Sub-Nanometer Structures of HIV-1 Envelope Trimers on Aldrithiol-2 Inactivated Virus Particles

Abstract

The HIV-1 envelope glycoprotein (Env) trimer, comprised of gp120 and gp41 subunits, mediates entry into cells. Recombinant Env trimers have been studied structurally, but characterization of Env embedded in intact virus membranes has been limited to low resolution. Here, we deploy cryo-electron tomography and sub-tomogram averaging to determine the structures of Env trimers on aldrithiol-2 (AT-2)-inactivated virions in ligand-free, antibody-bound, and CD4-bound forms at sub-nanometer resolution. Tomographic reconstructions document molecular features consistent with high resolution structures of engineered soluble and detergent-solubilized Env trimers. One of three conformational states predicted by smFRET was not observed, potentially due to AT-2 inactivation. We did observe Env trimers to open in situ in response to CD4 with an outward movement of gp120-variable loops and an extension of a critical gp41 helix. Overall features of Env trimer embedded in AT-2 treated virions appear well-represented by current engineered trimers.

The HIV-1 envelope protein (Env) is a type 1 viral membrane fusion machine that mediates virus entry into cells while evading the humoral immune response [reviewed in^1,2]. HIV-1 Env comprises a trimer of non-covalently associated gp120 exterior and gp41 transmembrane subunits that is embedded in the envelope membrane of virions. Interaction with the CD4 receptor induces gp120 structural rearrangements, which expose a coreceptor-binding site^3–5. Subsequent interactions with a coreceptor, either CCR5 or CCRX4, bring the Env trimer closer to the target membrane⁶ and are thought to induce the formation of a pre-hairpin intermediate of gp41 in which the fusion peptide of gp41 contacts the target cell. Downstream conformational rearrangements form the stable six-helix bundle of the post-fusion state^7,8, which drives fusion between viral and cellular membranes, allowing viral entry and initiation of infection.

High-resolution structures of ligand-free⁹, antibody-bound^10–13, and CD4-bound prefusion HIV-1 Env trimers^3,14–16 have been obtained for soluble gp140 SOSIP.664 trimers (sgp140 SOSIP.664), containing a disulfide bridge between gp120 and gp41 (SOS)¹⁷, an I559P substitution (IP)¹⁸, and the ectodomain truncation at residue 664 (.664)¹⁹. Structures have also been obtained with other types of Env trimer stabilization^20–23, some involving linked gp120 and gp41 subunits^24–26, as well as from detergent-solubilized trimers containing both membrane-proximal external region (MPER) and transmembrane (TM) region^27–29. These structures have provided considerable structural detail on the Env trimer and its entry-related conformational changes. In the ligand-free prefusion state, the V1V2 loops interact at the trimer apex⁹, wherein the coreceptor binding site in V3 loop is hidden beneath the V1V2 loops. Upon CD4 binding, the gp120 subunits rotate, the V1V2 loops move outwards from the apex to flank the CD4 receptor, and the V3 loops are released, primed to bind coreceptor^{3,4,6,14–16,30}. Trimer opening in response to CD4 likely proceeds through asymmetric intermediates^16,31–33.

Structural characterization of the conformations of Env trimers on the surface of intact virus, however, has been limited to low resolution^34–36. Insights into the conformational states adopted by HIV-1 Env on the surface of viruses, and the rates of exchange between them, have also been gained from single-molecule Forster resonance energy transfer (smFRET) experiments^37–39. smFRET has revealed three prevalent conformational states, a pre-triggered State 1, an intermediate State 2, and a fully CD4-engaged State 3. Comparison of the states observed by smFRET with extant Env structures has revealed that known high-resolution structures resemble States 2 and 3³⁷. A high-resolution structure of State 1 has not yet been determined.

As many broadly neutralizing antibodies exhibit a preference for the pre-triggered State 1 conformation^37,39, its structure is sought as a basis for vaccine design and to understand more fully the structural rearrangements HIV-1 Env undergoes. Based on recent investigations of Env conformations in solution and on the surface of intact virus³⁷, we reasoned that the pre-triggered State 1 conformation of the Env trimer may be present at higher occupancy on the surface of intact virus particles and applied cryo-electron tomography (cryoET) and sub-tomogram averaging to visualize Env trimers on HIV-1_BaL virions. Viruses were obligatorily treated with aldrithiol-2 (AT-2) to eliminate retroviral infectivity by covalent modification of internal viral proteins, including the nucleocapsid protein, while preserving overall Env structure and functional fusion competence⁴⁰. Tomographic reconstructions were obtained in ligand-free, antibody-bound and CD4- and 17b-bound states. Comparison of the conformational change from the closed to CD4-bound conformation was consistent with the outward movement of the V1V2 loops on gp120 and an extension of the gp41 HR1-C helix. These studies reveal the State 1 conformation of Env to be even more context-specific than previously anticipated and demonstrate that engineered recombinant soluble Env trimers accurately capture State 2 and 3 features of virus-embedded HIV-1 Env trimers.

Results

Ligand-free HIV-1 Env structure by cryoET and sub-tomogram averaging

CryoET combined with sub-tomogram averaging has been used to determine native structure of ligand-free native HIV-1 and SIV Env on virions (Extended Data Fig. 1a–c)^34–36. However, the resolution of these maps has been insufficient to resolve individual gp120 and gp41 subunits, domain features, or α-helices. To determine higher-resolution structures of the virus-embedded HIV-1 Env trimer, we used a Titan Krios with direct electron detector, energy filter, and Volta phase plate (VPP) to visualize the same AT-2-inactivated virus preparations of the HIV-1_BaL isolate previously studied at lower resolution^35,41,42. The AT-2 treated HIV-1_BaL virions produced from infected cell cultures incorporated full-length Env, including the transmembrane-spanning region and cytoplasmic tail (CT) domains, and retained functional CD4-dependent fusion capacity, with dual CCR5/CXCR4 tropism⁴⁰. From 683 virus reconstructions, we extracted 13,577 Env trimer particles for sub-tomogram averaging (Fig. 1a, b and Table 1) and determined the HIV-1 Env structure at 11.5 Å resolution estimated by map-model Fourier Shell Correlation (FSC) and local resolutions ranging from 7–14.5 Å estimated by Resmap (Table 1 and Extended Data Fig. 2).

Figure 1. In situ cryoET structure of ligand-free HIV-1 Env trimer.

(a) A slice through a representative tomogram of AT-2-treated HIV-1 virons. The scale bar is 100 nm. (b) A zoom-in view of one virion. Note that Env trimers on virus surface are readily visible. The scale bar is 50 nm. (c, d) Two perpendicular slices of a sub-tomogram average of the HIV-1 Env trimer. The scale bar is 5 nm. (e) Segmentation of the Env trimer is shown in front and top views, respectively. The gp120 subunit is colored in light blue, the gp41 ectodomain in orange, the MPER in purple, and the membrane in gray. (f) Fitted model of gp120 (blue) and gp41(orange) subunits from the crystal structure of ligand-free BG505 SOSIP.664 (PDB 4ZMJ).

Table.

CryoET data used in this study

HIV-1BaL	Number of Tomograms	Number of Virions	Number of Trimers
Ligand-free	70	683	13,577
10–1074 & 3BNC117	85	713	15,243
sCD4 & 17b	208	959	13,817

The in situ structure of the ligand-free HIV-1_BaL Env trimer adopted a closed conformation measuring 105 Å in height and 132 Å in width (Fig. 1c, d). The gp120 subunit was 78 Å in height, the ectodomain of gp41 57 Å in height, and the MPER of gp41 12 Å in height (Fig. 1e). Interactions between the neighboring protomers were concentrated at the apex of gp120 and within gp41. Notably, several secondary structure densities were visible in the cryoET map (Extended Data Fig. 3). Fitting the atomic structure of ligand-free BG505 SOSIP.664 (PDB 4ZMJ)⁹ into the cryoET map revealed characteristic features including the α1-helix of gp120, the gp41-interactive region of gp120, three helical bundles of densities corresponding to the HR1-C helix around the central axis of the spike, and density for a partial HR1-N helix at the exterior base of the ectodomain (Fig. 1e, Extended Data Fig. 3). gp41 showed clear electron density in the reconstruction that continued into the virion envelope membrane. Rod-shape densities from individual gp41 molecules bundled together, while HR1-N density appeared to interact with the HR2 helix from a neighboring gp41 subunit. Density corresponding to the MPER of gp41 was also observed in the cryoET map, and these appeared to insert obliquely in the virion envelope membrane, displaying a “twisting tripod” configuration (Fig. 1e and Extended Data Fig. 1d).

The in situ structure also revealed electron density corresponding to variable loop features and glycosylation (Extended Data Fig. 4). In the region corresponding to the V1, additional residues specific to the BaL sequence were apparent in the electron density. The other four variable loops in the BaL structure were observed as continuous stretches of electron density (Fig. 1f), with V2/V3 loops forming the trimer apex and V4/V5 loops residing at the spike periphery.

Antibody-bound Env trimer on virus resembles State 2 Env trimer

smFRET indicates that most known broadly neutralizing antibodies exhibit a preference for the State 1 conformation^37,39. We asked if two antibodies with State 1 preferences³⁷, 10–1074 recognizing the glycan N332 at the V3 loop glycan supersite and 3BNC117 recognizing the CD4-binding site, could stabilize a different conformational state of Env on AT-2 treated HIV-1_BaL. We produced antigen-binding fragments (Fabs) for both antibodies, complexed them with AT-2 treated HIV-1_BaL, and collected cryoET data as shown in Fig. 2a. From 713 virus reconstructions (Fig. 2b), 15,243 Env trimers were visually identified and extracted for sub-tomogram analysis. Classification was first used to remove Env trimers with poor contrast and Env trimers without Fabs. The Env trimers complexed with Fabs were aligned and averaged to generate a 10–1074 and 3BNC117-bound HIV-1 Env structure (Fig. 2c, d). The resolution of the structure was estimated by map-model FSC at 9.9 Å (Table 1, Extended Data Fig. 2). Both Fabs could be readily visiable in the cryoET map (Fig. 2c, d). The angle of approach for each of the two antibodies were similar in high-resolution structures in complex with engineered Env trimer^43–48, including the soluble BG505 SOSIP.664, indicating this structure to also be representative of State 2 (Extended Data Fig. 5)^37,39.

Figure 2. CryoET structure of HIV-1 Env with antibodies 10–1074 and 3BNC117 at sub-nanometer resolution reveals domain features of in situ trimer.

(a) A slice through a representative reconstruction of HIV-1 viruses in complex with 10–1074 and 3BNC117 Fabs. The scale bar is 100 nm. (b) A zoom-in view of one virion with Env trimers. The scale bar is 50 nm. (c) A central slice of a sub-tomogram average of the Env trimer bound with 10–1074 and 3BNC117. The scale bar is 5 nm. (d) A segmentation of the antibody-bound Env trimer shows 10–1074 (green), 3BNC117 (cyan), gp120 (blue) and gp41 (orange) subunits. (e) Enlarged views of gp41-interactive region of gp120 and gp41 (upper panel) with residue-level model fitting in the density map (Bottom panel). (f, g) Side and bottom views of gp41 at two different contour levels (0.156 and 0.121), respectively. The gp41 model from PGT151-bound JR-FL Env trimer (PDB 5FUU) fitted into the segmented maps, respectively (bottom); gp41 colored in orange, with the C-terminal region of the HR1-N highlighted in watermelon red.

The gp41-interactive region of gp120 could be seen branching into a fork, surrounded by the gp41-helical base (Fig. 2e). The C1 region of gp120, located close to the HR1-C region of gp41 in the BG505 SOSIP.664 structure (PDB 4TVP), fitted into the cryoET map well. By comparison, the C5 region of gp120 in the BG505 SOSIP.664 structure, which is located near to the HR1-N loop region and also proximal to the fusion peptide, was slightly stretched out and did not fit into the cryoET density well (Fig. 2e). These observations suggests that the C5 region in HIV-1_BaL Env trimers adopts a conformation different from extant high-resolution structures of the engineered Env trimers, perhaps related to opening of gp41 as indicated by recent DEER measurements⁴⁹ or to an altered conformation incompatible with SOS linkage in this region. However, at the current cryoET resolution, we could not build a residue-level model to ascertain the reason for this misplacement.

In SOSIP-stabilized Env trimers, the I559P substitution within the HR1-N loop stabilizes formation of the prefusion trimer, but most atomic models lack flanking densities. Intriguingly, the structure of PGT151 bound JR-FL ΔCT shows the intrinsic helical propensity of HR1-N (PDB 5FUU) with GS-linked BG505 having a more flexible loop (PDB 6B0N)^25,27. In the ligand-free cryoET map, except for the helical end (residues 527–543), the density for a HR1-N helix was not resolved (Fig. 1f). However, in the antibody-bound Env trimer map, three symmetric rod densities could be clearly distinguished between adjacent protomers albeit at a higher threshold of 0.156, spatially matching the HR1-N residue-level model from unbound protomer of the asymmetric PGT151-bound JR-FL ΔCT (Fig. 2f). However, at a lower threshold of 0.121, the HR1-N extended continuously from the gp41 base through the pivoting helix bundle (Fig. 2g). The HR1-N of gp41 appeared to twist around the central helical bundle, with extension of the helix sterically suppressed by overlaying gp120 subunits (Fig. 2f, g).

AT-2 treatment shifts HIV-1 Env from State 1 to State 2

The HIV-1_BaL viruses used in this study were generated from a chronically infected SUPT1-CCR5 cell clone and inactivated with AT-2 so that these highly concentrated virus preparations could be safely used^41,42. AT-2 reacts with the zinc finger cysteines of nucleocapsid to inactivate virus; Env on the surface of AT-2 inactivated virus, however, remains fusion-competent^41,42. Because AT-2 treatment was a key difference between viruses prepared for smFRET imaging versus cryoET, we asked if such treatment could possibly cause conformational effects. We incubated HIV-1_BG505, HIV-1_NL4–3, and HIV-1_JR-Fl viruses labeled with smFRET tags with 1 mM AT-2 for 1 h followed by 3–12 h incubation on ice, matching conditions used to inactivate the virus preparations for cryoET (see methods). Strikingly, smFRET imaging revealed that AT-2 treatment resulted in a profound conformational shift from State 1 to State 2 (Fig. 3).

Figure 3. AT-2-treatment increases the prevalence of State 2 HIV-1 Env on virus.

(a, b, c) AT-2 treatment shifts the State 1-predominance of the conformational landscape of ligand-free HIV-1_NL4–3 (a: left panel), HIV-1_BG505 (b: left panel), and HIV-1_JR-FL (c: left panel) towards State 2. FRET histograms of ligand-free (left panels) and AT-2 treated (middle panels), and the corresponding quantification of relative conformational states occupancy (right panels) for neutralization-sensitive lab-adapted isolate HIV-1_NL4–3 (a), and neutralization-resistant primary isolates HIV-1_BG505 (b) as well as HIV-1_JR-FL (c), respectively. FRET histograms compiled by a number (n) of individual FRET traces represent mean ± s.e.m., from three independent groups of smFRET traces. FRET histograms overview conformational profiles of native Env trimers in the context of intact viruses, while bar-graphs quantify the relative occupancy for each state.

In this context, we found that the binding of antibodies 10–1074 and 3BNC117, which have a preference for State 1 of Env on untreated virus (Extended Data Fig. 6)³⁷, was unable to shift the conformational landscape back to State 1. Env lacking the CT also showed a conformational shift in response to AT-2 treatment, suggesting the conformational impact of AT-2 to reside in the ectodomain (Extended Data Fig. 7). These data indicate Env on AT-2-treated virus to reside predominantly in State 2, explaining why high-resolution Env structures fit well into the lower resolution ~20 Å cryoET maps, and now fit well into the sub-nanometer resolution maps that we described here.

Structure of CD4- and 17b-bound Env at sub-nm resolution

As AT-2 treatment does leave Env functional viable^41,42, we employed cryoET and sub-tomogram averaging to explore in situ how HIV-1_BaL Env changes conformations in response to soluble receptor CD4 (sCD4_D1D2; 2-domain CD4) and coreceptor mimicking antibody 17b (Fig. 4a, b). We complexed virus-embedded Env with sCD4_D1D2 and Fab 17b and collected cryoET data. A total of 959 virus reconstructions were generated and 13,817 Env trimer particles on virus surface (Fig. 4a, b) were extracted for sub-tomogram averaging (Table 1). Classification was then used to remove particles with poor contrast and particles without CD4 or 17b. The final in situ structure of the HIV-1_BaL Env complexed with sCD4_D1D2 and Fab 17b was determined at 9.7 Å resolution (Fig. 4c–e, Table 1, Extended Data Fig. 2). Major components of the structure including gp120, gp41, sCD4_D1D2 and Fab 17b were well resolved in the cryoET map (Fig. 4c, d). Their densities also matched well with the residue-level structure (PDB 5VN3) of sCD4–17b bound B41 Env trimer³ (Fig. 4f).

Figure 4. In situ cryoET structure at 9.7 Å resolution of sCD4_D1D2- and 17b-bound Env trimer.

(a) A slice through a representative tomogram shows multiple virions. The scale bar is 100 nm. (b) A zoom-in view of one virion with Env spikes. The scale bar is 50 nm. (c, d) Sub-tomogram average shows sCD4_D1D2- and 17b-bound Env trimer from side and top views, respectively. The scale bar is 10 nm. (e) Segmentation of the Env trimer shows sCD4_D1D2 (purple), 17b (green), gp120 (blue) and gp41(orange) subunits in both side and top views. (f) Atomic models of gp120, gp41, 17b and sCD4 fit well into the map (PDB 5VN3). (g) V1V2 loop fitted from sCD4(D1D2), 21C-bound, and 8ANC195-bound B41 SOSIP.664 (PDB 6EDU).

Notably, the gp41 subunits exhibited three central helices and surrounding rod-shape densities, which were not completely consistent with the structure of B41 SOSIP.664 in complex with CD4 and 17b (Fig. 4e, f). We therefore compared structures of the gp41 subunits in ligand-free Env, 10–1074 and 3BNC117-bound, and sCD4_D1D2 −17b bound Env trimers in detail (Fig. 5a–c). The two pre-CD4 binding conformations were similar with density for the HR1-N projecting towards the HR1 loop as shown in the two-antibody stabilized Env (Figs. 2f, g, 5a, b). Upon interaction with sCD4_D1D2 and 17b, the trimer apex unraveled, with V1V2 loops undergoing large structural rearrangements and the HR1-C of gp41 to extend by approximately 12 Å (Fig. 5f). The HR1-N helices appeared to be relieved from the interactions with the adjacent HR2 and moved further from the trimer axis (Fig. 5c bottom view).

Figure 5. The central helical density of gp41 in situ resembles that observed in high-resolution SOSIP structures.

(a) gp41 subunit of ligand-free BaL HIV-1 Env. (b) gp41 subunit of 10–1074 and 3BNC117 bound Env. (c) sCD4–17b bound BaL HIV-1 Env. (d) sCD4–17b bound B41 SOSIP.664 (EMD 8713) and (e) sCD4-E51 bound BG505 SOSIP.664 (EMD 20605). The high-resolution single particle cryoEM maps were low-pass filtered to 10Å. Segments of gp41 are depicted as below, in the cryoET maps, HR1-N and fusion peptide (FP) are colored in red, HR1-C of gp41 in orange; for both reported EM maps, HR1-N densities are colored in pink, HR1-C are colored in yellow. The base portion of all the gp41s are colored in orange. (f) Each corresponding atomic structure of HR1-C helices was fitted into the maps that were viewed from the top and side views.

To understand the in situ structure of the sCD4_D1D2 and 17b bound Env, we compared it to the B41 SOSIP.664 in complex with sCD4 and 17b (PDB 5VN3)^3,16 at ~10 Å resolution. The central HR1-C helices all appeared similar (Fig. 5f), with the I559P substitution designed by Sanders and Moore¹⁸ ideally positioned to allow the CD4-induced extension of the HR1-C helix on the surface of virions. We also observed similar densities for HR1-N and fusion peptide (FP), but the orientation of HR1-N connecting FP was different in the cryoET map from the HR1-N in the sCD4- and 17b-bound B41 SOSIP.664 trimer. However, the separation between HR1-N and HR2 was wider in the in situ trimer, suggesting an opening or loosening of the gp41 base in preparation for conformational transitions that promote membrane fusion (Fig. 5c–e).

To better understand the conformations of HR1 and HR2, we further compared our cryoEM map with extant high-resolution structures at the gp41 base (Fig. 6). In the cryoET map, the HR1-N and FP were not well resolved, resulting in the V-shaped densities (Fig. 6a). By contrast, the HR1 helix lies at different positions relative to HR2 in single particle cryoEM maps (Fig. 6b–e): HR1-N from sCD4- and 17b-bound B41 SOSIP.664 adopts a horizontal conformation at the base of the trimer (Fig. 6b); and in the 8ANC195 bound-trimer, both HR1-N and HR2 helix are embedded in the interior of the envelope trimer (Fig. 6c). Intriguingly, the asymmetric protomers of the sCD4- and E51-bound trimer¹⁶ displayed features that more closely resembled those observed in the cryoET map (Fig. 6d, e). We docked both residue-level models into the cryoET map, and observed HR1-N and HR2 regions from each protomer to display discrepancies: FP exhibited an unstructured loop; and the distance of HR1-N and HR2 appeared smaller, similar to the sCD4–17b bound B41 model (Fig. 6i, j). There were also similarities: HR2 fitted well into the cryoET map, and HR1-N and FP curved into “V-shape” density (Fig. 6i, j). Consequently, the sCD4- and 17b-bound in situ Env trimer described here adopted a more open conformation at the base of gp41 than previously observed in some of the high-resolution sCD4-bound engineered Env structures, and was similar to the more open structure observed in the asymmetric structures of sCD4-E51 bound trimers¹⁶.

Figure 6. sCD4- and 17b-bound Env trimer adopts a more open conformation in situ in the HR1-N and HR2 regions of gp41.

(a) One copy of HR1-N and adjacent HR2 are segmented in sCD4–17b bound Env map. HR1-N is colored in red and HR2 colored in orange. (b, c, d, e) Segmentation of HR1-N and HR2 from four different open-state structures, respectively. (b) sCD4–17b bound B41 SOSIP.664 (EMD 8713). (c) sCD4–17b-8ANC195 bound BG505 SOSIP.664 (EMD 7516). (d, e) Two classes of sCD4-E51 bound BG505 SOSIP.664 (EMD 20605 and EMD 20607). (f, g, h, i) Each of the four atomic models are docked into the cryoET map, respectively. HR1-N and HR2 of gp41 subunits are highlighted. HR1-N and HR2 are colored in green (f, g), in purple (h) and in magenta (i, j). (f) sCD4–17b bound B41 SOSIP.664 structure (PDB 5VN3). (g) A zoom view of panel f. (h) sCD4–17b-8ANC195 structure (PDB 6CM3). (i, j) Two structures of E51-sCD4 bound BG505 SOSIP.664 (PDB 6U0L and 6U0N).

Discussion

The unique “double lock” strategy of HIV-1 Env entry, employing conformational changes induced by CD4 and co-receptor to foil host defenses⁵⁰, has been the subject of investigation for over 20 years. CryoET and smFRET have provided important insights into the nature and timing of these critical conformational processes. In 2008, cryoET revealed the architecture of Env trimers for ligand-free, antibody-bound, and CD4- and 17b-bound conformations on the surface of BaL virions at ~20 Å resolution³⁵. In 2014, smFRET delineated at least three prevalent prefusion conformations of the Env trimer on the surface of virus particles^37–39 and in 2019 revealed that the pretriggered structure of the State 1 conformation had not yet been determined³⁷. Here we took advantage of VPP-enhanced contrast and ligand-assisted image alignment to study the in situ structure of AT-2-inactivated Env trimer by cryoET and sub-tomogram averaging. In so doing, domain features on virus-embedded Env trimers and conformational changes induced by CD4 binding could be visualized at sub-nanometer resolution.

In the ligand-free Env trimers on the surface of HIV-1_BaL virions, we were able to delineate not only gp120 and gp41 subunits, but also density corresponding to variable loop features and N-linked glycans. Density corresponding to the gp41-interactive region of gp120 clearly extended to the membrane-proximal bottom of the ectodomain, clamped by the helix-rich region of gp41. Furthermore, from in situ complexes with antibodies 10–1074 and 3BNC117 we could observe alterations in conformation of the gp120 extension in the C5 region (β-strand) when compared to high-resolution antibody-bound structures of BG505 SOSIP.664 (e.g. PDB 4TVP). We also observed density for the HR1-N region between neighboring protomers. Whether ligand-free or in complex with antibodies, the prefusion-closed Env trimers appeared to be similar to existing State 2 structures derived from soluble SOSIP.664 engineered proteins and detergent-solubilized recombinant Env proteins.

To assess the discrepancy between the cryoET and smFRET data, we evaluated the conformational impact of AT-2 inactivation of HIV-1_BaL, performed to make them safe for handling in cryoEM facilities. Surprisingly, smFRET indicated that AT-2 treatment resulted in a marked shift in Env conformation from State 1 to State 2, an effect that could not be reversed by antibodies that preferentially stabilize State 1 on untreated viruses. These findings indicate global agreement between cryoET and smFRET data and that AT-2 treatment of virions directly or indirectly alters the energy landscape of the Env trimer to reduce State 1 occupancy.

Quantitative distinctions between cryoET and smFRET data are, however, observed that will require further, in-depth investigations. By smFRET, we observe AT-2 treatment does not quantitatively deplete State 1 and State 3 occupancy. By contrast, cryoET only identifies the State 2 conformation. The underlying reasons for this disparity may relate to the distinct sample preparations employed for the samples examined by cryoET and smFRET (BaL from chronic infected cell line versus HEK293, respectively), the site-specific labeling methods required for smFRET and/or distinctions in the data processing methods employed by cryoET and smFRET. The precise reason for the conformational impact of AT-2 inactivation on Env to depopulate State 1 to favor State 2 occupancy will also require further inquiry. AT-2 is a mild oxidizing agent that is thought to react with cysteine residues: cysteine residues are disulfide bonded in the Env ectodomain; free sulfhydryls are only found in the CT region. The observation that AT-2 treatment of virus-embedded Env lacking CT showed conformational shifts similar to those observed for the native Env trimer suggests a CT region-independent impacts.

Soluble CD4 and coreceptor-mimicking antibody 17b binding to the membrane embedded Env on the surface of viruses induced structural rearrangements within the gp120 subunits, where the variable loops that normally interact to form the apex of the closed trimer unravel and rotate to its exterior. Again, current high-resolution structures of sCD4-bound SOSIP.664 trimers generally fit well into the cryoET map. Particularly striking, the HR1-C helix was observed to extend in situ in response to CD4 binding up to the helix-breaking I559P substitution¹⁸, emphasizing the precision to which this designed substitution was compatible with State 2 and 3 features of the native Env trimer. At the base of gp41, we also observed CD4 to disrupt the interaction between HR1-N and HR2, resulting in the in situ vertical displacement of the HR1-N. These alterations were echoed by the more recent asymmetric structures of the sCD4- and E51-bound Env trimer¹⁵, likley reflecting preparation for the proposed spring-loaded adaptation of a pre-hairpin intermediate to initiate the fusion reaction. The CD4-induced shedding observed in our samples is consistent with increased instability (Table 1), a conformational transition that is prevented by the structural restraint imposed by I559P substitution and SOS disulfide bridge in SOSIP.664 trimers or by antibody PGT151 in the antibody-stabilized detergent-solubilized Env trimer.

The present findings reveal several new challenges for understanding the structure and dynamics of the native Env trimer on intact virus. First, our results indicate that efforts should be made to find alternatives to AT-2 treatment as the method of choice for virus inactivation since it exhibits conformational effects precluding the capture of the State 1 structure that predominates on the surface of virus in the absence of chemical modification. Second, approaches should be advanced to characterize conformational intermediates in the fusion reaction in the absence of conformational constraints. Advances in this area may benefit from initiatives to embed CD4 and coreceptor in target membranes. Finally, focused efforts are needed to perform measurements on identical samples to gain further insights into the quantitative relationships between orthogonal structural methods employed to gain insights into fundamentally dynamic molecular systems.

Methods

Cell lines and viruses

HIV-1_BaL virus was produced by the HIV-1_BAL/SUPT1-CCR5 cell line. The cell line ID for it is “CLN204”. Its derivation was described previously⁵¹ and the virus that it produces is dual tropic. HIV-1_BaL CL.1 virus was produced by the HIV-1_BAL/A66-R5 CL.1 limiting dilution clone [CLN445]. It was derived by infecting A66-R5 cells⁵² with a HIV-1_BAL infectious stock obtained from the NIH AIDS Reagent Program (catalog # 510 – lot 100219) followed by preparation of limiting dilution clones. HIV-1 BaL CL.1 virus is R5 tropic.

Virus production and purification

Up to 30 L batches of cell culture propagated in roller bottles were serially filtered using 5.0 μm capsule filters to remove the cells followed by filtration using either 0.45 or 0.50 μm capsule filters to remove a fraction of the contaminating microvesicles⁵³. The filtrate was treated with a final concentration 1mM Aldrithol-2 (AT-2) to inactivate the virus^54,55 and the virus was purified by continuous flow sucrose density gradient ultracentrifugation⁵¹. HIV-1_BAL virus lot P4311 and HIV-1_BAL CL.1 virus lot P4338 were imaged by cryoET.

CryoET sample preparation

AT-2 treated BaL strain HIV-1 virus⁵⁶ was incubated for 30 minutes at room temperature with ligands 10–1074 and 3BNC117, or sCD4_D1D2 and 17b. 6 nm colloidal gold solution was added to the mixture and 4 μl of them were placed onto freshly glow discharged holey carbon grids for 1 minute. Grids were blotted with filter paper from the back, and rapidly frozen in liquid ethane using a homemade gravity-driven plunger apparatus.

CryoET data collection

The frozen-hydrated grids were imaged at −170 °C on a cryo-electron microscope (Titan Krios, Thermo Fisher Scientific), equipped with a field emission gun, an energy filter, Volta phase plate, and a direct detection device (K2 Summit, Gatan). Volta phase plate enhances the contrast of the tomogram, helping a better visualization of the Env on virus surface (Fig. 1a and b). The microscope was operated at 300 keV with a magnification of 64,000×, resulting in 2.245 Å/pixel at the specimen level. The tomographic package SerialEM⁵⁷ was utilized to collect 35 image stacks at a range of tilt angles from −51° to +51° with increments of 3° for each tilt series. The raw images were collected from single-axis tilt series at focus with accumulative dose of ~50 e⁻/Ų. For every single tilt series collection, the does-fractionated mode was used to generate 8–10 frames per projection image.

Image processing and tomogram reconstruction

Collected dose-fractionated data was subjected to the motion correction using Motioncorr2⁵⁸ to generate drift-corrected stack files, which were aligned using gold fiducial makers by IMOD^59,60. The reconstructions were generated by TOMO3D⁶¹. In total, 431 tomograms (3,838×3,710 ×1,600) were generated for HIV-1 viruses.

Sub-tomogram analysis

HIV-1_BaL virus particles and then Env trimers were manually picked and extracted from the tomograms. In total, 59,716 Env trimers were used for sub-tomogram averaging by using I3 (0.9.9.3)^62,63. First, 4 × 4 × 4 binned Sub-tomograms were subjected to “alignment by classification” method to generate initial low-resolution structure without using any initial model as reference^63,64. Second, we used unbinned sub-tomograms for further refinement. We explored the possibility of asymmetry (C1), but the trimers stayed at low resolution. Higher resolution in situ structures were determined when C3 symmetry was imposed. We used Resmap⁶⁵ to estimate the local resolution of cryoET maps.

Modeling and visualization

To obtain models for cryoET density maps at different states, we referred to several crystal structures and single particle cryoEM structures to individually generate gp120 and gp41 subunit model of BaL strain. Swiss model workspace⁶⁶ was used to generate homology models with BaL sequence. We used the ligand-free crystal structure of BG505 SOSIP.664 gp120 and gp41 subunits (PDB 4ZMJ) for the ligand-free maps; the PGT151-bound JR-FL ΔCT(PDB 5FUU) ligand-free gp41 atomic model for the 10–1074/3BNC117 cryoET map, and CD4–17b/21c-8ANC195 (PDB 6CE3 and 6EDU) for the CD4–17b bound cryoET map. Atomic structures of Fabs and ligand including 10–1074, 3BNC117, sCD4, 17b were docked into cryoET maps through the function “Fit in map” in UCSF Chimera⁶⁷, retaining the binding interface within the trimers and ligands. We also used Phenix package⁶⁸ to refine the models and to estimate the resolution of the cryoET maps by comparing with the fitted models. All the density maps are segmented in the UCSF Chimera⁶⁷, and ChimeraX⁶⁹ was used for surface rendering and visualization of cryoET maps and models.

Preparation of dye-labeled HIV-1 Env trimer on the virus for smFRET

The preparation of fluorescently labeled HIV-1_NL4–3, HIV-1_JR-FL (with and without cytoplasmic tail), and HIV-1_BG505 virus Env was as previously described^70,71. Two short enzymatic labeling peptides Q3 and A1 (Q3: GQQQLG and A1: GDSLDMLEWSLM) were inserted into variable loops V1 and V4 (V1V4) of Env gp120 subunit, respectively. These peptides allow Env to be site-specifically labeled with organic dyes through enzymatic reactions. HIV-1_NL4–3 virus carrying labeled Env was generated by transfecting 80% confluent HEK293 cells with a 40:1 ratio of replication-defective (RT-deleted) wild-type full-length HIV-1 NL4–3 plasmid to V1V4-tagged Env HIV-1 NL4–3 plasmid. HIV-1_BG505 virus Env was generated in a similar fashion in which plasmids encoding RT-deleted full-length HIV-1 Q23 BG505 (Q23 as the backbone) and V1V4-tagged Env HIV-1 Q23 BG505. HIV-1_JR-FL virus Env were produced by pseudo-typing a NL4–3 backbone (RT-deleted, Env-deleted) with JR-FL wild-type or cytoplastic tail-deleted (ΔCT) Env at a ratio of 40:1. Viruses were harvested 40 hours post-transfection from the culture supernatant, concentrated through a 15% sucrose cushion at 25,000 rpm for 2 hours, and then resuspended in 50 mM HEPES pH 7.5 buffer with 10 mM MgCl₂ and 10 mM CaCl₂. Virus suspensions were then incubated with customized conjugated organic dyes (0.5 μM Cy3B(3S)-cadaverine and 0.5 μM LD650-CoA, Lumidyne Technologies), and enzymes (5 μM AcpS and 0.65 μM transglutaminase, Sigma Aldrich) at room temperature overnight. The next day, viruses were first incubated with PEG2000-biotin (0.02 mg/ml, Avanti Polar Lipids) at room temperature for 30 min and then purified by density gradient centrifugation to remove free dyes.

smFRET imaging, data acquisition, and analysis

All smFRET experiments were performed on a customized prism-based total internal reflection fluorescence (TIRF) microscope, and data were analyzed using the SPARTAN software package⁷². Briefly, fluorescently labeled viruses were immobilized on passivated streptavidin-coated quartz slides. A continuous-wave 532-nm laser was equipped to excite donor (Cy3) fluorophores. Fluorescence signals of both donor (Cy3) and acceptor (Cy5) were collected by a 60 × (1.27-NA) water-immersion objective (Nikon), spectrally split, and then simultaneously recorded on two synchronized ORCA-Flash4.0v2 sCMOS cameras (Hamamatsu) at a speed of 40 milliseconds per frame. Each recorded FRET movie consists of 2000 frames that reflect the conformational motions of HIV-1 Env in real time. Viruses were imaged in Tris (pH 7.4, 50 mM) NaCl (50 mM) buffer, which contains a cocktail of triplet-state quenchers as well as oxygen scavengers consisting of protocatechuic acid (2 mM, PCA) and protocatechuic 3,4-dioxygenase (8 nM, PCD). When indicated, fluorescently labeled virus were pretreated by AT-2, matching the conditions used for the virus preparations used for cryoET. Specifically, viruses were incubated without and with 1 mM AT-2 at 37 °C for one hour, followed by 18 hours incubation on ice before imaging. For measurements of the conformational effects of antibodies, viruses were pre-incubated with 50 μg ml⁻¹ V3-glycan patch targeting bNAb 10–1074 (NIH AIDS reagent program) and 50 μg ml⁻¹ CD4 binding site-targeting 3BNC117 (NIH AIDS reagent program) for 30 mins at room temperature and remained present during the following imaging.

Real-time fluorescence intensities of both donor (I_D) and acceptor (I_A) fluorophores were then extracted from recorded FRET movies. FRET efficiency was calculated based on FRET= I_A/(γI_D+I_A), where γ is the correlation coefficient that corrects for the discrepancy between donor and acceptor in quantum yields and detection efficiencies. FRET time trajectories or FRET traces that pass strict criteria of sufficient signal-to-noise (S/N) ratio and negative cross-correlation between donor and acceptor intensity were then compiled into FRET histograms (probabilities vs. FRET values). FRET histograms were further fitted into a sum of three Gaussian distributions based upon the direct observation of state-to-state transitions in each of the FRET traces and the following idealization of those transitions by Hidden Markov Modeling. The relative occupancy of each conformational state was estimated as the area under each Gaussian curve, and the standard error was derived from thousands of individual data points associated with a number (n) of molecules in the FRET histogram.

Reporting Summary is available

Further information on experimental design is available in the Nature Research Reporting Summary linked to this article.

Data availability statement

In situ structures of HIV-1 Env were deposited to the EMDB with the following codes: ligand-free Env: EMD 21412, CD4–17b bound Env: EMD 21411, and 10–1074&3BNC117 bound Env: EMD 21413.

Reporting Summary statement

Further information on experimental design is available in the Nature Research Reporting Summary linked to this article.

Extended Data

Extended Data Fig. 1. Comparison of cryoET maps of the ligand-free HIV-1/SIV-1 Env trimer.

(a) In situ structure of Env trimer on the surface of SIV-1 mac329 virus with truncated cytoplasmic tail (EMD 1246). (b) Structure of Env trimer of SIV-1mneE11S virus (EMD 1216). (c) Structure of Env trimer of HIV-1_BaL virus (EMD 5019). (d) Structure of ligand-free HIV-1_BaL Env trimer (current work).

Extended Data Fig. 2. CryoET resolution estimations.

(a) Resolution estimation using FSC 0.5 as cutoff value. (b) Local resolution estimation of ligand-free, 10–1074 and 3BNC117 double antibodies, and sCD4–17b bound HIV-1_BaL Env trimer is estimated with Resmap. For each map, two projections are shown: Z plane projection is shown as a top view, Y plane projection is shown as a side view.

Extended Data Fig. 3. Model fitting.

Env trimers from other two strains, omitting the binding antibodies, are rigid body docked into the cryoET map. (a) 3H109L and 35O22 bound B41 SOSIP.664 (PDB 6MUF) (b) PGT122 and PGV19 bound BG505 NFL.664 (PDB 6B0N). (c) Segmented characteristic helical densities of gp41 subunits in the cryoET map. (d-f) Trimeric HR1C, partial HR1N and HR2 from 4ZMJ fit well into the rod-shape densities.

Extended Data Fig. 4. Variable loop regions and glycosylation sites on the top of the ligand-free Env trimer map.

(a) Variable loop regions on the top of the Env trimer are differently colored in the map. (b) Potential glycosylation densities at the apical surface of BaL strain virus. Ligand-free monomer of fully glycosylated JR-FLΔCT (PDB 5FUU) was docked in the cryoET map. The glycan moiety was shown as sticks.

Extended Data Fig. 5. Comparison between the cryoET model for HIV-1_BaL Env and BG505 SOSIP.664 in complex with different antibody combinations.

Each ligand and its binding protomer were projected into central plane and the dihedral angles were measured with UCSF Chimera. (a) CryoET model of 10-1074-3 and BNC117 bound Env. (b) 10–1074 bound BG505 SOSIP-based immunogen RC1 (PDB 6ORN). (c) 3BNC117 bound BG505 SOSIP.664 (PDB 5V8M). (d) 10–1074 and IOMA bound BG505 SOSIP.664 (PDB 5T3Z). All angular values are listed in panel (e).

Extended Data Fig. 6. 10–1074 and 3BNC117 cannot stabilize State 1 of AT-2 treated HIV-1_BaL Env.

(a, b) 10–1074 and 3BNC117 Fabs exhibit a preference for State 1 of virus Env in the absence of AT2-inactivation (a) Representative fluorescence (top, donor Cy3 in green and acceptor Cy5 in red) and FRET trace (bottom, resulting FRET in blue and hidden Markov model idealization in red) of individual V1V4 labeled ligand-free HIV-1_NL4–3 virus Env. (b) FRET histogram of HIV-1_NL4–3 Env in the presence of broadly neutralization antibodies 10–1074 (50 μg ml⁻¹) and 3BNC117 (50 μg ml⁻¹), overlaid with that of ligand-free HIV-1_NL4–3 Env. (c, d) 10–1074 and 3BNC117 are unable to restore State 1-predominance observed on native virus Env after AT-2 treatment shifted the conformational equilibrium towards State 2. (c, d) experiments as in (a, b) of AT-2 chemically inactivated HIV-1_NL4–3 virus Env, respectively. (e) Quantification of relative state occupancy of HIV-1_NL4–3 virus Env, derived from FRET histograms in (b, d).

Extended Data Fig. 7. The conformational effect of AT-2 lies in the ectodomain.

FRET histograms of HIV-1_JR-FL Env without cytoplasmic tail under conditions: ligand-free and untreated (a); ligand-free Env after AT-2 treatment (b); and in the additional presence of 10–1074 (50 μg ml⁻¹) and 3BNC117 (50 μg ml⁻¹) (c). (d) Quantification of the corresponding relative state occupancies.