Biochemically Defined HIV-1 Envelope Glycoprotein Variant Immunogens Display Differential Binding and Neutralizing Specificities to the CD4-binding Site*

Background: Broadly neutralizing antibodies to the gp120 CD4-binding site (CD4bs) validate this region as a target for immunogen design. Results: HIV-1 envelope glycoprotein (Env) variant immunogens display differential biochemical and biophysical properties and elicit differential neutralizing specificities to the CD4bs. Conclusion: Engineered Env proteins may more efficiently direct responses toward the conserved CD4bs. Significance: This study provides parameters to consider associated with the elicitation of broader neutralizing responses. HIV-1 gp120 binds the primary receptor CD4. Recently, a plethora of broadly neutralizing antibodies to the gp120 CD4-binding site (CD4bs) validated this region as a target for immunogen design. Here, we asked if modified HIV-1 envelope glycoproteins (Env) designed to increase CD4 recognition might improve recognition by CD4bs neutralizing antibodies and more efficiently elicit such reactivities. We also asked if CD4bs stabilization, coupled with altering the Env format (monomer to trimer or cross-clade), might better elicit neutralizing antibodies by focusing the immune response on the functionally conserved CD4bs. We produced monomeric and trimeric Envs stabilized by mutations within the gp120 CD4bs cavity (pocket-filling; PF2) or by appending heterologous trimerization motifs to soluble Env ectodomains (gp120/gp140). Recombinant glycoproteins were purified to relative homogeneity, and ligand binding properties were analyzed by ELISA, surface plasmon resonance, and isothermal titration microcalorimetry. In some formats, the PF2 substitutions increased CD4 affinity, and importantly, PF2-containing proteins were better recognized by the broadly neutralizing CD4bs mAbs, VRC01 and VRC-PG04. Based on this analysis, we immunized selected Env variants into rabbits using heterologous or homologous regimens. Analysis of the sera revealed that homologous inoculation of the PF2-containing, variable region-deleted YU2 gp120 trimers (ΔV123/PF2-GCN4) more rapidly elicited CD4bs-directed neutralizing antibodies compared with other regimens, whereas homologous trimers elicited increased neutralization potency, mapping predominantly to the gp120 third major variable region (V3). These results suggest that some engineered Env proteins may more efficiently direct responses toward the conserved CD4bs and be valuable to elicit antibodies of greater neutralizing capacity.

During the course of natural infection, the HIV-1 Env elicits both type-specific and broadly cross-reactive neutralizing antibodies (23)(24)(25). However, it has been difficult to elicit potent and cross-reactive neutralizing antibody responses by immunization with Env-based vaccine candidates. In part, this difficulty may lie in the immune dominance of gp120 variable regions and the apparent limited immune recognition of conserved regions of the viral Env (23,24). To gain better insights into how the humoral immune system recognizes Env, investigators have studied HIV-1-infected individuals that mount broadly neutralizing antibody responses to conserved regions of the viral Env (26 -32) and especially to the CD4bs, to which several broadly neutralizing mAbs are directed (33,34). Guided by these studies, and extrapolated to the context of Env-based subunit vaccine design, the functionally conserved CD4bs of gp120 presents an attractive target for the elicitation of broadly neutralizing antibodies (35,36). However, attempts to re-elicit broad and potent CD4bs-directed antibodies using monomeric gp120 or trimeric, soluble gp120/gp140 spike mimetics (gp140 refers to gp120 plus the gp41 ectodomain) have met with limited success (37)(38)(39)(40)(41)(42)(43)(44)(45). With that said, several studies have reported the presence of CD4-blocking antibodies or weakly neutralizing CD4bs antibodies in the sera of Env-vaccinated subjects (46 -48). Generally, however, the conserved CD4bs in the context of the functional viral spike appears relatively poorly immunogenic, perhaps due to glycan occlusion, the greater intrinsic immunogenicity of the Env variable elements, or the multiple conformations of the flexible gp120 molecule that may be presented to the polymorphic B cell receptor. In fact, recent studies reveal that the subregion of the CD4bs recognized by the broadly neutralizing mAb, VRC01, lies more distal to the Env trimer central axis than that generally recognized by the non-neutralizing CD4bs mAbs. 4 These results suggest that alterations of the CD4bs, either by conformational fixation or steric occlusion, may be required to shift B cell responses to elicit antibodies that more closely resemble VRC01 recognition rather than the type of CD4bs-directed antibodies that are more commonly elicited by HIV-1 infection or vaccination.
Previously, structure-based approaches were implemented to rationally modify the soluble gp120/gp140 molecules in attempts to improve immunogen design. One of these approaches is to truncate the immunodominant major variable (V) elements, V1, V2, and V3, as well as the immunogenic but non-neutralizing N and C termini to generate monomeric core gp120 immunogens (49 -51) for attempting to shift B cell recognition to more conserved, recessed elements, such as the CD4bs. Another approach entailed stabilizing gp120 into the CD4-bound state by filling the CD4-interactive "Phe-43 cavity" with structure-guided T257S and S375W substitutions (PF2) (46). In a related path of rational immunogen design, internal cysteines were generated in the gp120 core for conformational stability, locking the core into the structurally defined CD4bound state (49). To mimic the trimeric functional spike, a heterologous trimerization motif, such as those derived from GCN4, ATCase, or foldon (F), was appended at the C terminus of gp140 and in one case gp120 (46,(52)(53)(54). Although the PF2 and cysteine-stabilized core immunogens do shift the specificity of the elicited B cell immune response, and the trimeric gp140 immunogens do elicit slightly more potent neutralizing antibodies than do the gp120 monomers, the antibodies that are elicited so far do not display demonstrable neutralization breadth against circulating HIV-1 primary isolates (i.e. so-called Tier 2/3 viruses) (43,44,(55)(56)(57).
In this study, we explored new approaches to attempt to better elicit responses to the CD4bs. We produced and characterized gp120 core protein constructs and gp120 trimers deleted of the immunogenic V1, V2, and V3 regions, each stabilized in the CD4-state with the Phe-43 cavity-filling PF2 mutations. The PF2 substitutions should increase CD4 affinity for this conserved binding site and, along with the V deletions, might focus the humoral immune response toward the conserved CD4bs as was seen for shifts to the co-receptor binding site in our previous study (49). Finally, we biochemically and immunogenically analyzed gp140 trimers, with and without the PF2 mutations, in a side-by-side manner. For the gp140-F trimers, we selected Envs derived from three resistant clade B or C isolates, YU2, CAAN, and ZA012, the first two without and with the PF2 modifications. We present a detailed analysis of the characterization of the biochemical and biophysical properties of the Env variants to define rigorously the products used for the preclinical immunogenicity studies performed here. Such analysis may potentially reveal associations between the biophysical properties of a modified Env and immunogenicity, as shown previously (49). We report that each of these immunogens was well recognized by the broadly neutralizing CD4bs antibodies, VRC01 and VRC-PG04, especially the YU2 gp120⌬V123/PF2-GCN4 trimers. We assessed sequential immunization with different Env formats to investigate two approaches. First, as a heterologous approach, the sequential immunizations beginning with the minimal and monomeric HXBc2 gp120/PF2 cores were followed by the YU2 gp120⌬V123/PF2-GCN4 trimers and then by the full-length gp140-F trimers (Ϫ/ϩPF2) to potentially better focus the response on conserved elements. Second, is the homologous immunization of the trimeric immunogens. Generally, the homologous regimens of the gp140-F trimers better elicited neutralizing activity regarding the potency and breadth against Tier 1 isolates, and this increased potency could be mapped largely to the V3 region. Interestingly, the YU2 gp120⌬V123/PF2-GCN4 trimers more efficiently elicited CD4bs neutralizing antibodies against sensitive isolates, suggesting that the dual modifications of CD4state stabilization and variable-region deletion may be of benefit to better target the CD4bs. The data suggest that immunogens designed to focus the immune response on this functionally conserved region have the potential to improve elicitation of CD4bs-directed neutralizing antibodies but likely still require improved immunogenicity regimens to elicit more potent neutralizing activity.
Expression and Purification of Monomeric and Trimeric Env-All proteins were expressed in serum-free medium by transient transfection of Freestyle 293-F cells (Invitrogen) as described previously (60). In brief, cells were transfected at a density of 1.2 ϫ 10 6 /ml in Freestyle 293 expression media (Invitrogen) using 293Fectin. Supernatants were collected 4 days after transfection, filtered through a 0.22-m filter, and supplemented with Complete TM EDTA-free protease inhibitor mixture (Roche Applied Science). All proteins were purified over affinity columns, and in selected cases, the trimeric fractions were isolated by size exclusion chromatography. As described previously (46), HXBc2 gp120 core/PF2 and YU2 gp120⌬V123/PF2-GCN4 were purified by the b12 or 17b affinity column, respectively. The YU2-F, CAAN-F (Ϫ/ϩPF2), and ZA012 gp140-F trimers were purified with lentil-lectin affinity chromatography followed by immobilized metal ion affinity chromatography. In brief, trimeric glycoproteins were captured by lentil-lectin-Sepharose TM 4B (GE Healthcare). After extensive washing with PBS, the glycoproteins were eluted with 0.5 M methyl ␣-D-mannopyranoside and captured in a second purification step via the His tag by chelating Sepharose TM Fast Flow (GE Healthcare) medium charged with nickel ions. After washing with 40 mM imidazole and 0.5 M NaCl in PBS, proteins were eluted with 300 mM imidazole in PBS. Eluted fractions containing proteins were pooled, concentrated with Amicon ultracentrifugal filter devices (Millipore), and dialyzed extensively against PBS. Affinity-purified trimeric proteins were further subjected to size exclusion chromatography using a Superdex 200 26/60 prep grade column (GE Healthcare) in PBS containing 0.35 M NaCl. The flow rate was set to 1 ml/min to allow the separation of specific oligomeric species. Relevant fractions containing trimers were pooled, concentrated, dialyzed against PBS, flash-frozen in liquid nitrogen, and stored at Ϫ80°C.
Native Gel and FPLC Analysis-The purified proteins were run on blue native gels and fast protein liquid chromatography (FPLC) system (GE Healthcare) to analyze their oligomeric states. For blue native gels, the running buffer in the outer chamber contained 50 mM Tris-HCl plus 50 mM MOPS. The inner chamber contained the same running buffer with 10 mg of SERVA-G (Invitrogen) per 0.5 liters of buffer. Protein samples were diluted 1:1 in 2ϫ sample buffer (100 mM Tris-HCl, 100 mM MOPS, 40% glycerol, 0.1% Serva-Blue G). Novex gel system (Invitrogen) was used to run the gel for 4 h at 4°C at 100 mV. Following electrophoresis, the gel was stained using Coomassie and de-stained by normal procedures used for SDS-PAGE. The proteins, before and after size exclusion chromatography, were also analyzed by FPLC using an AKTA system. 100 g of each protein was loaded onto 10/300 GL Tricorn TM high performance columns pre-packed with Superdex 200 medium and resolved at a flow rate of 0.3 ml/min to separate the different oligomeric glycoprotein species.
ELISA-To determine the antigenic profiles of the Env protein variants (Fig. 3) and the triple mutant core probes (Fig. 5), standard ELISA was conducted as described previously (46). In brief, high protein-binding ELISA plates (NUNC) were coated with 200 ng per well of Galanthus nivalis lectin (Sigma) at 4°C overnight. The wells were blocked for 2 h at room temperature (RT) with PBS containing 5% fat-free milk, 5% fetal calf serum, and 0.2% Tween 20. 200 ng per well of gp120 and gp140 protein variants were added and incubated for 1 h at RT. After four washes, the wells were then incubated for 1 h with selected monoclonal antibodies that were 5-fold serially diluted starting with an initial concentration of 40 g/ml. Following a 1-h incubation with 1:5,000 diluted horseradish peroxidase (HRP)-conjugated anti-human IgG (Jackson ImmunoResearch), the colorimetric HRP substrate (3,3Ј,5,5Ј-tetramethylbenzidine; Bio-Rad) was added into each well, and the reaction was stopped by 1:35 diluted sulfuric acid. Absorbances of each well were analyzed at 450 nm using an ELISA plate reader.
To determine the CD4bs-binding antibodies, ELISA analysis comparing inhibition of VRC01 binding to coreV3S protein using pooled immune sera from individual rabbits from each immunogen group was performed. ELISA plates, coated with 1 g/ml protein, were preincubated with 3-fold dilution of rabbit sera (or ligands) for 45 min at RT, reacted with biotinylated VRC01 (0.15 g/ml final concentration) for 30 min at RT, and detected with 1:250 dilution of HRP-conjugated streptavidin. Serum from BSA/adjuvant-immunized rabbits was used as a negative control.
SPR Kinetic Binding Analysis-All kinetic reactions were performed at RT on a Biacore 3000 SPR spectrometer. For the gp140-F trimers containing the His 6 tag, an indirect capture method was used to determine the kinetic constants of their interaction with CD4 and b12 Fab. To prepare the reference and test cell surface, anti-His 6 mAb (R&D Systems, 25 g/ml in 10 mM acetic acid, pH 5.0) was immobilized on CM5 chip by the amine coupling method following the manufacturer's protocol. The ligand, His 6 -tagged trimers (10 g/ml in HBS-EP buffer), was captured on the chip surface. The analyte, four-domain sCD4 or b12 Fab, was serially diluted in HBS-EP buffer at concentrations ranging from 10 nM to 2.56 M and flowed over the chip. Association was allowed for 6 min at 30 l/min. Dissociation was determined by washing off bound analyte over the next 6 min. In each reaction cycle, the chip surface was regenerated with one injection (30 s) of 10 mM glycine, pH 2.5, at 50 l/min. The HXBc2 gp120 core/PF2 and YU2 gp120⌬V123/ PF2-GCN4 were directly immobilized on a CM5 chip with ϳ500 response units, and their binding with CD4 and b12 Fab was determined as described above. The kinetic rate constants were obtained by fitting the curves to 1:1 Langmuir binding model using BIAevaluation software.
Animal Inoculations-Female New Zealand White rabbits were inoculated with 50 g of affinity-purified protein formulated in Adjuplex TM (Advanced BioAdjuvants, Omaha, NE). Protein/adjuvant was injected intramuscularly by splitting the protein/adjuvant mixture in the two hind legs at 4-week intervals. The prime/boost immunization regimen was as follows: groups of animals, composed of six rabbits per group, were primed two times with HXBc2 gp120 core/PF2, followed by two inoculations with YU2 gp120⌬V123/PF2-GCN4, and all animals finally received two inoculations with CAAN gp140-F, CAAN/PF2 gp140-F (data not shown), YU2 gp140/PF2-F, and ZA012 gp140-F trimers, respectively. Another set of animals was primed twice with CAAN gp140-F, CAAN gp140/PF2-F, YU2 gp140/PF2-F, and ZA012 gp140-foldon trimers in each group and were inoculated two times with YU2 gp120⌬V123/ PF2-GCN4, followed by two inoculations with HXBc2 gp120 core/PF2, but they showed no improvement in neutralization and are not shown. Another five groups of animals composed of four rabbits per group were inoculated six times with the YU2 gp120⌬V123/PF2-GCN4, CAAN gp140-F, YU2 gp140-F, YU2 gp140/PF2-F, and ZA012 gp140-F trimers, respectively ( Fig.  4A). Two additional rabbits were inoculated six times with BSA in adjuvant as controls. Serum was collected 7-8 days after each inoculation. All rabbits were housed and maintained in the AAALAC-accredited Biocon, Inc (Rockville, MD) under specific pathogen-free conditions. All experiments were approved by the Animal Care and Use Committee of the Vaccine Research Center and Biocon, Inc.
Neutralization Assays-The production HIV-1 Env pseudotypes and the TZM-bl neutralization assay were described previously (44,61). In brief, TZM-bl cells expressing huCD4, huCXC4, and huCCR5 were used for HIV-1 infection. A panel of 10 pseudoviruses from clade B and clade C were analyzed. The dilution at which the serum could neutralize 50% of virus entry (ID 50 ) was calculated by fitting the dose-response curve with a nonlinear function (four parameter logistic equation) using GraphPad Prism software (San Diego). For further control of nonspecific neutralization, sera from rabbits immunized with BSA were analyzed, and all sera were tested against a pseudovirus expressing the amphotropic murine leukemia virus envelope (62).
To examine the contribution of CD4bs antibodies to the serum neutralizing activity, a neutralization assay was performed using TriMut or TriMut368/70 as gp120-specific antibody-adsorbing probes. This assay is a modified version of the standard neutralization assay described above. Before addition of pseudovirus, 100 l of 3-fold serial dilutions from each serum were preincubated with 12.5 l of TriMut (100 g/ml), TriMut368/70 (100 g/ml), or cell culture medium, respectively, for 1 h at 37°C. For each serum, three neutralization curves derived from the assays performed in parallel were analyzed to reveal the neutralization directed to the CD4bs by differential inhibition of neutralization using the isogenic probes. To compare the percentage of CD4bs-directed neutralization between animals grouped by regimen, differences of neutralization absorbed by TriMut and TriMut368/70 at 1:12 serum dilution were plotted. Statistical analysis of CD4bs-directed neu-tralization detection was done using the Mann-Whitney nonparametric test (GraphPad Prism 5.0).
Peptide inhibition neutralization assays were done as described previously (44). Briefly, the neutralization assays were done as described above, except that the control or test peptide was added to the serum 30 min prior to the addition of pseudotyped HIV-1 isolates. The effect of the peptide on virus neutralization was reported as the percent inhibition of neutralization: ((% neutralization with scramble peptide Ϫ % neutralization with YU2 V3 peptide)/percent neutralization without peptide) ϫ 100. The YU2 V3 peptide (TRPNNNTR-KSINIGPGRALYTTG) was synthesized by SynPep (Dublin, CA) and the scramble peptide (IGPGRATRPNNNFYTTGTR-KSIH) was purchased from Sigma.

RESULTS
Purification and Binding Analysis of gp120/gp140 Variant Proteins-The gp120 and gp140 Env variants (Fig. 1) were purified over affinity columns and analyzed by gel electrophoresis and FPLC (see "Experimental Procedures" and Fig. 2). The migration of HXBc2 gp120 core/PF2 monomers, YU2 gp120⌬V123/PF2-GCN4 trimers, and the YU2, CAAN, and ZA012 gp140-F trimers (Ϫ/ϩPF2) were assessed under reducing conditions on SDS-polyacrylamide gels (Fig. 2B). Under these conditions, the gp120/PF2 core monomer (lane 1) and YU2 gp120⌬V123/PF2-GCN4F trimers (lane 2) migrated more rapidly through the gel, although the wild-type (WT) and the gp140/PF2-F trimers (lanes 3-6) exhibited similar but slower motilities. The oligomeric status of the cores, the YU2 gp120⌬V123/PF2-GCN4 trimers, and gp140-F trimers was then analyzed by size exclusion chromatography using FPLC (Fig. 2). As reported previously, the cores migrated predominantly as a single peak. The YU2 gp120⌬V123/PF2-GCN4 trimers exhibited multiple oligomeric states consisting predominantly of trimers, which were purified to homogeneity by size exclusion chromatography. The HIV-1 gp140-F Envs were also predominantly trimeric, along with the presence of higher order oligomeric forms, likely dimers of trimers, as well as higher molecular weight aggregates. The trimers were separated from the higher order oligomeric forms by FPLC to isolate the trimer-containing fractions. The purity of the trimers after fractionation was confirmed by blue native gels (Fig. 2D) and by a second round of gel filtration chromatography (Fig. 2C). To limit complications introduced by molecular heterogeneity, the FPLC-purified trimer-fractionated immunogens were utilized for all subsequent biochemical and biophysical analysis.
To characterize the antigenic profile of the Env variants, in particular the recognition of the CD4bs on each variant, we performed ELISA analysis with conformationally dependent ligands specific for CD4bs, including CD4-Ig and the broadly neutralizing mAbs VRC01, VRC-PG04, and b12. Consistent with our previous studies, the PF2 substitutions slightly increased recognition by CD4-Ig in the YU2 gp140-F context (Fig. 3). Interestingly, and perhaps more relevant for vaccine design, the PF2 modifications increased the recognition of the HXBc2 gp120 core proteins (gray curves), the YU2 gp120⌬V123-GCN4 trimers (red curves), and the YU2 (blue curves) and CAAN gp140-F trimers (black curves) by both VRC01 and VRC-PG04 (Fig. 3, upper and middle panels). Note also that the YU2 gp120⌬V123/PF2-GCN4 trimers were the best recognized of all constructs by CD4-Ig, VRC01, and VRC-PG04 (Fig. 3, red curves, solid symbols). All the proteins, except for the two gp140 trimers derived from the b12-resistant CAAN virus (63), were well recognized by the neutralizing FIGURE 1. Schematic to illustrate the different Env formats and inoculation rationale. A, schematic representation of the recombinant gp120 core/PF2, the gp120⌬V123/PF2 trimer, and gp140 trimer illustrating the different Env formats. The recombinant uncleaved gp140 trimer is composed of gp120 (gray), the gp41 ectodomain (brown), the heterologous trimerization motif (gold or green), and His 6 tag. The HXBc2 core protein and the YU2 gp120⌬V123 trimers are color-coded similarly to the gp140 trimer schematic. B, schematics of the experimental rationale for the prime/boosting of core-to-trimer or trimer-to-core which increases or decreases the elements of Env contained in each immunogen, as illustrated Env format alteration. The goal of the experiment is to focus the B cell immune response on the conserved CD4-binding site shared by all immunogens (red). Several of the constructs contained substitutions in the Phe-43 cavity (pocket-filling) that enhances CD4 affinity, with the goal to more efficiently engage B cells with enhanced B cell receptor affinity for the CD4-binding site.
Dashed lines indicate the Env loop and N-and C-terminal truncations. The cross-clade gp140-F trimers for boosting used constructs derived from the clade B YU2 and CAAN strains or the clade C ZA012 strain. CD4bs antibody b12 (Fig. 3). Also, in agreement with our previous studies, the PF2 substitutions, when introduced into either the gp120 core or the gp120/gp140 trimers, eliminated or greatly reduced recognition by the non-neutralizing CD4bs antibody F105 (14,46). This is part of the design strategy to eliminate activation of B cells that would recognize the CD4bs in manner similar to, for example, the non-neutralizing CD4bsdirected mAb F105. We next tested recognition by the co-receptor-binding site (or CD4-induced) mAb 17b. As expected from previous studies (50,64), the HXBc2 gp120 core/PF2 monomers were not well recognized by 17b, although all the other gp120/gp140 trimers were recognized by 17b, even in the absence of CD4 (data not shown). We also assessed the recognition by the non-neutral-izing antibody C11, whose binding specificity is directed against the gp120 N and C termini, to determine exposure of this region on the trimer constructs. C11 did not recognize the HXBc2 gp120 core/PF2 monomer due to deletion of the N and C termini nor did C11 recognize the ZA012 gp140-F trimer. The decreased C11 recognition of the ZA012 trimers may be because this clade C Env possesses an amino acid sequence different from clade B viral Env in the C terminus (residues 497-499, HXBc2 numbering), eliminating C11 recognition, or it may due to increased occlusion of the epitope on these trimers.
Binding Kinetics Determined by Surface Plasmon Resonance (SPR) of Selected Ligands to gp120/gp140 Variants-To determine the accessibility, on-and off-rates, and precise affinity of ligands to the CD4-binding site of each of the gp120/gp140 variants, we performed SPR using sCD4 and the Fab, b12, as representative CD4bs-directed ligand analytes. A representative set of curves is shown in Fig. 3B, with the bulk of the SPR data presented in Table 1 (individual binding curves at different concentration are shown in supplemental Fig. S2). The on-rates (K a ) of sCD4 binding gradually decreased as the Env format was altered from the V region-deficient core to the gp120⌬V123/ PF2-GCN4 trimers, to the gp140-F V region-containing trimers, likely due to occlusion by the V regions present on the gp140 constructs as reported previously (65). The PF2 mutations significantly increased the affinity of CD4 binding to YU2 gp140-F trimer, primarily by reducing the off-rate. We postulated that the increase in CD4 affinity might make the CD4bs on the PF2-containing Env variants easier to engage by naive B cells, a major criterion for including this modification in the proteins selected for immunogenicity analysis. The increase in sCD4 affinity was not observed in the CAAN gp140-F trimeric context by SPR, which already displayed a relatively high affinity. For b12 Fab recognition of the Env constructs, neither the on-rates nor the off-rates were greatly changed for any of the Env V region-containing, V region-lacking, or Ϫ/ϩPF2 formats.
Thermodynamic Analysis of the gp120/gp140 Variants by ITC-To evaluate the impact of V-region deletion in conjunction with the PF2 substitutions in the gp120 trimers, and the impact of the PF2 substitutions on conformational stabilization of the CD4-bound state of all of the Env formats used in this study, we performed ITC analysis with sCD4. The ITC data were also used to assess the stoichiometry of sCD4 binding to each of the Env variants (see below). As expected, proteins possessing the PF2 substitutions showed decreased entropy change in all cases, and the degree of stabilization, when comparing the WT to the PF2 Env versions, differed in magnitude depending upon the protein context. In the HXBc2 gp120 core context, stabilizing effects generated by the PF2 substitutions were minimal, consistent with data reported previously (46). In the gp120⌬V123-GCN4 trimeric context, the PF2 substitutions reduced entropy by 42%; in the YU2 gp140-foldon trimer, the PF2 substitution reduced entropy by ϳ40%; and, unexpectedly, in the CAAN gp140-F trimer context, the PF2 substitutions increased entropy by nearly 49%. Moreover, in the context of the YU2 gp120⌬V123/PF2-GCN4 trimers, the deletion of V1, V2, and V3 regions reduced the entropy by 26% (25 kcal/mol compared with the previously reported 33 kcal/mol with intact V regions). In this context, the combination of PF2 and V-region deletion reduced the entropy by 54%, compared with the full-length gp120/PF2-GCN4 trimer. The YU2 gp140-F trimer showed similar entropic change and much less enthalpic change upon CD4 binding than did the full-length YU2 gp120-GCN4 trimer. Interestingly and unexpectedly, the CAAN gp140-F trimers showed significantly reduced enthalpy changes compared with all other variants upon interaction with CD4.
ITC is considered one of the most reliable methods available to determine ligand stoichiometric values. Therefore, from the existing thermodynamic data, we examined the interaction of sCD4 with the variant Env constructs used here (see Table 2). Consistent with previous analyses (46,59), the stoichiometry of sCD4 for the gp120 core/PF2 interaction was close to unity (n ϭ ϳ1). The YU2 gp120⌬V123/PF2-GCN4 trimers also displayed a stoichiometry value close to n ϭ 1, indicating that sCD4 bound each ⌬V123/PF2 protomer within a given GCN4-trimerized oligomer. These results are consistent with our previous data indicating that removal of the major variable regions in the gp120-GCN4 trimeric context also demonstrated a 1:1 stoichiometry for interaction with sCD4 (59,66). The stoichiometric interaction of sCD4 with the CAAN gp140-F trimers was only 0.36 per protomeric subunit within the trimer, suggesting a relatively restricted binding stoichiometry of CD4 to these trimers.
Immunogenicity of the gp120/gp140 Variants in Rabbits-We next performed sequential inoculations of the Env variant constructs into rabbits to address several specific questions (Fig. 4). We first sought to test whether priming with the "most" open state of Env (the core/PF2 monomer) and boosting with progressively more "closed" states (gp120⌬V123/PF2-GCN4 trimer followed by the gp140-F trimers) might better elicit neutralizing antibodies by focusing the B cell response on the more conserved elements, such as the CD4bs. Within this approach, we hoped to better engage B cells specific for the CD4bs by including the PF2 mutations in the individual immunogens. We also reversed the order to prime with the most "closed" state and to boost with the "most" open, but this did not alter any of the neutralization outcomes relative to the open-to-closed regimen (data not shown). Because of the varying CD4-interactive properties revealed by the biophysical analysis, we also changed the Env strains (and Env clades) in an attempt to focus the immune response on the conserved and "in common" CD4binding site or other conserved Env regions (see the rationale in Fig. 1). For comparison, the YU2 gp120 ⌬V123/PF2-GCN4 trimers and the well studied YU2 gp140-F trimers were also inoculated in homologous immunization regimen. The YU2 gp140-F trimers served as a model immunogen, used in our previous studies (44,46,56,67), to allow comparisons with the new immunogens and regimens tested here.
All proteins purified by affinity chromatography were emulsified in Adjuplex adjuvant and inoculated two times for each protein as described under "Experimental Procedures" and Fig.  4. Besides pre-bleed sera, two rabbits inoculated with BSA in the Adjuplex adjuvant served as additional negative controls for the study. Following the inoculation regimen, sera were tested for binding activity to WT YU2 gp120 protein by ELISA. After  FEBRUARY 17, 2012 • VOLUME 287 • NUMBER 8 two to three inoculations, all of the animals inoculated with gp120/gp140 variants achieved peak titers of anti-gp120 IgG, with the EC 50 titers ranging from 3.1 ϫ 10 4 to 6.3 ϫ 10 5 (supplemental Fig. S3). Sera from BSA-immunized rabbits had no detectable binding titer in regard to gp120-reactive antibodies (data not shown). Because we were interested in CD4bs-directed responses, we first assessed if any of the regimens elicited B cell responses  directed toward the CD4bs. To identify CD4bs binding specificities present in the antisera, we performed a cross-competition ELISA with the post-6 sera pooled from each group of animals. As described previously by Dey et al. (46), biotinylated VRC01 antibody was used in the ELISA to compete with the serially diluted sera for binding to the HXBc2 coreV3S protein coated on the ELISA wells. As shown in supplemental Fig. S1, by this means of analysis, all animals in each group inoculated with either heterologous or homologous Env-based regimens elicited CD4bs-directed binding antibodies. To confirm the presence of CD4bs antibodies by direct binding analysis, we analyzed pooled sera derived from each group of animals and performed direct binding to the previously described resurfaced HXBc2 core (RSC3) (35). These analyses revealed a trend toward increased RSC3 recognition by the sera derived from the YU2 gp120⌬V123/PF2-GCN4 inoculated animals, perhaps suggesting that there were more CD4bs-directed binding antibodies elicited in these animals, but the difference was not statistically significant (data not shown). Next, we assessed the serum neutralizing activity using a panel of 10 HIV-1 Env pseudo-viruses (and an SIV Env control virus) to examine the breadth and specificity of neutralization elicited by the heterologous and homologous inoculation strategies. As seen in Fig. 4B, neutralizing activity, reported as ID 50 values (50% inhibitory dilution value), was detected against the Tier 1A viruses (MN, MW965, and SF162) in all sera derived from Env-inoculated animals following the 6th immunization. Note that although the V region-deleted trimers elicited neutralizing activity against these viruses, the YU2 gp140-F and CAAN gp140-F trimers, containing the intact V regions, elicited much higher titers against MW965 and SF162 (p ϭ 0.0286). Against the HXBc2 virus, moderate levels of neutralization were detected in most sera. Against the moderately sensitive isolates such as BaL.01 or SS1196, regimens that included the gp140-F trimers elicited low levels of neutralization compared with the barely detectable neutralization elicited by the V-region-deleted trimers, suggesting involvement of variable region-directed activity. Against the Tier 2 isolates, the neutralizing activity was generally sporadic and/or weak, with the CAAN gp140-F trimers displaying the broadest activity compared with the other regimens, albeit at quite modest levels of activity (Fig. 4).

Stable HIV Trimers Elicit CD4-binding Site Neutralization
Increased Tier 1 Neutralization Potency Elicited by the gp140-F Trimers Maps to V3-The pattern of neutralization suggested that the increased potency elicited by the gp140-F trimers to MW965 and SF162 compared with the YU2 gp120⌬V123/PF2-GCN4 trimers (lacking V regions) involved the V regions, most likely neutralizing determinants within V3. Previous studies, including our own, reported that the V3 region elicited much of the neutralizing response against selected viruses, such as 89.6, SF162, or MN (44,67,68). Therefore, we performed peptide inhibition assays by adding V3 peptide to the sera in the neutralization assay as described previously (43,44). By this method we were able to map much of the neutralizing activity elicited by the YU2 and CAAN gp140-F trimers against the Tier 1A isolates, SF162 predominantly to V3 (see Table 3). Similarly, the lower level of neutralizing activity elicited against BaL.01 and JRCSF by the YU2 and CAAN trim-ers was also somewhat V3-directed (see Table 3). As expected, the V3-lacking YU2 gp120⌬V123/PF2-GCN4 trimer-elicited neutralizing activity, which was limited to sensitive viruses, was unaffected by the addition of the V3 peptide.
Mapping CD4bs Neutralization Elicited by Different Trimer Formats-We then sought to determine whether the neutralizing activity elicited by the YU2 gp120⌬V123/PF2-GCN4 trimers, lacking the V regions, was directed against the CD4bs and to compare such CD4bs-directed neutralization to those elicited by the YU2 and CAAN gp140-F, V region-containing trimers, using the differential neutralization inhibition assay. Because the YU2 gp120⌬V123/PF2-GCN4 trimers eliciting neutralizing activity was limited to sensitive isolates, we restricted that analysis to the HXBc2 and MW965 viruses. Because HXBc2 is mismatched in the V3 region compared with the YU2 or CAAN gp140-F trimers, we observed previously that it was not generally neutralized by V3-directed antibodies elicited by YU2 Env. Previously, differential absorption of antibody subpopulations by beads coupled with gp120 proteins was used to identify the neutralizing specificities presented in serum (28,29). However, this is a labor-intensive assay that is difficult to perform with vaccine-induced antisera possessing modest neutralizing activity. To develop a higher throughput "protein inhibition neutralization assay" (similar to our previous differential adsorption binding assay (28,29) and similar to the strategy used to map sera from HIV-1ϩ individuals (35)), we designed a pair of probes, referred to here as TriMut and TriMut368/70, to potentially inhibit the neutralizing capacity of any core-directed neutralizing antibodies, as well as to map CD4bs-directed neutralizing antibodies. Both of the probes were based on HXBc2 coreV3S protein and contained three mutations in the gp120 bridging sheet region (423I/M, 425N/K, CD4bs probes were incubated with sera. Percent reduction in titer of sera with scramble V3 peptide or YU2 V3 peptide was calculated compared with mock. Net reduction was calculated by subtracting scramble V3 peptide effect from the YU2 V3 peptide effect. Net neutralization reduction by CD4bs probes was calculated the same way. If 50% neutralization was not reached, a titer of 2.5 was assigned for calculation of the percent reduction and net reduction. b The pseudotyped HIV-1 isolates that were used for mapping the neutralizing specificity of the serum. c ND means not determined. d Data were not calculated due to lack of measurable neutralization. and 431G/E), which eliminates CD4 binding but does not affect recognition by any of the known CD4bs antibodies (69). Because the TriMut core protein is rendered CD4 binding-defective by these substitutions, it can be directly added into the HIV entry assay without interfering with the CD4-dependent virus entry process that is detected in the entry/neutralization assay. In addition, one of the probes, TriMut368/70, contains two extra mutations in the CD4 binding loop (368D/R and 370E/F, known to specifically eliminate recognition by most CD4bs-directed antibodies). The antibody binding profile of the probes is shown in Fig. 5A. In brief, TriMut was recognized by all CD4bs antibodies analyzed, whereas the TriMut368/70 lost recognition of binding of most CD4bs antibodies, with HJ16 and b6 being the exception. Binding of 2G12, which is directed to core moieties outside the CD4bs, demonstrated that both probes were native and correctly folded. Specificity to detect CD4bs-directed neutralization by direct addition to the neutralization assay was verified using VRC01, a neutralizing antibody directed to CD4bs, and by 17b, a neutralizing antibody directed to co-receptor binding site, which served as a negative control (Fig. 5B). As shown, VRC01-mediated neutralization was inhibited by the TriMut core but not by the TriMut368/70, validating the differential neutralization absorption process. In contrast, and as expected, neutralization of HXBc2 by the co-receptor-binding site mAb 17b was not inhibited by either protein. Human HIV immunoglobulin and normal human IgG were also used as assay controls (see supplemental Fig. S4).
Using this differential neutralization inhibition process, CD4bs-directed neutralization of selected sera was mapped. We analyzed selected regimens to determine whether any regimen was exceptional in regard to the elicitation of CD4bsdirected neutralizing antibodies. Generally, we found that after two inoculations the weak neutralizing activity elicited by the HXBc2 gp120 core/PF2 did not map to the CD4bs, whereas nearly all of the more potent HXBc2 neutralizing activity elicited by the YU2 gp120⌬V123/PF2-GCN4 trimers did map to the CD4bs (see Fig. 5C for representative curves). After two inoculations, the YU2 gp140-F trimers elicited much less potent neutralization activity compared with the YU2 gp120⌬V123/PF2-GCN4 trimers, but such activity did map to the CD4bs (Fig. 5C). Following six inoculations, the YU2 gp140-F trimers elicited more neutralizing activity and most mapped to the CD4bs. In contrast, CAAN trimers elicited very low levels of HXBc2 neutralization activity after two inoculations, and although there was detectable neutralization activity following six inoculations, little mapped to the CD4bs. As well, the heterologous prime boosting, culminating in the two CAAN gp140-F inoculations, did not elicit much CD4bs directed neutralization activity (Fig. 5C, upper right panel). The results from the homologous regimens are summarized in Fig.  5D. In addition, compared with the homologous YU2 gp120⌬V123/PF2-GCN4 trimer group of animals, there was less CD4bs-directed neutralizing antibodies elicited in the post-4 and post-6 inoculation sera from selected heterologous prime/boost groups of animals (see supplemental Figs. S4 and S5), implicating other Env regions as targets of this neutralizing activity.

DISCUSSION
In this study, we generated a set of novel immunogens to test two related hypotheses in a well controlled manner. A primary focus was to test if sequential immunization of selected Env formats would focus the elicited B cell immune response on conserved elements of Env to better elicit broadly neutralizing antibodies. The CD4-state stabilizing PF2 mutations was included to potentially attract naive B cells bearing B cell receptors with higher affinity for the CD4bs, and format alterations included increasing the oligomeric status from monomer to trimer and, beginning with immunogens that had the V regions deleted, to better prime for responses to the CD4bs. Because the conserved CD4 binding region is the major immune target on the core/PF2 construct, we reasoned that it was not critical to use a primary isolate Env as the basis for this immunogen. We also performed homologous regimens using V-region-deleted, PF2-stabilized trimers and gp140-F trimers derived from highly resistant primary isolates. Following production and purification of stabilized monomers and gp120/gp140 trimers, we performed an analysis of these glycoprotein immunogens to characterize their differential antigenic profile, to evaluate their molecular homogeneity, and to examine differences in their interactions with soluble forms of the primary virus receptor CD4. Finally, we inoculated selected immunogens into rabbits, assessed their neutralization capacity, and mapped neutralization elicited by several of the immunogens and regimens.
Interestingly, the YU2 gp120 ⌬V123/PF2-GCN4 trimers were the best recognized proteins by the CD4bs ligands CD4-Ig, VRC01, and VRC-PG04, suggesting that both V region deletion and CD4-state stabilization did alter antigenicity. The CAAN gp140-F trimers displayed the highest intrinsic affinity for CD4 but the lowest level of stoichiometric interaction. Despite multiple alterations of each Env format, recognition by the CD4bs neutralizing mAb b12 was maintained, with the exception of trimers derived from the b12-resistant virus CAAN. In general, and as part of our prime/boost rationale, recognition by the non-neutralizing antibodies was decreased in the core/PF2 context (and should not be primed; i.e. F105 or C11) but did increase in the more complete gp140 trimers. We demonstrated that although high titer and high affinity antibodies were elicited following selected prime/boost Env/adjuvant vaccination regimens, the greatest level of in vitro neutralization breadth was elicited by the CAAN gp140-F trimers. However, elicitation of CD4bs-directed Tier 1 virus neutralizing antibodies was most rapid when using the V region-deleted, CD4-state conformationally fixed YU2 gp120 trimers, as demonstrated by differential inhibition of neutralization using the new and novel Env-based probes.
All Env immunogens used here elicited antibodies capable of cross-competing with the high affinity and recently isolated broadly neutralizing CD4bs antibody VRC01. This is consistent with previous reports using sCD4 or b12 in similar assay formats (35). Of particular interest was the finding that the V region-deleted, CD4-state gp120 trimer YU2 ⌬V123/PF2-GCN4 was better recognized compared with all other constructs by the potent and broadly neutralizing CD4bs-directed mAbs VRC01 and VRC-PG04. Not only was recognition of these constructs the best, but the YU2 gp120⌬V123/PF2-GCN4 trimers also most efficiently elicited the highest levels of CD4bs-directed antibodies against the Tier 1 viruses of HXBc2 and MW965. This is a very interesting association and is reminiscent of our earlier work in which we reported that an increase in 17b affinity for the co-receptor-binding site and stabilization of this region resulted in the greatly increased elicitation of antibodies to this site. This association points to a potential avenue to better elicit CD4bs mAbs based upon improvements of the YU2 gp120⌬V123/PF2-GCN4 "platform" or immunization regimen. However, in this study, this CD4bsdirected neutralizing response was not capable of neutralizing the more resistant viruses, suggesting that either the elicited CD4bs neutralizing titers present in the serum are at a level too low to be detected against the more resistant isolates or more likely that the type of antibodies that can access the CD4-binding site on the more sensitive isolates cannot do so on the more resistant viral isolates at any reasonable concentration. This is perhaps due to the overall antibody "shape" (i.e. relatively short CDR loops or angle used to access the CD4bs), the low levels of somatic hyper-mutation or some limitations due to the actual antibody-Env footprint contacts, or the interplay of all three of these factors. Further analysis of the type of CD4bs neutralizing antibodies elicited by the YU2 gp140-F trimer would provide an answer to this limitation, and such studies are ongoing in nonhuman primates.
Our data suggest that perhaps the combinations of engineered modifications, such as loop deletion, cavity-filling substitutions, and perhaps glycan-masking, along with as yet untested cysteine-pair stabilizing substitutions, which, although not sufficient as individual modifications, may in combination enhance the CD4bs-directed neutralizing response. However, much remains to be learned at the basic B cell biology level to understand the effects of prime/boosting to better enhance responses to the relatively poorly immunogenic gp120 CD4bs and to elicit CD4bs antibodies possessing the neutralizing properties of b12 or VRC01, rather than the limited neutralizing properties elicited by the Env immunogens used here and elsewhere. Note that the heterologous immuni- FIGURE 5. Characterization of probes for differential neutralization inhibition assay and mapping of nonpotent CD4bs neutralizing antibodies in antisera. Based on HXBc2 coreV3S, we designed two probes, TriMut with triple mutations (I423M, N425K, and G431E) and TriMut368/70 with two additional mutations (D368R and E370F), to map the CD4bs neutralizing antibodies present in sera by differential neutralization adsorption. The probes were preincubated, respectively, with serially diluted antisera in the neutralization assay, and differences between TriMut compared with the CD4bs "knock-out" mutations at 368/70 revealed the extent of neutralization directed to the CD4bs. Sera from rabbits of four groups were selected; one group of animals had been inoculated with heterologous prime/boost regimens (two inoculations with HXBc2 core/PF2, two inoculations with YU2 gp120⌬V123/PF2-GCN4, and two final inoculations with CAAN gp140-F), and three groups of animals were inoculated with a homologous regimen (six inoculations with YU2 gp120⌬V123/PF2-GCN4, or YU2 gp140-F, or CAAN gp140-F). A, binding profile of the probes with a panel of CD4bs antibodies and CD4-Ig. Binding of antibodies to TriMut and TriMut368/70 was determined by ELISA and SPR. B, validation of the differential neutralization absorption assay. Neutralization absorption of the CD4bs mAb, VRC01, the non-CD4bs, and co-receptor binding site mAb 17b was measured. C, differential adsorption neutralization assay against the HXBc2 virus by selected sera. Differential neutralization absorption curves of post-2 and post-6 sera from four rabbits are shown as representative examples. D, comparison of CD4bs neutralizing antibodies elicited with the different immunization regimens. Difference of neutralization absorbed by the two probes at 1:12 serum dilution was plotted by data derived from each respective group of animals to determine relative content of HXBc2 neutralizing CD4bs-directed antibodies. zation regimens that we used here involved two inoculations of each immunogen in series, and it is possible, although perhaps unlikely, that a single inoculation per immunogen might yield a more efficient elicitation of broadly neutralizing antibodies.
Here, we demonstrated that the CAAN gp140-F trimers, and to a certain extent the YU2 gp140-F trimers, elicited the most breadth in vitro but predominantly against viruses that display some level of sensitivity to V3-directed mAbs. Also these data suggest that the increased neutralization breadth elicited by homologous inoculation of the gp140-F trimers, compared with the heterologous prime/boost regimens involving the loop-truncated immunogens, is likely due to V region-directed antibodies (55,68,70). Similarly, for homologous neutralization of the YU2 isolate, some V region-directed responses are likely contributing to this effect, consistent with our previous immunogenicity/mapping studies (44,67). Although the CAAN trimers also weakly neutralized YU2 (and JR-FL), we were not able to map the specificity of the elicited neutralizing antibodies.
In sum, this comprehensive study provides parameters to consider associated with the elicitation of broader neutralizing responses, such as CD4 affinity and better recognition by the relatively new tools of the broadly neutralizing CD4bs-directed mAbs VRC01 and VRC-PG04. The data suggest that if we are able to improve CD4, VRC01, or VRC-PG04 recognition of and interaction with candidate immunogens by additional design and coupled with V region deletion, we may better focus the response on the CD4bs. In addition, with prime/boost Env regimens aimed to promote B cell responses and enhanced somatic hypermutation of B cells directed against the conserved CD4bs neutralizing target of HIV-1 Env, we may be able to increase the potency of the elicited CD4bs-directed antibodies.