Near-Pan-neutralizing, Plasma Deconvoluted Antibody N49P6 Mimics Host Receptor CD4 in Its Quaternary Interactions with the HIV-1 Envelope Trimer

ABSTRACT The first step in HIV-1 entry is the attachment of the envelope (Env) trimer to target cell CD4. As such, the CD4-binding site (CD4bs) remains one of the few universally accessible sites for antibodies (Abs). We recently described a method of isolating Abs directly from the circulating plasma and described a panel of broadly neutralizing Abs (bnAbs) from an HIV-1 “elite neutralizer” referred to as patient N49 (N49 Ab lineage [M. M. Sajadi, A. Dashti, Z. R. Tehrani, W. D. Tolbert, et al., Cell 173:1783–1795.e14, 2018, https://doi.org/10.1016/j.cell.2018.03.061]). Here, we describe the molecular details of antigen recognition by N49P6, an Ab of the N49 lineage that recapitulates most of the neutralization breadth and potency of the donor’s plasma IgG. Our studies done in the context of monomeric and trimeric antigens indicate that N49P6 combines many characteristics of known CD4bs-specific bnAbs with features that are unique to the N49 Ab lineage to achieve its remarkable neutralization breadth. These include the omission of the CD4 Phe43 cavity and dependence instead on interactions with highly conserved gp120 inner domain layer 3. Interestingly, when bound to BG505 SOSIP, N49P6 closely mimics the initial contact of host receptor CD4 to the adjacent promoter of the HIV-1 Env trimer to lock the trimer in the closed conformation. Altogether, N49P6 defines a new class of near-pan-neutralizing, plasma deconvoluted CD4bs Abs that we refer to as the N49P series. The details of the mechanisms of action of this new Ab class pave the way for the next generation of HIV-1 bnAbs that can be used as vaccine components of therapeutics.

B roadly neutralization antibodies (bnAbs) capable of neutralizing diverse circulating HIV-1 strains are considered the key for a successful vaccine or passive prophylaxis against HIV-1. High mutation rates, polymorphisms, altered glycosylation patterns forming a protective "glycan shield," and conformational heterogeneity of the envelope glycoprotein (Env) trimer all contribute to HIV-1 variability both within a patient and, to a greater extent, within a population driving viral escape from the immune system (1). Despite these obstacles, potent bnAbs have been isolated that target various distinct vulnerable epitopes within Env, including the CD4-binding site (CD4bs) (2), the V1/V2 loop, the V3 glycan patch (3), the membrane-proximal external region (MPER) (4), and the gp120-gp41 interface (5). Of particular interest is the CD4bs because the infectivity of HIV-1 largely relies on the successful engagement of Env with the host CD4 receptor. Due to this critical functional constraint, conservation of the CD4bs remains relatively high among circulating viruses, rendering them susceptible to neutralizing antibodies. Indeed, members of a group of antibodies targeting CD4bs exhibit both cross-clade neutralization and good potency. Many of them share similar characteristics in heavy chain gene usage and gp120 recognition mode and thus have been categorized as "VRC01 class" antibodies, after the first isolated member (6). The bnAbs within this class typically take longer to develop in natural infection than other bnAb specificities and contain, on average, a high degree of somatic hypermutation (SHM) and a short 5-residue CDR L3 (complementarity-determining region light chain 3) (7,8). The emergence of single B cell sorting and monoclonal antibody (mAb) isolation enabled the discovery of many potent VRC01 class bnAbs from HIV-infected donors. Of note, some of the most potent CD4bs bnAbs isolated from B cells included N6, 1-18, 3BNC117, and VRC07 (9)(10)(11)(12). However, memory B cell repertoires have been shown to differ from those from circulating plasma (13,14), and there is often a discordance between memory B cell pools and anti-Env circulating antibody responses, which shows that the HIV neutralization profiles of memory B cell-derived mAbs do not always match those found in plasma (15)(16)(17).
Recently, by using proteomic and genomic analyses, we described a panel of nearpan-neutralizing antibodies, including N49P7, from the plasma of an HIV-1 "elite neutralizer," donor N49. In addition to being an elite neutralizer, donor N49 belonged to a cohort of viremic controllers within a natural viral suppressor (NVS) cohort of HIV subtype B-infected donors (18,19). N49P7 was distinct from other bnAbs in the VRC01 class as it made substantially more contact with conserved residues within the gp120 inner domain, contributing to its near-pan-neutralizing ability, with a median 50% inhibitory concentration (IC 50 ) of 0.10 mg/ml (20). Also isolated from donor N49, mAb N49P6 exhibits slightly less potency with a similar remarkable neutralization breadth. A unique feature of both bnAbs N49P6 and N49P7 is that they completely recapitulate the neutralization breadth of the donor N49 anti-Env polyclonal plasma IgG, exhibiting broad and high potency with the ability to neutralize all 117 pseudoviruses in an HIV "global panel." In this study, we characterize the unique mechanisms of the neutralizing activity of bnAb N49P6 by describing the molecular details of N49P6's interaction with monomeric and trimeric Env antigens. The antigen-binding features of N49P6 are discussed in the broad context of antigen recognition of known CD4bs-specific bnAbs, including the VRC01-like class bnAbs, to describe the unique characteristics of Abs of the N49 lineage that allow them to achieve high potency and near-pan-neutralization breadth.
N49P6 recognizes highly conserved residues within gp120 93TH057 core e . In order to better understand the features of N49P6 that give it its extraordinary breadth and potency, we determined the crystal structure of N49P6 Fab-gp120 93TH057 core e to a 2.55-Å resolution ( Table 1). As shown in Fig. 2, N49P6 primarily binds through its V H 1-2 heavy chain by mimicking many of the gp120-binding characteristics of CD4. Heavy chain CDR H2 and CDR H3 contribute the most to binding, representing 39% and 27% of the total buried surface area (BSA) of the Fab, respectively (see Table S1 in the supplemental material). CDR H1, in contrast, contributes only 5 Å 2 (0.7%) to the heavy chain BSA. Of these two major anchoring elements, CDR H2 contributes the most to the BSA of the interface (373 Å 2 of the total 733-Å 2 heavy chain BSA), which includes a short antiparallel b-sheet (N49P6 residues 55 to 57) that interacts with residues in the gp120 CD4-binding loop (gp120 residues 365 to 368) (Fig. 2B). These contacts are "chaperoned" by a salt bridge between Asp 368 of gp120 and a heavy chain framework region 3 (FWR3) residue of N49P6, Arg 71 , found in many VRC01-like bnAbs (Fig. 2C). In addition, CDR H2 also interacts with gp120 loop D (through a hydrogen [H] bond formed between Trp 50 [CDR H2] and Asn 280 [loop D]) and gp120 loop V5 (through contacts with gp120 residues 455 to 459). Interestingly, although N49P6 mimics many of CD4's binding characteristics, Gly 54 in CDR H2 of N49P6 bypasses the Phe 43 (CD4)-binding cavity, a hydrophobic pocket used by CD4 and many CD4 mimetics to bind to Env (Fig. 2B). N49P6 shares this feature with N49P7, which indicates that this Ab lineage developed an alternate strategy to compensate for the loss of binding energy by using Gly instead of an amino acid with a bulky hydrophobic side chain at this position (20); some, but not all, VRC01-like Abs use the Phe 43 cavity for binding Env, although this may decrease the breadth in the presence of cavity-filling mutations such as Ile 375 in the N6-resistant pseudovirus CT565_C7_48. An important anchoring element that may FIG 2 Crystal structure of the N49P6 Fab-gp120 93TH057 core e complex. (A) Overall structure of the complex shown as a ribbon diagram. The complementarity-determining regions (CDRs) of N49P6 Fab are colored as follows: CDR H1 is black, CDR H2 is yellow, CDR H3 is red, CDR L1 is dark green, and CDR L3 is blue. The outer and inner domains of gp120 are dark and light gray, respectively. The outer domain D and E, CD4 binding, and V5 loops are shown in cyan, purple-blue, magenta, and green, respectively. Carbohydrates at positions N 276 (loop D) and N 355 (loop E) are shown as sticks. (B) N49P6 Fab-gp120 93TH057 core e interface with coloring as described above for panel A. (Top) N49P6 Fab is shown as a molecular surface, and the gp120 contact residues are shown as sticks. (Bottom) A 180°view reveals the detailed interaction of N49P6 Fab and the gp120 surface. Contact residues of N49P6 Fab are shown as sticks, and all Fab contact residues are listed. gp120 is shown as a molecular surface and colored according to its electrostatic potential, with red, blue, and white representing negative, positive, and neutral electrostatic potentials, respectively. (C) Blowup view of the hydrogen bond network of the interaction between N49P6 Fab and the gp120 surface. Three hydrogen bonds (CDR H2 W 50 and loop D N 280 , CDR H3 W 100J and loop D N 279 , and CDR L3 E 96 and loop D N 280 ) and a salt bridge (heavy chain residue R 71 and CD4-binding loop D 368 ) are formed at the interface. Residues contributing to the interaction are shown as sticks, and the hydrogen bond network is shown with red dashed lines.
Structural Basis of Neutralizing Activity of N49P6 ® compensate for this loss in N49P6 is the 20-amino-acid (aa)-long CDR H3 ( Fig. 2B and C). N49P6's CDR H3 contacts gp120 loop D (gp120 residues 275 to 276 and 279 to 283), establishing a Trp 100J (CDR H3)-Asp 279 (loop D) hydrogen bond, and contacts the conserved gp120 inner domain layer 3 (gp120 residues 474 and 476). As we discuss below, the inner domain contacts are unique to the N49 lineage.
The contribution of the N49P6 lambda light chain to the gp120 core complex interface is minimal, representing only 25% of the total Fab BSA (18% if excluding the contribution from the Asn 278 glycan). N49P6's light chain contains both deletions in its 8-amino-acidlong CDR L1 and a shortened CDR L3 to minimize light chain-mediated viral escape. The truncated CDR L1 contacts only the N-acetylglucosamine linked to Asn 276 on gp120 loop D. The short CDR L3 also contacts the Asn 276 glycan as well as neighboring residues in loop D (gp120 residues 276 and 278 to 280) and loop V5 (gp120 residues 458 to 459), with CDR L3 Phe 91 making van der Waals contacts with gp120 Asn 279 and Thr 278 and CDR L3 Glu 96 forming a hydrogen bond with the gp120 Gly 459 main chain. Steric clashes between the light chain and the Asn 276 glycan are a major limiting factor in breadth for CD4bs bnAbs, with most having deletions in CDR L1 in addition to a 5-residue CDR L3; N49P6 has one of the shortest CDR L1s, 1 amino acid shorter than VRC01 and 3 amino acids shorter than N6 (Fig. S1). In addition, N49P6 and N49P7 have a Cys at framework residue 36 of the light chain, normally Tyr, which allows the light chain to rotate further away from gp120 and better accommodate the Asn 276 glycan (20).
N49P6 closely resembles N49P7 in binding to gp120 93TH057 core e . N49P6 is clonally related to N49P7, another antibody from the same donor. N49P6 and N49P7 share similar breadths, with N49P7 being slightly more potent (Fig. 1). The available high-resolution crystal structure of the complex of the N49P7 Fab with gp120core e of the same clade A/E 93TH057 strain determined by us previously (20) allowed us to compare their mechanisms of attachment and determine what contributes to the breadth and potency of this antibody class. Figure 3 shows a structural comparison of both complexes and a detailed analysis of specific antibody-antigen contacts. Superimposition of the N49P6 and N49P7 complexes based upon gp120 93TH057 core e (Fig. 3A) revealed no significant structural differences (C a carbon root mean square deviations of 1.53 Å for the complex and 0.89 Å for the complex minus the constant part of the Fab), although N49P6 has a 1-amino-acid-longer CDR H3 than N49P7 (Arg 100C represents an insertion in N49P6 relative to N49P7). In the N49P6 gp120 complex, Arg 100C points away from gp120 and does not contribute to binding. However, we noticed a slight difference in the relative orientations of V L and V H in the N49P6 structure, with V L tilting further away from loop D, which results in a slightly smaller light chain footprint on gp120, as shown in Fig. 3B. Both N49P6 and N49P7 recognize a mixed inner domain/CD4-binding-site epitope referred to as the iCD4bs epitope, which is composed of highly conserved regions in gp120 (Fig. 3D). As shown in other studies, the gp120 inner domain, consisting of three mobile layers and a 7-or 8-stranded b-sandwich, contributes to Env integrity as well as to conformational transitions during viral fusion and therefore harbors some of its most highly conserved residues (21,22). Specifically, CDR H2 and CDR H3 of N49P6 interact with residues 97, 102, and 124 of layer 2 and residues 473, 474, 476, and 480 of layer 3.
Interestingly, structural analyses indicate that the slightly lower neutralization potency of N49P6 than of N49P7 can be attributed to a few sequence changes in CDR H2 and CDR H3 (lower BSA due to Ser 53 -versus-Met, Gly 56 -versus-Gln, and Thr 57 -versus-Val differences). In addition, the insertion of an extra Arg in CDR H3 of N49P6 (Arg 100C ) relative to N49P7 results in a slightly lower BSA for Glu 100F and Val 100H , even though their side chain positions are largely superimposable ( Fig. 3C and D). The only other significant difference in BSAs between the two epitopes resides on the light chain, with N49P7 having a slightly higher BSA in its CDR L3 due to a Phe 91 -versus-Tyr difference in their sequences.
N49P6 and N49P7 are unique among VRC01-like CD4bs Abs in how they extend their gp120 footprint and increase their neutralization breadth. The CD4-binding site is recessed in the HIV-1 trimer and provides only limited access to binding partners. Only a few classes of Abs that target this site can meet these requirements and effectively neutralize HIV-1, i.e., those derived from the heavy chain germ line allele V H 1-2 (VRC01 class) or V H 1-46 (8ANC131 class) and those that predominately bind with their CDR H3 (6). To better understand how N49P6 and also N49P7 fit into these categories, we compared their structures to those of bnAbs of this class whose gp120core structures are available, specifically, VRC01, VRC03, VRC07, NIH45-46, and N6 ( Fig. 4 and Fig. S1). As can be expected due to their shared germ line V H 1-2 allele, there is a high degree of similarity in how CDR H2 of each of these Abs binds gp120. The Thr 57 -to-Val difference between N49P6 and N49P7 that results in an approximately 12-Å 2 increase in the BSA for N49P7 seems common for all other members examined here (Fig. S1); Thr 57 in N49P6 forms a weak hydrogen bond with gp120 Gly 366 and may not pack as well against Ser 365 as does Val 57 in N49P7, which forms van der Waals contacts with both residues. Similarly, the choice of a larger hydrophobic residue than serine at position 53 seems the norm, with Met for N49P7 and Leu or Arg for all others except N6 with Gln.
A common strategy used by VRC01-like Abs to increase binding energy is the use of the CD4 Phe 43 -binding cavity. Both VRC03 and N6 do this by having a Trp or Tyr at heavy chain position 54, respectively. N49P6 and N49P7 but also VRC01, VRC07, and NIH45-46 have a FIG 3 Structural comparisons of N49P6 Fab-and N49P7-gp120 93TH057 core e complexes. (A) Blowup view into the Fab-gp120core e interface. The molecular surface is displayed over gp120 93TH057 core e , and the CDRs of N49P6 are shown as sticks, with colors as described in the Fig. 2 legend. The heavy chain and light chain of N49P7 are shown in dark and light blue, respectively. All structures were superimposed based on gp120 93TH057 core e . The N49P7 Fab-gp120 93TH057 core e structure is from PDB accession number 6BCK. (B) Binding footprints for N49P6 and N49P7 on gp120 93TH057 core e are outlined in yellow. The gp120 surface is colored as described in the Fig. 2 legend. (C) Contact residues of CDRs of N49P6 and N49P7 with gp120 93TH057 core are mapped onto the primary sequences. Contact residues are defined by a 5-Å cutoff and marked above the sequence with 1 for the side chain and 2 for the main chain to indicate the type of contact. Contact types are colored as follows: green for hydrophilic, blue for hydrophobic, and black for both. Buried surface residues were determined by PISA and are shaded in green. Residues that differ between N49P6 and N49P7 are highlighted with dashed boxes. (D) Buried surface area (BSA) contributions to binding of gp120 residues (top) and N49P6/N49P7 Fab residues (bottom) as calculated by PISA. (Top) The gp120 sequence is color-coded in a gradient based on its sequence conservation: dark blue corresponds to the percentage of sequences in the HIV sequence compendium (https://www.hiv.lanl.gov/content/sequence/HIV/COMPENDIUM/compendium.html) that differ at that position from the Hxbc2 reference sequence 0.2 to 7% of the time, and red corresponds those that differ at that position from Hxbc2 87 to 99.9% of the time. Intermediate colors correspond to intermediate percentages on a roughly linear scale. Inner domain layers 2 and 3 and the outer domain D, V5, and CD4-binding loops are colored as described in the legend of Fig. 2. BSA contributions for N49P6 and N49P7 contact residues are shown. (Bottom) CDRs H1, H2, H3, L1, and L3 are colored as described in the legend of Fig. 2. BSA contributions to binding for N49P6 and N49P7 CDR residues are shown in darker and lighter shades of the CDR color, respectively, and those for FWR residues shown in orange and blue, respectively.
Structural Basis of Neutralizing Activity of N49P6 ® glycine at position 54 and leave this cavity empty (Fig. S1). Phe 43 cavity-filling mutations such as those found in many clade A/E HIV-1 strains, e.g., His 375 , may interfere with a bulky side chain at this position, as is seen for many small-molecule CD4-mimetic compounds that rely heavily on the Phe 43 cavity. Having Gly at position 54 may therefore contribute to increased breadth at the expense of potency, although for CD4bs Abs, this explanation may not be as straightforward since His 375 , at least in the case of N6, does not prevent binding and neutralization (9).  Fig. 3. BSA values for N49P6, N49P7, VRC01, VRC03, VRC07, NIH45-46, and N6 contact residues are shown. (D) Contact residues of gp120 mapped onto the gp120 93TH057 core sequences. Contact residues are defined by a 5-Å cutoff and marked above the sequence with 1 for the side chain and 2 for the main chain to indicate the type of contact. Buried surface residues were determined by PISA and are shaded blue.
Tolbert et al.

®
N49P6 and N49P7 are unique in their use of long 20-and 19-aa CDR H3s, respectively. These are the longest found among VRC01 class Abs; the next longest, found in VRC07 and NIH45-46, have 16-aa CDR H3s (Fig. S1). As shown in Fig. 4, the elongated CDR H3s allow N49P6 and N49P7 to increase their binding footprint to include the gp120 inner domain, specifically gp120 inner domain layer 3. N49P6 and N49P7 have the largest BSA contribution with gp120 inner domain layer 3 to the total BSA of the complex among all VRC01 class Abs (113 and 157 Å 2 for the N49P6 and N49P7 complexes, respectively, compared to a range of 34 to 99 Å 2 for the other CD4bs bnAb complexes) (Fig. 4A). These include contacts with conserved and highly conserved residues at positions 474, 476, and 480 of gp120 layer 3 mediated by Ser 100 , Lys/Arg 100B , and Arg 100D of the extended CDR H3 of N49P6/P7 (Fig. 4B and C); VRC07, NIH45-46, and N6 but not VRC01 and VRC03, reach residues in gp120 inner domain layer 3 through CDR H3 contacts mediated by Ala 100 /Ala 99A and Arg 100A /Arg 99B of VRCO7/NIH45-46 and Tyr 98 of N6. VRC07 and NIH45-46 also rely on gp120 inner domain layer 2 contacts in an area also recognized by N49P6 and N49P7 but with noticeably lower BSAs. In contrast, the more typical VRC01-like bnAbs, including VRC01, VRC03, and, to a lesser extent, N6, use their CDR H3 to almost contact loop D residues exclusively, with limited contact with the layers in the gp120 inner domain (Fig. 4) Figure 5A shows the structure of the full complex generated by crystallographic symmetry, which consists of three gp120s, three gp41s, and three N49P6 Fabs. Each CD4bs of the trimer is occupied by an N49P6 Fab, and the primary contact for the Fab is the CD4bs, which largely mirrors the interaction of the Fab and gp120 in the monomer complex structure (Fig. 5B). The total CD4-binding loop BSAs are 188 Å 2 and 184 Å 2 for gp120 and trimer complexes, respectively; the loop D BSAs are 297 Å 2 and 310 Å 2 ; and the loop V5 BSAs are 180 Å 2 and 240 Å 2 (Table S1). Differences in BSAs for the CD4bs between the monomer and trimer structures can largely be attributed to differences in sequence between clade A/E 93TH057 and clade A BG505 gp120 and differences in conformation due to deletions in the V1/V2 and V3 loops in the gp120 core structure, which can be seen in the BSA plot versus residue positions (Fig. 5C). The difference in sequence in loop V5 (residues 460 and 461) and layer 3 (residues 474 and 476) influences the BSAs for these residues and also likely neighboring residues, i.e., Gly 459 and Gly 473 . A major conformational change at Gln 428 represents the largest single-residue difference in BSAs between the two structures. Gln 428 sits on the b20-b21 turn on the outer domain half of the bridging sheet, which is formed in the monomer but not the trimer complex. The inner domain half of the bridging sheet, the V1V2 loop, extends to the top of the trimer, which disrupts the bridging sheet in the trimer, while a deletion in the V1V2 loop facilitates its formation in the monomer. The better packing of N49P6 around Gln 428 in the trimer in the absence of the bridging sheet is likely also an explanation for the slightly higher BSA for the trimer residue E 102 than for the monomer due to their proximity in the two structures. Conversely, the higher BSA for the monomer in loop D residues around Asn 276 may reflect differences in the glycans between the two structures. The gp120 complex was made with HEK 293 GnT1-grown protein with all but the initial N-acetylglucosamine removed by Endoglycosidase H (EndoH f ) (New England BioLabs). The glycan on Asn 276 in the trimer structure is present but disordered past the initial N-acetylglucosamine; the trimer complex was also made from GnT1-grown protein, but the  Fig. 2. (B) Structural comparison of the N49P6 Fab-gp120 93TH057 core e complex and the N49P6 Fab-gp120 (primary contact) protomer of the BG505 trimer. The complexes are superimposed based on gp120 93TH057 core e . The N49P6 Fab-gp120 93TH057 core e complex is colored as described in the legend of Fig. 2, and the heavy and light chains of N49P6 Fab in the N49P6-BG505 protomer complex are shown in dark and lighter shades of blue, respectively. The enlargements show the contacts between N49P6 Fab and gp120 93TH057 core (top) and the BG505 protomer (bottom). (C) BSA contributions to binding of gp120 (93TH057 core and BG505 gp120 [primary contact] protomer) residues (top) and N49P6 residues (bottom) as calculated by PISA. (Top) The gp120 sequence is color-coded based on its sequence conservation as described in the legend of Fig. 3. Inner domain layers 2 and 3 and the outer domain D, V5, and CD4-binding loops are colored as described in the legend of Fig. 2. BSA values for N49P6 and N49P7 contact residues are shown in blue and red, respectively. (Bottom) All CDRs, H1, H2, H3, L1, and L3, are colored as described in the legend of Fig. 2. BSA contributions to binding of N49P6 contact residues to gp120 93TH057 core e and the primary BG505 protomer are shown in darker and lighter shades of the CDR color, respectively. Tolbert et al. high-mannose N-glycans were left intact in the protein used for crystallization. The truncated glycan may permit N49P6 to have a more stable light chain interaction in the monomer than in the trimer, although the total light chain contributions to the BSA are largely identical between the two, 245 Å 2 for the monomer and 258 Å 2 for the trimer; the light chain's contribution to binding is almost exclusively to the glycan at Asn 276 and nearby residues in both structures.
Although the primary CD4bs for both complexes are similar, contacts with the adjacent protomer increase the BSA of the complex formed by each N49P6 Fab with the trimer to make it significantly larger (Table S1). These interprotomer interactions are identical for each Fab due to crystal symmetry and are made mostly by the framework region immediately preceding CDR H1 and one CDR H1 residue (Gly 26 , Tyr 27 , Asp 28 , and Tyr 32 ). CDR H3 adds two arginines to the interface (Arg 96 and Arg 100C ), including the one (Arg 100C ) inserted in the CDR H3 relative to N49P7 (Fig. 5A and B). Contacts within the adjacent promoter map to mobile layer 1 on the gp120 inner domain and include Tyr 61 , Glu 64 , Lys 65 , and His 66 with Fab-gp120 hydrogen bonds (the carbonyl oxygen of Gly 26 (Fig. S1).
N49P6 mimics CD4 in its initial quaternary interaction with BG505 SOSIP.664. It has been shown previously that the first CD4 that binds a BG505 SOSIP trimer has both a primary and a secondary binding site (23). The primary site is essentially identical to that used in CD4 structures with monomeric gp120 (24). The secondary CD4bs resides on the adjacent gp120 within inner domain layer 1 (23). Interestingly, a structural alignment of the complexes formed between the BG505 SOSIP.664 trimer and three N49P6 Fabs or a single CD4 molecule (Protein Data Bank [PDB] accession number 5U1F) (23) reveals that the layer 1 gp120 residues used by N49P6 to contact the adjacent protomer are largely identical to the ones that CD4 uses in its initial quaternary interaction with the trimer (Fig. 6A through C). Although the resolution of the monovalent CD4-BG505 SOSIP.664 trimer complex was low, which introduces some ambiguity into the details of the interface, the CD4 contacts with the adjacent protomer were mapped to residues 62 to 66 of layer 1 and within the coreceptor-binding site, Lys 207 . (Fig. 6B) (23). Interestingly, N49P6's interprotomer contacts involve three (Glu 64 , Lys 65 , and His 66 ) of five residues utilized by CD4 within layer 1. Of note, residues Glu 64 , His 66 , and Lys 207 are almost universally conserved across HIV-1 isolates, and mutational studies confirm their importance in proper spike function and viral infectivity (23). N49P6 capitalizes on the functional importance and the sequence conservation of these residues to increase its affinity and breadth.
It has been shown previously that the one CD4-bound BG505 SOSIP trimer transitions very quickly to a more open conformation with three CD4s bound that fully exposes the coreceptor-binding site (25). In this state, the contacts of CD4 with the adjacent protomer are lost (26) (Fig. 7). The binding of coreceptor or coreceptor-bindingsite antibody stabilizes this open conformation with a few additional changes (26,27). Charge inversion within the quaternary CD4-binding site in HIV-1 Env of His 66 to Glu or Glu 62 to Lys and Glu 64 to Lys abrogates viral infectivity, highlighting the importance of this region (23). It is interesting to note that N49P6 interacts with these residues much like a charge-inverted virus to CD4, with Asp 28 (N49P6) forming a hydrogen bond to His 66 of the adjacent gp120 in the trimer and Arg 96 (N49P6) forming a hydrogen bond to Glu 64 (gp120) (Fig. 6C). The inferred binding based on distance in the CD4 complex is between Lys 21 (or Lys 22 ) (CD4) and His 66 (gp120) and between Gln 20 (CD4) and Glu 64 (gp120), which would seem less stabilizing. This may reflect CD4's role in infection, which is to drive the transition to an open conformation and expose the coreceptor-binding site. In contrast, N49P6's role in neutralization is to bind and lock the trimer in a closed state.
The N49P6 quaternary interaction is different from the quaternary interactions of other potent CD4bs Abs. N49P6 is not the only CD4bs bnAb to extend its footprint onto the adjacent protomer. Many of the broader and more potent CD4bs bnAbs use different interprotomer contact strategies to lock the trimer in a more closed conformation; some target elements of the gp120 inner domain, while others target the coreceptor-binding site at the base of the V3 loop. Figure 6C and D show the interprotomer contact details of many of these bnAbs compared to N49P6 and CD4. The most like N49P6 is the prototypic VRC01 class Ab VRC01, which binds to inner domain layer 1 of the adjacent gp120 with a total BSA of 102 Å 2 , compared to 206 Å 2 for N49P6, 199 Å 2 from layer 1 and 7 Å 2 from layer 2 (2,28). VRC03, another VRC01 class Ab from the same donor, uses heavy chain framework residues and a unique insertion in the heavy chain framework region to contact gp120 inner domain layer 1, 42 Å 2 , but relies more heavily on inner domain layer 2, 90 Å 2 , and the base of the V3 loop, 47 Å 2 , for a total gp120 BSA of 179 Å 2 (29,30). NIH45-46 also contacts inner domain layer 1 on the adjacent trimer, 38 Å 2 , but without the heavy chain framework insertion contacts only layer 1. 3BNC117 contacts the adjacent gp120 inner domain layer 1, 40 Å 2 , and the base of the V3 loop, 125 Å 2 , for a total of 165 Å 2 (31). Similarly, CH31 uses both heavy chain framework and CDR H1 residues to bind the adjacent gp120 inner domain layer 1, 100 Å 2 ; layer 2, 140 Å 2 ; and the base of the V3 loop, 92 Å 2 ; for a total of 332 Å 2 (29,30). And finally, 1-18 uses its heavy chain framework and CDR H1 residues to contact inner domain layer 2, 104 Å 2 , and the base of the V3 loop, 175 Å 2 , for a total BSA of 279 Å 2 (Table S2) (10).
An important consideration when comparing quaternary contacts in the trimer is the degree to which the trimer "opens" upon Ab binding. The HIV-1 BG505 SOSIP trimer goes from a "closed" trimer with distances between protomers as measured by the a-carbon of residue 375 at the base of the Phe 43 cavity of 54.5 Å to an average of 69.6 Å in the fully open trimer with three CD4s bound (Fig. 7). The degree to which a trimer opens determines the degree to which quaternary contacts are possible. It is worth noting that all the CD4bs bnAbs that make contacts with the adjacent gp120 have BG505 SOSIP trimers in a more closed conformation, more closed than even the BG505 SOSIP trimer with only one CD4 bound, which has an average distance of 57.4 Å. The average distance between protomers ranges from 54.5 to 55.9 Å, with N49P6 having an average interprotomer distance identical to that of the unbound BG505 SOSIP trimer, 54.5 Å. Some BG505 SOSIP trimers are asymmetric, like NIH45-46, which has quaternary contacts for only two of the three possible adjacent protomers in the trimer, but most are largely symmetric. NIH45-46 also has the smallest adjacent protomer footprint, which may explain its inability to lock its BG505 SOSIP trimer in a closed conformation.
N49P6's total adjacent protomer footprint is larger than those of VRC01, VRC03, NIH45-46, and 3BNC117 but smaller than those of 1-18 and CH31. However, N49P6 has the largest BSA for the inner domain layer 1 region that overlaps CD4's interprotomer footprint, twice the layer 1 footprints of VRC01 and CH31 and roughly four times those of VRC03 and NIH45-46. VRC03, 3BNC117, 1-18, and CH31 focus their epitopes more toward inner domain layer 2 and the base of the V3 loop of the adjacent trimer. This difference in strategy may ultimately determine neutralization breadth since breadth is a function of residue conservation as well as affinity.

DISCUSSION
The primary CD4-binding site (CD4bs), the major contact interface between the HIV-1 Env spike and host receptor CD4, can be broken down into three interacting regions, the CD4-binding loop, loop D, and loop V5 (32). Host receptor CD4 makes a number of important contacts with each of these regions that are highly conserved across HIV-1 strains (24). Broadly neutralizing CD4bs antibodies utilize many, if not all, FIG 7 Structural rearrangements of the HIV-1 Env trimer upon binding to host receptor CD4 and CD4-binding-site-specific bnAbs. (A) Overall structure of the unbound BG505 SOSIP.664 trimer (referred to as the "unbound" BG505 SOSIP conformation) (PDB accession number 4ZMJ) (left) and the trimer with 3 CD4s bound (referred to as the "soluble CD4 [sCD4]-bound" BG505 SOSIP conformation) (PDB accession number 5VN3) (right). Trimers were aligned based on the a-carbon positions of the central gp41 a7 helices, residues 570 to 595, and a center (Centr) (gray sphere) for the alignment calculated from the combined a-carbon positions. The relative position for each gp120 was then calculated based on the a-carbon position for residue 375 at the base of the CD4 Phe 43 -binding pocket in each gp120 of the trimer ( 375 C a ) (shown as blue spheres). The distances between the center and the 375 C a of each protomer (a, b, and c for "unbound" and a9, b9, and c9 for "sCD4 bound") and the 375 C a atoms of neighboring protomers (d, e, and f for unbound and d9, e9, and f9 for sCD4 bound) are shown to indicate the extent of the protomer rearrangement after CD4 binding. (B) Comparison of unbound, sCD4-bound, and bnAbbound conformations of the HIV-1 BG505 SOSIP trimer. ) were aligned as described above for panel A, and the 375 C a atom of each complex is shown as colored spheres. Distances for sCD4-bound and unbound BG505 SOSIP trimers are labeled as described above for panel A, with the equivalent N49P6 distances added as red dashes. The clockwise rotation for each gp120 in the sCD4-bound trimer relative to the equivalent gp120 in the closed unbound trimer is also shown and labeled a-a9, b-b9, and c-c9. (C) a, b, c, d, e, and f distances and a-a9, b-b9, and c-c9 rotation angles for each bnAb and CD4 complex relative to the unbound BG505 SOSIP trimer as shown in panel B.
Tolbert et al.

®
of these residues to achieve their neutralization breadth (6,29,32,33). Increasing neutralization potency, on the other hand, also requires the use of residues outside the CD4bs for effective competition with CD4. Which residues and their conservation across HIV-1 strains are often limiting factors in an antibody's neutralization breadth, while the number and strength of the interactions outside the CD4bs are limiting factors in determining their neutralization potency.
HIV-1 has developed several strategies to prevent antibodies from binding the CD4bs and to pivot away from those that do bind to prevent neutralization. The CD4bs is recessed and shielded by heavy glycosylation on Env, which prevents all but a few antibodies from binding (34,35). Broadly neutralizing antibodies that bind within the CD4bs have had to develop strategies to overcome this hurdle. It has been shown for the prototypical VRC01 class of CD4bs-specific bnAbs that changes, alone or in combination, in the lengths of gp120 loops D and V5 and the glycosylation pattern in loop D constitute major viral escape mechanisms through steric clashes with antibody CDR H2, the N terminus of the light chain, and/or antibody CDR L3 (36)(37)(38). VRC01-like antibodies adopt a short 5-amino-acid (aa)-long CDR L3 and often deletions in CDR L1 to accommodate changes in these regions (6,29,33). N49P6 together with N49P7 constitute a new branch of the VRC01-like class of CD4bs-specific bnAbs that, in addition to the strategy of shortening CDR L1 and L3, also utilize the light chain's rotation/tilt to provide additional space and prevent steric clashes with loops D and/or V5. The latter is possible due to a mutation in light chain framework region 2 (FWR2), Tyr 36 to Cys, which allows greater mobility of the heavy and light chains relative to one other. N49P6 also contains a deletion in its CDR L1 (8 aa) in addition to the extremely short CDR L3 (5 aa) found in VRC01-like bnAbs. The combination results in a minimal light chain footprint, which can be seen by the low buried surface area (BSA), 245 Å 2 (157 Å 2 without the contribution from the Asn 276 glycan), compared to those of most other Abs in its class (see Table S2 in the supplemental material). In addition, N49P6's CDR H2 also contains a Pro 60 ArgPro 62 motif, which preferentially adopts a left-handed polyproline helical conformation (with phi and psi angles for Pro 60 of 259°and 145°, Arg 61 of 247°and 244°, and Pro 62 of 258°and 211°, respectively) that points away from loop V5, further minimizing steric clashes with V5. Although N49P6 utilizes Asn 276 and its attached glycan in its epitope, this interaction seems to be mainly limited to the initial N-acetylglucosamine. This is in contrast to VRC01 and many other VRC01-like bnAbs, which utilize more of the glycan in their epitope (39). A change in the glycan position from position 276 to position 279 in gp120, which keeps steric pressure on the CD4bs but removes many of the beneficial glycan interactions, results in neutralization resistance to many of these other CD4bs bnAbs. This change may be less of a problem for N49P6 due to its limited dependence on the Asn 276 glycan.
The glycan at Asn 276 prevents the binding of all but a few germ line antibody alleles to the CD4bs. Those that do either bind by utilizing a long CDR H3 to bypass light chain-mediated steric clashes or bind with one of two germ line alleles, V H 1-2 (VRC01 class) or V H 1-46 (8ANC131 class), in combination with light chains with deletions in CDR L1 and L3 (6,29,33). N49P6, as a member of the VRC01 class, utilizes V H 1-2 with a relatively long CDR H2 and CDR H3, composed of 17 and 20 aa, respectively. The V H 1-2 germ line gene has several characteristics of its CDR H2 that permit it to mimic CD4. For example, both the typical V H 1-2 CDR H2 and CD4 form a short antiparallel b-sheet with residues in the gp120 CD4-binding loop. In addition, V H 1-2 germ line-derived bnAbs also often mimic the salt bridge between Arg 59 (CD4) and Asp 368 (gp120) and utilize the hydrophobic cavity that binds Phe 43 in CD4. The Phe 43 -binding cavity constitutes an important anchoring point for CD4bs-specific antibodies and small-molecule CD4-mimetic compounds that rely on hydrophobic contacts with residues lining the cavity (40). N49P6 preserves many of these features along with the Arg-Asp 368 salt bridge mediated by heavy chain framework residue Arg 71 ; however, N49P6, like N49P7, has a glycine at heavy chain position 54, which leaves the Phe 43 cavity empty.
The Phe 43 cavity is a key feature used by many small-molecule CD4-mimetic compounds to modulate the Env conformation and in many cases to drive Env to the CD4bound, open conformation (40)(41)(42). To compensate for its loss, N49P6 establishes an extensive network of interactions with loop D, loop V5, and, most importantly, the highly conserved residues within the gp120 inner domain that include inner domain layers 2 and 3. In addition, N49P6 also contacts inner domain layer 1 on the adjacent gp120 in the trimer, largely mimicking the interprotomer contact of host receptor CD4. The inner domain residues involved in the N49P6 epitope are highly conserved among HIV-1 clades, and N49P6 takes advantage of this. The mobile layers of the gp120 inner domain are involved in structural rearrangements of the HIV-1 trimer after CD4 attachment and coreceptor binding. They form the gp41-interacting face of gp120, and their conformational change serves as a driving force in trimer disassembly and cellular fusion. Their mutation comes at a fitness cost to the virus, which enables N49P6 breadth.
Finally, one of the most interesting features of how N49P6 engages HIV-1 Env is in its contact with the adjacent protomer. The mode of binding to the trimer and the scope of interprotomer contact residues largely resemble those of the interaction of host receptor CD4 in its initial binding to the HIV-1 Env trimer. Structural alignment of the N49P6-BG505 SOSIP trimer complex to the complex of the BG505 SOSIP trimer with one CD4 molecule bound (23) reveals a large overlap of the Fab/CD4 contact surface on the adjacent protomer. Other CD4bs-specific bnAbs use the adjacent protomer to increase their epitope footprint but tend to focus instead on the coreceptor-binding site at the base of the V3 loop and less on layer 1 at the site used by CD4. In this regard, N49P6 is unique in sharing many of the same contact residues used by CD4 in its initial binding to the trimer. Interestingly, it is known that following the initial contacts of one CD4 molecule to the HIV-1 BG505 SOSIP trimer, the engagement of a second and a third quickly follows, which leads to the full opening of the trimer (26). This is in contrast to the case of N49P6, where the BG505 SOSIP trimer is "locked" in the closed conformation with interprotomer distances almost identical to those of the unbound trimer.
Neutralization of the global panel of HIV-1 pseudoviruses by N49P6 is broad and potent but still leaves room for improvement, especially when IC 80 values are used for comparison. Are there any clues from the more resistant sequences that could suggest other means to improve neutralization? Sequences from the panel that are better neutralized by N49P7 are largely conserved across N49P6 contact residues, suggesting that resistance is more subtle than those due to single point mutations (Fig. S2). Indeed, the generally high degree of conservation among epitope residues may imply that there is a large fitness cost for their mutation. Instead, one common theme among more resistant sequences in the panel and the sequences from donor N49, which are resistant to N49P6/7, is a larger V5 loop. Although most of the N49P6 epitope residues in the V5 loop reside at its base, insertions in this region may result in steric hindrance or conformational rearrangement that may lower Ab affinity. Examination of viral sequences from donor N49 shows exceptionally long V5 loops with the possible addition of a glycan, which may be a mechanism of resistance. The only other obvious difference in N49 donor sequences is the exceptionally long V1 region in the V1V2 loop (Fig. S2). To compensate, N49P7 has increased its binding to the trimer within the primary CD4-binding site. N49P6 has opted instead to increase its binding to the secondary CD4-binding site on the adjacent protomer in the trimer. Both approaches seem to have merit when applied to the global pseudovirus panel.
In conclusion, the N49 lineage of near-pan-neutralizing antibodies deconvoluted from the plasma of an HIV-1-infected individual represents a new class of CD4-bindingsite-specific antibodies that we refer to as the N49P series. These antibodies omit the Phe 43 cavity and rely strongly on the gp120 inner domain for binding to the primary gp120 protomer, and they mimic the initial contact of host receptor CD4 with the adjacent promoter of the HIV-1 Env trimer. When bound to the HIV Env trimer, N49P series antibodies lock the trimer in the closed conformation using residues from the initial contact of the CD4 receptor with the adjacent protomer in the trimer.

MATERIALS AND METHODS
Protein production and purification. N49P6 Fab was produced by papain digestion of N49P6 IgG as described previously (43,44). BG505 SOSIP.664 was a kind gift from John Moore (Weill Cornell Tolbert et al.