Most antibodies elicited against the HIV-1 envelope glycoproteins during natural infection or after vaccination are incapable of neutralizing HIV-1 infectivity in vitro (6, 25, 37, 40, 45, 57). While several such antibodies effectively neutralize viruses that are adapted to replicate in immortalized T-cell lines, only three monoclonal antibodies, IgG1b12, 2G12, and 2F5, neutralize a wide range of primary HIV-1 isolates (7, 43, 50). These three monoclonal antibodies exhibit a higher affinity for oligomeric HIV-1 envelope glycoproteins on viruses or cell surfaces than do most antibodies directed against the envelope glycoproteins (44, 45). To date, most recombinant HIV-1 glycoproteins tested as vaccine candidates have been gp120 monomers. The antibody responses to gp120 are not usually effective in neutralizing primary HIV-1 isolates (3, 4, 9, 25, 37, 52, 57). To attempt to mimic the native HIV-1 envelope glycoprotein oligomer, soluble gp140 glycoproteins containing gp120 and the gp41 ectodomains have been created (6, 16, 17). When the gp120-gp41 junction is modified to reduce proteolytic cleavage, these soluble gp140 glycoproteins assemble into dimers and tetramers in addition to the monomeric forms (6, 16, 17, 51). The elicitation of neutralizing antibodies by oligomeric forms of soluble gp140 has been disappointing, perhaps because these oligomers do not fully resemble the biologically relevant envelope glycoprotein trimers (16, 51).

Attempts to produce HIV-1 envelope glycoprotein trimers for structural and immunologic analysis have been frustrated by the lability of these glycoprotein complexes. Both the intersubunit interactions that promote trimer formation and the association between gp120 and gp41 are labile (24, 39). Modifications of the gp120-gp41 cleavage site and introduction of cysteine cross-links between gp120 and gp41 have been employed to address the latter problem (5, 17). However, as alluded to above, these approaches do not deal with the instability of the oligomeric associations or with the tendency of the HIV-1 envelope glycoprotein ectodomains to form dimers and tetramers. Two strategies for stabilizing trimeric interactions among HIV-1 envelope glycoprotein subunits have been devised (20, 60). The first strategy involves the introduction of cysteine residues into the gp41 N36 coiled coil, allowing the formation of disulfide bonds between the monomeric subunits (20). This approach results in the cross-linking of the membrane-anchored gp160 envelope glycoprotein precursor only when the cysteines are positioned along the N36 coiled coil in locations that allow disulfide bond formation in the context of a trimer. These cysteines are not sufficient to stabilize soluble forms of the HIV-1 envelope glycoproteins, probably because the lability of the soluble trimer is such that disulfide bonds do not have an opportunity to form (X. Yang, et al., unpublished data). The second strategy is the addition of a trimeric GCN4 motif to the carboxyl terminus of the soluble envelope glycoproteins (26, 60). In initial studies, these helical GCN4 sequences were added in register with the N36 helices, presumably extending the trimeric gp41 coiled coil and thus enhancing the stability of the trimer (60). Particularly when gp120-gp41 cleavage was eliminated, these soluble gp130 glycoproteins containing GCN4 sequences formed relatively stable and homogeneous trimers. In these molecules, introduction of cysteines into appropriate locations in the N36 coiled coil resulted in intersubunit disulfide bonds and even greater stability of the trimers. Although these soluble gp130 trimers may be useful for some studies, they lack gp41 ectodomain structures that are thought to play a role in the function and antigenicity of the HIV-1 envelope glycoproteins. One example is the disulfide-linked loop at positions 598 to 604 of gp41, which probably contributes to noncovalent interactions with the first (C1) or fifth (C5) conserved region of the gp120 glycoprotein (8, 27). A second example is the heavily glycosylated gp41 region from residues 611 to 637. This region is longer in the transmembrane envelope glycoproteins of lentiviruses than it is in those of other retroviruses, consistent with a potential role in immune evasion. A third example is the gp41 C34 helix, which is an integral part of the six-helix bundle thought to mediate virus-cell membrane fusion (9, 48, 54). Finally, the deleted regions in the soluble gp130 trimers encompass almost all of the major gp41 epitopes, including that recognized by the 2F5 neutralizing antibody (16, 43).

In this study, we used combinations of the three approaches described above (addition of trimeric GCN4 helices, modification of the gp120-gp41 cleavage site, and introduction of cysteines into the gp41 N36 region) to create stable, soluble gp140 trimers containing most or all of the gp41 ectodomain. Figure Figure1A1A shows the composition of the soluble glycoproteins used in this study. In the nomenclature used herein, “gp140” refers to an HIV-1 envelope glycoprotein (YU2 strain) truncated near the carboxyl terminus of the gp41 ectodomain. A minus sign in the parentheses after a construct name indicates the presence of arginine-to-serine changes at residues 508 and 511, which impair proteolytic cleavage between gp120 and gp41 sequences (60). “CCG” in the parentheses indicates the substitution of two cysteines and a glycine for residues 576 to 578, which substitution has been shown to allow disulfide cross-linking of HIV-1 envelope glycoprotein trimers (20, 60). “GCN4” in the parentheses indicates the addition of the trimeric GCN4 sequence (MKQIEDKIEEILSKIYHIENEIARIKKLIGEV) to the carboxyl terminus of the glycoprotein.

To examine the expression and properties of these soluble gp140 glycoproteins, transfected 293T cells transiently expressing these proteins were radiolabeled with ³⁵S-methionine and ³⁵S-cysteine. The cell supernatants were precipitated with 3 μl of pooled sera from HIV-1-infected individuals, and the precipitated proteins were resolved on a sodium dodecyl sulfate (SDS)–7.5% polyacrylamide gel after boiling for 3 min in 1× Laemmli sample buffer. As shown in Fig. Fig.1B,1B, the gp120 and gp140(−) glycoproteins migrated on the gel mainly as monomers. A small amount of dimeric gp140(−) glycoprotein was visible under these conditions. The addition of the trimeric GCN4 sequences resulted in the production of high-molecular-weight forms of the soluble gp140 glycoproteins that remained stable even after boiling and electrophoresis. These higher-order species migrated with an apparent molecular mass of approximately 400 kDa, based on extrapolation from the migration of the molecular weight markers and the apparent dimeric forms of the glycoproteins. Under identical conditions, the soluble gp130(−/GCN4) glycoprotein migrated primarily as a monomer, indicating that the soluble gp140 oligomers are more stable than the previously described soluble gp130 trimers (60). The soluble gp140 glycoproteins with a proteolytic cleavage site modification [gp140(−/GCN4) and gp140(−/CCG/GCN4)] exhibited few or no monomeric forms on the gel. The gp140(GCN4) and gp140(CCG/GCN4) glycoproteins, which contain an intact gp120-gp41 proteolytic cleavage site, were resolved into both oligomeric and monomeric forms. Significantly, the monomeric forms of the gp140(GCN4) and gp140(CCG/GCN4) glycoproteins migrated comparably to gp120, not gp140. This result suggests that the proteolytically cleaved soluble gp140 oligomers are less stable than the uncleaved forms of the same glycoprotein. Consistent with this interpretation, when the oligomeric forms of gp140(GCN4) and gp140(CCG/GCN4) glycoproteins were disrupted by being heated to 100°C for 8 min in sample buffer containing 5% β-mercaptoethanol (β-ME), they migrated alongside the gp140(−) glycoprotein (data not shown). Thus, as was previously observed for soluble gp130 trimers (60), elimination of gp120-gp41 cleavage contributes to the stability of the soluble oligomer.

To investigate whether intersubunit disulfide bonds formed in the gp140(−/CCG/GCN4) glycoprotein, compared with the negative and positive control glycoproteins [gp140(−/GCN4) and gp130(−/CCG/GCN4), respectively], the proteins were analyzed after boiling in sample buffer containing 2% β-ME. Under these reducing conditions, both the oligomeric gp140(−/CCG/GCN4) and gp130(−/CCG/GCN4) glycoproteins were more stable than the gp140(−/GCN4) oligomers (Fig. (Fig.1C).1C). This result is consistent with the formation of intersubunit disulfide bonds in at least some of the gp140(−/CCG/GCN4) glycoproteins.

To analyze the soluble gp140 glycoproteins under conditions more gentle than those described above, the radiolabeled glycoproteins were concentrated and resolved by velocity centrifugation in 10 ml of 10 to 25% continuous sucrose gradients. Ten fractions were collected from each gradient, and the envelope glycoproteins in each fraction were detected by precipitation by pooled sera from HIV-1-infected individuals. The precipitates were analyzed on reducing SDS-polyacrylamide gels. As shown in Fig. Fig.1D,1D, the gp140(−) glycoprotein sedimented in fractions 4 to 8, with the majority of the protein in fractions 6 and 7. The uncleaved portion of the gp130(−/GCN4) glycoprotein, which has been shown to be trimeric and was included as a positive control, sedimented in fractions 1 to 5, with most of these forms in fractions 2 and 3. The proteolytically cleaved gp130(−/GCN4) glycoprotein, which is known to consist of monomeric gp120, sedimented in fractions 6 and 7. Likewise, the uncleaved forms of the gp140 (GCN4) and gp140 (CCG/GCN4) glycoproteins sedimented in fractions 1 to 4, whereas the cleaved forms of these glycoproteins sedimented in fractions 6 and/or 7. The vast majority of the gp140(−/GCN4) and gp140(−/CCG/GCN4) glycoproteins, which contain gp120-gp41 cleavage site modifications, sedimented in fractions 1 to 4. These results support the notion that the stability of higher-order forms of the soluble gp140 glycoproteins is enhanced by the GCN4 sequences and by alteration of the proteolytic cleavage site.

To obtain a more accurate estimate of the molecular weight of the higher order products, the gp120 and gp140(−/GCN4) glycoproteins were affinity purified and analyzed by molecular size exclusion chromatography. The proteins were highly pure, as determined by Coomassie blue staining of an SDS-polyacrylamide gel (Fig. (Fig.1E,1E, left panel). The gp120 monomers were eluted at an apparent size of about 175 kDa, larger than the expected molecular weight but consistent with previously published results (5) (Fig. (Fig.1E,1E, middle panel). The gp140(−/GCN4) proteins were eluted as one minor and two major peaks (Fig. (Fig.1E,1E, right panel). The minor peak was eluted at a rate close to that of the gp120 monomers. The first major peak was eluted earlier than the largest molecular size marker (a 669-kDa protein) and is likely composed of higher-order aggregates. The second major peak was eluted at a rate expected for a 410-kDa protein, consistent with the molecular weight of a trimer.

The functional integrity of the gp120 subunits in the gp140 trimers was assessed by testing their ability to bind the natural ligands, CD4 and CCR5. As previously shown (50, 58), the HIV-1 YU2 gp120 glycoprotein bound CCR5 on a cell surface much more efficiently in the presence of soluble CD4 (sCD4) (Fig. (Fig.2).2). Similarly, the gp140(−/GCN4) glycoprotein bound CCR5 in a manner dependent on the presence of sCD4. The CCR5 binding was not as efficient for the gp140(−/GCN4) protein as for the gp120 monomer. Qualitatively, the gp120 subunits of the soluble gp140 trimer retained the ability to bind the CD4 receptor and undergo the structural changes necessary for subsequent CCR5 binding.

To examine whether soluble gp140 trimers could be created for another HIV-1 strain, the gp140(−/GCN4) glycoprotein was constructed using the envelope glycoproteins derived from the X4 virus HXBc2. The properties of the gp140(−/GCN4) glycoproteins of the YU2 and HXBc2 strains were indistinguishable on SDS-polyacrylamide gels and sucrose density gradients (data not shown). These results suggest that modification of the gp120-gp41 proteolytic cleavage site and addition of GCN4 trimeric sequences can be used to create stable trimers from several HIV-1 strains.

Several studies have suggested that HIV-1-neutralizing antibodies bind the oligomeric envelope glycoprotein complex more efficiently than nonneutralizing antibodies. To examine the relative exposure of neutralizing and nonneutralizing antibody epitopes on the soluble gp140 glycoproteins, the recognition of the YU2 gp140(−/GCN4) glycoprotein by a panel of anti-gp120 monoclonal antibodies was compared with the recognition of the gp120 and gp140(−) glycoproteins. The radiolabeled glycoproteins were precipitated by saturating amounts of either pooled sera from HIV-1-infected individuals or the monoclonal antibodies. The precipitation with pooled sera, which recognize a variety of HIV-1 envelope glycoprotein epitopes, controls for the relative amount of the three glycoproteins available for precipitation by antibodies (Fig. (Fig.3A,3A, left panel). The monoclonal antibodies could be divided into three groups, based on their recognition of the gp140(−/GCN4) glycoprotein, relative to that of the gp120 and gp140(−) glycoproteins. The first group included neutralizing antibodies directed against the CD4 binding site (F105 and F91) or against CD4-induced epitopes (17b and 48d). These antibodies bound at least as well to the gp140(−/GCN4) glycoprotein as to the gp120 and gp140(−) glycoproteins (Fig. (Fig.3A,3A, middle panel and data not shown). The second group included nonneutralizing antibodies (C11, A32, 522-149, M90, and #45). These antibodies bound the gp140(−/GCN4) glycoprotein at slightly reduced levels compared with the gp120 or gp140(−) glycoprotein (Fig. (Fig.3A,3A, right panel and data not shown). The third group included antibodies (30D, 60D, and 522-149) directed against the extreme N and C termini of gp120. As had been previously observed for soluble gp130 glycoproteins, these antibodies recognized the gp140(−/GCN4) glycoprotein as efficiently as they did the gp120 glycoprotein (data not shown).

The above assays were conducted under conditions of high antibody concentration, which can obscure differences in affinity. To analyze the relative affinity of anti-gp120 antibodies more precisely, mixtures of equivalent amounts of the gp120 and gp140(−/GCN4) glycoproteins were precipitated either by pooled sera from HIV-1-infected individuals or by monoclonal antibodies. The ³⁵S-methionine- and ³⁵S-cysteine-labeled gp120 and gp140(−/GCN4) glycoproteins to be added to the mixture were first quantified by precipitations with an excess of pooled sera from HIV-1-infected individuals followed by PhosphorImager (Molecular Dynamics) analysis. Equivalent amounts of the two glycoproteins were mixed and precipitated at room temperature with either 3 μl of pooled HIV-1-positive sera, 1 μg of monoclonal antibody, or 1 μl of ascites in a total volume of 500 μl. The precipitated glycoproteins were run on an SDS-polyacrylamide gel, and the ratio of gp140(−/GCN4) to gp120 glycoprotein was calculated after PhosphorImager analysis. The relative affinity represents the gp140/gp120 ratio normalized to that obtained by precipitation with the pooled sera from HIV-1-infected individuals. Parallel studies were also conducted for the gp140(−) glycoprotein. Figure Figure3B3B shows that there were, at best, modest differences between neutralizing (F105, F91, 17b, and 48d) and nonneutralizing (C11, A32, and 30D) antibodies in their relative affinity for the gp140(−) glycoprotein. By contrast, the relative affinity of the neutralizing antibodies for the gp140(−/GCN4) glycoprotein was substantially higher than that of the nonneutralizing antibodies. These results suggest that, compared with the monomeric gp120 glycoprotein, the trimeric gp140(−/GCN4) glycoprotein exhibits some degree of occlusion or disruption of its nonneutralizing epitopes.

In our previous analysis of soluble gp130 trimers, some linear epitopes near the extreme N and C termini of gp120 were more accessible to monoclonal antibodies in the gp130 trimers than in gp120 monomers (60). This group of monoclonal antibodies includes 135/5 and 133/290, which are directed against linear sequences in the C1 region, and CRA-1 and M91, which recognize linear epitopes in the C5 region. These antibodies were tested for the ability to precipitate the soluble gp140 glycoproteins at saturating antibody concentrations (Fig. (Fig.3C).3C). With similar amounts of input glycoproteins, as judged by precipitation by pooled sera from HIV-1-infected individuals (Fig. (Fig.3C,3C, left panel), these monoclonal antibodies, including 135/9 (Fig. (Fig.3C,3C, middle panel), M91 (Fig. (Fig.3C,3C, right panel), 133/290, and CRA-1 (data not shown), preferentially precipitated gp140(−/GCN4) trimers compared with gp140(−) monomers. Of note, the efficiency of precipitation of the trimeric gp140(−/GCN4) and gp130(−/GCN4) glycoproteins by these antibodies was very low compared with that seen for the pooled HIV-1 positive sera (the left panel of Fig. Fig.3C3C was exposed to film overnight, whereas the middle and right blots were exposed to film 6 days). Thus, although these antibodies exhibit preferential recognition of trimeric forms of the HIV-1 envelope glycoproteins, the absolute efficiency of this recognition is very low.

To explore the integrity and exposure of gp41 ectodomain epitopes in the soluble gp140 glycoproteins, a panel of monoclonal antibodies directed against linear and discontinuous epitopes in the gp41 ectodomain (16) was used to precipitate the YU2 gp140(−) and gp140(−/GCN4) glycoproteins (Fig. (Fig.4A).4A). Compared with the pooled sera from HIV-1-infected individuals, the anti-gp41 monoclonal antibodies precipitated the gp140(−) and gp140(−/GCN4) glycoproteins inefficiently (the pooled serum precipitates were exposed to film overnight, whereas the others were exposed for 3 days). Nonetheless, the recognition of the gp140(−/GCN4) glycoprotein, relative to that of the gp140(−) glycoprotein, was equivalent for the T4 and D12 antibodies and significantly greater for the T3 and D50 antibodies. Although the specific nature of these epitopes remains to be characterized, the results do indicate that some previously defined gp41 ectodomain epitopes are present and exposed on at least a fraction of the gp140(−/GCN4) glycoprotein preparation. We also attempted to precipitate the gp140(−) and gp140(−/GCN4) glycoproteins with the 2F5 neutralizing anti-gp41 antibody. However, neither glycoprotein was efficiently recognized by the 2F5 antibody, probably due to the polymorphism in the gp41 epitope (ALDKWA instead of ELDKWA) in the YU2 strain.

The formation of intersubunit disulfide bonds in the gp140(−/CCG/GCN4) glycoprotein (see above) suggested that some elements of the gp41 N36 coiled coil may be formed in the soluble gp140 trimers. We previously showed that the N36 coiled coil not only was formed in the soluble gp130 trimers but also was accessible to a peptide corresponding to the C34 region of gp41 (60). To examine this aspect of the soluble gp140 trimers, the recognition of the gp140(−), gp140(−/GCN4), and gp130(−/GCN4) glycoproteins by the C34-like peptide, DP178, was tested. For this purpose, we used the PK-C299 peptide which, in addition to the C34-like DP178 sequence, contains a C-terminal C9 tag that can be recognized by the 1D4 monoclonal antibody. Similar amounts of the radiolabeled glycoproteins were incubated with a 50-fold molar excess of the PK-C299 peptide and the 1D4 antibody. Figure Figure4B4B shows that only the gp130(−/GCN4) glycoprotein was efficiently precipitated by this procedure. These results indicate that the carboxy-terminal gp41 ectodomain sequences in the gp140(−/GCN4) glycoprotein either directly or indirectly limit the binding of the PK-C299 peptide–1D4 antibody complex to the soluble trimer.

To examine whether any of the soluble gp140 glycoproteins can generate six-helix bundles corresponding to the fusogenic conformation, the recognition of the glycoproteins by the NC-1 monoclonal antibody was examined. The NC-1 antibody was elicited by immunization with an HIV-1 six-helix bundle peptide complex and specifically recognizes the six-helix bundle structure (30). The Δ528 glycoprotein was included as a positive control in these experiments. The Δ528 glycoprotein contains a heterologous signal sequence from tissue plasminogen activator fused with residues 529 to 679 of the HIV-1 HXBc2 envelope glycoproteins. Thus, the Δ528 glycoprotein represents a soluble form of the gp41 glycoprotein, including the N36 and C34 helices. A similar construct derived from the simian immunodeficiency virus envelope glycoproteins has been shown to form a six-helix bundle with high stability (61). The radiolabeled Δ528, gp120, gp130(−/GCN4), and soluble gp140 glycoproteins were precipitated at room temperature by either the pooled sera from HIV-1-infected individuals or the NC-1 antibody. Figure Figure4C4C shows the percentage of the glycoproteins precipitated by the NC-1 antibody relative to the amounts precipitated by the pooled HIV-1-positive sera. More than 87% of the positive control Δ528 glycoprotein recognized by the pooled sera was precipitated by the NC-1 antibody, as expected. The gp120 and gp140(−) monomers were not able to bind to the NC-1 monoclonal antibody, also as expected. The gp130(−/GCN4) glycoprotein was not detectably precipitated by the NC-1 antibody, consistent with the expectation that the absence of the C34 helices would preclude formation of the six-helix bundle. Approximately 16% of the gp140(−/GCN4) glycoprotein [labeled gp140Δ675(−/GCN4) in Fig. Fig.4C]4C] recognized by the pooled sera was precipitated by the NC-1 antibody. This result indicates that at least a small fraction of the gp140(−/GCN4) glycoprotein assumes a fusogenic conformation that includes the gp41 six-helix bundle.

To examine whether the precise position of the trimeric GCN4 sequence influences the characteristics of the soluble gp140 glycoproteins, the original gp140(−/GCN4) construct, also designated gp140Δ675(−/GCN4), was compared with two new constructs, gp140Δ655(−/GCN4) and gp140Δ683(−/GCN4). In the gp140Δ655(−/GCN4) glycoprotein, the GCN4 sequences are placed carboxy terminal to residue 655, within the C34 sequence of gp41 (Fig. (Fig.5A).5A). In the gp140Δ683(−/GCN4) glycoprotein, the GCN4 sequences are placed carboxy terminal to residue 683, which is believed to represent the boundary of the gp41 ectodomain and the transmembrane region. Precipitates of these glycoproteins by pooled HIV-1-positive sera on a nonreducing SDS-polyacrylamide gel are shown in Fig. Fig.5B.5B. Higher-order forms consistent with dimers and trimers were evident for the gp140Δ675(−/GCN4), gp140Δ655(−/GCN4) and gp140Δ683(−/GCN4) glycoproteins under these conditions. Unexpectedly, when the precipitates were boiled in sample buffer containing 2% β-ME, the reduced gp140Δ683(−/GCN4) monomers migrated similarly to the gp140(−) monomers and the reduced gp140Δ675(−/GCN4) and gp140Δ655(−/GCN4) monomers migrated faster than the gp140(−) monomers (Fig. (Fig.5C,5C, left panel). When the precipitates were treated with a mixture of endoglycosidase F and N-glycosidase F, all four proteins migrated at their expected relative positions (Fig. (Fig.5C,5C, right panel). These results indicate that the gp140(−) glycoprotein contains a higher proportion of N-linked carbohydrate than the other three glycoproteins. The gp140Δ675(−/GCN4), gp140Δ655(−/GCN4), and gp140Δ683(−/GCN4) glycoproteins sedimented indistinguishably on sucrose density gradients (data not shown). The recognition of the gp140Δ683(−/GCN4) glycoprotein by the NC-1 antibody was less than that of the gp140Δ675(−/GCN4) glycoprotein (Fig. (Fig.4C),4C), indicating a slightly decreased propensity for the former glycoprotein to form six-helix bundles. As expected from the disruption of the C34 helix, the gp140Δ655(−/GCN4) glycoprotein was not precipitated by the NC-1 antibody (Fig. (Fig.4C).4C).

The recognition of the gp140Δ675(−/GCN4), gp140Δ655(−/GCN4), and gp140Δ683(−/GCN4) glycoproteins by saturating concentrations of anti-gp120 and anti-gp41 antibodies was examined. No differences among the three glycoproteins were detected (data not shown). As was observed for the gp120Δ675(−/GCN4) glycoprotein, the gp140Δ655(−/GCN4) and gp140Δ683(−/GCN4) glycoproteins were not efficiently precipitated by the PK-C299 peptide–1D4 antibody complex (data not shown).

The relative recognition of the soluble gp140 variants and the YU2 gp120 monomer by neutralizing and nonneutralizing antibodies was assessed in a direct competition assay similar to that used to generate the data shown in Fig. Fig.3B.3B. The relative affinities of antibodies for the gp140Δ655(−/GCN4) glycoprotein were identical to those of antibodies for the gp140Δ675(−/GCN4) glycoprotein (data not shown). Compared with the gp140Δ675(−/GCN4) glycoprotein, the gp140Δ683(−/GCN4) glycoprotein exhibited slight increases in the relative affinities of neutralizing antibodies and slight reductions in the relative affinities of some nonneutralizing antibodies (compare Fig. Fig.5D5D with Fig. Fig.3B).3B). Although the significance of these small differences is unclear, the results indicate that the gp140(−/GCN4) variants exhibit only subtle conformational deviations from a common general structure.

Our previous work demonstrated that stable, soluble gp130 trimers of HIV-1 glycoproteins could be created by the addition of the trimeric GCN4 motif to the carboxyl terminus in combination with modification of the gp120-gp41 cleavage site (60). The introduction of cysteines in specific sites within the N36 region of these molecules resulted in the formation of intersubunit disulfide bonds that further stabilized the soluble gp130 trimers. In this study, similar approaches were applied to stabilize trimers containing the complete HIV-1 envelope glycoprotein ectodomains. In the context of the soluble gp140 glycoproteins, as was seen for the soluble gp130 constructs, the addition of carboxy-terminal GCN4 sequences and disruption of the gp120-gp41 cleavage site were necessary and sufficient for the production of relatively homogeneous trimers. When the wild-type cleavage site was present in the glycoprotein, only molecules that bypassed the proteolytic cleavage event remained oligomeric under the conditions employed in our assays. The stability of the gp140(−/GCN4) trimers is impressive; trimers were observed even after boiling in nonreducing buffers, SDS-polyacrylamide gel electrophoresis, and elution in high salt (3 M MgCl₂) during immunoaffinity purification (Fig. (Fig.1B1B and E). The soluble gp140 trimers exhibited superior stability compared with the soluble gp130 trimers, suggesting the presence of additional intersubunit molecular contacts in the former glycoproteins. Based on previous mutagenic and structural studies of the gp41 glycoprotein (8, 61), the gp140 constructs, relative to the gp130 constructs, would also be expected to retain more of the regions important for interaction with the gp120 moieties in the oligomer.

It is likely that, in the absence of the GCN4 sequences, cleavage-defective gp140 glycoproteins can form oligomeric structures, especially when such proteins are produced at high concentrations favoring weak intermolecular interactions (16). Although gp140(−) glycoproteins produced in our system and analyzed under our conditions were mainly monomeric, we occasionally observed some stable dimer formation (shown, for example, in Fig. Fig.5B).5B). This is consistent with several studies that report the production of dimeric and tetrameric soluble, cleavage-defective HIV-1 envelope glycoproteins (16, 17). Thus, while the trimeric GCN4 sequences may not be required in all contexts for the formation of soluble envelope glycoprotein oligomers, they are apparently critical for the assembly and maintenance of relatively homogeneous trimers.

Several observations support the assertion that a major fraction of the gp140(−/GCN4) glycoprotein is indeed trimeric. First, by molecular size exclusion chromatography and, less accurately, by SDS-polyacrylamide gel electrophoresis, the oligomers exhibited an apparent molecular mass of 400 to 410 kDa, consistent with the presence of three gp140 subunits. Moreover, in velocity gradients, the gp140(−/GCN4) glycoprotein sedimented at rates similar to those of the gp130(−/GCN4) glycoprotein, which is documented to be trimeric (60). Second, the presence of the CCG substitution at positions 576 to 578 in the gp41 N36 sequence resulted in increased stability of the gp140(−/CCG/GCN4) oligomers compared with the gp140(−/GCN4) oligomers in the presence of a reducing agent (Fig. (Fig.1C).1C). This observation is consistent with the formation of oligomer-stabilizing, intersubunit disulfide bonds, as has been previously seen for the HIV-1 gp160 envelope glycoprotein precursor and for soluble gp130 trimers (20, 60). Because the cysteine residues at positions 576 and 577 (at the d and e positions of a coiled-coil heptad repeat) are likely to form intersubunit disulfide bonds only in a trimeric context (20), the observed stability associated with the presence of these cysteines supports a trimeric model. Third, a detectable portion of some of the soluble gp140 oligomers was precipitated by the NC-1 antibody, which recognizes a six-helix bundle stabilized by the trimeric association of gp41 N36 helices. The NC-1 antibody recognized only the soluble gp140 glycoproteins [gp140Δ675(−/GCN4) and gp140Δ683(−/GCN4)] that both formed stable oligomers and contained intact N36 and C34 regions. Fourth, other antibodies, such as 135/9 and M91, which were previously shown to exhibit a strong preference for trimeric gp130 forms (60), also recognized the soluble gp140 oligomers. Finally, all of the above observations are consistent with the well-documented propensity of the modified GCN4 sequences used herein to form trimers (26). In the context of the fusion proteins studied here, the GCN4 sequences probably promote the weak trimeric interactions among the HIV-1 envelope glycoprotein components of the oligomeric complex.