A Native-Like SOSIP.664 Trimer Based on an HIV-1 Subtype B env Gene

ABSTRACT

Recombinant trimeric mimics of the human immunodeficiency virus type 1 (HIV-1) envelope glycoprotein (Env) spike should expose as many epitopes as possible for broadly neutralizing antibodies (bNAbs) but few, if any, for nonneutralizing antibodies (non-NAbs). Soluble, cleaved SOSIP.664 gp140 trimers based on the subtype A strain BG505 approach this ideal and are therefore plausible vaccine candidates. Here, we report on the production and in vitro properties of a new SOSIP.664 trimer derived from a subtype B env gene, B41, including how to make this protein in low-serum media without proteolytic damage (clipping) to the V3 region. We also show that nonclipped trimers can be purified successfully via a positive-selection affinity column using the bNAb PGT145, which recognizes a quaternary structure-dependent epitope at the trimer apex. Negative-stain electron microscopy imaging shows that the purified, nonclipped, native-like B41 SOSIP.664 trimers contain two subpopulations, which we propose represent an equilibrium between the fully closed and a more open conformation. The latter is different from the fully open, CD4 receptor-bound conformation and may represent an intermediate state of the trimer. This new subtype B trimer adds to the repertoire of native-like Env proteins that are suitable for immunogenicity and structural studies.

IMPORTANCE The cleaved, trimeric envelope protein complex is the only neutralizing antibody target on the HIV-1 surface. Many vaccine strategies are based on inducing neutralizing antibodies. For HIV-1, one approach involves using recombinant, soluble protein mimics of the native trimer. At present, the only reliable way to make native-like, soluble trimers in practical amounts is via the introduction of specific sequence changes that confer stability on the cleaved form of Env. The resulting proteins are known as SOSIP.664 gp140 trimers, and the current paradigm is based on the BG505 subtype A env gene. Here, we describe the production and characterization of a SOSIP.664 protein derived from a subtype B gene (B41), together with a simple, one-step method to purify native-like trimers by affinity chromatography with a trimer-specific bNAb, PGT145. The resulting trimers will be useful for structural and immunogenicity experiments aimed at devising ways to make an effective HIV-1 vaccine.

INTRODUCTION

Immunogens capable of inducing protective titers of broadly neutralizing antibodies (bNAbs) are being widely sought for use in vaccine design strategies for human immunodeficiency virus type 1 (HIV-1) (1). The basis of this approach is that bNAbs can prevent globally diverse HIV-1 strains from infecting target cells. They do so via binding to the envelope glycoprotein (Env) complex on the virion surface, an event that is both necessary and sufficient to neutralize HIV-1 infectivity (2, 3). One of the more common strategies to induce bNAbs involves the design of soluble, recombinant protein mimics of the native Env complex, a meta-stable structure comprising three gp120 and three gp41 subunits. The production of soluble Env trimers involves introducing a stop codon to truncate the gp41 ectodomain (gp41_ECTO) subunit prior to the transmembrane region to yield soluble gp140 proteins (4–6). The fragility of the Env complex is, however, a substantial problem from a protein engineering perspective, as the natural noncovalent interactions between the six subunits are not robust enough to allow soluble trimers to be expressed and purified without the use of stabilizing changes.

The most commonly used method to stabilize soluble gp140 proteins has been to eliminate, by mutagenesis, the natural cleavage site between gp120 and gp41_ECTO (7–19). It is now clear that the resulting uncleaved gp140 (gp140_UNC) proteins, when purified by size exclusion chromatography (SEC), contain the right number of gp120 and gp41_ECTO subunits (i.e., 3 of each). However, these various SEC-purified gp140_UNC protein populations mostly (90 to 100%) adopt nonnative configurations whether or not they contain additional “trimerization motifs” at the C terminus of gp41_ECTO (20–26; L. K. Pritchard, S. Vasiljevic, G. Ozorowski, G. E. Seabright, A. Cupo, R. W. Sanders, K. J. Doores, D. R. Burton, I. A. Wilson, A. B. Ward, J. P. Moore, and M. Crispin, submitted for publication). Electron microscopy (EM), other biophysical measurements, and glycan profiling have shown or implied that three semidissociated gp120 moieties remain linked to a postfusion form of trimeric gp41_ECTO via the uncleaved, intersubunit peptide (20–22, 24, 25; Pritchard et al., submitted). Accordingly, this category of gp140_UNC proteins does not form native trimers on a population basis and lacks the antigenic characteristics of native Env spikes; they express many non-NAb epitopes but not those for various key bNAbs, and their gp120 subunits contain aberrant inter- and intramolecular disulfide bonds (21, 24, 25, 27–32). Similar concerns apply to uncleaved, full-length gp140 proteins expressed on the cell surface (27–30, 33).

An alternative approach to soluble trimer design involves actively promoting the natural gp120-gp41_ECTO cleavage event while also introducing specific stabilizing mutations, i.e., an engineered disulfide bond that covalently links the two subunits and an Ile-to-Pro change at residue 559 that helps maintain the gp41_ECTO moieties in the prefusion form (4, 6). In addition, the truncation of gp41_ECTO at residue 664 eliminates a hydrophobic region that tends to cause trimer aggregation (34, 35). The resulting trimers are designated SOSIP.664 gp140s. The paradigm of this particular trimer design is based on the BG505 subtype A pediatric founder virus (36–38). The BG505 SOSIP.664 trimers have antigenic properties and morphologies that mimic those of native Env complexes and were the trimers used for determination of high-resolution X-ray crystallography and cryo-electron microscopy (cryo-EM) Env structures (23, 38–41). Their potential as immunogens is currently under evaluation.

We now seek to increase the repertoire of native-like soluble trimers available for structural and immunogenicity studies. Here, we describe the B41 SOSIP.664 trimer based on a subtype B env gene. Such trimers appear fully native-like when viewed by negative-stain electron microscopy (NS-EM) and have antigenicity properties comparable to those of their subtype A BG505 counterparts. We also report on how to overcome production issues associated with proteolytic clipping of the V3 region, which generally arises with subtype B Env proteins (42–45), including via use of positive-selection affinity columns based on the trimer-specific bNAb PGT145. NS-EM images show that the purified B41 trimers contain two distinguishable subpopulations. We propose that the subpopulations represent equilibrium between the fully closed (BG505-like) and a more open conformation that is an intermediate between the fully closed form and the fully open, CD4-bound configuration.

MATERIALS AND METHODS

Construct design.

The B41 env gene was derived from a subtype B founder virus isolated from an HIV-1-infected serial plasma donor; its more formal designation is 9032-08.A1.4685, and its GenBank accession number is EU576114 (46). To make the SOSIP.664 gp140 construct, we introduced the following sequence changes (HxB2 numbering system): A501C and T605C (gp120-gp41_ECTO disulfide bond [4]), I559P in gp41_ECTO (trimer stabilization [6]), REKR to RRRRRR in gp120 (cleavage enhancement [5]), tissue plasminogen activator (TPA) leader peptide (increase of gene expression [38, 47]), and a stop codon to terminate the protein after gp41_ECTO residue 664 (improvement of homogeneity and solubility [34, 35]). The resulting, codon-optimized B41 SOSIP.664 env gene was obtained from GenScript (Piscataway, NJ) and cloned into pPPI4 using PstI and NotI (4). Trimer variants containing a D7324 epitope tag sequence at the C terminus of gp41_ECTO were also made, as previously described, by adding the sequence GSAPTKAKRRVVQREKR after residue 664 in gp41_ECTO, followed by a stop codon (38). The resulting trimers are designated B41 SOSIP.664-D7324.

A monomeric B41 gp120 protein with the same sequence (other than the changes listed below) as that subunit of the trimer was designed by introducing a stop codon into the SOSIP.664 construct at residue 512, reverting the optimized cleavage site to the wild type (RRRRRR→REKR at residues 508 to 511), reverting the A501C change, and making an A500K substitution to optimize the D7324 epitope that is present in the C5 domain at the gp120 C terminus.

The B41 gp160 clone for generating Env-pseudotyped viruses for neutralization assays has been described elsewhere (46).

Env protein expression by transient transfection.

Env proteins were expressed in wild-type, adherent HEK293T (referred to as 293T) cells by transient transfection of env genes using polyethyleneimine (PEI), essentially as described elsewhere (38, 48). The furin gene was cotransfected with all SOSIP.664 trimer-encoding env genes to maximize gp120-gp41_ECTO cleavage (4, 5). This method involves culturing the transfected 293T cells in the presence of 10% fetal calf serum (FCS). Under these conditions, we found that serum proteases clip the V3 region of gp120 subunits of some of the B41 SOSIP.664 trimers, and of the B41 gp120 monomers, damaging the proteins (see Results). To overcome this problem, we used the 293Fectin lipofection system to transiently transfect 293F cells in suspension cultures with Env and Furin expression plasmids (4:1 ratio) and cultured the transfected cells in serum-free medium, following the manufacturer's (Invitrogen) recommendations (see Results).

Env protein production from stable cell lines.

We have described elsewhere how to make stable CHO and 293T cell lines that express fully cleaved BG505 SOSIP.664 trimers (49). We used the same method to make lines that produce B41 SOSIP.664 trimers. To reduce the extent of V3 clipping, it was necessary to try to adapt the stable lines to growth at lower serum concentrations than the standard culture conditions (i.e., with 10% FCS present). With the stable 293T line, we found that the FCS concentration could only be reduced to 5% without a substantial decrease in cell viability; at 5% FCS, some V3 clipping still occurred, so we discontinued our work with this line. However, the stable CHO cell line could be successfully adapted to 1% serum and under these conditions, V3 clipping was negligible (see Results). The stable CHO cells were then propagated under hygromycin resistance in PRO-CHO-AT medium (Lonza) containing 1% FCS.

B41 SOSIP.664 trimer and gp120 monomer purification.

We used two methods to purify B41 SOSIP.664 trimers (and the D7324-tagged variants) from the transfection supernatants. The first procedure is described in detail elsewhere (21, 38). Briefly, Env proteins were isolated via a bNAb 2G12 affinity column and MgCl₂ elution, and the trimer fraction was purified by SEC on a Superdex 200 26/60 column (GE Healthcare) (21, 38); this method is referred to as 2G12/SEC.

The second method involved an affinity column based on the PGT145 bNAb that recognizes a trimer-specific epitope (25, 38, 50). PGT145 was coupled to CNBr-activated Sepharose 4B beads (GE Healthcare). Env-containing culture supernatants were flowed through the resulting column, the beads were washed with 2 column volumes of buffer (0.5 M NaCl, 20 mM Tris [pH 8.0]), and the bound trimers were eluted using 1 column volume of 3 M MgCl₂. The eluted trimers were immediately buffer exchanged into 75 mM NaCl–10 mM Tris, pH 8.0, using SnakeSkin dialysis tubing (molecular weight cutoff [MWCO]of 10,000 [10K]) (Thermo Scientific). The purified trimers were then concentrated using Vivaspin columns with a 30-kDa cutoff (GE Healthcare). Unless otherwise indicated, the trimers eluted from the PGT145 column were then further purified by SEC; this method is referred to as PGT145/SEC.

We made B41 gp120 by transfecting 293F cells in serum-free medium, followed by purification via the 2G12 affinity column and MgCl₂ elution. Further fractionation by SEC was not required. The resulting gp120 proteins were not V3 clipped (<5%).

In all cases, the concentrations of purified proteins were determined using either a bicinchonic acid-based assay (BCA assay; Thermo Scientific, Rockford, IL) or UV₂₈₀ absorbance using theoretical extinction coefficients (51).

SDS-PAGE, BN-PAGE, and Western blotting.

Env proteins were analyzed using either the 10% Tris-glycine, 4 to 12% bis-Tris NuPAGE SDS-PAGE, or blue native PAGE (BN-PAGE) system followed by staining with Coomassie blue dye (38, 52), according to the manufacturer's recommendations (Invitrogen). When reducing conditions were required, e.g., to assess gp120-gp41_ECTO cleavage or V3 clipping, the Env proteins were mixed with 0.1 M dithiothreitol (DTT) prior to loading onto the SDS-PAGE gels. For assessments of V3 clipping, we tested more rigorous denaturing and reducing conditions, specifically by boiling the samples in 0.1 M DTT and SDS for 10 min prior to analysis on a Tris-glycine gel. This protocol yielded an outcome similar to the one described above, confirming that we were not underestimating the extent of clipping due to incomplete gp120 denaturation (data not shown).

Western blotting of SDS-PAGE gels was carried out as described previously, using the anti-gp120 monoclonal antibody (MAb) ARP3119 (NIBSC Reagent Repository, United Kingdom) or a pool of serum derived from individuals infected with subtype B HIV-1 strains (HIV-Ig) (49). ARP3119 recognizes a conserved, linear epitope (residues EDIISLW) in the C1 region of gp120 and hence detects the 70-kDa fragment that is produced when V3-clipped gp120 is fractionated on a reducing SDS-PAGE gel (the other, 50-kDa, fragment does not contain the ARP3119 epitope).

Antibodies and Fabs.

Antibody concentrations are generally recorded in micrograms per milliliter for neutralization assays and trimer binding enzyme-linked immunosorbent assays (ELISAs) but as nanomolar concentrations for isothermal titration calorimetry (ITC) experiments. Since the molecular mass of an average IgG molecule is approximately 150,000 Da, the conversion factors for IgG are as follows: 1 μg/ml = 6.7 nM and 1.0 nM = 150 ng/ml.

MAbs were obtained as gifts, or purchased, from the following sources: John Mascola and Peter Kwong (VRC01 and F105), the International AIDS Vaccine Initiative (PGV04, PG9, PG16, PGT121, PGT126, PGT135, PGT145, b6, b12, and F240), Polymun Scientific (447-52D and 2G12), Michel Nussenzweig (8ANC195), James Robinson (39F, 17b, A32, 19b, 14e, F91, and CO11), and Mark Connors (35O22).

Neutralization assays.

To determine the extent of HIV-1 neutralization by MAbs, we used Env-pseudotyped viruses and a single-cycle infection assay based on TZM-bl cells, as described previously (38). All infections were performed in duplicate. Uninfected cells were used to correct for background luciferase activity. The infectivity of each mutant without inhibitor was set at 100%. Nonlinear regression curves were determined and 50% inhibitory concentrations (IC₅₀s) were calculated using a sigmoid function in Prism software version 5.0. Celia LaBranche and David Montefiori (Duke University Central Laboratory, Durham, NC) have classified the B41 (referred to as 9032-08.A1.4685) Env-pseudotyped virus as having a tier 2 neutralization profile (46).

D7324 capture ELISA.

ELISAs using 2G12/SEC-purified or PGT145/SEC-purified B41 SOSIP.664-D7324 trimers, or B41 gp120 monomers, were performed essentially as described previously (21, 38). The input trimer concentration was 300 ng/ml, and the corresponding gp120 concentration (generally ∼900 ng/ml) was calibrated to yield similar binding signals for MAb 2G12. The sheep polyclonal Ab D7324 to a gp120 C5 epitope used for coating the ELISA wells was obtained from Aalto Bioreagents, Dublin, Ireland.

Correlations between the neutralization IC₅₀ and ELISA 50% effective concentrations (EC₅₀s) were investigated by nonparametric Spearman's rank correlation test with two-tailed P values for significance (Prism 6 for Mac OSX; GraphPad). The α-level was set to 0.05.

Differential scanning calorimetry (DSC).

Thermal denaturation was probed with a MicroCal VP-Capillary differential scanning calorimeter (Malvern Instruments) (38). Before the experiments were carried out, all samples were extensively dialyzed against phosphate-buffered saline (PBS). The protein concentration was subsequently adjusted to 0.1 to 0.3 mg/ml, as described above. After loading of the protein sample into the cell, thermal denaturation was probed at a scan rate of 90°C/h. Buffer correction, normalization, and baseline subtraction procedures were applied before the data were analyzed using Origin 7.0 software. The data were fitted using a non-two-state model, as the asymmetry of some of the peaks suggested the presence of unfolding intermediates.

ITC.

ITC binding experiments were carried out using a MicroCal Auto-ITC200 instrument (GE Healthcare) and a protocol similar to one previously described elsewhere (50, 53). Briefly, Env proteins were dialyzed against Tris-saline buffer (150 mM NaCl, 20 mM Tris [pH 8.0]) prior to conducting the titrations. The absorbance at 280 nm, derived using calculated extinction coefficients, served to determine and adjust the input protein concentrations. The ligand present in the syringe was PGT145 Fab, PGT151 Fab, or 19b IgG at a concentration between 10 and 100 μM, while the B41 SOSIP.664 trimer was present in the cell at a concentration between 3 and 5 μM. In each binding experiment, the reference power was 5 μcal. For PGT151 and 19b binding experiments, the first injection of 0.5 μl was followed by 15 injections of 2.5 μl each, at intervals of 180 s. Because of the low binding enthalpy, PGT145 binding experiments were carried out with a first injection of 0.5 μl, followed by 11 injections of 3.5 μl each, at intervals of 180 s. Origin 7.0 software was used to derive the stoichiometry of binding (N), the affinity constant (K_d) and the molar reaction enthalpy (ΔH), by fitting the integrated titration peaks via a single-site binding model. (All measured and derived thermodynamic parameters of binding are reported in Table 2.)

TABLE 2.

Thermodynamic parameters of 19b, PGT151, and PGT145 binding to B41 and BG505 SOSIP.664 trimers measured by ITC^a

Antibody group and epitope	Binding expt	ΔG (kcal mol−1)	ΔH (kcal mol−1)	−TΔS (kcal mol−1)	Kd (nM)	N
Non-NAb, V3	19b IgG into BG505	−10.6	−41.4	30.8	21	0.1
	19b IgG into B41	−11.9	−39.5	27.6	3	0.2
bNAb, gp120-gp41ECTO interface	PGT151 Fab into BG505b	−9.3	−23.8	14.5	130	1.3
	PGT151 Fab into B41	No detectable binding	No detectable binding	No detectable binding	No detectable binding	No detectable binding
bNAb, trimer apex	PGT145 Fab into BG505	−9.9	−3.3	−6.6	63	0.5
	PGT145 Fab into B41	−10.9	−13.2	−2.3	9	0.4

^aReported values are averages from at least two independent measurements; associated errors are ∼10% of the average. Representative isotherms can be found in Fig. 4.

^bPreviously reported in reference 28.

NS-EM.

B41 SOSIP.664 trimers were purified by either PGT145 or 2G12 affinity chromatography, followed by SEC, and then prepared for negative-stain electron microscopy (NS-EM) analysis as previously described for BG505 trimers (21, 38). Briefly, a 3-μl aliquot containing ∼0.01 mg/ml of trimer was applied for 5 s onto a carbon-coated 400 Cu mesh grid that had been glow discharged at 20 mA for 30 s and then was negatively stained with 2% (wt/vol) uranyl formate for 60 s. Data were collected using an FEI Tecnai T12 electron microscope operating at 120 keV, with an electron dose of ∼25 electrons/Ų and a magnification of ×52,000, which resulted in a pixel size of 2.05 Å at the specimen plane. Images were acquired with a Tietz TemCam-F416 CMOS camera using a nominal defocus range of 900 to 1,300 nm.

Image processing and 3D reconstruction.

Data processing methods were adapted from those used previously (38). Particles were picked automatically using DoG Picker and put into a particle stack using the Appion software package (54, 55). Initial, reference-free, two-dimensional (2D) class averages were calculated using particles binned by two via iterative multivariate statistical analysis (MSA)/multireference alignment (MRA) and sorted into classes (56). Particles corresponding to trimers were selected into a substack and binned by two before another round of reference-free alignment was carried out using the iterative MSA/MRA and Xmipp Clustering and 2D alignment software systems (57).

The 2D class averages were visually inspected, and the classes were segregated into one of three trimer structural groups designated “closed,” “open,” or “nonnative.” Trimers visually similar to those previously described for BG505 SOSIP.664 (38), namely, compact triangular propeller shapes with no additional density surrounding them, were classified as in the closed conformation. Classes containing triangular propeller density of an intensity and size similar to the closed conformation group, but displaying one, two, or three smaller spheres of density at the distal ends of the triangular density, were classified as in the open conformation (see Results). We consider both these categories of trimer as having, overall, a native structure. In contrast, particles that did not clearly show a central, triangular mass, but instead resembled previously described images of uncleaved, non-SOSIP gp140 proteins (21), were classified as nonnative forms. Any contaminating gp140 monomers, dimers, or noise particles identified were not included in further analysis. (Noise particles are inconsistencies in the carbon-coated nitrocelluose support and/or staining pattern that are mistaken for protein by the automated particle picking software.) Such contaminants made up <1% of total particles, as SEC purification removed most gp140 monomers and dimers, and care was taken to image only grid regions with a uniform stain density. The relative abundance of each subpopulation was then calculated as the percentage of particles belonging to a particular group divided by the total number of all trimeric particles. To ensure that the presence or absence of additional density was not stain dependent, three data sets were collected using three distinct squares of variable stain thickness (thin, medium, and thick) on the same grid. The calculated percentages for a specific subpopulation varied by <5% between the three data sets (data not shown). In addition, a new substack was created from only the open and nonnative classes and a further round of reference-free alignment was carried out. The same quantities of open and nonnative particles were calculated from the resulting 2D class averages, confirming that the features are true and not a result of misalignment. These quantities were consistent between the two alignment software systems tested.

To generate 3D NS-EM reconstructions, B41 SOSIP.664 trimers were incubated with either VRC01 or b12 Fab fragments, in a 10-fold molar excess compared to the trimer, at room temperature for 1 h. Grid preparation and data collection methods were similar to those used for unliganded trimers, with the addition of 10° tilt increments up to 50°. The tilts provided additional particle orientations to improve the image reconstructions.

Fabs were clearly visualized in the 2D class averages if they were bound to the trimer, allowing the percentage of Fab-bound versus unbound trimers to be tabulated. Ab initio common lines models were calculated from reference-free 2D class averages in EMAN2 (58) without imposing symmetry. One of those models was then refined against raw particles for an additional 30 cycles. EMAN (59) was used for all 3D reconstructions. The resolutions of the final models were determined using a Fourier shell correlation (FSC) cutoff of 0.5.

The cryo-EM model of the BG505 SOSIP.664 trimer complex with PGV04 (PDB code: 3J5M) was fit manually into the EM densities and refined by using the UCSF Chimera “Fit in map” function (60).

Reconstruction data codes.

The reconstruction data reported in this paper have been deposited in the Electron Microscopy Data Bank (http://www.emdatabank.org) (EMDB ID code EMDB-6256).

RESULTS

Biochemical and biophysical characterization of B41 SOSIP.664 trimers.

The B41 env gene, previously described as 9032-08.A1.4685, is derived from a subtype B founder virus (Fiebig stage III) isolated from a serial plasma donor (46). We introduced the sequence changes necessary to create SOSIP.664 trimers (see Materials and Methods) and transiently transfected the modified env gene into 293T cells, cotransfecting the furin gene to maximize cleavage of gp120 from gp41_ECTO (4, 5). In pilot experiments (data not shown), we used a combination of endpoints to judge that the B41 SOSIP.664 construct had suitable properties for additional studies, specifically, high-level Env expression (SDS-PAGE gels), efficient trimer formation (BN-PAGE gels), and a high percentage of native-like trimers (NS-EM).

For more detailed studies, the B41 SOSIP.664 construct was expressed transiently in 293T cells under standard conditions that involved the use of culture media containing 10% FCS. The secreted Env proteins were first affinity purified using the 2G12 bNAb, followed by SEC via a Superdex 200 26/60 column to isolate trimers (Fig. 1). A BN-PAGE analysis of the B41 SOSIP.664 Env proteins eluted from the 2G12 column showed that >40% were trimers, while dimers, monomers, and aggregates were each present at relative abundances of ∼20% (Fig. 1A). The SEC column removed the unwanted Env forms, yielding pure (>95%) trimers (Fig. 1B and C). A comparative reducing versus nonreducing SDS-PAGE gel analysis showed that the purified trimers were fully (>95%) cleaved into their gp120 and gp41_ECTO subunits (data not shown). These results broadly mirror our experiences with BG505 SOSIP.664 trimers (38). However, in contrast to our experiences with BG505, we observed some degradation products when the B41 SOSIP.664 trimers were analyzed by reducing SDS-PAGE followed by Coomassie blue staining or Western blotting (Fig. 1D). The anti-gp120 MAb ARP3119 detected a band of ∼70 kDa that is characteristic of a degradation event arising when the V3 region of gp120 is proteolytically clipped at a scissile site near the tip of its β-hairpin loop (42–45). The other gp120 fragment generated by V3 clipping, a band of ∼50 kDa, is not recognized by MAb ARP3119. This fragment and other less prominent bands (i.e., other degradation products) were detected when an HIV-Ig (subtype B) preparation was used to probe the blots (data not shown). Of note is that the extents of V3 clipping varied between B41 SOSIP.664 trimer preparations but never exceeded more than ∼25%, as judged by the relative intensities of the 70-kDa and 120-kDa bands in semiquantitative assessments. In contrast, monomeric B41 gp120 proteins were much more extensively V3 clipped (usually >50%) when produced and analyzed under the same conditions (Fig. 2C). Compared to other subtypes, Env proteins from subtype B are known to be generally quite vulnerable to V3 clipping, because a scissile site for thrombin family proteases (↓) is present in their consensus V3 sequence (GPGR↓AF) that is absent from the corresponding non-B consensus sequence (GPGQAF) (42–45). Hence, the subtype A BG505 SOSIP.664 trimers are not vulnerable to V3 clipping (Fig. 1D) (49).

FIG 1.

Biochemical characterization of B41 SOSIP.664 trimers. The env and furin genes were expressed in 293T cells in the presence of 10% FCS. (A) Env proteins purified on a 2G12 affinity column were analyzed on a Coomassie blue-stained BN-PAGE gel. The bands corresponding to aggregates, trimers, dimers, and monomers are indicated. The molecular mass marker proteins were thyroglobulin (669 kDa) and ferritin (440 kDa). (B) The same Env proteins were analyzed by SEC on a Superdex 200 26/60 column. (C) Coomassie blue-stained BN-PAGE analysis of B41 SOSIP.664 trimers purified via 2G12/SEC columns. (D) BG505 and B41 SOSIP.664 trimers were produced in 293T cells in the presence of 10% FCS, purified via 2G12/SEC columns, and then analyzed by reduced SDS-PAGE followed by Western blotting with MAb ARP3119. The 70-kDa fragment of gp120 that is characteristic of V3 clipping is indicated with an arrow.

FIG 2.

Preventing proteolytic clipping of B41 SOSIP.664 trimers. (A) The trimers were produced in 293F cells in serum-free medium and treated with thrombin (100 μg/ml) for the specified times before analysis by reducing SDS-PAGE and Western blotting with MAb ARP3119. (B) B41 SOSIP.664 trimers were produced either transiently in 293T cells in the presence of 10% FCS, in stable CHO cell lines in the presence of 1% FCS, or transiently in 293F cells under serum-free conditions. Trimers were purified via the 2G12 affinity column and SEC and then analyzed by SDS-PAGE and Western blotting with MAb ARP3119. (C) B41 gp120 monomers or SOSIP.664 trimers were produced in 293T cells in the presence of 10% serum, purified via either a 2G12 or a PGT145 affinity column, and analyzed by reducing SDS-PAGE and Western blotting with MAb ARP3119. The leftmost lane shows molecular mass markers. In all three panels, the 70-kDa fragment of gp120 that is characteristic of V3 clipping is indicated with an arrow.

Production of non-V3-clipped B41 SOSIP.664 trimers in low-serum medium.

To eliminate, or at least minimize, V3 clipping, we first explored different culture conditions for B41 SOSIP.664 trimer production. An alternative approach, based on the use of a PGT145 affinity column to positively select nonclipped trimers, is described below. As noted above, V3 clipping is commonly mediated by tryptic proteases of the thrombin family. Prior studies have shown that the most common source of these proteases is the FCS component of culture media (44, 45). While an alternative source may be proteases released from cells that are damaged or killed during transient transfection, we noted that our initial clones of stably transfected CHO and 293T cells, cultured in 10% FCS, also yielded trimers that were ∼20% V3 clipped in the absence of significant cell damage or death (data not shown). We therefore produced B41 SOSIP.664 proteins by transient transfection of 293F cells in serum-free medium. The resulting trimers, purified by 2G12/SEC, were fully intact with no evidence for V3 clipping in a Western blot analysis, implying that FCS was indeed the source of the proteolytic activity (Fig. 2A). When the nonclipped, 293F cell-derived trimers were treated with thrombin, a subpopulation (∼20%) became clipped (Fig. 2A). Hence, 293F cells do not produce Env variants that are intrinsically resistant to V3 clipping, but rather it is their ability to be cultured in serum-free medium that yields nonclipped B41 gp120 monomers and SOSIP.664 trimers.

To facilitate the production of B41 SOSIP.664 trimers on a larger scale, we made stable 293T and CHO lines using procedures previously described for the corresponding BG505 trimers (49). We then attempted to adapt these lines to growth in media containing lower serum concentrations, and we succeeded with the CHO line (see Materials and Methods). The B41 SOSIP.664 trimers produced when this line was cultured in 1% serum were only minimally (<5%) V3 clipped (Fig. 2B). We expect that we will eventually be able to adapt the CHO line to serum-free culture conditions, which should completely eliminate all V3 clipping.

Neutralization of the parental B41 virus.

To gain an understanding of the antigenicity of B41 Env, we first assessed the ability of MAbs against multiple epitopes to neutralize the B41 Env-pseudotyped virus in a TZM-bl cell-based assay. The Duke University Central Laboratory has independently classified this virus as having a tier 2 neutralization profile (46). The bNAbs tested targeted CD4bs (b12, VRC01, and PGV04) as well as the CD4-IgG2 protein, the N332 glycan-dependent V3 cluster (PGT121 and PGT126), the N332 glycan-dependent outer domain cluster (PGT135 and 2G12), the quaternary structure-dependent V1V2 epitopes (PG9, PG16, and PGT145), and the gp120-gp41 interface (PGT151, 35O22, and 8ANC195). bNAbs to the membrane-proximal external region (MPER) were not included in the test panel, as this region is not present in the B41 SOSIP.664 construct. With the exception of 35O22, all tested bNAbs neutralized the B41 virus, although with a wide range of potencies (Table 1).

TABLE 1.

Relationship between MAb binding to B41 SOSIP.664-D7324 trimers in ELISA and neutralization of the B41 Env-pseudotyped virus^a

Epitope and antibody group	Antibody	IC50 (μg/ml)	EC50 (μg/ml)
V1/V2 glycan, bNAb	PG9	0.56	0.49
	PG16	0.08	0.21
	PGT145	0.06	0.20
gp120-gp41ECTO interface, bNAb	PGT151	0.30	>20
	8ANC195	1.89	5.68
	35O22	>50	>20
CD4bs, bNAb	CD4-IgG2	8.49	>20
	VRC01	1.38	1.04
	b12	11.3	1.17
	PGV04	0.93	8.26
CD4bs, non-NAb	b6	>50	>20
	F91	>50	>20
	F105	>50	>20
V3 glycan, bNAb	PGT121	0.92	0.45
	PGT126	1.89	2.19
OD-glycan, bNAb	2G12	12.8	0.22
	PGT135	20.3	4.83
V3, non-NAbb	447-52D	>50	0.03
	19b	>50	0.02
	14e	>50	0.15
	39F	>50	0.04
	CO11	>50	0.05
CD4i, non-NAb	17b	>50	>20
gp41ECTO, non-NAb	F240	>50	>20

^aMidpoint neutralization concentrations (IC₅₀) and half-maximal binding concentrations (EC₅₀) were calculated using the nonlinear regression function in Prism software version 5.0.

^bNote that V3 non-NAbs have a high affinity for SOSIP.664-D7324 trimers under ELISA conditions (35). However, as shown when 19b was studied by ITC, the stoichiometry of binding to nontagged B41 SOSIP.664 trimers is low (Fig. 4), which is consistent with previous observations made using BG505 trimers (38).

Several MAbs that are generally found to be nonneutralizing for tier 2 viruses did not neutralize B41 (IC₅₀ >50 μg/ml), including b6, F91, and F105 to CD4bs; 17b to a CD4-induced epitope; 447-52D, 39F, CO11, 19b, and 14e to V3; and F240 to gp41_ECTO (Table 1). We confirmed that their epitopes were present on at least one form of B41 Env protein (e.g., gp120 monomers or gp41_ECTO), showing that their inability to neutralize the virus was not due to a sequence-dependent lack of the epitope (data not shown, but see below).

Antigenicity of B41 SOSIP.664 trimers by ELISA.

For antigenicity studies by ELISA, we used D7324-tagged trimers and, for comparison, gp120 monomers. Both proteins were produced in 293F cells in the absence of serum and purified by 2G12 affinity chromatography and SEC. Thus, they were not significantly V3 clipped. The MAbs tested are listed in Table 1. The amounts of gp120 monomers and SOSIP.664-D7324 trimers captured onto the ELISA wells were calibrated to yield comparable 2G12 binding curves that serve to normalize the data derived using the other test MAbs (Fig. 3A).

FIG 3.

Comparative antigenicity of B41 trimers, B41 gp120 monomers, and BG505 trimers in ELISA. (A) Representative binding curves of the indicated MAbs to B41 SOSIP.664-D7324 trimers (closed circles) or gp120 monomers (open circles). Both antigens were produced in 293F cells in serum-free medium and purified using 2G12 and SEC columns. The extent of V3 clipping was <5%. (B) Reactivity of the indicated MAbs with B41 (open circles) and BG505 (closed circles) SOSIP.664-D7324 trimers under comparative conditions.

bNAbs such as PG9, PG16, and PGT145 against quaternary structure-dependent or influenced epitopes are valuable for gauging whether soluble trimers adopt an appropriate conformation. The relevant epitopes span the V1V2 domains of two gp120 protomers at the trimer apex (50). Although PG9 and PG16 can bind a small subset of monomeric gp120s or uncleaved gp140s, including ones based on the BG505 sequence, they do so rarely and then only weakly compared to their reactivity with native-like trimers (25, 38, 50). PGT145, in contrast, is completely specific for trimers (25, 38, 50). Under ELISA conditions, PG9, PG16, and PGT145 all bound well to the B41 SOSIP.664-D7324 trimer but only weakly or not at all to the corresponding gp120 monomer (Fig. 3A and data not shown). The trimers also presented the epitopes for all tested bNAbs to CD4bs (e.g., VRC01 and PGV04), the N332 glycan-dependent V3 cluster (PGT121 and PGT126), and the N332 glycan-dependent outer domain cluster (2G12 and PGT135) (Fig. 3A and Table 1). One bNAb that neutralized the B41 Env-pseudotyped virus but did not bind to the corresponding trimers was PGT151 against a labile epitope at the interface between gp120 and gp41_ECTO (28, 29). However, a second bNAb against an adjacent epitope in the same interface, 8ANC195 (61), both neutralized the virus and bound the trimer efficiently (Table 1). A third member of the gp120-gp41_ECTO bNAb group, 35O22 (30), also failed to bind the trimer, but as it also did not neutralize the B41 virus, its epitope is probably absent due to sequence variation (Table 1).

Non-NAbs fail to neutralize HIV-1 because their epitopes are either absent or not accessible on a sufficient proportion of native trimers at the right time (62, 63). A soluble trimer that mimics these functional spikes should expose few or, ideally, no non-NAb epitopes (38). Accordingly, we assessed various non-NAbs for their abilities to bind B41 SOSIP.664-D7324 trimers and gp120 monomers. Compared to the monomers, most non-NAbs bound to the trimers much more weakly and, in several cases, not at all, with examples being F91 and b6 against CD4bs epitopes and 17b to a CD4-induced epitope (Fig. 3A and Table 1). The limited or lack of reactivity of these non-NAbs therefore indicates that their epitopes are indeed poorly accessible on the trimer due to native quaternary constraints. In contrast, the V3 MAbs 447-52D, CO11, F2A3, 39F, 19b, and 14e, which are generally non-NAbs for tier 2 viruses and did not neutralize B41, bound with high affinity to a subpopulation of the B41 SOSIP.664-D7324 trimers in ELISA, although much less so than to the gp120 monomer (Fig. 3A, Table 1, and data not shown). Hence, as is the case for the corresponding BG505 proteins (38), the V3 region of the B41 SOSIP.664-D7324 trimers becomes partially accessible for MAb binding under ELISA conditions while being inaccessible on the B41 virus. The accessibility of V3 to non-NAbs like 19b is, however, found to be much lower when trimer antigenicity is studied in other assay systems such as ITC (see below).

We compared the antigenicities of the B41 and BG505 SOSIP.664-D7324 trimers directly, using 2G12 binding to normalize the amounts of each trimer captured on the ELISA wells (Fig. 3B). Compared to BG505, the V3 non-NAb and the trimer-influenced or -specific bNAbs PGT145 and PG16 bound more strongly to the B41 trimers. Moreover, whereas the CD4bs bNAb VRC01 bound equivalently to the two trimers, b12 reacted only with the B41 version (Fig. 3B). The basis for the different reactivity profiles for these two CD4bs NAbs and the B41 and BG505 SOSIP.664 trimers is discussed further below (see Fig. 8). Finally, we confirmed in the comparative ELISA that PGT151 bound to the BG505 trimers but not B41 and that a second gp120-gp41_ECTO interface bNAb, 8ANC195, bound to both trimers, slightly better to B41 than BG505 (data not shown).

FIG 8.

NS-EM analyses of B41 SOSIP.664 trimers in complex with Fabs. (A) The left side shows top and side views of a 3D reconstruction (gray) of VRC01-liganded B41 trimers with the high-resolution model of PGV04-liganded BG505 trimers docked (PDB code 3J5M). The gp120 subunits are in blue, V1V2 is in magenta, V3 in green, gp41_ECTO in brown, and PGV04 in red. The right side shows the VRC01-liganded B41 trimer density (gray) superimposed on the volume map of the PGV04-liganded BG505 SOSIP.664 trimer (pink) (EMD-2427). (B) EM 2D class averages of VRC01-bound and b12-bound B41 SOSIP.664 trimers (left and right, respectively). (C) Projection matching (left) and Fourier shell correlation (right) for the VRC01-liganded B41 trimers. All images are of non-V3-clipped trimers produced in 293T cells and purified by the PGT145/SEC method.

Antigenicity of B41 SOSIP.664 trimers by ITC.

As noted above, V3 non-NAb epitopes are better displayed on D7324-tagged SOSIP.664 trimers under ELISA conditions than is the case when other antigenicity assays are used (25, 38). As we considered it possible that the same might apply to the trimer-dependent PGT145 epitope, we used ITC to further assess PGT145 and 19b binding to nontagged B41 SOSIP.664 trimers and, for comparison, their BG505 counterparts (Fig. 4). Antigen-antibody thermodynamic parameters derived from the ITC studies are summarized in Table 2.

FIG 4.

Antigenic analysis of B41 SOSIP.664 trimers by ITC. The top portions show the raw data and the bottom portions the binding isotherms for representative ITC binding experiments. Binding of B41 SOSIP.664 trimers to PGT145 Fab (A) and 19b IgG (B) was assessed. The thermodynamic parameters of binding are listed in Table 2. Although an IgG molecule was used in the calorimetry experiment, NS-EM imaging showed that it did not cause the trimers to aggregate under these experimental conditions.

PGT145 Fab bound with ∼7-fold-higher affinity to the B41 trimers than to BG505 but with a slightly lower stoichiometry (K_d = 9 nM and N = 0.4 versus K_d = 63 nM and N = 0.5) (Fig. 4). The thermodynamic parameters also differed in that Fab PGT145 binding to the BG505 trimers was more entropically favorable, whereas the B41 binding process was highly dominated by enthalpy. The explanation for these different thermodynamic contributions must await detailed structural insights into how the PGT145 paratope interacts with its epitope on the subtype A and subtype B SOSIP.664 trimers. The V3 non-NAb 19b also had an ∼7-fold-higher affinity for B41 SOSIP.664 trimers than for BG505 (K_d = 3 nM versus 21 nM) (Fig. 4). In both cases, however, the stoichiometry values for 19b binding were quite low (N = 0.2 versus 0.1 for B41 versus BG505). Thus, as previously reported, the V3 region can become exposed on a minor subpopulation of BG505 SOSIP.664 trimers (25, 38). We now show that this is also true of the corresponding B41 trimers.

An ITC experiment provided further confirmation that the PGT151 Fab did not bind to the B41 SOSIP.664 trimers, whereas it bound with modest affinity to the corresponding BG505 trimers (Table 2) (28). Overall, the ITC and ELISA data derived using PGT145, 19b, and PGT151 are consistent (Fig. 3, Fig. 4, and Table 2).

Correlation between B41 SOSIP.664 trimer antigenicity and virus neutralization.

We plotted the ELISA EC₅₀s for NAb and non-NAb binding to the B41 SOSIP.664-D7324 trimers (V3 nonclipped, from 293F cells) against the IC₅₀s for neutralization of the B41 Env-pseudotyped virus (Fig. 5). Taken together, the binding and neutralization data for the entire panel of MAbs did not correlate (r = 0.06, P = 0.78). As noted above, V3-directed MAbs are, however, known outliers under ELISA conditions (38). When the V3 non-NAbs were excluded, the resulting correlation was strong and significant (r = 0.69; 95% confidence interval = 0.33 to 0.87; P = 0.0011; n = 19). The corresponding values in a similar analysis of the BG505 SOSIP.664 trimers were an r of 0.82 and P value of <0.0001 (38). Overall, the B41 SOSIP.664 trimers broadly resemble their BG505 counterparts in their antigenic mimicry of native Env, albeit with some differences that are discussed further below.

FIG 5.

Correlation between neutralization of the B41 Env-pseudotyped virus and ELISA reactivity of B41 SOSIP.664-D7324 trimers. Midpoint neutralization concentrations (IC₅₀) were derived from single-cycle infections of TZM-bl cells with the B41 Env-pseudotyped virus. The values represent the averages of 2 to 5 independent titration experiments, each performed in duplicate. Half-maximal binding concentrations (EC₅₀) were derived from D7324-capture ELISAs using B41 SOSIP.664-D7324 trimers prepared in 293F cells and purified using 2G12 and SEC columns (V3 clipping was <5%). The EC₅₀s for MAb binding to the B41 SOSIP.664-D7324 trimers (y axis) were plotted against the IC₅₀s for neutralization of the B41 Env-pseudotyped virus (x axis). When accurate midpoint concentrations could not be calculated because of lack of binding or neutralization, the highest concentration tested was included in the correlation analysis (i.e., when the IC₅₀ of neutralization was >50 μg/ml, a value of 50 μg/ml was used). Spearman's correlation coefficient, r, was calculated using Prism software version 5.0. V3 non-NAbs excluded from the statistical analysis are shown enclosed in a dashed oval. Identical values obtained with multiple NAbs (EC₅₀ > 20 μg/ml; IC₅₀ > 50 μg/ml) are plotted as a single data point that is indicated by an arrow. These non-NAbs are F91, F105, b6, F240, 17b, and 35O22. A different arrow highlights the PGT151 bNAb that neutralizes the B41 virus but does not bind the SOSIP.664-D7324 trimers.

Removal of V3-clipped B41 SOSIP.664 trimers via a PGT145 bNAb affinity column.

The trimer apex is formed by an association between the V2 and V3 variable regions and is recognized by the PGT145 bNAb, which is exquisitely specific for native-like, closed trimers (25, 38, 50). Thus, PGT145 does not bind gp120 monomers, gp120-gp41_ECTO protomers, or nonnative gp140_UNC proteins (25, 38, 50). We therefore made a PGT145 affinity column to test the hypothesis that V3 clipping damages the variable loop-dependent structures at the trimer apex and that, accordingly, clipped trimers would flow through the column. Moreover, because of the selectivity for PGT145 for native trimers, we were also interested in evaluating whether any nonnative Env forms present (monomers, dimers, aggregates, uncleaved gp140s, etc.) would fail to bind to the PGT145 column, flow through, and be discarded. A one-column, single-step purification method that eliminated the need for a later SEC step could create substantial efficiency benefits if and when production of this and other trimers is scaled up.

When B41 SOSIP.664 Env proteins were produced in 293T cells in the presence of 10% serum and then passed down the PGT145 column, the bound and then eluted trimers were not detectably V3 clipped (Fig. 2C). This outcome contrasts with the purification of the same 293T cell-produced proteins via the 2G12 and SEC columns, which yielded trimers that were ∼25% V3 clipped (Fig. 1D and 2C). Thus, by inference, V3 clipping does indeed destroy the PGT145 epitope at the trimer apex, allowing the use of the PGT145 column to positively select for non-V3-clipped trimers. Moreover, as we had hypothesized, PGT145 column-purified trimers were found to be essentially homogeneous (i.e., free of contaminating Env aggregates, dimers, or monomers) when analyzed by BN-PAGE or SEC (Fig. 6A). Thus, in principle, it may be possible to eliminate the use of SEC to purify B41 and other SOSIP.664 trimers, but in the experiments described below, we elected to purify the trimers using both the PGT145 and SEC columns.

FIG 6.

Purification of B41 SOSIP.664 trimers using a PGT145 affinity column. (A) Analysis of the PGT145 column eluate by SEC on a Superdex 200 26/60 column. (B) Representative binding curves of the indicated MAbs to B41 SOSIP.664-D7324 trimers purified either by the 2G12 affinity column followed by SEC (closed circles) or by the PGT145 affinity column (open circles). In both cases the trimers were produced in 293F cells in serum-free medium, and the extent of V3 clipping was <5%.

We purified B41 SOSIP.664-D7324 trimers by the PGT145/SEC method and analyzed them by ELISA. Compared to non-V3-clipped trimers purified by the 2G12/SEC method, 2G12, VRC01, PG16, and PGT145 each bound indistinguishably, whereas the binding of the non-NAbs b6 and F91 to their CD4bs epitopes was modestly reduced (Fig. 6B). It is possible that the PGT145 column eliminates a very minor subset of nonnative B41 trimers that is present in preparations purified by the 2G12/SEC method or that arises over time.

Negative-stain EM reveals two subpopulations of native-like B41 SOSIP.664 trimers.

We used NS-EM to study the overall morphology of 2G12/SEC-purified, 293T cell-derived (i.e., partially V3-clipped) B41 SOSIP.664 trimers (Fig. 7A). The reference-free 2D class averages revealed that almost all (>95%) of the trimers adopted a native-like morphology with the three protomers closely associated. However, a significant subset of B41 trimers (∼35 to 75% in different preparations) displayed a subtly different, more open conformation characterized by the presence of a small, circular density at the distal end of one or more protomers (Fig. 7A). That feature distinguishes this subset of B41 SOSIP.664 trimers from their BG505 counterparts, which are almost all in the closed conformation. Various control experiments validated that this feature was intrinsic to the trimers and not a result of stain thickness variation or misalignment (see Materials and Methods). We emphasize that the more open subpopulation is native-like, is distinct from the fully open, CD4-bound conformation, and is completely distinguishable from the nonnative, semidisintegrated proteins that dominate gp140_UNC preparations (20–22, 24, 25; Pritchard et al., submitted). Thus, the more open subpopulation retains a 3-fold symmetric appearance, with the primary difference being small satellite densities emanating from the vertices of the trimer. Semidisintegrated gp140_UNC proteins, on the other hand, yield no regular shapes indicative of an ordered structure. The implications of the presence of two subpopulations of native-like trimers are explored in more detail below.

FIG 7.

NS-EM analyses of unliganded B41 SOSIP.664 trimers. Shown are 2D reference-free class averages of B41 SOSIP.664 trimers produced in 293F cells and purified via 2G12/SEC showing the closed and more open subpopulations (A), B41 SOSIP.664-D7234 trimers produced in 293F cells (i.e., non-V3 clipped) and purified by 2G12/SEC (B), the same trimers as in panel B but after exposure to thrombin (C), B41 SOSIP.664 trimers produced in 293T cells and purified via the PGT145 column (i.e., V3 nonclipped) (D), B41 SOSIP.664 trimers produced in the stable CHO cell line and purified via 2G12/SEC (extent of V3 clipping < 5%) and analyzed immediately after a sample of the stock was thawed (E), and the same trimers as in panel E but after incubation for 1 h at 37°C with no annealing prior to grid preparation (F). The percentages of closed versus more open trimers were derived from the total number of particles belonging to each subpopulation and were not simply based on the number of classes representing each subpopulation (see Materials and Methods).

We considered the possibility that V3 clipping could cause the B41 trimers to open up. However, non-V3-clipped, 293F cell-produced B41 trimers were similar in appearance to partially clipped ones made in 293T cells, in that two subpopulations of native-like trimers, with the closed and more open conformations described above, were again visible (Fig. 7A and B). After exposure of the non-V3-clipped trimers to thrombin, which clips ∼20% of the V3 regions present (Fig. 2A), the proportion of the trimers in the closed conformation was reduced from ∼60% to ∼30% (Fig. 7B and C). However, variations in the proportions of open versus closed B41 trimers between preparations, and the influence of thermal effects and incubation and storage conditions (see below), preclude us from judging whether thrombin has a genuine effect on trimer conformation. When B41 SOSIP.664 trimers produced in 293T cells, and hence partially V3 clipped (Fig. 1D and 2C), were purified via the PGT145 column, which eliminates the V3-clipped subset, a mixture of closed and more open trimer forms was again present (Fig. 7D). The two conformations are, therefore, an intrinsic feature of the B41 trimers, independent of the purification method (2G12/SEC versus PGT145/SEC) and of whether or not the trimers are partially V3 clipped. We confirmed that the B41 SOSIP.664-D7324 trimers used in ELISA had an appearance comparable to that of their nontagged counterparts, in that the overall population was native-like, with both open and closed conformations present (data not shown).

We investigated whether the relative abundance of the closed and more open trimer classes is instead driven by thermodynamics. A freshly thawed sample of B41 SOSIP.664 trimers stored at 4°C was compared with one that was incubated for 1 h at 37°C and then immediately blotted onto an EM grid to preserve the equilibrium state of the sample. The number of more open classes increased from 52% to 84% during the hour at 37°C (Fig. 7E and F).

To further study their overall 3D-dimensional structure, we incubated the B41 SOSIP.664 trimers with a VRC01 Fab fragment and collected a series of NS-EM images (Fig. 8). The 2D class averages were similar in appearance to complexes of BG505 SOSIP.664 trimers with CD4bs bNAbs, and a majority of classes were fully saturated (3 Fabs per trimer) (Fig. 8B). A 3D reconstruction of the data set at an ∼17-Å resolution shows that the VRC01-liganded B41 SOSIP.664 trimer is compact and highly similar to the PGV04 complex with the corresponding BG505 trimer (EMDB 5779; correlation coefficient, 0.95) (Fig. 8A). Hence, like their BG505 counterparts, the B41 trimers have a high structural integrity.

We next probed these different forms of the B41 trimers using the b12 Fab, as this CD4bs antibody only, or preferentially, binds to more open trimers, whereas VRC01 and PGV04 efficiently recognize and stabilize the closed form (64, 65). Despite multiple attempts, we could obtain no evidence from NS-EM and ELISA studies that b12 is able to bind to BG505 SOSIP.664 trimers, whereas it reacts efficiently with BG505 gp120 monomers, gp120-gp41_ECTO protomers, and gp140_UNC proteins (Fig. 3B and data not shown) (21, 25, 38, 64). Thus, the b12 epitope is present on BG505 Env proteins but is not exposed on the compact, closed BG505 trimers (21, 25, 38, 64). In contrast, the 2D class averages generated after overnight incubation of B41 SOSIP.664 trimers with b12 yielded clear evidence for complex formation (Fig. 8B). The implication is that the B41 trimers are more structurally plastic than their BG505 counterparts in that they spontaneously, and reversibly, open up over time and, in doing so, expose the b12 epitope. Once b12 has bound, the trimer appears to open further, such that its gp120 subunits now resemble the CD4-bound conformation.

Thermal stability of B41 SOSIP.664 trimers assessed by DSC.

We used DSC to assess the thermal stability of B41 SOSIP.664 trimers produced in 293F cells and purified via the PGT145 and SEC columns. The resulting profile showed a single unfolding peak with a thermal denaturation midpoint (T_m) of 57.0°C (Fig. 9). 293F cell-expressed SOSIP.664-D7324 trimers, purified via the 2G12 and SEC columns for use in ELISA, had a very similar T_m value of 57.6°C, implying that the D7324 epitope tag does not affect their thermal stability. These T_m values are ∼10°C lower than the 68.1°C found for BG505 SOSIP.664 trimers (38). Overall, we conclude that the B41 trimers are relatively stable, albeit less so than their BG505 counterparts. The decreased stability is likely related to the transient structural remodeling at the trimer apex observed by NS-EM that was noted above. In other words, the B41 trimers may be less thermally stable than BG505 because of their greater propensity to transit between alternative conformations.

FIG 9.

Thermal stability of B41 SOSIP.664 trimers measured by DSC. B41 SOSIP.664 trimers were purified using the PGT145 and SEC columns. The non-two-state best-fit curve is depicted as a broken line. T_m, thermal denaturation midpoint temperature; T_onset, temperature at which the peak can be differentiated from the baseline; T_1/2, width at half peak height.

DISCUSSION

The high-resolution structures of BG505 SOSIP.664 trimers and their general antigenicity and stability properties have created interest in the use of these native-like Env proteins as potential components of an HIV-1 vaccine (38–41). Moreover, the evidence that uncleaved gp140 proteins predominantly adopt improperly folded, nonnative conformations has compromised the original rationale for their use as vaccine components (20–25; Pritchard et al., submitted). For SOSIP.664 trimers to be truly useful in the vaccine context does, however, require that the available inventory be expanded beyond BG505. Thus, it would only be possible to immunize with multiple different trimers, simultaneously or as a longitudinal, sequence-related series, if all the required proteins could be made to an appropriate standard. In addition, having access to trimers based on multiple genetic subtypes, particularly B and C, would be useful for structural studies. Here, we describe the identification of a high-quality subtype B trimer, B41 SOSIP.664.

Taken together, the antigenicity assays show that most MAbs that neutralize the B41 Env-pseudotyped virus efficiently also bind strongly to the soluble B41 SOSIP.664 trimers (wild type and/or D7324 tagged). One trivial exception applies to MAbs to MPER epitopes that are not present in the SOSIP.664 trimer construct, but a more noteworthy one involves the gp120-gp41_ECTO interface bNAb PGT151, which binds the B41 SOSIP.664 trimers poorly or not at all while neutralizing the B41 virus (Table 1). As PGT151 binds to BG505 SOSIP.664 trimers, this epitope can be successfully presented on soluble, native-like trimers (Fig. 3B and Table 2) (25, 28). A different gp120-gp41_ECTO interface epitope, for bNAb 8ANC195, was present on the B41 SOSIP.664 trimers, implying that the perturbation to the PGT151 epitope is quite localized; it may be associated with how these trimers sample different conformations, as discussed below. Moreover, we can fully restore the PGT151 epitope to B41 SOSIP.664 trimers via a mutagenesis approach conducted as part of ongoing studies intended to increase the stability of the BG505, B41, and other trimers; this work will be described elsewhere. As with the BG505 trimers, nonneutralizing V3 epitopes (e.g., for 19b) are accessible on the D7324-tagged B41 trimers under ELISA conditions, but an ITC study showed that the stoichiometry of 19b binding to nontagged B41 trimers was low. A minor subpopulation of B41 SOSIP.664 trimers, like their BG505 counterparts, has V3 regions exposed, perhaps when the trimers open up (see below). Overall, the antigenicity of the B41 SOSIP.664 trimers is generally comparable to that of their BG505 counterparts, with respect to both bNAb epitope exposure and non-NAb epitope occlusion. The stronger binding of the PG16 and PGT145 bNAbs to their quaternary epitopes at the apex, compared to their BG505 counterparts, suggests that these highly conformation-dependent epitopes are particularly well presented on the B41 trimers.

A complication of working with almost all subtype B Env proteins is V3 clipping by proteases, much of which is attributable to the action of thrombin family enzymes present in serum (42–45). Although Env proteins from other subtypes can sometimes be V3 clipped and/or further degraded by various proteases, the clipping problem is greatest with subtype B proteins such as B41, whose V3 consensus sequence GPGR↓AF contains a scissile site (42–45). We found that a minor subset of the gp120 subunits of B41 SOSIP.664 trimers produced in the presence of 10% serum were V3 clipped. Only a subset (∼25%) of the trimers ever became V3 clipped, either during production in the presence of serum or when deliberately exposed to thrombin. In contrast, B41 gp120 monomers were more extensively damaged, with clipping usually exceeding 50%. The transient opening of the closed form of the trimer (see below) may expose the V3 region to any proteases present in the cell culture, but we do not yet know why the damage is limited to a subset of trimers. The key point is that V3 clipping can be almost completely prevented by eliminating serum from the cell cultures (44, 45). Accordingly, we made the B41 SOSIP.664 trimers in 293F cells in serum-free medium, and in an ongoing process we have successfully adapted our stable CHO cell line to growth at low serum levels. As growth in low serum may, however, reduce trimer yields, we are also investigating the consequences of eliminating the scissile site by mutagenesis (e.g., GPGRAF to GPGQAF).

We show here that V3-clipped trimers can also be removed by passage down a PGT145 column, as the damage to the V3 region destroys the PGT145 epitope. A more general reason why we evaluated the PGT145 column is because a one-step purification process for native-like trimers would offer significant yield and time advantages. PGT145 has an exquisite specificity for native trimers, failing to recognize gp120 monomers, dimers, gp120-gp41_ECTO protomers, or nonnative gp140_UNC proteins. The PGT145 column was indeed able to affinity purify homogeneous, native-like B41 SOSIP.664 trimers by positive selection, which speaks to the likelihood that this type of column, and/or ones based on other trimer-specific bNAbs such as PGT151, will be useful at the preclinical level. Studies on the use of PGT145 columns for purifying other SOSIP.664 trimers are in progress, as it is possible that PGT145 binding or elution may have sequence-specific effects on trimer conformation. Whether an additional purification step, such as SEC, would always be required when producing trimers for human trials, for example, to remove non-Env contaminants, will need to be considered.

However expressed and purified, the B41 SOSIP.664 trimers differ subtly from their BG505 counterparts when viewed by NS-EM. Almost 100% of the BG505 trimers have identical, propeller-like structures, but in contrast, a much lower proportion (range, 35 to 75%) of the B41 trimers are in this form, the remainder having a more open conformation. We again emphasize that the more open subpopulation is native-like and distinguishable from the nonnative, semidisintegrated proteins that dominate gp140_UNC preparations (20–22, 24, 25; Pritchard et al., submitted). Furthermore, this partially open state is not the same as the fully open CD4-bound conformation that has been described previously (66, 67). We interpret our NS-EM data as showing movement in the V1V2 region, while the core of each gp120 subunit remains in the prefusion conformation. The presence of the closed and more open subpopulations of the B41 SOSIP.664 trimers likely represents a static snapshot of a conformational equilibrium that is shifted further toward open forms for B41 than for BG505. The greater reactivity of the more open B41 trimers with the b12 bNAb, compared to that of BG505, seen in this study is consistent with conclusions drawn from biophysical studies of how different CD4bs bNAbs interact with native trimers (64). Thus, whereas VRC01 reacts with the fully closed trimer (B41 and BG505), b12 can do so only once the trimer opens up, which explains why b12 binds very poorly to the BG505 trimers (64).

Whether the greater propensity for B41 trimers to open up reflects their lower T_m, compared to that of BG505, is something worth considering. In ongoing studies, we have recently made B41 SOSIP.664 trimers that are both fully closed and with substantially greater thermal stability (i.e., a much higher T_m value), both by targeted mutagenesis and via the use of different producer cells (data not shown). Hence, the two properties may indeed be connected. The concept that Env trimers can equilibrate or breathe between closed and open conformations was first developed almost 20 years ago based on the differing NAb resistance phenotypes of primary and cell line-adapted viruses (68–73). These old ideas have been validated recently by the application of biophysical techniques to membrane-associated and soluble Env proteins and by the use of the new generation of bNAbs in neutralization-based studies (64, 65, 74, 75). We believe that our findings with the B41 SOSIP.664 trimers reflect and reinforce this rapidly evolving concept.

Strategies to make new SOSIP.664 trimers, and arguably any native-like trimers of other designs, should take into account both the temperature dependence of bnAb binding and conformational transitions that may not be detected by binding assays such as ELISA. Thus, the reaction temperature influences antigen-antibody interaction kinetics, which may be particularly relevant to the B41 SOSIP.664 trimers. These highly dynamic proteins exist as primarily two interconverting subpopulations with an equilibrium position that is affected by the entropy of the system (i.e., the incubation temperature). Ideally, the transient opening of all native-like trimers should still preferentially expose bNAb epitopes while occluding non-NAb epitopes. For the B41 SOSIP.664 trimers, this seems the likely outcome, as the gp120 core and gp41_ECTO components likely remain in a conformation close to the prefusion form when closed trimers do open up. Thus, the conformational motions involved in trimer breathing are localized to the variable loops and the timescale of non-NAb epitope exposure may be too short to allow such antibodies to bind. Accordingly, the B41 SOSIP.664 trimers may reflect the conformational dynamics that take place with native, virion-associated Env (65). Indeed, they might even do so better than their more fully closed and thermally stable BG505 counterparts. However, even the apparently fully closed BG505 SOSIP.664 trimers probably also breathe and open up, although in a way that is difficult to capture by NS-EM. We surmise that the BG505 trimers flicker between the two conformations more rapidly than the B41 trimers and spend less time in the open form.

As noted above, it will be valuable for both immunogenicity and comparative structural studies to identify many more high-quality, native-like SOSIP.664 trimers similar to those based on the BG505 and, now, B41 sequences. From ongoing, structure-guided studies, we have obtained and are now applying knowledge of some critical sequence determinants that improve the formation of native-like trimers as well their antigenicity and stability. Specific sequence changes can, of course, be combined with bNAb positive-selection columns (e.g., PGT145) to further increase both the range and quality of SOSIP.664 trimers. At present, it is a matter of judgment what SOSIP-modified env genes are worth pursuing for trimer production. Relevant variables include the overall amount of Env protein produced and the fraction that is trimeric, which should both be as high as possible. However, how the purified trimers appear when viewed by NS-EM is critical information. Our practice is to discard any SOSIP.664 candidate that yielded a substantial proportion of nonnative, misfolded, or semidisintegrated trimers with an appearance similar to that of gp140_UNC proteins when viewed by NS-EM. We have found that negative-selection columns using non-NAbs can remove such nonnative Env proteins from some SEC-purified SOSIP.664 preparations and thereby enrich the percentage of native-like trimers. Our concern is that the initial presence of abundant nonnative Env proteins is a hallmark of trimer instability. We also regard melting temperature as an important parameter; a thermal stability profile with multiple transitions or with a high value for the width at half peak height (T_1/2) is undesirable, as these parameters are probably indicative of unwanted heterogeneity.

The immunogenicity of the B41 SOSIP.664 trimers is now being investigated, as is that of stabilized (i.e., more closed, BG505-like) variants that we will describe elsewhere. The comparison should allow us to determine whether the equilibrium between closed and more open forms of B41 trimer influences immunogenicity. The resulting information will further guide our long-term goal of inducing bNAbs via the use of native-like trimers.