New Member of the V1V2-Directed CAP256-VRC26 Lineage That Shows Increased Breadth and Exceptional Potency

Study subject.

Centre for the AIDS Programme of Research in South Africa (CAPRISA) participant CAP256 was enrolled into the CAPRISA Acute Infection Study (30) that was established in 2004 in KwaZulu-Natal, South Africa, for follow-up and subsequent identification of HIV seroconversion. CAP256 was one of the seven women in this cohort who developed neutralization breadth (31). The CAPRISA 002 Acute Infection Study was reviewed and approved by the research ethics committees of the University of KwaZulu-Natal (E013/04), the University of Cape Town (025/2004), and the University of the Witwatersrand (MM040202). CAP256 provided written informed consent for study participation. Samples were drawn between 2005 and 2009.

B cell cultures.

Peripheral blood mononuclear cells (PBMCs) isolated from CAP256 blood draws at week 193 were stained with LIVE/DEAD Fixable Aqua (Invitrogen), CD19-Cy7-phycoerythrin (PE), IgM-Cy5-PE, IgD-PE, CD16-Pacific Blue, and CD3-Cy7-allophycocyanin (APC) (BD Pharmingen). The IgD-negative (IgD⁻) IgM-negative (IgM⁻) B cells were bulk sorted on a BD FACSAria II flow cytometer as described previously (19). Cells were plated at 2 B cells per well in 384-well plates and cultured for 14 days in the presence of CD40L-expressing irradiated feeder cells, interleukin-2 (IL-2) (Roche), and interleukin-21 (IL-21) (Gibco), as described previously (32, 33). Culture supernatants were screened by microneutralization against HIV-1 ZM53.12 and CAP210.E8 Env pseudoviruses as described in reference 34.

Expression and purification of trimeric HIV-1 Env BG505 SOSIP.664 probes.

An Avi tag (GLNDIFEAQKIEWHE) was inserted at the C terminus of the previously described construct BG505 SOSIP.664.T332N gp140, referred to here as BG505 SOSIP (35). For BG505 SOSIP.K169E-Avi, lysine (K) 169 was mutated to a glutamic acid (E) by site-directed mutagenesis. Both constructs were transfected and purified as described previously (12). Briefly, HEK 293 cells were cotransfected with BG505 SOSIP and furin plasmid DNAs at a 4:1 ratio. Transfection supernatants were harvested and purified through either a VRC01 or a 2G12 antibody affinity column. Proteins were eluted with 3 M MgCl₂, 10 mM Tris, pH 8.0. The eluate was concentrated and applied to a Superdex 200 column equilibrated in 5 mM HEPES, pH 7.5, 150 mM NaCl, 0.02% azide or phosphate-buffered saline. The fractions corresponding to the trimeric proteins were pooled, concentrated, flash-frozen in liquid nitrogen, and stored at −80°C. The proteins were biotinylated at the Avi tag sequence using BirA ligase and then conjugated to streptavidin-APC or streptavidin-PE (Invitrogen) as described previously (36).

The probes were assessed for the correct antigenicity by binding to antibody-coated beads. Anti-mouse kappa light chain beads (Becton Dickinson, San Jose, CA) were incubated with anti-human IgG (Becton Dickinson), washed, incubated with 2 μg/ml antibody CAP256-VRC26.09, PGT128, or F105, and washed again. BG505 SOSIP-streptavidin-APC or BG505 SOSIP.K169E-streptavidin-PE (1.5 μg) was then incubated with the beads, and the beads were washed and analyzed on an LSR II flow cytometer (Becton Dickinson).

Cell sorting.

PBMCs from week 159 were stained for IgG-positive (IgG⁺) B cells using LIVE/DEAD Fixable Aqua (Invitrogen), IgG-fluorescein isothiocyanate, CD19-Cy7-PE, IgM-Cy5-PE, CD14-PE-Texas Red (ECD), CD4-ECD, CD3-Cy7-APC, and CD8-Brilliant Violet 711 (BD Pharmingen). At the same time, they were stained with BG505 SOSIP-streptavidin-APC and BG505 SOSIP.K169E-streptavidin-PE. Cells that were BG505 SOSIP-streptavidin-APC positive (APC⁺) IgG⁺ CD19⁺ were sorted using a BD FACSAria II flow cytometer into single wells of a 96-well plate containing RNase inhibitor (New England Biolabs), SuperScript reverse transcriptase buffer (Invitrogen), dithiothreitol, and Igepal as described previously (36).

Isolation and expression of CAP256-VRC26 family genes.

Kappa and lambda light chain gene and IgG heavy chain gene variable regions were amplified from neutralization-positive B cell culture wells as described in reference 33 or from 96-well sort plates as described in reference 36. Briefly, cells were lysed and subjected to reverse transcription-PCR (RT-PCR) as described in reference 37 using a modified primer set (see Table S1 in the supplemental material). These primers were developed by our group or were described previously (see Table S1 in the supplemental material) (37–39). Briefly, cells were lysed and subjected to RT-PCR as described previously (37) using a modified primer and a round of nested PCR with 3 different multiplex primer pools, which were used to amplify either the heavy chains or the light chains. The amplicons were subcloned, expressed, and purified as described in reference 19. For antibodies CAP256-VRC26.13 to CAP256-VRC26.21, wells that previously yielded VRC26-lineage lambda chains but no heavy chains were subjected to RT-PCR using IgA-specific 3′ primers. All heavy chains were subcloned and expressed as IgG1 regardless of the class of the original amplicon.

Neutralization assays.

Single-round-of-replication Env pseudoviruses were prepared, titers were determined, and the pseudoviruses were used to infect TZM-bl target cells as described previously (40, 41). The neutralization breadth of CAP256-VRC26.25, CAP256-VRC26.26, and CAP256-VRC26.27, as well as that of previously described antibodies, was determined using a previously described panel (19, 36) of up to 200 geographically and genetically diverse Env pseudoviruses representing the major subtypes and circulating recombinant forms. The remaining CAP256-VRC26 antibodies were assayed on a 46-virus subset of this panel, along with the autologous virus strains CAP256.2.00.C7J (a primary infecting [PI] strain) and CAP256.206.c9 (a superinfecting [SU] strain). The data were calculated as a reduction in the number of luminescence units compared with the number for the control wells and are reported as the 50% inhibitory concentration (IC₅₀; in micrograms per microliter) for monoclonal antibodies (MAbs).

N160 glycan mutant Env pseudoviruses were generated by site-directed mutagenesis as described previously (28, 42). All N160 glycan mutants had the N160K mutation, except for ConC, which had the N160A mutation. The N156 mutants (this study) were generated as N156A mutants by site-directed mutagenesis (GeneImmune, New York, NY). The BG505 pseudovirus had the wild-type sequence (threonine) at position 332, unlike the BG505 SOSIP probe, which bears a T332N mutation (35).

Neutralization and autoreactivity data for CAP256-VRC26.25 were generated with a variant that had the amino acid change K126Q in the light chain, which was inserted during subcloning. We confirmed that this single amino acid change did not alter the neutralization data.

Autoreactivity.

CAP256-VRC26.01 to CAP256-VRC26.32 and UCA antibodies were assayed at 25 and 50 μg ml⁻¹ for autoreactivity to HEp-2 cells (Inverness Medical Professional Diagnostics) by indirect immunofluorescence and for binding to cardiolipin as described previously using a Quanta Lite ACA IgG III assay (INOVA Diagnostics Inc., San Diego, CA) (43). The UCA and seven of the antibodies with the broadest activity were further evaluated on a panel of autoantigens using an AtheNA Multi-Lyte ANA-II Plus test system (Alere, Orlando, FL) to detect semiquantitatively IgG antibodies to 9 separate analytes, SSA, SSB, Sm, RNP, Scl-70, Jo-1, centromere B, double-stranded DNA (dsDNA), and histone, as reported previously (43); values of >120 at 25 μg ml⁻¹ were considered positive (43).

Next-generation sequencing analysis.

454 pyrosequencing was performed on PBMCs from week 34, the results for which are reported here, as well as PBMCs from weeks 38, 49, 59, 119, 176, and 206, the results for which were described previously (21). mRNA was prepared from 10 million to 15 million PBMCs as follows: cells were lysed in 600 μl buffer RLT (Qiagen) per 10 million PBMCs and run over a QIAshredder (Qiagen). The flowthrough was applied to the DNA column from an Allprep kit (Qiagen), and that flowthrough was used for subsequent purification using an Oligotex kit (Qiagen). cDNA was synthesized using SuperScript II reverse transcriptase (Invitrogen) and oligo(dT)_12–18 primers with incubation at 70°C for 1 min, chilling on ice, and then synthesis at 42°C for 2 h. Individual PCRs were performed with Phusion polymerase (Thermo) for 30 cycles. Primers (21) consisted of pools of 5 to 7 oligonucleotides specific for all lambda chain gene families or for VH3 family genes and had adapters for 454 next-generation sequencing. For PBMCs from week 176 only, heavy chain PCR was performed with primers for all VH families and mixed lambda chain and kappa chain primers were used for the light chain (21). PCR products were gel purified (Qiagen). Pyrosequencing of the PCR products was performed on a GSFLX sequencing instrument (Roche-454 Life Sciences, Bradford, CT, USA) on a half chip per reaction (full chips were used for PBMCs from week 176). On average, ∼250,000 raw reads were produced.

Next-generation sequencing data from each time point were processed through an Antibodyomics pipeline (https://github.com/scharch/SONAR) as previously described (21, 44). Briefly, a series of Python scripts was used to check transcripts for appropriate length (300 to 600 nucleotides), provide calls for germ line V and J gene assignments by BLAST analysis, and check for in-frame junctions and open reading frames. Reads for which the V and J genes were successfully assigned and that had an in-frame junction and no stop codons were clustered using the CDHit program (45) at 97.25% identity, and singletons were discarded. In order to better account for possible errors introduced during sequencing, we restricted our analysis to sequences appearing at least twice in any one data set. This resulted in approximately 50% fewer final heavy chain sequences than we previously reported, but the phylogenetic structure of the lineage remained the same. In addition, we used a CDRH3 signature contained in all previously reported sequences as a less computationally intensive way of finding related heavy chain sequences. Thus, heavy chain sequences that matched the CAP256-VRC26 VH and JH assignments and that had CDRH3s that were at least 30 amino acids long and that had a YY motif were selected and combined across all time points. Light chains that matched the CAP256-VRC26 VL and JL assignments and that had a light chain complementarity-determining region 3 (CDRL3) with at least 92% sequence identity to a known CAP256-VRC26 family member were manually inspected, and those with a recombination pattern matching that of the known antibodies were combined across all time points. Selected heavy and light chains from all time points were reclustered so that sequences observed at multiple time points are represented only once in the final data set. A final manual inspection was used to remove sequences with apparent PCR crossover or homopolymer indel errors.

Heavy and light chain trees were built using the FASTML program (46), and the light chain tree was manually edited to match the heavy chain arrangement as previously described (21). Ancestral sequences were inferred from these trees using the DNAML program (47) as previously described (21).

CAP256-VRC26.25 Fab crystallization and structure determination.

Fab was prepared by inserting an HRV3C recognition site (GLEVLFQGP) after Lys 235. Purified IgG was incubated with HRV3C protease overnight at 4°C, and the digested protein was passed over protein A agarose to remove the Fc fragment and subsequently purified over a Superdex 200 gel filtration column. The Fab preparation was then screened against 576 crystallization conditions using a Mosquito crystallization robot. Initial crystals were grown by the vapor diffusion method in sitting drops at 20°C by mixing 0.1 μl of protein with 0.1 μl of reservoir solution. Crystals were manually reproduced in hanging drops by mixing 1.0 μl protein complex with 1.0 μl reservoir solution. Crystals grown in a reservoir solution of 23% polyethylene glycol (PEG) 8000 and 0.1 M HEPES, pH 7.5, were flash frozen in liquid nitrogen with 20% PEG 400 as a cryoprotectant. Diffraction data were collected at a wavelength of 1.00 Å at the SER-CAT beamline ID-22 (Advanced Photon Source, Argonne National Laboratory). All diffraction data were processed with the HKL2000 suite (48), and model building and refinement were performed in the COOT (49) and PHENIX (50) programs, respectively. Ribbon diagram representations of protein crystal structures were made with the PyMOL program (51), and electrostatics were calculated and rendered with the UCSF Chimera program (52). Data collection and refinement statistics are shown in Table S2 in the supplemental material.

Amino acid frequency analysis.

The resistance score for a given amino acid (or a gap) was defined as the ratio of its number of occurrences in sequences from resistant strains to its overall number of occurrences for the given residue position (53). A higher score indicates that the amino acid was preferentially found among sequences from resistant strains, with a score of 1 indicating that the amino acid was found only among sequences from resistant strains. Associations were analyzed using Fisher's exact test with the Bonferroni correction for multiple comparisons. Alignments were generated using the MUSCLE algorithm implemented in Geneious software, version 8.1.7 (54). Logo plots were generated with the Weblogo application (http://weblogo.berkeley.edu/) (55) and manually colored by the use of Inkscape software (https://inkscape.org).

Accession numbers.

Sequences from this study are available in GenBank under the following accession numbers: for the heavy and light chain sequences for CAP256-VRC26.13 to CAP256-VRC26.33, GenBank accession numbers KT371076 to KT371117; for 454 pyrosequencing data sets from CAP256 at week 34, Sequence Read Archive accession numbers SRR2126754 and SRR2126755; and for pyrosequencing data sets of bioinformatically identified lineage members from week 34, GenBank accession numbers KT371118 to KT371320. Coordinates and structure factors for CAP256-VRC26.25 have been deposited with the Protein Data Bank (PDB) under accession number 5DT1.

Isolation of new antibodies.

To further understand the CAP256-VRC26 lineage and in an attempt to find broader and more potent members, we isolated antibodies from time points of peak serum neutralization potency and breadth (Fig. 1A). We used several methods to isolate an additional 21 antibodies. These included repeated PCR amplifications with cDNA from wells from the original B cell cultures, using the new primers described below; a B cell culture of PBMCs from week 193 (Fig. 1B); and single-cell sorting with antigen-specific probes (Fig. 1C), using a PBMC sample from week 159 (at the peak of neutralization breadth in plasma [28, 29]). The recovered antibodies are named as follows: donor-lineage.clone, for example, CAP256-VRC26.25. For brevity, in some figures we refer to them as donor.clone, for example, CAP256.25.

(i) B cell cultures.

The B cell culture method used here involves sorting of IgM⁻ IgD⁻ B cells, followed by culture for 2 weeks and then microneutralization assays of culture supernatants (32, 34) (Fig. 1B). Previous B cell cultures from donor CAP256 (21) yielded multiple wells from which VRC26-lineage lambda chains, but not IgG heavy chains, were recovered. We used IgA-specific 3′ primers (39) with new sets of multiplex 5′ primers (see Materials and Methods and Table S1 in the supplemental material) and recovered nine VRC26-lineage heavy chains from these wells. Some of these may have derived from IgA⁺ B cells, as IgA⁺ B cells were not excluded in the sorting. Unexpectedly, two wells yielded both IgG and IgA amplicons with identical variable regions but distinct constant regions (Fig. 2A). We concluded that class switching had occurred in the well during the 2-week culture. The combination of CD40L and IL-21 was previously shown to support IgG-to-IgA class switching in human B cells (56–58). We therefore made the assumption that all such wells originally contained IgG⁺ B cells. All antibody sequences were subcloned and expressed as IgG1. This strategy yielded 9 antibodies (CAP256-VRC26.13 to CAP256-VRC26.21) derived from the PBMCs from weeks 119 and 206. We subsequently performed a B cell culture with B cells derived from week 193 and, using the new primer sets described in the Materials and Methods and Table S1 in the supplemental material, recovered four additional antibodies (CAP256-VRC26.22 to CAP256-VRC26.25).

(ii) Single-cell sort with trimer probes.

The recent development of the trimer mimic BG505 SOSIP (35) provides a probe that can be used to sort B cells expressing antibodies to quaternary epitopes (59). The broader antibodies of the CAP256-VRC26 lineage neutralize BG505, and thus, we chose to use this probe with the goal of isolating lineage members with broader neutralization capacities (Fig. 1C). To increase the specificity of the probe for the CAP256-VRC26 lineage, we assessed potential mutations that would knock out V1V2 antibody binding and neutralization: using an Env pseudovirus of BG505, we made a variant with a deletion of the N160 glycan (the N160K mutant) and a variant with a mutation in V2 strand C (the K169E mutant). While the N160K mutation rendered BG505 resistant to PG9-, CH01-, and PGT145-lineage antibodies, neutralization by CAP256 plasma and CAP256-VRC26.09 was more affected by the K169E mutation than by the N160K mutation (Fig. 2B). We therefore made a K169E mutation in the BG505 SOSIP trimer. Wild-type and mutant molecules were fluorescently labeled, and their antigenicity was assessed by binding them to beads coated with known antibodies (Fig. 2C). The BG505 SOSIP wild-type probe bound strongly to PGT128 and CAP256-VRC26.09 and did not bind to the weakly neutralizing antibody F105. The K169E mutant had similar binding, but as expected, it did not bind to CAP256-VRC26.09. As a control, YU2-gp140-foldon, which does not present quaternary epitopes (60, 61), was observed to bind PGT128 and F105 but not CAP256-VRC26.09 (Fig. 2C). Using the BG505 SOSIP probe pair, we sorted single IgG⁺ B cells from week 159 (Fig. 2D), a time point at which the donor's plasma showed peak neutralization breadth and potency (28, 29). We amplified and subcloned Ig genes from cells that were positive for the wild-type probe but did not bind the K169E probe. Fourteen wells yielded heavy chain amplicons, light chain amplicons, or both. In 11 of the 14 wells, the amplicon(s) matched the VRC26 lineage, and 8 such wells yielded heavy chain-light chain pairs. All eight pairs were expressed and had neutralizing activity, thus producing lineage members CAP256-VRC26.26 to CAP256-VRC26.33.

Characteristics of the new CAP256-VRC26 antibodies.

The 21 new CAP256-VRC26 antibodies showed a range in the levels of nucleotide mutations from the sequences of the germ line genes: mutations in 4.2 to 18% of the nucleotides compared with the sequences of the VH3-30*18 alleles and 2.5 to 15% of the nucleotides compared with the sequences of the VL1-51*02 alleles (Table 1). Similarly, they varied in the rates of mutations from the inferred sequence of the UCA: 7 to 23% of the nucleotides compared with the sequence of the UCA heavy chain and 3.6 to 17% of the nucleotides compared with the sequence of the UCA lambda chain (see Table S3 in the supplemental material). All had the long CDRH3s that are characteristic of this lineage (Table 1; see Table S3 in the supplemental material): 35 or 37 amino acids, as seen previously (21), or 36 amino acids in the case of CAP256-VRC26.25. As was observed with the initial set of MAbs, the new MAbs showed no or marginal autoreactivity when tested in multiple assays (see Fig. S1 in the supplemental material).

Like the original 12 lineage members, the new CAP256-VRC26 antibodies showed a range of neutralization breadth and potency. With a 46-virus multiclade panel, their neutralization breadth varied from 2% to 63% (Table 1; see also Table S4 in the supplemental material). The member of the lineage with the broadest neutralization breadth and the most potency was CAP256-VRC26.25, which had a median IC₅₀ of 0.003 μg/ml for this panel, which makes it 10-fold more potent than the most potent family members described previously (21). Two others, CAP256-VRC26.26 and CAP256-VRC26.27, also had broader neutralization than previous relatives but were not as potent as CAP256-VRC26.25. The antibodies CAP256-VRC26.19, CAP256-VRC26.22, and CAP256-VRC26.29 were similar to CAP256-VRC26.08 in neutralization breadth, neutralizing 46% of this panel (Table 1).

We also isolated antibodies that showed less neutralization breadth and potency than previously isolated family members. The sequence of the antibody with the lowest level of somatic mutation, CAP256-VRC26.24, showed only 4.2% divergence from the sequence of the heavy chain germ line and 2.5% from the sequence of the light chain. Its neutralization capacity was less broad than that of other early lineage members and it was less potent than other early lineage members, such as CAP256-VRC26.01 (Table 1). Other poor neutralizers included CAP256-VRC26.13; notably, it differed from CAP256-VRC26.12 by only 2 nucleotides or a single amino acid, even though they originated from different culture wells. An independent well yielded sequences that were 100% identical to the sequence of CAP256-VRC26.13. While these closely related lineage members were not broadly neutralizing, they did potently neutralize the autologous CAP256 superinfecting (SU) strain (see Table S4 in the supplemental material).

Like the previously identified members of the lineage (21), the new CAP256-VRC26 antibodies are highly dependent on a quaternary epitope and target V1V2. We assayed for binding to gp120 derived from HIV-1 strain CAP210, which is potently neutralized by all lineage members except CAP256-VRC26.20. None of the antibodies bound to this gp120 (see Fig. S2 in the supplemental material). CAP256-VRC26.25 was further tested on 10 additional clade C gp120s, with no binding being observed; however, it bound strongly to the BG505 SOSIP molecule, confirming the trimer preference of these antibodies. In addition, none of the 33 antibodies were able to neutralize R166A and K169E mutants in HIV-1 strain ConC (see Fig. S3A in the supplemental material), BG505.w2, or ZM53.12, and the 33 antibodies showed modest effects on ConC with mutations at amino acids 167 and 168, similar to the donor CAP256 plasma (28). Thus, the epitope recognized by the new CAP256-VRC26 antibodies is similar to that recognized by the previously described relatives.

Glycan independence of CAP256-VRC26 antibodies.

The PG9-like broadly neutralizing antibodies (PG9/PG16, CH01-04, and the PGT145/PGDM1400-family antibodies) target a V1V2 epitope with quaternary epitope specificity (20, 22, 23, 59). These antibodies have been shown to require both glycan and protein contacts on Env (62, 63), with a strict dependence on the glycan at N160 (20, 22, 23, 42) and with contacts to other glycans, typically at N156 (62). We therefore investigated the glycan requirements of the CAP256-VRC26 antibodies. Removing the glycan at N156 did not abrogate neutralization; it conferred a slight decrease in potency against HIV-1 strains BG505 and ZM53 and variable effects against Q461.e2, similar to the effect that we observed with PG9 (see Fig. S3B in the supplemental material). However, removing the glycan at N160 had more dramatic effects. As expected, V1V2-directed antibodies PG9, PG16, CH01, and PGT145 were unable to neutralize glycan N160 mutants (Fig. 3A). In contrast, CAP256-VRC26 antibodies were affected in a potency-dependent manner. The N160 mutations had little effect on the susceptibility of viruses that were potently neutralized by CAP256-VRC26 antibodies but did reduce or abrogate the sensitivity of viruses that were less potently neutralized (Fig. 3B). When the data for all antibodies were combined, wild-type viruses with an IC₅₀ of >0.1 μg/ml were more likely to show a neutralization reduction of at least 10-fold than the more potently neutralized viruses (P = 0.029, Fisher's exact test). Thus, the CAP256-VRC26 antibodies can maintain potent neutralization of viruses, despite the lack of the N160 glycan.

Phylogenetic analysis of the lineage.

To deepen our understanding of the CAP256-VRC26 lineage, we placed the 12 previously cloned antibodies, together with the 21 new lineage members, within the context of total B cell transcripts. A revised algorithm (see Materials and Methods) was used to analyze 454 pyrosequencing-derived sequences from week 34 (64) and reanalyze all sequences from weeks 38 to 206 (21). Phylogenetic trees derived from maximum likelihood analysis of the transcripts of all members of the lineage, along with the 33 cloned antibody sequences, are displayed in Fig. 4. As noted previously (21), the tree for heavy chain sequences bifurcates into an upper branch and a lower branch. Sequences that map to the upper branch are plentiful among B cell transcripts from weeks 34 to 59 but are rare among B cell transcripts obtained at week 119 or later. In contrast, the lower branch contains sequences from all time points. One notable feature of the sequences in the lower branch is a Cys-Cys which may stabilize the CDRH3 structure (21); it appears at week 49 and sequences from all subsequent times of the lower branch. Week 34 sequences do not contain Cys-Cys and appear in both the upper and lower branches of the tree.

The placement of the cloned antibodies on these trees led to several observations. The CAP256-VRC26.24 sequence is 4% mutated from the germ line sequence, clustered with week 59 sequences, and is located in the top (CAP256-VRC26.01) branch, yet it was isolated from week 193 PBMCs; we speculate that it is derived from a long-lived memory cell. CAP256-VRC26.25, the antibody with the broadest neutralization and greatest potency, has a 1-amino-acid insertion in the CDRH3 sequence compared to the UCA sequence (see Table S3 in the supplemental material) and lacks close relatives in the pyrosequencing data set, and its closest relatives are weakly neutralizing. Six of the eight antibodies that were derived from the probe sort, including the CAP256-VRC26.26 and CAP256-VRC26.27 antibodies with very broad neutralization capacities, cluster with other broad neutralizers, CAP256-VRC26.08, CAP256-VRC26.09, and CAP256-VRC26.22; all of these bear a 2-amino-acid insertion in the CDRH3 sequence compared to the UCA sequence. However, weak neutralizers CAP256-VRC26.20, CAP256-VRC26.30, and CAP256-VRC26.31 also appear in this cluster and have this insertion. Thus, the antibodies that are the most closely genetically related do not always display the same level of neutralizing activity.

CAP256-VRC26.25 structure.

To investigate the differences in neutralization potency and breadth among these antibodies, we solved the crystal structure of the CAP256-VRC26.25 antigen-binding fragment (Fab) to 2 Å. The structure was well defined, including the electron density for the entire CDRH3. Like the other lineage members, this antibody has a long CDRH3 projecting away from the body of the Fab (Fig. 5A). A disulfide bond was observed near the base of the CDRH3, as expected, and density confirming the two computationally predicted sulfated tyrosines was also visible (Fig. 5B). We compared this structure to the structures solved previously for seven other lineage members; of these, CAP256-VRC26.03 had a complete CDRH3 structure, while several others contained disordered regions of various lengths in the CDRH3 (21). Notably, the CDRH3 of CAP256-VRC26.25 bends in a considerably different direction than the others. The angle of rotation between the CDRH3s of Fabs CAP256-VRC26.03 and CAP256-VRC26.25, aligned by framework regions, is 79 degrees (Fig. 5B). When all structures are overlaid (Fig. 5C), we observe that the UCA and the early antibody CAP256-VRC26.01 bend in a similar direction, CAP256-VRC26.03, CAP256-VRC26.04, CAP256-VRC26.06, CAP256-VRC26.07, and CAP256-VRC26.10 bend in nearly the opposite direction, and CAP256-VRC26.25 bends in a direction different from the direction of all of these. The angle likely depends on the identity of amino acid at position 100V or 100W. The Gly of CAP256-VRC26.25 inserted at the base of the CDRH3 allows additional flexibility in the initial projection of the CDRH3, providing a different range of angles for the antibody to initially engage the viral spike (Fig. 5D). Thus, stabilization of the overall CDRH3 structure combined with a local flexibility at the base may contribute to the improved neutralization breadth and potency of CAP256-VRC26.25 compared to those of the other lineage members.

CAP256-VRC26.25 has broad neutralization and is highly potent.

CAP256-VRC26.25 is the most potent member of the lineage. On a multiclade panel of 183 Env pseudoviruses, it neutralized 57% of isolates with a median IC₅₀ of 0.001 μg/ml (Fig. 6; see also Table S5 in the supplemental material), which makes it 10-fold more potent than the previously published CAP256-VRC26.08 variant. We also compared CAP256-VRC26.08 and CAP256-VRC26.25 to the V1V2-directed antibodies PGDM1400, PG9, and CH01 (Fig. 6A) and additional antibodies to other neutralization epitopes (Fig. 6B and C; see also Table S5 in the supplemental material). A remarkable characteristic of CAP256-VRC26.25 is the highly potent neutralization of some viruses. This can be seen as the leftward shift of the potency-breadth plots compared to the locations of the results for the other antibodies (Fig. 6A) and as dots close to an IC₅₀ of 0.0001 μg/ml on the scatter plot (Fig. 6B).

We also examined the completeness of neutralization. Like other V1V2-directed broadly neutralizing antibodies (23, 65), the CAP256-VRC26 antibodies neutralize some viruses to a maximum of less than 90%. About 30% of sensitive viruses were neutralized with this plateau effect. This fraction is very similar to that for the V1V2-directed bNAb PGDM1400 (see Fig. S4 in the supplemental material).

To further characterize and compare the new CAP256-VRC26-lineage antibodies, we tested variants CAP256-VRC26.08, CAP256-VRC26.25, CAP256-VRC26.26, and CAP256-VRC26.27 on a multiclade panel of Env pseudoviruses (see Table S5 in the supplemental material). CAP256-VRC26.26 and CAP256-VRC26.27 have neutralization capacities nearly as broad as the neutralization capacity of CAP256-VRC26.25 at a cutoff of 50 μg/ml but less so at an IC₅₀ of <0.01 μg/ml (Fig. 7A), confirming the superior potency of CAP256-VRC26.25. We also analyzed the neutralization breadth within different HIV-1 clades (Fig. 7B). Against clade C, CAP256-VRC26.25 had a neutralization breadth of 70% at the IC₅₀ level (Fig. 7B; see also Fig. S5 in the supplemental material), with even greater coverage of strains from clade G and circulating recombinant forms AG and BC. While CAP256-VRC26.25 neutralized 70% of non-clade B strains overall, it neutralized only 15% of clade B strains. To understand the basis of this preference, we performed an amino acid frequency analysis (53) for positions 154 to 184 in V2. Specifically, based on sequence alignments, we searched for amino acids that were preferentially found among CAP256-VRC26.25-resistant strains versus CAP256-VRC26.25-sensitive strains for each residue position in V2 (Fig. 8A). Amino acids that were associated with resistance were found in at least one position among positions 166, 167, and 169 in 59% of resistant strains. These amino acids were observed in 71% of clade B strains and in 80% of resistant clade B strains. Conversely, the amino acid K or R at position 169 occurred in 86% of sensitive strains but in only 5% of clade B strains, with similar proportions occurring for D or E at position 164 (Fig. 8B). The occurrence of resistance-associated amino acids at one or more of these positions was statistically significant (P < 0.001 for each comparison, Fisher's exact test), as was the occurrence of the sensitivity-associated amino acids at positions 164 and 169. Thus, the characteristics of clade B sequences at V2 positions 164, 166, 167, and 169 may explain much of the resistance found in clade B strains.