monomeric gp120 moiety from an N6-bound structure (PDB ID: 5TE6) with a single protomer of a trimeric structure (PDB ID: 4TVP)).

(B) Sequence alignment of the heavy chain variable regions of antibodies N6, VRC03, 1–18 and CH103.

(C) Structural modeling of CH103 bound to the HIV-1 Env trimer created by aligning the monomeric gp120 moiety from an experimentally-determined CH103-bound structure (PDB ID: 4JAN) with a single protomer of a trimeric structure (PDB ID: 4TVP). The inset shows a magnified view of the predicted regions of quaternary interaction with a neighboring gp120 protomer.

Two acidic FRH3 residues predicted to mediate quaternary trimer interaction were sequentially acquired during CH103 lineage evolution

Analysis of the CH103 lineage evolution ²⁰ showed that the two acidic FRH3 residues E75 and D76 were not present in the inferred CH103 UCA, but were sequentially acquired during the lineage maturation of this bNAb (Figure 2A). To get more precise insights into the evolution of these sites, we analyzed 52 paired heavy and light chains of the CH103 clonal lineage isolated by antigen-specific B-cell sorting from donor CH505, and computationally reconstructed a genealogical tree that included the inferred UCA and maturation intermediates. The E75 mutation was shown to arise in the first intermediate and to be present in 37 of 52 clonal members (71%) (Figure S1). Among all the clonal family members of CH103, E75 was found in 77% of total heavy chain sequences (derived from both NGS sequencing of the patient B-cell repertoire and isolated mAbs), which is markedly higher than the expected frequency of 4% in the absence of antigenic selection (as estimated based on simulations of AID activity with the program ARMADiLLO ^31,32), indicating a strong positive selection. To estimate the selection strength at every position and for each amino acid mutation, we systematically compared the observed frequency of mutations to their expected frequencies in the absence of selection as calculated by the program ARMADiLLO. Positions or amino acids observed at much higher mutation frequencies than expected were estimated to be under strong positive selection. We found that position 75 ranks second for estimated positive selection strength among all positions in the heavy chain, with position 14 being the first (Figure 2B). Interestingly, selection-strength analysis of amino acid mutations showed that 4 of the top 11 strongest selected mutations were to glutamic acid and were located in the heavy chain CDRH2 and FRH3 regions, i.e., E56, E70, E75 and D76, suggesting a strong selective pressure for negatively-charged amino acids in putative paratope regions (Figure 2C). The D76 mutation was also estimated to be under strong positive selection, ranking 66^th among 593 observed mutations in the heavy chain. Taken together, these results demonstrated that E75 and D76 were sequentially acquired during the lineage evolution of CH103 and that there was a strong positive selection for acquisition of acidic residues at positions 75 and 76. We hypothesized that this tendency to select for negatively-charged side chains in the loop regions may have facilitated the progressive establishment of a quaternary contact with the HIV-1 Env trimer, thereby increasing the antibody potency.

Figure 2. Acquisition of two key acidic FRH3 residues by CH103 during its lineage maturation. See also figure S1.

(A) Amino acid sequence alignment of the maturation pathway of CH103 from its inferred UCA sequence to the mature form (see Figure S1 for the complete CH103 clonal lineage tree) ²⁰.

(B) Ranking of the estimated selection pressure at different positions in the CH103 heavy chain. Position 75 is ranked 2^nd among all positions. The “0” indicates a gap in the sequence due to insertions/deletions.

(C) Amino acid-based selection strength analysis showing 4 of the top 11 strongest selected mutations to glutamates in the FRH3 and CDRH3 of CH103, suggestive of strong selective pressure for negative charges in these regions. The “0” indicates a gap in the sequence due to insertions/deletions.

Cryo-EM structures of mature CH103 and a partially UCA-reverted CH103 mutant in complex with soluble HIV-1 Env trimers

The molecular interface of CH103 with the HIV-1 Env was partially defined in a previous crystal structure of CH103 in complex with the outer domain of gp120 from a clade-C isolate, ZM176.66 ²², but this structure lacked the Env trimer context. To obtain high-resolution details of the complete Env trimer-CH103 interface and to elucidate the role of residues E75 and D76, we determined the cryo-EM structure of a stabilized soluble Env trimer, BG505 DS-SOSIP, bound to either the mature CH103 Fab or to a partially UCA-reverted CH103 mutant (KN) where residues E75 and D76 were reverted to their respective UCA sequence (i.e., E75K and D76N) (Figures 3 and S2–S4; Table S1). From the mature CH103 dataset, we identified particle populations where either one or two Fabs were bound to the Env trimer. From the CH103 KN dataset, we identified particle populations that had either no Fab, one Fab or two Fabs bound to the trimer. Similar trends for the Fab-binding stoichiometries were also observed by negative stain electron microscopy (NSEM) analysis of the BG505 SOSIP trimer bound to mature CH103 or CH103 KN, which showed two mature CH103 Fabs bound to the trimer versus only one CH103 KN Fab bound to the trimer (Figure S2E).

Figure 3. Cryo-EM structures of mature and partially UCA-reverted CH103 and their interactions with the primary CD4-BS. See also figures S2–S6 and tables S1 and S2.

(A) Cryo-EM reconstructions of mature CH103 and CH103 KN bound to the BG505 Env trimer.

(B) Interaction of mature CH103 (left) and CH103 KN (middle) with the primary protomer (CD4-BS). The inset on the right shows the previously published crystal structure of mature CH103 bound to the gp120 outer domain. The gp120 outer domain that was included in the published crystal structure is shown in dark brown. The gp120 regions included in the cryo-EM structures in this study are colored in grey.

(C) Details of the mature CH103 interactions.

(D) Details of the CH103 KN interactions.

The global resolutions of the cryo-EM maps were between 3 and 4 Å, while the local resolutions at the Env-Fab interfaces, estimated at ~3 Å, were among the highest resolved regions in the structures (Figures 3, S3, and S4). Comparing the 1-Fab versus 2-Fab bound structures, we found that in the mature CH103 Fab-bound structures, the gp120 protomer positions were unchanged between the 1-Fab and 2-Fab bound structures (Figure S5). In the CH103 KN-bound structures, we observed a shift in the secondary protomer of the 2 Fab-bound structure, when the secondary protomer was itself bound to a Fab at its primary binding site. This shift pushed the CD4-BS2 closer to the FRH3 of the first Fab, while the distance between the second Fab FRH3 and the CD4-BS2 of the third protomer remained the same as in the 1-Fab bound structure. In all the structures, the protomer-bound Fabs, when aligned by their gp120 chains, overlaid well with each other.

We examined the Env-Fab interfaces at all the mature CH103 and CH103 KN binding sites (Figures 3, 4 and S6). At the primary CD4-BS contact site, the interactive surfaces of the mature CH103 and the CH103 KN mutant were essentially identical (Figures 3B–3D), with both the antibody heavy and light chains contributing to the Env interactions. While the gp120 inner domain had not been visualized in the published structure (PDB ID: 4JAN) ²², our structures provide a complete definition of the CH103 epitope, where the antibody interactions encompass contacts with both the outer and inner domains of gp120.

Figure 4. Cryo-EM structures of the quaternary interaction of mature and partially UCA-reverted CH103 with the neighboring gp120 protomer. See also figure S6 and table S2.

(A) Zoomed-in view of quaternary interactions of the FRH3 of mature CH103 with the neighboring protomer in the 1 Fab-bound structure. The dashed lines indicate hydrogen bonds. The solid lines denote the distance between the indicated residues.

(B) Zoomed-in view of quaternary interactions of the FRH3 of CH103 KN with the neighboring protomer in the 1-Fab bound structure. The dashed lines indicate hydrogen bonds. The solid lines denote the distance between the indicated residues.

(C) Differential scanning fluorimetry (DSF) of mature CH103 (brown line) and CH103 KN (cyan line) Fabs. Three technical replicates are shown for each measurement.

(D) Binding of mature CH103 and CH103 KN to the soluble BG505 DS-SOSIP trimer, as assessed by immunoassays. The data represent the mean (± standard deviation) from two replicate wells for each dose tested. The results are from one of two independent experiments that yielded similar findings.

(E) Binding of mature CH103 and CH103 KN to monomeric gp120 from BG505, as assessed by immunoassays. The data represent the mean (± standard deviation) from two replicate wells for each dose tested. The results are from one of two independent experiments that yielded similar findings.

The interactions with the primary CD4-BS were mediated by all three CDRs in both the heavy and light chains (Figures 3B–3D). We observed similar interactions with the gp120 outer domain as were previously defined in the structure of CH103 bound to the ZM176.66 ²². In the heavy chain, the CDRH3 played a central role with residue R97 forming an extensive H-bonding network with gp120 residues D368 and E370 in the CD4-binding loop. Interactions of the CDRH2 with the gp120 inner domain residues 427–429 via H-bonding between the antibody residue H53 side chain and gp120 residue N427 were among the newly defined contacts between CH103 and the Env trimer. Light chain interactions included contacts of the CDRL2 and L3 with gp120 Loops D and V5. A new light chain interaction was identified with glycan 363, with the side chain of the CDRL3 residue F97 stacking against the glycan.

The mature CH103 heavy chain FRH3 region stacks against the CDRH1 residue M29 and bridges between two protomers by making van der Waals interactions of its residue T73 with the inner domain residues T198 and I430 of the primary protomer, while the acidic residues E75 and D76 face the neighboring protomer and facilitate quaternary contact (Figures 3C, 4A; Table S2). As predicted by our structural modeling, the side chain of mature CH103 residue E75 engages in a salt bridge with residue R308 of the adjacent gp120 protomer, as well as a hydrogen bond with residue Y318 (Figure 4A). The close approach of this region also enables additional hydrogen bonding of residue R308 with residues D72 and S74 of mature CH103 FRH3. The FRH3 residue D76 is positioned to engage in hydrogen bonding and salt bridge interactions with residue K207 of the adjacent protomer as indicated by density for the side chain of K207 in the mature CH103 structure (Figure S4). In contrast, for the CH103 KN mutant, due to the E75K substitution, residue 75 loses its quaternary interaction with both Y318 and R308, and instead forms an intra-Fab hydrogen bond with the side chain of residue D72 (Figure 4B). Similarly, the mutated residue N76 forms an intra-Fab hydrogen bond with the backbone of residue V24 (Figure 3D). There is no density for the side chain of K207 in the CH103 KN-bound structure, indicating variability in the positioning of the side chain (Figure S4). A hydrogen bond was retained between residue D72 in the CH103 KN mutant and residue R308 of the adjacent gp120 protomer (Figure 4B).

Differential scanning fluorimetry (DSF) analysis of the mature CH103 and CH103 KN Fabs demonstrated that the CH103 KN Fab is more thermostable than the mature CH103 Fab (Figure 4C), consistent with the intra-Fab K75-D72 and N76-V24 hydrogen bonds observed in CH103 KN that are absent in mature CH103 due to the K75E and N76D substitutions. Thus, the Fab stability decreased during CH103 lineage affinity maturation due to the K75E and N76D substitutions as the antibody swapped internal contacts in favor of making quaternary contacts with Env. The loss of quaternary interactions in the CH103 KN mutant leads to greater separation and overall weakening of the interactions between the bound Fab and the neighboring gp120 protomer.

The cryo-EM structures thus indicate that the mature CH103 antibody establishes stable quaternary contacts with the neighboring protomer of the HIV-1 Env. Partial reversion to the UCA-sequence via the framework mutations E75K and D76N results in disuption of these quaternary contacts. Consistent with these observations, the CH103 KN mutant exhibited diminished binding to the BG505 SOSIP trimer by immunoassays compared to the mature CH103 (Figure 4D). However, similar binding of both antibodies was observed to monomeric BG505 gp120 (Figure 4E), confirming that the KN mutations did not impact antibody binding to the primary contact site located in the outer domain of gp120.

CH103 adopts a unique binding angle compared to other quaternary-interacting anti-CD4-BS antibodies

Structural alignment of VH gene-restricted quaternary-interacting antibodies VRC03, VRC-CH31, 1–18 and 1–55, performed by aligning the gp120 protomers in their respective Env-bound structures with the gp120 protomer in our Env-bound CH103 structure, showed that CH103 engages the Env trimer with an angle of approach that is different by approximately 17° from that of other quaternary-interacting bNAbs (Figure 5A). The VH gene-restricted quaternary-interacting anti-CD4-BS bNAbs contact the CD4-BS2 either via an extended CDRH1 loop (VRC-CH31, 1–18, 1–55) or via an extended FRH3 loop (VRC03) (Figure 5A). Our results demonstrate that CH103 is able to interact with the secondary binding site despite lacking an extended CDRH1 or FRH3 loop, by virtue of its different angle of approach, which is consistent with its CDRH3-dominated mode of interaction with the primary CD4-BS ^18–25.

Figure 5. Comparison of the CH103 angle of approach with that of other quaternary-interacting anti-CD4-BS bNAbs.

(A) Angle of approach and details of quaternary interaction of VH gene-restricted bNAbs 1–18, 1–55, VRC03, VRC-CH31 and 3BNC117 to the Env trimer, compared to that of CH103. The primary CD4-BS is shown in magenta and CD4-BS2 is shown in dark blue.

(B) Models of trimer interaction of CDRH3-dominated anti-CD4-BS bNAbs. The following structures were aligned to the structure of CH103 bound to the BG505 DS-SOSIP trimer determined here: VRC13 (4YDJ), VRC16 (4YDK), HJ16 (4YE4), VRC33.01 (7LL2), 179NC75 (7LLK), and VRC01 (3NGB) shown for reference. For VRC13, VRC16 and HJ16, we aligned the gp120-bound structures with the CH103-bound BG505 trimer structure solved in this study. The CD4-BS is shown in magenta and CD4-BS2 in blue; gp120 is shown in grey and gp41 in black. The primary binding gp120 protomers were used for structural alignments.

Next, we compared the angle of approach of CH103 with that of other anti-CD4-BS bNAbs that exhibit a CDRH3-dominated mode of recognition, including VRC13, VRC16, HJ16, 179NC75 and VRC33.01. Since structural data in complex with the Env trimer are available only for 179NC75 and VRC33.01 ³³, the binding angle for the other antibodies was derived from structural alignment. The angles of approach of VRC13, HJ16, 179NC75 and VRC33.01 preclude their contact with the adjacent gp120 protomer (Figure 5B). In contrast, we observed that the angle of approach of VRC16 would allow its FRH3 to be situated in close proximity to the adjacent gp120 protomer, thus making a quaternary interaction possible. However, no structures of VRC16 in complex with an HIV-1 Env trimer has been reported and, therefore, experimental validation of this model is needed.

The quaternary interaction of CH103 is functionally relevant

To establish the functional relevance of the observed quaternary interaction of CH103 with the HIV-1 Env trimer, we tested the sensitivity of targeted Env mutants to CH103-mediated neutralization in the context of the clade-A Env BG505. Both alanine and glutamic acid substitutions were introduced, although at most positions negative charges could not be evaluated because they abrogate infectivity, as previously reported ²⁷. As controls, we tested the sensitivity of the BG505 mutants to neutralization by three VH-gene-restricted quaternary-interacting anti-CD4-BS antibodies, VRC03, 1–18, 3BNC117, as well as chimeric 3BNC117, in which the extended FRH3 loop of VRC03 was engrafted into 3BNC117 (3BNC117 FR3–03). As previously observed ²⁸, the effect of negative-charge substitutions was generally more profound than that of alanine substitutions, consistent with the acidic nature of the interacting FRH3 and CDRH1 loops in the bNAbs (Figure 6A). While minimal changes in potency were observed with 3BNC117, VRC03 and the chimeric 3BNC117 FR3–03 showed the most significant loss of neutralization potency against nearly all mutants, demonstrating an extensive, functionally-relevant quaternary interface. Neutralization by CH103 was not affected by alanine substitutions in CD4-BS2, i.e., H66A and K207A; however, mutations at the base of the V3 loop, i.e., R304A/E, R308A, and especially the double mutant R304A/R308A and Y318E, caused a marked decrease in CH103-mediated neutralization (Figure 6A). The neutralization pattern of CH103 on the BG505 mutants was similar to that of bNAb 1–18, which utilizes an extended CDRH1 to establish quaternary contact ²⁴.

Figure 6. Functional relevance of the quaternary interaction of CH103 for trimer binding and neutralization.

(A) Neutralization of the clade-A HIV-1 strain BG505 and mutants thereof by CH103, VRC03, 1–18, 3BNC117 and chimeric 3BNC117 FR3–03.

(B) Neutralization of a small global panel of HIV-1 strains by mature CH103 and CH103 KN.

(C) Neutralization of HIV-1 BG505 and mutants thereof by CH103 and CH103 KN.

The two FRH3 acidic residues E75 and D76 are critical for HIV-1 neutralization and trimer interaction

To elucidate the functional role of the two acidic residues in the FRH3 of CH103, we tested the ability of the mature CH103 antibody and the CH103 KN mutant to neutralize a small reference global panel of 12 HIV-1 Envs of different clades and geographic origins ³⁴ (Figure 6B). We also tested the clade-A isolate BG505 and clade-C isolate CH505. All the strains tested were sensitive to the mature CH103 antibody with IC₅₀ values ranging from 0.3 to 15 μg/mL. In contrast, neutralization by the KN mutant was completely lost against 4 of the 12 strains (IC₅₀ >50 μg/mL) and markedly decreased (>3-fold) against the remaining 4 strains as well as against BG505 and CH505 (Figure 6B).

Next, we tested the neutralization activity of the CH103 KN mutant against the same panel of the BG505 mutants tested in Figure 6A. Compared to the mature CH103 antibody, the potency of the KN mutant was decreased against all the mutants (Figure 6C). The most substantial reduction was seen against the unmutated Env, while the mutations in the CD4-BS2 region and the V3 base partially restored the neutralization sensitivity. As previously observed ^28,29, glutamic acid mutants, i.e., R304E, R308E and Y318E, were more effective than alanine mutants in the restoration of neutralizing capacity, with only a ~3-fold or lesser reduction compared to the mature CH103 antibody. These data suggest that in these mutants the basic residue K75 in the antibody heavy chain can facilitate the re-establishment of partial quaternary contact with the acidic glutamic acid of the mutated Env. Taken together, the above results highlight the functional relevance of the quaternary interaction between CH103 and the HIV-1 Env trimer and the importance of the two acidic FRH3 residues acquired during the lineage maturation of CH103.

DISCUSSION

To acquire broad and potent neutralizing activity, HIV-1 bNAbs undergo a long process of affinity maturation, which in many instances demands a high degree of somatic hypermutation. In the case of the VRC01 lineage, one of the most extensively characterized, somatic mutation rates reach 32% in the heavy chain variable region and 20% in the light chain variable region, which are far in excess of the average 8–10% mutation frequency seen with typical affinity-matured human antibodies ³⁵. Hypermutation of the variable regions may be required to optimize the binding interface with discontinuous epitopes, to refine the angle of engagement, to cope with the hurdle of interposed glycan chains, and/or to adapt to the continuous mutation of the HIV-1 Env ^20,23,36,37. In this report, we provide evidence of an alternative mechanism of affinity maturation in the course of the lineage evolution of a CDRH3-dominated HIV-1 bNAb directed to the CD4-BS, aimed at the acquisition of quaternary-binding capacity with a second gp120 protomer. Only a selected group of CD4-BS-specific bNAbs have developed the ability to establish a quaternary interaction with the HIV-1 Env trimer, by making direct contact with two neighboring gp120 protomers simultaneously ^24,28, thus mimicking the quaternary binding mode of CD4 ²⁷. All of the quaternary-binding antibodies described thus far belong to the VH gene-restricted group of anti-CD4-BS bNAbs. Interaction with a second protomer was shown to enhance the stability of binding and to potentiate the neutralizing activity ²⁸. To reach the secondary binding site across the inter-protomer groove, these bNAbs have typically acquired unusual structural features such as elongated CDRH1 or FRH3 loops. Unlike these bNAbs, we discovered that bNAb CH103 has developed a unique ability to contact a second gp120 protomer without unusual elongations of its heavy chain loops. As we confirmed with the cryo-EM structure, the quaternary contact of CH103 is mediated by a regular-length FRH3 loop owing to a different angle of Env engagement of this antibody onto the Env trimer, which brings the FRH3 loop in close proximity to the CD4-BS2 and the base of the V3 loop of a neighboring gp120 protomer. In common with other known quaternary-interactive bNAbs, however, CH103 displays multiple negatively-charged amino acids in its FRH3 loop, which we demonstrated to functionally interact with the positively-charged region in the distal portion of CD4-BS2 and especially at the base of the V3 loop. Thus, CH103 is an example of a quaternary-interacting CD4-BS bNAb that has not developed unusual structural features to reach a second gp120 protomer.

The maturation history of CH103 has been characterized in depth thanks to the availability of sequential samples obtained from the same infected individual (donor 505) starting from the acute phase of HIV-1 infection until the development of broad neutralization more than 3 years later ²⁰. Of note, the T/F virus began to diversify in key CD4-contact regions as early as 3 months after the initial infection and remained under sustained positive selective pressure throughout the observation period. When we revisited the evolution of the CH103 lineage in light of the discovery of the quaternary-binding ability of the mature antibody, we noticed that the FRH3 loop of the inferred UCA precursor is not negatively charged since it lacks two acidic residues at the apex of the loop, with one of them (D75) even being replaced by a positively-charged amino acid (K75). Thus, the net charge of the UCA loop was neutral. Our cryo-EM structure of the partially UCA-reverted CH103 mutant confirmed that reversion to the UCA sequence at these two positions is sufficient to abrogate the quaternary contact. In-depth analysis of the lineage evolution documented the early appearance of the E75 mutation by the first intermediate and its presence at significant frequency (71%) in mature clonal members. Moreover, we found evidence that this mutation was subjected to a strong positive selection. In contrast, lineage reconstruction showed that the second mutation, D76, arose later in clonal lineage development and was estimated to have arisen independently in two phylogenetically distinct regions of the clonal tree. The fact that D76 arose independently twice during the maturation process is indicative of a high fitness mutation. Interestingly, selection strength analysis showed that nearly 50% of the top selected mutations were to glutamic acid and occurred in the CDRH2 and FRH3 regions, suggesting a strong selective pressure for introducing negatively-charged amino acids into these two regions. Previous work has documented the beneficial effects of somatic mutations in the framework regions to enhance the breadth and potency of bNAbs by providing increased flexibility and/or direct antigen contact ²⁶. The strong positive selection for the two acidic residues that we detected in the FRH3 region of CH103 provides a new paradigm for the scope of hypermutation of framework regions in the heavy chains of bNAb.

Experiments of trimer binding and especially of neutralization using the partially UCA-reverted CH103 support a role of the quaternary interaction in stabilizing the Env-antibody interface and, thereby, in enhancing the neutralization potency of mature CH103. In particular, reversion of the two acidic residues in the FRH3 to the UCA sequence was sufficient to dramatically reduce the neutralization potency of CH103 against a panel of diverse HIV-1 strains. The increased binding affinity of the mutated intermediates likely provided the selective force to promote the survival and expansion of these variants in the germinal centers of secondary lymphoid organs during CH103 lineage evolution in patient CH505. The majority of CD4-BS-specific bNAbs do not engage a second gp120 protomer, deriving their high binding affinity exclusively from their interaction with the primary CD4-binding site in the gp120 outer domain. The significant loss of function that we documented upon abrogation of the secondary contact for CH103 implies that the interaction of this bNAb with the primary binding site is not sufficient to provide an optimized interface compared to that of other, non-quaternary-binding bNAbs. Whether the requirement for quaternary interaction for optimal neutralization activity is due to the angle of trimer engagement by CH103 or simply to a lack of further affinity maturation after the lineage reached a sufficient level of potency remains to be established. It is noteworthy that CH103 belongs to the CDRH3-dominated groups of anti-CD4-BS bNAbs and is the first antibody in its class to be shown to establish quaternary interaction. Among known CDRH3-dominated anti-CD4-BS bNAbs, we predicted by structural modeling that another antibody, VRC16, might engage the trimer with a suitable angle to interact with the neighboring gp120 protomer, although such interaction has not been experimentally validated.

In summary, the present study provides a novel example of an HIV-1 bNAb whose high potency of neutralization is mediated in part by the establishment of a quaternary contact with two distinct gp120 protomers in the Env trimer. Unlike other quaternary-interacting bNAbs, all of which belonged to the VH gene-restricted class, CH103 is CDRH3-dominated and exploits a unique angle of engagement rather than elongation of its heavy chain loops to reach a second gp120 protomer. While the present findings corroborate the possibility of engineering natural antibodies to improve their functional properties by conferring quaternary-binding capability ²⁸, the implications of the present findings for vaccine development are less clear. So far, the case of CH103 appears to be unique in its acquisition of quaternary-interactive capability during its lineage evolution. However, we cannot exclude that other quaternary-binding bNAbs may have similarly acquired such ability in the course of their lineage evolution. Likewise, it is difficult to infer how frequently bNAbs mature through the acquisition of quaternary interaction. A phase I clinical trial (HIV Vaccine Trials Network trial, HVTN 300) is in progress to evaluate the ability of the CH103 UCA-targeting, transmitted/founder Env CH505 to induce CH103-like bNAbs in humans. It will be especially interesting to monitor the frequency of somatic hypermutations induced by the vaccine in the FRH3 loops region of such antibodies. Further studies on the antibody repertoire of HIV-1-infected individuals and Env vaccinees who eventually develop bNAbs will be important to address these questions.

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Paolo Lusso (plusso@niaid.nih.gov).

Materials availability

New materials generated in this study include recombinant antibodies, HIV-1 Env trimers and plasmids for their production, which can be obtained upon request to the lead contact.

Data and code availability

• All data generated in this study are available within this paper and the supplementary files. Cryo-EM maps have been deposited in Electron Microscopy Database (EMDB) under accession numbers: EMD-44613, EMD-44614, EMD-44604, EMD-44605 and EMD-44606. The model coordinates have been deposited in the Protein Data Bank (PDB) under accession numbers: 9D7O, 9D7P, 9D7G, 9D7H and 9D7I. They are publicly available as of the date of publication. Accession codes are also listed in the key resources table.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Goat Anti-Human IgG-PE	Southern Biotech	Cat# 2040-09
IgM PerCp-Cy5.5 (Clone# G20-127)	BD Biosciences	Cat# 561285
IgD FITC (Clone# IA6-2)	BD Biosciences	Cat# 555778
CD235a PE-Cy5 (Clone# GA-R2)	BD Biosciences	Cat# 559944
CD10 PE-CF594 (Clone# HI10A)	BD Biosciences	Cat# 562396
CD3 PE-Cy5 (Clone# HIT3a)	BD Biosciences	Cat# 555341
CD27 PE-Cy7 (Clone# O323)	eBioscience	Cat# 25-0279
CD38 APC-AF700 (Clone# LS198-4-2)	Beckman Coulter	Cat# B23489
CD19 APC-Cy7 (Clone# LJ25C1)	BD Biosciences	Cat# 561743
CD14 BV605 (Clone# M5E2)	Biolegend	Cat# 301834
CD16 BV570 (Clone# 3G8)	Biolegend	Cat# 302035
HRP-conjugated goat anti-human IgG	Jackson ImmunoResearch	Cat# 109-035-008
Monoclonal anti-HIV-1 gp120 CH103	This study	N/A
Monoclonal anti-HIV-1 gp120 CH103 KN	This study	N/A
Monoclonal anti-HIV-1 gp120 VRC03	This study	N/A
Monoclonal anti-HIV-1 gp120 1-18	This study	N/A
Monoclonal anti-HIV-1 gp120 3BNC117	This study	N/A
Monoclonal anti-HIV-1 gp120 3BNC117 FR3-03	This study	N/A
Biotin-conjugated anti-HIV-1 gp120 447-52D	This study	N/A
Biotin-conjugated anti-HIV-1 gp120 17b	This study	N/A
Biotin-conjugated anti-HIV-1 gp120 48d	This study	N/A
Bacterial and Virus Strains
NEB® 5-alpha Competent E. coli (High Efficiency)	New England BioLabs	Cat# C2987H
Chemicals, Peptides, and Recombinant Proteins
Human soluble CD4-183 (sCD4-183)	NIH AIDS Reagent Program	Cat# 7356
CD4-M48U1 mini-protein	Loic Martin, CEA	N/A
Lectin from Galanthus nivalis	Sigma-Aldrich	Cat# L8275-5MG
BG505.T332N.SOSIP.664 trimer	This study	N/A
JR-FL SOSIP.664 trimer	This study	N/A
CH505 SOSIP.664 trimer	Kevin O. Saunders, Duke	N/A
BG505 DS-SOSIP trimer	Priyamvada Acharya, Duke	N/A
BG505 gp120 monomer	This study	N/A
BaL gp120 monomer	This study	N/A
Critical Commercial Assays
Luciferase Assay System	Promega	Cat# E4550
QuickChange II Site-Directed Mutagenesis Kit	Agilent	Cat# 200524
Plasmid Mega Kit (25)	Qiagen	Cat# 12181
TransIT®-293 Transfection Reagent	Mirus Bio	Cat# MIR 2705
FreeStyle™ MAX Reagent	Gibco	Cat# 16447500
Effectene transfection reagent	Qiagen	Cat# 1054250
ExpiFectamine	Thermo Fisher Scientific	Cat# A14525
EasySepTM Human B cell enrichment Kit	StemCell	Cat# 19054
Archer Immunoverse BCR IGH/K/L kit	IDT	Cat# DB0221
RNeasy Plus Mini Kit	Qiagen	Cat# 74134
StepOne Real-Time PCR system	Thermo Fisher Scientific	Cat# 4376357
KAPA Library Quantification Kit	Roche	Cat# KK4824
MiSeq Reagent Kit v3	Illumina	Cat# 20075295
Deposited Data
Cryo-EM structure of BG505 DS-SOSIP.664 with 1 CH103 Fab bound	This study	EMD-46613; PDB: 9D7O
Cryo-EM structure of BG505 DS-SOSIP.664 with 2 CH103 Fabs bound	This study	EMD-46614; PDB: 9D7P
BG505 DS-SOSIP.664 apo structure from the CH103 KN cryo-EM dataset	This study	EMD-46604; PDB: 9D7G
Cryo-EM structure of BG505 DS-SOSIP.664 with 1 CH103 KN Fab bound	This study	EMD-46605; PDB: 9D7H
Cryo-EM structure of BG505 DS-SOSIP.664 with 2 CH103 KN Fabs bound	This study	EMD-46606; PDB: 9D7I
Experimental Models: Cell Lines
Expi293F cells	ThermoFisher Scientific Inc	Cat# A14527
FreeStyle 293-F cells	ThermoFisher Scientific Inc	Cat# R79007
HEK 293T cells	ATCC	Cat# CRL-3216
TZM-bl cells	NIH AIDS Reagent Program	Cat# 8129
Recombinant DNA
pVRC8400 BG505 SOSIP.664	Peter D. Kwong, NIH	N/A
pVRC8400 JR-FL SOSIP.664	Peter D. Kwong, NIH	N/A
pVRC8400 DS-SOSIP.664	Priyamvada Acharya, Duke	N/A
pCH103 HC in CMVR expression vector	Kevin O. Saunders, Duke	N/A
pCH103 LC in CMVR expression vector	Kevin O. Saunders, Duke	N/A
pCH103 KN HC in CMVR expression vector	This study	N/A
CH505 env expression vector	Nicole Doria-Rose, NIH	N/A
Plasmids pSG3Δenv	NIH AIDS Reagent Program	Cat# 11051
Panel of Global HIV-1 Env Clones	NIH AIDS Reagent Program	HRP-12670
Oligonucleotides
gctgacagactaacagactg (forward primer for sequencing the variable region of CH103 heavy chain)	This study	N/A
cagacaatgcgatgcaatttcctc (reverse primer for sequencing the variable region of CH103 heavy chain)	This study	N/A
atcagtggacacgtccaagaaccagttctccctgag (forward primer for inserting the KN mutation in the CH103 heavy chain)	This study	N/A
ctcagggagaactggttcttggacgtgtccactgat (reverse primer for inserting the KN mutation in the CH103 heavy chain)	This study	N/A
Software and Algorithms
FlowJo Version 10.8.1	FlowJo, LLC	N/A
QuikChange Primer Design	Agilent	www.agilent.com/store/primer DesignProgram.jsp
Pymol Version 2.5.5	Schrödinger, LLC	www.pymol.org/
GraphPad Prism 9 Version 9.5.0	GraphPad Software Inc.	www.graphpad.com/scientific-software/prism/
Clonalyst Program	N/A	www.bu.edu/computationalimmunology/research/software/
ARMADiLLO	Beem et al. 32	www.github.com/WieheLab/AR MADiLLO
ChimeraX Version 1.8	Meng et al. 38	www.rbvi.ucsf.edu/chimerax/
cryoSPARCv4.01 Version 4.01	Punjani et al. 39	www.cryosparc.com
Coot Version 0.9.8.7	Emsley et al. 40	www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
Phenix Version 1.20.1	Liebschner et al. 41	www.phenix-online.org/

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

The cell lines used in this work were Expi293F, FreeStyle 293-F, HEK 293T and TZM-bl (see details in the Key Resources Table). HEK293T and TZM-bl cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM; Gibco) supplemented with 10% heat-inactivated fetal bovine serum (FBS; Gibco), 1% Penicillin-Streptomycin (Gibco), 1X GlutaMAX^™ Supplement (Gibco), 1 mM Sodium Pyruvate (Gibco), 1X MEM Non-Essential Amino Acids (NEAA) (Gibco) and 24 mM HEPES (Gibco). HEK293FS (ThermoFisher) cells were cultured in FreeStyle^™ 293 Expression Medium (Gibco). Neither sex nor gender had any influence on the results of this study. All work related to the human subject from whom CH103 was isolated was previously reported ²⁰, and was performed in compliance with Institutional Review Board protocols approved by the Duke University Health System Institutional Review Board.

METHOD DETAILS

Structural modeling

As the previously published structure of CH103 bound to the ZM176.6 gp120 outer domain (PDF: 4JAN) lacked the Env trimer context, we overlayed this CH103-bound structure on the structure of a soluble Env trimer, BG505.SOSIP.664, bound to antibodies PGT122 and 35O22 (PDB ID: 4TVP), using the gp120 outer domains for structural alignment. Similarly, we modeling the interation of CD4-BS antibody N6 in the Env trimer context by overlaying the gp120 from a previously described N6-bound structure (PDB: 5TE6) on the 4TVP structure.

Antibody isolation from peripheral blood memory B cells

Envelope-specific memory B cells were isolated by fluorescence activated cell sorting as described previously ⁴². Cryoperserved peripheral blood mononuclear cells (PBMCs) for B cell isolation were collected at visit 113 and 201 from the person living with HIV-1 called CH505. PBMCs were stained with IgM PerCp-Cy5.5 (Clone# G20–127, BD Biosciences, Catalog# 561285), IgD FITC (Clone# IA6–2, BD Biosciences, Catalog# 555778), CD235a PE-Cy5 (Clone# GA-R2, BD Biosciences, Catalog# 559944), CD10 PE-CF594 (Clone# HI10A, BD Biosciences, Catalog# 562396), CD3 PE-Cy5 (Clone# HIT3a, BD Biosciences, Catalog# 555341), CD27 PE-Cy7 (Clone# O323, eBioscience, Catalog# 25–0279), CD38 APC-AF700 (Clone# LS198–4-2, Beckman Coulter, Catalog# B23489), CD19 APC-Cy7 (Clone# LJ25C1, BD Biosciences, Catalog# 561743), CD14 BV605 (Clone# M5E2, Biolegend, Catalog# 301834), CD16 BV570 (Clone# 3G8, Biolegend, Catalog# 302035), and biotinylated CH505 transmitted/founder chimeric stabilized SOSIP trimer (TF chTrimer) conjugated to either streptavidin-BV421 or streptavidin-AF647. The cells were then labeled with Fixable Aqua Live/Dead Cell Stain Kit (Invitrogen, Catalog# L34957). On a BD FACSAria II flow cytometer (BD Biosciences), Env-specific memory B cells were gated as viable CD3⁻/CD14⁻/CD16⁻/CD235a⁻/CD19⁺/IgD⁻/ Env-BV41⁺/Env-AF647⁺ cells. Single Env-specific memory B cells were sorted into 96-well plates containing lysis buffer. Plates containing cells were immediately frozen in a dry ice/ethanol bath, and stored at −80 °C. Flow cytometric data were analyzed using FlowJo version 10. Antibody genes were amplified by RT-PCR from flow cytometry-sorted single B cells using the methods described previously ^43,44. The PCR primer sequences have been reported elsewhere ⁴². The PCR-amplified genes were purified and sequenced with 10 μM forward and reverse primers by Azenta Life Sciences. Contigs were built from each forward and reverse sequence and used for immunogenetic annotation. Immunogenetics and clonal relatedness of antibody sequences was determined using the human library in Clonalyst ^20,45. Purified PCR amplicon were used to construct linear expression cassettes using overlapping PCR as previously described ⁴³. The linear cassettes for heavy ahd light chains were co-transfected into 293T cells using effectene transfection reagent (Qiagen, Catalog# 1054250). Three days after transfection, cell culture supernatant was harvested, concentrated, and filtered. The cell culture supernatants containing recombinant IgGs were analyzed for binding to HIV-1 Env by ELISA. Antibodies of interest were selected based on immunogenetic and reactivity. Plasmids encoding the heavy and light chain genes of the antibodies were synthesized by GenScript and used for antibody production in Expi293F cells transfected with ExpiFectamine (Thermo Fisher Scientific, Catalog# A14525).

B-cell receptor repertoire sequencing

Isolated B cells using EasySepTM Human B cell enrichment Kit (StemCell, Cambridge, MA; Catalog #19054) with negative selection from CH505 PBMC sample were used to prepare NGS gene libraries using the Archer Immunoverse BCR IGH/K/L kit (IDT, Boulder, CO; Catalog #DB0221). In brief, total RNA was extracted from 220,000 B cells using RNeasy Plus Mini Kit (QIAGEN, Germantown, MD; Catalog #74134) and RNA was eluted in 30 μl nuclease-free water, 100ng total RNA was used for cDNA synthesis, BCR amplification, and generation of sequence-ready libraries following the manufacturer’s protocol. The quantities of BCR gene libraries for sequencing was assessed on StepOne Real-Time PCR system (ThermoFisher, Catalog #4376357) with KAPA Library Quantification Kit (Roche, Indianapolis, IN; Catalog #KK4824). BCR libraries were sequenced using the MiSeq Reagent Kit v3 (600-cycles) (Illumina, Catalog #20075295) to generate 2,301 paired end reads following the manufacturer’s protocol (Illumina, Document #15027617 v05) with loading concentration of libraries at 15pM plus 30% PhiX (Illumina, Catalog #FC-110–3001), generate around 32 million pair-end reads for heavy and light chains. NGS sequences were combined with the 52 CH103 member sequences recovered by antigen-specific B cell sorting and were then grouped into clones using Cloanalyst ^45,46 (https://www.bu.edu/computationalimmunology/research/software/). A total of 1,225 heavy chain sequences from bulk NGS repertoire sequencing were determined to be clonally-related to CH103.

Antibody sequence analysis

Reconstruction of the 52-member CH103 clonal genealogy was performed using the Cloanalyst software package using paired heavy and light chain sequences from previously isolated CH103 clonal members ²⁰ and newly isolated heavy and light chain members from this study as input. Selection strength estimates for mutations were calculated both at the position level (strength of selection at a specific amino acid position in the CH103 heavy chain) and amino acid level (strength of selection of a specific mutation at a specific position). To estimate selection strength, we compared observed mutation frequencies in recovered CH103 clonal member sequences to the expected frequencies of those mutations in the absence of selection ³¹ following the principle that positively selected mutations should occur at higher frequencies than expected based on intrinsic AID activity alone. Observed mutation frequencies were computed from the total CH103 clonal member sequences recovered both by bulk high-throughput B cell receptor repertoire sequencing and by antigen-specific single B cell sorting. Expected mutation frequencies were calculated using the ARMADiLLO program which simulates somatic hypermutation in the absence of antigenic selection and accounts for targeting and substitution biases in the somatic hypermutation process. First, we used ARMADiLLO to generate simulated mutated sequences that matched the distribution of the number of heavy chain nucleotide mutations found in the 1277 sequences in the clone. We generated 100 simulated mutated sequences for each observed member of the clone. We then computed the frequencies of all amino acid mutations in the simulated sequences set to estimate the expected frequency of any amino acid mutation in the absence of selection. For position level selection strength estimates, we compared the observed mutation frequency at each position to the expected frequency of a mutation to any of the 19 amino acids at that position as estimated from the simulations. For amino acid level selection strength estimates, we compared the observed frequency of each of the 19 possible mutated amino acids at each position to the expected frequency of that specific amino acid at that specific position as estimated from the simulations. To quantify selection strength, we calculated a chi-squared statistic using the observed and expected frequencies and ranked positions and mutations by their chi-squared statistics.

Cryo-EM structure determination

Cryo-EM data were collected on a FEI Titan Krios with an acceleration voltage of 300 kV and a magnification of 81000x. The data were processed in cryoSPARCv4.01 ³⁹. Movies were aligned using Patch Motion Correction and the non-dose weighted aligned micrographs were used for the CTF correction with PatchCTF Estimation. In the first round of particle selection, particles were picked using either the AI picker Topaz or the Blob picker with a minimum particle diameter of 150 Å and a maximum diameter of 350 Å. These particles were extracted with a diameter of 320 pixels and subjected to 2D classification. Non-representative classes were discarded, and the remaining particles underwent heterogeneous refinement. Non-garbage particles were used to generate consensus structures of mature CH103 + BG505 DS SOSIP and CH103 KN + BG505 DS SOSIP. Non-uniform refinement was used to obtain final electron density maps. To obtain structures of multiple states of the Fab binding, we performed further heterogeneous refinement on the particles used for the consensus structures. We were able to separate out 1-Fab bound and 2-Fab bound classes for mature CH103 + BG505 DS SOSIP and no Fab bound, 1-Fab bound and 2-Fab bound for CH103 KN + BG505 DS SOSIP. Non-uniform refinement was again used on the final groups of particles to obtain the final structures from heterogeneous refinement. ChimeraX ³⁸ was used to fit a previously solved structure of BG505 DS-SOSIP (PDB ID:4ZMJ) as well as a previously solved structure of CH103 Fab (PDB ID: 4JAM) into each of the densities. For the CH103 KN structures, the residues at positions 75 and 76 of the Fab were mutated to K75 and N76 in Coot ⁴⁰. A combination of real space refinement in Phenix ⁴¹ and manual refinment in Isolde and Coot were used to refine the model. Cryo-EM data collection and refinement statistics are summarized in Table S1. The contacts between CH103 WT and KN Fab HC residues 75 and 76 and the CD4-BS2 on the adjacent protomer are summarized in the contact map generated by the MAPIYA server ⁴⁷ in Table S2.

Protein mutagenesis, expression and purification

Site-directed mutagenesis of HIV-1 Env and antibody-expressing genes was carried out with QuikChange II site-directed mutagenesis kit (Agilent Technologies; 200524). Soluble HIV-1 Env trimers were expressed in HEK293FS (ThermoFisher) by cotransfecting a Env-expressing plasmid and a furin-expressing plasmid. Cells were cultured for 6 days, and the supernatant was harvested and passed through a 0.22 μm filter. A Galanthus nivalis lectin column (Vector laboratories) was used to purify the proteins. The filtered supernatants were loaded onto the lectin column, followed by washing with PBS. The bound proteins were eluted with 1 M methyl α-D-mannopyranoside in PBS. The eluted samples were dialyzed to PBS, followed by two rounds of size-exclusion chromatography in PBS. The trimer peak was collected and loaded onto a mAb 447–52D affinity column. The flow-through was collected and concentrated to 1~5 mg/mL and stored in PBS supplemented with 10% glycerol at −80 °C.

Antibody expression and purification

The amino acid sequence of the heavy and light chain variable regions of antibody CH103 and its KN mutant are reported in table S3. Antibodies were expressed in HEK293FS cells by cotransfecting equal amount of heavy chain- and light chain-expressing plasmids. After 5 days incubation at 37 °C, the supernatant was harvested and passed through a 0.22 μm filter. Proteins were loaded onto a protein A column, washed with PBS, and eluted with IgG elution buffer (Thermo Fisher Scientific; 21009). The samples were neutralized with 0.1 M Tris-HCl pH 9.0, followed by dialyzed into PBS.

Enzyme immunoassays

Binding of bNAbs to HIV-1 Env was tested by ELISA. If not specified, 1 h incubation at RT was carried out after each reagent was added to the wells. Plate wash was performed three times after each incubation with 1X wash buffer (R&D Systems). 4 μg/mL of Galanthus nivalis lectin (Sigma; L8275–5MG) was coated onto 96-well ELISA plates (Corning) at 4 °C overnight. Plates were blocked with 0.2% casein in PBS, followed by addition of HIV-1 Env trimers at 1 μg/mL or monomers at 2 μg/mL. Serially diluted antibodies were added to duplicate wells followed by incubation with HRP-conjugated goat anti-human IgG (Jackson ImmunoResearch; 109–035-008). Binding signal was revealed using substrate reagent and stop solution (R&D Systems). Light absorption was measured at a wavelength of 450 nm on a luminometer (PerkinElmer).

HIV-1 pseudovirus neutralization

The neutralizing activity of the antibodies was evaluated by TZM-bl assays. HIV-1 pseudoviruses were produced by co-transfecting 4 μg of a backbone plasmid pSG3^ΔENV and 2 μg of Env-expressing plasmids into HEK293T cells using TransIT^®-293 Transfection Reagent (Mirus). Serially diluted HIV-1 antibodies were incubated with pseudoviruses for 30 min at RT, followed by adding 10,000 TZM-bl cells to the duplicate wells. After 48 h incubation at 37 °C, the luminescence signal was revealed by a luciferase assay kit (Promega). IC₅₀ and IC₈₀ values were obtained by calculating the antibody concentration that yielded 50% or 80% neutralization with Prism 9 software.

QUANTIFICATION AND STATISTICAL ANALYSIS

Cryo-EM data collection and refinement statistics are summarized in Table S1. Pseudovirus neutralization titers (IC₅₀ and IC₈₀) were obtained by calculating the antibody concentration that yielded 50% or 80% neutralization using Inhibitor vs. Response – Variable slope (four parameters) on the Prism 9 software. Selection strength estimates for mutations were calculated by comparing observed mutation frequencies in recovered CH103 clonal member sequences to the expected frequencies of those mutations in the absence of selection. Expected mutation frequencies were calculated using the ARMADiLLO program.

Supplementary Material

SUPPLEMENTAL INFORMATION

Supplemental information includes 6 figures and 2 tables, and can be found with this article online at https://doi.org/...

Highlights.

• Antibody CH103 makes quaternary contact with the HIV-1 envelope

• The quaternary contact was progressively acquired during lineage maturation of CH103

• Two acidic FHR3 residues are critical for this functionally relevant quaternary contact

• CH103 can reach two gp120 protomers owing to its peculiar angle of engagement