Mutations That Destabilize the gp41 Core Are Determinants for Stabilizing the Simian Immunodeficiency Virus-CPmac Envelope Glycoprotein Complex*

The human and simian immunodeficiency viruses (HIV and SIV) envelope glycoprotein consists of a trimer of two noncovalently and weakly associated subunits, gp120 and gp41. Upon binding of gp120 to cellular receptors, this labile native envelope complex undergoes conformational changes, resulting in a stable trimer-of-hairpins structure in gp41. Formation of the hairpin structure is thought to mediate membrane fusion by placing the viral and cellular membranes in close proximity. An in vitro-derived variant of SIVmac251, denoted CPmac, has acquired an unusually stable virion-associated gp120-gp41 complex. This unique phenotype is conferred by five amino acid substitutions in the gp41 ectodomain. Here we characterize the structural and physicochemical properties of the N40(L6)C38 model of the CPmac gp41 core. The 1.7-Å resolution crystal structure of N40(L6)C38 is very similar to the six-helix bundle structure present in the parent SIVmac251 gp41. In both structures, three N40 peptides form a central three-stranded coiled coil, and three C38 peptides pack in an antiparallel orientation into hydrophobic grooves on the coiled-coil surface. Thermal unfolding studies show that the CPmac mutations destabilize the SIVmac251 six-helix bundle by 15 kJ/mol. Our results suggest that the formation of the gp41 trimer-of-hairpins structure is thermodynamically coupled to the conformational stability of the native envelope glycoprotein and raise the intriguing possibility that introduction of mutations to destabilize the six-helix bundle may lead to the stabilization of the trimeric gp120-gp41 complex. This study suggests a potential strategy for the production of stably folded envelope protein immunogens for HIV vaccine development.


For human and simian immunodeficiency viruses (HIV 1 and SIV), infection of cells is initiated by fusion of the viral and
cellular membranes. This membrane fusion process is mediated by the viral envelope glycoprotein and receptors on the target cell. The envelope glycoprotein is synthesized as the precursor gp160 that is processed by proteolytic cleavage to yield two noncovalently associated subunits, gp120 and gp41, which are organized on the viral surface as oligomeric spikes (1,2). gp120 determines viral tropism by binding to the cell surface receptor CD4 and a chemokine coreceptor such as CCR-5 or CXCR-4 (Refs. 3 and 4; see also Ref. 5). The transmembrane gp41 subunit is responsible for mediating fusion between the viral and cellular membranes, a process that results in penetration of the viral genome into the host cell. The native (prefusogenic) envelope conformation, as found on the surface of the infectious virion, is inactive in fusion. Binding of gp120/gp41 to CD4 induces initial conformational changes in gp120 that expose the chemokine receptor binding site, and the subsequent binding of gp120 to the chemokine receptor initiates structural rearrangements in gp41 that promote viruscell membrane fusion (reviewed in Ref. 6). Recently, there has been an influx of structural information on core fragments of gp120 in the CD4-associated configuration and the gp41 ectodomain core (3,6). However, it has not yet been possible to crystallize the native envelope glycoprotein complex, and thus the structural consequences of its conformational activation are not well understood.
The HIV-1 envelope glycoprotein is a primary target for vaccine development because it is the major virus neutralization antigen during natural viral infection. For example, the three most potent and broadly neutralizing anti-HIV-1 antibodies yet identified, IgG1b12, 2G12, and 2F5, are known to interfere with virus-cell attachment and fusion (7)(8)(9)(10)(11)(12)(13)(14)(15)(16). However, attempts to induce neutralizing antibodies after vaccination with individual gp120 and gp41 subunits have been disappointing. Recent studies suggest that an immune response to the native trimeric envelope complex rather than to viral debris (i.e. nonnative forms of the envelope glycoprotein such as uncleaved gp160 precursor, dissociated gp120, or gp41 ectodomain) leads to the production of neutralizing antibodies (17)(18)(19)(20). Since the gp120 molecule is readily dissociated or shed from virions because of its labile, noncovalent association with the gp41 ectodomain (21)(22)(23)(24), a major challenge in the development of an effective HIV immunogen is to preserve the native envelope glycoprotein structure or crucial components thereof in vaccine preparations (reviewed in Ref. 25). Such efforts have to date relied on approaches to stabilize the trimeric envelope protein complex through incorporation of disulfide bonds between the gp41 ectodomains or between the gp120 and gp41 subunits (26 -29). A safe and effective HIV-1 vaccine is not yet available, making it important to understand the factors that influence the native envelope conformation.
Recent work has shown that a laboratory-adapted variant of SIVmac251, termed CPmac, exhibits increased numbers of envelope glycoprotein spikes on virions and a remarkably stable association of the gp120 and gp41 subunits (31). In contrast to previously described HIV isolates, gp120 and gp41 remained associated during detergent lysis of virions and could be coimmunoprecipitated by monoclonal antibodies to either gp120 or gp41. Quantitative analysis of virions by high performance liquid chromatography using chimeras between CPmac and an isogenic derivative of SIVmac251, termed NCmac, revealed a high gp120 content on CPmac virions (31). 2 A comparison of the CPmac and NCmac envelope glycoproteins mapped changes responsible for this phenotype to five amino acid mutations in the gp41 ectodomain (31). These mutations are Leu-557 to Ser, Val-571 to Leu, Lys-573 to Thr, Thr-586 to Ile, and Glu-641 to Lys substitutions (Fig. 1). Interestingly, the five mutations are either in the N-terminal heptad-repeat region or the C-terminal heptad-repeat region (see below). The locations of these mutations suggest that interactions between the two ␣-helical coiled-coil segments may contribute to the biological phenotype of the CPmac envelope glycoprotein. To lay the groundwork for efforts to define the structural determinants of the stable gp120-gp41 association in CPmac, we sought to delineate the role of the five CPmac mutations in the folding, thermodynamics, and conformation of the gp41 ectodomain.
The transmembrane gp41 protein contains several notable features within its ectodomain that are critical for membrane fusion activity (Fig. 1A). A hydrophobic, glycine-rich sequence referred to as the fusion peptide is located at the extreme NH 2 terminus of gp41. This fusion peptide region is known, in the case of influenza HA 2 , to insert into the target membrane during the fusion process (32,33). A region following the fusion peptide has a high ␣-helical propensity and a 4-3 heptad repeat of hydrophobic residues, a sequence feature characteristic of coiled coils (34 -36). This N-terminal heptad-repeat region is followed by a disulfide-bonded loop and then a C-terminal heptad repeat that precedes the transmembrane domain anchoring the envelope glycoprotein to the viral membrane (34 -36). Biochemical and structural analyses of a protease-resistant core of gp41 indicate that these two heptad-repeat regions form a six-helix bundle in which three antiparallel C-terminal ␣-helices pack into hydrophobic grooves on the surface of an N-terminal trimeric coiled coil (37)(38)(39)(40)(41)(42)(43)(44)(45). This six-helix bundle structure probably corresponds to the core of the fusion-active trimer-of-hairpins conformation (37,46,47), since the hairpin formation is structurally coupled to the apposition of viral and cellular membranes and the actual membrane fusion event (41, 48 -50). Interestingly, the trimer-of-hairpins motif is emerging as a general feature of many viral membrane fusion proteins (cf. Refs. 51 and 52). Peptide inhibitors that interfere with formation of gp41 hairpin structure can inhibit HIV-1 infection (37,(53)(54)(55)(56)(57)(58)(59). Thus, it would appear that the trimer-of-hairpins element plays a critical role in promoting membrane fusion and controls the key process of viral entry.
In this study, we have characterized the effects of the CPmac mutations on the overall structure and stability of the gp41 ectodomain core. The crystal structure of the CPmac gp41 core solved to 1.7-Å resolution reveals that the mutant sequences are accommodated in the six-helix bundle structure. The stability of the CPmac six-helix bundle decreases by 15 kJ/mol, relative to the isogenic SIVmac 251 complex. Formation of the gp41 hairpin structure has been proposed to provide an important thermodynamic driving force for the conformational activation of the envelope glycoprotein (37, 60). Our results are 2 J. A. Hoxie and E. Chertova, unpublished data. consistent with the hypothesis that the native envelope conformation can be stabilized by mutations that destabilize the fusion-active gp41 core structure. By extension, our results may also offer a simple and effective approach to enhancing thermostability of soluble HIV envelope glycoprotein trimers that might serve as useful immunogens in efforts to develop an AIDS vaccine.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-The gp41 ectodomain segment (representing residues 555-674 of full-length gp160) from the env gene of the SIVmac239 isolate (61) was amplified using primers containing NdeI and BamHI restriction sites and ligated into the NdeI and BamHI sites of the pAED4 vector (62) to make pSIVegp41. Plasmid pN40/ C38 NC , encoding the SIVmac251 N40(L6)C38 protein, was derived from pSIVegp41 by the insertion of the appropriate DNA sequences encoding the linker residues Ser-Gly-Gly-Arg-Gly-Gly between the COOH terminus of the N40 segment (residues 555-594) and the NH 2 terminus of the C38 segment (residues 637-674). Five CPmac substitutions were simultaneously and individually introduced into the N40 and C38 segments in pN40/C38 NC using the method of Kunkel et al. (63) and verified by DNA sequencing. All recombinant proteins were expressed in Escherichia coli strain BL21(DE3)/pLysS (Novagen). Cells were grown at 37°C in LB medium to an optical density of 0.8 at 600 nm and induced with isopropylthio-␤-D-galactoside for 3-4 h. Cells were lysed at 0°C by glacial acetic acid. The bacterial lysate was centrifuged (35,000 ϫ g for 30 min) to separate the soluble fraction from inclusion bodies. The soluble fraction, containing denatured N40(L6)C38, was dialyzed into 5% acetic acid overnight at room temperature. Proteins were purified from the soluble fraction to homogeneity by reverse-phase high performance liquid chromatography (Waters) on a Vydac C-18 preparative column (Hesperia, CA), using a water-acetonitrile gradient in the presence of 0.1% trifluoroacetic acid and lyophilized. The molecular weights of each protein were verified by using matrix-assisted laser desorption ionization-time-of-flight mass spectrometry (PerSeptive Biosystems). The concentrations of each protein solution were determined by using tyrosine and tryptophan absorbance at 280 nm in 6 M guanidinium hydrochloride (GdmCl) (64).
Circular Dichroism Spectroscopy-CD experiments were performed on an Aviv 62A DS circular dichroism spectrometer. The wavelength dependence of molar ellipticity, [], was monitored at 4°C on a 10 M protein solution in Tris-buffered saline (TBS) (50 mM Tris-HCl (pH 8.0), 100 mM NaCl) from 200 to 260 nm in a 0.1-cm path length cuvette. Helix content and the residue ellipticity, [], were calculated by the method of Chen et al. (65). Thermal stability was determined by monitoring the change in [] 222 as a function of temperature. Thermal melts were performed in 2°C increments with an equilibration time of 2 min at the desired temperature and an integration time of 30 s, using a 1-cm path length cuvette. All melts were reversible. Superimposable folding and unfolding curves were observed, and Ͼ90% of the signal was regained upon cooling. The melting temperatures, or midpoints of the cooperative thermal unfolding transitions (T m ) were determined from the maximum of the first derivative, with respect to the reciprocal of the temperature, of the [] 222 values (66). The error in estimation of T m is Ϯ0.5°C. For most of the proteins with a melting temperature of 100°C or higher in TBS, thermal denaturation experiments were performed in TBS in the presence of 2 or 3 M GdmCl.
Sedimentation Equilibrium-Sedimentation equilibrium measurements were performed on a Beckman XL-A (Beckman Coulter) analytical ultracentrifuge equipped with an An-60 Ti rotor (Beckman Coulter). Protein solutions were dialyzed overnight against TBS (pH 8.0); loaded at initial concentrations of 10, 30, and 100 M; and analyzed at rotor speeds at 17,000 and 20,000 rpm at 20°C. Data sets were fitted simultaneously to a single-species model of ln(absorbance) versus (radical distances) 2 using the program NONLIN (67). Protein partial specific volume and solvent density were calculated as described by Laue et al. (68). The molecular weights of N40(L6)C38 variants, except for T586I, were all within 10% of those calculated for an ideal trimer, with no systematic deviation of the residuals.
Crystallization, Data Collection, and Structure Determination-Crystals were obtained using the handing drop method of vapor diffusion by equilibrating 2-l drops (protein solution mixed 1:1 with reservoir solution) against a reservoir solution at room temperature. Initial crystallization conditions were screened by using sparse matrix crystallization kits (Crystal Screen I and II; Hampton Research, Inc.) and then optimized. Hexagonal crystals of NCmac-N40(L6)C38 were grown from 20 mg/ml protein, 0.6 M sodium chloride, 0.04 M barium chloride dihydrate, and 0.01 M hexadecyltrimethylammonium bromide. Rhombohedral crystals of CPmac-N40(L6)C38 were obtained from 15 mg/ml protein and 3.8 M sodium formate. All crystals were transferred to a cryoprotectant solution containing 15% (v/v) glycerol in the corresponding mother liquor. Crystals were mounted in nylon loops (Hampton Research, Riverside, CA) and flash-frozen in liquid nitrogen. Diffraction data on NCmac-N40(L6)C38 were recorded at 100 K at the Cornell High Energy Synchrotron Source on beamline F1 that incorporates an ADSC Quantum-4 CCD detector. Data on CPmac-N40(L6)C38 were collected at 100 K at beamline X25 at the Brookhaven National Laboratory National Synchrotron Light Source using a Brandeis B4 CCD detector.
Diffraction intensities were integrated by using DENZO and SCALE-PACK software (69) and reduced to structural factors with the program TRUNCATE from the CCP4 program suite (70). The structures of the N40(L6)C38 complexes were solved by molecular replacement with the program AMORE (71). A monomeric model of the N40-C38 complex derived from the crystal structure of the SIV gp41 ectodomain (45) was used as a search model. Density interpretation and model building were carried out with the O and CNS programs (72,73). Iterative rounds of manual rebuilding and refinement increased the quality of the initial electron density maps and served to reduce model bias, as guided chiefly by improvement in the free R factor. Crystallographic refinements were done with the program CNS 1.0 (74). Noncrystallographic symmetry constraints and restraints were applied during the first few cycles of refinement of the NCmac-N40(L6)C38 structure but then were removed and not used in the final refinement. Since molecular replacement methods can cause a final model to be biased by the search model, the current N40(L6)C38 models were verified by simulated annealing omit maps. The quality of coordinates was examined by PROCHECK (75). No residues were in disallowed regions, and 99% of the amino acids were in the most favored regions of the Ramanchandran plot. Residues 555-559, 588 -594, and 668 -674 of NCmac-N40(L6)C38 and residues 555-557, 592-594, and 672-674 of CPmac-N40(L6)C38 were left out of the model because of the absence of interpretable electron density for these atoms. The refined coordinates for NCmac-N40(L6)C38 (entry 1JPX) and CPmac-N40(L6)C38 (entry 1JQ0) have been deposited in the Research Collaboratory for Structural Bioinformatics Protein Data Bank.

RESULTS
CPmac N40(L6)C38 -Our focus is on how the five CPmac substitutions affect the structure and stability of the NCmac gp41 ectodomain core. A protein fragment corresponding to the intact ectodomain of gp41 tends to aggregate under physiological conditions (37,38) and is therefore intractable for biophysical analysis. Accordingly, we designed a recombinant model for this soluble gp41 core. The design was based on the N36(L6)C34 polypeptide in which the N36 and C34 helices are connected via a short peptide linker in place of the disulfidebonded loop region (60). The N36 peptide consists of residues 555-590, and the C34 peptide consists of residues 637-670 of the SIVmac251 gp41 protein sequences. For this study, we designed analogous polypeptides that were extended by four residues at the NH 2 terminus and four residues at the COOH terminus in order to accommodate all five CPmac mutations. The NCmac construct, designated NCmac-N40(L6)C38, was identical to parent SIVmac251, while the CPmac construct, designated CPmac-N40(L6)C38, contained the five CPmac mutations noted above. For convenience, we refer to the amino acids in N40(L6)C38 by their positions in the context of fulllength gp160.
The recombinant N40(L6)C38 proteins were produced by bacterial expression and purified by reverse-phase high performance liquid chromatography. Purified proteins were solubilized in 6 M GdmCl and 50 mM Tris-HCl (pH 8.0), and refolded by dilution into TBS. On the basis of CD measurements under physiological conditions, both the NCmac-and CPmac-N40(L6)C38 proteins are well folded and extremely stable, with greater than 95% ␣-helical content (see Table II). Sedimentation equilibrium experiments indicate that each protein exists in a discretely trimeric state across a concentration range from 10 to 100 M (Fig. 2). Taken together, these results indicate that CPmac-N40(L6)C38, like its NCmac counterpart and analogous polypeptides from other SIV and HIV strains (37)(38)(39), forms a stable ␣-helical trimer in solution. We conclude that, to a good approximation, the introduction of the five CPmac mutations into the N-and C-terminal heptad-repeat regions of the gp41 ectodomain does not affect the overall protein fold of the six-helix bundle.
Measurements of Thermal Stability-The thermal stability of the NCmac-and CPmac-N40(L6)C38 trimers was assessed using CD by monitoring the ellipticity at 222 nm as a function of temperature. The CD spectra of both of the proteins show a typical loss of ␣-helical content when the temperature is raised ( Fig. 2A). Before the unfolding transition, the molar ellipticity is the same for the two proteins and increases linearly with temperature. The pre-and post-transitional slopes and the shape of the main transition are very similar for the NCmac and CPmac six-helix complexes, indicating that the CPmac mutations do not detectably alter the mechanism of thermal unfolding. At a concentration of 10 M, the midpoint of thermal denaturation of CPmac-N40(L6)C38 is 92°C, as compared with a midpoint (T m ) exceeding 100°C for NCmac-N40(L6)C38 at the same concentration ( Fig. 2A; see Table II). In the presence of the denaturant GdmCl at 3 M concentration, both the NCmac-and CPmac-N40(L6)C38 proteins undergo cooperative and reversible unfolding transitions, with T m values of 72 and 63°C, respectively ( Fig. 2A; see Table II). The change in thermal stability between the two molecules is ϳ9°C in the presence of GdmCl. By using the Gibbs-Helmholtz equation, together with the thermodynamic parameters determined for the SIV N34(L6)C28 protein (76), the corresponding free energy difference is 14.9 Ϯ 1.5 kJ/mol for the CPmac-N40(L6)C38 trimer relative to the trimeric NCmac complex. These results support the hypothesis that the conformational instability of the CPmac gp41 trimer-of-hairpins may enhance the stability of its native envelope glycoprotein complex.
Crystal Structures of the NCmac-and CPmac-N40(L6)C38 Trimers-To better understand the chemical basis of destabilization of the six-helix bundle by the CPmac mutations, the high resolution x-ray crystal structures of the NCmac-and CPmac-N40(L6)C38 proteins were determined. Crystals of NCmac-N40(L6)C38 contain three monomers in the asymmetric unit, wherein each monomer is related by a crystallographic 3-fold axis. Crystals of CPmac-N40(L6)C38 contain a monomer in the asymmetric unit, with the trimer formed around the crystallographic 3-fold axis. Both of the structures were solved by a molecular replacement approach. In each case, the final 2F o Ϫ F c electron density map and simulated annealing omit map are readily interpretable except for a few disordered residues at the helix termini and in the six-residue loop region. Representative examples of omit electron density maps calculated with the final model phases are shown in Fig. 3. The NCmac-N40(L6)C38 structure was refined to an R value of 23.7%, with an R free value of 28.3% over a resolution range of 50.0 to 2.3 Å. Because of asymmetric crystal contacts, the three individual chains in the NCmac-N40(L6)C38 structure have slight differences in conformation and degree of order, with the B chain having a higher overall B factor (43.0 Å 2 ) than the A and C chains (35.3 and 40.8 Å 2 , respectively). The root mean square (r.m.s.) deviations between the individual chains in the NCmac-N40(L6)C38 noncrystallographic trimer vary from 0.62 Å (chains A and B) to 0.83 Å (chains B and C). The CPmac-N40(L6)C38 structure was refined to 1.7 Å to yield an R value of 20.9% with an R free value of 25.2%. Details of data collection and refinement statistics are listed in Table I. As anticipated, the overall architecture of the CPmac-N40(L6)C38 trimer is the same as that of the wild-type NCmac complex. CPmac-N40(L6)C38 forms a rod-shaped structure of ϳ55 Å in length and 40 Å in diameter (Fig. 3C). Each polypeptide chain has an ␣-helical hairpin conformation, in which two antiparallel helices are connected by a loop region. The N40 helices (residues 558 -591) form an interior, three-stranded ␣-helical coiled coil. This coiled-coil core includes ϳ34 residues (558 -591) (the three most N-terminal and C-terminal residues are disordered). The C38 helices (residues 637-671) pack in an antiparallel manner into hydrophobic grooves on the surface of the trimeric coiled coil (the three most C-terminal residues are not well defined in the electron density map). The N terminus of N40 and the C terminus of C38 are oriented at the same end of the rod. This packing arrangement would place the fusion peptide, located immediately before N40, and transmembrane segment, located immediately after C38, close together. The structural differences between the NCmac and CPmac six-helix bundles are small; the average r.m.s. deviation in C ␣ positions of the N40 coiled coil between the two molecules is 0.31 Å. The C38 helices can also be superimposed, with an r.m.s. deviation of 0.48 Å. Thus, the five CPmac mutations do not significantly alter the six-helix bundle structure.
Effects of CPmac Mutations on the Hairpin Structure-The N-terminal Leu-557 and Ser-557 residues are disordered in the NCmac-and CPmac-N40(L6)C38 structures, respectively. This could mean that end effects in the N40(L6)C38 construct cause these residues to unfold. We focused our structural analysis on the other four well defined CPmac substitution regions. On the surface of the N40 coiled-coil trimer, the grooves that are the sites for C38 interaction are lined with predominantly conserved hydrophobic residues at positions e and g (Fig. 1B). In the NCmac N40(L6)C38 structure, Val-571, at a g position,  packs against the C38 helix through hydrophobic contacts with the Leu-654 and Ile-655 residues (Fig. 4A). This hydrophobic effect buries 195 Å 2 of solvent-accessible surface area. In the CPmac core structure, Leu-571 interacts similarly with the C38 peptide, with a buried surface area of 213 Å 2 (Fig. 4B). Thus, the Val-571 to Leu substitution has little effect on the interface interaction and thus the gp41 core stability. In addition, Val-571 is conserved among different SIV strains, but it is changed to alanine in HIV-1. The hydrophobic residue at position 571 is expected to stabilize the six-helix bundle structure.
The solvent-exposed b, c, and f positions of the N40 helix are enriched with polar residues, most of which are limited to conservative polar mutations. These polar residues seem likely to be essential for maintaining the solubility of the gp41 core. Lys-573 is located at the b position, and its side chain is shown clearly in the NCmac-N40(L6)C38 electron density (Fig. 4C). On the basis of distance criteria, however, the Lys-573 side chain has no potential to form electrostatic and/or polar interactions on the surface of the six-helix bundle. In the CPmac-N40(L6)C38 structure, the substituted Thr-573 residue forms a hydrogen bond to the backbone carbonyl group of Leu-569 (2.76 Å) in the preceding turn of the same helix (Fig. 4D). Our data suggest that the stabilization energy gained from this hydrogen-bonding interaction may compensate the unfavorable effect of low helix propensity of threonine relative to lysine in the wild-type NCmac core. By extension, the Lys-573 to Thr mutation per se does not appear to alter greatly the gp41 core stability.
Whereas hydrophobic packing interactions involving residues at the a and d positions of the N40 peptide are important determinants of the high thermal stability of the gp41 core, occasional buried polar residues at this coiled-coil interface impose specificity for the trimer structure at the expense of stability (60,77). An interplay between these hydrophobic and polar interactions is thought to ensure that gp41 conformational changes proceed in a concerted fashion (60,76). In CPmac gp41, the a-position Thr-586 residue is mutated to isoleucine, a residue that is in the corresponding position of HIV-1. In both NCmac-and CPmac-N40(L6)C38 structures, three N40 helices are packed together in the classical acute "knobs-intoholes" arrangement characteristic of trimeric coiled coils, in which the C ␣ -C ␤ bonds in the side chains (knobs) of the a and d residues make an acute angle with respect to the recipient holes formed by spaces between four residues on the neighboring helices (78,79). The Thr-586 side chains face each other across the molecular 3-fold symmetry axis, but their C ␥ atoms are too far from each other to make van der Waals contacts (Fig. 4E). By contrast, the Ile-586 residues pack against each other to form favorable hydrophobic packing interactions (Fig.  4F). Thus, the Thr-586 to Ile substitution contributes to the thermodynamic stability of the N40(L6)C38 trimer.
The C-terminal end of the hydrophobic cavity on the surface of the N40 trimer features a hydrophobic cavity that has been identified as a potentially attractive drug target (56 -58). This cavity accommodates three hydrophobic residues from the abutting C38 helix: Trp-637, Trp-640, and Val-644. The Glu-641 residue projects outward and does not make contacts with the coiled coil. In both NCmac-and CPmac-N40(L6)C38 crystal structures, the Glu-641 and Lys-641 side chains have not been modeled for lack of connected electron density, presumably indicating disorder (Fig. 4, G and H). Interestingly, Glu-641 and its conservative aspartic acid substitution are completely conserved in 260 of 263 fully sequenced HIV-1, HIV-2, and SIV isolates (HIV-1 sequence data base, July 2001, Los Alamos National Laboratory). The remaining three isolates possess a lysine substitution at this position. The Glu/Asp-641 residue may be critical for membrane fusion due to the requirement of electrostatic complementarity in a different conformational state of gp41.
NCmac-N40(L6)C38 Variants with Single CPmac Mutations-Our crystallographic analysis suggests that the Lys-573 to Thr and Thr-586 to Ile mutations probably stabilize the NCmac gp41 core, while its stability is essentially unaffected by the Val-571 to Leu and Glu-641 to Lys mutations. Given that the CPmac hairpin structure is less stable than the corresponding NCmac structure, it would appear that the Leu-557 to Ser mutation markedly destabilizes the six-helix bundle. To directly test this hypothesis, we introduced each of the five CPmac mutations into NCmac-N40(L6)C38. The resulting sin-gle-point mutants were named by the position of the substitution. Sedimentation equilibrium measurements indicate that the L577S, V571L, K573T, and E641K variants sediment as clean trimers (Table II). T586I is also trimeric in solution and exhibits no systematic dependence of molecular weight on concentrations between 10 and 100 M (Table II). However, a systematic trend is observed in the residuals between the data and the linear fit (Fig. 5A), suggesting that T586I is prone to aggregation. On the basis of CD measurements at 10 M protein concentration in TBS, V571L, K573T, T586I, E641K, and NCmac-N40(L6)C38 (wild type) have a thermal stability that exceeds 100°C, whereas the L557S mutant exhibits a cooperative melt with a T m of 73°C (Table II). In the presence of 2 M GdmCl, L557S has a melting temperature 30°C lower and is ϳ49.5 kJ/mol less stable than the wild type (Table II). The V571L and E641K mutants have the same stability as the wild type ( Fig. 5B; Table II). The Lys-573 to Thr mutation has a small favorable residual stabilization of ϳ6.6 kJ/mol ( Fig. 5B; Table II). This may be due to a hydrogen-bonding interaction between the Leu-569 and Thr-573 residues seen in the CPmac-N40(L6)C38 crystal structure. The T586I mutant shows stabilization of ϳ38.0 kJ/mol, with an apparent elevation of the melting temperature relative to the wild type of about 23°C in the presence of 3 M GdmCl ( Fig. 5B; Table II). This greater stabilization effect can be ascribed to the van der Waals and hydrophobic interactions associated with the Ile-586 residues. By taking the sum of interactions in the five single mutants combined, the overall residual destabilization amounts to 4.9 kJ/mol (L557S (49.5 kJ/mol) Ϫ (K573T ϩ T586I) (44.6 kJ/mol)). This value is below that in CPmac-N40(L6)C38 (14.9 kJ/mol), suggesting that coupling interactions including structural relaxation in a coiled coil complicate parsing the free energy in this manner. DISCUSSION The production and characterization of the native trimeric gp120-gp41 complex have been hindered by the lability of intersubunit interactions. Initial attempts at making stable envelope glycoprotein trimers involved the disruption of the proteolytic cleavage site between gp120 and gp41 (80 -83). Such proteins are also truncated N-terminal to the membrane-spinning region of gp41, so they are efficiently secreted as soluble proteins. Although the uncleaved gp140 proteins form stable oligomers (Ref. 84 and references therein), their conformations are different from those of the mature envelope glycoproteins, as inferred from changes in antibody binding (18,85,86). Immunogenicity studies carried out to date with these gp140 proteins have not been particularly encouraging, as broadly neutralizing antibodies to HIV-1 have not been elicited (28, 29, 83,[87][88][89]. Recent attempts have been made to stabilize soluble gp140 proteins with the natural gp120-gp41 cleavage site preserved, but with a disulfide bond introduced between gp120 and gp41 (27). Notwithstanding the labile nature of the cleaved gp140 oligomers, they have antigenic properties that resemble those of the virion-associated envelope complex (27). Clearly, one of the major obstacles in the design of vaccines inducing humoral immunity is to develop envelope protein preparations that have a correctly folded native conformation and oligomerization state.
Interestingly, CPmac, unlike many other laboratory-adapted HIV-1 and SIV strains, displays an unusually stable gp120-gp41 interaction in the envelope glycoprotein spike (31). This tight intersubunit association is correlated with an increase in gp120 retention and high envelope spike count on virions (30,31). In addition, gp120 and gp41 isolated from CPmac virions are still associated in the presence of detergent. Importantly, the CPmac envelope glycoprotein is fully cleaved and fusioncompetent (31). It is surprising that this phenotype results from five point mutations within the ectodomain of gp41 (31). 2 Hence, the CPmac envelope glycoprotein provides an excellent system for conformational and functional studies of the native gp120-gp41 complex as well as for understanding the physical basis of the gp120-gp41 interaction. We have investigated the role of the CPmac mutations in conferring structural specificity and conformational stability to the gp41 core. Our results show that CPmac-N40(L6)C38, like the wild-type molecule, forms a trimer of helical hairpins in solution. The x-ray crystallographic analysis indicates that the CPmac mutation sequences are well accommodated in the six-helix bundle structure by forming helical packing interactions with different sets of atoms. The NCmac and CPmac six-helix bundles structures can be superimposed, with an r.m.s. difference of 0.51 Å. The major differences appear in the central N-terminal coiled-coil region, where four amino acid substitutions occur in CPmac gp41. Thermal unfolding studies show that the CPmac mutations destabilize the six-helix bundle by 15 kJ/mol, an apparent decrease of the melting temperature relative to wild-type of about 9°C. Furthermore, the buried Ser-557 residue, located at the interhelical interface g position of the N40 coiled coil, appears to be the key determinant for the reduced stability of the CPmac gp41 core. It is possible that the destabilizing effects observed here may be related to the biological properties of the CPmac envelope complex (see below).
Current models for virus-mediated membrane fusion suggest that viral fusion proteins undergo conformational changes to become fusion-active (reviewed in Ref. 6). The most notable example is the "spring-loaded" mechanism for activation of the hemagglutinin (HA) protein of influenza virus (90 -92). The native HA 1 -HA 2 complex on virions exists in a metastable conformation, distinct from the fusion-active state (90). The low pH environment of endosomes induces a major refolding of HA 2 , in which a nonhelical loop converts to a three-stranded coiled coil, projecting the fusion peptide region on the top of the molecule, where it can reach the target membrane (91,92). It has became apparent that the fusion potential of HA is activated by destabilization of the metastable native conformation (93,94).
For HIV-1 and SIV, membrane fusion is also thought to be promoted by a series of conformational changes in the envelope glycoprotein, although few details are understood (reviewed by Refs. 47, 51, and 95). Considerable evidence now suggests that gp41 exists in at least three different conformations: (i) a metastable native state, which is stabilized by extensive interactions with the gp120 surface subunit, (ii) a "prehairpin" intermediate, formed by exposure of the fusion peptide region and concurrent formation of the N-terminal coiled coil (49, 96 -98), and (iii) the fusogenic hairpin form, where the C-terminal helix associates with the N-terminal coiled coil to induce apposition of membranes for fusion (41,48,49). It is thought that free energy made available from hairpin formation can be harnessed to overcome the activation energy barrier for the fusion of two bilayers (e.g. Refs. 47 and 60), although it is unclear how the thermodynamics of the gp41 structural rearrangements are coupled directly to membrane fusion events. Several lines of evidence suggest that the conformational stability of the trimer-of-hairpins structure determines the membrane-fusion properties of the gp120-gp41 complex. First, the folding and stability of the gp41 core bearing mutations in one of the a positions of the N-terminal heptad-repeat correlate with severity of the in vivo phenotypes observed in cells expressing the mutant HIV-1 envelope proteins (98,99). These results are consistent with the notion that formation of the N-terminal three-stranded coiled coil occurs as an early event in the gp41 refolding process. Second, replacement of a conserved glutamine (Gln-652) at the buried face of the C-terminal helix by leucine increases HIV-1 infectivity (100). This fusion-enhancing mutation strengthens packing interactions between the Nand the C-terminal helices, stabilizing the six-helix bundle structure (101). Third, interhelical packing interactions appear to underlie the antiviral activity of peptides derived from the C-terminal helical region (called C-peptides) (43,101,102). Studies from several groups support a mechanism of dominant negative inhibition in which exogenous C-peptides bind to the N-terminal region of the prehairpin structure and prevent formation of the fusogenic gp41 trimer of hairpins (37,49,97,103,104).
The detrimental effects of the CPmac mutations on the gp41 core structure identified here, together with other available evidence, can be interpreted to support a minimal hypothesis: namely, mutations that destabilize the fusogenic hairpin structure might block gp41 activation and shift the conformational equilibrium of the envelope glycoprotein toward the native state, possibly leading to stabilization of the native envelope complex (60,77). Therefore, the CPmac mutations could tip the balance between the metastable native and stable fusogenic envelope conformations by destabilizing the six-helix bundle. Alternatively, these mutations could directly stabilize the native envelope complex. It is widely believed that in the native state, the N-terminal heptad-repeat region of gp41 exists in a non-coiled-coil conformation, such that gp120 shedding or displacement may allow formation of the N40 coiled coil (54,98,99,105,106). Thus, gp120 may serve as a clamp that binds to this N-terminal region, repressing formation of the six-helix bundle structure. The CPmac substitutions could enhance gp120-gp41 interaction, thereby stabilizing the native envelope structure. Nonetheless, our data show that an apparent increase in the stability of the CPmac envelope glycoprotein complex is accompanied by a decrease in the stability of the fusogenic gp41 hairpin structure. If this proves to be generally true, this finding may lead to a novel approach to stabilize the HIV-1 envelope glycoprotein complex for structural and immunogenicity studies.