Previous studies have produced soluble HIV-1 envelope glycoproteins lacking a transmembrane region and containing a modified proteolytic cleavage site between gp120 and gp41 (4
). These soluble glycoproteins were expressed in mammalian cells using vaccinia virus vectors and were reported to form dimers and tetramers (4
). Equivalent gp140(−) constructs made in this study from both the primary YU2 and laboratory-adapted HXBc2 HIV-1 isolates were mainly monomeric. The reasons for this difference are unknown. The formation of gp140(−) oligomers may depend upon the high levels of expression achieved by vaccinia virus vectors; higher concentrations of the gp140(−) glycoproteins could allow weak interactions favoring dimer and tetramer formation.
In contrast to the gp140(−) glycoproteins, some of the soluble gp130 glycoproteins created in this study form relatively stable trimers. The functionally relevant form of the HIV-1 envelope glycoproteins is thought to be a trimer (10
). Several lines of evidence indicate that the gp130(−/GCN4) molecules are trimeric. First, the gp130(−/GCN4) glycoprotein, but not the gp130(−) glycoprotein, reacts with the DP178 peptide. The DP178 peptide intercalates into the hydrophobic grooves formed by a trimeric coiled coil (11
) but would not be expected to bind efficiently to dimeric or tetrameric coiled coils. Second, the placement of the cysteine-cysteine-glycine residues at gp41 positions 576 to 578 would favor intersubunit disulfide bond formation only in the context of a trimeric coiled coil (26
). In dimeric or tetrameric coiled coils, the distances between the Cβ atoms of the cysteines would be unfavorable for cross-linking of the subunits. Finally, gel filtration analysis indicates that the molecular weight of the gp130(−/GCN4) glycoprotein is consistent with that expected for a trimer.
Our work provides insights into the factors that influence the stability of soluble HIV-1 envelope glycoprotein trimers. These insights, in addition to providing practical guidance for producing tractable, soluble trimers, might also apply to the native, membrane-anchored envelope glycoprotein complex. The extension of the N-terminal gp41 coiled coil by the trimeric GCN4 sequence was required for the efficient production of stable, soluble trimers. The success of this approach suggests that, within the trimers, elements of the N-terminal coiled coil are formed and participate in intersubunit contacts. This assertion is supported by the observation that intersubunit disulfide bonds form when cysteines are placed in the d and e positions of the N-terminal gp41 heptad repeat (26
). These positions are located on the inner, hydrophobic face of the trimeric coiled coil and are separated by distances that are acceptable for disulfide bond formation (11
). That the N-terminal gp41 coiled coil is well formed along its length in the soluble gp130 trimers is supported by the ability of DP178, which corresponds to the C-terminal α-helix in the gp41 ectodomain, to bind these oligomers. C-terminal α-helical gp41 peptides have been shown to bind within a long hydrophobic groove created by the interaction of two N-terminal gp41 helices in the trimeric coiled coil (10
). The accommodation of the N-terminal gp41 coiled coil within a stable, soluble trimer supports a model in which this coiled coil exists, at least in part, within the complete HIV-1 envelope glycoprotein precursor and contributes to oligomerization. Disulfide bonds among the trimer subunits form as a result of the introduction of the cysteine pair at positions 576 and 577 of the complete HIV-1 gp160 envelope glycoprotein precursor, further supporting this model (26
). Perhaps the coiled coil of the HIV-1 fusion protein, unlike that of influenza virus hemagglutinin, does not need to undergo extensive conformational changes from a precursor state in order to form (5
Besides the addition of trimeric GCN4 sequence, another factor that exhibited a major influence on the stability of soluble HIV-1 envelope glycoprotein trimers was proteolytic cleavage at the gp120–gp41 junction. Regardless of the means by which the trimer subunits were associated, including the presence of covalent disulfide bonds in the gp41 subunit, cleaved proteins were monomeric. This was unexpected because at least a portion of the cleaved, membrane-associated HIV-1 envelope glycoproteins retains trimeric structure on native virions. The basis for this difference is unknown, but it may simply reflect the greater lability of soluble envelope glycoprotein trimers or, alternatively, it might be due to differences in the accommodation of the cleaved segments in the two contexts.
In the case of the gp130(−/GCN4) and gp130(−/CCG/GCN4) glycoproteins, a low level of proteolytic cleavage was observed despite alteration of two basic residues N-terminal to the cleavage site. It is uncertain whether cleavage occurred precisely at the natural site in these mutants, although the recognition of the cleaved gp120 glycoprotein by antibodies against gp120 C-terminal regions was comparable to that of wild-type gp120. Efforts to reduce the observed residual cleavage by further alteration of basic residues near the natural cleavage site did not succeed (data not shown).
Covalent linkage of the trimeric subunits through disulfide bond formation resulted in extremely stable oligomers that remained associated on SDS-polyacrylamide gels run under nonreducing conditions or in the presence of 1.5% β-ME. Although covalent linkage was neither necessary nor sufficient for the production of stable soluble trimers, it may prove useful in circumstances where trimers are subjected to harsh conditions.
During natural infection, the humoral response to the HIV-1 envelope glycoproteins consists of both nonneutralizing and neutralizing antibodies. Many of the nonneutralizing antibodies appear to be generated against shed, monomeric gp120 glycoproteins and do not bind efficiently to the functional envelope glycoprotein trimer (48
). The vast majority of the gp120 epitopes, including all of the neutralization epitopes examined, were present on the soluble gp130 trimers, where their exposure was similar to that seen on the monomeric gp120 glycoprotein. Differences between monomeric and trimeric envelope glycoproteins involved epitopes in the first (C1) and fifth (C5) conserved regions of gp120, which have been previously implicated in the interaction with gp41. The C11 antibody, which recognizes a discontinuous gp120 epitope with C1 and C5 components, precipitated the proteolytically cleaved, monomeric forms of the soluble glycoproteins more efficiently than any of the other antibodies studied. This is consistent with the idea that the C11 antibody was generated to a monomeric, soluble gp120 glycoprotein shed from virions or infected cells during natural HIV-1 infection (45
Recognition of the gp130 proteins by some antibodies directed against linear C1 and C5 epitopes was actually increased relative to recognition of the gp120 or gp140(−) monomers. Previous studies of the native HIV-1 gp120 monomer suggested that C1 and C5 sequences at the very N and C termini of the protein, respectively, were well exposed, whereas more-interior N- and C-terminal residues were less accessible to antibodies (44
). This is consistent with the known involvement of the interior C1 and C5 sequences in secondary structural elements of the gp120 core domains (37
). In our study, recognition of the mature, fully glycosylated gp120 monomer by antibodies against interior C1 regions (residues 61 to 70 and 111 to 120) and an interior C5 region (residues 461 to 470) was minimal (43
). By contrast, these regions were accessible to antibodies on the trimeric, but not the monomeric, forms of the soluble gp130 glycoproteins. These observations suggest that, in the formation of these trimers, the N- and C-terminal regions of gp120 are extended into more-exposed conformations than those assumed in the gp120, gp140, and gp130 monomers. Interestingly, the influenza virus HA1
glycoprotein N and C termini, which make extensive contacts with the HA2
transmembrane protein, also exhibit extended structures in the trimeric hemagglutinin complex (13
Another explanation for the differential recognition of the monomers and trimers by the C1-directed antibodies 133/290 and 135/9 and by the C5-directed antibodies M91 and CRA-1 is a potential difference in the glycosylation of monomeric and trimeric soluble glycoproteins. Although these antibodies did not recognize the fully glycosylated gp120 monomer, they did precipitate a faster-migrating form of gp120 that is presumably incompletely glycosylated. It is possible that soluble trimers are glycosylated differently than the monomeric proteins, contributing to better recognition by these antibodies.
Surprisingly, the formation of stable gp130 trimers was not sufficient to render all of the nonneutralizing gp120 epitopes inaccessible to antibodies. In fact, the linear C1 and C5 epitopes that are accessible only on the soluble gp130 trimers are not thought to be available for antibody binding in the context of the functional virion spike and, consequently, are not neutralization targets. Some of these differences between soluble gp130 glycoproteins and membrane-associated, native envelope glycoprotein trimers may be due to the presence of the glycosylated, C-terminal portion of the gp41 ectodomain or the viral membrane in the latter. Alternatively, the soluble gp130 glycoproteins may be trapped in a conformation different from that normally assumed by the envelope glycoproteins on the virion spike. Future studies of these and other issues will be expedited by the availability of stable, tractable forms of HIV-1 envelope glycoprotein trimers.