Sequential Immunization with a Subtype B HIV-1 Envelope Quasispecies Partially Mimics the In Vivo Development of Neutralizing Antibodies^,,

Abstract

A major goal of human immunodeficiency virus type 1 (HIV-1) vaccine efforts is the design of Envelope (Env)-based immunogens effective at eliciting heterologous or broad neutralizing antibodies (NAbs). We hypothesized that programming the B-cell response could be achieved by sequentially exposing the host to a collection of env variants representing the viral quasispecies members isolated from an individual that developed broad NAbs over time. This ordered vaccine approach (sequential) was compared to exposure to a cocktail of env clones (mixture) and to a single env variant (clonal). The three strategies induced comparable levels of the autologous and heterologous neutralization of tier 1 pseudoviruses. Sequential and mixture exposure to quasispecies led to epitope targeting similar to that observed in the simian-human immunodeficiency virus (SHIV)-infected animal from which the env variants were cloned, while clonal and sequential exposure led to greater antibody maturation than the mixture. Therefore, the sequential vaccine approach best replicated the features of the NAb response observed in that animal. This study is the first to explore the use of a collection of HIV-1 env quasispecies variants as immunogens and to present evidence that it is possible to educate the B-cell response by sequential exposure to native HIV-1 quasispecies env variants derived from an individual with a broadened NAb response.

INTRODUCTION

Human immunodeficiency virus type 1 (HIV-1) evolves rapidly within the host, resulting in the accumulation of diverse HIV-1 variants called a viral quasispecies population. Most variable in the genome is the env gene, which encodes the 160-kDa glycoprotein designated Envelope (Env). Env is embedded in the membrane of HIV-1 as it buds from infected cells and is the only target of neutralizing antibodies (NAbs), which have the capacity to bind to the virus and prevent the infection of target cells in vitro. There are two types of NAbs: autologous, which are directed against viral variants present in the host's quasispecies, and heterologous, which can neutralize viruses that the host did not encounter, presumably by binding to conserved determinants shared with the infecting isolates. The development of an autologous NAb response usually occurs during the first year of infection, whereas it can take several years to develop heterologous NAbs (20, 41), and only a subset of subjects develop NAbs capable of neutralizing heterologous isolates (45). Previous studies with animal models have shown that the passive transfer of HIV-specific IgG and/or neutralizing monoclonal antibodies (MAbs) can be protective against intravenous and mucosal viral challenges, including protection against heterologous viruses of the same genetic subtype (23), thereby demonstrating that NAbs can be protective against HIV-1 infection (37, 47). A major goal of HIV vaccine efforts is the design of Env-based immunogens that are effective at eliciting broad, heterologous NAbs (7).

Longitudinal studies of HIV-infected patients and simian immunodeficiency virus (SIV)-infected macaques have demonstrated that the immune system gradually recognizes viral variants that emerge over time (20, 32, 41, 58). During the course of infection, antibodies directed to Env undergo immunological maturation, increasing in avidity, conformation dependence, and neutralizing capacity (11). The diversification of the env gene in the viral quasispecies may drive Env-specific antibody maturation both by presenting new epitopes in escape variants and by focusing the response on more conserved epitopes, such as the conformation-dependent regions of Env involved in CD4 and chemokine receptor binding. Mutations associated with changes in susceptibility to autologous NAbs are located in regions of Env that are exposed and may be accessible to antibodies (33). NAbs target these relatively exposed regions of Env, as shown by the isolation of human MAbs that target these regions from HIV-infected subjects (46). Escape from autologous NAbs (41, 58) is due to alterations to Env characterized by entropic masking (27), flexibility in size and the positioning of the variable loops (10, 16), amino acid sequence variation (25), and glycosylation changes (8, 58). Indeed, during the course of infection, the location of potential N-glycosylation sites (PNG) is altered (3) and the number of PNGs is increased (44). Recent studies (36, 43) demonstrated that multiple pathways are involved in escape from autologous NAbs in clade C-infected patients and the pathways are context dependent, as they vary from patient to patient and during the course of infection. These pathways include the evolution of the V3 to V5 region of env, the cooperation of V1/V2 with the ectodomain of gp41 (43), or the evolution of the V1/V2 and C3 regions (36). Other proposed escape mechanisms include the expression of a nonfunctional decoy Env (13, 35, 39). We recently showed that in viremic controllers there is continuous viral escape and selection by autologous NAbs (33). Therefore, the resulting NAbs appear to exert selective pressure on the viral quasispecies, resulting in the fixation of env mutations that alter Env charge, shape, or epitope exposure, in turn resulting in a dynamically changing B-cell response.

A number of approaches have been attempted to design Env immunogens capable of eliciting broad, heterologous NAbs (reviewed in reference 22). These designs include inactivated viruses, monomeric secreted Env, stabilized Env trimers, the stabilization of Env intermediate fusion states, structural analogs of conserved Env epitopes grafted onto scaffolds, and polyvalent or consensus/ancestral Env sequences. To date, only low levels of NAbs have been detected in vaccine studies using these immunogens, with antibodies typically neutralizing only a subset of easier-to-neutralize tier 1 viruses. Previous studies showed that trimeric gp140 is more efficient at inducing an immune response than monomeric gp120 (2, 5, 54, 61), but only marginally so. Because NAbs target native Env trimers on the surface of virions, it may be necessary to recapitulate native Env conformation in vaccines. One such strategy is the use of DNA vaccines based on expression plasmids injected intramuscularly or intradermally. The antigen of interest then is made in vivo with the concomitant development of both humoral and cellular immunity directed to the full-length native trimer. Several studies have shown that weak autologous and heterologous NAbs can be elicited by a combination of DNA prime/protein boost immunization (9, 29, 34, 51, 56), thereby suggesting that structures in the native Env are important for eliciting NAbs (50). Our prior studies exploring the use of ancestral DNA vaccines delivered intradermally by a Gene Gun showed that binding antibodies (BAbs) were elicited in a DNA dose-dependent manner (19) and that DNA vaccination followed by a boost with monomeric gp120 protein elicited weak NAb responses (17) that were poorly cross-reactive. We sought to improve upon these results by exploring a novel idea for env-based vaccines.

Understanding how the virus can elicit a broad NAb response may provide clues as to how to generate these responses during vaccination. We hypothesized that immunization with a collection of env quasispecies variants would result in the development of broader NAbs than immunization with any single env gene alone. To test this hypothesis, we utilized a collection of env quasispecies variants that evolved sequentially in a macaque that developed a moderate cross-NAb response following infection with SHIV_SF162P4. This unvaccinated animal (A141) developed autologous NAbs quickly and showed evidence of gradual broadening, with its serum neutralizing eight heterologous variants, including four standard subtype B viruses, by the end of the study at day 670 (26). The major goal of the present study was to test the quasispecies vaccine concept and to investigate whether exposing rabbits to the same quasispecies in sequential order as it developed over time in A141 could elicit the same moderate breadth of response by immunization as that seen in A141. We compared antibody responses of rabbits exposed to the sequential evolution of env clones (sequential group) to those of rabbits receiving a cocktail of env clones present in the quasispecies (mixture group) or repeated exposure to the same env sequence (clonal group). Clonal and sequential vaccination resulted in greater antibody maturation than the mixture approach. We found that the three vaccine strategies led to the development of comparable levels of binding and neutralizing antibodies. A slight enhancement of heterologous neutralization against HIV-1 SS1196.1 was observed in the sequential group. In addition, NAbs elicited by the mixture and sequential strategies targeted the same epitope on Env as A141. Taken together, our data indicate that the sequential vaccine strategy best replicated the features of the NAb response observed in macaque A141.

MATERIALS AND METHODS

Animals.

Female New Zealand White rabbits were housed at R&R Research, Marysville, WA. Immunizations were performed on rabbits of 6 pounds of weight or greater. All procedures followed IACUC-approved protocols by the Seattle Biomedical Research Institute (Seattle, WA) and R&R Research (Marysville, WA). A rhesus macaque of Indian origin (A141) was part of previously described studies (9, 26), and only serum samples from this animal were used in the current study.

Site-directed mutagenesis.

The mutations of interest in A141 were introduced in the codon-optimized gp160 plasmid expression vector (Novartis, Emeryville, CA) by site-directed mutagenesis with the QuikChange multisite-directed mutagenesis kit (Stratagene, La Jolla, CA). Two hundred twenty mutations (labeled according to HxB2 Env numbering) were introduced to generate the 15 plasmid constructs of interest and are listed in Fig. S1 in the supplemental material. The PCR conditions were 1 min at 95°C; 30 cycles of 1 min at 95°C, 1 min at 55°C, and 12 min at 65°C; and a hold at 4°C. The parental DNA then was digested with DpnI, and XL-10 Gold cells were transformed with the mutated DNA. Bacteria were plated on LB plus kanamycin and grown overnight at 30°C. Individual colonies were picked and grown in small liquid culture (LB plus kanamycin) overnight at 30°C. The plasmid DNA clones were isolated by miniprep (Qiagen, Valencia, CA) and screened by double enzymatic digest with EcoRI and XmnI (New England BioLabs, Ipswich, MA). The presence of the mutation of interest was verified by DNA sequencing using Prism dye terminator kits (ABI, Foster City, CA) on an Applied Biosystems 3730XL genetic analyzer. Nucleotide alignments were manually edited using BioEdit 5.0.9, and sequences were analyzed with Vector NTI (Invitrogen, Carlsbad, CA) or Sequencher (Gene Codes Corporation, Ann Arbor, MI) software programs.

Gold bullets.

DNA was precipitated onto 1-μm-diameter gold beads, and bullets were prepared according to the manufacturer's instructions (Bio-Rad, Hercules, CA). Each bullet was loaded with a total of 2 μg DNA of interest (clonal, sequential, or mixture). To verify that the bullets were functional, COS-7 cells were transfected with the DNA carried by the gold beads. Briefly, COS-7 cells were seeded in two-chamber culture slides (Falcon, Franklin Lakes, NJ) at 3 × 10⁵ cells/chamber in medium (DMEM [Dulbecco's modified Eagle's medium], 10% fetal calf serum, 1% l-glutamine, 1% penicillin-streptomycin) and transfected the next day by a Gene Gun (Bio-Rad, Hercules, CA) at a pressure of 125 lb/in². Cells were incubated at 37°C for 2 days and were fixed and labeled with 0.5 μg/ml monoclonal antibody IgG1 b12 (gift of D. Burton) and a 1:50 dilution of fluorescein isothiocyanate (FITC)-conjugated goat anti-human IgG (Zymed, Invitrogen, Carlsbad, CA). The presence of envelope-transfected cells was revealed by FITC staining visualized by fluorescence microscopy.

Recombinant gp140 proteins.

Wild-type recombinant gp140 from HIV-1 SF162 was prepared as described elsewhere (48). A8-9 late protein from serum of macaque A141 at day 670 postinfection (designated d670 A8-9) was derived from the gp160 expression plasmid by site-directed mutagenesis to insert the following mutations (labeled according to HxB2 Env numbering): E509S and R511S (primary protease cleavage site) and K502I, R503S, and R504S (secondary cleavage site). The EcoRI-XmnI fragment then was introduced into the gp140 vector (pCMVlink140.SF162mod; Novartis, Emeryville, CA) by replacement cloning. DH10B cells (Invitrogen, Carlsbad, CA) were transformed with the plasmid DNA, and a large-scale endotoxin-free plasmid preparation (Qiagen, Valencia, CA) was used for stable expression in CHO cells for protein production as previously described (48).

Immunizations.

New Zealand White rabbits were immunized intradermally by a Gene Gun (Bio-Rad, Hercules, CA) at a pressure of 400 lb/in². A total of 36 μg DNA was delivered 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). Animals were vaccinated with DNA at weeks 0, 4, 10, 20, 24, and 32. Protein boosts, consisting of 50 μg recombinant gp140 protein mixed with an equal volume of MF-59 adjuvant (Novartis, Emeryville, CA), were delivered intramuscularly by needle injection at weeks 41 and 47. Blood was collected 2 weeks after each immunization, and serum was separated and stored at −20°C until IgG purification.

IgG purification.

IgG from individual rabbit sera was purified by affinity chromatography at weeks 12, 22, 26, 34, 43, and 49. Briefly, individual samples (9 ml) of rabbit sera were mixed with 1 ml 10× Tris-buffered saline (TBS). The samples each were mixed with 10 ml of protein G Sepharose (Pharmacia, Uppsala, Sweden) and rotated end-over-end for 1 h at room temperature in Handee spin columns (Pierce, Rockford, IL). The protein G Sepharose-serum mixture was washed three times in 1× TBS by centrifugation (150 × g, 1 min). IgG was eluted with 3 ml of 0.5 M acetic acid, pH 3, by centrifugation (1,000 × g, 2 min), and the eluate was immediately neutralized with 250 μl of 1 M Tris, pH 9. The elution step was repeated three times, and protein concentrations were determined via Nanodrop (ND-1000; Wilmington, DE). For each sample, IgG fractions with the highest protein concentration were pooled and dialyzed three times at 4°C against 1× phosphate-buffered saline (PBS) in Slide-a-lyzer cassettes with a molecular mass cutoff of 10 kDa (Pierce, Rockford, IL). IgG purity was assessed by the Coomassie staining of 4 to 12% gradient SDS-PAGE gels (Invitrogen, Carlsbad, CA) and was found to be equal to or greater than 90%. In addition, a pool of prebleed IgG was prepared, in a manner similar to that described above, and used as a negative control in the enzyme-linked immunosorbent assay (ELISA) and neutralization assays.

gp120-specific ELISA.

The binding antibody response to SF162 recombinant monomeric gp120 (rgp120) envelope protein was measured by kinetic ELISA. Briefly, Immunosorp plates (Nunc, Rochester, NY) were coated with 2 μg/ml SF162 rgp120 in 1× carbonate/bicarbonate buffer overnight at 4°C. Blotto (1× PBS, 1% normal goat serum, 5% bovine serum albumin [BSA]) was used as a blocking step for 1 h at room temperature. IgG samples were diluted in 1× PBS plus 0.1% Triton X-100 and incubated for 1 h in the wells at room temperature. The bound antibodies were detected with horseradish peroxidase-conjugated recombinant protein G (Zymed, Invitrogen, Carlsbad, CA) at a dilution of 1:5,000 in blotto and revealed by the addition of TMB substrate (Sigma, St. Louis, MO). The plates were read at 650 nm in a SpectraMax190 plate reader every 30 s for 30 min (Molecular Devices, Sunnyvale, CA). Values were standardized to a positive-control rabbit IgG sample from the Doria-Rose et al. study (17), which was included with every assay.

Pseudoviruses.

Pseudoviruses were produced using the pSG3ΔEnv DNA plasmid encoding the HIV backbone and a plasmid encoding either homologous (SF162, d56 B3, d417 D15, d487 A6-1, and d670 A8-9) or heterologous (SS1196.1, TRO11, QH0692.42, REJO.67, 7165, 89.6, 6535.3, and 6101.1) envelope variants as previously described (33). Briefly, 293T cells were plated at 5 × 10⁵ cells/well in six-well plates in medium (DMEM, 10% fetal calf serum, 1% l-glutamine, 1% penicillin-streptomycin). The following day, 293T cells in DMEM were transfected with 1 μg total DNA/well at a 20/1 ratio of backbone-to-Env plasmids in the presence of 10.5 μg polyethylenimine (PEI; Polysciences, Inc., Warrington, PA) or 3 μl Fugene 6 (Roche Applied Science, Indianapolis, IN) per well. Cells were incubated for 48 h at 37°C. Virus was harvested by the centrifugation of the supernatant at 2,000 rpm for 10 min at 4°C. The pseudovirus stock was aliquoted and stored at −80°C until use. The pseudovirus was titrated on TZM-bl cells; dilutions of the pseudovirus were used to infect TZM-bl cells in the presence of 7.5 μg/ml DEAE-dextran (Fisher Biotech, Fair Lawn, NJ). Two days later, infection was measured by luciferase expression, and 200 50% tissue culture infective doses (TCID₅₀) (determined by the Reed-Muench formula) were added to each neutralization assay.

Neutralization assay.

IgG purified from the following time points from each animal were tested: 12, 22, 26, 34, 43, and 49 weeks. Serum samples from week 49 also were assessed for heterologous neutralization. Pooled purified rabbit IgG from a previous immunization study (17) was included as a control. The TZM-bl neutralization assay was performed in flat-bottom 96-well plates as previously described (57). Briefly, 200 TCID₅₀ of virus was added to serum samples or serial dilutions of rabbit IgG and incubated in a total volume of 150 μl medium (DMEM, 10% fetal calf serum, 1% l-glutamine, 1% penicillin-streptomycin) for 1 h at 37°C. Each well then received 100 μl of TZM-bl cells resuspended in medium at 1 × 10⁵ cells/ml in the presence of 7.5 μg/ml DEAE-dextran. Forty-eight hours later, 150 μl of medium was removed from each well, and cells were lysed for 2 min directly in the neutralization plate, using 100 μl of Bright-Glo luciferase assay substrate (Promega, Madison, WI), and immediately analyzed for luciferase activity with a luminometer (Victor; PerkinElmer, Waltham, MA). The IgG concentration necessary to achieve 50% neutralization was reported as the 50% inhibitory concentration (IC₅₀), and an increase in neutralization was measured by a decrease in the IC₅₀. As a negative control, a pool of prebleed IgG was used, which never neutralized the tested pseudoviruses (data not shown). All values were calculated with respect to virus-only wells with the following formula: [(value for virus only minus value for cells only) minus (value for serum minus value for cells only)] divided by (value for virus minus value for cells only).

Peptide competition neutralization assay.

Serum from macaque A141 at day 670 postinfection was tested for the neutralization of SF162 in the presence of linear peptides derived from each of the variable regions of Env or in the presence of a control peptide (V3 scrambled) as previously described (15). IgG samples purified from the week 49 time point from each rabbit were tested in the presence of the V1 peptide (TNLKNATNTKSSNWKEMDRGEIK) or the V3 peptide (PNNNTRKSITIGPGRAFYATGD) from SF162 (Invitrogen, Carlsbad, CA) or the V3 scrambled peptide (PNNNTRKSIFYRGAPGITATGD) (Genscript, Piscataway, NJ). Briefly, serial dilutions of rabbit IgG (30 μl) were incubated with 10 μg/ml of the peptide of interest (30 μl) for 1 h at 37°C before the addition of 200 TCID₅₀ of virus in a total volume of 90 μl medium (DMEM, 10% fetal calf serum, 1% l-glutamine, 1% penicillin-streptomycin) for an additional 1 h at 37°C. Each well then received 100 μl of TZM-bl cells resuspended in medium at 1 × 10⁵ cells/ml in the presence of 7.5 μg/ml DEAE-dextran. Forty-eight hours later, 90 μl of medium was removed from each well and cells were lysed for 2 min directly in the neutralization plate, using 100 μl of Bright-Glo luciferase assay substrate (Promega, Madison, WI), and immediately analyzed for luciferase activity with a luminometer (Victor, PerkinElmer, Waltham, MA). The IgG concentration necessary to achieve 50% neutralization was determined, and data were reported as the fold increase in the IC₅₀ above the neutralization measured without peptide (which was set at 1). All values were calculated with respect to virus only: [(value for virus only minus value for cells only) minus (value for serum minus value for cells only)] divided by (value for virus minus value for cells only).

SPR assay.

All surface plasmon resonance (SPR) experiments were performed at 25°C on a Biacore T200 (GE Healthcare, Piscataway, NJ) using a CM4 sensor chip and HBS-EP+ buffer (10 mM HEPES, 150 mM NaCl, 1 mM EDTA, 0.05% P20). Protein A (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 950 resonance units (RU). IgG samples from four of the highest responders in each group were tested at weeks 12 and 43. The antibodies were captured onto the protein A surface on flow cells 2 to 4, leaving flow cell 1 as a reference. The antibodies were noncovalently immobilized to a level of 1,500 (week 12 IgG), 1,000 (week 43 IgG), and 80 RU (positive-control MAb VRC01) by being flowed for 20 to 30 s at a concentration of 15 μg/ml. This resulted in an R_max of 7 to 10 for the week 12 samples and 30 to 50 for the week 43 samples, because only a fraction of the total IgG was antigen specific. Monomeric SF162 gp120 was injected over the sensor surface at the following concentrations: 0, 0.41, 1.23, 3.7, 11.11, 33.33, and 100 nM. All concentrations had a 6-min association period and a 2-min dissociation period, except for the 33.33 and 100 nM concentrations, which had a 1-h dissociation (24). To ensure the accurate quantitation of drift and assay reproducibility, replicate blank injections for 1 h were used and gp120 at 11.11 nM was tested twice, respectively. In all experiments, the binding of gp120 to the reference surface was negligible, and the binding to the reference surface did not change with concentration. Samples were maintained at 15°C before injection. The regeneration of the capture complex was achieved using a 30-s pulse of 10 mM glycine, pH 1.75, 0.05% P20. The data were analyzed using T200 evaluation software. All data were double reference subtracted using the averages from two blank injections. Data were fit to a 1:1 model without any bulk refractive index term despite the inherent multiphasic shape of some sensorgrams due to the immobilization of a polyvalent ligand. Therefore, the measured off-rate is an apparent k_d (off-rate; s⁻¹) because the ligand is polyclonal IgG.

Statistical analyses.

Statistical analysis was done using SAS statistical software, version 9.2. Analyses of BAb levels, SF162 and d56 B3 neutralization from samples drawn in weeks 12 to 49, and k_d rates were determined using mixed-effect models to account for the serial correlation of the repeated measures obtained from each animal. These analyses used log-transformed BAb levels, log-transformed SF162 and d56 B3 IC₅₀ levels, and log-transformed k_d rates as their outcome variables. Due to the presence of floor effects (no observed neutralization, which resulted in the substitution of the maximum tested concentration for a measured IC₅₀), the log-transformed IC₅₀ for d417 D15 and SS1196.1 were not normally distributed. These outcomes were modeled as ordinal variables using a generalized estimating equation with a multinomial link function, with the correlation between repeated measures taken into account in the analysis. These ordinal regression analyses modeled the association of immunization group with the probability of having a lower IC₅₀ (higher neutralization). The targeting of peptides V1 and V3 was analyzed through a mixed-effect model of the log-transformed SF162 IC₅₀s that treated the assessments of the four different preincubations as repeated measures for each animal, specifying an unstructured correlation matrix for the repeated measures to allow for nonuniform variances. The association of the immunization strategy with the neutralization of the heterologous variants at week 49 was analyzed using Welch's analysis of variance (ANOVA) of the log-transformed IC₅₀. Neutralization breadth scores were computed as described by Piantadosi et al. (38) and are the sum of the number of viruses (SF162, d56 B3, d417D15, d487 A6-1, and d670 A8-9) for which the neutralization IC₅₀ was below the median for that antigen. A single median was computed for each antigen using all measurements from all animals and time points. The association-of-breadth scores with immunization strategy and time point was analyzed using ordinal regression as described above.

Nucleotide sequence accession numbers.

All A141 sequences included in this study have been deposited in GenBank (accession numbers JF419573 to JF419587).

RESULTS

Selection of clones and generation of env variants.

Fifteen variants among 41 env clones that represented the major Envelope variants present during the development of the heterologous NAb response in macaque A141 (26) were selected using the following inclusion criteria: (i) dominant sequence at each time point; (ii) multiple unique sequences per time point; (iii) more variants at time points with more viral sequence diversity; (iv) all mutations that became fixed over time; (v) key changes in PNG sites; and (vi) the sequence at the last time point that was the most divergent from the inoculum. These clones (Fig. 1) include the inoculum at day 0 (d0; SF162), two variants at day 56 (d56 A4-5 and d56 B3), three variants at day 215 (d215 A5-1, d215 A5-6, and d215 A5-C7), three variants at day 417 (d417 D8, d417 D15, and d417 D16), three variants at day 487 (d487 A6-1, d487 A6-2, and d487 A6-6), and four variants at day 670 (d670 A8-2, d670 A8-9, d670 A8-11, and d670 A8-15). Each clone has between 4 and 27 amino acid (aa) changes compared to the sequence of the inoculum and contains between 20 and 25 PNGs (Table 1). As seen in Fig. 1, nonsilent mutations accumulated through the quasispecies and accrued in the variable regions as well as in the constant regions of envelope, such as V1/V2, C3, and gp41. Site-directed mutagenesis was used to reconstruct each mutation that was found in the natural variants arising over time in macaque A141 at the five different time points. The selected Envs were expressed in codon-optimized gp160 DNA plasmid vectors. The expression of Env was assessed by the transfection of 293T cells, and all 15 Envs were expressed at the cell surface as measured by IgG b12 binding in an immunofluorescence assay (data not shown), and most of them were able to form syncytia and entry-competent pseudovirions (data not shown).

Fig. 1.

Phylogenetic and Highlighter analyses of A141 envelope sequences included in the quasispecies vaccine. The DNA sequences were aligned using the Sequencher program. The neighbor-joining phylogenetic tree and the Highlighter analysis were generated using the HIV Sequence Database from the Los Alamos National Laboratory (http://www.hiv.lanl.gov/content/sequence/HIGHLIGHT/highlighter.html). In the Highlighter plot, nucleotide differences from the SF162 inoculum/master sequence are indicated by tic marks (red, nonsilent; green, silent; gray, gaps due to alignment with SF162 of partial env coding sequences obtained from A141). The variable loops V1/V2, V3, V4, and V5 are shaded yellow.

Table 1.

Potential N-linked glycosylation sites in the 15 chosen Envelope clones^a

^aAll protein sequences were aligned to HXB2 (accession no. AF033819) and were analyzed with the N-GlycoSite program using the HIV Sequence Database from the Los Alamos National Laboratory (http://www.hiv.lanl.gov/content/sequence/GLYCOSITE/glycosite.html). Green boxes indicate the addition of a PNG compared to the number of PNGs found in HIV_SF162, and purple boxes indicate the removal of a PNG. 1, Presence of PNG at a given amino acid position; 0, absence of PNG at a given amino acid position.

Envelope-specific binding antibody responses are elicited by the three vaccine strategies.

Female New Zealand White rabbits were immunized intradermally by a Gene Gun with 36 μg of plasmid DNA/rabbit/time point during a period of 32 weeks to approximate, on an accelerated scale, proportional exposure to the quasispecies in A141 (Fig. 2). Because all sequences were derived from an individual animal infected with SHIV_SF162P4, all rabbits were immunized with env from the viral inoculum at the first vaccination time point. For the following five time points the rabbits received the clonal, sequential, or mixture vaccine strategy. In the clonal group, rabbits received the SF162 env used in the original viral inoculum. The mixture group was immunized with a cocktail of all 15 quasispecies env sequences given at each of the five remaining time points. Rabbits in the sequential group received an ordered immunization with five cocktails of env sequences derived from each of the five different time points to recapitulate the changes in the viral quasispecies over time. The clonal group was boosted with the trimeric clonal protein SF162 gp140, and both the sequential and the mixture group were boosted with the late variant trimeric protein d670 A8-9 gp140. This late variant was chosen as a representative Env for the d670 clones because it has 25 PNGs (mean, 24.5 PNGs for d670 variants) and 24 amino acid changes compared to the sequence of SF162 (mean, 26.75 aa changes for d670 variants) (Table 1). This boosting strategy was used to maximize potential neutralizing antibody responses in each group. The two gp140 protein boosts (50 μg/animal) were delivered at weeks 41 and 47 in the presence of MF-59 adjuvant.

Fig. 2.

Immunization strategies. Rabbits were vaccinated with six DNA primes at weeks 0, 4, 10, 20, 24, and 32, followed by two protein boosts in the presence of MF-59 at weeks 41 and 47. All three groups were immunized with SF162 env at week 0. The clonal group then was given SF162 env five additional times and was boosted twice with SF162 gp140 protein. The mixture group was immunized with a cocktail of all 15 quasispecies env sequences given five times. The sequential strategy was an ordered immunization with five cocktails of env sequences derived from five different time points to recapitulate the changes in the viral quasispecies over time. Both the sequential and the mixture groups were boosted with d670 A8-9 gp140 protein.

The binding antibody (BAb) responses in rabbits were assessed by measuring SF162 gp120-specific binding antibodies using a kinetic ELISA (Fig. 3). IgG samples purified by affinity chromatography were tested after the third DNA prime, subsequent DNA primes, and protein boosts. There was similar variability among the BAb titers in each vaccine group, with some animals generating higher titers than others. Overall, BAb titers increased during the DNA primes in the sequential group (P = 0.0015) and in the clonal group (P < 0.001) but not in the mixture group (P = 0.57). BAb titers were significantly increased in all groups by the protein boosts (P < 0.001) regardless of the boosting agent used. The second protein boost did not further increase these titers, which decreased slightly (clonal, P = 0.002; mixture, P = 0.004; sequential, P = 0.035). Overall, BAb titers were not statistically different between the three vaccine strategies (P > 0.05).

Fig. 3.

SF162 Envelope-specific binding and neutralizing antibodies elicited by the three vaccine strategies. The binding antibodies (BAb) against SF162 gp120 protein were detected following three to six DNA primes and two protein boosts in each vaccine group (clonal, mixture, and sequential). (A) Antibody titers (μg/ml) were measured by kinetic ELISA. (B) Serially diluted IgG samples from the three vaccine groups were tested for the neutralization of SF162 pseudovirus in a TZM-bl cell assay. The IgG concentration (μg/ml) required to reach 50% neutralization at each time point is reported as the IC₅₀ (μg/ml).

Efficient neutralization of clonal HIV SF162 is achieved by all three vaccine approaches.

Purified IgG samples were tested for the neutralization of HIV SF162, which was neutralized by A141 serum samples at day 670 postinfection (26). Similarly to the BAb data, neutralization was variable among individual rabbits, with weak responders and strong responders in each vaccine group (Fig. 3). All IgG samples except 7207 in the mixture group neutralized SF162 efficiently after the DNA primes. In the clonal group, SF162 neutralization improved moderately during the DNA primes (P = 0.003), and it was significantly higher after the last DNA prime than at the earlier time points (P < 0.004), as measured by a decrease in IC₅₀. Neutralization was enhanced by protein boosting in all three groups (clonal, P = 0.0005; mixture, P = 0.0027; sequential; P = 0.0002), with an IC₅₀ of less than 10 μg/ml achieved by five or more animals in each group regardless of the protein used for boosting. The second protein boost significantly reduced SF162 neutralization in the clonal group (P < 0.004) but not in the other two groups. Overall, there was no statistically significant difference in the neutralization of SF162 between the three vaccine regimens (P > 0.05). However, there was a strong correlation between binding antibody and neutralizing antibody levels against SF162 (P < 0.0001), and this correlation does not differ between groups (P = 0.3).

Variable levels of autologous neutralization by all three vaccine groups.

To investigate the autologous antibody response generated by the three vaccine strategies, we measured the neutralization of an early Env variant, d56 B3, two middle variants, d417 D15 and d487 A6-1, and a late variant, d670 A8-9 (Fig. 4), by constructing pseudoviruses with each of these env genes. The neutralization pattern of d56 B3 was very similar to that obtained with SF162. For some rabbits, such as 7222, 7180, and 7203, the neutralization of d56 B3 was more effective than that of SF162, and we suggest that these results are due to the sequence similarities between the two viruses (only a 5-aa difference). As seen with the neutralization of SF162, we observed both weak and strong responders in each vaccine group. In addition, statistical analysis demonstrated that there was a strong positive correlation (P < 0.0008) at every time point between d56 B3 and SF162 neutralizations. For all groups the neutralization of d56 B3 was effective during the DNA primes and was significantly enhanced by the protein boosts (clonal, P < 0.0001; mixture, P < 0.0001; sequential, P = 0.0007). Four to eight animals in each group reached an IC₅₀ of 10 μg/ml or less after the protein boosts, but the second protein boost decreased the neutralization of d56 B3 in all groups (P < 0.0001). There was no statistically significant difference between vaccine strategies (P > 0.05). The neutralization of d417 D15 was not as efficient as the neutralization of the clonal SF162 or the early variant d56 B3, as the IC₅₀s were 1 to 2 log₁₀ higher at all time points tested (Fig. 4). However, some similar patterns were observed, including the presence of weak and strong responders in each group, and efficient boosting by the proteins (clonal, P = 0.011; mixture, P = 0.018; sequential, P = 0.013). The reduction of NAbs directed to d417 D15 after the fourth DNA prime were not statistically significant when the groups were analyzed individually. The boosting effect was pronounced for d417 D15, with an at least 1 log₁₀ improvement for most samples. Six to seven rabbits in each group had an IC₅₀ of less than 500 μg/ml after the protein boosts. The neutralization of d417 D15 was not significantly different among the vaccine strategies (P > 0.05). The neutralization of d487 A6-1 and d670 A8-9 was weak in all vaccine groups (Fig. 4). The sequential group appeared to neutralize d487 A6-1 better than the clonal or the mixture group; however, there was no statistical difference between groups regarding the neutralization of these last two variants. For the autologous viruses d56 B3, d417 D15, d487 A6-1, and d670 A8-9, neutralization breadth scores were generated (38). Protein boosts significantly increased the breadth scores in all vaccine groups (clonal, P = 0.012; mixture, P = 0.033; sequential, P = 0.019), and this increase was significantly larger in the clonal group than in the mixture group (P = 0.034). There was no significant difference in autologous breadth scores between vaccine groups, but autologous neutralization breadth was associated with SF162 BAbs (P < 0.03) and SF162 NAbs (P < 0.005).

Fig. 4.

Elicitation of autologous neutralizing antibodies. Serially diluted IgG samples from the three vaccine groups were tested for the neutralization of the pseudovirus of interest (d56 B3, d417 D15, d487 A6-1, and d670 A8-9) in a TZM-bl cell assay. The IgG concentration (μg/ml) required to reach 50% neutralization at each time point is reported as the IC₅₀ (μg/ml).

Weak heterologous neutralization elicited by all three vaccine strategies.

We assessed the broadening of the NAb response of the three vaccine strategies by measuring the neutralization of heterologous viruses among those neutralized by animal A141 (26), such as SS1196.1, 89.6, TRO.11, REJO4541.67, QH0692.42, 7165.18, 6535.3, and 6101.1. The neutralization of SS1196.1 (Fig. 5) was weak but detectable after the DNA primes for several samples in each group. Protein boosts were efficient at improving the neutralization of SS1196.1 in all groups (P < 0.03). Indeed, an IC₅₀ of 2,000 μg/ml or less after the first protein boost was achieved by 6/8 animals in the sequential group but by only 3/8 animals in the clonal and mixture groups. Out of these responders in the sequential and clonal groups, 67% reached an IC₅₀ of <500 μg/ml, whereas only 33% reached this level in the mixture group. Furthermore, some samples achieved an IC₅₀ of 80 μg/ml or less after the first protein boost (7168 in the sequential group and 7213 in the mixture group). However, the protein boosts did not induce neutralization in rabbits that did not respond to the DNA primes, such as 7186 in the clonal group, 7109 in the sequential group, and 7207 in the mixture group. Statistical analysis demonstrated that the effect of the first protein boost on the neutralization of SS1196.1 was stronger in the sequential group than in the clonal and mixture groups (P = 0.0064). In addition, the neutralization of SS1196.1 was associated with the neutralization of SF162 (P = 0.004), d56 B3 (P = 0.002), and d417 D15 (P = 0.0008) and the autologous breadth score (P = 0.0009). When modeled together, both the strength and the breadth of the neutralization of autologous viruses predicted the neutralization of SS1196.1 (P = 0.02).

Fig. 5.

Weak heterologous neutralizing antibodies against SS1196.1. Serially diluted IgG samples from the three vaccine groups were tested for the neutralization of the heterologous pseudovirus SS1196.1 in a TZM-bl cell assay. The IgG concentration (μg/ml) required to reach 50% neutralization at each time point is reported as the IC₅₀ (μg/ml).

To further our investigation of the broadening of the NAb response, we measured the neutralization of heterologous isolates 89.6, TRO.11, REJO4541.67, QH0692.42, 7165.18, 6535.3, and 6101.1 by serum samples after the last protein boost (data not shown). Serum from macaque A141 neutralized these viruses, but it did so less efficiently than SS1196.1 (26). Little to no neutralization of these specific heterologous viruses was observed in any vaccine group.

Preferential targeting of the V3 loop by macaque A141 and rabbits immunized with sequential and mixture strategies.

Peptide competition neutralization assays were performed to determine which variable regions of Env are targeted by the antibodies generated in animal A141. We performed these experiments with linear peptides derived from the different variable loops of SF162 Env, which is the original inoculum virus for macaque A141. Serum from A141 at day 670 postinfection was tested for the neutralization of SF162, and among the tested epitopes, only V3 was targeted by NAbs in A141 (Fig. 6A), whereas V2, V4, and V5 loops were not (data not shown). However, the V3 peptide did not compete out all of the neutralization activity, thereby indicating that NAbs target other epitopes as well. Given these results, we performed similar experiments for the three vaccine strategies with two peptides derived either from the first variable loop (V1 peptide) or from the third variable loop (V3 peptide) of SF162 Env. There are up to four amino acid changes (out of 23 aa) in the sequence of V1 and up to two amino acid changes (out of 22 aa) in the sequence of the V3 region of the 15 variants compared to the SF162 amino acid sequences used for the linear peptides. These differences may affect antibody affinity toward the peptide of interest. In addition, competition experiments performed with linear peptides may not identify neutralizing antibodies that target conformational epitopes. Despite these limitations, we observed a differential epitope targeting among vaccine regimens when IgG samples purified after the last protein boost were tested against SF162. Antibodies targeting the V3 loop were found in the mixture group (V3 peptide versus V3 scrambled peptide, P = 0.0011) and in the sequential group (V3 peptide versus V3 scrambled peptide, P = 0.0001), whereas antibodies targeting V1 were found in the clonal and mixture groups (V1 peptide versus no peptide, P < 0.0001). In the clonal group, V1 and V3 were equally targeted by NAbs (P > 0.05). In contrast, both the mixture and the sequential strategies preferentially targeted the V3 loop over the V1 loop (mixture, P = 0.0002; sequential, P = 0.0039) (Fig. 6B). Furthermore, the targeting of V3 over V1 is statistically different between the groups that were immunized with the quasispecies and with SF162 (mixture versus clonal, P = 0.0002; sequential versus clonal, P = 0.0027); however, due to the experimental design, this cannot be separated from the effect of the immunization strategy. The inhibition of neutralization by the V3 peptide was not complete (data not shown), implying that NAbs target the V3 loop region as well as other conformational and/or linear epitopes. Taken together, these results suggest that NAbs elicited by the three vaccine strategies differentially targeted exposed epitopes on Env and that the sequential strategy elicits epitope targeting similar to that obtained with A141.

Fig. 6.

Epitope mapping by peptide competition assays. (A) Epitope mapping in A141. A serially diluted d670 serum sample from macaque A141 was preadsorbed with Env-derived peptides and tested for the neutralization of SF162 pseudovirus. V3 scr, V3 scrambled peptide. The data are presented as 50% infectious doses (ID₅₀) in the presence of these peptides. (B) Epitope mapping in immunized rabbits. Serially diluted IgG samples obtained after the last protein boost were preincubated with peptides corresponding to the V1 or V3 region of SF162 Env before use in TZM-bl cell assays against SF162 pseudovirus. The IC₅₀ (μg/ml) was recorded, and the data are presented as fold increase in IC₅₀ in the presence of V1, V3, or V3 scrambled (V3 scr) peptide. The colors identifying each sample are the same as those in Fig. 3, 4, and 5. ‡, P < 0.0001; *, P < 0.005.

Greater antibody maturation to monomer gp120 induced by the clonal and sequential strategies.

Antibody maturation is likely to be important for effective vaccination, as mature antibodies generally have higher affinity and avidity for their antigen. To investigate whether the three different vaccine strategies induced antibody maturation, surface plasmon resonance (SPR) experiments were performed. Since the evolution of the off-rates (k_d) is indicative of the antibody maturation as measured by a decrease in k_d over time (49), we assessed the binding of SF162 gp120 protein to purified IgG of four of the highest responders in each group at an early time point (week 12) and a late time point (week 43). All groups showed a statistically significant decrease in k_d between week 12 and week 43 (clonal, P < 0.0001; mixture, P = 0.0015; sequential, P < 0.0001) (Fig. 7), thereby indicating that all vaccine strategies induced antibody maturation. The mean k_d values at week 12 were not different between groups, but they were significantly higher at week 43 in the mixture group than in the clonal and sequential groups (clonal, P = 0.0249; sequential, P = 0.0417), thus demonstrating that the clonal and sequential approaches induced greater antibody maturation than the mixture group. In addition, the shape of the sensograms was indicative of a homogenous/clonal response in the clonal and in the sequential groups, whereas it was more indicative of a heterogeneous/polyclonal response in the mixture group (see Fig. S4 in the supplemental material). Binding responses (R_max) increased between week 12 and week 43 in all groups, showing that a greater fraction of the purified IgG was antigen specific (see Fig. S4), a finding that is in agreement with the increase in antibody titers as measured by ELISA (Fig. 3A). VRC01 was used as a positive control, and the values (k_a [on-rate], 3.54E+04 M⁻¹ s⁻¹; k_d, 4.28E-05 s⁻¹; and K_D [equilibrium dissociation constant], 1.21E-09 M) are in line with the data published by Wu et al. (60). Taken together, our data show that the clonal and sequential strategies induced greater antibody maturation than the mixture approach.

Fig. 7.

Antibody maturation elicited by all three vaccine strategies. IgG samples from four of the highest responders in each group (clonal [black]: 7202, 7205, 7222, 7227; mixture [blue]: 7123, 7128, 7198, 7213; sequential [red]: 7168, 7173, 7180, 7223) were tested for binding to SF162 gp120 protein at two different time points (weeks 12 and 43) by surface plasmon resonance. The off-rate data are reported as k_d (s⁻¹). *, P < 0.001; **, P = 0.0015; #, P = 0.0417; ##, P = 0.0249.

DISCUSSION

This study is the first, to our knowledge, to utilize clones of native, sequential HIV-1 env variants arising in vivo during infection to elicit NAbs. We explored a new vaccine strategy in which rabbits were sequentially exposed to Env-encoding DNA expression plasmids in an ordered manner designed to recapitulate the changes in the viral quasispecies experienced in vivo over time. Our goal was to determine how closely, if at all, the resulting NAb responses resembled those observed in the SHIV-infected macaque from which the sequences were derived. Fifteen variants representing the env variants from macaque A141 showing a broadened NAb response were used to determine if these sequences could program the B-cell response in a more effective manner than an individual env (clonal strategy) or a cocktail of these same env variants (mixture strategy). Our data demonstrate that this sequential vaccine approach induced the production of binding antibodies, autologous and weak heterologous NAbs. Furthermore, NAbs targeted the V3 loop of Env among other epitopes. This sequential strategy was comparable to or better than vaccination with a single env clone or with a cocktail of env variants. In addition, this strategy partially reelicited a similar NAb response profile and the same epitope targeting as those obtained from macaque A141, from which the env sequences were cloned.

The three immunization strategies induced gp120-specific BAbs in all rabbits, and as previously demonstrated (18, 19), several DNA primes were required to mount a measurable immune response. In addition, BAb titers were improved by the protein boosts, which is consistent with the published literature (51). As predicted, the BAb levels were not statistically different between the three vaccine groups. The Env-specific BAb response also gradually increased in macaque A141 (data not shown) and reached very high levels as a consequence of the high viremia during both the acute and chronic phases of infection (6). We assessed the longitudinal development of the neutralization of several autologous viruses included in the vaccine. We hypothesized that if the sequential immunization strategy were effective at educating B cells, statistical differences in the NAb responses, but not in the BAb responses, would be observed among the three vaccines groups, which is consistent with the results described above. Our data show that both SF162 and the earliest variant, d56 B3, were efficiently neutralized by sera from all rabbits. The neutralization of the middle variant, d417 D15, was far less efficient than the neutralization of SF162 and d56 B3. The immunization strategies did not induce NAbs against the late clones d487 A6-1 and d670 A8-9, which appear to be more difficult to neutralize. Interestingly, when four different subtype A envelopes with differing neutralization sensitivity profiles were used to immunize rabbits (vaccinia viruses with env DNA plus protein boosts), NAbs were elicited only against the most easily neutralized virus and were elicited equally with env variants that had relatively exposed neutralizing determinants compared to levels for those that were more occluded (4). In addition, the sequential group shows that a later variant was specifically recognized by the immune system, albeit quite weakly, despite being introduced in the host after other variants with shared epitopes. These data suggest that the sequential vaccine strategy does not follow the paradigm of original antigenic sin, in which the only recognized epitopes are those shared with the first presented immunogen (1, 40). Similar findings were obtained with the influenza vaccine from healthy human individuals, for whom the highest antibody response targeted the current vaccine strain and not previous ones (59).

To assess the broadening of the neutralizing antibody response, neutralization specificities obtained after immunization with the three vaccine strategies were compared with those developed in the SHIV-infected animal. The neutralization of SS1196.1 was similar in all groups at all time points except after the first protein boost, where it was more efficient in the sequential group than the clonal and mixture groups, suggesting that the sequential vaccine strategy induces a broadening of the antibody response. However, we observed no neutralization of the other heterologous variants by any of the vaccine groups. Thus, the neutralization of one heterologous isolate did not predict the breadth of the heterologous neutralizing antibody response. These findings are consistent with the data obtained from macaque A141, from which the env variants were cloned (26), with better neutralization of the isolates that were easier to neutralize; therefore, our sequential vaccine strategy replicated some of the features of the NAb response obtained from an SHIV-infected animal with a broadened response. In contrast, in a rabbit immunization study with env derived from clade A-infected patients who showed broad NAbs, the vaccines did not induce any heterologous NAbs and were not able to reproduce the breadth measured in the individuals from whom the env variants were obtained (4). In addition, Derby et al. demonstrated that a concentration of 10 mg/ml of purified IgG collected after the immunization of macaques with SF162 DNA prime/protein boost was necessary to neutralize SF162 potently and 89.6, 6535, and SS1196.1 moderately, and it showed some hint of the neutralization of REJO4541.67 (15). Therefore, our results are consistent with other immunization studies.

Differences in neutralization sensitivity in some HIV Env proteins have been correlated to changes in variable loop lengths (21, 44) or variation in PNG numbers (44). The tested heterologous viruses had the same V3 length as the A141 Env proteins. Neutralization resistance has been shown to be associated with V1/V2 length (21, 44), and NAbs have correlations with shorter V1-V4 lengths in clades other than B; thus, these effects do not explain our observations. Previous studies demonstrated that neutralization resistance correlated with an increase in the number of PNGs (44). We found that both the late autologous variants (d417 D15, d487 A6-1, and d670 A8-9) and the heterologous variants had increased numbers of PNGs in gp120 compared to that of the parental SF162 (Table 2). Indeed, the autologous variants had more PNGs in V1/V2 and V4 regions, and this increase in PNG sites could explain why they are more difficult to neutralize. Some heterologous variants had more PNGs in C1, V1/V2, C2, C3, V4, and V5 regions. Interestingly, SS1196.1 and REJO4541.67 had fewer PNGs than TRO.11 and 7165.18, but only SS1196.1 is neutralization sensitive; thus, variation in PNG number alone cannot explain the measured variation in neutralization resistance as tested in our study. The influence of specific PNG sites and amino acid residues on neutralization sensitivity in the A141 quasispecies is being investigated further.

Table 2.

PNGs in autologous and heterologous variants^a

^aEach row represents a viral Env variant. All protein sequences were aligned to HXB2 (accession no. AF033819) and were analyzed with the N-GlycoSite program using the HIV Sequence Database from the Los Alamos National Laboratory (http://www.hiv.lanl.gov/content/sequence/GLYCOSITE/glycosite.html). An increase in the number of PNGs compared to that of SF162 is indicated by a green box, and a decrease is indicated by a purple box.

Neutralization competition assays were performed with linear V1 and V3 SF162 peptides to determine which variable regions of Env are involved in the neutralizing activity of the polyclonal antibodies generated by our vaccine strategies. The targeting of V1 and V3 loops was expected, since both loops are known to be immunogenic (14, 15). Interestingly, the mixture and sequential strategies preferentially targeted the V3 loop over the V1 loop; however, the clonal strategy did not significantly target either of these loops. Rabbits immunized by the mixture or sequential strategy were exposed to 15 different env variants, but rabbits in the clonal group were exposed only to SF162 env. Therefore, we propose that exposure to this quasispecies focuses the immune response toward V3. Furthermore, in serum from macaque A141, NAbs targeted V3, among other epitopes, but not V1. It was recently demonstrated that MAbs derived from animal A141 neutralized SF162, and one of these MAbs targeted a linear epitope in V3 (42). Several additional MAbs targeted quaternary epitopes involving V2 and V3, but none of them targeted the V1 loop (42). In addition, anti-V3 antibodies are generated in most HIV-1-infected patients, and PG9, PG16, and HGN194, three very potent MAbs recently isolated, target epitopes located in V2 and V3 (12, 55). Interestingly, Li et al. showed that SS1196.1 also was susceptible to anti-V3 NAbs (28), and SS1196.1 is a heterologous variant neutralized by both macaque A141 and our vaccinated rabbits. Therefore, taken together our data show that the sequential and mixture strategies replicated the epitope targeting observed in an SHIV infection, suggesting that it is due to the Env evolution in the quasispecies.

Finally, it was shown recently that antibody maturation is important to elicit a protective antibody response in infected subjects. In SHIV-infected macaques, an increase in antibody avidity was inversely correlated with peak viremia (62). In addition, the broad and potent neutralizing activity of MAb VRC01, which was isolated from an HIV-1-infected subject, was the result of extensive antibody maturation (63). Surface plasmon resonance experiments (using Env-SF162 monomer binding to captured IgG on the chip) showed that our three vaccine strategies induced antibody maturation. The data obtained with the clonal group are not surprising, since the rabbits in this group were exposed only to SF162 DNA and protein. Importantly, significantly greater antibody maturation was achieved by the clonal and sequential strategies than the mixture approach. These data suggest that including multiple variants in a vaccine to broaden the antibody response is a viable approach, and that the variants are best delivered sequentially rather than in a cocktail.

In conclusion, immunization with a collection of env quasispecies variants resulted in the development of autologous and weakly heterologous NAbs that were comparable to the ones elicited by the clonal strategy. In addition, our quasispecies vaccine strategy replicated some of the NAb response features obtained from an SHIV-infected animal (from which the quasispecies env variants were cloned), such as the neutralization of two of the same HIV envelope variants and V3 loop immunofocusing. Therefore, our data suggest that it is possible to program humoral responses in part by exposure to a subset of native HIV-1 quasispecies derived from an individual with a broadened NAb response. This strategy did not result in the induction of broad heterologous NAbs, but this may have been due to the vaccine delivery system used and/or to the fact that sequences were derived from an SHIV-infected macaque after less than 2 years of infection. Naked DNA plasmids have long been considered a weak delivery system to induce protective immune responses (30, 31), but recent studies showed that DNA prime-protein boost approaches can induce neutralizing antibodies in rabbits and in humans (52, 53). Additional experiments are in progress using env variants cloned from the plasma of HIV-infected human subjects who developed more potent NAbs than did macaque A141 to determine the broader utility of this approach, as well as to further explore the use of key individual variants that arise during infection as immunogens. Finally, an ideal HIV Env-based vaccine would generate antibody maturation, specific epitope targeting, and broad neutralization. Our study emphasizes the importance of assessing all of these different aspects of the immune response in immunized animals to determine which immunization approaches are beneficial to the vaccinee.