High thermostability improves neutralizing antibody responses induced by native-like HIV-1 envelope trimers

Design and generation of hyperstable BG505 SOSIP trimers

We previously described the BG505 SOSIP.v6 immunogen³⁵, a thermostable SOSIP trimer that contained additional inter- and intraprotomeric disulfide bonds and folded as a covalent trimer with a remarkably high thermostability. However, SOSIP.v6 still showed some reactivity to non-NAbs and was not produced as efficiently as preceding SOSIP versions. Our initial goal here was to generate hyperstable SOSIP trimers that do not present these drawbacks. Thus, we selected a set of mutations that have been shown to improve the antigenicity, stability, trimerization, and purification yields of HIV-1 Env trimers^31,33,34,35 and introduced them in the BG505 SOSIP.v6 design in different combinations to generate BG505 SOSIP.v9.1-v9.4 (Fig. 1a, Fig. S1). A subset of the MD39 mutations (519S, 568D, 570H, and 585H)³³ were introduced in all four SOSIP.v9.1 -v9.4 trimers in order to improve trimerization and purification yields. Newly incorporated mutations 306L-308L³⁴, 304V and 319Y³³, as well as the 316W mutation previously introduced in SOSIP.v4³², sequester the immunodominant V3 loop region by hydrophobic interactions. Due to the proximity of these mutations in the three-dimensional structure of the trimer (Fig. 1b), we included them in three different combinations (316W+306L-308L in SOSIP.v9.1 and SOSIP.9.4, 304V+319Y in SOSIP.v9.2 and 316W+304V+319Y in SOSIP.v9.3) to avoid undesired interactions or interferences. The disulfide bond 201C-433C, which reduces sampling of the CD4-induced state³¹, was introduced in SOSIP.v9.4 only.

We first purified His-tagged versions of BG505 SOSIP.v9.1-v9.4 expressed in HEK293F suspension cells by PGT145 affinity chromatography, as previously described^32,51. All SOSIP.v9.1-v9.4 proteins showed increased purification yields compared to their SOSIP.v6 predecessor, with a 3.8 – 4.3-fold increase in the case of SOSIP.v9.1-v9.3 and a more subtle but consistent ~1.3-fold increase for SOSIP.v9.4 (Table 1). These yields were ~1.5-fold higher than those of non-hyperstabilized SOSIP.v4 and SOSIP.v5 trimers³¹. As expected, after PGT145 purification, all SOSIP versions tested formed trimers, as assessed by the BN-PAGE analysis (Fig. S2). An SDS-PAGE analysis showed that all SOSIP.v9.1-v9.4 purified proteins remain trimeric under non-reducing conditions, while they separate into monomers under reducing conditions (Fig. 1c). This confirms that SOSIP.v9 trimers contain interprotomer disulfide bonds that covalently link the three protomers, similarly to the SOSIP.v6 predecessor³⁵. Furthermore, 2D class averages of negative-stain electron microscopy (NS-EM) images revealed that all four SOSIP.v9 variants assumed a >95% native-like conformation (Fig. 1d, Table 1). A Ni-NTA ELISA experiment with the purified proteins revealed a favorable antigenic profile for all the SOSIP.v9 trimers compared to SOSIP.v5 and SOSIP.v6 (Fig. 1e, Fig. S3), with similar (or higher) binding to bNAbs and decreased binding to non-NAbs. BG505 SOSIP.v9.3 showed the most desirable antigenic profile, with increased binding to all the bNAbs tested, in particular the quaternary-dependent bNAbs PG16, PGT145, and VRC026.25, and decreased binding to all non-NAbs tested (17b + sCD4, 14e, 19b, and F105) (Fig. 1e, Fig. S3). In summary, the four BG505 SOSIP.v9 proteins fold as covalent native-like Env trimers and are improved compared to SOSIP.v6 in terms of production yields and antigenicity. Overall, the SOSIP.v9.3 construct displayed the most promising features, especially regarding its antigenicity profile.

SOSIP.v9 trimers show superior global and epitope-specific thermostability

In order to evaluate the influence of the stabilizing mutations incorporated on the thermostability of SOSIP.v9 trimers, we used Nano Differential Scanning Fluorimetry (nanoDSF), which provides information on the overall thermostability, but does not reveal how stable individual bNAb epitopes are. The SOSIP.v9.1-v9.4 trimers presented superior global thermostability compared to all their precursors, including the hyperstable SOSIP.v6 (Fig. 2a, Table 1). The SOSIP.v9.3 and SOSIP.v9.4 proteins displayed the highest melting temperatures (T_m) among the SOSIP.v9 variants, with T_m values of 80.7 °C and 84 °C, respectively (Fig. 2d, Table 1). These T_m values are 13.1 °C and 16.4 °C higher than the ones of first-generation BG505 SOSIP.664 trimers³⁵.

To gauge the thermostability of specific bNAb epitopes, we performed thermostability ELISA assays with bNAb 2G12 (Fig. 2b), which depends on conformational structure but not on quaternary structure, and PGT145 (Fig. 2c), which depends on proper quaternary conformation at the trimer apex. All SOSIP.v9 trimers displayed more stable 2G12 and PGT145 epitopes than the previous SOSIP versions (Fig. 2b–d, Table 1). The T_m values of the 2G12 and PGT145 epitopes of SOSIP.v9.3 (78.3 °C and 76.1 °C, respectively) and SOSIP.v9.4 (85.0 °C and 82.0 °C, respectively) were the highest among the SOSIP.v9 proteins.

The results of all three techniques (nanoDSF, 2G12, and PGT145 thermostability ELISA) were strongly correlated (Fig. 2e, Table 1). The T_m values measured by nanoDSF were ~2 °C higher than the ones measured by 2G12 thermostability ELISA, and ~5 °C higher than those measured by PGT145 thermostability ELISA. These trends are expected, as trimer dissociation probably precedes loss of gp120 conformation and, finally, unfolding.

Chemical crosslinking further stabilizes SOSIP.v9.3

To add an additional layer of stabilization, we used glutaraldehyde (GLA) to chemically crosslink the SOSIP.v9.3 immunogen (SOSIP.v9.3.XL)³⁷. After purification of the SOSIP.v9.3.XL protein by PGT151 affinity chromatography, we visualized it by NS-EM, confirming that >95% of the species presented a native-like conformation (Fig. S4). Trimer antigenicity was evaluated in a BLI experiment with a small panel of relevant Abs (Fig. S5). Chemically crosslinked and PGT151-purified SOSIP.v9.3 showed similar 2G12 and PGT151 binding compared to normal SOSIP.v9.3, whereas PGT145 and VRC01 binding was decreased, consistent with previous crosslinking experiments^37,52. The decrease in PGT145 binding has previously been hypothesized to be related to the direct modification of the lysine residues 168, 169, and 171, rather than to a loss of native-like apex conformation³⁷. Notably, chemical crosslinking increased the T_m of SOSIP.v9.3 from 80.7 °C to 91.3 °C, which is an increase of 23.7 °C compared to the SOSIP.664 prototype³⁵ (Fig. 2d, Table 1). Thus, we generated a panel of SOSIP trimers with a gradient of stabilities (SOSIP.v9.3.XL > SOSIP.v9.4 > SOSIP.v9.3 > SOSIP.v9.2 ≃ SOSIP.v9.1 > SOSIP.v6 > SOSIP.v5 > SOSIP.v4 > SOSIP.664) (Fig. 2d, Table 1).

Addition of PNGS at position 241 and 289 fills the 241/289 glycan hole on SOSIP.v9 trimers

Considering the immunodominance of the N241/N289 BG505 glycan hole, we aimed to immunosilence this epitope and redirect the NAb responses towards other, potentially non-strain-specific, epitopes. Thus, we created Glycan hole Masked (GM) versions of the SOSIP.v9.3 and SOSIP.v9.4 trimers (SOSIP.v9.3.GM and SOSIP.v9.4.GM) by introducing PNGS at positions 241 and 289 (241N, 291S) together with changes that counteract the destabilizing effects result of the incorporation of such PNGS motifs (240T, 271I, 288L, 290E)⁴⁵ (Fig. 3a).

We expressed the His-Tagged versions of SOSIP.v9.3.GM and SOSIP.v9.4.GM in 293 F suspension cells and purified them by PGT145 affinity chromatography. SOSIP.v9.3.GM and SOSIP.v9.4.GM showed slightly decreased purification yields (2.4 mg/L and 0.6 mg/L, respectively) as compared to the respective parental proteins (Table 1). They also expressed as covalently linked trimers, as observed on SDS-PAGE (Fig. S6). When imaged by NS-EM, we observed a reduction in the percentage of native-like trimers of SOSIP.v9.4.GM ( > 85% vs >95%), while SOSIP.v9.3.GM conserved a >95% native-like conformation (Fig. 3b, Table 1). Site-specific glycan analysis revealed that PNGS occupancy and glycan composition was similar for SOSIP.v9.3.GM (Fig. 3d, Fig. S7) when compared to previous SOSIP versions. The occupancy at the newly incorporated 241 and 289 PNGS was 90% and 66%, respectively, thus effectively filling the hole with at least one glycan in the majority of molecules (Fig. S7).

Both GM trimers showed an overall similar antigenic profile as SOSIP.v9.3 and SOSIP.v9.4 (Fig. S8). Furthermore, an antibody targeting the 241/289 glycan hole of BG505 (10 A)⁴² showed negligible reactivity to GM trimers, confirming the effective filling of the 241/289 hole (Fig. 3d).

The inclusion of new PNGS slightly influenced thermostability, with nanoDSF-measured T_m values of 79.9 °C and 83.2 °C for SOSIP.v9.3.GM and SOSIP.v9.4.GM, respectively (i.e. 0.8 °C lower compared to the parental proteins) (Fig. S9).

Ultrastable SOSIP.v9 trimers induce autologous NAb responses

To evaluate the immunogenicity of the SOSIP.v9 trimers and gauge their ability to generate NAb responses, we performed an immunization experiment in rabbits (Fig. 4a). We selected the following immunogens to be tested: (1) the BG505 SOSIP.v5 and SOSIP.v6 predecessors, to serve as control groups; (2) BG505 SOSIP.v9.3 and SOSIP.v9.4, as they showed the best antigenic profile and thermostability, respectively, among the SOSIP.v9 trimers; (3) BG505 SOSIP.v9.3.XL trimer, i.e. the trimer with the highest thermostability; and 4) glycan masked versions of BG505 SOSIP.v9.3 and SOSIP.v9.4 (SOSIP.v9.3.GM and SOSIP.v9.4.GM), to study redirection of responses away from the 241/289 hole. Untagged versions of these immunogens were purified by PGT145 affinity chromatography and an additional size exclusion chromatography (SEC) purification step to remove aggregated species. BN- and SDS-PAGE analyses were performed for quality control (Fig. S10).

Seven groups of animals (5 animals per group) were immunized at three timepoints (weeks 0, 4 and 20) with 30 µg of each of the test proteins formulated with Squalene Emulsion (SE) adjuvant. Sera were obtained from the rabbits at weeks 0, 4, 6, 16, 20 and 22 to characterize the antibody responses (Fig. 4a).

First, we measured the binding of the sera at the different timepoints to a His-tagged BG505 SOSIP.v5 trimer using a Ni-NTA ELISA assay. As expected, substantial increases in binding titers occurred at week 6 and week 22, two weeks after the second and third immunizations, respectively (Fig. 4b). We observed no significant differences in the binding titers induced by the different immunogens (Fig. S11a) at any timepoint. There were also no differences between the binding titers induced by the GM immunogens and their parental constructs (Fig. S11b).

Next, we measured the development of the NAb response against an autologous BG505 pseudovirus (Fig. 4c, Fig. S12). We observed that these autologous NAb responses evolved differently between weeks 6 and 20 for the different groups (Fig. 4c, Fig. S13). In some groups, in particular, the ones receiving SOSIP.v5, SOSIP.v9.3 and SOSIP.v9.3.GM, NAb responses peaked at week 6 and then declined resulting in lower NAb responses at week 20. In contrast, the NAb titers increased from week 6 to week 20 in the groups receiving SOSIP.v6, SOSIP.v9.4, SOSIP.v9.4.GM and SOSIP.v9.3.XL (Fig. 4c, Fig. S13). As a result, there was a trend that the groups receiving more stable proteins maintained higher NAb titers until week 20 (Fig. 4d). While this trend was not statistically significant, it would be consistent with the hypothesis that more stable immunogens might persist for longer thereby enhancing the durability of the response. Animals vaccinated with SOSIP.v6, SOSIP.v9.4 and SOSIP.v9.4.GM showed delayed responses, with no NAb activity detected at week 6 (two weeks after the second immunization), while a NAb response was detected by that time in the other rabbit groups. Nonetheless, all immunogens showed similar responses at week 16 and developed strong autologous NAb responses at week 22, after the third immunization (Fig. 4c, Fig. S12).

To assess the breadth of the NAb responses, we evaluated the neutralization of a panel of Tier 1 and Tier 2 heterologous viruses, including the viruses from the global HIV-1 panel⁵³, at week 22 (Fig. S14). Except for some low titer responses, no consistent Tier 2 heterologous neutralization was observed. Remarkably, rabbits vaccinated with SOSIP.v9.4 and SOSIP.v9.4.GM immunogens showed significantly decreased NAb titers to the Tier 1 A MW965.26 virus (Fig. S15). Neutralizing responses to this virus are dominated by V3-targeting NAbs that are unable to neutralize Tier 2 primary HIV-1 isolates⁵⁴, and the V3 might be less accessible in the SOSIP.v9.4 and SOSIP.v9.4.GM trimers as a consequence of the stabilization. Hence, NAb responses induced by these immunogens might be less hindered by competing V3-targeting narrow-NAb responses.

Autologous NAb responses correlate with SOSIP trimer stability

To assess the influence of the thermostability of the immunogen on the induction of autologous NAb responses, we used the neutralization titers determined for the BG505 pseudovirus virus at week 22 (Fig. 4e, Fig. S12). When we plotted the autologous NAb titers, expressed as the geometric mean (Geomean) of the ID₅₀ values of each group, versus the nanoDSF-measured T_m values of the corresponding immunogen, we observed a positive correlation (p = 0.0167), indicating that increased thermostability is associated with the induction of stronger NAb responses (Fig. 4f). We also observed an inverse trend (p = 0.0833) between the nanoDSF-measured T_m values and the spread of the ID₅₀ values within each group of animals, which suggests that higher thermostability increases the consistency of the NAb response (Fig. 4g). These trends were also observed when 2G12 and PGT145 thermostability ELISA-measured T_m values were used (Fig. S16a) and were found irrespective of the presence or absence of the cytoplasmic tail in the BG505 gp160 used for pseudovirus production (Fig. S16b).

Glycan masking redirects immunodominant glycan-hole responses to subdominant neutralizing epitopes

SOSIP.v9.3.GM and SOSIP.v9.4.GM immunogens were included in the rabbit immunization experiment to test the potential of the glycan masking strategy to redirect the NAb responses towards epitopes different than the 241/289 glycan hole. Surprisingly, we found that animals vaccinated with the glycan masked versions of SOSIP.v9.3 and SOSIP.v9.4 displayed autologous NAb responses of similar magnitude compared to those vaccinated with conventional SOSIP.v9.3 and SOSIP.v9.4, despite the lack of the dominant 241/289 NAb epitope (Fig. 4h). To map the neutralizing epitope of these animals we compared the ID₅₀ titers of wild-type BG505 pseudovirus to the ID₅₀ titer of a panel of mutant BG505 variants that contain PNGS sites at positions 241 and 289 (“241N/289N”), in the variable loop 1 at positions 133 and 136 (“133aN/136aA”) and at position 465 in the constant region 3 (465N) (Fig. S17). The relative ID₅₀ titers (RID₅₀) revealed that responses of most BG505 SOSIP-vaccinated rabbits induced a response to the immunodominant 241/289 glycan hole, consistent with previous studies^42,55. However, few, if any, of NAbs elicited in recipients of GM-modified SOSIP trimers focused on the 241/289 glycan hole (p < 0.0001). Instead, NAb responses from these animals were significantly more focused on the 465 position than those in the non-GM groups (p = 0.0181) (Fig. 4i, Fig. S17). There was no difference for other epitopes, such as the one involving positions 133 and 136 (Fig. 4i). This indicates that the glycan masking strategy efficiently redirected the responses from the immunodominant 241/289 epitope to the subdominant C3/465 epitope, and possibly to other epitopes.

While the autologous neutralization in most animals could be attributed to Abs targeting either the 241/289 hole or the C3/465 epitope, we used a wider panel of mutant viruses to map the responses in animals in which these epitopes were not the major targets (i.e. animals UA139, UA142, UA155, and UA157). Only rabbit UA142 showed decreased neutralization to one of the mutants in the panel (BG505 G458Y), pointing at a CD4bs-targeting NAb response (Fig. S18). The NAb targets in the other three animals remain unidentified.