Processing, Stability, and Receptor Binding Properties of Oligomeric Envelope Glycoprotein from a Primary HIV-1 Isolate*

The envelope glycoprotein of human immunodeficiency virus type 1 (HIV-1) is thought to exist on the virion surface as a trimer of non-covalently associated gp120/gp41 molecules. We expressed trimeric envelope glycoprotein from three primary, macrophage tropic HIV-1 isolates in baby hamster kidney cells and analyzed the furin-mediated cleavage, stability, and receptor binding properties of the oligomers. The envelope glycoprotein was secreted in a soluble form deleted of its transmembrane anchor and the intracytoplasmic domain (gp140). A mixture of trimers, dimers, and monomers of gp140 as well as monomeric gp120 was detected on polyacrylamide gels. Analysis by sucrose gradient centrifugation revealed that trimers and dimers were essentially composed of uncleaved gp140, whereas most of the gp120 was found in the monomeric fraction. To analyze the effect of the cleavage of gp140 to gp120/Δ41 on trimerization, we co-expressed the furin protease along with gp140. Surprisingly, furin expression changed the subcellular localization of the envelope glycoprotein, which became in majority sequestered in the major furin compartment, the trans-Golgi network, as judged by confocal laser microscopy. The envelope glycoprotein secreted from furin-co-expressing cells was almost completely cleaved to gp120 and Δgp41, but gp120 was found exclusively in the monomeric fraction, with a few residual oligomers being composed of uncleaved gp140. Secreted uncleaved gp140 trimers were purified to homogeneity and analyzed for their capacity to interact with cellular receptors CD4 and CC chemokine receptor 5 (CCR5). Receptor binding was analyzed on CD4- and CCR5-expressing cells as well as on peripheral blood mononuclear cells. Trimers showed greatly reduced binding to CD4 as compared with monomers. Neither monomers nor trimers bound directly to CCR5. In conclusion, our results show that the cleaved form of the envelope glycoprotein does not form stable trimers, suggesting that gp120/gp41 oligomers on the virion surface might be stabilized by a yet to be identified mechanism and that the virion might attach to CD4 via a monomeric form of gp120. These results are relevant to the development of an envelope-based vaccine against AIDS.

The HIV-1 1 envelope glycoprotein is synthesized as an approximately 845-870-amino acid precursor in the rough endoplasmic reticulum. High-mannose sugar chains are added to asparagine residues to form the gp160 glycoprotein, which assembles into oligomers (1)(2)(3)(4)(5)(6). The gp160 oligomers are transported to the Golgi apparatus, where cleavage by the cellular protease furin (7) or a related enzyme of the subtilisin family generates the mature envelope glycoproteins: gp120, the exterior envelope glycoprotein, and gp41, the transmembrane glycoprotein.
The gp41 glycoprotein includes a fusion peptide and an ectodomain that is implicated in trimerization (8 -10), a membrane-spanning anchor, and a long cytoplasmic tail. Crystallization studies suggest that the core of the fusion-active envelope oligomer is a trimeric structure of 36 amino acids located toward the N terminus of gp41 (10). This sequence forms an interior, parallel coiled-coil trimer, whereas three helices of a sequence closer to the C terminus of gp41 pack in an oblique, antiparallel manner into highly conserved, hydrophobic grooves on the surface of this trimer. This structure resembles the low pH-induced conformation of influenza hemagglutinin.
The gp120 and gp41 glycoproteins are held together by noncovalent interactions between the gp41 ectodomain and discontinuous structures composed of N-and C-terminal gp120 sequences (11). When they reach the infected cell surface, a fraction of these envelope glycoprotein complexes is incorporated into budding virus particles. gp120/gp41 complexes on the virion surface as well as at the plasma membrane of infected cells are functional as they can mediate fusion with CD4 ϩ /CCR5 ϩ cells. Given their non-covalent association, these complexes can disassemble, releasing gp120 and exposing the previously buried gp41 ectodomain. Most of the surface-exposed elements of the mature, oligomeric envelope glycoprotein complex (reviewed in Ref. 12), such as the CD4 (13) and coreceptor binding domains (14,15) as well as major neutralization epitopes, are contained on the gp120 glycoprotein (16). Evidence from studies of virion-associated (5,6) or recombinant (2,(17)(18)(19) envelope glycoprotein support the idea that the HIV envelope glycoprotein is a trimer or tetramer. However, these studies did not show purified trimeric envelope complexes composed of cleaved gp120/gp41.
HIV entry requires the interaction of the gp120 subunit of the envelope glycoprotein with specific receptors: the CD4 glycoprotein (13,20,21) and members of the chemokine receptor family (22)(23)(24). Binding to CD4 is a major determinant of HIV tropism in vivo, illustrated by the presence of CD4 on the surface of the main target cells, e.g. T lymphocytes, monocytes, macrophages, dendritic cells, and brain microglia. Binding of gp120 to CD4 results in a conformational change in gp120, which contributes to the formation or exposure of the binding site for the coreceptor (25,26). The major coreceptor for syncytium-inducing strains appearing at the late stages of AIDS progression as well as T-cell line-adapted strains is CXCR4, a receptor for the CXC chemokine stroma cell-derived factor (SDF) 1␣ (27), although all primary M-tropic strains of HIV-1 described to date have been shown to be capable of using CC chemokine receptor (CCR) 5 (28,29), a receptor for the ␤-chemokines macrophage inflammatory protein (MIP) 1␣ and ␤ and RANTES (regulated upon activation, normal T cell-expressed and -secreted) (23). These binding events result in the exposure of the previously hidden fusion peptide of gp41 and its penetration into the target cell membrane (30), leading to fusion of the virion with the plasma membrane.
We report here on the production, purification, and characterization of soluble trimeric envelope glycoprotein from primary macrophage-tropic isolates of HIV-1. Purified trimers are important tools to study the interaction of native envelope with cellular receptors and to analyze their immunogenic potential in view of the development of an anti-HIV subunit vaccine.
Recombinant defective SFV particles were synthesized as described previously (36,37). Briefly, plasmids pSFV-helper2, pSFV⌬envBX08, pSFV⌬envBX17, pSFV⌬envE402, pSFVCD4, pSFVCCR5, pSFVlacZ, and pSFVhfur were digested with SpeI or SphI (pSFV⌬envE402), purified, and in vitro transcribed using SP6 RNA polymerase in a buffer containing cap analog. After analysis of RNAs on agarose gels, pSFV-helper2-derived RNA was admixed to equal quantities of RNA derived from plasmids pSFV⌬envBX08, pSFV⌬envBX17, pSFV⌬envE402, pS-FVCD4, or pSFVCCR5. The RNA mixtures were added to 8 ϫ 10 6 BHK-21 cells in 800 l of PBS and immediately transferred to a 0.4-cm electroporation cuvette (Eurogentec). The RNA-cell mixture was subjected to two 0.4-ms pulses at 830 V and 25 microfarads in a Bio-Rad gene pulser and plated into 75-cm 2 flasks in GMEM medium containing 5% FCS. 24 h later the supernatant containing recombinant defective SFV particles was harvested, concentrated by ultracentrifugation, and separated into aliquots. Before infection aliquots were activated by chymotrypsin digestion. For infection with recombinant SFV, cells were washed with serum-free GMEM and incubated with dilutions of viral particles in 2% GMEM for 1 h at 37°C followed by incubation in 5% FCS GMEM. At 6 h post-infection, the medium was replaced by 0% FCS GMEM for standard protein purification and by 0% FCS methionine free-DMEM containing 35 S-labeled methionine for the production of radiolabeled gp140. Synthesis of gp140 was continued up to 24 h post-infection.
Purification of Oligomeric Soluble gp140 -Supernatants of SFV⌬envinfected BHK cells were harvested at 24 h post-infection, centrifuged at low speed to pellet cellular debris, and concentrated approximately 30-fold in Vivaspin concentration units (cut off 100 kDa). To cross-link gp140 oligomers, the supernatant was incubated with 5 mM ethylene glycolbis(sulfosuccinimidylsucccinate) (sulfo-EGS) (Pierce) in 20 mM Hepes for 30 min at room temperature. The reaction was stopped by incubation in 50 mM Tris-HCl (pH 7.5) for 15 min at room temperature. Sucrose gradients (5-20%) were prepared as described previously (2). Approximately 600 l of concentrated untreated or cross-linked supernatant was carefully added to the top of the gradient followed by ultracentrifugation in a SW41 rotor at 37,000 rpm for 18 h at 4°C. Fractions were collected at 2 ml/min from the bottom of the tube using a capillary hooked on a peristaltic pump (Gilson). Fractions were deposited on SDS-PAGE gradient gels and analyzed by Western blot. Protein concentration was determined by Western blot using gp140 MN as a standard. If necessary, corresponding fractions were pooled, diluted in PBS, and concentrated in low binding Centriplus units (Amicon).
Subcellular Localization of CD4, CCR5, gp140, and Furin-BHK cells were infected with SFVCD4 or SFVCCR5 for 12 h and fixed with 4% paraformaldehyde before incubation with PBS containing 5% normal horse serum. Subsequently, cells were incubated with mAb Q4120 directed against CD4 (40) or mAb 182 directed against CCR5 (R&D Systems). Specific binding was detected by FITC-labeled secondary antibodies.
BHK cells were infected separately (MOI 5) or co-infected with SFV⌬envBX08 (MOI 5) and SFVhfur (MOI 1) for 12 h and then fixed with 4% paraformaldehyde at 4°C for 15 min and permeabilized by incubation with 0.1% Triton in PBS at room temperature for 4 min. Then cells were washed 3 times with PBS and incubated with 5% normal horse serum in PBS. Subsequently, cells were incubated simultaneously with human mAb 2G12 (2 g/ml) and mouse mAb MON-152 (1 g/ml) directed against human furin (Alexis Biochemicals, San Diego, CA). Specific binding was detected by simultaneous incubation with anti-human IgG FITC-labeled antibody and anti-mouse IgG biotinlabeled antibody ϩ avidin-rhodamine-labeled antibody (Vector). Labeled cells were analyzed with a Zeiss confocal laser microscope.
Receptor Binding Studies-For binding studies, BHK cells were infected at a MOI 5 with SFVCD4, SFVCCR5, or SFVlacZ for 12 h, washed 2ϫ with GMEM 0% FCS, and incubated with ligands for 30 min at 37°C. After incubation with ligand, cells were washed 3 times with serum-free medium and incubated in lysis buffer (1% SDS, 500 mM NaCl, 10 mM EDTA, 1%Triton X-100 with 1% sodium deoxycholate, 200 mM Tris-HCl, pH 7.5), and cell-associated radioactivity was counted in biodegradable counting scintillant (Amersham Pharmacia Biotech) in a liquid scintillation counter (Rackbeta, Amersham Pharmacia Biotech). Purified unlabeled gp120 IIIB (Medical Research Council) or gp140 MN (Aventis-Pasteur) were incubated on cells at 1 to 20 g/ml in serum-free medium, washed, and fixed with methanol for 5 min at Ϫ20°C. To detect bound glycoprotein, cells were incubated with mAb K24 at 2 g/ml followed by incubation with 125 I-labeled anti-mouse IgG F(abЈ)2 fragment. Radiolabeled secondary antibody was detached by incubation in lysis buffer. 35 S-Labeled purified oligomers and monomers of gp140 were incubated on cell monolayers at various concentrations for 30 min at 37°C, after which the cells were washed, lysed, and counted. BHK cells infected with SFVCCR5 or SFVlacZ were incubated with 125 Ilabeled MIP-1␤ (12000 cpm; PerkinElmer Life Sciences) alone or together with various concentrations of cold MIP-1␤ for 30 min at 37°C and then washed, lysed, and counted.
Human PBMC used for binding studies were isolated according to standard procedures on Ficoll-Paque ® (Amersham Pharmacia Biotech) cushions, stimulated with 1 g/ml phytohemagglutinin (Murex) for 3 days. PBMC (10 6 cells) were incubated with 35 S-labeled gp140 BX08 monomers and dimers or trimers at various concentrations in a total volume of 200 l of PBS for 30 min at 37°C. Control cells were preincubated for 30 min with 10 g/ml mAb Q4120 directed against CD4. Binding was further controlled by preincubating 35 S-labeled gp140 BX08 monomers, dimers, or trimers at 1 g/ml with soluble CD4 at 20 g/ml.

Expression, Purification, and Antigenicity of Soluble
Oligomeric gp140 -Soluble envelope glycoproteins (gp140) from three different HIV-1 isolates corresponding to clades A, B, and E were cloned and expressed in the Semliki Forest virus expression system (36,38). The expressed sequences, which include the entire gp120 region and the extracellular domain of gp41, correspond to gp160: amino acids 1 to 692 of isolate BX17, amino acids 1 to 666 of isolate BX08, and amino acids 1 to 679 of isolate E402. Envelope glycoproteins were expressed in BHK cells infected with recombinant vectors SFV⌬envBX17, SFV⌬envBX08, or SFV⌬envE402. gp140 could be detected in infected cells (data not shown) and in supernatants at approximately 1 to 5 g/10 6 cells. Five different forms of the envelope glycoprotein could be visualized by SDS-PAGE under non-reducing conditions. The size of these different molecules was calculated based on migration of molecular weight markers identifying the main forms as trimers and dimers as well as monomers of uncleaved gp140, cleaved gp120, and ⌬gp41 (Fig.  1A). Oligomerization patterns of gp140 from the different isolates were very similar, although the gp140 BX17 and gp140 E402 bands migrated slightly faster than gp140 BX08 , possibly due to reduced glycosylation of these envelopes. Furthermore, cleavage of gp140 BX17 and gp140 E402 seemed to be less efficient, as demonstrated by the low level of gp120 in supernatants (Fig. 1A).
When supernatants from gp140 BX08 -secreting cells were analyzed on sucrose gradients (5 to 20%), different oligomeric and monomeric forms could be resolved. In addition to the main molecular entities (trimers, dimers, monomers), the existence of intermediate forms between monomers and dimers and dimers and trimers could be observed. Oligomers with a higher molecular weight than trimers, possibly tetramers, could also be detected (Fig. 1B). About half of the secreted glycoprotein was present in the cleaved form (gp120).
To find out whether gp140 BX08 was secreted from cells as oligomers and subsequently disassembled into monomers before harvest of the supernatant at 24 h post-infection (p.i.), we performed the following time-course experiment. Secreted oligomers were collected in 3-h intervals until 36 h p.i. Samples from different time points were analyzed by SDS-PAGE and immunoprecipitated with mAbs F5.5 and 41A (data not shown). At all time points, the supernatant was composed of a mixture of different forms of envelope glycoprotein-containing trimers, dimers, and monomers at the same ratio. No increase of monomeric forms was found in samples corresponding to the late time points (24 -36 h p.i.) compared with samples from early time points (0 -6 h p.i.), suggesting that monomers did not stem from disassembled oligomers but were already secreted as monomers.
Mixed preparations of oligomers and monomers of gp140 BX08 , gp140 BX17 , and gp140 E402 were probed with monoclonal antibodies known for their specificity for either linear or conformation-dependent epitopes. gp140 BX08 and gp140 E402 , but not gp140 BX17 , could be immunoprecipitated by anti-V3loop mAbs K24 and F5.5. All three isolates could be immunoprecipitated by anti-CD4 binding site antibody 110K as well as conformation-dependent antibodies 2G12 (16) and 670D (39), suggesting that oligomers possessed a native conformation (data not shown). Consequently, recognition of gp140 or gp120 by 2G12 and 670D was greatly diminished when proteins were denatured and analyzed under reducing conditions.
Composition and Stability of Soluble Oligomeric Envelope Glycoprotein-To identify whether oligomers were composed of uncleaved gp140 BX08 or cleaved gp120/⌬gp41, sucrose gradient fractions corresponding to trimers or dimers were treated with reducing agents and subjected to SDS-PAGE analysis. This experiment revealed that the trimers did not contain gp120 and that they were exclusively made of uncleaved gp140 (Fig.  1C). Monomeric ⌬gp41 was not detected because of filtration of the supernatants through membranes with a 100-kDa cut off. However, after treatment in Laemmli buffer and analysis by SDS-PAGE, some monomeric ⌬gp41 was detected in fractions corresponding to monomers, dimers, and trimers of gp140 (Fig.  1B). The quantity of ⌬gp41 increased when furin was co-expressed with gp140 BX08 (Fig. 2). We conclude from this experiment that the ⌬gp41 subunits were associated with each other as non-covalently linked dimers and trimers. Such ⌬gp41 oligomers might account for the intermediate forms between monomers and dimers and dimers and trimers, e.g. a dimer of gp140 associated to one ⌬gp41 would migrate slightly higher than a gp140 dimer. Composition of trimers was further analyzed by treating fractions of sucrose gradients corresponding to trimers or dimers by SDS (2%) in Laemmli buffer and heat (3Ј, 90°C) before analysis on SDS-PAGE. Under non-reducing conditions (Fig. 1B, lower panel) only a partial dissociation of oligomers to monomers could be observed. We therefore asked the question of whether dissociation of oligomers was due to an intrinsic lability of oligomers in the supernatant of gp140secreting cells or whether dissociation of oligomers to monomers occurred during SDS-PAGE. Comparison of untreated supernatants containing gp140 BX08 oligomers (Fig. 1B, lower  panel) with supernatants treated with the cross-linking agent EGS (Fig. 1B, upper panel) revealed that oligomers in the supernatant were stable and that partial dissociation was due to heat/SDS treatment before SDS-PAGE. The stability of oligomers that resisted SDS and heat treatment was further investigated by subjecting these molecules to low pH, high ionic strength (5 M NaCl) or urea (8 M) and reducing agents dithiothreitol and ␤-mercaptoethanol. Only treatment with reducing agents allowed complete dissociation of oligomers into monomers (Fig. 1A). Very similar characteristics of stability were observed for oligomeric gp140 E402 and gp140 BX17 as well as for gp140s from different sources, e.g. vaccinia virus-expressed gp140 MN (Aventis-Pasteur) and gp140 89.6 (R. W. Doms, University of Pennsylvania).
These results show that the trimeric and dimeric fractions were composed of a mixture of heat/SDS-sensitive and heat/ SDS-resistant oligomers, the latter being sensitive to denaturation in the presence of reducing agents. To determine the composition of the trimeric fraction with respect to heat/SDSsensitive and heat-resistant oligomers, we analyzed 35 S-labeled gp140 BX08 trimers or dimers (Fig. 1B, lower panel) with a PhosphorImager. This analysis revealed that 50% of the envelope glycoprotein contained in the trimeric fraction was present as heat/SDS-resistant trimers, another 30% as heat/SDS-resistant dimers, and 20% as monomers. If one accounts for the monomers that must have been associated to the heat/SDSresistant dimers, the overall composition of the trimeric fraction is 50% as heat/SDS-resistant trimers, 45% as partially heat/SDS-resistant trimers, i.e. heat/SDS-resistant dimers associated to one monomer, and 5% as heat/SDS-sensitive trimers. The dimeric fraction was composed of 70% heat/SDS-resistant dimers and 30% heat/SDS-sensitive dimers.
To test whether such heat/SDS-resistant oligomers were present on HIV-1 BX08 virions, we analyzed concentrated virion preparations by SDS-PAGE and Western blot (data not shown). Treatment by SDS and heat under non-reducing conditions showed the presence of heat/SDS-resistant dimeric, but not trimeric nor tetrameric, molecules in concentrated virion preparations. A significant amount of envelope glycoprotein did not enter the gel, suggesting that these were aggregates. Only upon treatment with reducing agents could all of the glycoprotein be reduced to gp160 and gp120 monomers.
Furin-cleaved Oligomers Are Labile-The observation that the oligomeric fractions were essentially composed of gp140 (Fig. 1C) and that gp120 was exclusively found in the monomeric fraction (Fig. 1B) suggested that the cleavage of gp140 to gp120 and ⌬gp41 destabilized oligomers. To further investigate this phenomenon, we co-expressed the furin protease with gp140 and analyzed the cell supernatant on sucrose gradients. One-half of the supernatant was left untreated (Fig. 2, lower  panel), and the other half was cross-linked at the end of the 24-h synthesis in order to avoid potential shedding of gp120 from oligomeric complexes during centrifugation (Fig. 2, upper  panel). Quantification of 35 S-labeled gp140/gp120-specific bands in all fractions by PhosphorImager analysis showed that more than 90% of the envelope glycoprotein was present in the monomeric fraction. This was observed with the untreated as well as the cross-linked supernatant, demonstrating that disassembly of furin-cleaved oligomers was not due to the purification procedures but rather to the intrinsically unstable nature of furin-cleaved oligomers resulting in complete shedding of gp120. The residual oligomers that were secreted from gp140 BX08 /furin-co-expressing cells were composed of uncleaved gp140 (data not shown).
Colocalization of gp140 and Furin-The main subcellular localization of furin is in the trans-Golgi network where furin co-localizes with the endogeneous glycoprotein TGN38 (41). We analyzed the subcellular location of gp140 and furin in BHK cells infected with SFV⌬envBX08 or SFVhfur; co-infected with SFV⌬envBX08 and SFVhfur or co-infected with SFV⌬envBX08 and SFVlacZ as control. At 12 h p.i., cells were fixed, permeabilized, and labeled with mAbs 2G12 (anti-gp120) and MON-152 (anti-human furin). Cells expressing only gp140 BX08 or both gp140 and ␤-galactosidase displayed a diffuse staining throughout the cytoplasm, whereas cells expressing furin alone showed a discrete labeling in a subcellular compartment reminiscent of the trans-Golgi network. When cells expressed both gp140 and furin, the furin localization remained unchanged, but the gp120-specific label was detected in the same compartment as furin (Fig. 3). This finding suggests that furin targets or retains the HIV envelope glycoprotein during processing to the trans-Golgi network. In addition to its localization in the trans-Golgi network, Schä fer et al. (41) show that furin can be secreted. In our co-expression studies a secreted protein of approximately 80 kDa co-sedimented in sucrose gradients with monomeric gp120 (Fig. 2). This protein could not be detected in the supernatant when gp140 BX08 was expressed alone (Fig.  1B). We have confirmed the identity of this protein as furin by immunoprecipitation with the anti-human furin mAb MON-152.
Receptor Binding Studies-CD4 and CCR5 were expressed in BHK cells using SFVCD4 and SFVCCR5. Using specific antibodies to detect either receptor and confocal laser microscopy, we were able to show that the receptors were expressed at the plasma membrane ( Fig. 4A and C). To show that receptor molecules exhibited physiological and functional properties, receptor-expressing cells were incubated with specific ligands at 12 h p.i. Functional expression of CCR5 in BHK cells was analyzed by incubating these cells with 125 I-labeled ligand MIP-1␤. Fig. 4D shows that MIP-1␤ specifically bound to CCR5 and that binding of radiolabeled ligand could be inhibited by admixture of unlabeled MIP-1␤ to the culture supernatant. No binding of 125 I-labeled MIP-1␤ to control cells was observed.
CD4-expressing cells were probed with recombinant gp120 IIIB , and bound protein was detected using the V3 specific mAb K24 and 125 I-labeled anti-mouse IgG F(abЈ)2 fragment. Only CD4-expressing cells and not uninfected cells or cells infected with the SFVlacZ construct showed specific labeling (Fig. 4B). Comparable results were obtained when gp120 IIIB was incubated for 30 min at 37°C or for 2 h at 4°C.
To evaluate binding of gp140 BX08 oligomers to plasma membrane-expressed receptors, cells were infected with SFVCD4 or SFVCCR5 and then incubated with 35 S-labeled gp140 BX08 monomers, dimers, or trimers previously separated on a 5-20% sucrose gradient. Control protein gp120 IIIB at 1 g/ml bound equally well to CD4 when incubated in medium containing 20, 10, 5, or 0% sucrose, demonstrating that sucrose concentrations had no influence on envelope binding to CD4 (data not shown). Results in Fig. 5A show that trimers and dimers at 7.2 nM possess reduced binding capacity to membrane-expressed CD4 in comparison to monomers. When increasing concentrations of trimers and monomers were used for CD4 binding studies, we observed specific binding of trimers to CD4-expressing cells, albeit severalfold lower than that of monomers (Fig. 5B). To verify if this phenomenon could also be observed in primary cells, activated PBMC were incubated with trimers, dimers, and monomers in the presence or absence of the anti-CD4 mAb Q4120. Fig. 6 shows that monomers bound more efficiently to PBMC than trimers and dimers at 7.2 nM concentration and that this binding could be blocked by anti-CD4 antibody Q4120, known to interfere with the interaction of gp120 with CD4 (40). Neither monomeric nor trimeric soluble envelope glycoprotein bound directly to plasma membrane-expressed CCR5 (data not shown). This observation is consistent with data reported previously by others (15). DISCUSSION The present report confirms the oligomeric nature of the envelope glycoprotein from a primary R5 strain of HIV-1 (BX08) and illustrates the influence of furin cleavage on the stability of the oligomers. The transmembrane-deleted envelope glycoprotein (gp140) was secreted under the form of trimeric, dimeric, and monomeric gp140 as well as monomeric gp120 and ⌬gp41. The oligomerization profile of gp140s from different isolates belonging to clades A, B, and E were very similar.
Oligomer formation, however, was almost completely abolished upon gp140 cleavage whether by an endogeneous subtilisin protease or by co-expressed furin protease. The role of furin in the transformation of gp160 to the fusion-competent form gp120/gp41 (7,42) and its subcellular localization in the trans-Golgi network have been shown earlier (41). However, these studies did not address the question of subcellular localization in cells co-expressing gp140 or gp160 and furin. We found that in addition to cleaving gp140, furin sequestered and changed the secretion pathway of gp140, which became concentrated in the major furin compartment, the trans-Golgi network. The molecular interactions underlying the furin-mediated change in subcellular localization of the envelope glycoprotein are not known and are subject to further investigation.
The view that the envelope glycoprotein of HIV-1 is a trimer is supported by structural analyses of peptide sequences corresponding to the extracellular domain of gp41, which comprises the N-terminal amphipathic helix (8,10). Evidence from expression studies of soluble envelope glycoprotein containing gp120 fused to the gp41 ectodomain (2,18,19,43) as well as of membrane-bound full-length gp160 (17,44,45) support the view that the envelope glycoprotein is a trimer or a tetramer. Although the trimeric nature of vesicular stomatitis virus (46) and rabies virus (47) envelope glycoproteins has been determined using virion-derived envelope, relatively little is known about the oligomeric nature of the HIV envelope on virions (5, 6). Using cross-linking agents, Schawaller et al. (5) were able to detect dimeric, trimeric, and tetrameric gp160 on virions of a T-cell line-adapted strain. At present, available data do not allow the conclusion of whether the glycoprotein on the virion of primary HIV isolates is trimeric or tetrameric or whether it exists as a mixture of oligomeric and monomeric forms. The results presented here support the view that envelope glycoprotein oligomerization via the gp41 ectodomain is not sufficient to maintain cleaved gp120/gp41 as an oligomer. Although In A and C, plasma membrane expression of receptors was verified by immunofluorescence using mAbs for CD4 (Q4120, FITC) and CCR5 (mAb182, FITC). Cells were analyzed with a ZEISS confocal laser microscope. Intensity of staining is shown using artificial rainbow colors ranging from blue (low) to yellow/ red (high). Ligand binding was analyzed in B and D. gp120 IIIB specifically binds to CD4 but not to ␤-galactosidase-expressing cells (B). Cells were incubated with 1 g/ml of gp120 IIIB , and bound protein was detected with 125 I-labeled anti-mouse F(abЈ)2 fragment. Specific binding of MIP-1␤ to CCR5-expressing cells is shown in D. Cells were incubated with a constant amount of 125 I-labeled MIP-1␤ as tracer and increasing amounts of cold MIP-1␤. Radioactive tracer in the absence of cold MIP-1␤ did not bind to ␤-galactosidase expressing cells (data not shown). gp41 remains associated as oligomers, as shown by the presence of gp41 in the oligomeric fractions, the gp120 subunit is entirely shed from soluble complexes. This is in accordance with studies that showed detection of trimeric and tetrameric gp41 in mammalian cells expressing full-length gp160 LAI (3,45,48). In contrast to results obtained for Friend murine leukemia virus (49) and recently for HIV-1 (50), demonstrating dimeric association of the envelope glycoprotein subunit, HIV-1 BX08 gp120 dissociated exclusively as a monomer from gp120/ ⌬gp41 complexes.
The lability of the oligomeric envelope raises the question of how such complexes are held together on the virion and the surface of the infected cell. This could occur through the formation of heterotrimers composed of the stable, uncleaved gp160 and the cleaved gp120/gp41. However, our results do not support this hypothesis, as we observed only very little gp120 dissociation from trimeric or dimeric complexes. Another hypothesis could be the stabilization of the gp120⅐gp41 complex by a cellular protein such as the putative chaperon cyclophilin A (51), which has recently been reported on the HIV virion surface (52).
Several approaches have been developed to stabilize the trimeric envelope. Sodroski and co-workers (44) report on artificial stabilization of gp160 HXB2 trimers by introduction of a cysteine pair in the N-terminal gp41 helix. This protein was poorly cleaved by furin, and when this mutation was introduced into a transmembrane-deleted soluble envelope (gp140), no stabilization of trimers could be observed. In turn, these gp140s could be successfully stabilized by the addition of an oligomerization peptide from transcription factor GCN4 at the C-terminal end of the truncated gp41 sequence (53). Binley et al. (54) attempted to stabilize furin-cleaved gp140 oligomers by introduction of an artificial disulfide bond between the C terminus of gp120 and the N terminus of ⌬gp41. This allowed expression of soluble gp140 that cannot shed its gp120 subunit. Although this disulfide-linked gp120/⌬gp41 has native antigenicity, as judged by immunoprecipitation of monomers, the oligomeric nature of these molecules as well as their ability to bind CD4 has yet to be proven experimentally. Several lines of evidence allow the hypothesis of a physiological role for uncleaved gp160 in the fusion process. Large quantities of gp160 can be found at the plasma membrane of infected cells (17,55) and, to a lesser extent, on virions (56,57). Our own results and those of others show that furin can be targeted to the plasma membrane and secreted (41) (Fig. 2). Cell-surface gp160 can be cleaved to gp120 and gp41 by soluble furin added to the supernatant (42). A recent finding shows that Sindbis virus binds to cell surface heparan sulfates via the intact furin cleavage site of the E2 envelope glycoprotein (58). One could thus speculate that uncleaved gp160 at the virion or infected cell surface represents a stable form of the envelope oligomer, with the CD4 binding site hidden from recognition by antibodies. This oligomer could attach to plasma membrane heparan sulfates via its intact furin cleavage site and subsequently be activated to the fusion-competent form gp120/gp41 upon contact with cell surface-expressed furin.
Analysis of oligomers revealed that dimeric and trimeric fractions of all the gp140s studied contained mixtures of oligomers possibly stabilized by interchain disulfide bonds and non-covalently linked oligomers. A similar result has been obtained earlier with gp140 from the T-cell line-adapted strain FIG. 5. CD4 binding of trimers, dimers, and monomers of soluble envelope glycoprotein. BHK cells were infected with SFVCD4 for 12 h. In A, cells were incubated with purified trimeric, dimeric, and monomeric fractions of 35 S-labeled gp140 BX08 at 1 g/ml (7.2 nM), and in B, with increasing concentrations of purified trimeric and monomeric fractions of radiolabeled gp140 BX08 . Cells were washed three times before measuring cell-associated radioactivity. Results are shown as the mean of two determinations and are representative of four independent experiments.
FIG. 6. Binding of trimers, dimers, and monomers of soluble envelope glycoprotein to PBMC. Activated PBMC were preincubated (gray columns) or not (black and white columns) with mAb Q4120 directed against CD4 (␣-CD4) at 10 g/ml before incubation with purified radiolabeled trimeric, dimeric, and monomeric fractions of gp140 BX08 at 1 g/ml (7.2 nM). In parallel, radiolabeled trimers, dimers, and monomers of gp140 BX08 at 1 g/ml were preincubated with soluble CD4 at 20 g/ml before adding the mixture to PBMC (white columns). In positive controls (black columns), trimers, dimers, and monomers of gp140 BX08 at 1 g/ml were preincubated in PBS before adding the mixture to PBMC preincubated in PBS (white columns). Cells were washed four times before measuring cell-associated radioactivity. Results are shown as the means of two determinations and are representative of three independent experiments. HXB2 (19). In addition to HIV, interchain disulfide bond formation has been described for envelope glycoprotein subunits of other viruses, such as guinea pig cytomegalovirus (59), Friend murine leukemia virus (49), and Newcastle disease virus (60). The formation of intra-and interchain disulfide bonds by members of the protein disulfide isomerase family constitutes an integral part of the maturation of most secretory and membrane-bound proteins in the endoplasmic reticulum (51,61). Before targeting to the cell surface, the envelope glycoproteins of the alphavirus Semliki Forest virus undergo a covalent liaison with protein disulfide isomerase (62), which catalyzes disulfide formation of cysteine residues situated in close proximity during folding. In HIV, correct folding is supposed to go along with formation of intrachain disulfide bonds (63) followed by non-covalent oligomerization. In our system, only about 5% of all trimers and 30% of dimers were present as non-covalently linked molecules. The reason for the high proportion of disulfide-linked oligomers versus non-covalently linked trimers is unclear, and we currently do not know which cysteine residues could be involved in interchain disulfide formation. However, approximately one-half of the trimers were composed of interchain-linked dimers that are non-covalently bound to monomeric gp140. Furthermore, trimers showed reactivity with conformation-dependent monoclonal antibodies such as 2G12 and displayed a low, albeit significant, binding to CD4, indicating that their conformation was not aberrant. A physiological role of disulfide-linked oligomers could be considered if one takes into account plasma membrane expression of protein disulfide isomerase and rearrangement of disulfide bonds at the cell surface (64). Ryser et al. (65) described earlier that membrane-impermeable inhibitors of protein disulfide isomerase inhibit HIV infection and suggested rearrangement of disulfide bonds of the envelope glycoprotein subsequent to contact with target cells. We could confirm these results by showing that cell-to-cell fusion of gp120/gp41-expressing BHK with PBMC could be blocked by protein disulfide isomerase inhibitors 5,5Ј-dithiobis(nitrobenzoic acid) and 4-chloromercuribentenesulfonate (data not shown). However, we do not know whether disulfide interchange occurs at the level of inter-or intrachain disulfide bonds of the envelope.
The three-dimensional structure of gp120 complexed to a CD4 fragment, and the Fab fragment of a neutralizing antibody has been determined by x-ray diffraction (25). The model of a trimeric gp120/gp41 that was established from these data suggests that the CD4 binding site of the gp120 subunit is occluded by the V1 and V2 loops (1,12). Our results are in accordance with this model, as soluble trimeric envelope glycoproteins showed weak binding to plasma membrane expressed CD4 when compared with monomers. The fact that trimers have a lower binding capacity to CD4 has been inferred by earlier studies using soluble CD4 binding to gp120/41-expressing cells. These studies were based on the assumption that all plasma membrane-expressed gp120/gp41 is trimeric (67,68). Sattentau and co-workers (69) proposes that trimers might have a flexible structure that oscillates from an open (high affinity CD4 binding) to a closed state (low affinity CD4 binding). Alternatively, trimers on the virion or infected cell surface could transform to high affinity CD4 binding monomers upon contact with the target cell, followed by coreceptor binding and gp41-mediated fusion events. This hypothesis is coherent with recent data obtained by analytical ultracentrifugation studies of the gp41 ectodomain of Simian immunodeficiency virus, demonstrating a dynamic equilibrium between the monomeric and trimeric state of gp41 (66).