Structural delineation of a quaternary, cleavage-dependent epitope at the gp41-gp120 interface on intact HIV-1 Env trimers

Abstract

All previously characterized broadly neutralizing antibodies to the HIV-1 envelope glycoprotein (Env) target one of four major sites of vulnerability. Here, we define and structurally characterize a unique epitope on Env that is recognized by a recently discovered family of human monoclonal antibodies (PGT151-PGT158). The PGT151 epitope is comprised of residues and glycans at the interface of gp41 and gp120 within a single protomer and glycans from both subunits of a second protomer and represents a neutralizing epitope that is dependent on both gp120 and gp41. Because PGT151 binds only to properly formed, cleaved trimers, this distinctive property, and its ability to stabilize Env trimers, has enabled the successful purification of mature, cleaved Env trimers from the cell surface as a complex with PGT151. Here we compare the structural and functional properties of membrane-extracted Env trimers from several clades with those of the soluble, cleaved SOSIP gp140 trimer.

Figure 1. PGT151 and PGT152 binding to SOSIP gp140 Env trimer is dependent on glycosylation and cleavage and requires the pre-fusion conformation of gp41

(A) PGT151 binds strongly to BG505 SOSIP.664 gp140 trimers produced in untreated HEK 293T cells but not to the same trimers produced in HEK 293S cells (Man₅-₉ glycans only) or in HEK 293T cells treated with kifunensine (kif) (Man₉ glycans only). PGT151 also does not bind BG505 gp120 monomers. 2G12, which binds high mannose glycans, served as a control MAb. 2G12 binding to trimers produced in kifunensine-treated 293T cells or in 293S cells, was modestly increased as a result of the elevated amount of high mannose glycans. The data were derived using a Ni-NTA ELISA and His-tagged Env proteins. Data are representative of at least 2 independent experiments. (B) In an ITC assay, PGT151 and PGT152 Fabs bind with low affinity (K_D ≈ 1 mM) to complex type asialo-tri-antennary glycans. The raw binding (top) and integrated titration (bottom) curves are shown. (C) PGT151 binds His-tagged BG505 SOSIP.664 and BG505 IP.664, but not to BG505 WT.664 as shown by Ni-NTA-capture ELISA. Binding of the anti-gp120 MAbs VRC01 and 2G12 to the BG505 IP.664 and BG505 WT.664 variants is greatly reduced (left). The anti-gp41 MAbs F240 and 7B2 do not bind to SOSIP.664 trimers because their epitopes on gp41 are occluded by the gp120 subunits, but they do bind to IP.664 and WT.664 proteins from which gp120 has dissociated (middle). PGT151 binds the SOSIP.664 trimers and the IP.664 proteins, but does not bind cleaved WT.664 (right). These data suggest that the PGT151 epitope is only present in the pre-fusion conformation of gp41. Data presented here were chosen from 2-6 independent experiments. Each experiment was done in duplicate and averaged values are shown.

Figure 2. Crystal structures of PGT151 and PGT152 Fab and use of PGT151 and PGT145 to purify and characterize JR-FL and BG505 EnvΔCT trimers from cell membranes

(A) Variable domain from the crystal structure of PGT151 Fab. The HCDR loops 1-3 are labeled and colored in blue, red and magenta, respectively, while the LCDR loops 1-3 are labeled and colored in green, yellow and cyan, respectively. The long 26-residue CDR3 towers above the combining site. (B) Superposition of crystal structures of PGT151 Fab (cyan) and PGT152 Fab (orange). See also Table 1 for data statistics. (C) Scheme for purification of JR-FL EnvΔCT from membranes using PGT151. (D) SEC profile of the purified JR-FL EnvΔCT: PGT151 complex; fractions containing the Fab-trimer complex are labeled in blue. (E) SDS PAGE (reducing and non-reducing) and BN PAGE analysis of the purified JR-FL EnvΔCT: PGT151 complex. (F) Western blot analysis of cell surface expressed BG505 EnvΔCT and JR-FL EnvΔCT trimers, after incubation with PGT151, cell lysis and fractionation on a reducing SDS-PAGE. Gp120 was detected on a western blot to assess the amount of Env captured by PGT151. The cell lysate and protein A flow through (FT) fractions are compared (from step 3 in the purification scheme (C)). Band intensities were analyzed by densitometry in the panels below the gel lanes (absolute intensities: black numbers, relative percentages: blue numbers). Overall, ∼10% of the total gp120 detected (cell surface and intracellular) was captured on protein A as a PGT151-trimer complex, while the remaining ∼90% was in the FT fractions and, hence, non-reactive to PGT151. (G) Blue native PAGE analysis showing the relative stability of JR-FL EnvΔCT or BG505 EnvΔCT complexes with either PGT151 Fab or PGT145 Fab after incubation at 4°C. From left to right: JR-FL EnvΔCT: PGT151 is stable for 40 days (40d) with no dissociation; JR-FL EnvΔCT: PGT145 partly dissociates after 7 days (7d) and is fully dissociated after 40 days (40d); BG505 EnvΔCT: PGT151 is stable for 30 days (30d), whereas BG505 EnvΔCT: PGT145 is partly dissociated. (H) BN-PAGE assessment of the stability of JR-FL EnvΔCT in complex with either PGT151 Fab or PGT145 Fab after incubation at different temperatures for 1h (left and middle). In a densitometric analysis, the intensities of the trimer bands are normalized against the total Env protein content of each lane (right).

Figure 3. Membrane-derived, PGT151-purified Env trimers from clades A, B and C all share similar structural features with soluble, cleaved SOSIP trimers and are bound by PGT151 in a sub-stoichiometric manner

(A-C) Top (left) and side views (right) of three different EnvΔCT: PGT151 Fab EM reconstructions, showing two Fabs bound per trimer. From left to right, the Env trimers are: (A) BG505 (clade A), (B) JR-FL (clade B), and (C) IAVI C22 (clade C). (D) For comparison, top and side views are shown for the BG505 SOSIP.664: PGT151 Fab complex reconstruction. (E) Comparison of the PGT151-liganded JR-FL EnvΔCT trimer reconstruction (pink (same as B)) with the cryo-electron tomographic reconstruction of the unliganded BaL-1 virion-associated Env spike (EMDB ID: 5019) on the viral surface (gray). See also Figure S4 and Table S1. (F) ITC analysis of high affinity PGT151 binding to BG505 SOSIP.664 trimers. The affinity and stoichiometry were calculated from an average of two experiments. The N=1.3 value is higher than the previously reported 0stoichiometry of N=0.6-0.8 for one PG9 binding to the BG505 SOSIP.664 trimer (Julien et al., 2013b) and lower than the N= 2.3-2.4 for PGT121, PGT128 and 2G12 binding to the same trimer (Sanders et al, 2013). The inference is that ∼2 PGT151 Fabs are bound per trimer. Data analysis was carried out using a single-site binding model. The ITC data do not fit a 2-site model, which argues against a cooperative binding modality. The K_D was determined as an average from 3 measurements. Binding affinity was further assessed by bio-layer interferometry in comparison with two other bnAbs (Fig. S1C) (G) Densitometric analysis of a reducing SDS PAGE of JR-FL EnvΔCT trimers purified from cell membranes using either PGT151, PGT128 or PGT145. The black numbers denote the respective peak intensities. The blue numbers in brackets indicate the relative intensity of the respective peak compared to the corresponding peak on the PGT151 panel. While the intensities of the gp120 bands are comparable (0.94-1.0) in each case, the relative intensities of the Fab bands are 0.6 and 1.5 for the PGT145 and PGT128 purifications respectively, indicating that the PGT151 stoichiometry lies between PGT145 (which binds with a stoichiometry of 1 per trimer (McLellan et al, 2011; Sanders et al, 2013; Walker et al, 2011)) and PGT128 (which binds with a stoichiometry of 3 per trimer (Pejchal et al., 2011)). Note that PGT128 only allowed for extraction of a very small amount of Env, which was only sufficient for SDS-PAGE. Neither the yield nor the purity of the sample was in the range of that achieved with PGT151 or PGT145.

Figure 4. The PGT151 epitope on BG505 SOSIP.664 trimer consists of residues from gp141 and gp120

(A) PGT151 binding, assessed by Ni-NTA ELISA, is reduced when His₆-tagged BG505 SOSIP.664 variants contain certain substitutions in the C5 region of gp120. In the control, 2G12 binding is only mildly affected by the same substitutions. (B) PGT151 binding, assessed by ELISA, is reduced when BG505 SOSIP.664 variants contain certain substitutions in gp41. Point substitutions of N611 and N637 that eliminate glycan sites are shown in the lower panel. For comparison, most substitutions had no effect on 2G12 binding, although the K601A and N607K changes did cause a modest reduction. Assays were done using His- or D7324-tagged BG505 SOSIP proteins (as indicated in the figure). (C) Summary of substitutions in BG505 SOSIP.664 trimers that affect PGT151 binding. Binding ability is scored on a scale from (−, no binding) to (++++, strong binding). The analysis is derived from mean values of 3-5 experiments for each of the variants. Color-code: red: weak to no binding, yellow/orange: moderate binding, green: strong binding. Trimerization and cleavage were comparable to WT SOSIP as assessed by SDS-PAGE and BN-PAGE (data not shown). We cannot exclude the possibility that mutations proximal to the SOS disulfide bond (residues 501 and 605) change the local stability of the trimer and, therefore, binding by PGT151. Both direct and indirect effects are consistent with PGT151 interactions at the gp120-gp41 interface.

Figure 5. Modeling of the PGT151 epitope in the context of the high-resolution EM structure of the soluble, cleaved BG505 SOSIP.664 trimer

(A) Top and side view of the reconstruction of the JR-FL EnvΔCT: PGT151 Fab complex, fitted with the atomic-level EM model of the BG505 SOSIP.664 trimer (PDB ID: 3J5M) (Lyumkis et al., 2013). The gp120 protomers are in gray and the gp41 HR1 and HR2 helices in purple. On the 2 copies of PGT151 Fab, the heavy chain (HC) is in dark blue and the light chain (LC) in light blue. (B) Top view as in (A). The PGT151 LC is predicted to interact with glycans from two gp120 protomers (labeled gp120 A and B), and amino-acid residues from one gp120 protomer. The putative interacting glycans are depicted as spheres in shades of yellow, and the glycosylation site position (Asn) is shown in orange. LCDR coloring is shown as in Figure 2A. (C) The same view as in (B), showing regions of C1 that could potentially interact with LCDR1. In light pink are residues 56-63, and dark pink 75-82. (D) In a side view as in (A), putative PGT151 HC interactions with one gp120 protomer (gray) and two gp41 protomers (purple) are shown. The map shows a single gp120 and the gp41 trimer, segmented from the high-resolution EM structure 3J5M. The HR helices of gp41 B are depicted as purple ribbons. HCDR3 appears to interact with gp140 protomer A at the gp41-gp120 interface, near the fusion peptide proximal region (FPPR) shown in brown. The N637 glycosylation site from protomer B shown in orange is in close proximity to the PGT151 HC. (E) The glycan at residue N637 is modeled as a tetra-antennary GalNac complex glycan that was shown to interact with all members of the PGT151 family (Falkowska et al., 2014). The EM density in cyan, not modeled in the high-resolution structures, is predicted to encompass the putative location of the HR2 residues 611-624 as depicted by the dotted line. The orange circle and arrow respectively indicate the putative position and orientation of N611 and its glycan based on the neutralization assay. See also Figure S2, and movie S1.

Figure 6. PGT151 binds to a new site of vulnerability that does not overlap with any other bnAb epitopes

The JR-FL EnvΔCT: PGT151 Fab reconstruction of the new PGT151 epitope compared with the docking of representative Fabs that recognize three of the four previously known bNAb epitope clusters (V1V2 plus N160 glycan, N332 glycan, CD4 binding site). We cannot exclude the possibility that access to the CD4bs might be blocked partly by the two bound PGT151 Fabs; accordingly, PGV04 binding to the CD4bs is shown on the third protomer interface that remains does not have PGT151 bound. The PGT151 binding site on the open face is also shown in blue for reference, with the PGT151 Fab density removed for clarity (right panel). A similar analysis of BG505 SOSIP.664 trimers produces nearly identical results. See also Figures S3 and S4.