Envelope Variants Circulating as Initial Neutralization Breadth Developed in Two HIV-Infected Subjects Stimulate Multiclade Neutralizing Antibodies in Rabbits

ABSTRACT

Identifying characteristics of the human immunodeficiency virus type 1 (HIV-1) envelope that are effective in generating broad, protective antibodies remains a hurdle to HIV vaccine design. Emerging evidence of the development of broad and potent neutralizing antibodies in HIV-infected subjects suggests that founder and subsequent progeny viruses may express unique antigenic motifs that contribute to this developmental pathway. We hypothesize that over the course of natural infection, B cells are programmed to develop broad antibodies by exposure to select populations of emerging envelope quasispecies variants. To test this hypothesis, we identified two unrelated subjects whose antibodies demonstrated increasing neutralization breadth against a panel of HIV-1 isolates over time. Full-length functional env genes were cloned longitudinally from these subjects from months after infection through 2.6 to 5.8 years of infection. Motifs associated with the development of breadth in published, cross-sectional studies were found in both subjects. We compared the immunogenicity of envelope vaccines derived from time points obtained during and after broadening of neutralization activity within these subjects. Rabbits were coimmunized four times with selected multiple gp160 DNAs and gp140-trimeric envelope proteins. The affinity of the polyclonal response increased as a function of boosting. The most rapid and persistent neutralization of multiclade tier 1 viruses was elicited by envelopes that were circulating in plasma at time points prior to the development of 50% neutralization breadth in both human subjects. The breadth elicited in rabbits was not improved by exposure to later envelope variants. These data have implications for vaccine development in describing a target time point to identify optimal envelope immunogens.

IMPORTANCE Vaccine protection against viral infections correlates with the presence of neutralizing antibodies; thus, vaccine components capable of generating potent neutralization are likely to be critical constituents in an effective HIV vaccine. However, vaccines tested thus far have elicited only weak antibody responses and very modest, waning protection. We hypothesized that B cells develop broad antibodies by exposure to the evolving viral envelope population and tested this concept using multiple envelopes from two subjects who developed neutralization breadth within a few years of infection. We compared different combinations of envelopes from each subject to identify the most effective immunogens and regimens. In each subject, use of HIV envelopes circulating during the early development and maturation of breadth generated more-potent antibodies that were modestly cross neutralizing. These data suggest a new approach to identifying envelope immunogens that may be more effective in generating protective antibodies in humans.

INTRODUCTION

A human immunodeficiency virus type 1 (HIV-1) envelope (Env) component that effectively stimulates the humoral arm of the adaptive immune response will be a critical element in future HIV vaccine candidates. Reduction in risk of acquisition of infection was associated with HIV-1 Env-specific antibodies in the RV144 human trial (1–3), but vaccine efficacy was modest. In animal models, Env-specific antibodies have been shown to protect from infection and to control viremia (4, 5); these results are summarized in reference 6. Passive protection studies of macaques with neutralizing antibodies (NAbs) have demonstrated that infection can be blocked and sterilizing immunity can be achieved (7). Thus, a high degree of research effort has been devoted to understanding the development of NAbs in HIV infection (8–10) in order to inform development of an Env-based vaccine to elicit a broad cross-clade neutralizing response in animal models (7, 11). Unfortunately, NAbs elicited by vaccination so far have shown weak to moderate potency and a low degree of neutralization breadth, despite the use of diverse envelopes (Envs) in the vaccines tested (12–19) and novel approaches to develop recombinant Env proteins that mimic the native trimeric Env protein (20). In addition, Envs isolated from different human subjects have divergent antigenic and immunogenic properties (11), further complicating the selection criteria for candidate immunogens. Thus, a major recent undertaking has aimed to discover specific protein sequences or motifs within Env that are associated with neutralization potency and breadth.

Emerging evidence from studies of HIV-infected subjects from the earliest time of infection suggests that founder and subsequent progeny viruses contribute to the broadly neutralizing antibody (bNAb) developmental pathway (17, 21, 22) and that accumulating amino acid changes driven by autologous NAbs (aNAbs) can account for the extensive variability of env (23–25), facilitating viral escape (26). Studies of neutralizing human monoclonal antibodies cloned from elite neutralizers reveal unusual characteristics employed to circumvent the evasive mechanisms of HIV Env. These characteristics include long CDR3 regions and extensive somatic hypermutation (27), surface quaternary epitopes formed by neighboring variable loops (28, 29), and exquisite simulation of interactions with CD4 (30). Prolonged antigenic exposure is a key clinical parameter associated with the natural development of bNAbs in elite neutralizers (24, 31–33), suggesting that the dynamic interactions occurring between viral quasispecies and host B cells result in continuously changing antibody specificities in vivo (24, 25) and likely contribute to the generation of bNAbs (34). In fact, multiple pathways to neutralization breadth have been shown in different subjects (9, 33, 35–38), whereas in some subjects, antibodies with a single specificity or a few specificities can account for much of the neutralization activity (10, 29, 33, 39–42). Thus, further understanding of the humoral response in infected individuals who naturally develop bNAbs could guide the selection of candidate Env immunogens and the design of vaccine strategies that elicit breadth (43). Significant effort has been devoted to understanding this relationship in elite neutralizers, the rare few percentage of infected subjects with very potent bNAbs. In addition, about half of HIV-infected subjects show cross-neutralization in the range of 50% of isolates tested (44), and nearly one fifth to a third eventually develop potent bNAbs (45). Hence, there are likely to be multiple “solutions” to the puzzle of which Env sequences can generate bNAbs in the polyclonal milieu, where antibodies that target different regions can be additive (46).

To address the role of the host quasispecies in the development of bNAbs, our laboratory has been exploring for several years the concept of using sequential immunizations with individual or multiple Env variants isolated from subjects with neutralization breadth to stimulate B cells to develop bNAbs (47, 48). We hypothesize that B cells target conserved, functional regions on Env by sequential exposure to viral variants and that this process can be recapitulated by sequential immunizations using specific circulating Envs isolated from these subjects. We previously reported that sequential immunizations with Env vaccines isolated from a simian-human immunodeficiency virus (SHIV)-infected macaque mimicked more closely the humoral response developed in vivo (47). We expanded this approach to using immunogens from a subtype A-infected human subject, using a bioinformatic approach to select vaccine clones that recapitulate the evolution of quasispecies variants (48). Moreover, we also showed that coimmunization with Env-expressing plasmid DNA and recombinant Env proteins more effectively generated Env-specific antibodies in rabbits (49, 50) and CD8 T cell responses in mice (49). In the current study, we built upon these recent findings and used longitudinally cloned quasispecies gp160 env genes isolated from plasma from two subjects (VC10014 and VC20013). These two subtype B-infected subjects gradually developed moderate bNAbs (33) within a few years of infection by targeting two distinct regions of Env, an epitope overlapping the CD4 binding site in VC10014 and the membrane-proximal external region (MPER) in VC20013 as noted in the accompanying article by Sather et al. (51). We assessed the phylogenetic and neutralization profiles and epitope exposure of the Envs based on binding to neutralizing monoclonal antibodies (NMAbs) and used these data to select a subset of variants to design vaccine strategies in rabbits. The goal was to examine the relationship between epitope exposure, neutralization motifs, neutralization sensitivity, and the time of env isolation relative to early breadth development in the host with their ability to elicit bNAbs.

MATERIALS AND METHODS

VC10014 and VC20013 subjects.

VC10014 and VC20013 are two clade B, antiretroviral therapy (ART)-naive, HIV-1-infected individuals from the Vanderbilt/Center for AIDS Research (CFAR) Cohort (33). During the time of observation, they were not undergoing HIV antiretroviral therapy, and their CD4 T cell counts were above 250 counts/μl. They did not show signs of AIDS-defining illnesses. Plasma samples were available over a period of 5.8 years postinfection (YPI) for VC10014 and over a period of 2.6 YPI for VC20013.

Single-genome amplification and envelope cloning.

Viral RNA was extracted from plasma from HIV-infected subjects VC10014 and VC20013 using the QiaAmp viral RNA extraction kit (Qiagen, Valencia, CA) per the manufacturer's instructions. Any genomic DNA contamination was removed by DNase I treatment with the DNA-free kit (Ambion, Life Technologies, Grand Island, NY). cDNA was generated with the envB3out specific 3′ primer (TTGCTACTTGTGATTGCTCCATGT) using the SuperScript III first-strand synthesis system (Invitrogen, Carlsbad, CA). Single-genome amplification (SGA) of full-length gp160 envelope was performed according to Salazar-Gonzalez et al. (52) using high-fidelity Platinum Taq (Invitrogen, Carlsbad, CA). First-round primers for gp160 SGA amplification are envB5out (TAGAGCCCTGGAAGCATCCAGGAAG) and envB3out (TTGCTACTTGTGATTGCTCCATGT). The first-round conditions are as follows: denaturation at 94°C for 2 min; 35 cycles, with 1 cycle consisting of 94°C for 15 s, 58°C for 30 s, and 68°C for 4 min; and a final elongation at 68°C for 15 min. Second-round primers for gp160 SGA amplification are envB5in NheI (GATCGCTAGCTTAGGCATCTCCTATGGCAGGAAGAAG) and envB3in MluI (GATCGACGCGTGTCTCGAGATACTGCTCCCACCC). The second-round conditions are as follows: denaturation at 94°C for 2 min; 45 cycles, with 1 cycle consisting of 94°C for 15 s, 58°C for 30 s, and 68°C for 4 min; and a final elongation at 68°C for 15 min. Upon satisfying the SGA criteria of fewer than 30% positive results, the 2.95-kb PCR products were treated with exoSAP (Affymetrix, Cleveland, OH) and analyzed by DNA sequencing using BigDye Terminator v3.1 sequencing kits on an Applied Biosystems 3730XL DNA analyzer. The sequencing primers were as follows: 218 (ATCATTACACTTTAGAATCGC), ED5P3mod (ATGGGATCAAAGTCTAGAGCCATGTG), KK1 (GCACAGTACAATGTACACATGGAA), env8R (CACAATCCTCGCTGCAATCAAG), and env6For (GAATTGGATAAGTGGGCAAG). Sequences with early stop codons or with double peaks were discarded. Full-length contigs were built with the Geneious Pro software 5.4.6 (Biomatters, Auckland, New Zealand). Duplicate sequences were identified by ElimDupes (http://www.hiv.lanl.gov/content/sequence/ELIMDUPES/elimdupes.html) and were removed from further analysis. Unique full-length nucleotide sequences were aligned to HXB2 (GenBank accession no. K03455) with HIVAlign (http://www.hiv.lanl.gov/content/sequence/VIRALIGN/viralign.html).

In order to maximize the number of envelope clones generated, the 2.95-kb inserts were first cloned into a TOPO cloning vector (Invitrogen, Carlsbad, CA) per the manufacturer's instructions. Escherichia coli TOP10 bacteria were transformed and grown at 30°C for 24 h. Colonies were restreaked to obtain a clonal population. Restreaked colonies were screened by colony PCR with GoTaq DNA polymerase (Fermentas, Glen Burnie, MD) and primers ED5P3mod (ATGGGATCAAAGTCTAGAGCCATGTG) and ED8 (CACTTCTCCAATTGTCCCTCA). The conditions are as follows: denaturation at 95°C for 5 min; 35 cycles, with 1 cycle consisting of 95°C for 30 s, 60°C for 30 s, and 72°C for 1.5 min; and a final elongation at 72°C for 10 min. Positive colonies giving a 1,078-bp band were grown in small liquid cultures at 30°C for 24 h. Glycerol stocks and plasmid minipreps were made (Promega miniprep kit; Promega, Madison, WI). Once stably inserted in the cloning vector, envelopes were amplified by PCR with Phusion HS II Taq (Thermo Fisher Scientific, Pittsburgh, PA) and primers 014N envB5in NheI pEMC* (GATCGCTAGCACAGAAAATTTATGGGTCACAGTCTAC) and envB3in MluI (GATCGACGCGTGTCTCGAGATACTGCTCCCACCC) to introduce the NheI restriction site for in-frame cloning with the tissue plasminogen activator (t-PA) signal contained in the pEMC* vector. The conditions are as follows: denaturation at 98°C for 30 s; 35 cycles, with 1 cycle consisting of 98°C for 10 s, 62°C for 30 s, and 72°C for 1.5 min; and a final elongation at 72°C for 10 min. The 2.5-kb fragment was gel extracted and digested with NheI and MluI enzymes (New England BioLabs, Ipswich, MA) before ligation to the shrimp alkaline phosphatase (SAP)-treated pEMC* vector with a Roche rapid DNA ligation kit (Roche Diagnostics, Indianapolis, IN). MAX Efficiency Stbl2-competent cells (Invitrogen, Carlsbad, CA) were transformed and grown at 30°C for 24 h. Clonal populations were screened by colony PCR with GoTaq DNA polymerase (Fermentas, Glen Burnie, MD) and primers ED5P3mod and ED8 as described above. Positive colonies (1,078-bp band) were grown in small liquid cultures at 30°C for 24 h. Glycerol stocks were made, and plasmid minipreps were generated with the Promega miniprep kit per the manufacturer's instructions.

Phylogenetic analyses.

Unique nucleotide sequences were aligned to HXB2 (GenBank accession no. K03455) with HIVAlign (http://www.hiv.lanl.gov/content/sequence/VIRALIGN/viralign.html) and manually edited in Geneious to remove indels. The DIVEIN program (http://indra.mullins.microbiol.washington.edu/DIVEIN/) was used to build maximum likelihood (ML) phylogenetic trees rooted on the HXB2 sequence using the HKY85 model. The ML trees were then visualized with the Figtree program. The genetic distance to the most-recent common ancestor (MRCA) was determined with the DIVEIN program (http://indra.mullins.microbiol.washington.edu/DIVEIN/). Positively selected sites in both quasispecies during early and later neutralization breadth development were identified with the TAU Selecton webserver (http://selecton.tau.ac.il/index.html), and the positions are identified according to HXB2 numbering.

Motif optimization.

Each envelope gene was motif optimized using the Robins-Krasnitz algorithm as previously described (48, 53).

Gold bullets.

DNA was precipitated onto 1-μm-diameter gold beads, and bullets were prepared as described previously (47) according to the manufacturer's instructions (Bio-Rad, Hercules, CA). Each bullet was loaded with 2 μg total DNA. To verify that the bullets were functional, COS-7 cells were transfected with the DNA carried by the gold beads and assessed for Env protein expression (47).

Recombinant gp140 proteins.

The gp140 DNA was derived from the gp160 envelope sequence by site-directed mutagenesis (QuikChange multisite-directed mutagenesis kit; Stratagene, La Jolla, CA) to insert the previously described mutations (47, 54) in the primary and secondary protease cleavage sites: REKR was mutated to RSKS, and KAKRR was mutated to KAISS. A large-scale endotoxin-free plasmid preparation (Qiagen, Valencia, CA) was used for stable expression in 293F cells for protein production as previously described (55).

Animals.

Female New Zealand White rabbits (Western Oregon Rabbit Company, Philomath, OR) were housed at the Oregon National Primate Research Center (ONPRC) in Beaverton, OR. Immunizations were performed on rabbits weighing 6 pounds or greater. All procedures were performed according to the rules and protocols approved by the Institutional Animal Care and Use Committee at Oregon Health & Science University (OHSU).

Immunizations.

New Zealand White rabbits (six per group) were coimmunized with DNA and protein at weeks 0, 4, 12, and 20. A total of 36 μg DNA was delivered intradermally by Gene Gun (Bio-Rad, Hercules, CA), at a pressure of 400 lb/in², in 18 immunizations of 2 μg DNA each given in clusters of three nonoverlapping positions at six shaven sites (lower back, inside back legs, and abdomen). Fifty micrograms of recombinant gp140 trimeric protein were delivered intramuscularly by needle injection with polyethylenimine (PEI) (Sigma, St. Louis, MO) as the adjuvant. Blood was collected every 2 weeks after the first immunization; serum was separated and stored at −20°C until the assays were performed. Polyclonal IgG was purified as previously described (47).

ELISA antibody assays.

The binding antibody response to SF162 gp140 trimeric envelope protein was measured by kinetic enzyme-linked immunosorbent assay (ELISA) as described previously (47). Values were standardized to a positive-control rabbit IgG sample from the Malherbe et al. study (47) that was included with each assay. The binding antibody response to gp70(MLV)-V1V2 (MLV stands for murine leukemia virus) (HIV-1/clade B/case A2) protein (Immune Technology, New York, NY) was measured by endpoint ELISA with week 22 serum samples as described previously (56). The peptide ELISA mapping was performed on week 22 serum samples pooled by group with overlapping linear 15-mer peptides for the gp160 clade B consensus sequence as previously described (57).

Antigenic characterization of trimers. (i) Surface plasmon resonance (SPR).

The characterization of gp140 trimers was performed at 25°C on a Biacore T200 instrument (GE Healthcare, Piscataway, NJ) using a CM5 sensor chip and HBS-P+ buffer (10 mM HEPES, 150 mM NaCl, 0.05% surfactant P20 [pH 7.4]). Protein A/G (Pierce, Rockford, IL) was covalently immobilized on all flow cells at a concentration of 50 μg/ml in acetate buffer (pH 4.5) using standard amine coupling to a level of 1,500 to 2,000 resonance units (RU). The nine monoclonal antibodies (VRC01, b12, HJ16, PG9, PGT121, 2G12, 2F5, b6, and F240) were captured onto the protein A/G surface on flow cells 2 to 4, leaving flow cell 1 as a reference. The antibodies were noncovalently immobilized to a level of 400 ± 50 RU by flowing for 25 to 60 s at a concentration of 2 μg/ml. Trimeric gp140 proteins were injected over the sensor surface using single-cycle kinetics at the following concentrations: 1.23, 3.70, 11.11, 33.33, and 100 nM. The association phase was 180 s, and the dissociation phase was 300 s. Samples were maintained at 15°C before injection. Regeneration of the capture complex was achieved using a 60-s pulse of 10 mM glycine (pH 1.7). The data were analyzed using T200 evaluation software, with all data being double reference subtracted and normalized to the level of captured antibody.

(ii) Biolayer interferometry (BLI).

The gp140 trimers were tested at eight concentrations starting at 500 nM or 844.8 nM and titrating 2-fold. Anti-human IgG Fc capture (AHC) octet sensors were used to capture the different antibodies (20 μg/ml) for 300 s. After a 60-s equilibration period in KB buffer (1× phosphate-buffered saline [PBS], 0.01% bovine serum albumin [BSA], 0.02% Tween 20, 0.005% NaN₃), they were dipped into gp140 solutions for a 300-s association phase. The dissociation phase was 600 s. Data points were collected approximately every second.

(iii) ELISA.

The antigenic characterization of gp140 trimeric proteins was performed as follows. Immunosorp plates (Nunc, Rochester, NY) were coated overnight with gp140 trimers at 10 or 20 μg/ml. After the blocking step, monoclonal antibodies were tested in 3-fold dilutions with a starting concentration of 50 μg/ml. Horseradish peroxidase-conjugated recombinant protein A (Invitrogen, Carlsbad, CA) was used as the detection reagent, and the assay was developed as described previously (47).

Binding analyses of polyclonal antibodies. (i) SPR.

Antibody concentrations in rabbit sera and in human plasma to autologous trimeric gp140 proteins 014_092603_G4 and 014_080406-C4a and to heterologous trimer SF612 were determined on a Biacore T200 instrument (GE Healthcare, Piscataway, NJ) as previously described (50).

(ii) BLI.

HIVSF162 gp140 trimeric Env was biotinylated at a 1:1 molar ratio using NHS-PEG4-Biotin system (Thermo Scientific, Rockford, IL) per the manufacturer's instructions. Zeba desalt spin columns (Thermo Scientific, Rockford, IL) were then used to remove free biotin and buffer exchange into PBS. The binding affinities between SF162 gp140 trimer and purified polyclonal rabbit IgG were measured using a ForteBio Octet Red system (ForteBio, Menlo Park, CA). The biotinylated SF162 trimeric gp140 was immobilized on streptavidin biosensors (ForteBio) at a single concentration of 1.25 μg/ml and loaded onto the streptavidin sensors to 50% saturation to obtain a 1:1 binding ratio between antibody and antigen. HIV Env-specific monoclonal antibodies b12 and 447-52D and sCD4 were used to optimize and verify availability of binding sites after biotinylation and binding to the streptavidin surface. After immobilization of the SF162 gp140 trimer to the surface of the sensor, sensors were dipped into six, 2-fold serial dilutions starting at 1,000 nM of the purified rabbit IgG samples for 900 s. The sensors were then moved into kinetic buffer (ForteBio) for an additional 1,800 s for dissociation of the Env-bound IgG. Association and dissociation rates were measured in real time and were calculated using the Octet Molecular Interaction System software, which fit the observed global binding curves to a 1:1 binding model. A buffer-only reference was subtracted from all curves, and pooled preimmune IgG was used as a negative control. Association and dissociation rate constants were calculated using at least three different concentrations of total IgG, to achieve a X² < 1 and R² > 0.90. Equilibrium dissociation constants (K_D) were calculated as the kinetic dissociation rate constant divided by the kinetic association rate constant.

Pseudoviruses.

Pseudoviruses were produced using the pSG3ΔEnv DNA plasmid encoding the HIV backbone and a plasmid encoding either homologous or heterologous envelope variants as described previously (24).

Neutralization assay.

The TZM-bl cell neutralization assay was performed as previously described (26). All values were calculated with respect to virus-only wells [(value for virus only minus value for cells only) minus (value for serum minus virus for cells only)] divided by the value for virus minus value for cells only. The peptide competition assay was carried out with the SF162 V3 peptide as described previously (47).

Statistical analyses.

For the genetic distances from the MRCA, linear regression with VC20013/VC10014 as the indicator variable was used to test differences between regression slopes. For the longitudinal neutralization of SF162 and for the ELISA avidity assays, repeated-measure analysis of variance (ANOVA) with vaccine group as a between-group factor and time after vaccination as a within-group factor was used to explore the effect of vaccine over postinfection time points. Since there is no assumption on within-subject correlation structure over time, unstructured was chosen to be within a covariance structure. Tukey-Kramer adjustment was used for multiple comparisons for the binding kinetic analysis. For SF162 NAbs, the first auto regressive order [AR (1)] was chosen to be within a covariance structure and false discovery rate (FDR) adjustment was used to adjust for multiple comparisons. For the neutralization breadth score analysis, multivariate Poisson regression was used to examine the vaccination variation groups and time effects. A Poisson regression model fits a count or the number of occurrences of an event or the rate of occurrence of an event (breadth response) as a function of some predictor variables (vaccination scheme and time). A generalized estimating equation (GEE) is used to estimate the parameters of a Poisson regression with a possible unknown correlation between outcomes due to repeated measures. The Mann-Whitney test was used to compare neutralization breadth scores across studies. For the gp70-V1V2 ELISA and peptide ELISA mapping studies, the Kruskal-Wallis test followed by Dunn's multiple comparison was used. For the potential N-glycosylation site (PNGS) analyses, a paired t test was used for the individual quasispecies, and due to the unequal variance (P = 0.006), a Welch adjusted two-sample test was used for analyses between quasispecies. The statistical analyses were performed either with GraphPad Prism or with SAS V9.3 software packages.

Nucleotide sequence accession numbers.

All VC10014 and VC20013 sequences included in this study have been deposited in GenBank (accession numbers KJ698244 to KJ698348).

RESULTS

Sources of envelope clones from two subjects with broad NAbs.

VC10014 and VC20013 are both clade B HIV-1-infected subjects from the Vanderbilt/CFAR Cohort (33) who developed broad NAbs (bNAbs) within a few years of presumed time of infection. In this context, we are defining bNAbs as those antibodies that can neutralize 40% or more of the tested pseudovirus panel. The measurements of plasma viral loads (pVL) over a period of 5.8 years (subject VC10014) and 2.6 years postinfection (YPI) (subject VC20013) are shown in Fig. 1. The VC10014 pVL decreased to values below 500 copies/ml for 11 months (Fig. 1A), but this dip was not due to ART treatment, according to available clinical data. Subject VC10014 is coinfected with hepatitis C virus and was treated with interferon during this time frame. Previous studies have shown the beneficial effect of interferon in controlling HIV replication (58), and we speculate that the decrease in plasma viral load observed in subject VC10014 was due to the interferon treatment. Over the 2.6-year observation period, the VC20013 pVL varied within a 1.5-log₁₀-unit range (Fig. 1B).

FIG 1.

VC10014 and VC20013 plasma viral loads and neutralization breadth. (A and B) HIV plasma viral load (pVL) was determined longitudinally and is expressed as the number of viral copies per milliliter of plasma for subjects VC10014 (A) and VC20013 (B). (C and D) Neutralization breadth for subjects VC10014 (C) and VC20013 (D) was assessed longitudinally in a TZM-bl assay against a panel of clade A (4), B (11), and C (5) viruses. Neutralization breadth is expressed as percent pseudoviruses neutralized. YPI, years postinfection.

Breadth of neutralization in plasma at selected time points from each of these subjects was assessed using the TZM-bl assay against a panel of HIV-1 pseudovirus clones. The panel was composed of 20 viruses: 4 clade A tier 2 viruses; 11 clade B viruses (1 tier 1A virus, 1 tier 1B virus, 8 tier 2 viruses, and 1 tier 3 viruses); and 5 clade C viruses (1 tier 1B virus and 4 tier 2 viruses) (51). Plasma from subject VC10014 had developed 20% neutralization breadth within 0.74 YPI, and breadth continued to expand at each time point until the end of the observation period, 5.81 years postinfection (Fig. 1C). In contrast, only 5% breadth was measured in VC20013 plasma at 0.51 YPI, a comparable postinfection time point compared to the 0.74 YPI of VC10014 with 20% breadth (Fig. 1D). VC20013 subsequently developed 40% breadth by 1.01 YPI compared to 40% breadth in VC10014 plasma at 1.79 YPI, indicating a more rapid development of breadth in VC20013 in this time frame. In both subjects, viruses in all three clades were neutralized, but breadth elicited against clade B viruses outpaced overall breadth (Fig. 1C and D). Plasma from each of these time points was also tested against seven sensitive viruses (Fig. 2), and a similar pattern of neutralization rates from time of infection was observed for the two patients. Therefore, the breadth of these two subjects are comparable at approximately 1 to 2 YPI, but the kinetics during early broadening of NAb development varied between subjects.

FIG 2.

Longitudinal heterologous neutralizing antibodies in subjects VC10014 and VC20013. Longitudinal plasma samples were tested for neutralization of four clade B heterologous viruses, one clade A heterologous virus, and two clade C heterologous viruses in a TZM-bl assay. Neutralization data are expressed as ID₅₀, the plasma dilution that neutralized 50% of the infecting virus. Neutralization breadth scores were determined as previously described (72).

Characterization of VC10014 quasispecies env clones.

A total of 120 full-length env genes were cloned from plasma from subject VC10014 obtained at nine longitudinal time points spanning approximately 6 years. After the env genes were tested for functionality by measuring their ability to produce infectious pseudoviruses (data not shown), 50 functional env genes were carried forward for further analysis. The maximum likelihood phylogenetic tree (Fig. 3) shows the evolution of env divergence and diversity over time in the VC10014 quasispecies. The env divergence (or genetic distance) from the most-recent common ancestor (MRCA) was quantified as the rate of env evolution since infection (see Fig. S1A in the supplemental material). Env diversity can be visualized in the Highlighter plot (Fig. S2) displaying the accumulation of synonymous and nonsynonymous changes in both variable and constant regions of gp120 and gp41 starting from the earliest cloned Env nearest to the estimated time of infection. The accumulation of mutations was especially noticeable at late time points (22 March 2006 [3/22/06] and thereafter). In addition, over time, changes in potential N-glycosylation sites (PNGS) accumulated in previously identified hot spots (V1, V2, V4, and V5) (59, 60) and in the C2 region of Env and in gp41 as neutralization breadth increased (data not shown).

FIG 3.

VC10014 maximum likelihood (ML) phylogenetic tree. VC10014 envelopes were generated by single-genome amplification. Sequences of VC10014 functional clones were aligned with HIVAlign (http://www.hiv.lanl.gov/content/sequence/VIRALIGN/viralign.html) to clade B reference HXB2 in order to generate the ML tree rooted to HXB2 with the DIVEIN program (http://indra.mullins.microbiol.washington.edu/DIVEIN/). Bar, scale for genetic difference.

VC10014 envelopes were characterized for their neutralization sensitivity to well-known broadly neutralizing monoclonal antibodies (NMAbs) (VRC01, b12, 2G12, PG9, PG16, 2F5, and 4E10) and to several autologous plasma specimens (autologous NAbs [aNAbs]) from different time points (51). Based on these neutralization assays, VC10014 env genes were sensitive to NMAbs targeting the CD4 binding site and varied over time in sensitivity to NMAbs to the MPER region but were not sensitive to V1V2-targeting NMAbs PG9/16 (29). Further, VC10014 env quasispecies were shown to follow the patterns previously described with earlier env genes being more sensitive than later env genes to aNAbs (23, 25) indicative of escape from immune pressure (data not shown).

Characterization of VC20013 quasispecies env clones.

We isolated a total of 70 full-length env sequences from five longitudinal VC20013 plasma samples: 9/2/04 and 9/23/04 (estimated at 0.5 to 0.6 YPI) before the appearance of bNAbs; and 3/3/05, 4/12/06, and 10/5/06 (three samples collected after the appearance of bNAbs estimated at 1.01, 2.12, and 2.6 YPI, respectively). We first characterized these clones in silico (48), and the relatedness of the 55 functional clones is shown in the maximum likelihood phylogenetic tree (Fig. 4). Similar to VC10014, the env divergence (quantified by the genetic distance from the MRCA) shows a steady rate of env evolution since the time of infection (see Fig. S1B in the supplemental material), but the slope (rate of change) in VC20013 was significantly greater than that in VC10014 (P = 0.0001). The Highlighter plot displays an accumulation of silent and nonsilent mutations in multiple regions of envelope (V1V2, C2, C3, V4, V5, and gp41) since infection (Fig. S3), particularly at late time points (4/12/06 and 10/5/06). Further, PNGS remained relatively stable (data not shown), which is in contrast to VC10014 and some studies associating the addition, subtraction, and relocation of PNGS with resistance to neutralization (26, 61).

FIG 4.

VC20013 maximum likelihood (ML) phylogenetic tree. VC20013 envelopes were generated by single-genome amplification. Sequences of VC20013 functional clones were aligned with HIVAlign (http://www.hiv.lanl.gov/content/sequence/VIRALIGN/viralign.html) to clade B reference HXB2 in order to generate the ML tree rooted to HXB2 with the DIVEIN program (http://indra.mullins.microbiol.washington.edu/DIVEIN/).

No aNAbs were detected against pseudoviruses expressing the earliest Envs (51), suggesting that VC20013 quasispecies env genes were naturally evolving without the selective pressure of aNAbs until the appearance of bNAb activity (3/3/05 time point). Plasma from subsequent time points displayed increased neutralization potency against previous variants and detectable activity against concurrent variants, consistent with the continuous selection and escape described in the literature (24, 62, 63) and for subject VC10014.

The characterization of VC20013 Env clones by NMAb neutralization showed a range of sensitivities (51). VC20013 Envs were sensitive to NMAbs targeting the CD4bs and the MPER region. Most VC20013 Envs from the later time points were resistant to 2G12 due to the absence of necessary PNGS (N386, N392, N397, or N448) (64, 65). Only five VC20013 Env-based pseudoviruses were sensitive to PG9/PG16 even though all but one clone had the critical residues for interaction with PG9/PG16 (29). Interestingly, two of these clones, isolated from 9/23/04 plasma, were neutralized at low titers by plasma from an earlier time point (9/2/04), despite being resistant to contemporaneous neutralization (51). We propose that the rapid development of bNAbs in the VC20013 subject, as well as the minor divergence and diversity displayed by env genes isolated prior to this event suggest that clones from early neutralization breadth time points possess characteristics that generated the NAb activity detected in later plasma.

Motifs correlated with bNAb development found in both quasispecies.

We investigated whether the VC10014 and VC20013 quasispecies harbored motifs that correlated with neutralization breadth. All VC10014 Envs contain the NIS signature motif at PNGS 332; a motif that is associated with less neutralization breadth than the NLS motif (66), which is found in all VC20013 Envs. Furthermore, even though both VC10014 and VC20013 quasispecies have a higher number of NXT glycosylation sites than NXS sites (P < 0.0001 for both [Fig. S4 in the supplemental material]), the proportion of NXS-to-NXT PNGS is significantly higher in subject VC20013 (P < 0.0001 [Fig. S4]). The NXS motif is less likely to be glycosylated, and a lower level of glycosylation was recently associated with the development of breadth (66); thus, we speculate that this may have contributed to more rapid development of breadth in the VC20013 subject (less than 2 years to neutralize 50% of tested strains) than in the VC10014 subject (almost 3 years to neutralize 50% of tested strains).

Several cross-sectional studies have identified Env sequence motifs associated with the development of neutralization breadth (66, 67), and VC10014 and VC20013 quasispecies display some of these motifs. For example, VC10014 and VC20013 share two of the signatures previously identified (67): the R_K motif (positions 419 to 421 [HxB2 numbering]) in the coreceptor binding site and the arginine at position 419 which was also independently identified as a signature of breadth. In addition, 86% of VC10014 Envs and 84% of VC20013 Envs have an asparagine at position 186, a position also associated with breadth (67). However, one caveat is that these motifs were identified in cross-sectional studies of viral populations, whereas our data sets are two longitudinal quasispecies so the cross-sectional results may not be predictive for the individual quasispecies examined here. Indeed, Sather et al. (51) identified previously unknown amino acid changes associated with the development of breadth in the quasispecies of both subjects. Neutralization in VC10014 targets an epitope that overlaps the CD4 binding site, and the F277I and N279D mutations in C2 and a four-amino-acid insertion in V4 were associated with the development of breadth. In contrast, neutralization in VC20013 targets the MPER region and a K677N change correlated with the development of breadth in this subject.

To analyze for positive selection, we compared the sequences from the early neutralization breadth time points (defined as neutralization breadth of <40%) (9/26/03 to 10/15/04 in subject VC10014 and 9/2/04 to 3/3/05 in subject VC20013) with later time points (defined as breadth of >40%) (3/22/06 to 10/22/08 in VC10014 and 4/12/06 to 10/5/05 in VC20013). In both quasispecies, distinct sites were under pressure early and later in the development of neutralization breadth (see Fig. S5 in the supplemental material). The number of positively selected sites in the early versus later breadth in the VC10014 quasispecies did not increase significantly (19 versus 22 sites), whereas in the VC20013 quasispecies, this number increased greatly (6 versus 22 sites). This difference in number of positively selected sites suggests that the earliest VC10014 clones may already display escape mutations.

Vaccine clone selection and experimental design.

Sequences from each time point were selected as vaccines (Table 1) based on the in silico and in vitro analyses described above and in the accompanying article (51). The chosen env genes display several motifs associated with neutralization breadth (Fig. 5) identified in the literature (67) or in the accompanying article by Sather et al. (51). We used these criteria to choose representative clones from the full set of gp160 clones for each subject. As noted above, all selected VC10014 envelopes and 92% of selected VC20013 envelopes display an asparagine at position 186, and all possess the R_K motif at position 419 to 421. In contrast to VC10014, the VC20013 Envs have the NLS PNGS at position 332, which was recently linked with breadth (66). Other motifs associated with the development of breath are specific to each quasispecies; VC10014 Envs display the F277I and N279D changes in C2 and the four amino acid insertion in V4, whereas VC20013 Envs have the K677N mutation in MPER.

TABLE 1.

VC10014 and VC20013 vaccine immunogens and strategies

Subject and group (n = 6)	Vaccine immunogen
	Wk 0		Wk 4		Wk 12		Wk 20
	DNA	Protein	DNA	Protein	DNA	Protein	DNA	Protein
VC10014
Clonal	092603 G4	092603 G4	092603 G4	092603 G4	092603 G4	092603 G4	092603 G4	092603 G4
Sequential	092603 G4	092603 G4	041504 F8	041504 F8	032206 H5a	032206 H5a	080406 C4a	080406 C4a
	092603 D9		041504 G10a	101504 C6a	032206 D10		111006 E1
	101603 F1		041504 G6a		032206 E9a		102208 B4a
	101603 H7		101504 C6a
	011504 B9		101504 E5a
	011504 C10		101504 H10
Simplified sequential	092603 G4	092603 G4	041504 F8	041504 F8	032206 H5a	032206 H5a	080406 C4a	080406 C4a
			101504 C6a	101504 C6a
Early breadth	041504 F8	041504 F8	041504 F8	041504 F8	041504 F8	041504 F8	041504 F8	041504 F8
	041504 G10a	101504 C6a	041504 G10a	101504 C6a	041504 G10a	101504 C6a	041504 G10a	101504 C6a
	041504 G6a		041504 G6a		041504 G6a		041504 G6a
	101504 C6a		101504 C6a		101504 C6a		101504 C6a
	101504 E5a		101504 E5a		101504 E5a		101504 E5a
	101504 H10		101504 H10		101504 H10		101504 H10
VC20013
Sequential	092304 c5	092304 c5	092304 c5	092304 c5	041206 c13	041206 c13	041206 c13	041206 c13
	092304 c19	030305 c5	092304 c19	030305 c5	041206 c16	100506 c19	041206 c16	100506 c19
	030305 c5		030305 c5		100506 c1		100506 c1
					100506 c19		100506 c19
Later sequential	030305 c11	030305 c11	030305 c11	030305 c11	041206 c13	041206 c13	100506 c1	100506 c19
	030305 c14		030305 c14		041206 c16		100506 c19
Early breadth S/R	092304 c5	092304 c5	092304 c5	092304 c5	092304 c15	030305 c5	092304 c1	092304 c1
	092304 c19		092304 c19		030305 c5		092304 c13
Early breadth R/S	092304 c1	092304 c1	092304 c1	092304 c1	092304 c15	030305 c5	092304 c5	092304 c5
	092304 c13		092304 c13		030305 c5		092304 c19

FIG 5.

Motifs associated with neutralization breadth in VC10014 and VC20013 vaccine clones. Protein sequences of vaccine clones were aligned to clade B HXB2 (HIVAlign [http://www.hiv.lanl.gov/content/sequence/VIRALIGN/viralign.html]), and the motifs associated with breadth are boxed and identified according to HxB2 numbering. The sequence logo at the top of the alignment is representative of the proportion of sequences that harbor a given amino acid residue. Gaps introduced to maximize sequence alignment are indicated by dashes.

The goal of the current study was to investigate whether the sequential delivery of Envs is needed to stimulate neutralization breadth or whether using a subset of Envs from carefully chosen time points could elicit heterologous neutralizing antibodies in vaccinated animals. Therefore, we compared four immunization strategies for each subject, each incorporating different combinations of Envs (Table 1). For subject VC10014, we compared a clonal strategy (a single env from the earliest time point as a control), a sequential strategy (multiple env genes spanning all time points cloned and delivered sequentially), and a condensed sequential strategy, simplified sequential method (only one or two env genes per time point) as a comparison to using multiple clones at each time point. Finally, we also tested only the env clones derived from the time point when breadth first appeared against clade B viruses and the time point immediately preceding it (early breadth [time points 4/15/04 and 10/15/04]). The 4/15/04 Envs have either FTN or ITN motifs at positions 277 to 279, and all of the 10/15/05 Envs have ITN and the F277I change that was associated with the development of neutralization breadth. In addition, all 10/15/04 Env immunogens have an insertion of four amino acids in V4 that is also associated with the development of breadth in this subject (Fig. 5).

We also designed four VC20013 strategies to compare immunogenicity in rabbits (Table 1). These strategies included a sequential group to address whether sequential delivery of env genes representing the timeline of bNAb development could stimulate bNAbs. This group is most similar in concept to our prior vaccines (47). The second group (later sequential group) addressed whether circulating Envs isolated after bNAb appearance would be effective in generating bNAbs if used alone without the earlier clones to prime the responses. Two groups were coimmunized with DNA and proteins derived from early neutralization breadth time points (early breadth group) to assess the immunogenicity of env variants that existed before bNAb appearance. Considering that variants isolated from 9/23/04 and 3/3/05 displayed minimal env sequence diversity, we reasoned that differences in sensitivity to NMAbs PG9 and PG16 might reflect structural differences. We immunized one of these groups of rabbits sequentially with clones of increasing neutralization resistance to these NMAbs (early breadth S/R group [S stands for sensitive, and R stands for resistant]). The second group received the same clones in reverse order, i.e., from resistant to sensitive (early breadth R/S group) to assess whether a progressive exposure of the epitope would better guide the immune system toward eliciting broader NAbs. These VC20013 early breadth immunogens display either a lysine at position 677 (in 9/23/04 Envs) or an asparagine (3/3/05_c5 Env), and the K677N change was associated with the development of breadth in the VC20013 subject (Fig. 5) (51).

We performed an analysis of the antigenic profiles of the Envs selected as protein immunogens to determine whether VC10014 and VC20013 purified recombinant uncleaved gp140 trimers displayed specific epitopes as determined by MAb binding using surface plasmon resonance (SPR), ELISA, or biolayer interferometry (BLI) (Fig. 6; see Fig. S6 in the supplemental material). All gp140s bound to weak or nonneutralizing antibodies b6 and F240. The MAb b6 also bound BG505 SOSIP while F240 did not, suggesting that cluster I gp41 epitopes, which become exposed during gp120 shedding, are more easily accessed on these trimers than on BG505 SOSIP (20). Interestingly, nonneutralizing MAb F240 has been shown to be protective against vaginal transmission of SHIV to macaques largely due to its strong ability to capture infectious virions (68). Eliciting a potent neutralizing antibody response is primary to HIV vaccine design. However, presentation of the F240 epitope for potential elicitation of protective, but weakly neutralizing antibodies may not be detrimental and may be additive to a polyclonal response induced by vaccination.

FIG 6.

Antigenic characterization of VC10014 and VC20013 gp140 trimers by monoclonal antibodies 2G12, PG9, VRC01, and 2F5. (A) SPR binding responses of VC10014 and VC20013 gp140 trimers to four monoclonal antibodies. The different antibodies were captured onto protein A/G. The gp140 trimers were injected over the captured antibodies using a single-cycle kinetic method with gp140 concentrations ranging from 1.23 nM to 100 nM. The binding responses (expressed as resonance units [RU]) are adjusted to the antibody capture level. (B) BLI binding responses of VC10014 and VC20013 gp140 trimers to four monoclonal antibodies. The antibodies were captured on anti-human Fc sensors. The gp140 trimers were tested in a 2-fold dilution dose-response experiment with a starting concentration of 500 nM or 844.8 nM. The binding responses (expressed as nm shift) for the optimal gp140 concentrations are displayed.

All proteins bound to several potent and broad NMAbs that target conformational epitopes on the trimeric Env spike (27, 29). Of the NMAbs tested, the most variable binding was seen to HJ16, PG9, and 2F5. As expected, the NMAb 10E8 did not bind to these trimers (data not shown), all of which are missing part of the binding site due to the design of the trimer. In contrast, we measured strong 2G12 binding to all of the VC10014 and VC20013 trimers, as has been reported for BG505 SOSIP (20). In addition, binding of VRC01 and PGT121 to the VC10014 and VC20013 trimers was in many cases strong, as seen with another recently characterized gp140 functional mosaic trimer (69). Therefore, this antigenic analysis shows that VC10014 and VC20013 gp140 trimers display many of the known conformation-dependent NMAb epitopes. These antigenicity data are in line with the neutralization by NMAbs that was performed with the native Envs displayed on the surfaces of pseudoviruses (Fig. 4) (51), where neutralization with PG9 and PG16 was observed only for VC20013 092304_c5 and 030305_c5. Notably, these two gp140 trimers bound well to PG9 in SPR and BLI assays. This suggests that these gp140 trimeric protein immunogens present bNMAb epitopes and contain many characteristics of natively expressed Envs.

Binding antibody responses to VC10014 and VC20013 vaccines.

Using the four strategies per subject described above (Table 1), rabbits (six per group) were coimmunized with gp160 envelope DNA and gp140 trimeric Env protein, with the same delivery methods and schedule for both sets of vaccines. Binding antibody characteristics were monitored for the two sets of rabbits longitudinally against the heterologous neutralization-sensitive Env SF162 (Fig. 7A and B). Env-specific binding antibodies in all groups were detected after only two coimmunizations with DNA and trimeric protein. By this measure, there was no significant difference between groups immunized with either VC10014 or VC20013 Env immunogens.

FIG 7.

SF162-specific longitudinal binding and neutralizing antibodies. (A and B) Serum samples from VC10014-vaccinated (A) and VC20013-vaccinated (B) rabbits were tested longitudinally for binding antibodies to SF162 gp140 trimeric protein by kinetic ELISA. (C and D) Serum samples from VC10014-vaccinated (C) and VC20013-vaccinated (D) rabbits were tested longitudinally for neutralization antibodies to SF162 by TZM-bl assay. The P values were as follows: for VC10014 study, P = 0.0396 for early breadth group versus clonal group and P = 0.0396 for early breadth group versus sequential group; for VC20013 study at week 22, P < 0.05 for early breadth R/S group versus all three other groups.

Antibody affinity is a major characteristic of antibody-based vaccines (47, 70, 71), and due to the polyclonal nature of our samples, we assessed the apparent affinity of purified rabbit IgG samples to trimeric SF162 protein in the four highest responders of each vaccine group after the second, third, and fourth coimmunizations (weeks 6, 14, and 22, respectively) using biolayer interferometry. We measured each sample individually to determine overall kinetic binding characteristics, association constant (K_on), dissociation constant (K_dis), and K_D (see Table S1 in the supplemental material). The K_D values were not different between groups immunized with either VC10014 or VC20013 Env immunogens (Fig. 8A and B), but overall, the average affinity increased significantly between the second and third coimmunizations as well as between the second and fourth coimmunizations in both sets of experiments as shown by the decrease in K_D (P < 0.0001) between these time points to levels in the 10 nM range (Fig. 8C and D and Table S1). K_D values between the third and fourth coimmunizations in the VC10014 groups were similar (P = 0.9348), but the fourth immunization improved the affinity in the VC20013 groups (P = 0.0490), thus implying that the improvement of antibody affinity in this setting is influenced both by the number of immunizations and the composition of the vaccine.

FIG 8.

SF162 binding antibody affinity. (A and B) Polyclonal IgG samples from the four highest responders in each VC10014 (A) and VC20013 (B) immunization group were tested for antibody affinity to SF162 gp140 trimeric protein with the ForteBio Octet Red system. Data are K_D values (nanomolar) after the second, third, and fourth coimmunizations (weeks 6, 14, and 22, respectively). The colors of the group are the same as in Fig. 7. (C and D) The effect of the number of immunizations on the evolution of K_D values was assessed in VC10014-vaccinated (C) and VC20013-vaccinated (D) rabbits.

In addition, surface plasmon resonance was used to measure the amount of antigen-specific responses (50) toward autologous G4 (time point 9/26/03), C4a (time point 8/4/06), and heterologous SF162 trimeric proteins in serum samples from VC10014-vaccinated rabbits (rabbits vaccinated with DNA and protein from subject VC10014) after the second, third, and fourth immunizations (see Fig. S7A in the supplemental material). We found that regardless of the immunization strategy, similar levels of antigen-specific responses were elicited against autologous and heterologous proteins in all vaccine groups. However, the amount of antigen-specific responses increased over time for each antigen (P ≤ 0.0110 between immunizations 2 and 4 and P ≤ 0.0184 between immunizations 3 and 4). The antigen-specific responses measured longitudinally in plasma from human subject VC10014 were up to 20 times higher in magnitude than in the vaccinated rabbits (Fig. S7B) and demonstrate the difference between antigen-specific antibody responses elicited here by immunization compared to those resulting from natural infection.

Finally, to investigate which regions of Env were targeted by binding antibodies in rabbits vaccinated with DNA and protein from subjects VC10014 and VC20013 (VC10014- and VC20013-vaccinated rabbits), we assessed the responses to gp70-V1V2 and performed a peptide ELISA mapping study with gp160 linear overlapping peptides. We found that all VC10014 and VC20013 vaccine strategies elicited similar levels of V1V2-specific binding antibodies (see Fig. S8 in the supplemental material). In the VC10014-vaccinated animals, the clonal strategy sera had lower levels of V1V2-specific binding antibodies, but there was no statistical difference between the approaches (Fig. S8). Peptide ELISA mapping of VC10014-vaccinated rabbit sera revealed that some characteristics were shared by all strategies, such as a complete lack of V4, V5, and MPER targeting. In addition, all VC10014-vaccinated groups recognized C1, C4, the N-terminal region of V3, and the immunodominant region of gp41 (Fig. S9), but no differences were seen between groups (P > 0.05 for gp120, gp41, and MPER). Similar to what was observed in VC10014-vaccinated rabbits, there was little targeting of V4 and V5 regions by all VC20013 strategies, and there was no difference in overall gp120 targeting between groups. However, the VC20013 vaccine approaches resulted in stronger targeting of gp41 (P < 0.05 for sequential group versus all three other groups) but not of MPER, despite the higher responses observed in the sequential and later sequential groups (Fig. S9).

NAb responses elicited by VC10014 and VC20013 vaccines.

Neutralization of SF162 was used to monitor the longitudinal development of neutralization activity in rabbit immune sera (Fig. 7). Neutralization activity against SF162 was observed in all VC10014 groups after two immunizations, and the potency of neutralization increased after subsequent vaccinations (Fig. 7C). The early neutralization breadth strategy elicited more potent SF162 NAbs compared to the clonal and sequential approaches (P = 0.0396 versus clonal strategy group and P = 0.0396 versus sequential strategy group). In VC20013-vaccinated rabbits (Fig. 7D), the early breadth R/S strategy elicited significantly better SF162 neutralization at week 22 than the other three VC20013 vaccine groups (P < 0.05), all of which generated similar responses. Overall, the VC20013-vaccinated rabbits had slightly higher titers of SF162 NAbs than VC10014-vaccinated animals. However, SF162 neutralization targeted the V3 loop in all VC10014- and VC20013-vaccinated rabbits (determined by V3 peptide competition), and there was no significant difference between groups (data not shown).

Heterologous neutralization was assessed as a measure of the effectiveness of the different regimens in generating bNAbs. Serum samples collected after the third and fourth immunizations were tested against the panel of seven heterologous viruses (Fig. 9) that was also evaluated against human plasma samples from selected time points (Fig. 2). Titers against tier 1A viruses (SF162, Q461e2TAIV, and MW965) were higher in VC20013-vaccinated rabbits than in VC10014-vaccinated rabbits. The level of neutralization of tier 1B and tier 2 clade B viruses was similar in both vaccine studies, but more rabbits per group in VC20013 vaccine groups had NAbs against these viruses than in the VC10014 vaccine groups. In addition, for both studies, the fourth immunization improved the neutralization breadth only for a few rabbit samples, suggesting that shortening the immunization strategy to three administrations may be sufficient to elicit this level of heterologous neutralization. Breadth scores for each vaccine group (72) were not statistically greater between the third and fourth immunizations (P = 0.60 in the VC10014 study and P = 0.25 in the VC20013 study). Interestingly, the potency of the SF162 neutralization response correlates positively with the breadth scores in both VC10014 and VC20013 studies at both time points studied (in the VC10014 study, P = 0.0008 and P < 0.0001 after the third and fourth immunizations, respectively; in the VC20013 study, P = 0.0087 and P < 0.0001 after the third and fourth immunizations, respectively). Within the VC20013 study, breadth scores were not statistically different between groups, showing that the order of presentation in the early breadth strategies or the inclusion of later clones (escape mutants) did not improve or alter the neutralization breadth, which was moderate in all groups. In contrast, in the VC10014 study, breadth scores were significantly higher in the early breadth group than in the clonal and simplified sequential groups (P = 0.036 and P = 0.0015, respectively). In a final analysis performed to compare the effectiveness of env genes per subject in generating neutralization breadth, we found that the breadth scores were significantly higher in VC20013-vaccinated rabbits than in VC10014-vaccinated rabbits after three and four immunizations (Fig. 10), with rabbits from each vaccine group identified using distinguishable symbols and colors. By this criterion, the VC20013 quasispecies overall was a better source of immunogens than the VC10014 quasispecies env genes that primarily limited the strongest responses to rabbits receiving the early breadth VC10014 sequences.

FIG 9.

Longitudinal heterologous neutralizing antibodies elicited by VC10014 and VC20013 vaccine strategies. Rabbit serum samples after the third immunization (week 14) and fourth immunization (week 22) were tested for neutralization of four clade B heterologous viruses, one clade A heterologous virus, and two clade C heterologous viruses in a TZM-bl assay. Neutralization data are expressed as ID₅₀, the serum dilution that neutralized 50% of the infecting virus. Neutralization breadth is expressed as percent pseudoviruses neutralized (number of neutralized isolates divided by total numbers of viruses in panel). Breadth scores were determined, and for the VC10014 study, P values were as follows: P = 0.036 early breadth group versus clonal group and P = 0.0015 early breadth group versus the simplified sequential group.

FIG 10.

Higher breadth scores in VC20013-vaccinated rabbits. The breadth scores in VC10014- and VC20013-vaccinated rabbits were compared after three (A) and four (B) immunizations. Groups are color coded (groups assignments indicated in the keys below the panels), and the scores for individual rabbits are plotted (mean [black line] ± standard error of the mean [error bars]).

DISCUSSION

One of the many hurdles that HIV vaccinologists face is to design immunization strategies that address the low immunogenicity of conserved neutralization determinants on HIV Env that are the targets of bNMAbs. One active area of research for the past decade or more has been to design Env proteins that are more effective in presenting these neutralization determinants; to date, these approaches have not significantly increased the neutralization breadth of NAbs elicited through vaccination (73–75). An alternative approach that we have taken is to vaccinate with natural quasispecies sequences that have evolved during the development of breadth. We identified two clade B-infected subjects who developed moderate breadth within a few years of infection by targeting different regions of Env, and we cloned the dominant plasma Env gp160 genes at closely spaced time points during bNAb development. We hypothesized that the outgrowth of the major viral quasispecies in subjects with bNAbs results from the exposure of B cells to structural or sequence motifs that subsequently elicit bNAbs. We performed comparative vaccine studies designed to discriminate differences between types of sequential immunizations with these diverging Env clones in an effort to reveal relationships between the time of appearance of the variants and the gradual increasing development of breadth in the subject. To reduce the complexity of the in vivo experiment, we used neutralization sensitivity and bioinformatics to choose a subset of cloned Envs for testing in rabbits. These experiments revealed that early breadth time points are promising sources of Env immunogens and that quasispecies obtained from two unrelated HIV-infected subjects who developed moderate neutralization breadth could each generate modest cross-clade NAbs.

By keeping several factors constant (animal species, adjuvant, immunization schedule, and routes) and by using comparable vaccine strategies, we observed very similar responses using env genes from unrelated sources. The antigenicity analysis of VC10014 and VC20013 gp140 trimers showed that these trimeric proteins display conformational epitopes of many known broad NMAbs. The VC10014 and VC20013 DNA-plus-protein vaccine candidates were highly immunogenic in rabbits, as shown by the enhanced SF162 neutralization potency by all vaccine strategies compared to SF162 DNA and Env trimer coimmunization (50). The VC10014 early breadth group neutralized SF162 better than the clonal and sequential groups, and the VC20013 early breadth R/S strategy also elicited significantly better SF162 neutralization at week 22 than the other three VC20013 vaccine groups. Therefore, despite testing Env immunogens derived from different subjects who developed moderate cross-clade neutralization breadth through distinct pathways (51) and with different kinetics, our data show that a common vaccine strategy based on Envs derived from the time points preceding, or contemporaneous with, the development of breadth is a promising approach to elicit NAbs in vaccinated animals. The potential of sequential immunization with early variants from broad neutralizers was also observed in recent studies showing that concomitant antibody maturation and virus evolution occurred in a subject with breadth (21) and that transmitter/founder Env immunogens elicited low, but broader NAb responses than consensus or chronic Envs (17).

Despite the potent neutralization of SF162, only modest neutralization breadth and marginal autologous neutralization were elicited by any of the VC10014 or VC20013 vaccine strategies tested here. Only tier 1A viruses were neutralized at moderate to high levels, and similar to previous vaccine studies (12, 17, 19, 35, 36, 48), almost no autologous neutralization was elicited (data not shown). This lack of autologous neutralization and limited heterologous neutralization compared to the moderate cross-clade neutralization breadth in the infected VC10014 and VC20013 subjects could be explained by several factors. First, it is possible that our immunogens do not fully replicate the Env conformation as it appears on the virion during a natural infection (20, 76, 77) as previously observed with clade A Env immunogens (35) that may have resulted in altered immunogenicity. The potential to engage germ line B cell receptors (BCR) of broad NMAbs has recently emerged as an important feature to drive antibody maturation and elicit elite neutralization status, but the VC10014 gp140 proteins used in this study bind only to the mature form of b12 and not to the germ line-reverted b12 (78). In addition, we speculate that VC10014 and VC20013 envelopes would be unable to engage VRC01 germ line BCR due to a PNGS at position 276 (79), thus failing to induce B cell maturation that would lead to the generation of VRC01-like broad NAbs. Differences in neutralization potency and breadth in the human plasma and rabbit serum samples are likely due at least in part to the limited antigen exposure of this vaccine regimen and may be enhanced by alternate deliveries, including recombinant viral vectors. When influenza virus antibody titers were assessed in infected versus vaccinated subjects, a 3-fold difference was detected (80). Moreover, the lack of aNAbs and of bNAb breadth in rabbits compared to macaques and humans may be due to differences in CD4 molecules or in V_H gene usage in these species (81, 82).

Developing a shorter vaccination scheme is another important goal of HIV vaccinologists; we contend that using a DNA-plus-protein coimmunization rather than a DNA prime-protein boost will help achieve this objective without sacrificing the quality of the elicited humoral response. Our data show that three DNA-plus-protein coimmunizations with primary envelope genes and trimeric proteins induced the maximum responses in the vast majority of assays tested in this study. Antibody affinity data from these experiments indicate that the greatest increase in affinity was between the second and third immunizations, with additional modest increases in K_D to 2 to 10 nM between the third and fourth immunizations using VC20013 Envs. However, the neutralization breadth using this panel of HIVs was not statistically different between groups of rabbits immunized with Envs from either subject after three and four covaccinations. Nevertheless, breadth scores were significantly greater with the VC20013 Envs than the VC10014 Envs after both three and four immunizations. These results suggest that the immunization schedule could be shortened from 20 to 12 weeks. Our results are consistent with those of other groups exploring DNA-plus-protein coimmunizations with purified monomeric HIV gp120 or aldrithiol 2 (AT-2)-inactivated viral particles (83, 84). Additionally, the development of binding and neutralizing antibodies was similarly accelerated by DNA-plus-protein coimmunization in vaccine studies against dengue virus and Japanese encephalitis virus (85). The regimen that we describe here certainly could be further optimized, but the results already suggest that only a few vaccinations may be effective, even in complex multiple Env vaccine compositions.

In conclusion, Envs derived from two subjects at time points preceding or contemporaneous with the development of neutralization breadth generated potent tier 1 NAbs. Other vaccine studies in guinea pigs or rabbits comparing Envs derived from elite neutralizers, T/F sequences, and consensus Envs have resulted in similar, modest neutralization potency and breadth against panels of viruses and significant variation in individual animals (17, 19). These studies, as well as our prior vaccine studies using sequentially derived Envs (47, 48), have contributed to an appreciation of the antigenic properties of Env components and how combinations of Envs can be used, but there are still no clearly superior sequences that have been identified. It was encouraging to us that native Envs selected from two different subjects consistently generated NAbs in multiple rabbits that were broader and more potent than those generated in our hands using Envs from other sources of clones using the same delivery strategy (48, 50). Our prior approach using many more Envs to drive this response (47) may not be necessary if fewer and more effective sequences can be identified. In addition, using quasispecies from HIV-infected individuals who developed moderate neutralization breadth is a potentially valuable alternative approach to designing Env-based vaccine strategies, since the earliest development of NAbs may inform how breadth can intensify and target conserved determinants. Overall, these vaccine studies support the concept that certain viral clones present in the circulating quasispecies of infected individuals who develop modest breadth have sequence or structural motifs that facilitate bNAb development and that these sequences arise early in infection. Thus, if these results can be further expanded upon, it may not be necessary to engineer new Envs or to use Envs from elite neutralizers to reach the goal of eliciting significant potency and breadth. The obvious next step, beyond the scope of this study, is to investigate the polyclonal responses generated by these vaccines in macaques, where it is possible to examine individual B cell clones to better understand targeting and pathways of NAbs (86). A way forward is to combine these findings with the latest data on immunogen design to engage the germ line BCR (87) and to deconvolute the complexity of the polyclonal targeting of serum neutralization to predict component-antibody specificity (88) with the goal of unveiling new pathways to improve the immune response by vaccination.