Influences on Trimerization and Aggregation of Soluble, Cleaved HIV-1 SOSIP Envelope Glycoprotein

Abstract

We describe methods to improve the properties of soluble, cleaved gp140 trimers of the human immunodeficiency virus type 1 (HIV-1) envelope glycoproteins (Env) for use in structural studies and as immunogens. In the absence of nonionic detergents, gp140 of the KNH1144 genotype, terminating at residue 681 in gp41 (SOSIP.681), has a tendency to form higher-order complexes or aggregates, which is particularly undesirable for structure-based research. We found that this aggregation in the absence of detergent does not involve the V1, V2, or V3 variable regions of gp120. Moreover, we observed that detergent forms micelles around the membrane-proximal external region (MPER) of the SOSIP.681 gp140 trimers, whereas deletion of most of the MPER residues by terminating the gp140 at residue 664 (SOSIP.664) prevented the aggregation that otherwise occurs in SOSIP.681 in the absence of detergent. Although the MPER can contribute to trimer formation, truncation of most of it only modestly reduced trimerization and lacked global adverse effects on antigenicity. Thus, the MPER deletion minimally influenced the kinetics of the binding of soluble CD4 and a CD4-binding site antibody to immobilized trimers, as detected by surface plasmon resonance. Furthermore, the MPER deletion did not alter the overall three-dimensional structure of the trimers, as viewed by negative-stain electron microscopy. Homogeneous and aggregate-free MPER-truncated SOSIP Env trimers are therefore useful for immunogenicity and structural studies.

INTRODUCTION

One of the most substantial obstacles to the development of an effective vaccine to prevent infection by human immunodeficiency virus type 1 (HIV-1) is our collective inability to design immunogens that are able to induce broadly active neutralizing antibodies (bNAbs) at sufficient titers (1–4). However, NAbs do invariably target the envelope (Env) glycoprotein complexes that are present as spikes on the surface of virions, but often in a strain-specific manner. The Env spikes mediate virus-cell attachment and fusion, processes that are prevented by NAb occupancy. Hence one rational vaccine design strategy for bNAb induction involves the use of recombinant versions of Env spikes as immunogens.

The Env spike is a trimer, each subunit containing a gp120 surface glycoprotein linked noncovalently to a gp41 transmembrane glycoprotein. The three gp120/gp41 protomers are also noncovalently associated, predominantly via their gp41 components but with additional contributions from gp120-gp120 interactions near the spike apex (5–8). The native Env spike must undergo a complex series of conformational changes, triggered by receptor interactions, to fulfill its fusion functions. Accordingly, the intersubunit bonds are fairly weak, and the spikes can spontaneously decay or otherwise lose function over time (9–11). Recombinant Env proteins are often expressed in soluble form, as gp120 or as gp140, which lacks the transmembrane and cytoplasmic domains of gp41 (12–18). However, trimers of gp140 are labile: unless stabilizing mutations are added, they tend to disintegrate into constituent subunits. Two fundamentally different approaches have been taken to solve this problem. The most commonly used method involves elimination of the cleavage site between the gp120 and gp41 ectodomain (gp41_ECTO) subunits, thereby preventing the dissociation of the subunits of the heterodimer from each other, while also promoting trimer integrity by unknown mechanisms (12, 14, 15, 19–22). The resulting uncleaved gp140s are often further stabilized by the addition of exogenous trimerization domains, such as foldons, at the C terminus (12, 19, 21). Our unpublished results show that purified uncleaved trimers, irrespective of the genotype or specific design, predominantly adopt aberrant structures that do not resemble native Env spikes. An alternative strategy for soluble trimer design involves stabilizing fully cleaved gp140s via a disulfide bond between gp120 and gp41_ECTO (created by appropriately positioned Cys substitutions) and a mutation (I559P) that strengthens the trimerization of the gp41_ECTO moieties (17, 23). The disulfide bond, when introduced into membrane-associated gp140, is compatible with Env function, in that the mutant Env protein interacts with CD4 and coreceptors and mediates viral entry provided the disulfide bond is reduced at the appropriate time (24–27). However, the I559P change blocks fusion by preventing the refolding of gp41. The soluble cleaved trimers are designated SOSIP gp140s and are the focus of the present study. We have previously shown and also demonstrate in the accompanying article that these cleaved trimers closely resemble virion-associated Env spikes when viewed by negative-stain electron microscopy (EM) (5, 28–31).

There are two major reasons to make soluble gp140 trimers: immunogenicity trials and structural studies. Multiple immunogenicity experiments in animals have shown that gp140 trimers are modestly superior to the corresponding gp120 monomer at NAb induction but not to an extent that seems likely to solve the overall vaccine design problem (12, 13, 18, 22). The solution here may come from having additional knowledge of trimer structure, generated by either X-ray crystallography or high-resolution EM. For these techniques to be successful, soluble trimers need to be highly pure and homogenous. Aberrant forms of Env, including but not limited to aggregates, compromise the potential of both methods. Here, however, we describe a way to improve the homogeneity and general properties of SOSIP gp140 trimers, by truncating the hydrophobic membrane-proximal external region (MPER) from the C-terminal end of gp41_ECTO. The resulting, slightly truncated, trimers are designated SOSIP.664, whereas those containing all of the MPER are termed SOSIP.681. We also report on trimers from which the V1/V2 and V3 variable regions of the gp120 subunits have been deleted without compromising cleavage or trimerization. To compare the trimers, we used surface plasmon resonance (SPR) to measure the rate and extent of monoclonal antibody (MAb) and soluble CD4 (sCD4) binding. The reactivities of the bNAb PGV04, directed against the CD4-binding site (CD4bs), and sCD4 were somewhat greater with SOSIP.664 than SOSIP.681 gp140 trimers. In an accompanying article, we use negative-stain EM to further study the full-length, MPER-truncated, and V1/V2- and V3-deleted KNH1144 SOSIP gp140 trimers and their reactivity with sCD4 and the PGV04 bNAb (29).

MATERIALS AND METHODS

Constructs.

The full-length SOSIP gp140 based on the KNH1144 subtype A sequence has been previously described (32, 33). Here, that protein construct is named SOSIP.681 gp140. The truncated gp140 from which 17 residues, including most of the MPER, have been deleted is designated SOSIP.664. The previously described KNH1144 SOSIP.681 ΔV1V2.9VE construct (34, 35), in which amino acids 128 to 194 are replaced by Asp-Asn-Gly and Val-120 is replaced by Glu (V120E), was used as a template for making three different changes in the V3 region: (i) in mutant 1, residues 301 to 323 in V3 were deleted, (ii) in mutant 2, they were substituted for by a 6-residue polybasic linker, and (iii) in mutant 3, there was instead a poly-Gly-Ser linker, as described previously (36). Mutant 1 was selected for further studies on the basis of its higher degree of cleavage between gp120 and gp41_ECTO (see Results) and was termed SOSIP.681.ΔV1V2V3. The same deletions were made in SOSIP.664, yielding SOSIP.664.ΔV1V2V3. In addition, we generated a number of C-terminal truncation variants—SOSIP.680, SOSIP.679, SOSIP.678, and SOSIP.677—in the context of both the full-length and ΔV1V2V3 variants. All of the gp140 constructs were subcloned in the pPPI4 vector system (17). Single-amino-acid substitutions and deletions were made by using the QuikChange mutagenesis kit (Stratagene, La Jolla, CA). All of the above gp140s contain a hexa-arginine (R6) motif at the C terminus of gp120 that increases the efficiency of gp120-gp41 cleavage, the A501C and T605C substitutions that allow a disulfide bridge between gp120 and gp41_ECTO, as well as the trimer-stabilizing I559P substitution (17, 23, 33, 37). We also generated two variants in which the I559P substitution was reverted, yielding SOS.681 and SOS.681.ΔV1V2V3. The full-length SOSIP.681 sequence had the amino acid substitutions A662E, G664D, T671N, S676T, and N677K, creating the epitopes for monoclonal antibodies (MAbs) 2F5 and 4E10.

Reagents.

HIV Ig was obtained through the AIDS Research and Reference Reagent Program (ARRRP), Division of AIDS, NIAID, NIH. MAb 2G12 was provided by Hermann Katinger through the ARRRP. MAbs VRC01 and PGV04 were gifts from Dennis Burton (Scripps Research Institute, La Jolla, CA), MAb 412d was from James Robinson (Tulane University, New Orleans, LA), and MAb 1-863 was from Michel Nussenzweig (The Rockefeller University, New York, NY). The CD4-IgG2 and sCD4 proteins from Progenics Pharmaceuticals have been described elsewhere (38) and were a gift from William Olson. MAbs from the PGT series were obtained through the IAVI Neutralizing Antibody Consortium (NAC) repository.

Cell culture and transient transfection.

Small- and large-scale expression was performed in the HEK293T cells. 293T cells were transiently transfected with env genes by using linear polyethylenimine (PEI) (molecular mass, 25 kDa) as described previously (39). Briefly, DNA encoding Env proteins was diluted in Dulbecco's modified Eagle's medium (DMEM) (Invitrogen, Carlsbad, CA) to one-tenth of the final culture volume and mixed with PEI (0.15 mg/ml). After incubation for 20 min, the DNA-PEI mixture was added to the cells for 4 h before replacement with culture medium containing 10% fetal bovine serum (FBS; Invitrogen), penicillin, streptomycin, and minimal essential medium (MEM) nonessential amino acids (0.1 mM; Invitrogen). Supernatants were harvested 48 h after transfection. All protein samples were kept on ice prior to subsequent analyses.

Purification of KNH1144 SOSIP gp140 trimers.

Trimers of all designs were purified by 2G12 affinity chromatography as described elsewhere (40). Purified proteins were stored at −80°C and thawed only once before use. Protein concentrations were determined by a bicinchoninic acid (BCA)-based assay (Thermo Scientific, Rockford, IL). 2G12 affinity-purified trimer variants were fractionated by using a Superose-6 10/30 column (GE Healthcare, NJ) and an AKTA Avant fast protein liquid chromatography (FPLC) system (GE Healthcare, NJ).

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

Env proteins were detected by SDS-PAGE and Western blotting procedures involving the cross-reactive anti-gp120 MAb ARP3119/C13 recognizing the sequence EDIISLW in C1 (0.2 μg/ml; Programme EVA Centre for AIDS Reagents) or the anti-gp41 MAb D50, horseradish peroxidase (HRP)-labeled goat anti-mouse IgG (1:5,000 dilution), and the Western Lightning ECL enhanced chemiluminescence system (PerkinElmer, Groningen, The Netherlands). Blue native (BN)-PAGE was carried out with minor modifications to a published method (41). Purified proteins or cell culture supernatants were diluted with an equal volume of buffer [100 mM 4-(N-morpholino) propanesulfonic acid (MOPS), 100 mM Tris-HCl (pH 7.7), 40% glycerol, 0.1% Coomassie blue], immediately prior to loading onto a 4 to 12% Bis-Tris NuPAGE gel (Invitrogen). The gels were run for 2 h at 150 V (∼0.07 A) with 50 mM MOPS–50 mM Tris (pH 7.7) as the buffer.

Immunoprecipitation assays.

Twenty-fold-concentrated supernatants from transiently Env-transfected 293T cells were incubated overnight at 4°C with MAbs or CD4-IgG2 in 500 μl of radioimmunoprecipitation assay (RIPA) buffer (150 mM NaCl, 1 mM EDTA, 50 mM Tris-HCl [pH 7.0]). A 50-μl volume of protein G-coated agarose beads (Pierce/Thermo Fisher Scientific, Rockford, IL) was then added, with rotation mixing, for 2 h at 4°C. The beads were washed extensively with ice-cold RIPA buffer containing 0.01% Tween 20 before proteins were eluted by heating for 5 min at 100°C in 50 μl of SDS-PAGE loading buffer supplemented with 100 mM dithiothreitol (DTT). The immunoprecipitated proteins were analyzed on 8% SDS-PAGE gels (125 V, 2 h).

SPR.

SPR analysis of binding was performed at 25°C on a Biacore 3000 instrument (GE Healthcare). Trimers were immobilized to CM5 chips by amine coupling, reaching preset ligand immobilization level (R_L) values of 485 to 630 response units (RU). Control channels were created by the same treatment in the absence of trimer. Association was recorded during a 300- to 400-s period and dissociation for 600 s. We regenerated the interaction surface after each analyte exposure by one injection of 10 mM glycine·HCl (pH 3.0) at 75 μl/min for 1 min; then the chip was equilibrated in running buffer (30 μl/min, 5 min) before the next analyte injection. Nearly complete analyte dissociation and intact binding capacity were achieved by each regeneration, as demonstrated by the absolute and relative response plots for the ensemble of cycles. Analyte diluted to various concentrations (2 to 600 nM) in running buffer (10 mM HEPES [pH 7.4], 150 mM NaCl, 3 mM EDTA, 0.005% P20) was passed over trimer-containing and control surfaces at a flow rate of 30 μl/min.

The MAb-trimer interaction was largely kinetically limited. Thus, increasing the flow rate did not enhance the response (whereas reducing it to <20 μl/min lowered it); the ln(dy/dx) plots for the association phase were linear with negative slopes; the local and global fits largely agreed; the T values were in the range 10 to 400 for the initial interaction parameters, and accommodation for mass transport limitation when modeling the data did not improve the fitting.

Both control-channel and zero-analyte responses were subtracted before fitting, which was performed with local R_max and local refractive-index shifts. None of the binding curves fitted well with a simple Langmuir model. Two different kinetic models in BIAevaluation (GE Healthcare) were chosen according to goodness of statistical parameters and the molecular properties of the analytes. Because of the known conformational changes in Env induced by CD4, a two-state, conformational-change model was used for the fitting of sCD4 binding data. The fidelity was increased compared with a Langmuir model, and the model was further validated by varying the duration of the association phase (contact time): increasing the contact time from 75 to 175 s resulted in a 22-fold reduction in apparent dissociation constant [K_d(App)]; increasing it from 175 to 750 gave a 2.6-fold reduction. For IgG binding, a bivalent model gave superior fits. This model allows for the binding of two paratopes inter- or intratrimerically but without effects on the affinity of any induced conformational changes. All three models (Langmuir, conformational change, and bivalent) were applied in their basic forms as included in the BIAevaluation software.

Electron microscopy.

Samples of SOSIP.664 or SOSIP.681 trimers containing 0.025% (wt/vol) CYMAL-7 detergent were treated with Bio-Beads (Bio-Rad, CA) for 2 h prior to grid preparation to remove excess detergent. Negative-stained grids were prepared as previously described (28). Briefly, a 0.1-mg/ml solution of the purified SOSIP gp140 construct (liganded or not) was applied to a freshly glow-discharged, carbon-coated grid and stained with 2% uranyl formate. Grids were viewed by using an FEI Tecnai TF20 electron microscope operating at a magnification of 100,000× and a high tension of 120 kV. Images were acquired on a Gatan 4,000-by-4,000 CCD camera at a defocus range of 600 to 720 nm and a dose of less than 16 e⁻/Ų by using the Leginon system (42). The pixel size of the CCD camera was calibrated at this magnification to be 1.09 Å by using a two-dimensional (2D) catalase crystal with known cell parameters. For more details, see the accompanying article (29).

Data collection and image classification.

Particles were automatically selected from micrographs with DoG Picker (43). Contrast transfer functions (CTF) for the untilted and tilted micrographs were estimated with ctffind3 and ctftilt (44). Particles were binned by 4 (80-by-80 boxes), and reference-free 2D class averages were calculated by using the Sparx package (45). Class averages and their associated particles were discarded based on their morphology and quality. For more details, see the accompanying article (29).

RESULTS

KNH1144 SOSIP.681 gp140 trimers require detergent to prevent aggregation.

Our short-term goal was to make homogeneous, monodisperse, cleaved, soluble Env trimers for structural and vaccine studies. We have previously shown that cleaved, soluble SOSIP.681 gp140 trimers based on the subtype A strain KNH1144 can be produced efficiently in 293T cells (33). However, in follow-up experiments, we detected the presence of aggregated gp140 species when immunoaffinity-purified SOSIP.681 gp140 trimers were purified and concentrated in the absence of detergent (40). To study this phenomenon in more detail, we analyzed the same proteins by analytical size exclusion chromatography (SEC), using a Superose 6 column (Fig. 1A). The eluted proteins were then fractionated and analyzed by BN-PAGE and Western blot analyses (Fig. 1B). The main peak, eluting at 14.9 ml, represented trimeric gp140, while a minor shoulder (∼16.6 ml) contained monomeric gp140. Only small amounts of dimers were present, in contrast to what was observed in similar studies with SOSIP.681 gp140s from the JR-FL (23) and other (33; data not shown) genotypes, but consistent with earlier analyses of KNH1144 SOSIP.681 gp140 (33, 40). A large and broad peak (12.2 ml) representing Env protein of high molecular mass appeared before the elution of the trimer. We have not studied the antigenicity or exact composition of these larger species, which could include aggregates of misfolded protein or higher oligomeric forms, such as multimerized trimers or dimers.

Fig 1.

KNH1144 SOSIP.681 gp140 forms aggregates. (A) 2G12 affinity-purified KNH1144 SOSIP.681 gp140 expressed in 293T cells was fractionated by SEC on a Superose 6 column at a flow rate of 0.5 ml/min. (B) The indicated column fractions were analyzed by BN-PAGE followed by Western blotting with the anti-gp120 MAb ARP3119. The molecular masses (kilodaltons) of marker proteins are shown on both sides. The elution positions of gp140 aggregates, trimers, and monomers are marked on both panels.

We reanalyzed the SEC-purified trimer fraction using BN-PAGE after concentration to 1 mg/ml. In the absence of detergent, a sharp band of aggregated protein was clearly visible (Fig. 2). A similar band was faintly visible higher up on the gel; these sharp bands are more suggestive of multiples of trimers than of nonspecific aggregation of various forms of Env. Of note is that, whatever their exact composition, addition of increasing concentrations of the nonionic detergent n-dodecyl-β-d-maltopyranoside (DDM) prevented the formation of the larger complexes of Env (Fig. 2). This finding is consistent with previous studies using Tween 20 (40). Thus, the purified KNH1144 SOSIP.681 gp140 trimers have a tendency to aggregate, which can be prevented by adding modest amounts of nonionic detergents.

Fig 2.

KNH1144 SOSIP.681 gp140 aggregates can be dissociated by nonionic detergents. SEC-fractionated gp140 trimers were incubated with the indicated concentrations (% wt/vol) of n-dodecyl-β-d-maltopyranoside (DDM) and before analysis by BN-PAGE followed by Coomassie blue staining. The positions where aggregates and trimers migrate are indicated, as are the molecular masses (kilodaltons) of marker proteins (left lane).

KNH1144 SOSIP.681 gp140 aggregate formation does not involve the V1, V2, or V3 regions of gp120.

The presence of detergent in gp140 preparations is undesirable for both structural and immunogenicity studies. We, therefore, investigated why gp140s tend to aggregate and how to prevent it. Elements in the V1/V2 and V3 variable regions have been implicated in the formation of nontrimeric Env forms, notably dimers (16, 46). Accordingly, we wanted to make a gp140 construct from which V1/V2 and V3 were deleted. We were aware that elimination of the variable loops from cleaved gp140s can be problematic (47). Apparently, loop deletion can uncover patches of hydrophobic residues that compromise protein folding (34, 35). However, virus evolution techniques yielded V1/V2 deletion variants that lacked any newly exposed hydrophobic patches (34, 35).

In this study, one of these variants, the KNH1144 SOSIP.681 ΔV1V2.9VE construct, in which amino acids 128 to 194 are replaced by Asp-Asn-Gly and Val-120 is replaced by Glu (V120E), was used as a template for further modifications (34). First, variants were made in which the outmost segment of V3 (residues 301 to 323) was either deleted entirely (mutant 1) or substituted for by 6-residue polybasic (mutant 2) or poly-GS (mutant 3) linkers, as described previously (36) (Fig. 3A). BN-PAGE analyses showed that all of the constructs trimerized efficiently (Fig. 3B), while SDS-PAGE revealed that the two gp140s in which V3 was replaced by linkers were slightly less efficiently cleaved at the gp120-gp41 junction despite having an optimal cleavage site (R6) (37). Thus, some residual uncleaved gp140 was present (Fig. 3C). The latter finding is consistent with previous observations of other loop-deleted SOS gp140 proteins (47). In contrast, mutant 1, from which residues 301 to 323 were deleted without substitution, was cleaved efficiently. We therefore selected the latter construct, termed SOSIP.681.ΔV1V2V3, for follow-up analysis.

Fig 3.

KNH1144 SOSIP.681 gp140 aggregate formation does not involve the V1, V2, or V3 regions of gp120. (A) The V3 sequences of the full-length (FL) KNH1144 SOSIP.681 gp140 construct and mutants 1, 2, and 3 (see the text) are shown. The dashes indicate the absence of amino acids from the specified positions. Note that FL SOSIP.681 contains all of the variable loops, including V1/V2, but they are deleted from variants 1, 2, and 3. (B) The four gp140s defined in panel A were analyzed by BN-PAGE followed by Western blot analysis with the anti-gp120 MAb ARP3119. The positions of trimers and monomers are indicated. Note the lower molecular masses of the three mutants compared with FL gp140. (C) The same four gp140s were also analyzed by SDS-PAGE followed by Western blot analysis with the anti-gp120 MAb ARP3119. (D) The SOSIP.681.ΔV1V2V3 (mutant 1) gp140 was purified by 2G12 affinity and fractionated by SEC on a Superose 6 column. The elution positions of aggregates, trimers, and monomers are marked. (E) The SOSIP.681 ΔV1V2V3 gp140 (mutant 1) was incubated in the presence of the indicated concentrations of the DDM detergent before analysis by BN-PAGE and Coomassie blue staining. The positions where aggregates and trimers migrate are indicated, as are the molecular masses (kilodaltons) of marker proteins (left lane).

The SOSIP.681.ΔV1V2V3 gp140 proteins were purified in 2G12 affinity columns and analyzed by SEC (Fig. 3D). The elution profile contained three distinct peaks at 13.1 ml, representing aggregates, 15.1 ml, representing trimers, and 16.8 ml, representing monomers. Note that these elution volumes are slightly greater than those for the full-length SOSIP.681 trimers (Fig. 1A), which is consistent with the reduction in size caused by deletion of V1/V2 and V3. Furthermore, the monomer peak was more prominent in the SOSIP.681.ΔV1V2V3 gp140 than its full-length counterpart (SOSIP.681), suggesting that the triple loop deletion has a minor adverse effect on KNH1144 SOSIP.681 gp140 trimerization. The SEC studies also showed that the full-length triple and loop-deleted forms of SOSIP.681 gp140s contained similar amounts of Env aggregates (compare Fig. 3D with Fig. 1A). We reanalyzed the SEC-purified trimer fraction using BN-PAGE after concentration to 1 mg/ml and again observed aggregates (Fig. 3E). However, when nonionic detergents (here, CYMAL-7, but also several others) were added, neither gp140 preparation contained aggregates (Fig. 3E) (data not shown). Together, these findings indicate that the V1, V2, and V3 loops are not involved in aggregate formation.

A detergent micelle associates with the MPER of KNH1144 SOSIP.681 gp140.

We investigated the purified KNH1144 SOSIP.681 gp140 by negative-stain EM (Fig. 4) (see below and the accompanying article [29]). The images showed that a large mass was present at the bottom of the trimer, i.e., around the MPER region and approximately where the viral membrane would be located in the context of a membrane-associated spike. The size of this extra density is consistent with that of a CYMAL-7 detergent micelle (48). Furthermore, we observed that several SOSIP.681 trimers could be attached to the same micelle, forming rosette-like structures (Fig. 4, circled examples). The MPER is located in the general area of the SOSIP.681 gp140 trimer where the detergent micelle forms, i.e., at or near the bottom of the gp41_ECTO. As the MPER contains hydrophobic residues that normally associate with the viral membrane (49–52), we hypothesized that the detergent micelle was forming around the MPER on the soluble trimers, as a substitute for interaction with membrane lipids. By extension, we considered it likely that the MPERs of different trimers might associate via hydrophobic interactions when detergents were absent, leading to aggregate formation.

Fig 4.

A detergent micelle associates with the MPER of KNH1144 SOSIP.681 gp140. A negative-stain EM micrograph shows KNH1144 SOSIP.681 proteins, purified in the presence of CYMAL-7 nonionic detergent. The red arrow shows the position of the detergent micelle that forms around the base of the gp41_ECTO, where the MPER is located. Two SOSIP.681 trimers attached to same micelle are indicated by black arrows. In the lower two circled images, rosettes can be seen. The bar at the bottom measures 200 Å.

Truncation of the MPER prevents KNH1144 SOSIP gp140 from aggregating.

To explore the above hypotheses, we elected to truncate the MPER of KNH1144 SOSIP gp140. To define the optimal truncation point, we evaluated the structure and amino acid composition of the MPER (i.e., residues 658 to 681) (Fig. 5). The MPER contains many bulky, hydrophobic residues, most commonly tryptophan, particularly near its C terminus. The crystal structures of the MPER (as a peptide) in complex with NAbs 2F5, 4E10, Z13e1, and 10E8, together with nuclear magnetic resonance and simulation studies, show that two helical segments are separated by a hinge region formed by residues Phe-674 and Glu-675 (48, 53, 54). It has been proposed that the multiple tryptophan moieties anchor the MPER in the viral membrane by intercalating their indole groups between phospholipid molecules (53). Accordingly, the degree of lipid immersion rises as the tryptophan density increases toward the C terminus of the MPER (Fig. 5A) (51). To identify the optimal place to truncate the MPER, we analyzed its hydrophilicity (http://web.expasy.org/protscale/). The curves for that parameter change dramatically after Gly-664, suggesting this position could be a particularly suitable truncation point (Fig. 5B). The resulting construct has Gly-664 as its C-terminal residue and is designated SOSIP.664 gp140.

Fig 5.

Removal of MPER residues 665 to 681 prevents aggregation of KNH1144 SOSIP gp140s. (A) Schematic representation of a gp140 showing the mutations introduced to make the SOSIP.681 construct. The highlighted residues 658 to 681 constitute the MPER. The asterisks above the amino acid sequence indicate hydrophobic residues. The letters h (α-helix), h (3₁₀ helix), and t (turn) under the individual amino acids indicate the local structure of the MPER in complex with the 10E8 bNAb (54). The triangle illustrates the proposed increasing relative lipid immersion depth of the MPER amino acids toward the C-terminal end of the MPER domain in the context of a membrane (51). (B) The hydrophilicity score (http://web.expasy.org/protscale/) for the MPER is given, indicating the location of surfaces that may contribute to aggregation. Note that the score is determined over a range of residues, and therefore no score can be assigned to residues beyond position 674. In panels A and B, the arrows indicate Gly-664, where the SOSIP.664 construct terminates. (C) 2G12 affinity-purified KNH1144 SOSIP.681 (light gray plot) or SOSIP.664 (black plot) gp140s, produced in 293T cells, were fractionated by SEC on a Superose 6 column at a flow rate of 0.5 ml/min. The elution positions of aggregates, trimers, and monomers are marked. Note that the SOSIP.681 gp140 profile is the same as the one shown in Fig. 1A. (D) The SEC-purified SOSIP.664 gp140 trimer was analyzed by BN-PAGE followed by Western blotting with the anti-gp120 MAb ARP3119. The molecular masses (kilodaltons) of marker proteins are shown to the left. Note the absence of aggregates.

The KNH1144 SOSIP.664 gp140 was transiently expressed in 293T cells, purified by 2G12 affinity chromatography, and studied by analytical SEC (Fig. 5C). When the UV traces for the purified SOSIP.664 and SOSIP.681 gp140 trimers were superimposed, very prominent trimer peaks for both proteins were visible at the elution volume of ∼14.9 ml. However, the aggregate peak (12.2 ml) was entirely absent from the SOSIP.664 gp140 preparation, in contrast to SOSIP.681 (Fig. 5C). A shoulder, representing gp140 monomers (16.6 ml), was slightly more prominent for SOSIP.664 than SOSIP.681. A BN-PAGE analysis of 2G12-purified SOSIP.664 Env proteins confirmed that no aggregates were present in the SOSIP.664 preparation, while trimers were abundant and a gp140 monomer band was also visible (Fig. 5D). Thus, truncation of the MPER prevents SOSIP gp140 aggregation, but it modestly decreases trimerization efficiency (Fig. 5). In practice, the latter observation is not a substantive concern as the preparative SEC process cleanly separates the trimers and monomers; moreover, the SOSIP gp140 monomers can themselves be collected and purified for additional structural studies.

The MPER contributes to KNH1144 SOSIP.681 gp140 trimerization.

The modest increase in the proportion of gp140 monomers in SOSIP.664 preparations suggests that the MPER may be involved in trimer formation or stabilization. To further study this, we sequentially deleted the WLWY residues from the C terminus of SOSIP.681 gp140, in the context of both the full-length and ΔV1V2V3 variants (Fig. 6). A BN-PAGE analysis shows that when both Trp-680 and Tyr-681 were deleted from the former variant, the abundance of trimers was decreased at the expense of dimers. Similar results were obtained with ΔV1V2V3 gp140, except that only when all of the last three residues (Leu-679, Trp-680, and Tyr-681) were deleted did the proportion of dimers increase. These findings indicate that the MPER does indeed contribute to the trimerization of SOSIP gp140 proteins and, by extension, possibly that of membrane-associated Env proteins.

Fig 6.

The MPER contributes to KNH1144 SOSIP.681 gp140 trimerization. BN-PAGE and Western blot analyses of unpurified (A) Full-length (FL) and (B) V1V2V3-deleted SOSIP.681 gp140 proteins compared to variants of the same proteins truncated at residues 677, 678, 679, or 680. The molecular masses (kilodaltons) of marker proteins are shown to the left. The positions of the trimer and monomer bands are indicated to the right. The lower panels show SDS-PAGE analyses of the same samples.

KNH1144 SOSIP.681 trimers lacking V1V2V3 are highly dependent on Pro-559.

We show in the accompanying article (20) that EM reconstructions of the native, membrane-associated Env spike and the soluble SOSIP.681 gp140 trimers yield similar structures (29). The intersubunit contacts are located close to the membrane, presumably mediated by gp41_ECTO (perhaps including MPER residues), but also at the apex where the V1/V2 and V3 regions are positioned (5–8). We, therefore, made the SOSIP.664.ΔV1V2V3 construct to assess whether cleaved, soluble trimers would be stable when both the variable loops and the MPER were deleted. A BN-PAGE analysis of SEC-purified SOSIP.681 and SOSIP.664 and their respective ΔV1V2V3 variants showed that the SOSIP.664.ΔV1V2V3 construct could still form trimers, despite lacking the two regions that presumably do contribute to trimerization (Fig. 7A). Hence gp41_ECTO residues outside the MPER and V1V2V3 must play the major role in the formation of SOSIP gp140 trimers.

Fig 7.

V1V2V3-deleted KNH1144 SOSIP gp140s are highly dependent on Pro-559 for trimer formation. (A) BN-PAGE analyses of SEC-purified, full-length (FL) SOSIP.681 and SOSIP.664 trimers as well as SOSIP.681.ΔV1V2V3 and SOSIP.664.ΔV1V2V3 trimers. (B) BN-PAGE and Western blot analyses of unpurified SOSIP.681 and SOSIP.681.ΔV1V2V3 gp140 trimers as well as their counterparts lacking Pro-559 (SOS.681 and SOS.681.ΔV1V2V3). The positions where molecular mass (kilodaltons) marker proteins (not shown) migrate are indicated. The lower panels show SDS-PAGE analyses of the same samples.

One such region seemed likely to involve residue Ile-559, where the I559P substitution was designed specifically to improve the trimerization of SOSIP gp140s (23). In contrast to soluble Env of various other genotypes, KNH1144 SOSIP gp140 proteins have a high natural propensity to trimerize, such that the I559P substitution is not absolutely critical for trimer formation (23, 33, 55, 56; unpublished observations). We, therefore, removed the trimer-stabilizing Pro-559 from the KNH1144 SOSIP.681 and SOSIP.681.ΔV1V2V3 gp140s, yielding SOS.681 and SOS.681.ΔV1V2V3, respectively. In contrast to the full-length SOS.681 protein, the SOS.681.ΔV1V2V3 gp140 did not form any trimers (Fig. 7B). The loss of trimers was compensated for by a significant increase in gp140 dimers but not monomers. Thus, V1V2V3 elements do contribute to the formation of SOSIP gp140 trimers, although their influence is sometimes only apparent when other trimer-stabilizing sequences are absent.

MPER truncation has no global effect on the antigenic structure of KNH1144 SOSIP gp140.

We used a set of NAbs to probe the antigenic structure of various SOSIP gp140 trimer preparations, first by immunoprecipitation and then by using more quantitative techniques. In an immunoprecipitation assay, the full-length and ΔV1V2V3 versions of the SOSIP.681 and SOSIP.664 trimers all interacted efficiently with the tetravalent CD4 construct CD4IgG2 and with NAbs VRC01, PGV04, and 1-863 to CD4bs epitopes (36, 38, 57, 58) (Fig. 8). MAb 412d against a CD4-induced (CD4i) epitope bound better to SOSIP.664 gp140 trimers than to SOSIP.681, irrespective of whether sCD4 was present to induce the epitope (Fig. 8). The ΔV1V2V3 variants did not bind the CD4i NAbs (data not shown), which is consistent with the known involvement of V3 residues in the formation of these epitopes. As expected, the MPER deletion eliminated the binding sites for NAbs against this region (i.e., 2F5, 4E10, and 10E8) (data not shown).

Fig 8.

MPER truncation does not affect the antigenic structure of KNH1144 SOSIP gp140. Each panel shows an immunoprecipitate of SOSIP.681 (FL), SOSIP.664 (FL), SOSIP.681.ΔV1V2V3, and SOSIP.664.ΔV1V2V3 proteins by the indicated MAb or CD4-IgG2 protein.

Finally, we tested a number of MAbs from the PGT121 to -135 series, a group of bNAbs that interact with Asn 332 glycan-dependent epitopes on the outer domain of gp120 and the V3 base (59). PGT128 interacts with two glycans at positions Asn-301 and Asn-332, and elsewhere at the base of V3 (28). Both PGT128 and the related PGT125 MAb interacted efficiently with both SOSIP.664 and SOSIP.681 gp140 trimers, but not with either ΔV1V2V3 variant, which lacks critical V3 contact residues for these bNAbs. The PGT135 bNAb interacts with three glycans at Asn-332, Asn-392, and Asn-386 and penetrates the glycan shield to bind to the gp120 protein surface, including the V4 loop (60). It bound efficiently to all four SOSIP variants, consistent with the location of its epitope further from V1, V2 and V3. PGT123, which binds to an Asn-332-dependent epitope that overlaps that of PGT128 (30), also bound SOSIP.664 and SOSIP.681 gp140 trimers efficiently. The binding of PGT123 was less affected by the V1V2V3 deletion than was the case for PGT125 and PGT128, particularly with SOSIP.681. The precipitation of the trimers may thus be somewhat influenced by indirect conformational effects of the MPER (61, 62). In addition, differential glycosylation may have subtle effects. Overall, these data indicate that MPER deletion does not result in major perturbations of the antigenic structure of KNH1144 SOSIP gp140 trimers, a conclusion consistent with additional EM studies reported in the accompanying article (29).

MPER truncation does not affect the binding to CD4 and PGV04.

Mutations in gp41_ECTO have long been known to affect HIV-1 neutralization by MAbs against the CD4bs, quite plausibly by modulating their affinity for functional Env (63–67). Hence, it is also conceivable that the presence of residues at the C terminus of gp41_ECTO could influence the structure of conformationally sensitive epitopes in the gp120 components of SOSIP gp140 trimers. We therefore used SPR to quantify the binding of sCD4 and the CD4bs MAb PGV04 (VRC PG04) to immobilized SOSIP.681 and SOSIP.664 gp140 trimers. There were no marked kinetic differences in the binding of sCD4 or PGV04 between these trimers (Fig. 9); the initial kinetic and the calculated equilibrium dissociation constants for each analyte were comparable for the two trimers (Table 1). There was, however, a tendency toward lower on- and off-rate constants for sCD4 with SOSIP.664 than with SOSIP.681 trimers, although the ratios (i.e., the K_d1 values), were similar for the trimers. Both analytes gave higher S_m values on SOSIP.664 gp140 trimers than SOSIP.681 (mean ± standard error of the mean [SEM], 1.8-fold ± 0.41-fold). Although the amine coupling of trimer to dextran quite plausibly affects analyte binding and reduces the number of available binding sites (the S_m values are not supplied for stoichiometric interpretation but as a degree of quality control), we detected high-affinity interactions by the two analytes and affinities for the two trimers, which showed only modest differences in extent of binding. A more detailed, SPR-based study of MAb reactivity with SOSIP gp140 trimers and the relationship to virus neutralization will be described elsewhere.

Fig 9.

MPER truncation does not affect the kinetics of KNH1144 SOSIP gp140 binding to sCD4 and PGV04. Purified SOSIP.681 (left panels) and SOSIP.664 (right panels) gp140 trimers were immobilized on CM5 chips. The SPR sensorgrams record the differences in responses (Resp. Diff) between the ligand and control channel, in response units (RU), on the y axes as a function of time (s) on the x axes. They show the binding of (A) sCD4 at 3.4, 6.9, 14, 28, 56, and 110 nM for SOSIP.681, (B) 5.5, 11, 22, 44, 89, and 180 nM for SOSIP.664, and (C and D) MAb PGV04 at 5.5, 11, 22, 44, 89, 180, and 360 nM.

Table 1.

SPR parameters of sCD4 and PGV04 binding

Analyte (n)	Ligand	kon1 (M−1 s−1)a	koff1 (s−1)a	Kd1 (nM)a	Smb
sCD4 (4)	SOSIP.681	1.2 × 105 ± 3.5 × 104	5.1 × 10−3 ± 5.4 × 10−4	47 ± 17	0.79 ± 0.039
	SOSIP.664	4.7. × 104 ± 7.5 × 103	2.4 × 10−3 ± 6.5 × 10−4	57 ± 20	1.1 ± 0.090
PGV04 (4)	SOSIP.681	6.2 × 103 ± 1.4 × 103	2.7 × 10−4 ± 1.5 × 10−5	54 ± 14	0.68 ± 0.058
	SOSIP.664	5.2 × 103 ± 1.5 × 103	2.8 × 10−4 ± 3.0 × 10−5	76 ± 27	1.5 ± 0.021

^aThe values represent means of 4 replicates ± SEM. The kinetic constants k_on1 and k_off1 describe the initial interaction of analyte with ligand; they were obtained by fitting a two-state, induced-fit model to the sCD4 data and a bivalent model to antibody data (see Materials and Methods). The dissociation constant for this initial interaction is calculated from the kinetic constants: K_d1 = k_off1/k_on1.

^bS_m describes the stoichiometry of the binding: the number of analyte molecules bound per Env trimer. S_m = (R_max/R_L) × (M_L/M_A), where R_max is the modeled maximum response at equilibrium, R_L is the ligand immobilization level, M_L is the molecular mass of the Env trimer (ligand), and M_A is the molecular mass of sCD4 or PGV04 (analyte).

MPER truncation does not alter the overall morphology of KNH1144 SOSIP trimers.

We obtained negative-stain EM images of the unliganded KNH1144 SOSIP.664 and SOSIP.681 gp140 trimers. The 2D class averages again showed a CYMAL-7 micelle was attached to the bottom of the SOSIP.681 gp140 trimer. The micelle was absent in the SOSIP.664 trimers purified in the absence of detergent, but other than that, the trimers looked very similar (Fig. 10). Furthermore, although the micelles in the images obscure some parts of the SOSIP.681 gp140 trimers, it can be clearly seen that the CD4bs-directed PGV04 Fabs bind in a similar manner to these trimers and to the SOSIP.664 variants (Fig. 10). Additional images are provided in the accompanying article (29).

Fig 10.

MPER truncation does not alter the overall morphology of KNH1144 SOSIP trimers. Shown are 2D class averages of negative-stain EM images of KNH1144 SOSIP.664 and SOSIP.681 trimers with and without PGV04 (see reference 29 for more details). Note that both SOSIP.664 and SOSIP.681 exhibit a three-bladed fan and that micelles can be seen in the SOSIP.681 samples. The left- and right-hand images are two different views of the same sample. The white scale bar to the left above the bottom panel measures 100 Å.

DISCUSSION

The goal of this study was to improve the biochemical properties of cleaved SOSIP gp140 trimers to increase their utility for structural and, perhaps, immunogenicity studies. Using a variety of techniques, notably negative-stain EM, we observed that trimers containing the complete gp41_ECTO (SOSIP.681 gp140) had a tendency to aggregate unless nonionic detergents were present. Both aggregates and added detergents are undesirable in crystallization studies as they could impede crystal formation and packing; detergents may also affect the outcome of immunogenicity experiments, for example, by interacting unpredictably with some adjuvant components. Negative-stain EM showed that the added detergent formed a micelle around the base of the gp41_ECTO component of the SOSIP.681 trimer, presumably associated with the relatively hydrophobic MPER and perhaps substituting for the membrane lipids to which the MPER is known to bind (49–52). As we wished to avoid using detergents when making soluble trimers, we chose to truncate the MPER. We realized that this strategy would eliminate an epitope cluster for bNAbs, although the peptide sequences are insufficient for induction of such Abs, even when presented in a trimeric context (68–70). The precise choice of where to truncate the new trimers was based on a biophysical assessment of the hydrophilicity of residues in the MPER. As a result, we decided to truncate the new variant C terminally of residue 664.

By a variety of criteria, but notably SEC and negative-stain EM, the resulting SOSIP.664 gp140 trimers are a substantial improvement on their predecessors. In the absence of detergent, they are stable and homogenous with no tendency to form aggregates. In the accompanying article (29), we show that the SOSIP.664 and also SOSIP.681 gp140 trimers closely mimic virion-associated Env spikes when viewed by negative-stain EM, both alone and as complexes with sCD4 or the PGV04 bNAb (29).

We did note that there was some reduction in trimer formation for the SOSIP.664 gp140 construct compared with SOSIP.681, which suggests that the MPER contributes to trimerization of soluble, cleaved gp140s. The degree of instability was modest, and we deemed it acceptable given the other benefits that accrued. We have since applied the 664 truncation to next-generation trimers based on the BG505 subtype A sequence; the resulting BG505 SOSIP.664 trimers are also aggregate free, highly homogenous, and stable (30, 31; R. W. Sanders, R. Derking, A. Cupo, J. P. Julien, A. Yasmeen, N. de Val, H. J. Kim, C. Blattner, A. Torrents de la Peña, J. Korzun, M. Golabek, K. de los Reyes, T. J. Ketas, M. van Gils, C. R. King, I. A. Wilson, A. B. Ward, P. J. Klasse, and J. P. Moore, submitted for publication). Thus, we believe that the MPER truncation method of improving the properties of soluble gp140 trimers will be generalizable. An earlier mutagenesis study on Env trimers based on the subtype B isolate, YU2, showed that the MPER also modestly affected the trimerization of uncleaved, soluble Env (71). Hence, again, our present findings on KNH1144 Env may be generalizable to other genotypes. Whether truncation at residue 664 is optimal remains to be determined.

We found that the formation of aggregates of KNH1144 SOSIP.681 gp140 proteins does not involve the gp120 V1, V2, or V3 regions. These variable loops do contribute, however, to a trimerization site at the apex of the gp140 spike (5–8). Nonetheless, the SOSIP.664.ΔV1V2V3 gp140 construct could still form trimers, despite lacking contributions to trimer formation from residues located in the MPER and the variable loops. The major trimerization determinants must, therefore, reside elsewhere in gp41_ECTO. One such region involves Ile-559, where the I559P substitution was designed specifically to improve the trimerization of SOSIP gp140 proteins (23). We, therefore, made two gp140 variants in which the stabilizing I559P change was reversed. In contrast to the full-length SOS.681 construct (which has Ile-559), which formed gp140 trimers fairly efficiently, the SOS.681.ΔV1V2V3 gp140 (which also has Ile-559) formed none at all, with predominantly gp140 dimers but not monomers present. The V1/V2 and V3 regions can, therefore, contribute to the formation of SOSIP gp140 trimers, but their effect is not always apparent when other, stronger influences on trimer formation, such as the I559P change, are also present.

Truncation of the MPER had no globally adverse effect on the antigenic structure of the KNH1144 SOSIP.664 gp140 trimers, compared with SOSIP.681, other than the obvious loss of the MPER epitopes for bNAbs such as 2F5, 4E10, and 10E8. Thus, using SPR, we found no major kinetic differences in the binding of sCD4 or the CD4bs MAb PGV04 to immobilized SOSIP.681 and SOSIP.664 trimers: there was only a tendency toward a slower association and dissociation of sCD4 with SOSIP.664 gp140 trimers compared with SOSIP.681, whereas the extent of binding to SOSIP.664 was greater. Additional studies on SOSIP.664 trimers based on the BG505 sequence support the conclusion that the 664 truncation has a benign overall effect on antigenicity (A. Yasmeen, R. W. Sanders, J. P. Moore, and P. J. Klasse, unpublished results).

In conclusion, the truncation of the MPER by termination of SOSIP.664 gp140 proteins at residue 664 improves their biochemical and biophysical properties, in that the desired trimers can be purified and are stable in the absence of detergents. The resulting SOSIP.664 gp140 trimers are more suitable than their predecessors for structural and, perhaps, immunogenicity studies. Making any soluble gp140 trimer necessarily involves modification of the native trimer present on virions, because the soluble versions lack the transmembrane and cytoplasmic domains. As there are substantial advantages to working with soluble trimers, some compromises are justified. In other studies on the structural, antigenic, and immunogenic properties of the next-generation BG505 SOSIP.664 gp140 trimers, we will establish their validity as appropriate mimics of native, virion-associated Env spikes (unpublished data). Here, and in the accompanying article, we show that the KNH1144 SOSIP.664 gp140 trimers mimic the structures of virion-associated Env spikes when viewed by negative-stain EM at resolutions of 15 to 20 Å (29). Moreover, how sCD4 and PGV04 bind to SOSIP gp140 trimers is consistent with the native-like properties of the latter; for example, sCD4 binding is capable of inducing conformational changes that open up the SOSIP gp140 trimers (Table 1 and Fig. 8 and 9) (5, 29). In marked contrast, we will show elsewhere that soluble gp140 trimers that are uncleaved between gp120 and gp41_ECTO do not resemble cleaved SOSIP gp140 trimers and native Env spikes, either from the antigenicity perspective or when viewed by negative-stain EM. In other words, the cleavage status of soluble, as well as probably membrane-associated, trimers is absolutely critical to their consistent adoption of a functionally relevant structure.