Closing and Opening Holes in the Glycan Shield of HIV-1 Envelope Glycoprotein SOSIP Trimers Can Redirect the Neutralizing Antibody Response to the Newly Unmasked Epitopes

Design, production, and use of glycan-modified SOSIP trimer immunogens.

We based the design of the BG505 SOSIP.v4.1 trimer variants on our identification of the 241/289-glycan hole as the predominant autologous NAb epitope in rabbits (29). Accordingly, we knocked in a glycan at residue 241 in an attempt to block access to this epitope. At the time this study was initiated, we had not identified the C3/465 epitope as the next most immunogenic autologous NAb epitope (38). Hence, we did not change this site on the mutant trimers used in this study (although see the description of group 8, below). For B41, we knocked in a glycan at residue 289 in an attempt to close the only known epitope for autologous NAbs on this genotype of SOSIP trimer (13, 29). The amino acid sequence changes and designations of the various trimers used in this study are described in Tables 1 and 2. Data on the occupancy of the various knocked-in glycans are described below and were obtained by a proteomics-based method (37) only after the immunization experiments had been completed.

In our previous rabbit immunization studies, we used Iscomatrix adjuvant (14, 27, 29, 50, 51, 54). However, as this reagent was no longer available, we obtained GLA-LSQ from the Infectious Disease Research Institute (IDRI), Seattle, WA, via the Bill and Melinda Gates Foundation’s (BMGF) collaborative network and used it for the first 4 immunizations. While this adjuvant did support the immunogenicity of the various BG505 and B41 SOSIP.v4.1 trimers, we were concerned that, in this and other contemporaneous experiments not reported here, the incidence of non- or poorly responding rabbits was higher than we expected. For example, of the 5 rabbits immunized four times with the parental BG505 SOSIP.v4.1 trimers in GLA-LSQ, only 2 had autologous NAb titers of >20 (specifically, 150 and 700), which corresponds to fewer responses and lower titers than we typically see after 3 immunizations with Iscomatrix (14, 27, 29, 50, 51, 54) (see also the summary of group 1 below). For the parental B41 SOSIP.v4.1 trimers, the autologous NAb titers after three or four immunizations with GLA-LSQ also tended to be lower than those previously obtained with Iscomatrix (14, 29, 54). In an attempt to increase the NAb responses, we reboosted all the rabbits in every BG505 and B41 immunogen group twice with the same trimers and a different adjuvant, an Adjuplex-based formulation (see Materials and Methods). The response rate in the parental trimer and other groups then increased. Accordingly, the data presented below were all obtained 2 weeks after the sixth immunization (4 with GLA-LSQ and then 2 with the Adjuplex-based formulation).

The neutralization responses, including mapping data, are summarized below on a group-by-group basis for the BG505 trimers and then for B41. For BG505, the data were derived with the parental BG505.T332N pseudovirus, mutants corresponding to the immunogen trimers, and other mutants with substitutions at the 241/289 and C3/465 sites that are targeted by NAbs from most rabbits immunized with BG505 SOSIP trimers (29, 38). For B41, the data were obtained with the parental B41.R315Q and glycan site mutant pseudoviruses. In both cases, assessments of titer changes could only be made reliably when the half-maximal inhibitory dilution factor (ID₅₀) values against the parental pseudovirus or the autologous, i.e., immunogen-matching, virus were >100.

Group 1: the parental BG505 SOSIP.v4.1 trimer.

The neutralization data for all 8 of the BG505 trimer immunogen groups are summarized in Fig. 1. The abilities of the sera to neutralize the autologous (trimer immunogen-matched) pseudovirus and the parental (BG505.T332N) pseudovirus are shown in Fig. 1A and B, respectively. For group 1, the data are the same in the two plots, as the autologous virus is the parental.

After their sixth immunization, all group 1 rabbits developed detectable, albeit wide-ranging, autologous NAb titers, which was also the case for most of the other groups (Fig. 1A). For group 1, sera from the 5 rabbits generated ID₅₀ values in the range of 51 to 5,800 (median, 140) against the autologous, parental pseudovirus (Fig. 1A and B). Two sera, r2213-1 and r2214-1, neutralized this virus with high ID₅₀ values, 640 and 5,800, and the other three with low ID₅₀ values (Fig. 1). For both of the strongly neutralizing sera, the NAb titers were markedly reduced (relative ID₅₀ [RID₅₀] values, 5 to 32%) by the N241-KI and N241-KI/N289-KI mutations, while the C3/465 epitope was not targeted (Fig. 2A and B).

Group 2: the BG505 SOSIP.v4.1 + N241-KI trimer.

The N241-KI trimer has a glycan inserted at residue 241 to block the dominant 241/289-glycan hole site (Fig. 3A). Three out of 5 rabbits developed detectable but low NAb responses against the autologous N241-KI mutant pseudovirus (Fig. 1A). The NAb titers against the parental BG505.T332N pseudovirus and the autologous responses against the N241-KI mutant pseudovirus in the group 2 rabbits were indistinguishable and in the range of 20 to 260; hence, these weak responses tolerated the N241-KI mutation (Fig. 1A and B; see also Fig. A1). These titers, both autologous and against the parental virus, were lower than the autologous responses in group 1 (titers of 51 to 5,800). It is, therefore, possible that the added N241-glycan had suppressed the immunogenicity of the trimer, although the statistical trend was not significant (P = 0.14 in both cases). Only one rabbit (r2220-2) developed NAbs that neutralized the parental pseudovirus at a titer (ID₅₀ of 250) high enough for the cross-reactive epitope to be mapped. The reduced RID₅₀ value of this serum for the N365-KI/R360I double mutant implicated the C3/465 site as the NAb target (Fig. 3B).

Group 3: the BG505 SOSIP.v4.1 + N241-KI + N130-KI trimer.

The BG505 SOSIP.v4.1 + N241-KI + N130-KI mutated trimer contained knocked-in glycans at both residues 241 and 130, the latter to block a putative glycan hole in V1 (Fig. 4A). The design of this immunogen was suggested by comparative sequence analysis and structural modeling; in the Los Alamos database, 62% of the isolates have a glycan at residue 130 (https://www.hiv.lanl.gov). The design of this trimer predated our use of BG505.T332N virus mutants to map NAb responses, which showed that the N130-KI mutation did not reduce ID₅₀ values for sera from BG505 SOSIP.664 trimer-immunized rabbits (38). Four of the 5 rabbits developed detectable autologous NAb responses (20 to 1,900; median, 82) (Fig. 1A). In 2 of the animals, the responses against the parental virus were strong enough for the shared epitope to be analyzed (Fig. 1B). For r2225-3, the NAb response was clearly directed against the C3/465 site, but the evidence was less definitive for r2224-3 (Fig. 4B). Here, however, we noted that the NAb titer against the autologous N130-KI/N241-KI mutant pseudovirus was increased by 72% compared with that of the parental (i.e., an RID₅₀ of 170%) (Fig. 4B). Hence, a partial redirection of the neutralization response to an immunogen-specific, glycan-dependent neoepitope may have occurred in this rabbit.

Group 4: the BG505 SOSIP.v4.1 + N241-KI + N197-KO trimer.

In group 4, the immunogen was a trimer with the N241-glycan knocked in and the glycan at C2 residue 197 knocked out (Fig. 5A). This design was intended to answer whether the NAb response could be diverted from the 241/289-glycan hole epitope toward the lining of the hole that was created within C2 in proximity to the CD4bs. However, as the N241-KI + N197-KO double mutant virus was noninfectious, the analysis was restricted to the N197-KO mutant pseudovirus. The ID₅₀ values against this mutant pseudovirus ranged from 39 to 15,000 (median, 470) (Fig. 1A). Three rabbits raised NAb responses that strongly cross-neutralized the parental BG505.T332N pseudovirus (Fig. 1B). In sera r2228-4 and r2231-4, these NAbs were directed against the C3/465 site (Fig. 5B). However, the NAbs in r2229-4 were partly redirected to a neoepitope dependent on the N197-KO mutation (Fig. 5B). The effect on NAb recognition of this change when combined with the N241-KI mutation could not be tested, since that double mutant pseudovirus was not infectious.

One rabbit in group 4 (r2228-4) produced a particularly strong pan-neutralizing response across all of the BG505-based pseudoviruses that it was tested against (Fig. 5B). Against other mutants not included in Fig. 5B, the ID₅₀ values for this serum were in the range of 4,800 to 29,000. The lower values occurred with viruses that included mutations in the C3/465 site, such as the N356-KI and N465-KI changes (Fig. 5B). RID₅₀ values against the N355-KO and N241-KI + N355-KO mutants were ∼30%. We conclude that the N197-KO mutation had not created an immunogenic epitope in this rabbit and that the NAb response was directed to the C3/465 site, including the N355-glycan. It may or may not have been diverted there by the N241-KI mutation, as this site is immunogenic in a subset of rabbits given the parental trimer (38).

Group 5: the BG505 SOSIP.v4.1 + N241-KI + N234-KO trimer.

The group 5 immunogen had the N241-KI mutation and a new hole in the glycan shield at residue 234 that is close to the CD4bs (Fig. 6A). The parental NAb titers in group 5 were significantly lower than the autologous titers, suggesting that the responses in this group had been redirected (P = 0.0079) (Fig. 1A and B; see also Fig. A1). Thus, all 5 rabbits responded with substantial NAb titers (ID₅₀, 170 to 1,500; median, 800) against the autologous N241-KI + N234-KO mutant pseudovirus (Fig. 1A), but none of the sera cross-neutralized the parental pseudovirus (Fig. 1B). The consistently and markedly enhanced neutralization (RID₅₀ of 700 to 4,100%) of the N234-KO and N241-KI + N234-KO mutant pseudoviruses by all 5 sera implies that a consistently immunogenic neoepitope has been created by eliminating the N234-glycan from the trimer (Fig. 6B).

Group 6: the BG505 SOSIP.v4.1 + N241-KI + N276-KO trimer.

For group 6, the N241-KI mutation was accompanied by a newly opened hole at residue 276, where this CD4bs-flanking glycan was deleted (Fig. 7A). Sera from 4 rabbits immunized with this trimer variant neutralized the autologous N241-KI + N276-KO mutant pseudovirus detectably (ID₅₀ for the five sera of 20 to 1,800; median, 570) (Fig. 1A), while 3 sera (r2237-6, r2240-6, and r2241-6) strongly cross-neutralized the parental virus (Fig. 1B). In rabbits r2237-6 and r2241-6, the NAbs were directed against the C3/465 site, but we were unable to identify the epitope on the parental pseudovirus that was targeted by NAbs in the r2240-6 serum (Fig. 7B). For rabbits r2238-6 and r2240-6, the strongly enhanced neutralization (RID₅₀ of 950 to 6,400%) of the N276-KO and N241-KI + N276-KO mutants clearly showed that NAbs were generated against the newly created N276-glycan hole (Fig. 7B). It is possible that N276-KO mutation specifically facilitates the binding of antibodies with an approach angle similar to that of some CD4bs-targeting bNAbs, such as VRC01 and N6 (54–56). Indeed, VRC01 was 14-fold and N6 was 10-fold more potent against the N276-KO mutant than the parental pseudovirus (the respective IC₅₀ values were the following: VRC01, 0.014 ± 0.0011 versus 0.19 ± 0.025 μg/ml; N6, 0.012 ± 0.0010 versus 0.13 ± 0.0088 μg/ml).

Group-7: the BG505 SOSIP.v4.1 + N241-KI + N332-KO trimer.

For group 7, we tested the effect of knocking out the N332-glycan from the N241-KI mutant trimer (Fig. 8A). All 5 of the resulting group 7 sera neutralized the autologous N241-KI/N332-KO virus (ID₅₀ of 120 to 650, median of 330) (Fig. 1A), and 4 of them cross-neutralized the parental virus (Fig. 1B). The responses in the 4 cross-reactive sera were directed against the C3/465 site, although the N289-KI mutation also had a modest additional effect (Fig. 8B). Two of the rabbits (r2243-7 and r2246-7), however, also generated NAbs against the neoepitope created by the N332-glycan deletion, as shown by the increased neutralization (RID₅₀ of 220 to 1,500%) of the N332-KO and N241-KI + N332-KO viruses (Fig. 8B).

Group 8: the BG505 SOSIP.v4.1 + N241-KI + N355-KO trimer.

The final BG505 group received the N241-KI mutant trimer from which the N355-glycan was deleted (Fig. 9A). The NAb titers against the autologous virus varied widely, but two rabbits responded strongly (Fig. 1A). Four of the sera weakly neutralized the parental virus; the fifth was negative (Fig. 1B). There was a nonsignificant trend for the titers against the autologous N241-KI + N355-KO virus to be higher than those against the parental BG505.T332N virus (P = 0.15). The likely explanation for the low ID₅₀ values is that the N355-KO mutation also changed the immune response to the C3/465 site, as residue 355 is located within a stretch of C3 residues that is integral to that epitope (38). (As stated above, we were unaware of the existence of this NAb epitope at the time the mutant trimer was designed, so this outcome was fortuitous.) Hence, on the SOSIP.v4.1 + N241-KI + N355-KO trimer immunogen, the principal immunogenic NAb epitopes are either blocked when the N241 glycan is knocked in or affected when the N355 glycan is knocked out. It was notable, however, that sera from 3 of the immunized rabbits (r2248-8, r2249-8, and r2250-8) had enhanced neutralization activity (RID₅₀ of 460 to 25,000%) against the N355-KO and N241-KI + N355-KO viruses (Fig. 9B). Thus, although the N355-KO mutation eliminated any cross-reactivity with the parental C3/465 epitope, it apparently also created an immunogenic neoepitope. It seems, however, that the antibodies raised against this neoepitope can only neutralize mutant viruses that also lack the N355-glycan. Four different N355-KO mutant pseudoviruses were previously found to be resistant to neutralization by some sera from BG505 SOSIP.664 trimer-immunized rabbits that contain NAbs directed against the C3/465 epitope cluster (38). It is possible that the N355-glycan contributes to a subset of C3/465 region epitopes while a glycan resulting from the N356-KI mutation blocks access to all such epitopes (38). In contrast, neutralization by rabbit sera specific for the 241/289-glycan hole is frequently enhanced by the N355-KO mutations (38). An explanation may be the existence of an immunogenic trimer subpopulation on which the N355 glycan site is not occupied and that induces antibodies that can only neutralize viruses lacking that glycan (38).

Overall neutralization sensitivity of glycan-KO and non-glycan-KO mutants, tier 1A NAb responses, and evidence for redirection.

The ID₅₀ values for neutralization of glycan-KO mutant viruses by the nonautologous (i.e., group- and virus-mismatched) sera (n = 315) did not differ from those for neutralization of non-glycan-KO viruses by nonautologous sera (n = 140) (P = 0.73). This analysis shows that the glycan-KO mutants had not become globally neutralization sensitive. We then checked whether the glycan-KO mutant trimers (groups 4 to 8) had induced stronger neutralization titers against tier 1A viruses, MN.3 and MW965.26, than had the non-glycan-KO mutant trimers (groups 1 to 3; see Fig. A2). Again, there was no difference in titers (P = 0.98 and P = 0.63 against MN.3 and MW965.26, respectively). We also found general evidence that NAb responses were redirected to epitopes in the newly opened glycan holes. Thus, serum ID₅₀ values against the immunogen-matched glycan-KO mutant pseudoviruses (n = 45; median, 410) were significantly higher than those against non-immunogen-matched glycan-KO mutants (n = 315; median, 66; P < 0.0001; see Fig. A3).

Glycan occupancy on BG505 SOSIP.v4.1 trimers.

An assumption underlying this experiment was that glycans would be added efficiently at each knocked-in potential N-linked glycosylation site (PNGS) on the SOSIP trimer immunogens and the corresponding mutant pseudoviruses. Recent efforts to measure glycan occupancy at specific sites on trimers and viruses have shown that the above-described assumption is not always valid; thus, a small subset of PNGS on SOSIP trimers of three different genotypes are significantly underoccupied (<90%), whereas occupancy approaches 100% across the glycan sites on virion-derived Env (37,57,58). We applied the latest methodology to retrospectively quantify glycan site occupancy at the knocked-in sites on the BG505 SOSIP.v4.1 trimer immunogens. The N241 site was occupied to extents that varied from 81% (group 4) to 100% (group 5) on the various trimers for groups 2 to 8, almost exclusively by oligomannose glycans (Fig. 10A). For comparison, the N332 site was almost completely occupied (∼100%) on the trimers used to immunize groups 1 to 6 and 8, although the proportion of complex glycans varied from 2.2 to 14% (Fig. 10A).

Every glycan site on a lymphoid cell line-derived stock of fully infectious BG505.T332N virus has been found to be almost fully occupied (58). We were not able to produce enough pseudovirus for occupancy analyses, but we analyzed the available neutralization assay data in two ways relevant to N241-glycan occupancy. First, we identified sera from 13 rabbits in groups 2 to 8 for which the ID₅₀ values against the parental BG505.T332N pseudovirus were >100; the RID₅₀ values for those sera against the N241-KI mutant are listed in Fig. 3 to 9. The RID₅₀ values for these 13 sera against the N241-KI mutant and parental viruses were indistinguishable (P = 0.67 by Wilcoxon’s signed ranked test against the 100% RID₅₀ value for the parental). Thus, the N241-KI mutation in the pseudovirus did not consistently reduce the titers, and for one group 4 serum, r2241-4, the RID₅₀ value for the mutant virus was even increased to 210%. These observations suggest that the dominant NAbs can bind at least as well in the presence of the N241-glycan as in its absence. Second, sera from all 17 rabbits in groups 5 to 8 included in Fig. 6 to 9, which received different N241-KI + glycan-KO mutant immunogens, had ID₅₀ values of >100 against the corresponding single glycan-KO mutant. Those NAb responses were high enough to allow quantification of potential titer-reducing effects of the N241-KI mutation, but the RID₅₀ values against the double N241-KI + glycan-KO mutants did not differ significantly from those against the single glycan-KO mutants (P = 0.24 by Wilcoxon’s matched-pairs test).

Taken together, the findings described above suggest that the N241 site was highly occupied on the various BG505 SOSIP.v4.1 trimer immunogens that include the N241-KI mutation and also on the corresponding N241-KI mutant pseudoviruses. The glycan site occupancy data for the trimers are therefore consistent with the relatively strong neutralization, where measurable, of the N241-KI virus in groups 2 to 8. However, the observed variation in occupancy of 81 to 100% among the immunogen groups could have consequences; a low level of nonoccupancy might not confer neutralization sensitivity to a virus (59, 60), but the same degree of nonoccupancy (e.g., 20% of the Env proteins) on an immunogen trimer might induce a dominant antibody response that neutralizes only viruses that lack the same glycan, either naturally or created by a KO mutation. In this context, we note that the strongest evidence for redirection of the NAb response to neoepitopes, through the opening and closing of glycan holes, was from the trimer immunogen groups with the highest occupancy of the N241 site (100% in group 5 and 96% in group 8) (Fig. 6, 9, and 10A). Conversely, there were at most weak indications of a redirected NAb response in group 4, where the N241 site occupancy, 81%, was the lowest measured among the trimer immunogens. We also observed that the N130 site was poorly occupied, 41%, on the BG505 SOSIP.v4.1 + N241-KI + N130-KI trimer used to immunize group 3 rabbits, which would increase the difficulty of inducing N130 glycan-dependent NAbs (Fig. 4 and 10A).

Immunogenicity of the B41 SOSIP.v4.1 trimers.

All of the B41 trimer immunogens incorporated the R315Q change to eliminate proteolytic clipping within V3, but for simplicity, we omit its mention in the descriptions below. The rabbit experiment was based on 4 groups of 5 animals: group 9, B41 SOSIP.v4.1; group 10, B41 SOSIP.v4.1 + N289-KI; group 11, B41 SOSIP.v4.1 + N289-KI + N332-KO; group 12, B41 SOSIP.v4.1 + N289-KI + N356-KO (Table 2). All animals responded with detectable autologous neutralization except two in group 12 (Fig. 11A). Of note is that, for group 10, the autologous NAb response (i.e., to the B41 N289-KI virus) was markedly lower than the cross-neutralization of the parental B41 virus (ID₅₀ values of 44 to 670, median of 90, versus ID₅₀ values of 82 to 30,000, median of 4,300; P = 0.0079) (Fig. 11A and B). One explanation is that the glycan site was poorly occupied on the immunogen trimer.

The occupancy of the knocked-in N289 site on B41 trimer immunogens is discussed below. The autologous NAb responses in group 9 were, as expected, directed to the N289-glycan hole, as judged by the resistance of the B41 N289-KI virus to the sera (Fig. 12). However, unexpectedly, such resistance also occurred in the group 10 and group 11 rabbits, which all received trimer variants on which the N289-glycan was knocked in to fill that hole (Fig. 13 and 14). We also found that the B41 N230-KI mutant virus consistently tracked the resistance of the N289-KI mutant. These two glycan sites are highly proximate, such that a knocked-in glycan at either site can block the same hole in the glycan shield (38). In group 12, there were some similarities to but also differences from the other three groups (Fig. 15). Of the four responding rabbits, r2271-12 and r2275-12 developed NAbs to the 289-glycan hole, and the response in r2273-12 was partially sensitive to the nearby N230-KI substitution. However, r2272-12 developed NAbs only against a neoepitope created by the N356-KO change, while r2271-12 and r2273-12 also raised NAbs against that site as well as against the N289/N230-glycan hole epitope (Fig. 15). These findings contrast with the lack of evidence from group 11 that the N332-KO mutation at the base of V3 produced an immunogenic neoepitope.

Glycan occupancy on the B41 SOSIP.v4.1 trimers.

A retrospective glycan site occupancy analysis of the immunogens provided an explanation for the immunogenicity of the 289-glycan hole on trimers that contained the N289-KI mutation, i.e., groups 10 to 12. Thus, the N289 site was occupied on only 64% of the group 10 trimers, and it was even less occupied, at 18%, on the group 11 immunogens. Only for group 12 was occupancy of the N289 site high, at 82% (Fig. 10B). Substantial subpopulations of trimers that have the immunogenic N289-glycan hole account for the NAb responses in these groups, particularly groups 10 and 11. The differences between the group 11 (no NAbs to the N332-KO neoepitope) and group 12 trimers (NAbs elicited against the N356-KO neoepitope) might reflect either inherent differences in the immunogenicity of the two newly created glycan hole epitopes or the greater occupancy of the N289 site for group 12 (82% versus 18%). If the latter explanation is correct, it would argue for the necessity to substantially block the immunodominant N289 site when diverting the NAb response elsewhere.

Analyses of the neutralization data are consistent with the observations that the N289 site was poorly occupied on the B41 SOSIP.v4.1-N289-KI trimers. Thus, sera from the 12 rabbits with high enough titers against the B41 parental virus to detect a reduction (ID₅₀ of >100 in parental column; all sera except r2272-12 in Fig. 13 to 15) had significantly lower RID₅₀ values against the B41 N289-KI pseudovirus than against the parental (P = 0.0005 by Wilcoxon’s signed-ranked test against the 100% RID₅₀ of the parental). Furthermore, for sera with ID₅₀ values of >100 against the single glycan-KO mutants (all 8 sera included in Fig. 14 and 15), the corresponding values against the double N289-KI + glycan-KO mutants were significantly lower (P = 0.0078 by Wilcoxon’s matched-pairs test).

Neutralization assays with heterologous tier 2 viruses.

All 40 of the BG505 trimer immunogen sera were tested against a global panel of 9 tier 2 pseudoviruses (61). Of the 360 combinations of rabbit sera and pseudoviruses, 60 yielded background-corrected ID₅₀ values of >40, but only two of them exceeded 100 (see Fig. A4A). The frequency of ID₅₀ values of >40 differed among immunization groups and viruses, but there was no identifiable concurrence between the absence or presence of a glycan on the trimer immunogens and the pseudoviruses (see Fig. A5).

The B41 trimer immunogen sera were tested against the same global panel of 9 tier 2 pseudoviruses (61) (see Fig. A4B). Somewhat higher titers were recorded than in the BG505 experiment (P = 0.0002), although the median ID₅₀ values were <20 in both cases. Of the 180 serum-virus combinations, 19 gave ID₅₀ values of >40 (50 to 550). The strongest and most consistent hits againt the tier 2 viruses were recorded for rabbit r2257-9 from group 9, which was immunized with the parental B41 trimer. The virus most frequently neutralized was Ce1176 (7 of 20 sera with ID₅₀ of >40, range of 50 to 550). Although viruses Ce1176 and X1632 both lack a glycan at the exact N356 position (they have others nearby; see Fig. A5), they were not detectably neutralized by the group 12 sera from rabbits immunized with the N356-KO trimer.

Thus, although there were indications that the B41 trimers are somewhat superior to their BG505 counterparts at inducing sporadic heterologous cross-neutralizing antibodies, the glycan-KI and -KO mutations did not enhance neutralization breadth in either experiment.

Production and characterization of BG505 and B41 SOSIP.v4.1 trimer mutants.

The BG505 mutants were all based on the SOSIP.v4.1 design that contains the E64K and A316W substitutions to the original SOSIP.664 sequence to further stabilize the trimers and suppress the immunogenicity of non-NAb epitopes (14, 27). The B41 mutants were likewise based on the SOSIP.v4.1 design but also contained an R315Q substitution in V3 to eliminate a low level of proteolytic clipping of this region that can occur with clade B trimers (13). For both genotypes, the trimers also contained a C-terminal D7324 epitope tag. The substitutions used to knock in and knock out various glycan sites are listed in Tables 1 and 2. Site-directed mutagenesis was performed by the QuikChange (Agilent Technologies) method, as described previously (51). For both genotypes, the trimers were expressed by transient transfection of 293F cells, followed by PGT145 bNAb affinity chromatography (13, 15). Each PGT145-purified variant trimer was verified to be >90% native-like when visualized by negative-stain electron microscopy, as described previously (15, 51, 62).

Immunization of rabbits.

Female New Zealand White rabbits were immunized intramuscularly with 30 μg of the BG505 or B41 SOSIP.v4.1 trimer variants essentially as described previously but with the exceptions outlined below (14, 27, 29). Four immunizations were carried out at weeks 0, 4, 20, and 40 (for BG505), or 38 (for B41), each involving the GLA-LSQ adjuvant obtained from IDRI, Seattle, WA, via the BMGF’s collaborative network. The rabbits were then reboosted at weeks 67 and 71 with the same trimers in an adjuvant formulation based on the Adjuplex product that was obtained as a gift from Cheng Cheng at the Vaccine Research Center of the National Institutes of Health, Bethesda, MD (63–65). Sera were obtained for analysis 2 weeks after each immunization. The data presented were derived from sera obtained at week 73, i.e., 2 weeks after the sixth immunization.

Ethics statement.

The rabbit immunization study was approved and carried out in accordance with protocols provided to the Institutional Animal Care and Use Committee (IACUC) at Covance Research Products (CRP), Inc. (Denver, PA), study number C0026-17. The rabbits were housed, immunized, and bled in compliance with the Animal Welfare Act and other federal statutes and regulations relating to animals and in adherence to the Guide for the Care and Use of Laboratory Animals of the National Research Council (45).

Glycan site occupancy determinations.

A proteomics-based method was used to quantify the occupancy (and proportion of oligomannose- and complex-type glycans) of N-linked glycans on the BG505 and B41 SOSIP.v4.1 trimer mutants (37). Briefly, glycopeptides were sequentially treated with Endo H and PNGase F in the presence of ¹⁸O-labeled water. The masses of the resulting peptides are specific to each glycan type or reflect nonoccupancy of the relevant site, thereby allowing the extent of occupancy to be quantified. These analyses were performed after completion of the rabbit immunization experiments described above.

Production and use of BG505.T332N and B41 env-pseudotyped virus mutants.

Mutant BG505 env genes containing glycan site substitutions to match those in the various SOSIP trimer immunogens were made as previously described (13–15) (Table 3). The BG505.T332N virus with a full-length cytoplasmic tail was used as the parental for mutant design, because it contains a knocked-in N332 glycan to match its sequence with the corresponding trimer immunogens (14, 15, 29). The B41 virus mutants are described in Table 4. All these viruses contain the R315Q substitution in V3, i.e., the B41.R315Q Env sequence was the parental for the other mutant viruses. The infectivity of each Env pseudovirus for Tzm-bl cells was determined by titration. The dilution factors of the mutant pseudoviruses relative to that of the parental yielding infectivity signals of ∼2 × 10⁶  (RLU) are given in Tables 3 and 4.

Neutralization of pseudoviruses was quantified in the TZM-bl cell assay, as described previously (13–15); rabbit sera were analyzed at Duke University Medical Center (DUMC), a subset of them also at the Academic Medical Center, Amsterdam (AMC), and monoclonal antibodies (MAbs) at Weill Cornell Medical College (WCMC). The dilution factors reducing infectivity by 50% (half-maximal inhibitory dilution factor, or ID₅₀) for each pseudovirus and the relative dilution factors (RID₅₀) for mutants compared with the parental BG505.T332N or B41.R315Q pseudovirus, as reported in Results, were obtained as described elsewhere (38). These values are based on two intra-assay replicates and the requisite repeats of the assay to fulfill the following criteria: the average RLU (relative light unit) values of virus control wells is >10 times the average RLU of cell control wells; the percent coefficient of variation between RLU in the virus control wells is ≤30%; the percent difference between duplicate wells is ≤30% for sample dilutions that yield at least 40% neutralization; neutralization curves cross the 50% neutralization value on the y axis 0 or 1 time; ID₅₀ values based on sigmoidal curve fits and linear interpolation between the two data points on either side of 50% are within 3-fold of each other; and the value of the positive control (NAb) is within a 3-fold range of the established mean. The rabbit sera were also tested against a panel of 9 tier 2 pseudoviruses, as previously described (61). The ID₅₀ values against murine leukemia virus (MLV) were subtracted from those for the test viruses to control for any cytoxicity.

Structural modeling.

The three-dimensional model of the BG505 SOSIP.v4.1 trimer was based on the PDB entry 5V8M crystal structure (66). The gp120 glycans were modeled as Man₅GlcNAc₂; gp41 glycans were either derived from PDB entry 5FUU or modeled as Man₅ GlcNAc₂ (67). The three-dimensional model of the B41 SOSIP.v4.1 trimer was based on our unpublished crystal structure of B41 SOSIP.664 at 3.5 Å. BG505 models were visualized by UCSF Chimera (68) and B41 models by Pymol (Schrödinger LLC, USA).

Statistics.

Groups of rabbits were compared pairwise by the two-tailed Mann-Whitney U test. The effects of mutations on neutralization by the respective sera were first compared by the Wilcoxon matched-pairs test. If the pairing was not significant, as assessed by Spearman correlations, they were then compared by Mann-Whitney U tests (used unless otherwise stated). The relative neutralization of N241-KI mutants of BG505 and N289-KI mutants of B41 was compared with the neutralization of parental virus, defined as 100%, by the Wilcoxon signed-rank test (all in Prism; GraphPad).