Biophysical Characterization of the Structure of the Amino-terminal Region of gp41 of HIV-1

A peptide of 51 amino acids corresponding to the NH2-terminal region (5–55) of the glycoprotein gp41 of human immunodeficiency virus type 1 was synthesized to study its conformation and assembly. Nuclear magnetic resonance experiments indicated the sequence NH2-terminal to the leucine zipper-like domain of gp41 was induced into helix in the micellar solution, in agreement with circular dichroism data. Light scattering experiment showed that the peptide molecules self-assembled in water into trimeric structure on average. That the peptide molecules oligomerize in aqueous solution was supported by gel filtration and diffusion coefficient experiments. Molecular dynamics simulation based on the NMR data revealed a flexible region adjacent to the hydrophobic NH2 terminus of gp41. The biological significance of the present findings on the conformational flexibility and the propensity of oligomerization of the peptide may be envisioned by a proposed model for the interaction of gp41 with membranes during fusion process.

Fusion between enveloped viruses and host cells is an essential step for viral infection (1). The fusion domain within the transmembrane (TM) 1 proteins has been shown to be responsible for virus-cell fusion (2). For a majority of viruses, the fusion peptide, which penetrates into the target membrane, has been found at the NH 2 terminus of the TM subunit (3). A conformational change in the fusion domain of the virus, triggered by low pH (4) or binding to the receptors (5) leading to exposure of the fusion peptide, has been proposed to be an important intermediate step for virus-mediated fusions (for review, see Ref. 6). As suggested by many experiments, the fusion proteins oligomerize in the fusogenic state (7), and the fusion pore is believed to be composed of several oligomeric units of fusion proteins (8 -10). Thus, accumulation of the virus glycoproteins has also been found at the cell-cell contact region (11). Recently an early fusion active conformation of HIV-1 TM gp41 has been detected using an inhibitory synthetic peptide (12).
Recent structural studies by x-ray crystallography on the hemagglutinin of influenza virus (7) and the TM glycoprotein gp41 of human immunodeficiency virus revealed that the core of the fusion domain consists of an inner triple-stranded coiledcoil buttressed by three helices formed by the COOH-terminal region of the ectodomain of the envelope glycoproteins (13)(14)(15)). An anti-parallel core complex of the fusion domain of fusionmediating virus has also been deduced from a recent solution NMR study on a truncated protein derived from the ectodomain of simian immunodeficiency virus (16). The common features of these helix bundles are that the central coiled-coil trimer is formed by the leucine zipper-like heptad repeat sequence and that the height of the cylinder of the complexes is at least 60 Å. Thus, it is of interest to examine how the fusion complex containing several long rigid cylinders initiates approach of the merging membranes.
The function of the NH 2 -terminal fusion peptide of gp41 has been characterized, and the ensuing more polar region has been shown to associate with the surface subunit, gp120, of the external domain of HIV-1 env glycoprotein. These two regions of gp41 have also been implicated in the viral fusion by mutational studies (17). Despite elucidation of the core of the viral fusion domain complex by the x-ray diffraction and NMR studies, the structural role in the fusion event of the fusion peptide and the intervening sequence between it and the zipper-like region has not been addressed.
To better define the role of a given domain in the functions of a protein, a peptide corresponding to that segment can be studied (18 -20). In particular, the NH 2 -terminal hydrophobic peptides have been found to cause cell lysis, to induce electric current, and to promote lipid mixing in liposomes (21,22). A peptide corresponding to amino acids 5 through 16 of gp41 has been found to induce current under external applied voltage (21). The interaction with SDS micelle, on the atomic scale, of the 23-amino acid peptide corresponding to the NH 2 -terminal region of gp41 has been studied recently (23). It was found that Ala-Gly dipeptide, located 15 and 16 residues from the NH 2 terminus of gp41, is at the micellar-aqueous boundary and the region COOH-terminal to it is flexible.
In view of the highly conserved nature of the NH 2 -terminal region of gp41 implying its important structural role in the fusion (24), we report here the structure and conformations of a peptide encompassing amino acids 5 through 55 of gp41, under varying solution conditions. Using NMR, circular dichroism spectroscopies, light scattering, and gel filtration chromatography we attempt to further unravel the virus-mediated cell fusion in light of the reported oligomeric core structure of gp41 (13,14,16). The NMR methods are particularly suited for our purposes because they afford atomic resolution, the observation of conformational change in aqueous and membranemimic environments, and information regarding the flexibility of the specific regions in the peptides or proteins under study. In combination with the crystalline structural data on the core complex of gp41 external domain, the results show the physical properties for the NH 2 -terminal region of the TM glycoprotein gp41 and suggest the conformational transition during the fusion event and the hemifusion intermediate. Similar changes in conformation and assembly state may also occur in the fusion process of other enveloped viruses such as influenza virus.

EXPERIMENTAL PROCEDURES
Materials-The peptide (formyl-GALFLGFLGAAGSTMGARSMT-LTVQARQLLSGIVQQQNNLLRAIEAQQHLL-OH, gp41-N51) corresponding to amino acids 5-55 of gp41 and the peptide (NH 2 -YT-SLIHSLIEESQNQQEKNEQELLELDKWASLWNWF-OH, gp41-C36) corresponding to amino acids 127-162 from strain LAV 1a (LAV, lymph adenopathy-associated virus) were synthesized in an automated mode by a solid phase synthesizer from Applied Biosystems (Foster City, CA) model 431A using Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry. The peptides were cleaved from the resins by trifluoroacetic acids and purified by reverse phase high-performance liquid chromatography using a C8 column. The primary sequence of the peptide was ascertained by electrospray mass spectrometry as well as amino acid analysis.
SDS (sodium dodecyl sulfate) was acquired from Boehringer Mannheim (Mannheim, Germany) and d 25 -SDS from Cambridge Isotope (Andover, MA). 1,2-Ditetradecanoyl-sn-glycero-3-phosphocholine (PC) and 1,2-dihexadecanoyl-sn-glycero-3-phospho-L-serine (PS) were acquired from Calbiochem (San Diego, CA). Lysozyme (molecular mass: 14.4 kDa) was purchased from Sigma. All reagents were used in the experiments without further purification. Solutions containing vesicles are prepared by solubilizing the lipids in chloroform:methanol (4:1, v/v) mixture and drying the sample under nitrogen stream before dissolving in buffer solution. Mixtures of peptides and SDS or phospholipids were sonicated for at least 0.5 h before measurements.
NMR Experiments-Micellar solutions containing 1 mM gp41-N51 and 60 -100 mM d 25 -SDS were used for NMR measurements. Onedimensional and two-dimensional nuclear Overhauser effect spectroscopy and total correlation spectroscopy NMR experiments were performed in a Bruker AMX500 spectrometer as described previously (23). The NOE data were converted into interproton distance using 2H/3H cross-peak of the aromatic ring of phenylalanine as reference. A range of 0.6 -1.0 Å was allowed to vary in the distance constraints in the structural computation.
1 mM each of lysozyme and gp41-N51 in D 2 O solutions was used in pulsed field gradient experiments for the diffusion coefficient (D s ) determination. Temperature was kept at 298 K. The signal attenuation at time 2 relative to that at time 0 is given by the equation, where ␥ is the 1 H gyromagnetic ratio, ␦ is pulse field gradient duration, G is the gradient strength in gauss cm Ϫ1 , ⌬ is the diffusion delay time.
The diffusion coefficient of methanol was used to calibrate the gradient strength. In the experiments, ␦ and ⌬ were kept at 11 and 14 ms, respectively, while G was varied systematically. The maximum G was set at 45 gauss cm Ϫ1 . According to Stokes-Einstein equation, D s is inversely proportional to the hydrodynamic radius of a spherical particle, hence it is inversely proportional to the molecular mass to the one-third power.
Circular Dichroism Experiments-For CD measurements, cells with path lengths of 0.5-1.0 mm were employed in final 10 -50 M peptide concentrations. Vesicular samples were adjusted to pH 7.0 by phosphate buffer. All the CD experiments were carried on a Jasco 720 spectropolarimeter at ambient temperature. The spectra were recorded from 190 to 260 nm at scanning rate of 10 -50 nm⅐min Ϫ1 with a time constant of 2 or 4 s. Each of the CD data was obtained from an average of three to four scans with step resolution of 0.05 nm and bandwidth of 1 nm. CD data are represented by the mean residue ellipticity [] (degree⅐cm 2 ⅐mol Ϫ1 ), obtained from the observed ellipticity () according to the following equation [] ϭ ⅐l Ϫ1⅐ c Ϫ1 ⅐N Ϫ1 where l is the cell length in mm, c is the molar concentration, and N is the number of amino acid residues in the peptide. Quantitative prediction of the secondary struc-ture was accomplished by fitting CD data with Hennessey-Johnson algorithm (25), using 33 proteins of known secondary structure as the basis set.
Light Scattering Experiments-Measurements were performed on a DLS-7000 light scattering photometer (Photal Otsuka Electronics, Osaka, Japan) with control unit LS-71 using He-Ne laser light source of 10 milliwatts at 632.8 nm. The instrument was operated with a software controlled stepping motor driven goniometer. The measurements were made at 10°interval in the scattering angle over the range of 40°-140°at room temperature. Samples were dissolved in water and filtered twice before measurements. The concentration used in the study was 1.4 mg⅐ml Ϫ1 . Lysozyme was used as the molecular mass reference. The molecular weight of samples measured at low concentration can be obtained from the following equation, where K is the optics constant, 2 c is the concentration, Mw is the average molecular weight, R is the Rayleigh ratio, R G is the radius of gyration of the molecule, is scattering angle, and is the wavelength of incident light (for review, see Ref. 26).
Gel Filtration Chromatography-The experiments were performed in a Hitachi L-6000 liquid chromatography using a Superdex-75 HR 10/30 column (Pharmacia). The peptide and protein samples of 0.4 -0.8 mg⅐ml Ϫ1 concentrations were eluted with solution containing 22% acetonitrile and 0.1% trifluoroacetic acid with a flow rate of 0.5 ml⅐min Ϫ1 . Data were recorded at a wavelength of 214 nm. Transferrin, ovalbumin, ribonuclease A, and aprotinin were used as molecular mass markers (molecular mass: 81, 43, 13.7, and 6.5 kDa, respectively).
Structure Calculations-A total of 480, including 232 nonsequential, NOE interactions and 50 intraresidue dihedral angles, were utilized in the structural computations using DG/simulated annealing protocols of Biosym programs InsightII, Discover, and NMRchitect (version 2.3) from Biosym Technologies, Inc. (San Diego, CA). In simulated annealing protocol, the temperature was raised to 1000 K and a molecular dynamics run was carried out for 8 ps to allow more conformational space to be explored. The system was subsequently annealed to 300 K in 10 steps for a total 35 ps and minimized by the steepest-descent and conjugated gradients methods before final refined structures were obtained.  Table I). Resonance peaks of the ␣H of Gly 1 , Gly 6 , Gly 12 , and Gly 16 in SDS micellar solution are split, suggesting a certain degree of structural rigidity in the NH 2terminal region of the peptide. In contrast, in water ␣H of glycine residues, with the exception of Gly 6 , give rise to degenerate chemical shifts, implying more flexible conformation for the peptide in aqueous solution. This is likely due to the fact   that Gly 6 is in the FLG motif shown to form a type I ␤-turn in the fusion peptide of gp41 in aqueous solution (27).

NMR Data Indicate Enhanced Helical
To identify the secondary structure, the chemical shifts are compared with the random coil values as compiled by Wishart and Sykes (28), and the result is presented in Fig. 2. In water (Fig. 2B), the helical structure can be detected in the Gln 25 -Leu 30 stretch and is more prominent in the region COOHterminal to Gln 35 . This result is in agreement with the NOE results shown in Fig. 1D where d(i,i ϩ 3) interactions, characteristic of helical form, are observed in these regions. In the presence of SDS micelles ( Fig. 2A), on the other hand, the regions adopting helical structure encompass Leu 2 -Gly 12 , Val 24 -Val 34 , and Asn 38 -Leu 50 . The sequence following Ile 44 has the zipper 4-3 repeat pattern and thus is expected to have high propensity of forming helical structure. The NH 2 -terminal fusion peptide of gp41 has been shown to insert into membranous structure primarily as a helix (23). In addition to the helix formation of the NH 2 -terminal fusion peptide in SDS solution, enhancement of the helical form in the micellar medium is also observed for the region Gln 25 -Ser 31 and the region following Gln 35 . This is evident from the larger chemical shift deviation of ␣H and ␤H from the random values in SDS solution than those obtained in water as shown in Fig. 2. The present data for gp41-N51 thus corroborate with the idea that helix is induced in the micellar environment.
CD Data Also Reveal Enhancement in ␣-Helix for gp41-N51 in Vesicular and Micellar Solutions-The CD results displayed in Fig. 3 are consistent with the NMR data in that the helical content for the peptide is significantly increased upon association with SDS micelle. The helical structure is also enhanced in both PS and PC vesicular solutions. From the CD data, the secondary structural element analyzed by the algorithm of Hennessey and Johnson (25) for gp41-N51 is summarized in Table II Fig. 4. The intercept for gp41-N51 is 5.71 ϫ 10 Ϫ5 , which yields a particle mass of 17.5 kDa. The value is approximately three times the monomeric mass of 5.4 kDa for gp41-N51. The molecular weight determined by light scattering experiments is a weight-averaged value, hence the data indicate that the peptide molecules are predominantly trimeric or they are in an equilibrium between dimers and tetramers. The error estimated from the difference between the experimental and reported values of lysozyme is 9%.
Gel Filtration Experiment and Diffusion Coefficient Measurement Support Oligomerization of gp41-N51 Molecules in Aqueous Solution-The gel filtration data on gp41-N51 displayed in Fig. 5A may arise from dimerization of the peptide. Oligomerization of the NH 2 -terminal peptide of gp41 is also demonstrated by diffusion coefficient measurements from NMR pulsed field gradient experiments shown in Fig. 5B. The D s of the peptide is slightly larger than that of lysozyme with a molecular mass of 14.4 kDa. Thus, as in the light scattering experiments, diffusion coefficient measurements indicate that the peptide molecules exist in the aqueous solution as trimers, or there is equilibrium between dimers and tetramers for the peptide. It is of interest to observe that a particle mass of about 15 kDa is obtained for the peptide in the concentration on the order of 1 mg⅐ml Ϫ1 for light scattering and NMR experiments, while smaller size is detected at lower concentration employed in gel filtration experiment. The oligomerization of the NH 2 -terminal peptide of shorter length encompassing 33 amino acids has also been reported by Kliger et al. (29).

Three-dimensional Structures of gp41-N51 Derived from NOE Data Suggest a Flexible Segment Adjacent to the Hydrophobic NH 2 -terminal Region-
The three-dimensional structure of gp41-N51 in the presence of SDS micelles was calculated by employing the constraints derived from NOE data and hydrogen bonds suggested by the deuteron-hydrogen exchange experiments. Fig. 6A displays a typical computed structure and the superposition of 10 structures excluding residues between Ala 10 and Val 24 . As expected, the helical form can be observed in the NH 2 -terminal region up to Ala 11 , and the segments Leu 22 -Val 34 and Asn 38 -Leu 50 . The sequence between Ala 11 and Val 24 exhibits substantial flexibility as indicated by Fig.  6B, which shows larger root mean square deviation values of backbone atoms for these residues than residues in the remainder of the peptide as analyzed from the 10 conformers used in Fig. 6A. This flexible region overlaps with the gp120-associating site implicated in the mutagenesis studies (1).
The root mean square deviation values of atomic fluctuation for Asn 38 is substantially larger than those of the adjacent residues. In addition, a 3 10 helix is found for the Gln 35 ning Ala 11 -Leu 22 . Aside from both ends of gp41-N51 that are subject to the fraying effect, the larger fluctuation is seen for the Ala 11 -Leu 22 region. The result concurs with the idea that the Ala 11 -Gly 12 dipeptide is localized at the micellar-aqueous boundary, as noted previously (23), and determines the depth of penetration for the fusion peptide into the micelle. As will be discussed in the next section, the flexibility of this segment may have implication in the fusogenic activity of the NH 2terminal portion of gp41.

Conformational Change of the Hydrophobic Fusion Peptide in Membrane-mimic Environment and the Flexibility of the Adjacent Polar Region of gp41 Suggest an Essential Role for These Two Regions in the Virus-mediated Fusion Process-Our
CD data in water, the micellar and vesicular solutions for both the 23-amino acid fusion peptide sequence of gp41 in a previous report (23), and a longer gp41-N51 peptide indicate that, in the membranous environment, the helix form is induced at the expense of ␤-form or random coil for the NH 2 -terminal hydrophobic stretch of gp41. As shown previously (23), these data suggest that the fusion peptide penetrates into the membrane primarily as a helix. In the present study, conformational adaptability of gp41-N51 on the amino acid level exhibited in the presence and absence of membrane-mimic environment is provided by NMR data. However, conformation of the more polar sequence following the fusion peptide, Ala 12 -Leu 22 , was found more flexible from structural analysis of NMR data (Fig.  6). This conclusion is also consistent with scarcity of NOE cross-peaks involving protons in the residues of Ala 12 -Leu 22 segment, likely due to the smaller motional correlation times and/or more disordered structure, both of which are compatible with the idea of a flexible region. Because the segment was found to be essential for association with gp120 (2,30), the flexibility of this region in the membranous medium suggests that it plays a role in the conformational change during the native-to-fusogenic transformation of gp41, which is likely triggered by the dissociation of gp120 from gp41 upon binding to CD4 and the secondary receptor (for example, CCR5 or CXCR4).
As indicated in Figs. 2, A and B, the FLG motif (Phe 4 -Gly 6 ) undergoes a ␤ to ␣ transition as the peptide interacts with the SDS micelle. Moreover, the motif has been shown to serve as an  Top panel, the core oligomeric coiled-coils associating with the COOH-terminal helices of gp41 ectodomain. The NH 2 -terminal fusion peptide is primarily in ␤and random forms and the sequence between the fusion peptide and zipper-like domain, which may have a loose structure, is not completely dissociated from gp120 and/or the co-receptor on the target membrane. This is due to the highly hydrophobic nature of the fusion peptide, which does not favor its complete exposure to the polar solvent. Initial contact of gp41 with the target membrane is probably through the FLG motif. Middle, complete dissociation of gp41 from other components of fusion complex exposes the fusion peptide, which subsequently inserts into, and destabilizes, both the target and viral membranes primarily as a helix. Helical structure is also enhanced for the sequence NH 2 -terminal to the zipper-like region. According to the determined depth of penetration of fusion peptide, only the outer leaflet of the membrane bilayer of target cell has direct contact with the viral fusion peptide. Driven probably by the oligomeric coiled-coil zipper-like domain, the NH 2 -terminal region of the gp41 molecules self-associate. Propensity of the NH 2 -terminal region of gp41 to oligomerize, in turn, induces bulged surfaces on the two approximate membranes when the fusion peptides insert into these two membranes, thus facilitating the membrane merger. Alternatively, the interaction of the flexible region of gp41-N51 with the base region of the gp41 ectodomain can promote membrane fusion. Note that some hydrophobic NH 2 -terminal sequences may insert into the viral membrane. The straddling of the NH 2 termini from a core complex of gp41 over the target and viral membranes and the tendency to oligomerize by the NH 2 -terminal regions of gp41 provide the energy required to dehydrate the membranes to be merged. As a consequence, the two opposing membranes are brought into contact and dehydration occurs as the two membranes begin to attach. From the figure it can be visualized that the importance of the pliability of the segment Ala 11 -Leu 22 to allow the core coiled-coil rod-like structure with length initiation site for helix formation (27). Together with insertion into the micelle of the NH 2 -terminal Gly 1 -Ala 10 region, as suggested by the spin label attenuation data (23), these results suggest that the FLG motif may make initial contact with, and penetrate into, the membrane as part of the conformational change of gp41 following binding of the HIV envelope glycoprotein to its receptors in the fusion event (31,32).
The Ala 11 -Gly 12 region where the structures begin to diverge as shown in Fig. 6B coincides with the site found at the micellar-aqueous interface from spin label attenuation experiment for the fusion peptide in SDS solution. These results are reasonable in that the segment of the peptide in the micellar interior is more restricted than the region lying on the external surface of the micelle. Since the hydrocarbon chains of the cellular membrane are generally longer than that of SDS micelle, our data indicate that the inner leaflet of the target membrane bilayer is not likely to be reached by the viral fusion peptide during fusion. Therefore, it is probably that the direct effect of viral fusion peptide is to cause the hemifusion, as first suggested by White and co-workers on the fusion mediated by hemagglutinin of the influenza virus (33).
The increase in helicity for the segment Gln 25 -Ser 31 and the stretch Ser 19 -Val 24 (Fig. 2) in the presence of SDS micelles strongly suggests that these changes are part of the structural transition in gp41 as it transforms from fusion-inactive to fusion-active state. As already mentioned, the Gln 35 -Asn 38 stretch is likely to be helical in the fusogenic state. The enhancement of helical form for gp41-N51 in the membranous medium suggests that helix is the dominant form during membrane fusion.
Oligomerization of gp41-N51 in Water Detected by Light Scattering, Gel Filtration, and Diffusion Coefficient Measurements Indicates the Propensity of Self-assembly of the Peptide and Suggest a Functional Role of Merging Viral and Target Cell Membranes for the NH 2 -terminal Portion of gp41-The oligomerization state of the envelope glycoprotein of HIV-1 has been examined by various techniques, which yielded differing results of dimeric, trimeric, and tetrameric structures (13,14,34). It is, however, agreed that the membrane-anchoring subunit gp41 is primarily responsible for the oligomerization. In the present work, the peptide corresponding to the NH 2 -terminal region of gp41 was found to be oligomeric in water. Although the mass values deduced from gel filtration and diffusion coefficient data depend somewhat on the molecular shape, the experiments point to a non-monomeric state for gp41-N51. However, more definitive result on the oligomerization is provided by light scattering data which give the absolute molecular mass from the intercept of a typical Zimm plot (26,35). Dimers are apparently present in the more dilute solution for the gel filtration experiment, whereas trimers or mixtures of dimers and tetramers can account for data obtained in the more concentrated solution for NMR diffusion coefficient measurements. The latter conclusion is in agreement with the result obtained from the light scattering experiment in the same concentration range. Hence it is probable that different association states found in various studies arise from different conditions under which experiments are performed. Different oligomerization and assembly states might also be present at various stages of fusion event. Since gp41-N51 peptide molecules oligomerize in aqueous environment, it is likely that the NH 2 -terminal portion of gp41 penetrates into the target mem-brane in oligomeric form during some stage of fusion process.
Analogous examples for the zipper-like domain to induce oligomerization have been reported for the protein A/gp41(538 -593) chimeric protein (36) and for the maltosebinding protein/gp41(558 -595) chimeric protein (34). The zipper-like domain encompassing amino acids 555-583 of gp41 is responsible for oligomerization of the chimeric proteins, since both protein A and maltose-binding protein are monomeric (34,36). The stability of oligomerization is apparently critical in merging the phospholipids of the two apposing membranes. This is supported by the observation that a mutation in gp41 fusion peptide dominantly interfered with fusion and infectivity (37), which can be rationalized by participation of the defective mutant protein in the oligomer, rendering the fusion complex ineffective.
It has been shown by a number of investigators that the peptide corresponding to the hydrophobic NH 2 -terminal fusion peptide sequence is sufficient to cause leakage and phospholipid mixing of the vesicular bilayers (22,38). The vesicle and micelle are more prone to fuse than planar cellular membrane because of their smaller radius of curvature (39). Thus, to effect the fusion of planar bilayers of the viral and cellular membranes, it is possible that the core oligomeric complex formed by the zipper-like coiled-coil and COOH-terminal helices of gp41 ectodomain acts as the joint of a molecular nipper to hold the oligomeric structure, while the fusion peptide segments that penetrate into the membranes and the flexible polar border sequences would bring the two opposing membranes into contact by acting as the arms of a nipper. More explicitly, since the extended rod-like cylinder is at least 60 Å in length, the helical length would be extended to more than 90 Å if the intervening sequence between zipper-like domain and the hydrophobic fusion peptide also assumes rigid rod conformation, greater than 60 Å to perform its function in bringing the two membranes into contact. Bottom, the two membranes merge, and several units of fusion core oligomer of gp41 assemble to form the fusion pore to allow the flow of nucleocapsid of the virus. Initial lipid mixing probably involves only the outer leaflets of the two membranes, i.e. stalk formation, as suggested by the finding that the fusion peptides of gp41 do not reach the inner leaflet of either the target or viral membrane bilayer. making approach of the two apposing membranes impossible. The flexibility of the Ala 11 -Leu 22 stretch, illustrated in Fig. 6, A and B, and the propensity of oligomerization of the NH 2portion of gp41 render possible direct contact of phospholipids of the opposing membranes (Fig. 7). This is done by bending the Ala 11 -Leu 22 segment and the base of the gp41 ectodomain to make the core oligomeric complex parallel to the membrane surfaces. Flexibility of the sequence near the base of gp41 ectodomain was suggested by the crystallographic study of the NH 2 -and COOH-terminal helices of gp41.
Energetically, a driving force to merge the two approaching membranes is provided by the tendency for the NH 2 -terminal portion of gp41, represented by gp41-N51, to self-assemble and/or by the interaction between the NH 2 -terminal region and the COOH-terminal region of the gp41 external domain. The latter hypothesis is based on the fluorescence data on gp41-C36 incubated with gp41-N51 (see Fig. 8). The model illustrated in Fig. 7 emphasizes the essential role in bringing the two fusing membranes into approximation played by the segment NH 2terminal to the zipper-like domain of gp41. Because the core antiparallel helix bundle of gp41 functions as the joint or hinge of a nipper, it cannot exert force on the target or viral membranes. The latter function is then furnished by the arm of the nipper, the polar intervening sequence (Ala 11 -Leu 22 ) of gp41, and/or by its interacting with the segment near the base of gp41 ectodomain. It must be noted that the rigid structure such as the core hetero-helix bundle cannot directly exert force on the membranes. Rather, closing of flexible arms (Ala 11 -Leu 22 of gp41-N51), due to the tendency to oligomerize for the NH 2terminal portion of gp41 extending from the core structure and/or the base region of the gp41 ectodomain can promote membrane coalescence, as illustrated in Fig. 7. It is stressed that the fusion peptide, because of its hydrophobicity, is not likely to be completely exposed to solvent during the entire fusion process. The fusion peptide may be ushered by other component(s) of the fusion reaction, such as gp120 or the receptor or the co-receptor, to the target membrane.
Membrane fusion is an energy costing event, since it involves destabilization of the bilayers and dehydration of the interface regions of merging membranes (40). This notion is also consistent with the low efficiency of productive entry by the virus into the target cell (41). The insertion of the hydrophobic fusion peptide makes initial perturbation of the target membrane. Propensity of the region encompassing gp41-N51 for oligomerization and the interaction between the flexible, more polar NH 2 -terminal region outside the membranes and the segment preceding the membrane-anchoring sequence of gp41 probably helps to overcome the hurdle of membrane dehydration during the membrane merging.
In summary, study of the NH 2 -terminal region of gp41 complements the structural investigation of core complex formed by the oligomers of the zipper-like domain and the COOHterminal helix. The findings of flexibility of the Ala 11 -Val 24 segment and the tendency of forming oligomers by gp41-N51 and its interaction with the region near the base of the ectodomain of gp41 along with increased helicity in the membranous medium for the peptide suggests an structural role of the fusogenic activity of the NH 2 -terminal portion of gp41. Our data also imply a stalk and a hemifusion intermediate for the fusion reaction mediated by HIV. Similar mechanism may be used by other enveloped viruses in mediating cell fusion.