Regulation of epitope exposure in the gp41 membrane-proximal external region through interactions at the apex of HIV-1 Env

Glycoprotein Env of human immunodeficiency virus type 1 (HIV-1) mediates viral entry through membrane fusion. Composed of gp120 and gp41 subunits arranged as a trimer-of-heterodimers, Env adopts a metastable, highly dynamic conformation on the virion surface. This structural plasticity limits the temporospatial exposure of many highly conserved, neutralizing epitopes, contributing to the difficulty in developing effective HIV-1 vaccines. Here, we employed antibody neutralization of HIV-1 infectivity to investigate how inter- and intra-gp120 interactions mediated by variable loops V1/V2 and V3 at the Env apex regulate accessibility of the gp41 membrane-proximal external region (MPER) at the Env base. Swapping the V3 loop from EnvSF162 into the EnvHXB2 background shifted MPER exposure from the prefusogenic state to a functional intermediate conformation that was distinct from the prehairpin-intermediate state sensitive to gp41-targeted fusion inhibitors. The V3-loop swap had a profound impact on global protein dynamics, biasing the equilibrium to a closed conformation resistant to most anti-gp120 antibodies, stabilizing the protein to both cold- and soluble CD4-induced Env inactivation, and increasing the CD4 requirements for viral entry. Further dissection of the EnvHXB2 V3 loop revealed that residue 306 uniquely modulated epitope exposure and trimer stability. The R306S substitution substantially decreased sensitivity to antibodies targeting the gp41 MPER and, surprisingly, the gp120 V3-loop crown (residues 312–315), but had only modest effects on exposure of intervening gp120 epitopes. Furthermore, the point mutation reduced soluble CD4-induced inactivation, but had no impact on cold inactivation. The residue appeared to exert its effects by electrostatically modifying the strength of intra-subunit interactions between the V1/V2 and V3 loops. The distinct patterns of neutralization and stability pointed to a novel prefusogenic Env conformation along the receptor activation pathway and suggested that apical Env-regulation of gp41 MPER exposure can be decoupled from much of the dynamics of gp120 subunits.


Introduction
The Envelope glycoprotein (Env) from human immunodeficiency virus type 1 (HIV-1) is responsible for mediating viral entry through membrane fusion [1]. As the sole viral protein on the virion surface, Env is the primary focus of vaccine development efforts to generate neutralizing antibodies (NAbs) that block viral replication [2]. Unfortunately, these efforts have proven challenging for a number of reasons, including the dense glycan shield on Env providing a flexible barrier to the underlying protein, the breadth of Env sequence diversity in the general population, the speed at which resistant Env species emerge within an individual, and the conformational plasticity of Env that temporally limits exposure of conserved epitopes critical for function [3][4][5]. In this study, we further investigated the regulation of this conformational flexibility by exploring how inter-and intrasubunit interactions at the apex of Env control NAb epitope exposure at the base of the complex.
Env is a homotrimeric, single-pass transmembrane protein with each protomer consisting of two subunits: surface subunit gp120 responsible for binding cellular CD4 and chemokine receptors (CCR5 and/or CXCR4, collectively denoted CoR), and transmembrane subunit gp41 responsible for catalyzing membrane fusion [6]. In the native state, the gp120 subunits form a canopy that covers much the gp41 trimer, with extensions of each gp120 subunit tightly enmeshed with its own gp41 elements at the base (juxtamembrane portion) of the complex [7][8][9]. The trimeric organization is stabilized by gp120-gp120 interactions involving variable loops V1/V2 and V3 at the apex and by a coiled coil formed by part of the gp41 N-terminal heptad repeat (N-HR) at the base.
In the presence of target cell membranes, CD4 binding to the gp120 canopy disrupts V1/ V2-and V3-loop interactions, leading them to disengage from the conserved Env core as the top of each gp120 subunit rotates and relaxes radially outward [10][11][12][13][14]. This general opening of the Env structure releases steric constraints on several neutralizing epitopes and exposes a respectively, and differ at 10 residue positions ( Fig 1C). The largest physiochemical change is a net charge reduction from +9 for HXB2 to +3 for SF162. Unlike the parental H4 Env, chimeric H5 Env was poorly fusogenic on CD4 + CXCR4 + CCR5cells but mediated robust fusion on CCR5-expressing target cells (S1 Fig) [64].
Despite switching CoR tropism, the V3-loop swap did not exert a dramatic effect on late Env conformational changes triggered by CoR binding. H4 and H5 Envs were similarly sensitive to fusion inhibitors (FIs) C37-KYI and 5-Helix, implying that the spatiotemporal exposure of the gp41 N-HR and C-HR regions during the PHI were not substantially altered (Fig 1D  and 1E) [65,66]. Curiously, inserting the SF162 V3 loop into Env HXB2 led to a 29-fold reduction in the potency of NAb 2F5, whose extended epitope at the junction of the gp41 C-HR and MPER regions partially overlaps the binding site for 5-Helix ( Fig 1F) [67]. Comparable reductions in potency were observed for NAbs 10E8 and 4E10, which target MPER segments C-terminal to that for NAb 2F5 (Fig 1G). In fact, the IC50 values measured for these anti-MPER NAbs against H5 Env were close to or exceeded the geometric mean IC50 value reported for tier 2 clade B isolates [S1 Table, data from the Los Alamos National Laboratory (LANL) HIV Immunology Database, January 2022]. The loss of potency did not appear to reflect reduced binding strength, as neutralization was effectively irreversible for HIV-1 pseudotyped with either H4 or H5 Env in antibody dilution assays (S2 Fig). Rather, our results suggested H4 and H5 Envs substantially differ in temporospatial exposure of the MPER, mirroring previous observations that showed mutations at the Env apex can exert dramatic effects on the juxtamembrane region over 90 Å away [48].
To assess the global impact of the V3-loop swap on Env structure, we compared sensitivities of H4 and H5 Envs to a panel of neutralizing antibodies with epitopes throughout the glycoprotein (Fig 1B and S1 Table). H5 Env was significantly less sensitive to antibodies targeting most tertiary and quaternary epitopes, including the V3 crown (the -GPGR-sequence invariant to the V3-loop swap), the CD4 binding site, the bridging sheet exposed in the CD4-induced (CD4i) conformation, and the gp120-gp41 interface ( Fig 1H). Notably, the potency of anti-CD4i antibody 17b was reduced more than 300-fold by the V3-loop swap. Two exceptions to this trend were NAbs 2G12 and PG9. NAb 2G12 targets lateral surface glycans at residue positions unaffected by the V3-loop swap [68,69]; not surprisingly, the antibody neutralized H4 and H5 Envs similarly. NAb PG9 targets a quaternary epitope of V1/V2 loops most stabilized in apo-Env conformations [2,70,71]; interestingly, this antibody had more potent activity against H5 Env than H4 Env. Overall, the neutralization profile of H5 Env resembled that for tier 2 clade B primary isolate Envs with closed, epitope-occluded prefusogenic conformations CD4 binding preferentially stabilizes the CD4i conformations, while chemokine-receptor (CoR) binding triggers transition into the prehairpin intermediate state (PHI). The PHI ultimately collapses into the fusogenic trimer-of-hairpins (TOH) structure. The placement of entry inhibitors and NAbs along the timeline (bottom) indicates the approximate point at which each one blocks downstream conformational transitions. (B) Ribbon diagram of the Env JRFL SOSIP.664 structure (PDB 5FYK [79]) with each protomer illustrated in a different color. The V1/V2 and V3 loops are shown in isolation (enlargement). The approximate locations of neutralizing epitopes are indicated for the NAbs used in this study. (C) Sequence alignment of the V3 loops from Env HXB2 and Env SF162 (numbering according to the Env HXB2 sequence). Dash ("-") represents identically conserved amino acids, while dot (".") indicates a sequence gap. (D-F) Inhibitory titrations of fusion inhibitors C37-KYI (D) and 5-Helix (E) and anti-gp41 MPER bNAb 2F5 (F) against HIV-1 pseudotyped with either wild type Env HXB2 (H4, filled circles)) or V3-loop chimeric H5 Env (open circles). (G-H) Sensitivity of H4 (solid) and H5 (hatched) Envs to antibodies targeting the MPER (G) or other regions of the glycoprotein trimer (H). The average IC50 values for each NAb are compared to the geometric means of IC50 values against tier 2-clade B (magenta lines) and tier 3 (slate lines) Envs (compiled from LANL HIV Immunology Database; lines are solid or dashed depending, respectively, on whether >85% or <85% of isolates were reportedly neutralized). Arrows above the bars for antibodies F105, 447-52D and 17b against H5 Env indicate that 50% neutralization was not achieved at the highest concentrations tested. Data points in D-F and bars in G-H represent the mean±SEM from three or more independent experiments. Single asterisk in G-H indicates a significant difference as determined by a two-sample t-test (equal variance not assumed) with p < 0.05; ampersand (for NAbs PG9 and VRC01) denotes a p-value between 0.050 and 0.051.

The V3-loop swap modifies temporal properties of MPER exposure
The exact timing of anti-MPER antibody binding to Env when neutralizing HIV-1 infectivity remains unclear. Some studies demonstrate a correlation between neutralization potency and MPER accessibility in the absence of target cells, suggesting the interaction occurs with the prefusogenic state [17,22,48]. Other studies show that MPER exposure is transiently enhanced following receptor activation of Env, suggesting that the interaction occurs in a fusion-intermediate state like the PHI [72,73]. To test these models, we determined how NAb 10E8 potency depended on fusion rate, as modified by CoR antagonists (CoRAs). CoRAs reduce CoR binding stoichiometry with the Env trimer, thereby prolonging intermediate state lifetimes [64,74]. As shown previously, CoRAs enhanced the potency of bona fide intermediate-state FIs C37-KYI and 5-Helix in a dose-dependent manner (Fig 2A-2C). For NAb 10E8, however, the degree of CoRA-dependent enhancement varied depending on Env strain. Against H4 Env, the enhancement of 10E8 potency was significantly lower than that for FIs (Fig 2A). By contrast, against H5 Env and J5 Env (Env JRFL , a representative tier 2 clade B primary isolate strain), the enhancement of 10E8 potency matched that for FIs (Fig 2B and 2C). The data suggested that much of NAb 10E8 binding to H4 Env occurs in the prefusogenic state outside the context of viral membrane fusion, while the majority of 10E8 neutralization of H5 and J5 Envs occurs during a functional intermediate state of viral entry with a lifetime dependent on CoR binding. This shared CoRAdependence to NAb 10E8 neutralization for both H5 and J5 Envs underscored the tier 2-like characteristics conferred by the V3-loop swap in the Env HXB2 background.
To test whether NAb 10E8 and FIs interact with gp41 during the same intermediate state, we introduced mutations that intrinsically altered the lifetime of the PHI. The double substitution N637K/T639I in the gp41 C-HR segment shortened PHI lifetimes [64], leading to a 3-to 6-fold increase in the 5-Helix IC50 values for H4, H5, and J5 Envs ( Fig 2D). Conversely, the single substitution L565Q in the gp41 N-HR segment lengthened the PHI lifetime [64], leading to a 13-fold and 4.4-fold decrease in 5-Helix IC50 values for H4 and H5 Envs, respectively. Neither substitution significantly altered NAb 10E8 potency against H4 Env (Fig 2E), as we had expected since the MPER epitope appears to be accessible for antibody binding outside the context of membrane fusion. Curiously, the mutations also had minimal impact on NAb 10E8 potency against H5 Env; moreover, the rate enhancing N637K/T639I substitution actually increased J5 Env sensitivity to NAb 10E8 neutralization. These observations implied that either the functional intermediate state targeted by NAb 10E8 does not temporally overlap with the PHI or the PHI duration constitutes only a small fraction of the NAb 10E8-neutralization window.

The V3-loop swap significantly stabilizes the apo-conformation of the Env HXB2 trimer
To assess how the V3-loop swap impacted the prefusogenic Env HXB2 conformational equilibrium, we first explored Env sensitivity to sCD4. As an inhibitor of viral infectivity, sCD4 had more than 50-fold lower potency against H5 Env compared to H4 Env (Fig 3A), on par with the reduction observed in the potencies of NAbs targeting the CD4 binding site (Fig 1H). Likewise, the sCD4 dependence to gp120 shedding was shifted to greater than 10-fold higher concentrations for H5 Env (Fig 3B). These changes were quantitatively similar to the differences between H4 and J5 Env and, together, implied that the V3-loop swap made fluctuations into CD4-bound conformations significantly more unfavorable.

PLOS PATHOGENS
HIV-1 Env apex regulates gp41 MPER exposure Decreased sensitivity to sCD4-induced gp120 shedding is historically associated with increased Env thermal stability [43,55,75]. To test for this correlation, we examined how the V3-loop swap altered heat and cold inactivation of viral entry. While heat inactivation did not distinguish between H4, H5 and J5 Envs (S3 Fig), differences were readily apparent in cold inactivation experiments. H5 and J5 Envs closely tracked with one another, maintaining greater than 50% infectivity after 96 hours at 0˚C. By contrast, H4 Env lost the majority of its activity by 36 hours and showed only 10% activity at the 96-hour timepoint ( Fig 3C). Together, these data were consistent with the V3-loop swap energetically stabilizing a closed conformation with a suboptimal CD4-binding site configuration.
To examine how this energetic change altered CD4 utilization during HIV-1 entry, we measured viral infectivity as we titrated available CD4 levels on target cells using the CD4-antagonist DARPin 23.2 (D23.2) [76]. Since H4 and H5 Env engage different CoRs whose expression Note that CoRA concentration is expressed relative to the CoRA IC50 for each virus (94, 61 and 23 nM for H4, H5 and J5 Envs, respectively). Data points represent the mean±SEM of four independent experiments. For H4 Env, the NAb 10E8v4 data set is significantly different (p < 0.025) from the 5-Helix data set as determined by a two-way ANOVA; for the comparison of 10E8v4 and C37-KYI data sets, p = 0.029. No statistical difference was observed in any of the data set comparisons for H5 and J5 Envs. (D, E) Potencies of fusion inhibitor 5-Helix (D) and NAb 10E8v4 (E) against Envs with altered PHI lifetimes. IC50 values were determined against HIV-1 pseudotyped with H4 Env (solid black), H5 Env (hatched black) and J5 Env (red) containing either the rate-enhancing N637K/T639I substitution or the rate-retarding L565Q substitution. Bars represent the mean±SEM of three or more independent experiments. Single asterisks indicate significance as determined by one-way ANOVA with means comparisons using the Tukey method with p < 0.025 [109]; double asterisks indicate significance as determined by a twosample t-test (equal variance not assumed) with p < 0.01. https://doi.org/10.1371/journal.ppat.1010531.g002

PLOS PATHOGENS
HIV-1 Env apex regulates gp41 MPER exposure Env (red symbols). (B) Soluble CD4 dependence to gp120 shedding from H4, H5 and J5 Env-expressing HEK293T cells. Shed gp120 was detected by SDS-PAGE/Western blot, quantified by densitometry and normalized to the total gp120 detected in cell lysates. In the images of representative Western blots (bottom), L denotes a 1:3 dilution of cell lysate; the lines to the right of the images indicate levels and interaction properties could affect CD4 utilization, we simultaneously used CoRAs to titrate surface CXCR4 and CCR5 levels. At all functional CoR levels, H5 Env was greater than 3.5-fold (average 5.1-fold) more sensitive to D23.2 inhibition than H4 Env (Figs 3D and 3E and S4). Moreover, the steepness of each titration along the D23.2 axis was substantially greater for H5 Env (average Hill Coefficient n H = 1.6) than for H4 Env (n H = 0.73) (S4 Fig). In fact, the CD4-and CoR-dependence to H5 Env-mediated entry were almost identical to that for J5 Env (Figs 3F and S4). Together, the data implied that H5 Env requires a higher cellular CD4 concentration and a higher CD4 binding stoichiometry per Env trimer to support entry. The changes in CD4 dependence to neutralization, shedding and entry induced by the V3-loop swap paralleled previously reported differences between primary isolate and laboratory-adapted Envs [22,45,46,77].

Swapping V3 loops has a variable impact in other Env strains
We introduced the HXB2 V3 loop into CCR5-tropic Envs from viral strains JRFL, BG505 and SF162 ( Fig 4A). For the JRFL and SF162 strains, the chimeric Envs (hereafter denoted J4 and S4, respectively) adopted CXCR4-tropism and were poorly fusogenic when using CCR5 as CoR (S1 Fig). By contrast, the chimeric Env for the BG505 strain (denoted B4) displayed significant activity using either CXCR4 or CCR5 for viral entry. J4, B4 and S4 Envs were all neutralized more easily by anti-MPER NAb 10E8 compared to their wild type counterparts ( Fig  4B and S2 Table). For J4 Env, NAb 10E8 and FI potency showed the same CoRA dependence, suggesting that the higher sensitivity of some chimeric Envs to anti-MPER antibodies did not necessarily correlate with substantially enhanced MPER exposure in the prefusogenic state prior to receptor activation ( Fig 4C). In the case of Env JRFL , swapping in the Env HXB2 V3 loop conferred the same increases in potency to FI C37-KYI and 5-Helix (S5 Fig), suggesting that enhanced NAb 10E8 potency against J4 Env likely resulted from slower transit through the functional intermediate state that exposed the MPER (see Discussion).
The chimeric Envs were also more sensitive to NAbs targeting the CD4i-exposed bridging sheet (17b) and V3-loop crown (447-52D) (Fig 4D and 4E and S2 Table). The V3-loop swaps had a more nuanced effect on the potencies of antibodies targeting the CD4 binding site: NAbs that preferentially bound more activated prefusogenic conformations (b12 and F105) were more potent against chimeric Envs (Fig 4F and 4G), while the NAb that was less sensitive to prefusogenic conformational fluctuations (VRC01) displayed similar potency against the wild type and chimeric species (Fig 4H) [2,78]. As expected, the V3-loop substitution had minimal impact on NAb 2G12 potency with the exception of Env BG505 (Fig 4I, the disparity between B4 and B5 sensitivity might reflect a difference in the glycan at residue 295 that rendered B5 Env insensitive to NAb 2G12 [79]). Altogether, introduction of the HXB2 V3 loop appeared to globally increase accessibility to epitopes found in activated Env conformations, similar to the way that mutations in the V1/V2 and V3 loops have been shown to do [48,53,57,60].
We also performed a V3-loop swap in CXCR4-tropic Env NL4-3 , which is closely related to H4 Env. Inserting the SF162 V3 loop into the NL4-3 background generated the CCR5-tropic positions of 160 and 100 kDa molecular weight markers. (C) Time dependence to cold inactivation of HIV-1 pseudotyped with H4, H5 or J5 Env. Data for each virus were normalized to the infectivity detected after a one-hour incubation on ice. (D-F) Dependence of viral infectivity on CD4 and CoR levels for HIV-1 pseudotyped with H4 Env (D), H5 Env (E) and J5 Env (F). Target cells were U87.CD4. CXCR4 (D) and U87.CD4.CCR5 (E, F). Receptors levels were titrated by adding CD4 antagonist D23.2 and CoRA AMD3100 (D) or TAK-779 (E,F) alone or in combination during the infection. Data have been normalized to the infectivity level in the absence of either receptor antagonist. Data points in A-C and bars in D-F represent the mean±SEM from three or more independent experiments (error bars for D-F can be found in S4 Fig). In B and C, data sets for H4 Env are significantly different (p < 0.01) from data sets for H5 and J5 Env as determined by two-way ANOVA. The H5 and J5 data sets were not statistically different. https://doi.org/10.1371/journal.ppat.1010531.g003

PLOS PATHOGENS
HIV-1 Env apex regulates gp41 MPER exposure chimeric N5 Env (Figs 4A and S1). N5 Env was significantly less sensitive to NAbs b12, F105, VRC01 and 17b, similar to what we had ascertained for H5 Env (Fig 4D and 4F-4H and S2  Table). However, the anticipated decrease in NAb 10E8 potency was not observed (Fig 4B).  Table listing the V3-loop strains of the tested Env variants. The Env HXB2 V3 loop was substituted into Env BG505 (green), Env JRFL (red) and Env SF162 (orange), while the Env SF162 V3 loop was substituted into Env NL4-3 (blue). (B) IC50 values determined for anti-MPER NAb 10E8 against wild type (solid) and chimeric (hatched) Envs. (C) Dependence of C37-KYI, 5-Helix and NAb 10E8 inhibitory potencies on CoRA AMD3100 concentration for J4 Env, the Env JRFL V3-loop chimeric variant. Data are normalized to the IC50 values obtained in the absence of CoRA. Note that CoRA concentration is expressed relative to the AMD3100 IC50 for J4 Env (61 nM). (D-I) IC50 values determined for anti-bridging sheet NAb 17b (D), anti-V3-crown NAb 447-52D (E), anti-CD4 binding site NAbs b12(F), F105 (G) and VRC01 (H), and anti-glycan NAb 2G12 (I). In B and D-I, the solid and dashed black lines indicate the mean IC50 values measured against H4 and H5 Envs, respectively. Bars represent the mean ±SEM from three or more independent experiments. Arrows above the bars indicate that 50% neutralization was not achieved at the highest concentrations tested. Asterisks indicate significance as determined by a two-sample t-test (equal variance not assumed) with p < 0.05. In C, data points represent the mean ±SEM from four independent experiments; there was no significant difference between the three data sets (two-way ANOVA, p > 0.4). In E, "nd" denotes not done. https://doi.org/10.1371/journal.ppat.1010531.g004

V3-loop residue 306 is an important regulator of MPER exposure in Env HXB2 and Env NL4-3
Using site-directed mutagenesis, we discovered that residue 306 accounted for the differential impact the V3-loop swap had on MPER exposure in Env HXB2 and Env NL4-3 . This residue is an Arg in Env HXB2 , while it is a Ser in Env NL4-3 and Env SF162 (Fig 5A). Hence, residue 306 changed identity in the V3-loop swap that produced H5 Env but was preserved in the V3-loop swap that produced N5 Env. In H4 Env, the R306S substitution decreased NAb 10E8 neutralization potency by 16-fold, nearly the entire 23-fold difference between H4 and H5 Envs ( Fig  5B). By contrast, the substitution had almost no effect on NAb 10E8 inhibition of H5 Env. The box representation in Fig 5C (see legend) shows that NAb 10E8 resistance was achieved by either a V3-loop swap that maintained Arg at position 306 (top horizontal) or by an R306S substitution in the HXB2 V3-loop background (left vertical). Neither a V3-loop swap with Ser at position 306 (bottom horizontal) nor an R306S mutation in the SF162 V3-loop background (right vertical) significantly impacted NAb 10E8 potency. Together, the results implied that the R306S substitution worked in concert with the remainder of the HXB2 V3 loop to modify MPER accessibility, an effect that was blunted in context of the SF162 V3 loop (see Discussion).
Substituting Ser306 with an Arg enhanced N4 Env sensitivity to NAb 10E8 almost 4-fold, but had minimal impact on N5 Env sensitivity ( Fig 5D and S2 Table). In the box representation of Fig 5E, the diagonal represents the equivalent of the H4-to-H5 Env transformation: in the Env NL4-3 background, that transition results in a 5.7-fold increase in NAb 10E8 IC50. Again, major changes to NAb 10E8 potency occur along the top horizontal and left vertical, with the other transitions having only minor ramifications. The result matched what we observed for Env HXB2 , although the differences were quantitatively smaller. In contrast to the CXCR4-tropic Envs, the point mutation at residue 306 exerted no effect on antibody neutralization of either S4 or S5 Envs (Fig 5F and 5G and S2 Table). In other words, the increased NAb 10E8 potency against S4 Env occurred independent of the amino acid at residue 306. The same phenotype was observed in the Env JRFL background (S2 Table). The results implied that differences from the H4 and N4 Env sequences outside the V3 loop enabled S4 and J4 Envs to weather an Arg residue at position 306.

The R306S mutation destabilizes the MPER-accessible prefusogenic conformation of Env HXB2
The R306S substitution in H4 Env disrupted neutralization by anti-MPER NAbs 2F5 and 10E8 to the same extent without altering FI activity ( Fig 6A). The loss of antibody potency was not accompanied by an appreciable loss of binding affinity, as antibody neutralization remained irreversible (S2 Fig). Moreover, the point mutation altered the CoRA-dependence of NAb 10E8 neutralization, which now matched that for FI 5-Helix inhibition ( Fig 6B). The results suggested that the R306S substitution largely restricted anti-MPER antibody binding to a functional intermediate state, as had occurred with the V3-loop swap that produced H5 Env.
While the V3-loop swap imparted relative resistance to antibodies targeting epitopes exposed in more-activated Env conformations, the R306S substitution had a more varied

PLOS PATHOGENS
HIV-1 Env apex regulates gp41 MPER exposure (C) Box representation of the relative effects that residue 306 substitution (vertical) and V3-loop swap (horizontal) exert on the NAb 10E8 sensitivity of Env HXB2 . The numbers adjacent to the lines connecting vertical, horizontal and diagonal points indicate the fold-change in IC50 value for the indicated substitution. (D, E) As in B and C, except testing Env NL4-3 . The wild type Env is designated N4 (filled circles), while the V3-loop chimera is designated N5 (open circles). The S306R substitution was incorporated into both N4 and N5 Envs (filled and open triangles, respectively). (F, G) As in B and C, except testing Env SF162 . The wild type Env is designated S5 (filled circles), while the V3-loop chimera is designated S4 (open circles). The S306R substitution was incorporated S5 Env (filled triangles), while the R306S substitution was incorporated into S4 Env (open triangles). In each box representation, the wild type Env species is circled. https://doi.org/10.1371/journal.ppat.1010531.g005

PLOS PATHOGENS
HIV-1 Env apex regulates gp41 MPER exposure H4 R306S Env has an antibody sensitivity profile distinct from both wild type H4 Env and the chimeric H5 Env. (A) Anti-MPER antibody (2F5, 10E8) and FI (5-Helix, C37-KYI) neutralization potencies against H4 R306S Env. The solid and dashed gray lines reflect the average IC50 values obtained for H4 and H5 Envs, respectively. NAbs 2F5 and 10E8 were significantly less potent against H4 R306S Env than against wild type H4 Env; the antibody potencies against H4 R306S Env and H5 Env were not significantly different (one-way ANOVA with means comparison using the Tukey method, p < 0.025 for H4 / H4 R306S comparison [109]). (B) Impact of CoRA AMD3100 on the inhibitory potencies of C37-KYI (triangles), 5-Helix (squares) and NAb 10E8v4 (circles) against HIV-1 pseudotyped with H4 R306S Env. Data are normalized to the IC50 values obtained in the absence of CoRA. Note that CoRA concentration is expressed relative to the AMD3100 IC50 for H4 R306S Env (53 nM). The data sets for 5-Helix and 10E8v4 are not statistically different as assessed by two-way ANOVA (p = 0.9); the difference in data sets for C37-KYI and 10E8v4 does achieve significance (p = 0.022). (C) Anti-gp120 and anti-interface antibody neutralization potencies against H4 R306S Env. Arrows above the bars and dotted lines indicate that 50% neutralization was not achieved for H4 R306S and H5 Envs at the highest concentrations tested. Of these antibodies, only NAbs 447-52D, 35O22 and PG9 had significantly lower potencies in comparisons of H4 R306S and H4 Env (one-way ANOVA with means comparison using the Tukey method, p < 0.025). All data reflect the mean±SEM of three or more independent experiments. https://doi.org/10.1371/journal.ppat.1010531.g006

PLOS PATHOGENS
HIV-1 Env apex regulates gp41 MPER exposure impact ( Fig 6C). Like H5 Env, H4 R306S Env was significantly less sensitive to antibodies targeting the V3-loop crown (447-52D) and gp120-gp41 interface (35O22). However, H4 R306S Env remained mostly sensitive to antibodies targeting the CD4 binding site and CD4i-exposed bridging sheet. Moreover, the potency of NAb PG9, which preferentially binds the fully closed, apo-conformation of Env, was not enhanced by the R306S substitution, as it was by the V3-loop swap. These results indicated that the R306S substitution and V3-loop swap had different effects on the energy landscape of prefusogenic Env conformational fluctuations.
The varied impact of the R306S substitution also extended to functional characteristics regarding CD4 binding versus utilization for entry. The sensitivity of H4 R306S Env to sCD4 inhibition was much closer to that of wild type H4 Env than H5 Env (Fig 7A), suggesting that binding was not particularly affected by the point mutation. By contrast, the sensitivity of H4 R306S Env to CD4-antagonist D23.2 more closely resembled that for H5 Env than H4 Env (Fig 7B), indicating that the point mutation had a significant impact on CD4 utilization during entry. This shift in CD4 utilization was not accompanied by a significant change in CoR utilization, as wild type and mutant H4 Envs were similarly sensitive to CoRA AMD3100 ( Fig 7C). Thus, despite only minimally perturbing CD4 interaction with the Env trimer, the R306S substitution dramatically altered CD4 binding requirements for viral entry.
This dichotomy in structural and functional consequences of the R306S substitution was also reflected in the results from Env-stability measurements. The R306S variant had lower spontaneous gp120 shedding than wild type Env and was relatively resistant to sCD4-induced shedding; in fact, the data closely mimicked that for the stable H5 Env variant (Fig 7D). Surprisingly, the R306S substitution had no appreciable impact on cold inactivation, with the mutant variant sensitivity overlapping that for the relatively unstable wild-type H4 Env ( Fig  7E). These observations illuminated the different energy landscapes of H4, H4 R306S and H5 Envs and, additionally, showed that the two measures of Env stability report on independent biochemical processes.
To summarize, H4 R306S Env adopted some characteristics of H5 Env while retaining properties of its H4 Env background ( Fig 7F). Specifically, the R306S variant acquired resistance to antibodies targeting the MPER, gp120-gp41 interface and V3-loop crown, increased its requirement for CD4 to promote viral entry, and decreased its sensitivity to sCD4-induced shedding. By contrast, H4 R306S Env maintained (in)sensitivity to antibodies targeting several gp120 conformational epitopes, including the V1/V2 loop, CD4 binding site and CD4iexposed bridging sheet. In addition, the point mutation only minimally perturbed CD4 binding affinity and sensitivity to cold denaturation. These changes were quite distinct from the impact of the V3-loop swap, which appeared to stabilize the glycoprotein trimer in its closed conformation and to increase the CD4 requirement for activation. The sensitivity of H4 R306S Env to antibodies targeting activated conformations (e.g., F105, 17b) implied that the mutant trimer could still readily fluctuate into open structures, but not to ones that allowed easy access to gp41 epitopes. Curiously, exposure of the V3 crown in H4 R306S Env appeared to occur in these later MPER-accessible states, even though earlier fluctuations effectively exposed the bridging sheet that lies underneath the V3 loop.

Towards a mechanism of how residue 306 modulates MPER accessibility
Currently, only CCR5-tropic Envs with a Ser at position 306 have been structurally characterized to high resolution. Prior to CD4 activation, the amino acid is located on the lateral face of the V3 loop exposed to aqueous solution between gp120 subunits (Fig 8A and 8B).  CoRA AMD3100 (C) against HIV-1 pseudotyped with H4 Env (filled circles), H5 Env (open circles) and H4 R306S Env (triangles). Of these three inhibitors, only D23.2 showed significantly different potencies when comparing H4 R306S to H4 Env (one-way ANOVA with means comparison using the Tukey method, p < 0.025). (D) Soluble CD4 dependence to gp120 shedding from H4, H4 R306S and H5 Env-expressing cells. Shed gp120 was detected by SDS-PAGE/Western blot, quantified by densitometry and normalized to the total gp120 detected in cell lysates. In the images of representative Western blots (right), L denotes a 1:3 dilution of cell lysate; the lines to the right of the images indicate positions of 250, 150 and 100 kDa molecular weight markers. The H4 R306S Env data set is significantly different from the H4 Env data set (two-way ANOVA, p < 0.01) but not statistically different from the H5 Env data set (p > 0.20). (E) Time dependence to cold inactivation of HIV-1 pseudotyped with H4, H4 R306S or H5 Env. The H4 R306S data set is significantly different from the H5 Env data set (two-way ANOVA, p < 0.01) but not statistically different from the H4 Env data set (p > 0.40). Data points in A-E represent the mean±SEM of three or more independent experiments. (F) Sensitivity profile of H4 R306S Env scaled to the differences between H4 and H5 Env sensitivities. For each inhibitor, the IC50 value for H4 R306S Env was transformed according to the following formula: For gp120 shedding ("shedding"), the sCD4 concentrations that elicited a normalized band intensity of 0.16 (estimated from the data lines in D) were used in place of IC50 values in the equation above. For cold denaturation ("cold"), the infectivity fractions at 96 hours were used in place of IC50 values. The scaled sequenced HIV-1 Envs (Los Alamos National Laboratories HIV Sequence Database, January 2022, 6223 total Envs), but conservation significantly varies by clade (Figs 8C and S6B). Arg or Lys is found at that position with an overall frequency of 4.11%, but, again, this number shows high variability across clades (Figs 8C and S6B). When Arg306 was substituted with a Lys in H4 Env, the IC50 value for NAb 10E8 remained low (Fig 8D). By contrast, when Arg306 was replaced with either a neutral or acidic amino acid, NAb 10E8 IC50 values increased 9.5-fold or more, with the R306E variant showing the largest change (19-fold). These changes in NAb 10E8 sensitivity were independent of the side chain at residue 318, indicating that an interaction between residues 306 and 318 does not determine MPER exposure (S6C Fig). Based on these results and on the location of Ser306 in available Env structures, we hypothesized that the residue at position 306 exerted its effects on Env conformational equilibrium via through-space electrostatic interactions. To test this idea, we introduced charge-altering mutations at Arg306 and two spatially adjacent charged residues, Glu172 and Lys305 (Fig 8B). We reasoned that NAb 10E8 potency should track with net charge on this amino-acid triad if control of MPER exposure were mediated by long-range electrostatics that contributed to subunit-subunit interactions. Wild-type H4 Env has a net triad charge of +1 and high sensitivity to NAb 10E8 neutralization. However, the configuration of the individual charges could be rearranged to produce an Env variant with a triad charge of +1 that was relatively resistant to NAb 10E8 inhibition (Fig 8E). Similarly, Envs with net triad charges of 0 or -1 manifested both low and high NAb 10E8 sensitivity. These observations were inconsistent with long-range electrostatics between subunits governing MPER accessibility.
The dependence of NAb 10E8 sensitivity on triad charge configuration suggested that MPER exposure could depend on intra-subunit interactions mediated by the three apical residues. In the prefusogenic structure of clade B Env JRFL [79], the side chains of Glu172 and Lys305 appear to form an ionic bond that bridges the V1/V2 and V3 loops (Fig 8B). Of sequenced clade B isolate Envs, almost 78% have an acidic amino acid at position 172 and a basic amino acid at position 305 (Figs 8C and S6B). We systematically evaluated the role of this putative ionic bond in H4 Env by substituting Glu172 with either an Ala or a Lys in the context of an Arg, Ser or Glu at residue 306. The ionic interaction did not play a role in MPER exposure when residue 306 was charged: NAb 10E8 inhibited all Arg306 variants with high potency and all Glu306 variants with low potency (Fig 8F). However, in context of Ser at position 306, MPER occlusion was only observed with an intact 172-305 ionic bond, while Env HXB2 variants with a E172A, E172K, or K305A mutation were highly susceptible to NAb 10E8 neutralization (Fig 8F).
The observations above suggested that an electrostatic interaction between V1/V2 and V3 loops somehow controlled exposure of the MPER epitope. Such an interaction could potentially restrict access to the V3-loop crown and explain the correlation in potencies of antibodies targeting these spatially disparate epitopes. To explore this possibility, we compared the impact of disrupting the putative 172-305 ionic bond on the potencies of NAbs 10E8 and 447-52D. In the context of a Ser at position 306, the E172A mutation significantly enhanced susceptibility to both NAb 10E8 and NAb 447-52D neutralization (Fig 8G and 8H). By contrast, the E172A mutation exerted no effect on the low neutralization potency of either antibody when position value provides the relative similarity of H4 R306S Env to either H4 Env (scaled sensitivity = 0) or H5 Env (scaled sensitivity = 1, dotted line). Data in F were calculated using average IC50 (or other) values with errors formally propagated. Red symbols designate that the IC50 value for H5 Env (+), H4 R306S Env (#) or both ("^") exceeded what we could measure. In these situations, the scaled value shown is an overestimate (+), underestimate (#) or merely an estimate (^) of the true value. https://doi.org/10.1371/journal.ppat.1010531.g007

PLOS PATHOGENS
HIV-1 Env apex regulates gp41 MPER exposure 306 was Glu. The results supported an electrostatic interaction between V1/V2 and V3 loops that governed exposure of both the V3-loop crown and MPER epitope in Env HXB2 .
The potential for an ionic bond between residues 172-305 is not well conserved across HIV-1 clades (S6D Fig). However, the two positions show a high degree of coevolution, indicating energetic coupling between the amino acids [80]. Interestingly, in Env BG505 , which lacks an acidic residue at position 172, Lys305 appears to make a cation-π interaction with aromatic V2-loop reside Tyr173 (S6E Fig). Disruptions to this interaction have been shown to increase susceptibility to anti-gp120 antibody neutralization in several Env strains [61]. In fact, aromatic residues are highly conserved at position 173, which also strongly coevolves with position 305 [80]. The potential for this cation-π interaction remains high in most HIV-1 clades (S6F Fig). These findings suggest that an interaction between V1/V2 and V3 loops adjacent to position 306 is a common feature of Env. Based on our observations in H4 Env, we speculate that the charge on residue 306 modulates the strength of this loop-loop interaction, which, in turn, determines the significance of disrupting the 172-305 ionic bond or 173-305 cation-π interaction in regulating MPER accessibility (see Discussion).

Discussion
As one of two major regions involved in interprotomer interactions, the apex of HIV-1 Env has long been recognized as a key regulator of prefusogenic glycoprotein structure and dynamics. Destabilizing mutations in the V1/V2 and V3 loops of primary isolate Envs have been shown before to enhance exposure of neutralizing epitopes in both gp120 and gp41 [48,53,[57][58][59][60][61][62]. In this study, we have accomplished the reverse in whole or in part: swapping the Env SF162 V3 loop into the laboratory-adapted isolate Env HXB2 bestowed tier 2-like, primary isolate neutralization sensitivity to H5 Env, while introducing a point mutation at position 306 in the Env HXB2 V3 loop selectively conferred resistance to NAbs targeting the gp120 V3-loop crown and gp41 MPER. The apparent structural disparity in H4, H4 R306S and H5 Envs implies differential stabilization of states along the activation pathway that governs prefusogenic Env dynamics.
Historically, the overall stability of HIV-1 Env has been characterized through temperature-induced inactivation and gp120 shedding. Typically, the two measures show a high degree of correlation: Envs with low sensitivity to cold-or hot-induced inactivation are most often resistant to gp120 shedding, both spontaneous and CD4-induced [22,43,44,47,55,56]. Such Envs are relatively insensitive to broad spectrum antibody neutralization, require higher levels of cellular CD4 to trigger membrane fusion and mediate very little CD4-independent entry [22,39,43,45,46]. These properties are thought to reflect the high stability of a closed apo-structure that energetically limits fluctuations along the activation pathway in the absence of receptor binding [35,36,41]. High stability is a characteristic feature of Envs from primary isolate HIV-1 strains, likely selected in order to limit neutralizing immune responses. the LANL HIV Sequence Database using the Analyze Align tool. (D) NAb 10E8 potency against HIV-1 pseudotyped with mutant H4 Env variants containing substitutions at Arg306. The asterisk indicates that the 10E8 IC50 values of all H4 Env variants except the R306K mutant are significantly higher than that for wild type H4 Env (one-way ANOVA with means comparisons using the Tukey method, p < 0.001 [109]). (E, F) NAb 10E8 potency against HIV-1 pseudotyped with mutant H4 Env variants containing charge substitutions at triad residues Glu172, Lys305 and Arg306. Amino acids corresponding to the "-", "0" and "+" designations along the x-axis are as follows: Glu (-), Ala (0) and Lys (+) for residues 172 and 305; Glu (-), Ser (0) and Arg (+) for residue 306. In E, mutant Envs are grouped according to the total triad charge. In F, mutant Envs are grouped according to the charge at residue 306. (G, H) Inhibitory titrations of NAbs 10E8 (G) and 447-52D (H) against wild type H4 Env (gray filled circles) and variants H4 R306S (gray triangles), H5 (gray open circles), E172A/R306S (black diamonds) and E172A/R306E (black stars). Data bars and points represent the mean±SEM of three or more independent experiments, except for NAb 447-52D titrations of H4 E172A/R306S and H4 E172A/R306E Envs in H, where the data reflect the mean±range of mean of two independent experiments. https://doi.org/10.1371/journal.ppat.1010531.g008

PLOS PATHOGENS
HIV-1 Env apex regulates gp41 MPER exposure Laboratory adaptation of HIV-1 results in Envs of lower stability with less-pronounced energetic differences between apo-and receptor-induced conformations [35]. The resulting increase in receptor-independent conformational fluctuations along the activation pathway leads to enhanced exposure of neutralizing epitopes, increased sensitivity to temperature-induced inactivation, and decreased requirement of CD4 to trigger viral entry or gp120 shedding [21,39,43,45]. These characteristics well describe the properties of the laboratory-adapted clade B isolate H4 Env, which has considerable sequence divergence from clinical clade B primary isolates (86% identity/5 gaps with Env JRFL ; 89% identity/7 gaps with clade B Env consensus sequence, Los Alamos National Laboratory HIV Sequence Database). However, the properties of H5 Env imply that laboratory adaptation has not fundamentally altered the physiochemical properties of Env HXB2 : in addition to its tier 2-like sensitivity to antibody and sCD4 neutralization, H5 Env displays characteristics of high stability Envs including resistance to both sCD4-induced gp120 shedding and cold inactivation and a requirement for high levels of CD4 on target cells for entry. Interestingly, the single point mutation R306S in the H4 Env V3 loop confers partial stability to the glycoprotein trimer. Like H5 and J5 Envs, H4 R306S Env is resistant to sCD4-induced gp120 shedding and requires high levels of CD4 on target cells for entry. However, H4 R306S Env has the same sensitivity to cold inactivation as wild type H4 Env. These results indicate that the two measures of Env stability reflect different elements of the prefusogenic energy landscape.
Despite its resistance to antibodies targeting the V3-loop crown and MPER, H4 R306S Env remains sensitive to antibodies that selectively bind activated Env conformations, including the bridging sheet that lies underneath the V3 loop. Differential exposure of these epitopes necessitates at least two post-CD4 binding Env conformations. Comparison of the neutralization sensitivities of H4, H4 R306S and H5 Env implies that the V3-loop crown / MPER accessible conformation is downstream of gp120 fluctuations that expose the bridging sheet. Given the temporospatial properties of MPER exposure, we conclude that these two states lie on the fusion pathway prior to the transition into the PHI. Finally, based on our manipulations of the charged triad at the Env HXB2 apex, we posit that fluctuation into the MPER accessible state is tied to the loss of an interaction between V1/V2 and V3 loops regulated by the charge at position 306 and supported by an ionic bond between V1/V2 residue 172 and V3 residue 305. These considerations lead us to propose a new state along the activation pathway in which the V1/V2 and V3 loops disengage from the gp120 core, remaining associated with each other but abrogating intersubunit interactions with loops on adjacent protomers (State C in Fig 9). This partially opened conformation exposes previously occluded gp120 regions (e.g., the bridging sheet) while still shielding the V3-loop crown and MPER. Ultimately, the trimer transitions into its fully open, V3-loop crown and MPER-accessible conformation through splaying of the V1/V2 and V3 loops on either side of the gp120 subunit as seen in structures of CD4-17b-gp120/gp41 and CD4-CCR5-gp120 [11][12][13]18] (State D in Fig 9). How V1/V2-and V3-loop interactions are tied to MPER occlusion is currently unknown, but the splaying might increase Env flexibility, enhancing the ability of the trimer to tilt relative to the membrane (depicted in Fig 9) as observed in recent cryo-EM structures of Env trimers in viral membranes [81] and in artificial lipid bilayers bound to a single 10E8 antibody [82].
As a working model, we consider the 4-state Env-activation pathway depicted in Fig 9. A fully closed, unliganded conformation (state A, heretofore not structurally elucidated) transitions into the closed state depicted in many SOSIP structures (state B) after a single CD4 binding event [36,37,83,84]. Binding of multiple CD4s induces transition first into State C, with the V1/V2 and V3 loops flipped up in tandem, and then into State D, with the V1/V2 and V3 loops splayed. Many anti-gp120 antibodies that bind activated conformations (e.g., NAbs F105 and 17b) can interact with Env in states C and D, while anti-V3-crown and anti-MPER antibodies can only bind to state D.
We predict that state A is much more stable than the downstream states in H5 Env, as it appears to be in tier 2/3 primary isolates (Fig 9B) [35-37,47]. Effective transition for such Envs into states C and D would require stabilization from CD4 binding, causing the bridging sheet and MPER epitopes to be exposed only transiently during membrane fusion. Consequently, neutralization by NAbs 17b, 447-52D and 10E8 is kinetically restricted like bona fide PHI fusion inhibitors, accounting for the relatively poor potencies for these antibodies and the strict CoRA-dependence to NAb 10E8 inhibitory activity. By contrast, the energy differences between state A and downstream states in H4 Env are considerably lower, enabling frequent CD4-independent fluctuations into states C and D (Fig 9B). As a result, NAbs 17b, 447-52D and 10E8 have a chance to engage Env outside the context of target-cell interaction, thereby increasing their neutralization potencies and decreasing the CoRA-dependence to NAb 10E8 inhibitory activity. In our model, the R306S mutation in H4 Env primarily impacts the energy difference between states C and D (when bonds between the V1/V2 and V3 loops are broken), leaving the energy differences between states A, B and C largely unchanged (Fig 9B). Because CD4 binding is required to sufficiently expose the gp41 MPER, NAb 10E8 neutralization of H4 R306S Env becomes kinetically restricted (H5 Env-like) while neutralization by anti-gp120 antibodies, including 17b, is mostly unaltered (H4-Env like).
Our analysis clarifies why the CD4-dependence to viral entry (including Hill coefficient, S4  Fig) correlates with gp120 shedding but not necessarily with cold inactivation. According to our model, both H5 and H4 R306S Envs require higher CD4 binding stoichiometry compared to H4 Env in order to achieve the stabilization of state D needed for efficient CoR-triggered irreversible transition into the PHI (red line in Fig 9B). Our data suggest that gp120 shedding reports on this required CD4 binding stoichiometry. By contrast, cold inactivation seems to reflect the stability of state A relative to states B and C. Thus, CD4-induced gp120 shedding appears to follow on-pathway conformational changes through (at least) state D, whereas cold inactivation diverges from the activation pathway somewhere between states A and C.
Our results indicate that the Env HXB2 V3 loop is primarily responsible for the laboratoryadapted phenotype of H4 Env. The interprotomer electrostatic repulsion resulting from the additional six positive charges per V3 loop (compared to the Env SF162 V3 loop) is likely very destabilizing, especially since these cations are concentrated in the gp120-gp120 interface at the apex of Env. However, introduction of the Env HXB2 V3 loop into Env JRFL had only a modest effect on antibody neutralization and no effect on the CoRA-dependence to MPER exposure. It is clear that Env JRFL regions outside of the V3 loop function to blunt the impact of apical destabilization. Curiously, swapping the Env HXB2 V3 loop into Env BG505 did produce a substantial increase in global epitope exposure that rivaled and even exceeded changes observed for the V3-loop swap in Env HXB2 (with the notable exception of NAb VRC01, which has the ability to interact with the CD4-binding site in multiple states [78]). Ongoing work is focused on dissecting the differences between primary isolate Envs that make them more or less susceptible to apical disruptions. Our attention here on Env HXB2 , however, highlights the utility of using less stable laboratory-adapted isolates in parsing out more nuanced aspects of Env conformational dynamics that might not be as easily elucidated in more stable primary isolate trimers.

Molecular biology
All env genes were amplified by PCR from their parental plasmids and cloned into the Xho1 and Eag1 restriction sites of the pEBB expression plasmid [107] (oligos from Eurofins, PCR enzyme Pfu Turbo from Agilent, restriction and ligation enzymes from New England Biolabs). V3-loop swaps were performed as previously described [63]. Briefly, PCR amplification of the Env HXB2 or Env SF162 V3 loop using overhang primers produced short dsDNA segments that were subsequently used as primers in QuikChange PCR (Agilent) to introduce the desired changes. Point mutations in the Env sequences were achieved through QuikChange PCR using overlapping oligonucleotides (Eurofins and IDT). All Env constructs were confirmed through Sanger sequencing of the entire open-reading frame prior to use (Genewiz).

Peptide preparation
Preparation of C-peptide C37-KYI was performed as previously described [65]. Briefly, HIV-1 gp41 TOH construct NC1 containing N637K and T639I substitutions was expressed and purified from bacterial lysates using metal-chelate chromatography (Qiagen). Eluate was proteolyzed with trypsin (Sigma-Aldrich), and C37-KYI was purified to homogeneity by reverse phase HPLC using a Vydac C18 column and linear gradient of acetonitrile in water containing 0.1% trifluoroacetic acid. C37-KYI was lyophilized for storage and resuspended in water for use. The identity of the peptide was confirmed by mass spectrometry.
The 5-Helix protein was solubilized from bacterial inclusion bodies using 8M guanidine HCl in Tris-buffered saline (TBS) [65]. 5-Helix was recovered using metal-chelate chromatography and renatured in 4M guanidine HCl by a reverse thermal gradient (90˚C to room temperature over 4hrs). Properly folded monomeric species were separated from protein aggregates by size exclusion chromatography (Superdex 75, GE Healthcare). The identity of the protein was confirmed by SDS-PAGE (molecular weight 25 kDa) and HIV-1 inhibition (activity) assays.
The expression plasmid for CD4-antagonist D23.2 (kindly provided by Dr. Alexandra Trkola, University of Zürich) was transformed into XL1 Blue cells. A 200mL starter culture of XL1 cells in LB supplemented with 1% (w/v) D-Glucose and 100 μg/ml carbenicillin (Fisher) was grown overnight and used to inoculate 1-2 L of LB plus 1% (w/v) D-Glucose and 100 μg/ ml carbenicillin. When the culture achieved an OD 600 = 0.9-1.1, expression of the DARPin was induced with 1 mM IPTG (final concentration) and the culture was left to grow for 4 hours at 37˚C. Bacteria were isolated by centrifugation and resuspended in PBS500 buffer [1xPBS solution (Fisher) with an additional 500mM NaCl and 20 mM Imidazole, pH 7.2] supplemented with 100μg/mL DNaseI, 2mM PMSF and 2 μg/mL each of Leupeptin and Pepstatin (all chemicals from Sigma-Aldrich). Bacteria were lysed using a French press (500psi) and the lysate was diluted 1:2 with PBS500 containing 8 M urea. The solution was centrifuged (Sorvall SS-34) at 17,000 x g for 20 minutes at 4˚C. Supernatant was applied to Ni-NTA agarose (Qiagen), and protein was renatured by reverse urea gradient over 20 column volumes and ultimately eluted in PBS500 containing 250 mM imidazole. Eluate was concentrated and subjected to size exclusion chromatography (Superdex 75) in 1xPBS to separate monomers from aggregates. D23.2 was concentrated using a 9 kDa MWCO Amicon concentrator to greater than 100 μM, flash frozen in liquid nitrogen and stored at -80˚C until use. The identity of the protein was confirmed by SDS-PAGE (molecular weight 14.7 kDa) and HIV-1 inhibition (activity) assays. The concentrations of D23.2, 5-Helix and C37-KYI were determined at 280nm using the method of Edelhoch [108].
Neutralization potencies were determined by serial dilution of antibodies and inhibitors. Each condition was tested in duplicate or triplicate in each experiment. Averaged data were normalized to the infection level measured in the absence of inhibitors and fit to the Langmuir equation ) unless otherwise specified. For experiments exploring the CoRA-dependence to inhibitory potency, titrations of antibody and FIs were prepared in solutions containing CoRA at defined concentrations. Experiments were repeated three or more times using at least two different production batches of the same virus (unless otherwise indicated).
For the HIV-1 preincubation/dilution experiments in S2 Fig, virus was suspended in media with and without NAb 10E8v4 at 2-4x IC50 concentration to form full-strength samples. A portion of these samples (± antibody) was diluted 10-fold either immediately or after a 1-hour incubation at 37˚C. The sample diluted at 1 hour was further incubated at 37˚C for 3 hours to allow for antibody unbinding and recovery from neutralization. Culture media over target cells was completely replaced with full strength or diluted samples after 0, 1 and 4 hour incubations and viral infectivity was measured by luminescence 48 hours later.
sCD4-induced gp120 shedding assay HEK293T cells (5 x 10 6 ) were transfected with 2.5 μg each of pEBB_Env X and pCMV_Rev using the PEI method described above and incubated for 48 hours. Cells were lifted with EDTA (0.5 mM in 1xPBS) and extensively washed with 1xPBS before being resuspended in 350 μL of 1xPBS containing 2 mg/ml BSA (Sigma-Aldrich), 1 μM PMSF (Sigma-Aldrich), and Complete protease inhibitor (Roche). Aliquots (50 μL) of the cell suspensions were supplemented with sCD4 (50 μL) at various concentrations and incubated for 1 hour at 37˚C. Cells were pelleted by gentle centrifugation (10 minutes at 300 x g), and the clarified supernatant was subjected to SDS-PAGE (10% acrylamide, BioRad). Separated proteins were transferred to nitrocellulose paper (Amersham, GE) and probed with sheep anti-gp120 antiserum followed by HRP-conjugated donkey anti-sheep antibody (Jackson Immuno). Protein bands were visualized using ECL detection reagent (Thermo Fisher), imaged using a ChemiDoc system (BioRad), and quantified by densitometry using Adobe Photoshop. Band intensities of shed gp120 were normalized to the intensity of gp120 in a lysate (100 μL of assay buffer supplemented with 2% Triton X-100) of an equal aliquot of untreated cells.

Cold inactivation
HIV-1 samples were prepared and harvested as described above, and aliquots (200 μL) were immediately flash frozen in liquid nitrogen before being stored at -80˚C. At designated timepoints between 96 hours and 1 hour prior to infection, viral aliquots were removed from the -80˚C storage, quickly thawed at 37˚C for 1 min, and transferred to ice until being applied to target cells. At time of infection, aliquots on ice were collectively warmed to 37˚C for 1 min before addition to U87 cells. Viral infection was measured 48 hours later as described above. For each experiment, data were normalized to the infection level of the 1-hour timepoint.
Supporting information S1 Table. IC50 values (μg/ml) for different NAbs grouped by tier designation. Numbers reflect the geometric means obtained from viruses with both detected and undetected IC50 values. Orange designates geometric means calculated from groups in which less than 85% of the IC50 values were measurable. Data were procured from the Los Alamos National Laboratory HIV Immunology database (January 2022). For comparison, the mean IC50 values for H4, H5 and H4 R306S Envs found in this study are shown. Red numbers indicate that 50% neutralization was not achieved at the highest concentrations tested. (DOCX) S2 Table. IC50 values for different NAbs and sCD4 against Envs with V3-loop swaps and residue 306 substitutions. Numbers reflect the mean and SEM from three or more independent experiments. The units are ng/ml for all NAbs and nM for sCD4. Red numbers indicate that 50% neutralization was not achieved at the highest concentrations tested. Blank spaces indicate that the antibody or sCD4 was not tested against that Env. Note that B4 R306S Env was nonfunctional. (DOCX)

S1 Fig. Relative infectivity of HIV-1 pseudotyped with different wild type and variant Env species. (A)
For each virus, infectivity was measured on both U87.CD4.CXCR4 cells (grey bars) and U87.CD4.CCR5 (magenta bars) and normalized to viral content estimated by luciferase expression from proviral DNA in virus-producing cells (see B). Normalized infectivity is reported relative to the values for H4 Env on CXCR4 + cells and H5 Env on CCR5 + cells. Data represent the mean±range-of-mean from two independent experiments. (B) Relationship between luciferase expression from virus producing cells and HIV-1 p24 levels found in culture supernatants. For each experiment, cells were transfected (Lipofectamine Plus, Life Sciences) with differing amounts of pseudotyping vector pNL4-3 E -R -Luc + , Env-expressing vector pEBB_HXB2Env and GFP-expressing vector pBABE_GFP as indicated in the legend. After a 36 hour incubation, supernatants were collected and assayed for viral p24 content by ELISA (see reference [63] of text) while cells were lysed and assessed for luciferase activity. For each experiment, the measured solution p24 levels and luciferase activities were normalized to their respective values obtained from the 0.5 μg pNL4-3 E -R -Luc + transfection. Data from four separate experiments have been fit to a linear regression fixed to go through the origin; the slope was 0.96 and the R-value was 0.93. HIV-1 was suspended in media with and without NAb 10E8v4 at 2-4x IC50 concentration to form the full-strength samples. A portion of these samples (± antibody) was diluted 10-fold either immediately or after a 1-hour incubation at 37˚C. The sample diluted at 1 hour was further incubated at 37˚C for 3 hours to allow for antibody unbinding and recovery from neutralization. Culture media over U87 target cells was completely replaced with full strength or diluted samples at the 0-, 1-and 4-hour time points (red circles) to asses viral infectivity in the presence or absence of antibody. (B) Representative data from a single experiment using HIV-1 pseudotyped with H4 Env. Infectivity was measured by luciferase reporter expression for full strength (FS) or diluted (Dil) samples at the indicated time points. Bars and error bars represent the mean ± SEM of three experimental replicates for each sample. (C-F) NAb 10E8v4 neutralization of HIV-1 pseudotyped with H4 (C), H4 R306S (D), H5 (E) and J5 (F) Envs. Fraction infectivity was calculated as the ratio of the luciferase signals recorded in the presence and absence of NAb 10E8v4 (i.e., adjacent magenta and cyan bars in B) under each condition (FS or Dil) at the indicated time point. Antibody concentrations (ng/mL) in the FS and Dil conditions, as well as the IC50 value, are listed in the adjacent box. Data represent the mean ± SEM from four to five independent experiments. The lower fraction infectivity at the 1-hour timepoint compared to the 0-hour timepoint (more easily seen in the Dil samples) indicated that antibody binding to the prefusogenic state occurred during the 0-to-1 hour incubation period under FS conditions. Neutralization was enhanced by approximately the same amount for all Envs tested despite the fact that NAb 10E8 concentrations in H4 R306S, H5 and J5 Env viral samples were approximately 10-to 20-fold higher than that in wild type H4 Env samples. This finding suggested that the antibody association rate constant was much greater for H4 Env than for the other Env species. The lower fraction infectivity in the Dil samples at the 4-hour timepoint compared to the 1-hour timepoint implied that antibody dissociation was slow compared to 10E8-induced Env inactivation (see reference [22] of main text). The functional irreversibility of NAb 10E8 neutralization ruled out decreased binding strength (manifest in increased dissociation rate) as the mechanism behind the reduced antibody potency caused by the V3-loop swap or R306S substitution in Env HXB2 . Together, the results implied that disruption of anti-MPER antibody neutralization arose from a reduced association rate constant, consistent with a decreased lifetime of an MPER-accessible Env conformation. (TIF) S3 Fig. Heat inactivation of HIV-1 pseudotyped with H4, H5 or J5 Env. Viral aliquots were incubated at various temperatures for 45 minutes prior to addition to U87 cell cultures at 37˚C. Infectivity levels were normalized to the level measured following a 37˚C incubation. Data represent the mean±range-of-mean from two independent experiments. (TIF)