Glycine Substitution at Helix-to-Coil Transitions Facilitates the Structural Determination of a Stabilized Subtype C HIV Envelope Glycoprotein

Glycine Substitutions in gp41 Promote Soluble Env Trimer Formation

To generate clade C NFL Env glycoproteins amenable to crystallization, we pursued a strategy to promote well-ordered trimer formation. To disfavor the Env post-fusion state and consequently promote pre-fusion trimer formation, we introduced glycine substitutions into key coil-to-helix transitional areas of the gp41 subunit, guided by the pre-fusion high-resolution Env trimer structures (Figure 1A) (Garces et al., 2015, Gristick et al., 2016, Julien et al., 2013, Lyumkis et al., 2013, Pancera et al., 2014). Putative helical intermediates of Env are formed following receptor-coreceptor engagement, liberating gp41 from gp120-imposed conformational constraints. Because glycines disfavor helix formation, we substituted G residues at selected coil-to-helix junctions to inhibit conformational transitions to putative intermediate and structurally defined post-fusion, helix-dominated conformations. To test the glycine substitution strategy, we first generated NFL trimeric proteins based on the transmitted/founder subtype C South African HIV Env sequence, 1086c, introducing TD substitutions previously described (Guenaga et al., 2015b). TD stabilization alone was insufficient to generate native-like trimers for 1086c as determined by both SEC and EM (Figure 1B and Figure S1A), in contrast with their previous substantial impact on trimer formation for Indian clade C 16055 NFL and clade B JRFL NFL (Guenaga et al., 2015b).

To design more homogeneous 1086c NFL trimers and potentially for a wide range of trimers from multiple strains, we identified inferred coil-to-helix transitional regions in gp41 guided by the high-resolution structure of the pre-fusion BG505 SOSIP.664 and gp41 post-fusion structures (Buzon et al., 2010, Garces et al., 2015, Gristick et al., 2016, Julien et al., 2013, Liu et al., 2009, Lyumkis et al., 2013, Pancera et al., 2014). The pre-fusion gp41 regions of interest were located in the N-terminal HR1 loop (HR1_N), which connects the central HR1 helix to the fusion peptide proximal region (FPPR), and in the region proximal to the N-glycan at residue 637 in HR2, where the HR2 helices bend to wrap around the gp120 N and C termini (Figure 1A). Crystal structures reveal that these coiled regions in gp41 transition to helices during the pre-fusion to the post-fusion state (Buzon et al., 2010, Shu et al., 2000, Tan et al., 1997). Therefore, within the context of the 1086c NFL TD protein that displays no significant amount of ordered trimers, we generated selected glycine substitutions to test their effect on ordered trimer formation. Residues T536, L537, L544, L568, T569, N636, and Y638 were individually substituted with glycines and evaluated by size exclusion chromatography (SEC), EM, and immunoprecipitation (IP). Following initial NFL glycoprotein purification by lectin-affinity chromatography, SEC profiles revealed a marked change in the proportions of monomeric, dimeric, and aggregated 1086c glycoproteins carrying the selected glycine substitutions (Figure 1B). 1086c NFL TD glycoproteins with glycine substitutions at positions 568 and 569 displayed the most improved trimer formation relative to the TD variant, with reduced monomer, dimer, and higher-order forms (Figure 1B). Substitutions at 544 and 636 showed moderate improvement, whereas substitutions at 536, 537, and 638 showed no improvement (Figure 1B).

EM negative-stain analysis of the trimeric proteins showed a marked increase in native-like trimers for variants with glycine substitutions at residues 568, 569, and 636 and moderately for 544, generally consistent with the SEC analysis (Figure S1A). IP of NFL TD protein supernatants using the CD4 binding site (CD4bs), trimer-preferring broadly neutralizing antibody (bNAb) VRC03, and the trimer-independent CD4bs bNAb, VRC01, yielded results consistent with the EM analysis. For example, stronger intensity VRC03 IP bands were detected for NFL TD variants with glycine substitutions at residues 568 and 569 compared to isogenic trimers lacking these changes (Figure S1B). We also evaluated various single-, double-, and triple-glycine combinations and concluded that single-L568G or -T569G and double-L568G-N636G or -T569G-N636G substitutions markedly improved the generation of well-ordered 1086c NFL TD trimers. Based on the VRC03 IP band intensity, triple combinations (T569G-L544G-N636G and L568G-L544G-N636G) did not appear to be better than the double substitutions and were not pursued further in this study (Figure S1B).

To assess the transferability of this strategy to other HIV Env sequences, we introduced the 1086c-effective T569G substitution into the JRFL NFL TD and 16055 NFL TD glycoproteins (Guenaga et al., 2015b) and assessed its effect on trimer homogeneity by SEC and differential scanning calorimetry (DSC). The T569G substitution greatly increased trimer formation and homogeneity in both clade B and C NFL TD contexts (Figures 1C, S1C, S2A, and S2B). The L568G mutation conferred similar results in the context of 16055 NFL TD as assessed by SEC and DSC, displaying comparable levels of homogeneity and stability (only 0.4°C less) (Figure S1C). In fact, with no additional purification except for isolation of the trimer fraction from the SEC, 98% of the 16055 NFL TD (T569G) trimers displayed a native-like conformation as determined by negative-stain EM (Figure 1C).

These results revealed that helix-disrupting glycine substitutions at critical helix-to-coil interfaces disfavor transitions toward the receptor-triggered putative HIV Env transitional intermediate and, specifically, that glycine substitutions at the HR1_N residues 568 and 569 and HR2 residue 636 increased native-like Env trimer formation.

The Clade C Cleavage-Independent 16055 NFL TD CC T569G Trimer Shares Structural Homology with Clade A, B, and G Envs

Several HIV-1 subtype C Env-derived SOSIP and NFL trimers are well described biochemically and by low-resolution EM (Guenaga et al., 2015a, Guenaga et al., 2015b, Julien et al., 2015, Ringe et al., 2015, Sharma et al., 2015). However, no high-resolution structure of a soluble subtype C Env trimer has been determined. We therefore used the highly homogeneous T569G variant of 16055 NFL TD CC trimer to generate crystals that diffracted to 3.9 Å in complex with bNAbs PGT124 and 35O22 (Garces et al., 2014, Huang et al., 2014, Pancera et al., 2014). From this ternary complex, we were able to determine the crystal structure of a clade C Env NFL trimer by molecular replacement (see STAR Methods, Figure 2A, and Table 1).

Overall, the structure of 16055 NFL TD CC (T569G) in complex with PGT124 and 35O22 closely resembles those of soluble subtype A BG505 SOSIP.664, subtype G X1193.c1 SOSIP.664, and full-length native subtype B JRFL Env ΔCT Env trimers (Figure 2B) (Garces et al., 2015, Julien et al., 2013, Lee et al., 2016, Pancera et al., 2014, Stewart-Jones et al., 2016). The similarity strengthens the interpretation that the engineered soluble NFL trimers, like SOSIP, are close mimics of HIV Env. The largest differences among Env structures were localized to variable regions V1, V2, V4, and V5 of gp120 and the HR1_N region spanning residues 545–569 of gp41 (Figure S2C). With the exception of the V3, variable regions of gp120 differ both in length and conformation and contribute to antigenic variation with respect to antibody recognition (Figure S2C). The observed HR1_N differences may be relevant to trimer stability and, perhaps, to the capacity of different Env sequences to form stable, soluble trimers, since the well-documented I559P substitution (Sanders et al., 2002) and the T569G described here are located in this flexible region, and both affect trimerization propensity (Figure S2C). Another difference in these NFL trimers is the flexible (G₄S)₂ linker connecting gp120 to gp41 that eliminates the need for furin cleavage (Sharma et al., 2015), for which only the first three glycine residues are visible in the NFL structure (Figure 2A).

Structural alignment between the BG505 SOSIP.664 and the 16055 NFL TD CC (T569G) structures revealed low C_α RMSD values for the gp140 protomer (0.7 Å) and the respective gp120 (0.6 Å) or gp41 (0.6 Å) subunits (Figure S3A). Although residues 553–565 of the gp41 HR1_N were not resolved in the 16055 structure, L566, Q567, L568, and G569 downstream of the I559P substitution were interpretable and offer some insights into how the 569 glycine substitution contributed to the generation of stable clade C Env NFL trimers (Figure 2B). The orientation and location of these four residues (566–569) in 16055 NFL TD CC (T569G) deviate from that in the BG505 SOSIP.664 and X1193.c1 SOSIP structures (PDB: 5CEZ and PDB: 5FYJ) but follow an initial path more similar to that in the clade B JRFL Env ΔCT (PDB: 5FUU), although different in conformation (Figures 2B and S2C). The helix-disrupting capacity of the T569G mutation disfavors the extension of the HR1_C helix, helping HR1_N remain in its flexible coil form observed in the SOSIP structures (Figures 2B and S2C). The increase in NFL trimer formation may also result from generation of new local stabilizing contacts in the HR1_N region; however, no increase in the overall trimer thermal transition midpoint (T_m) was detected (Figure S2A). Perhaps the glycine substitution simply decreases the tendency of this region to adopt a transitional intermediate form of gp41, therefore favoring the pre-fusion state.

These analyses demonstrated that the crystal structure of the furin-independent clade C Env 16055 NFL TD CC (T569G) shares overall structural homology with other Envs, revealing differences in the surface-exposed variable regions and in the more occluded trimer-axis proximal region of gp41 HR1_N.

Env N-Glycosylation Differences between Strains Identify Potential Sites of Vulnerability for Antibody Recognition

HIV has evolved immune evasion mechanisms that pose a major challenge to vaccine design. Env surface N-glycosylation is an important evasion mechanism in which self-glycans occlude the underlying Env protein surface. With an average of 80–90 N-linked glycans per Env trimer, this protective coverage is often referred to as the “glycan shield” (Lasky et al., 1986, Wei et al., 2003). The 16055 NFL TD CC trimers were produced in a GnTI^−/− mutant cell line that results in processing of N-glycans to Man_5–9GlcNAc₂ oligomannose forms. The ternary complex of PGT124:16055 NFL TD CC:35O22 was deglycosylated by Endoglycosidase H (Endo H) treatment to facilitate crystallization. At least one sugar moiety for 22 glycans was visible in the electron density maps out of the 29 potential N-linked glycosylation sites (PNGS) in the 16055 Env sequence. For 15 N-glycans out of the 22 visible in the structure, we observed density beyond the peptide-proximal N-acetylglucosamine moiety, indicating protection from Endo H cleavage. The other glycan positions for which we did not observe interpretable density were mainly in disordered regions or loops. In comparison, 17 glycans had observable electron density in the 27 PNGS of the 3Å structure of BG505 SOSIP.664 (Figure 3) (Garces et al., 2015). The V1 and C2 regions of 16055 were more heavily glycosylated, possessing five additional PNGSs at N138, N147, N230, N241, and N289. In contrast, the C3 region of BG505 had two additional potential glycosylation sites at N339 and N355 (Figure 3). Breaches in the glycan shield, such as those corresponding to N230-N241-N289 and N339-N355 for BG505 and 16055, respectively, may define strain-specific sites of vulnerability as recently shown for Env vaccine-elicited antibodies in rabbits (Crooks et al., 2015, Klasse et al., 2016, McCoy et al., 2016). Some of these rabbit antibodies neutralize the HIV BG505 strain by exploiting a gap in the glycan shield at the N241 region, thereby more readily accessing the exposed protein surface. However, the majority of HIV strains are glycosylated at N241, as is 16055, or possess a different protein sequence in this region, making them resistant to neutralization by such antibodies (Figure 4A).

Almost all bNAbs isolated from infected patients, such as PGT124 that targets the 332 N-glycan and V3 base, incorporate glycans as part of their epitope, contacting a combination of sugar moieties and underlying protein residues (Figure 4B). In the 16055 crystal structure, PGT124 penetrates into a cleft surrounded by glycans N156, N301, and N332 (Figure 4B). The antibody mostly avoids the N-glycans at residues N156 and N301 but makes contact with N332 and the conserved V3 ³²⁴G(N/D)IR³²⁷ motif as previously demonstrated for PGT124 in the context of a gp120 core (Garces et al., 2014). The NFL trimer structure shows additional van der Waals interactions between PGT124 CDRL3 (Ser93, Ser95), CDRH3 (Phe100k), and gp120 V1 (Thr137) and backbone contacts between CDRL3 (Arg94) and the base of gp120 V3 (Ile322) (Figure 4B). Phe100g at the tip of CDRH3 that is important for the neutralization activity of PGT124 in a paratope scan (Garces et al., 2014) interacts with Env Gln328, Tyr330, Thr415, and Pro417 (Figure 4B).

These data were consistent with the limited neutralizing capacity of strain-specific antibodies that exploit gaps in the HIV Env N-glycan shield, while bNAbs evolve to accommodate glycan recognition as a means to more consistently penetrate the shield.

The High-Resolution 16055 NFL Structure Reveals the Stabilizing Effects of the TD and CC Substitutions

The 16055 NFL trimer contains eight engineered TD residues, the CC disulfide, and the T569G glycine substitution. The NFL structure elucidated the role of these design elements in generating native-like trimers with increased stability (Guenaga et al., 2015b). We observed interactions for six of the eight TD residues located in four distinct areas of Env: gp120-gp41 interface (Arg500), gp120 N terminus (Asp47 and Glu49), pre-bridging sheet (Arg429 and Gln432), and apex domain (Leu165) (Figure S3B). The Arg500 substitution at the C terminus of gp120 contacts the adjacent gp41 His619 and gp120 N-terminal residues Gly32 and Leu34, strengthening the association of gp120 with gp41 at the base of the trimer. Asp47 and Glu49 in the N-terminal region of gp120 contact Lys487 and Asp99, respectively, anchoring this flexible region. Leu165 at the tip of the trimer V2 apex contacts Thr128 and Cys126 of the adjacent protomer, presumably strengthening apical trimer interactions. Arg429 and Gln432 contact Asn425, Thr202, and Leu116, creating a network of interactions that may stabilize the pre-bridging sheet (Figure S3B). TD Lys65 and Thr106 do not contribute to any obvious productive interactions.

The 16055 NFL TD CC (T569G) crystal structure illustrates formation of the I201C-A433C disulfide (CC) that was engineered to prevent CD4-induced trimer rearrangements and subsequent exposure of non-neutralizing epitopes (Figure 5A) (Guenaga et al., 2015b, Kwon et al., 2015). As the 16055 NFL TD CC (T569G) and BG505 SOSIP.664 trimer structures displayed similar conformations, with a C_α RSMD of 0.6 Å for gp120 (Figure S3A), the engineered CC does not appear to significantly alter the conformation of the pre-CD4 form of the CD4bs. This region constitutes a conserved neutralizing determinant, so maintenance of an unperturbed CD4bs is an important consideration for HIV vaccine design. Relevant to this latter point, affinity and avidity measurements for bNAb Fabs VRC01 and VRC06b to NFL TD trimers with and without the CC disulfide were of similar magnitude (Figure S4). For sCD4, although the overall K_D values were the same, CD4 had a slightly faster on rate but faster off rate with the NFL TD CC trimers.

Comparison of the gp120 subunit of the 16055 NFL TD CC (T569G) trimer structure with the CD4-liganded gp120 core (PDB: 3JWO) (Pancera et al., 2010) revealed differences between these two gp120 conformations, as previously described (Pancera et al., 2014) (Figures 5B and 5C). This comparison suggested a receptor-triggered model, whereby gp120 conformational changes culminate in the release of restraints on gp41. In this model, CD4 interaction with gp120 induces a displacement of the β20 and β21 loop, triggering Env rearrangements that lead to co-receptor engagement. Serine 42 of CD4 would “push” gp120 glutamine 428 (β20), facilitating release of β21 contacts with β3 (Figure 5B). The subsequent swap of β3 and β2 facilitates rearrangement of the variable domains (V1/V2), liberating the V3 region to engage with the co-receptor. However, the engineered CC disulfide specifically prevents dissociation of β21 from β3, acting as a molecular tether that locks gp120 into its pre-receptor conformation (Figure 5A). Other structural differences were observed that connect the CD4-bound trigger point (β20-β21) with release of gp41 HR1 from gp120. Helix α-1 in the CD4-liganded core was tilted a few degrees toward the adjacent α-0 (Figure 5C), suggesting that CD4 receptor and subsequent CCR5 coreceptor binding trigger a chain of events that lead to the formation of a putative intermediate Env structure.

In sum, these analyses demonstrated that the TD, CC, and T569G mutations combine to preserve the pre-fusion state of the NFL trimers by disfavoring conformational changes in gp120 and gp41 that otherwise would favor the fusion-active form.

Targeting Additional HIV Env Metastable Regions Facilitates Generating NFL Trimers Derived from Several Subtypes

Metastability is likely a requisite for Env to mediate receptor-triggered viral entry at neutral pH. Regions of Env, such as the variable region 3 (V3) and the fusion peptide (FP), undergo large translational and conformational rearrangements following receptor-coreceptor engagement on target cells (Figure 6A) (Buzon et al., 2010, Luftig et al., 2006, Pancera et al., 2014). We sought to further stabilize the soluble NFL trimers beyond the stabilizing mutations described in the structure presented here by further reducing the inherent Env metastability. We reasoned that conformational stabilization of these trimeric immunogens might be important for preservation and presentation of quaternary-dependent neutralizing epitopes following vaccination and affinity maturation while avoiding Env states that expose non-neutralizing epitopes.

We selected JRFL NFL TD CC (T569G) as a test case because of its inherent lower thermostability in comparison with BG505 or 16055 native-like trimers (Guenaga et al., 2015a, Guenaga et al., 2015b, Sharma et al., 2015). Following a structure-guided approach, a number of V3 or FP residues were scanned for their capacity to increase inter- or intra-protomer interactions. We made substitutions of selected residues in these regions to large side chains—mainly, Trp, Tyr, and Arg—to fill pockets or make potential additional contacts to enhance subunit and trimer interactions. The initial scan identified three V3-FP stabilizing mutations (V3 N302Y and FP F519R and L520R) that resulted in increased trimer melting temperatures. DSC analysis of the triple mutant (N302Y F519R L520R) JRFL NFL TD CC (T569G) confirmed a 6°C increase in the T_m (Figure 6B). We introduced the V3-FP mutations into subtype A BG505 NFL CC and subtype C 16055 NFL TD CC (T569G) and determined that their respective T_m’s also increased by 6°C (Figure 6B). By applying all elements of the current design strategy (TD, CC, gp41 glycines, and V3-FP mutations, herein called NFL TD CC+) to other HIV Env sequences from clades B (SC422 and WITO) and C (1086c, 001428 and ZM197M), we sought to demonstrate the general applicability of our soluble NFL TD CC+ trimer platform (Figure S5). The NFL TD CC+ design made use of double glycine substitutions, either G568-G636 or G569-G636, as needed to increase homogeneity of the highly stabilized trimers. The resulting NFL TD CC+ trimers were purified via lectin affinity followed by SEC and then subjected to DSC analysis to assess their overall stability and homogeneity. SEC profiles revealed prominent trimer peaks at the expected elution volume (Figure 6C), resulting in final yields of well-ordered trimers ranging from 2–4.5 mgs per liter of transfected 293F cells (Figure S5). Trimer stability, measured by DSC and reported as the T_m, ranged from 66°C to 77°C, displaying single narrow symmetric thermal transition curves that are characteristic of a homogeneous molecular species (Figure S6). Recognition of trimer-preferring V2-apex bNAbs (PGT145, PGDM1400, PG16, and VRC26), which bind one quaternary epitope per trimer, suggested that these soluble trimers adopted native-like conformations. Binding affinities for these ligands were determined by BLI and ranged from 2 to 205 nM (Figure 6D and Table S1). Generally, higher affinities (< 100nM) for NFL TD CC+ trimers were associated with neutralization of the matched parental HIV strain by that same bNAb, while lower binding affinities (> 100nM) were associated with an inability of that bNAb to neutralize the parental isolate from which the NFL was derived (Table S1).

These results indicated that combining the HIV Env stabilizing elements (TD, CC, glycines, and the V3-FP mutations) facilitated the high-yield production of cross-clade stable soluble NFL trimers that were efficiently recognized by trimer-specific bNAbs.

Key Resources Table

Contact for Reagent and Resource Sharing

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Richard T. Wyatt (wyatt@scripps.edu).

Method Details

Data and Software Availability

The 16055 NFL TD CC (T569G) crystal structure has been deposited to the RCSB Protein Data Bank with accession number PDB: 5UM8.