Members of the oligomannose-patch specific PGT128/130 bnAb family bind avidly to CRM197-conjugated glycomimetic NIT211, with affinities approximating those reported for recombinant HIV.
We first evaluated by ELISA the binding of oligomannose-specific bnAbs from the PGT128/130 family (PGT125, 126, 128, 130) to the CRM197-conjugated glycoside, dubbed NIT211. We had used these bnAbs previously for antigenic characterization of the BSA conjugate of the same glycoside5. All four antibodies bound NIT211 (4.1 glycans per CRM197) avidly (EC50 0.1 - 7 nM; Fig. 1A). As reported recently19, these antibodies bind NIT211 at least as good as the BSA conjugate loaded at similar density.
We then used recombinantly expressed Fab fragments in conjunction with Biacore SPR to determine binding affinities of the four PGT antibodies for NIT211. A biotinylated version of NIT211 (4.1 glycans per CRM197), captured onto the sensor surface by immobilized streptavidin, served as the ligand. Results show that the Fabs bind NIT211 with average affinities ranging from 271 nM (PGT130) to 678 nM (PGT128) (Table 1, Fig. 2A). The KD values determined here for PGT125 and PGT128 are similar to the reported monovalent KD values of these antibodies for a Man9-V3 glycopeptide (PGT125: 706 nM; PGT128: 326 nM) and the estimated monovalent KD of PGT128 for BG505 SOSIP (303 nM)20. Similar to others21, we observed biphasic sensorgrams for all the PGT antibodies despite the use of monovalent Fab molecules, suggestive of a bivalent binding interaction with NIT211 (Fig. 2B).
Oligomannose-specific bnAbs such as those from the PGT128/130 family interact with two or more glycans on HIV Env22,23. To evaluate the relationship between NIT211 glycan density and antibody binding avidity, and given the above-noted observations of biphasic antibody binding in SPR experiments, we assayed by ELISA binding of the PGT antibodies to NIT211 conjugated with the oligomannose mimetic at densities of 2.6, 4.1, and 6.2 glycans per CRM197. We first assessed binding of the antibodies formatted as IgGs. Consistent with our previous report5, antibodies of the PGT128/130 family bound increasingly avid to NIT211 with increasing glycan density (Fig. 3A), with apparent maxima at a density of ~6 glycosides. Strikingly, the relative binding affinity of the Fab fragments of antibodies PGT125, 126 and 128 also increased substantially upon glycan density (Fig. 3B) suggesting a bivalent or greater interaction of these Fab fragments to NIT211 ligands. Taken together with the SPR analyses, these results suggest that presentation of the oligomannose mimetic at a density of 4-6 molecules per CRM197 reasonably approximates oligomannose presentation on HIV gp120 conducive to the binding of at least the PGT128/130 family of bnAbs.
Oligomannose patch-specific bnAbs outside of the PGT128/130 bnAb family bind less avidly to CRM197-glycoconjugate NIT211.
Having established that our lead glycomimetic conjugated to CRM197 retains favorable antigenicity for members of the PGT128/130 bnAb family as compared to the BSA conjugate, we next evaluated binding of representatives from other bnAb families specific for the oligomannose patch (BF520.1, BG18, PCDN-33A, PGDM12, PGDM21, VRC41.01)20,24–27. We found that bnAbs BG18, PGDM21, and VRC41.01 bind NIT211 (4.1 glycans per CRM197) with reasonably high avidity (EC50 14 - 70 nM) (Fig. 1B), albeit less than the PGT128/130 bnAbs. In contrast, bnAbs BF520.1, PCDN-33A and PGDM12 bound the CRM197 conjugate very poorly (EC50 1 - 5 μM) (Fig. 1B). Binding of these antibodies as Fab fragments was not detectable by SPR. Overall, these results likely reflect the varied degrees of specificity of the bnAbs for oligomannose, their capacity to engage other glycans, and their relative dependence on interaction with the V3 protein backbone. Nevertheless, the results suggest that the oligomannose mimetic and its presentation on CRM197 sufficiently imitates the binding epitope of not just the PGT128/130 bnAb family but also a few other oligomannose-specific bnAbs.
CRM197-conjugated glycoside mimic of oligomannose is bound also by inferred germline (gl) precursors of anti-HIV oligomannose-specific bnAb families.
A prevailing thesis in the HIV vaccine field currently is that an immunogen that can be bound with sufficient affinity in vitro by gl precursors of existing bnAbs might be better able to activate the appropriate B cells in vivo to yield similar bnAbs (reviewed in ref.1). We therefore assessed the binding of published inferred gl precursors from four oligomannose-specific bnAb families (BF520.1, BG18, PCDN-33A and PGT128/130) for binding in ELISA to NIT211 (4.1 glycans per CRM197). As shown in Fig. 1C, all four gl antibodies bound NIT211, albeit with the expected lower avidity (EC50 0.1 - 5 µM) compared to their mature cognates. Fab fragments of these gl antibodies did no bind detectably to NIT211 by SPR. Recognition of the conjugated oligomannose mimetic by various oligomannose bnAbs and, most importantly, some of their respective gl precursors provided encouragement in NIT211’s potential to trigger B cells with nominal specificity for oligomannose.
CRM197-conjugated glycoside NIT211 elicits glycan-specific antibodies in human-antibody transgenic Trianni mice.
Having determined that the NIT211 glycoconjugate is bound reasonably well by several oligomannose-specific bnAbs and a few gl versions thereof, we next sought to identify a possibly optimal adjuvant formulation for the elicitation of the desired glycan-specific antibody response. We chose to conduct our immunizations in the human-antibody transgenic Trianni mouse system to allow for an approximation of possible antibody responses in people. Before commencing, we confirmed that these mice do not develop an abnormal antibody response by subcutaneously administering a priming immunization with the model antigen KLH formulated in Alhydrogel plus CpG ODN1826. We found total IgG responses in Trianni mice to be as robust as similarly primed wild‐type C57BL/6 mice at one month post-immunization (Supplementary Fig. 1A), albeit that the IgG1 and IgG2c responses in Trianni were notably stronger (Supplementary Fig. 1B). Total T‐follicular helper (Tfh) cell frequencies (B220-CD4+CD44hiPD1+CXCR5+) in draining inguinal lymph nodes at day 8 post prime (4.9-13.1 %) were also similar to the average frequencies reported in the draining popliteal lymph nodes of non-transgenic mice28 after a single footpad immunization with KLH formulated with alum (Supplementary Fig. 2).
We then assessed and compared the immunogenicity of NIT211 formulated in three different types of adjuvants: Alhydrogel, AddaVaxTM, and GLA-SE. The NIT211 conjugate used for immunization carried an average of 6.2 glycosides per CRM197 molecule, which, as shown in Fig. 3A, was bound strongly by the PGT antibodies. Mice (n=5 per group) were primed at day 0 and boosted at days 21, 42 and 105, a schedule modeled on glycoconjugate vaccine schedules in humans29,30, and sera collected on days 0, 10, 28, 49, and 119.
All mice immunized with the GLA-SE formulation mounted a rapid IgG response to the CRM197 carrier protein by day 10 after priming (Supplementary Fig. 3), consistent with previous reports19 and the expected high immunogenicity of the carrier. The magnitude of the antibody response was somewhat slower in mice immunized with the Alhydrogel and AddaVax formulations. Even so, IgG levels in all three animal groups plateaued following the second boost (day 49; Supplementary Fig. 3). Overall, mice immunized with the GLA-SE formulation produced the highest IgG response to the CRM197 carrier protein, followed by the AddaVax and Alhydrogel groups, respectively.
To measure antibody responses to the oligomannose mimetic without interference from antibodies specific for the CRM197 carrier, we used the BSA-conjugated version of the glycoside (dubbed NIT82b5), which was conjugated at 4 glycosides per BSA molecule. We confirmed that none of the sera from the final bleed (day 119) bound unconjugated BSA (Supplementary Fig. 4), meaning that any measured binding to NIT82b should be due to the conjugated glycoside. Unexpectedly, we found that only mice immunized with the GLA-SE adjuvanted conjugate produced IgG that bound the oligomannose mimetic as presented on BSA (Fig. 4A). We found no evidence of glycoside-specific IgM antibodies in sera from any of the three groups of immunized animals (Supplementary Fig. 5). The level of IgG binding varied among the GLA-SE-immunized mice; two of the animals produced reasonably good anti-glycan antibodies, a third produced modest levels and two appeared to not produce much antibody (Fig. 4B). Strikingly, IgG3 was the most prevalent IgG subclass in the response of GLA-SE immunized animals to the oligomannose mimetic, whereas IgG1 and IgG2 (IgG2b and IgG2c) were the most prevalent against the CRM197 carrier (Fig. 4C).
In sum, the results above show that NIT211 formulated in GLA-SE induced the greatest antibody response to the glycomimetic compared to formulations with Alhydrogel or Addavax.
Anti-glycan serum antibodies elicited by CRM197-glycoconjugate NIT211 in GLA-SE bind recombinant HIV gp140 trimers but are unable to exert neutralizing activity.
We next assessed whether the elicited anti-glycan antibody response could recognize glycans on HIV. First, we measured serum IgG binding to a panel of SOSIP-based HIV gp140 trimers by ELISA. Unexpectedly, notable IgG binding was observed with the sera of all five animals in the GLA-SE group (Fig. 5A), including sera from the two animals that had bound only marginally to the BSA conjugate. This unexpected result prompted us to also assay sera from the animals immunized with the Alhydrogel and Addavax formulations. Sera from these animals did bind somewhat to a selection of SOSIP trimers, but this binding was similar to the binding of sera from unimmunized mice to the trimers, suggesting that it may have been caused by naturally occurring anti-glycan antibodies. We observed binding of all the GLA-SE sera to the B41 SOSIP.v4.1 trimer (Fig. 5A), which was chosen for this analysis.
IgG subclass analyses revealed that whereas IgG3 was the predominant IgG subclass responsible for binding to the oligomannose mimetic in sera from animals immunized with the GLA-SE formulation (cf. Fig. 4C), serum antibodies binding to the SOSIP trimers were predominantly IgG2 (Fig. 5B). These results suggest that the immunization with NIT211 yielded two different anti-glycan responses; one, mainly of the IgG3 subclass, specific only for the synthetic glycoside, and the other, mainly of the IgG2b/c subclass, capable of binding glycans on HIV.
To determine whether the NIT211 + GLA-SE formulation had truly stimulated antibodies with capacity to bind the high-mannose patch on HIV gp120, we performed an inhibition ELISA assay in which we measured the binding of bnAb PGT128 to B41 SOSIP trimer after incubation with GLA-SE sera vs immune sera from animals immunized with KLH, which itself carries a large number of high mannose-type glycans31. We observed a notable decrease in PGT128 residual binding following incubation with the NIT211 + GLA-SE sera compared to KLH immune sera (Fig. 5C). These results show that NIT211, when formulated with GLA-SE, had indeed elicited antibodies capable of binding glycans within or neighboring the high-mannose patch.
Given the observed binding of the NIT211 + GLA-SE sera to SOSIP trimers, we assessed the sera for neutralizing activity against a small panel of HIV-1 strains from different subtypes in the Monogram PhenoSense neutralization assay. However, we observed no significant neutralizing activity relative to unimmunized animals even at the lowest serum dilution tested (1:30) (Fig. 5D). We conclude from these results that the elicited antibodies are of insufficient affinity (or avidity) to bind glycans on the virus or at least not enough to exert neutralizing activity.