Two Classes of Broadly Neutralizing Antibodies within a Single Lineage Directed to the High-Mannose Patch of HIV Envelope

ABSTRACT

The high-mannose patch of human immunodeficiency virus (HIV) envelope (Env) elicits broadly neutralizing antibodies (bnAbs) during natural infection relatively frequently, and consequently, this region has become a major target of vaccine design. However, it has also become clear that antibody recognition of the region is complex due, at least in part, to variability in neighboring loops and glycans critical to the epitopes. bnAbs against this region have some shared features and some distinguishing features that are crucial to understand in order to design optimal immunogens that can induce different classes of bnAbs against this region. Here, we compare two branches of a single antibody lineage, in which all members recognize the high-mannose patch. One branch (prototype bnAb PGT128) has a 6-amino-acid insertion in CDRH2 that is crucial for broad neutralization. Antibodies in this branch appear to favor a glycan site at N332 on gp120, and somatic hypermutation is required to accommodate the neighboring V1 loop glycans and glycan heterogeneity. The other branch (prototype bnAb PGT130) lacks the CDRH2 insertion. Antibodies in this branch are noticeably effective at neutralizing viruses with an alternate N334 glycan site but are less able to accommodate glycan heterogeneity. We identify a new somatic variant within this branch that is predominantly dependent on N334. The crystal structure of PGT130 offers insight into differences from PGT128. We conclude that different immunogens may be required to elicit bnAbs that have the optimal characteristics of the two branches of the lineage described.

IMPORTANCE Development of an HIV vaccine is of vital importance for prevention of new infections, and it is thought that elicitation of HIV bnAbs will be an important component of an effective vaccine. Increasingly, bnAbs that bind to the cluster of high-mannose glycans on the HIV envelope glycoprotein, gp120, are being highlighted as important templates for vaccine design. In particular, bnAbs from IAVI donor 36 (PGT125 to PGT131) have been shown to be extremely broad and potent. Combination of these bnAbs enhanced neutralization breadth considerably, suggesting that an optimal immunogen should elicit several antibodies from this family. Here we study the evolution of this antibody family to inform immunogen design. We identify two classes of bnAbs that differ in their recognition of the high-mannose patch and show that different immunogens may be required to elicit these different classes.

INTRODUCTION

It is becoming increasingly apparent that the “high-mannose patch” on the outer domain of the human immunodeficiency virus (HIV) envelope glycoprotein, gp120, is a site of great vulnerability to HIV type 1 (HIV-1) broadly neutralizing antibodies (bnAbs) (1–7). The high-mannose patch is a region rich in high-mannose glycans centered around the N332 glycan in many isolates but replaced by an N334 glycan in other isolates (5). In still other isolates, both N332 and N334 are lacking, but nevertheless, the high-mannose region still persists (8–12). A prototype bnAb to the high-mannose patch, PGT121, protects against simian-human immunodeficiency virus (SHIV) challenge in macaques at low serum concentrations (13) and is highly effective at reducing viral load in established SHIV infection (14). Serum neutralization from many HIV “elite neutralizers” can be mapped to this high-mannose patch (4, 6, 7), and several families of bnAbs have been isolated from a number of different donors of which the prototypes include PGT121, PGT124, 10-1074, PGT128, and PGT135 (1–3, 15–17, 20). These antibodies recognize the high-mannose patch using distinct but overlapping modes of binding, each interacting with different combinations of N-linked glycans and protein components in addition to the N332 glycan (1–3, 15–17).

We have recently shown that bnAbs targeting the high-mannose patch can be promiscuous in their N-glycan site recognition and can, in certain cases, utilize the glycans at positions N136/N137, N295, or N334 in the absence of the N332 glycan (18) and, in some cases, bind to both high-mannose and complex sugars (2, 16). This ability to utilize other N-linked glycan sites may help counter neutralization escape mediated by shifting or elimination of glycosylation sites. We observed that not only were there differences in the extent of promiscuity of high-mannose patch recognition between donors but also within antibody families from individual donors. This was particularly apparent for bnAbs isolated from International AIDS Vaccine Initiative (IAVI) protocol G donor 36 (PGT125 to PGT131) (3, 4, 18). Surprisingly, PGT130 could neutralize 68% of viruses naturally displaying an N334 glycan compared to PGT128 that could neutralize only 39%. Further, PGT130 was able to neutralize 45% of a panel of 80 N332A/N334A mutant pseudoviruses unlike PGT128 that could neutralize only 23%. When combined, this family of bnAbs is able to neutralize an additional 14 viruses (12%) from a 120-pseudovirus panel compared to PGT128 alone, again suggesting complementary differences in epitope recognition (18). Phylogenetic analysis of PGT125 to PGT131 sequences has shown that these bnAbs cluster into two distinct classes, PGT125 to -128 and PGT130 and -131 (19). PGT128 in the first class has been the most studied due to its superior potency and slightly higher neutralization breadth (3, 15). However, little has been reported on the epitope recognition by the second class, PGT130 and PGT131, which are more dependent on N334. Further characterization of the PGT130 epitope and comparison of the evolution of the two classes of high-mannose patch binding bnAbs in this individual may help in the design of immunogens that elicit a diverse and promiscuous neutralizing antibody response similar to the serum of elite neutralizer donor 36 (18).

Here, we use next-generation sequencing (NGS) and a recently reported phylogenic method (ImmuniTree) to model lineage evolution and diversity of the bnAb response in donor 36 (20). We show that, while the two branches described originated from the same recombination event, the distinct structural features that evolved are critical for their N332 and/or N334 binding modes. A heavy-chain insertion in the CDRH2 of the PGT128 branch appears necessary to avoid V1 loop glycosylation, but this appears to limit the ability to bind to N334. We also show that N334 dependence of the PGT130 branch arose at an early time point and identify a new somatic variant that is predominantly N334 dependent. High levels of somatic hypermutation mutation (SHM) in both branches are required to accommodate the glycan heterogeneity present on the HIV envelope and to avoid the V1 loop glycans in specific HIV strains. However, PGT130 is most sensitive to Env glycan heterogeneity. A crystal structure of the unliganded PGT130 Fab fragment compared to the PGT128 Fab fragment shows that the CDRH2 insertion and CDRL1 deletion are likely in close proximity to glycans in the high-mannose patch when PGT130 binds to its epitope in support of the above conclusions. Our data suggest that in order to elicit a diverse bnAb response similar to that in donor 36, different immunogens may be required to elicit these two antibody classes.

MATERIALS AND METHODS

Ethics statement.

Peripheral blood mononuclear cells (PBMCs) were obtained from donor 36, an HIV-1-infected donor from the IAVI protocol G cohort (21). All human samples were collected with written informed consent under clinical protocols approved by the Republic of Rwanda National Ethics Committee, the Emory University Institutional Review Board, the University of Zambia Research Ethics Committee, the Charing Cross Research Ethics Committee, the Uganda Virus Research Institute (UVRI) Science and Ethics Committee, the University of New South Wales Research Ethics Committee, St. Vincent's Hospital and Eastern Sydney Area Health Service, Kenyatta National Hospital Ethics and Research Committee, University of Cape Town Research Ethics Committee, the International Institutional Review Board of the Mahidol University Ethics Committee, the Walter Reed Army Institute of Research (WRAIR) Institutional Review Board, and the Ivory Coast Comité National d'Ethique des Sciences de la Vie et de la Santé (CNESVS).

Cell sorting and RNA extraction.

Frozen vials of 10⁷ PBMCs were thawed and washed before staining with Pacific Blue-labeled anti-CD3 (UCHT1), Pacific Blue-labeled anti-CD14 (M5E2), fluorescein isothiocyanate (FITC)-labeled anti-CD19 (HIB19), phycoerythrin (PE)-Cy5-labeled anti-CD10 (HI10a), PE-labeled anti-CD27 (M-T271), and allophycocyanin (APC)-labeled anti-CD21 (B-ly4), all from BD Biosciences. Sorts were performed on a high-speed BD FACSAria cell sorter into mirVana lysis buffer (Ambion). Immature B cells, exhausted tissue-like memory B cells, activated mature B cells, resting memory B cells, and short-lived peripheral plasmablasts were stained using previously described markers (22). Total RNA from the pooled sorted cells was then extracted using the mirVana RNA extraction kit (Ambion) according to the manufacturer's instructions and quantitated on a 2100 bioanalyzer (Agilent).

454 sequencing library preparation (20).

Reverse transcription was performed with 10 μl total RNA and 2 μl reverse transcriptase (RT) primer mix [50 μM oligo(dT) and 25 μM random hexamer]. The mixture was heated at 95°C for 1 min and then at 65°C for 5 min and cooled on ice for 1 min. For each reaction, a mix was prepared with 4 μl of 5× First-Strand (FS) buffer (Invitrogen), 1 μl of 10 μM deoxynucleoside triphosphate (dNTP) mix, 1 μl of 0.1 M dithiothreitol (DTT), 1 μl RNase inhibitor (Enzymatics), and 1 μl SuperScript III RT (Invitrogen). This mix was added to the reaction mixture and incubated at 25°C for 10 min, 35°C for 5 min, 55°C for 45 min, and 85°C for 5 min. RNA/DNA hybrid was removed by adding 4 μl Escherichia coli RNase H (Enzymatics). PCR mixtures were assembled using 13.75 μl water, 5 μl cDNA, 5 μl of 5× HF buffer (Thermoscientific), 0.5 μl of 10 mM dNTP, 0.25 μl of each 100 μM primer stock, and 0.25 μl Phusion Hot Start. The reaction mixture was cycled as follows: (i) 98°C (60 s); (ii) 24 cycles, with 1 cycle consisting of 98°C (10 s), 62°C (20 s), and 72°C (20 s); (iii) final extension at 72°C (5 min). Samples were purified on a QIAquick column and run on a 2% agarose E-gel. The desired bands were purified using the Qiagen MinElute gel extraction kit, eluted twice with 10 ml Buffer EB (Qiagen), and quantitated on a 2100 bioanalyzer. Samples were sent to SeqWright for 454 sequencing (20), which was performed per the manufacturer's instructions.

Primers.

The primers for the heavy chain were 20111028-IGHV4 (CRGCTCCCAGATGGGTCCTGT) and 20080924-IGHG (CSGATGGGCCCTTGGTGG). The primers for the light chain were 20111028-IGLV2-8 (TCACTCAGGGCACAGGGTCCTG) and 20111028-IGLC (AGAGGAGGGYGGGAACAGAGTG).

Raw data processing: VDJ alignment and clone definition.

Raw sequencing data were processed using in-house tools written in Python. Reads were split into bar codes, size filtered, and aligned to the IMGT germ line VDJ reference database (www.IMGT.org). The scores were kept low, as we were interested in sequences that were very highly mutated. The variable (V) region was aligned first and then it was removed, the joining (J) region was aligned and then removed, and finally, removal the diversity (D) region was aligned and then removed. The IMGT-defined CDR3 sequence of each read was then extracted. The CDR3 sequences were sorted by abundance and clustered with USEARCH5.1 with the options “–minlen 0 –global –usersort – iddef 1 –id 0.9”. Finally, each CDR3 sequence was aligned to the “target” antibody sequences of PGT125 to PGT131 to determine a “divergence” value from these antibodies (20).

Antibody variant identification and analysis.

The divergence-mutation plots are used as a tool to identify reads that are similar to the known PGT125 to PGT131 antibodies (20). High-identity clusters of sequences and clusters that are above background are manually identified and used as input for a phylogeny inference with ImmuniTree. ImmuniTree is a dedicated algorithm for building immune receptor sequence trees from high-throughput DNA sequencing. It implements a Bayesian model of somatic hypermutation of clones, including probabilistic models of somatic hypermutation mutation (SHM) and sequencing error and performs Markov chain Monte Carlo over the tree structure, birth/death times of the subclones, birth/death, mutation, and sequencing error rates, subclone consensus sequences, and assignment of reads to nodes. The tree structure is also used for multiple computations and to overlay different information. We estimate the selection pressure that a given node has experienced using the BASELINe algorithm. It performs a Bayesian estimation of selection pressure by comparing the observed number of replacement/silent mutations in the CDRs/FWRs of the node consensus sequence. For a thorough description of the ImmuniTree algorithm, see reference 20.

Pseudovirus production and neutralization assays.

To produce pseudoviruses, plasmids encoding Env were cotransfected with an Env-deficient genomic backbone plasmid (pSG3ΔEnv) in a 1:2 ratio with the transfection reagent polyethylenimine (PEI) (1 mg/ml; 1:3 PEI/total DNA; Polysciences) into HEK 293T cells or GnT1-deficient 293S cells (GnT1^−/−). Pseudoviruses were harvested 72 h posttransfection for use in neutralization assays. Glycosidase inhibitors were added at the time of transfection at the following concentrations as described in reference 23: 25 μM kifunensine and 20 μM swainsonine.

Neutralizing activity was assessed using a single-round replication pseudovirus assay with TZM-bl target cells as described previously (3). Briefly, TZM-bl cells were seeded in a 96-well flat-bottom plate at a concentration of 20,000 cells/well. The serially diluted virus/antibody mixture, which was preincubated for 1 h, was then added to the cells, and luminescence was quantified 72 h following infection via lysis and addition of Bright-Glo luciferase substrate (Promega). To determine 50% inhibitory concentrations (IC₅₀s), serial dilutions of monoclonal antibodies (MAbs) were incubated with virus, and the dose-response curves were fitted using nonlinear regression.

Antibody expression and purification.

Antibody sequences for predicted precursors were synthesized (IDT DNA, San Diego, CA) and cloned into previously described heavy- and light-chain vectors (24). Antibody plasmids were cotransfected at a 1:1 ratio in 293 FreeStyle cells using 293fectin (Invitrogen). Transfections were performed according to the manufacturer's protocol, and antibody supernatants were harvested 4 days following transfection. Antibody supernatants were purified over a protein A column, eluted with 0.1 M glycine (pH 3.5), and buffer exchanged into phosphate-buffered saline (PBS).

To prepare the PGT130 Fab fragment for crystallization studies, the Fc C_H2-3 domains in the expression vector containing the full-length IgG were deleted, and soluble Fab was produced by transient transfection of 293S cells with the recombinant DNA. The Fab fragment was purified using an anti-human lambda affinity column followed by passage through a Mono S 10/100 GL cation exchange column (GE Healthcare). The Fab fragment was then treated with endoglycosidase H (New England BioLabs [NEB]) to cleave the N-linked glycans and purified through size exclusion using a Superdex 200 (GE Healthcare) column.

Antibody and envelope mutations.

Mutations in the PGT130 heavy and light chains and the HIV-1 envelope glycoprotein were introduced using QuikChange site-directed mutagenesis (Stratagene, La Jolla, CA). Mutations were verified by DNA sequencing (Eton Biosciences [La Jolla, CA] and MWG Eurofins [Germany]).

Crystallization.

The PGT130 Fab fragment was concentrated to 8 to 12 mg/ml using an Amicon centrifugal ultrafiltration unit (Millipore) with a 10-kDa-molecular-size cutoff. Fab samples were screened for crystallization using the 384 conditions of the JCSG Core Suite (Qiagen) at both 277 and 293 K using the Joint Center for Structural Genomics (JCSG)/IAVI robotic Crystalmation system (Rigaku). Crystallization was performed using the nanodroplet vapor diffusion method (25) with standard JCSG crystallization protocols (26). Sitting drops composed of 200 nl of protein sample were mixed with 200 nl of crystallization reagent in a sitting drop format and were equilibrated against a 50-μl reservoir of crystallization reagent. After approximately 3 days at 20°C, crystals of Fab 130 formed in 40% (wt/vol) polyethylene glycol 400 (PEG 400), 0.2 M NaCl, and 0.1 M Na/K phosphate, pH 6.2 (JCSG Core Suite 2, well C11).

Data collection.

The PGT130 crystals diffracted to 2.75 Å at beamline 08ID-1 of the Canadian Light Source (CLS). Data were collected at 100 K to a completeness of 91.4% with an overall R_sym of 0.13 (0.67 in the high-resolution shell) (see Table S3 in the supplemental material). Data were indexed, processed, and scaled with HKL-2000 (27) in orthorhombic space group P2₁2₁2₁ with unit cell parameters a = 67.0 Å, b = 69.1 Å, and c = 272.2 Å.

Structure refinement.

The structure was determined by the molecular replacement method using Phaser (28) with the PGT128 Fab (PDB identifier [ID] or accession no. 3TV3) as an initial model. Model building was carried out using Coot-0.7 (29), and refinement was implemented with Phenix (30). The final R_cryst and R_free values are 24.5% and 26.9%, respectively.

Protein structure accession number.

The coordinates and structure factors for PGT130 Fab have been deposited in the Protein Data Bank under 4RNR.

RESULTS

bnAbs in donor 36 arise from the same recombination event but differ by the presence of a heavy-chain insertion and light-chain deletion.

Six bnAbs have previously been isolated from donor 36 (PGT125 to PGT131), and the binding of all bnAbs to gp120 is strongly dependent on the N-linked glycans at positions N301 and N332 (3, 18). Analyses of the antibody sequences show that as well as sharing common germ line V and J genes (IGHV4-39, IGHJ5*02 and IGLV2-8, IGLJ3*02), they share long stretches of identical nucleotides and common mutated residues. The antibody sequences cluster into two distant but clonally related branches that have evolved from the same VDJ recombination event (3, 19). The two branches differ by the presence of a 6-residue insertion in CDRH2 and a 5-residue deletion in CDRL1 of bnAbs in the PGT125-PGT128 branch (Table 1; see Fig. S8 in the supplemental material). To determine the importance of these indels for divergence in high-mannose patch recognition, we constructed two sets of complementary mutant bnAbs, those in which the PGT128 CDRH2 insertion was removed or the PGT128 CDRL1 deletion was restored and those in which the insertion or deletion was incorporated into PGT130. Neutralization breadth and potency of the mutant nAbs were assessed on an indicator panel of six pseudoviruses (21) (Fig. 1). The CDRH2 insertion in PGT128 has previously been shown to be important for neutralization of JR-FL (15). This insertion is shown here to be critical for neutralization breadth and potency across the cross-clade indicator panel (Fig. 1C). However, restoration of the CDRL1 deletion in PGT128 had no effect on neutralization activity (Fig. 1B). PGT128 is more potent than its sister clone PGT130 (3), but the increased potency is not simply due to the CDRH2 insertion and heavy-chain disulfide bond between CDRH1 and CDRH2 in PGT128; when residues S50-A52C (insertion) and G32C mutation (permitting disulfide formation) were introduced into PGT130, no neutralization was observed (Fig. 1F). This result suggests that PGT130 has evolved distinct somatic mutations to enable recognition of the high-mannose patch region in the absence of the heavy-chain insertion.

TABLE 1.

Sequence characteristics of selected heavy and light chains from donor 36

bnAb	V gene	J gene	CDR3 lengtha	CDR3 sequence	Mutation frequency (%)b		Position and size of indels
					Nucleotides	Amino acids
Heavy-chain
glc	IGHV4-39	IGHJ5*02	19	FGGEVLVYNDWPKPAWFDP	0	0
3H	IGHV4-39	IGHJ5*02	19	SGGDILYYIDWQKPHWFDP	14	21
9H	IGHV4-39	IGHJ5*02	19	CGGDILYFHDWRKPHWFHP	24	32
74H	IGHV4-39	IGHJ5*02	19	SGGDILYYIEWQKPHWFYP	19	24
95H	IGHV4-39	IGHJ5*02	19	FGGEVLVYNDWPKPAWVDL	16	25	+6 in CDRH2
125H	IGHV4-39	IGHJ5*02	19	FDGEVLVYNHWPKPAWVDL	22	31	+6 in CDRH2
126H	IGHV4-39	IGHJ5*02	19	FDGEVLVYHDWPKPAWVDL	20	25	+6 in CDRH2
127H	IGHV4-39	IGHJ5*02	19	FGGEVLVYRDWPKPAWVDL	17	25	+6 in CDRH2
128H	IGHV4-39	IGHJ5*02	19	FGGEVLRYTDWPKPAWVDL	21	31	+6 in CDRH2
130H	IGHV4-39	IGHJ5*02	19	SGGDILYYYEWQKPHWFSP	22	30
131H	IGHV4-39	IGHJ5*02	19	SGGDILYYIEWQKPHWFYP	22	30
Light-chain
gl	IGLV2-8	IGLJ3*02	10	SSYAGNWDVV	0	0
3L	IGLV2-8	IGLJ3*02	10	SSYVGSWDVV	7	11
46L	IGLV2-8	IGLJ3*02	10	SSLFGRWDIV	13	19
71L	IGLV2-8	IGLJ3*02	10	SSLVGNWDVI	8	11	−5 in CDRL1
125L	IGLV2-8	IGLJ3*02	10	GSLVGNWDVI	16	21	−5 in CDRL1
126L	IGLV2-8	IGLJ3*02	10	SSLVGNWDVI	10	14	−5 in CDRL1
127L	IGLV2-8	IGLJ3*02	10	SSLVGNWDVI	9	13	−5 in CDRL1
128L	IGLV2-8	IGLJ3*02	10	GSLVGNWDVI	9	13	−5 in CDRL1
130L	IGLV2-8	IGLJ3*02	10	SSLFGRWDVV	12	19
131L	IGLV2-8	IGLJ3*02	10	SSLSGRWDIV	13	21

^aCDR3 lengths were determined according to the IMGT definition.

^bMutation frequency was calculated over the V gene and J gene as nucleotides or amino acids differing from the putative germ line sequence for heavy-chain sequences and light-chain sequences. The CDR3 regions and insertions and deletions were excluded from the analysis.

^cgl, germ line.

FIG 1.

Importance of indels for neutralization breadth and potency by bnAbs from donor 36. Neutralization breadth and potency of mutated and chimeric antibodies were determined on a 6-pseudovirus panel previously shown to be predictive of breadth and potency on a larger panel (21). (A) PGT128, (B) PGT128+del (5-residue CDRL1 deletion restored), (C) PGT128-insert (6-residue CDRH2 insertion removed), (D) PGT128-insert+del (6-residue CDRH2 insert removed and 5-residue CDRL1 deletion restored), (E) PGT130, and (F) PGT130+insert (PGT128 6-residue CDRH2 insert introduced into PGT130 CDRH2 and G32C permitting disulfide formation).

N332/N334 dependence arises early in the PGT130 branch.

To explore the evolution of the high-mannose patch bnAb response and the divergence in N332/N334 dependence in this donor, we used 454 deep sequencing and a recently published program, ImmuniTree, designed to model SHM (20). Unfortunately, longitudinal samples were not available from this individual, and therefore, only PBMCs from a single time point when the bnAbs were isolated were sequenced and used to model the lineage evolution of the PGT128 family. To produce libraries for 454 sequencing, gene-specific primers were used on the population of 78,000 sorted IgG⁺ memory B cells to amplify the IGHV4-39 and IGLV2-8 gene families from which PGT128 and PGT130 were derived (19). 454 sequencing of the library yielded 390,563 and 711,575 heavy- and light-chain reads, respectively. Heavy- and light-chain trees (see Fig. S1a and S1b in the supplemental material) were generated using a new Bayesian phylogeny algorithm, ImmuniTree, that models somatic hypermutation for the PGT125-to-PGT131 family of bnAbs (see Materials and Methods for full details). The trees consist of both inferred and observed nodes with the observed nodes being the most highly mutated. To address sequencing errors, the algorithm collapsed the 390,563 heavy-chain sequences and 711,575 light-chain sequences into 108 heavy-chain nodes (named 1H to 108H) and 136 light-chain nodes (named 1L to 136L). Of these nodes, 24 heavy-chain and 32 light-chain nodes were inferred statistically and did not have direct sequencing data. As we were interested in whether less-mutated inferred intermediates had neutralizing activity, heavy- and light-chain nodes with different levels of mutation at key branch points were selected and paired to generate a simplified tree (Fig. 2A and Table 1; see Table S1 in the supplemental material).

FIG 2.

Neutralization breadth and potency of PGT125 to PGT131 variants identified by next-generation sequencing and phylogenetic analysis. (A) Simplified antibody tree representing analysis of next-generation sequencing data using the ImmuniTree algorithm (for full analysis, see Fig. S1a and S1b in the supplemental material). The mutation level is listed for heavy (H) and light (L) chains individually and based on nucleotide changes from the germ line (gl) over the V and J segments. (B) Percent breadth and potency on the 6-virus panel of predicted precursor bnAbs of PGT128 and PGT130 and a new somatic variant (9H 46L). Values are reported as IC₅₀ (antibody concentrations required to inhibit HIV activity by 50%) and measured in micrograms per milliliter. The cells in the table are colored as follows; red, IC₅₀ of <0.01 μg/ml; orange, IC₅₀ of 0.01 to 0.1 μg/ml; yellow, IC₅₀ of 0.1 to 1 μg/ml; green, IC₅₀ of >1 μg/ml.

Neutralization breadth and potency of the purified antibodies were initially determined on the 6-pseudovirus indicator panel (Fig. 2B; neutralization data for a larger panel of chimeric bnAbs can be found in Fig. S2 in the supplemental material). As seen for other families of bnAbs, no HIV reactivity for the predicted germ line antibody was detected on the 6-virus panel (Fig. 2B), and no binding to the corresponding recombinant gp120 glycoprotein was observed in an enzyme-linked immunosorbent assay (ELISA) (data not shown) (31–35). Within the PGT128 branch, the least mutated predicted intermediate 95H 71L (12% mutated in the nucleotide sequence of the heavy- and light-chain V- and J-gene segments) was able to neutralize all 6 viruses albeit at a slightly lower potency than that of the fully mutated PGT128 (16%). In the PGT130 branch, intermediate 3H 3L (11%) showed no neutralization activity. This is in contrast to the PGT121 family of bnAbs where 46% neutralization was observed with 6% SHM (20). However, chimeric antibody 74H 3L (13%) showed very limited neutralization activity.

To investigate how the differences in N332 and N334 dependency arose between the two branches, we next tested intermediate and mature bnAbs on a cross-clade 54-pseudovirus panel (Fig. 3A and B; see Table S2 in the supplemental material) that included 41 viruses with a glycan at position N332, 12 viruses with a glycan at position N334, and 1 virus lacking both N332 and N334. Of clade CRF01_AE viruses, 95% have a glycan site at position N334, whereas the remaining subtypes have >75% of viruses with a glycan site at position N332 (36). For both the PGT128 and PGT130 branches, an increased level of SHM was seen to increase breadth and potency of neutralization as illustrated in the cumulative frequency distribution plot (Fig. 3C). Similar to that shown previously, PGT130 neutralized more N334 viruses (73%) than PGT128 did (58%) (18), and PGT128 neutralized more N332 viruses (98%) than PGT130 (73%). The less-mutated PGT130 branch bnAb 74H 3L, although able to neutralize only 13% of viruses, was able to neutralize 42% of N334 viruses compared to 5% of N332 viruses (Fig. 3A and B). This suggests that the PGT130 branch may have initially evolved to be more N334 specific before developing N332 reactivity. To further explore the N332/N334 dependency of the two bnAb classes, we next tested them against two mutated forms of the 6-virus panel where the glycan at N332 was either shifted to N334 or removed by an N332A mutation (Fig. 3D) (18). Notably, 74H 3L could now neutralize JRCSF N334 and 92RW020 N334, although not the N332A versions, further supporting a preference for N334 (Fig. 3D).

FIG 3.

Precursor bnAbs show distinct patterns of N332 and N334 neutralization. (A) Overall percent breadth and potency of precursor bnAbs against a cross-clade panel of 54 pseudoviruses. (B) Percent breadth and potency of precursor bnAbs separated for N332 and N334 viruses. (C) Cumulative frequency distribution of IC₅₀s by precursor bnAbs compared to fully mutated bnAbs. Antibodies in the PGT128 and PGT130 branch are shown in black and green, respectively. (D) Ability of predicted precursor bnAbs to neutralize N332A/N334A and N334-shifted mutant viruses. The cells in the tables in panel D are IC₅₀ values and are colored as follows: red, <0.01 μg/ml; orange, 0.01 to 0.1 μg/ml; yellow, 0.1 to 1 μg/ml; green, >1 μg/ml.

A somatic variant that is predominantly N334 dependent is observed in the PGT130 branch.

In addition to clonal variants that cluster into the PGT128 and PGT130 branches, new heavy- and light-chain sequences were identified within the PGT130 branch of the heavy- and light-chain trees. 9H (24% mutated) also contains no indels but is significantly divergent from PGT130, and it was paired with 46L. This new somatic variant, 9H 46L (19% mutated), was seen to be predominantly N334 dependent. This dependency was further confirmed on a small number of N334A variant viruses (Fig. 3D; see Fig. S3A and S3B in the supplemental material) and the N334-shifted panel (Fig. 3D). The data suggest that predominantly N334-dependent somatic variants exist within the polyclonal response against the high-mannose patch, although single B cell cloning would be required to confirm the correct heavy- and light-chain pairing. Of note, although 9H 46L is more N334 dependent, the epitope is still centered on the N301 glycan site as shown by its inability to neutralize a JRCSF N334 variant lacking the N301 glycan (Fig. S3C and S3D).

High levels of SHM are necessary to accommodate V1 loop glycans.

Many of the N332-dependent bnAbs require binding to additional neighboring N-linked glycans for neutralization (1, 15, 17). Sok et al. recently showed that putative precursors of PGT121 are dependent on both N301 and N332 glycans and that the dependence on the N301 glycan is reduced with increased SHM (20). To determine how SHM and indels impact on the dependence of these two classes of bnAbs on N-linked glycans within the high-mannose patch, we used a panel of JRCSF glycan mutant viruses and measured impact on neutralization by PGT128, PGT130, 95H 71L, 3H 130L, and PGT128-insert (Fig. 4A; see Fig. S4 in the supplemental material). Glycan sites N135, N295, and N301 were chosen on the basis of the recent HIV envelope trimer crystal structure that revealed their close proximity to the PGT128 epitope (17). Interestingly, 3H 130L and PGT128-insert showed enhanced neutralization for the N135A mutant virus, suggesting increased SHM and/or the heavy-chain insertion might allow better accommodation of this V1 loop glycan. To explore this hypothesis further, we measured neutralization of a small panel of V1 loop glycan mutant pseudoviruses (Fig. 4B to H). The results were somewhat strain dependent; for N332 viruses, removal of the N135/136/137 glycan enhanced neutralization, whereas for N334 viruses, removal of the N141, N149, or N151 glycan-enhanced neutralization. The difference in V1 loop glycan shielding between N332- and N334-containing viruses likely reflects the various proximities of the glycans arising from the different protein sequences. The PGT130 bnAb class, including 9H 46L, was much more sensitive to removal of V1 loop glycans than the PGT128 class (Fig. 4). Removal of a V1 loop glycan allowed some viruses to become sensitive to neutralization by PGT128-insert when the wild-type virus had been resistant. However, the germ line version of the antibody remained ineffective against all mutant viruses (data not shown). Overall, it appears that the V1 loop glycans limit the putative precursor bnAbs from fully accessing their epitopes but that this can be overcome by additional SHM and/or indels.

FIG 4.

bnAbs in donor 36 evolved to avoid V1 loop glycans in some virus strains. V1 loop glycans were removed by site-directed mutagenesis, and the effect on neutralization was measured. Wild-type (WT) viruses are shown as solid lines, and glycan mutant viruses are shown as dashed lines. (A) JRCSF and N135A variant, (B) SF162 and N136A variant, (C) 92RW020 and N131A variant, (D) JR-FL and N134A variant, (E) BG505 N332 and N137A variant, (F) BG505 and N137A variant, (G) 92TH021 and N151A variant, (H) CNE8 and N141A and N149A variants.

High levels of SHM are required to accommodate glycan heterogeneity on the HIV envelope.

Analysis of the neutralization profiles of a number of inferred intermediate bnAbs, and on occasion PGT130, showed that viruses neutralized incompletely with a low neutralization plateau and/or a shallow neutralization curve (see Fig. S5 in the supplemental material). We have previously shown that incomplete neutralization can be a result, at least in part, of envelope glycan heterogeneity giving rise to a population of viruses that are resistant to neutralization (23). This might suggest that less-mutated bnAbs and the PGT130 class of bnAbs are more restricted in the type of glycan they can recognize at the key glycosylation sites (N332/N334 and N301) and therefore are unable to neutralize the complete population of heterogeneously glycosylated virus particles. We therefore measured the impact of changes in envelope glycosylation on neutralization sensitivity for an antibody panel by making BJOX015000.11.5 (Fig. 5) and JR-FL (see Fig. S6 in the supplemental material) in the presence of the glycosidase inhibitors kifunensine and swainsonine and in the GnT1-deficient cell line, 293S (GnT1^−/−) (23). We observed different sensitivities toward changes in glycosylation profile between the classes of bnAbs and the virus strain.

FIG 5.

Increased somatic hypermutation of donor 36 bnAbs allows better accommodation of viral gp120 glycan heterogeneity. BJOX015000.11.5 pseudoviruses were prepared in the presence of glycosidase inhibitors kifunensine (Kif) (endoplasmic reticulum [ER]-mannosidase I inhibitor), swainsonine (Golgi-α-mannosidase II inhibitor) and in a GnT1-deficient cell line (HEK-293S), and the effect on neutralization was determined for 74H 3L (A), PGT130 (B), 95H 71L (C), PGT128 (D), and 9H 46L (E).

Neutralization by PGT128 remained largely unaffected by changes in the viral glycosylation profile, although some increase in neutralization sensitivity was observed for viruses grown in the presence of kifunensine, which leads to homogeneous Man₉GlcNAc₂ glycans. Within the PGT128 branch, the less-mutated 95H 71L showed increased potency against BJOX015000.11.5 when glycan heterogeneity was decreased by any of three conditions (Fig. 5). The increase brought the inferred precursor up to the level of PGT128 in terms of isolate neutralization potency suggesting that, for this isolate, PGT128 was better able to cope with glycan heterogeneity than the precursor antibody. The effects were less pronounced for the JR-FL isolate (see Fig. S6 in the supplemental material), which might already have a more homogeneous or optimal glycan profile. In contrast, the PGT130 branch antibodies were more sensitive to changes in glycosylation. PGT130 had a reduced potency against both the JR-FL pseudoviruses prepared with kifunensine or made in 293S cells but an increased potency against the virus prepared with swainsonine (Fig. S6). Very little effect was observed for the BJOX015000.11.5 viruses (Fig. 5). Less-mutated 74H 3L and the new somatic variant 9H 46L were unable to neutralize either the virus prepared with kifunensine or made in 293S cells, and virus made in the presence of swainsonine was neutralized with a slightly increased potency. This suggests that, for some donor 36 bnAbs, either certain glycan structures cannot be accommodated within the antibody-binding site, resulting in a greater sensitivity to glycan heterogeneity, or alternatively, the larger Man₉GlcNAc₂ structure present in the V1 loop might prevent engagement of the antibody with its epitope. For example, in a similar manner to PG9 (23, 37), it is possible that PGT130 or 74H 3L might not be able to accommodate a Man₉GlcNAc₂ glycan at position N301 and/or binding affinity is enhanced if N301 is a hybrid glycan for some viruses. Similarly, members of the PGT121 family have shown differing abilities to recognize both complex and high-mannose sugars (2, 16). Overall, SHM affects the ability of bnAbs to accommodate glycan heterogeneity on HIV envelope, and this may impact protective efficacy in vivo.

A crystal structure of the PGT130 Fab fragment highlights differing modes of binding by the antibody classes.

To better understand how the different structural features of the PGT128 and PGT130 branches of the bnAb family might impact the antibody combining site and subsequent recognition of the high-mannose patch and N332/N334 dependence, we next determined the crystal structure of the PGT130 Fab fragment to 2.75-Å resolution (Fig. 6A). In the unliganded state, the CDR loops of PGT130 appear highly flexible, resulting in poorly defined electron density, especially for the CDR H3 loop, which is largely disordered. Of note, electron density for single GlcNAc moieties are defined at the N-linked glycosylation sites N24^LC and N82b^HC. The antigen combining sites of the unliganded PGT130 Fab and PGT128 structures overlay very well except at the 6-amino-acid heavy-chain insertion and the 5-amino-acid light-chain deletion in PGT128 (Fig. 6B). The CDRL1 of PGT130 adopts a more helical conformation consistent with Chothia canonical L1 loop class 6 compared to the L1 loop of PGT128, which is not within any Chothia canonical class, presumably due to the 5-amino-acid deletion (38). A model of the PGT128 Fab epitope on the BG505 SOSIP trimer crystal structure (17) (Fig. 6C) reveals the close proximity of the CDRL1 loop to the N137 V1 loop glycan and the tight association of the CDRH2 loop insertion with the N332 glycan on the gp120 core. Thus, the CDRH2 loop may skew the PGT128 interaction toward the V3 loop and away from the V1 loop, and the 5-amino-acid deletion in the PGT128 CDRL1 may allow better accommodation of V1 loop glycosylation (Fig. 6D). This is corroborated by PGT128 being less tolerant of the N334 glycan shift than PGT130, which does not include the indels. Additionally, neutralization for the V1 loop glycan mutant viruses by PGT128 lacking the heavy-chain insertion supports its function to avoid the V1 loop glycans (Fig. 4). However, reintroducing the CDRL1 deletion into the fully mutated PGT128 does not have a significant impact on neutralization potency for the fully mutated bnAb or the PGT128 variant lacking the heavy-chain insertion (Fig. 1). Thus, these structural differences indicate diverging strategies for recognition of the high-mannose patch: PGT128 focuses on the more-conserved elements of the high-mannose patch and V3 loop region at the cost of reduced or no tolerance for variation at V3 glycans, while PGT130 is more sensitive to V1 glycans at the cost of less neutralization breadth and potency in general.

FIG 6.

Crystal structure of PGT130 Fab highlights differing modes of binding by the antibody classes. (A) The crystal structure of the PGT130 Fab variable region is represented as a gray ribbon. The CDR loops are individually labeled and colored, with small red circles indicating parts of the CDR H3 loop that are disordered in the structure. The two N-linked glycosylation sites are also labeled and shown in green ball-and-stick representations. To the right, the combining site of PGT130 Fab is displayed in more detail. Residues that are important for binding gp120, as determined by alanine-scanning mutagenesis, are labeled, and their side chains are shown as ball-and-sticks. The side chain of Trp^H100E is not shown because it is disordered in the crystal structure. (B) The crystal structures of the PGT130 (gray ribbon) and PGT128 (thin blue ribbon) (PDB ID or accession no. 3TV3) variable regions are shown superposed. CDR loops with significant differences between the structures are labeled. (C) The CDR loops of PGT128 are shown as individual ribbons at its extended epitope on the SOSIP trimer extending across portions of the V3 loop (yellow surface) and the V1 loop (gray surface). The V1 component of the epitope is modeled here based on superposing the PGT128-gp120 outer domain structure (PDB ID 3TYG) on the BG505 SOSIP trimer structure (PDB ID 4NCO) only on gp120. Glycans are shown as colored spheres. Insertions/deletions on the CDR loops are colored red. (D) The glycan shift from position N332 to N334 is modeled in relation to the PGT128 CDR H2 loop colored as in panel C. (E) Alanine-scanning mutagenesis of the PGT130 paratope. Amino acids at positions known to be important for PGT128 neutralization were mutated to alanine in PGT130, and the effect on neutralization of JR-FL and JR-CSF was determined. The fold increase in the neutralization IC₅₀ compared to WT IgG is depicted in the table with different colors as follows: green, 3- to 5-fold; yellow, 6- to 20-fold; orange, 21- to 100-fold; red, >100-fold. The residues mutated in PGT130 and PGT128 are shown in red and black in the variant column, respectively. N/A, not applicable, as that particular residue not present at that position in the somatic variant; n.d., not determined.

Antibody residues that contact the N332 glycan are conserved across all somatic variants and predicted precursor bnAbs.

To map the key interactions between the PGT130 combining site and gp120, we performed paratope mapping of PGT130. Residues shown to be critical for PGT128 binding to high-mannose and eODmV3 protein (15) were mutated to alanine in PGT130, and their impact on neutralization of JR-CSF and JR-FL pseudovirus (Fig. 6E) and on gp120 binding (see Fig. S7 in the supplemental material) was measured. Residues shown to be critical for PGT128 interaction with Man_8/9GlcNAc₂ at position N332 (HCDR3 residues W100eA and K100gA and LCDR3 W95A and D95aA) were also found to be critical for PGT130 neutralization and gp120 binding, suggesting similar modes of recognition for this high-mannose glycan. These residues are conserved between all family members and precursor antibodies indicating that these mutations arose before divergence of the two branches (Fig. S8). This may suggest that the interaction with the N332 glycan was important for initial selection of this bnAb family. Residues H59 and K64 in PGT128 that also contact the N332 glycan are conserved in PGT130, but these residues had little effect on neutralization when mutated in PGT128 (15). Residues within the CDRH2 of PGT130 (Ile⁵¹ to Thr⁵⁵) were also identified as important for neutralization by PGT130. Based on the PGT130 Fab structure, His⁵⁰ and Ile⁵¹ likely play a structural role in maintaining the antibody-combining site, whereas Tyr⁵³ and Thr⁵⁴ are more likely forming key contacts with the target epitope. Unlike PGT128, PGT130 has N-linked glycans in the variable regions of its heavy and light chains. When these glycans were removed using site-directed mutagenesis, no significant effect on either neutralization breadth or potency was observed (Fig. 6E), suggesting that these glycans are not important for the difference in epitope recognition by the two classes of bnAbs.

DISCUSSION

Insight into the development of broadly neutralizing antibodies (bnAbs) in HIV-infected individuals is a major goal in HIV vaccine research (39–43). bnAbs targeting the glycan shield must either bind the N-linked glycans within their binding site for increased affinity or evolve to accommodate or avoid obstructing glycans. The bnAb response against the high-mannose patch in IAVI protocol G donor 36 uses a combination of these strategies. Unlike less-mutated forms of PGT121 that use additional N-linked glycans for neutralization (17, 20), this family of bnAbs evolves to better accommodate or avoid the V1 loop glycans (Fig. 4) while gaining affinity from binding to both the N332/N334/N295 and N301 glycans. From the small number of putative germ line precursor bnAbs tested in this study, it can be deduced that increased SHM and indels in this family of bnAbs have several effects: (i) an increased breadth and potency of neutralization (Fig. 3C), (ii) an ability to better accommodate changes in glycan heterogeneity within the antibody combining site (Fig. S5), (iii) an ability to better accommodate glycans within the V1 loop for some virus strains (Fig. 4), and (iv) an increased ability to utilize other N-linked glycans within the high-mannose patch in the absence of the N332 glycan (Fig. 3D). Here we describe two related classes of bnAbs against the high-mannose patch arising from a single recombination event in donor 36 that achieve these advantageous features differently. The most potent class, PGT128, has a critical insertion in the CDRH2. The second more-mutated class, PGT130, although less potent, is more able to tolerate different glycan patterns in the high-mannose patch (Fig. 3) (18).

We first consider the PGT128 class of bnAbs. An important difference between the PGT128 and PGT130 classes of bnAbs from donor 36 is the CDRH2 insertion and CDRL1 deletion in the PGT128 branch that appears to occur at a relatively early time point (Fig. 1 and Table 1; see Fig. S1 in the supplemental material). We have shown that a PGT128 variant lacking the CDRH2 insert has very limited neutralization activity, but neutralization can be increased by removal of certain V1 loop glycans. These data, along with docking of PGT128 on the BG505 SOSIP trimer (17), suggest that the PGT128 CDRH2 insertion is necessary to accommodate V1 loop glycosylation (Fig. 6C and D). However, the tight association between the CDRH2 loop insertion and the N332 glycan, although increasing neutralization potency, appears to constrain the PGT128 interaction toward the V3 loop, thus reducing the ability of PGT128 to use alternate glycans in the vicinity of the high-mannose patch in the absence of the N332 glycan compared to PGT130 (Fig. 3) (18). SHM also plays an important role in epitope recognition in the PGT128 bnAb class. Antibody 95H 71L, which also has both the heavy-chain insertion and light-chain deletion but a lower level of SHM, was found to be more sensitive to changes in glycan heterogeneity and removal of V1 loop glycosylation and had a reduced ability to neutralize N334 and N332A/N334A mutant viruses.

The bnAbs in the PGT130 class have no indels; therefore, they rely solely on SHM for neutralization breadth and potency and use of alternate glycans in binding to the high-mannose patch. A common feature of many HIV bnAbs is the presence of long insertions and deletions that appear critical for HIV reactivity (1, 20, 44–46), and it is a challenge for immunogen design to develop strategies capable of eliciting antibodies with these unusual features (47). As demonstrated here, PGT130 is able to reach a level of neutralization breadth similar to that of PGT128 without indels, albeit with a slightly higher degree of somatic mutation and a slightly lower potency. The decreased potency of PGT130 may reflect its reduced ability to accommodate V1 loop glycosylation near the CDRL1 loop within the binding site and its greater sensitivity to changes in glycan heterogeneity (Fig. 4 and 5; see Fig. S6 in the supplemental material). However, the subsequent benefit of PGT130 lacking the PGT128 indels is its noticeably increased ability to neutralize N334-containing viruses. Here we show that the putative precursor of PGT130, Ab 74H 3L, is able to neutralize 42% of N334 viruses compared to 5% of N332 viruses. This may suggest that the PGT130 branch initially evolved to be predominantly N334 dependent and upon increased SHM, it was able to bind more different configurations of the high-mannose patch and neutralize N332-containing viruses. In support of this hypothesis, we also identified an additional subbranch (9H 46L) that neutralized 50% of N334 viruses (with a median IC₅₀ of 0.08 μg/ml) compared to only 12% of N332 viruses.

The findings described here have implications for vaccine design. This study and our previous work (18) suggest a spectrum of N332 and N334 dependency exists for bnAbs that bind the high-mannose patch. At one extreme, from a single donor, donor 36, PGT127 shows strong dependence on N332 (3, 18), and at the other extreme, 9H 46L shows strong dependence on N334 (Fig. 3B). PGT130 occupies the middle ground and neutralizes both N332 and N334 viruses efficiently. A combination of bnAbs with differing N332/N334 dependence will likely be most effective at limiting viral escape and providing protection against all subtypes. Without longitudinal samples, it is not possible to determine in detail how the two antibody classes diverged. However, despite the differences described above, the crystal structures of PGT128 and PGT130 Fab are remarkably similar except for the PGT128 indels. Paratope mapping (Fig. 6E) and sequence alignment (see Fig. S8 in the supplemental material) has shown conservation of residues critical for N332/N334 binding across both bnAb classes and predicted precursor nAbs suggesting that a common pathway of maturation occurred before divergence and introduction of the PGT128 indels. As shown by Moore et al., the differing N332/N334 specificity may have arisen from the different selective pressures on the virus from the two antibody branches within this donor that drove the shift of the N332/N334 glycan site (5). Therefore, in terms of immunogen design, both immunogens with a glycan at position N332 and immunogens with a glycan at position N334 may prove essential for eliciting a bnAb response similar to that in donor 36.

Further, these findings suggest a series of immunogens that first lack key V1 loop glycan sites (N137 or N149/N151), and then inclusion of these glycan sites, might favor elicitation of bnAbs that can better accommodate V1 loop glycosylation. The complexity of the epitope of the donor 36 bnAbs, and similarly, other high-mannose patch binding bnAbs (1, 16, 18, 20), indicates that immunogens focused only on the key contacts with the mature bnAbs will likely elicit bnAbs with limited neutralization breadth and potency. Finally, the ability to accommodate the natural glycan heterogeneity of virion-associated envelope may prove essential for in vivo protection against 100% of virus particles. Therefore, although some bnAbs have a higher binding affinity for certain glycoforms, immunogens designed to reelicit this family of bnAbs may need to incorporate glycan heterogeneity similar to that on viral gp120 such that elicited bnAbs bind all possible envelope glycoforms and achieve 100% neutralization.

In summary, somatic hypermutation and indels in a family of bnAbs to the high-mannose patch of HIV gp120 increases neutralization breadth and potency through more complete recognition of glycan heterogeneity, accommodation of V1 loop glycans, and the ability to utilize alternate glycans. The level of SHM required for HIV reactivity is higher than for donor 17 bnAbs (20). Two classes of bnAbs within a single lineage directed to the high-mannose patch exist in donor 36. The PGT128 class is superior in neutralization breadth and potency; however, PGT130 would likely be more effective at countering potential escape through mutation of glycan sites. Therefore, a vaccine capable of eliciting both classes of bnAbs would be highly advantageous (18). In terms of immunogen design, a combination of immunogens displaying either the N332 or N334 glycan sites, with carefully positioned V1 loop glycans and glycan heterogeneity which resembles that of the virus may be needed to maximize the potential of eliciting antibodies of this specificity and a high degree of neutralization breadth and potency.