The Six-Helix Bundle of Human Immunodeficiency Virus Env Controls Pore Formation and Enlargement and Is Initiated at Residues Proximal to the Hairpin Turn

Abstract

Residues that create the grooves of the human immunodeficiency virus type 1 (HIV-1) Env triple-stranded coiled coil (HR1) and the residues that pack into the grooves (HR2) to complete the formation of the six-helix bundle (6HB) were mutated. The extent and kinetics of fusion as well as pore enlargement were measured for each mutant. Mutations near the hairpin turns of each monomer of the 6HB were more important than those far from the turn, for both HR1 and HR2. This result is consistent with the idea that binding of HR2 to the HR1 grooves is initiated near the hairpin turn of each monomer. Mutations at the distal portions also reduced fusion, albeit to a smaller extent. An intermediate of fusion (temperature-arrested state [TAS]) was formed, and the consequences of mutation were compared; a mutant that exhibited less fusion also showed slower kinetics from TAS. This suggests that formation of the bundle is a rate-limiting step downstream of the intermediate state. The rate of enlargement of a fusion pore also correlated with the extent and kinetics of fusion. The rate of pore enlargement was severely reduced by mutation. This supports our prior conclusion that formation of the 6HB occurs after pore creation and strongly suggests that the free energy released by bundle formation is directly used to promote pore growth.

All class I viral fusion proteins are composed of three identical monomers that together fold into a six-helix bundle (6HB) in the protein's final, postfusion state (33, 50). Each monomer of human immunodeficiency virus (HIV) Env (which is a class I fusion protein) consists of two subunits, gp120 and gp41. The three gp41 subunits of each Env trimer form the 6HB. The crystallographic structure of gp41 has been determined for its final state; its 6HB reveals that the three N-terminal segments of gp41—each segment referred to as heptad repeat 1 (HR1)—have folded into a central triple-stranded coiled coil of α-helices, and the three C-terminal segments (HR2) have packed, antiparallel, as α-helices into the three grooves of the coiled coil. The 6HB of HIV Env is thermally stable (43) and can confidently be considered a final structure.

The crystallographic structures for both the initial and final states have been obtained for several other class I viral fusion proteins (2, 7, 22, 26, 43, 54). They reveal that the protein must undergo large-scale conformational changes in transiting from their initial state to final 6HB state, including changes in secondary structure. Although the initial structure of gp41 has not been determined, it is known that grooves are not exposed in its initial conformation; they become exposed as part of the gp41 conformational changes during fusion (13). The configurations of Env during which grooves are exposed are referred to as prebundles. It is assumed that gp41 undergoes large-scale reconfigurations during the fusion process.

It is certain that formation of the 6HB is a central event in infection; 6HBs form in all class I proteins; peptides that prevent formation of the bundle block infection (6, 17, 20, 21, 36, 38, 51); and many residues of HIV's 6HB are conserved, even though Env is a rapidly mutating protein. In fact, the residues that comprise the grooves are almost perfectly conserved, and the groove-packing residues of the C-terminal HR2 segments are highly conserved (8). The importance of the bundle is underscored by the finding that point mutations that weaken the associations between HR2 and their grooves are deleterious to fusion (23).

Because each monomer of gp41 has a hairpin turn in the final structure, bundle formation requires that each monomer fold in the “middle,” like a jackknife. For HIV Env, HR1 is near the fusion peptide, and HR2 is close to the membrane-spanning domain (MSD); thus, bundle formation brings the fusion peptides and MSDs into proximity. At the time 6HBs were being identified as a common structure, it was assumed that their formation merely brought membranes into contact (8, 49), but it has since been shown that for HIV Env and for simian virus 5 not all of the fusion proteins that participate in fusion fold into 6HBs prior to membrane merger (35, 41) and, in fact, bundle formation is not completed until the viral envelope and cell membrane have become continuous upon the creation of a fusion pore (28).

The amino acids at the “e” and “g” positions of HR1 comprise the grooves of the central triple-stranded coiled coil (Fig. 1). The “a” and “d” positions of the HR2 domains fit into these hydrophobic grooves to create the 6HB (Fig. 1). Alanine-scanning mutagenesis has been carried out for the “e” and “g” positions of HR1 to identify sites deleterious to fusion, and it has been shown that reductions or increases in bundle stability caused by mutation at these sites are not good predictors for reductions or increases in extents of fusion (23).

FIG. 1.

Depiction of a 6HB. Three HR1 segments, one from each gp41 monomer, combine to form a central triple-stranded coiled coil. Three HR2 segments of gp41 pack into the grooves to complete the bundle (six-circle figure). Residues at positions “a” and “d” of HR2 pack into the grooves. All were mutated, one at a time, to alanine. The “e” and “g” of HR1 create the groove. Residues at these positions were mutated to alanine, except for residue A558 (a “g” position) which is naturally alanine.

In the present study, we mutated all “a” and “d” residues of HR2 and all “e” and “g” residues of HR1. We then measured the extent and rate of fusion, as well as the rate of pore growth, and compared results for the various mutants. From these comparisons, we were able to infer the order of interactions between residues of HR1 and HR2 in bundle formation.

MATERIALS AND METHODS

Cells and labeling.

HelaT4+ cells expressing CD4 and CXCR4 as chemokine receptor (25) were obtained from the NIH Research and Reference Reagent Program and were used as target (Tg) cells. They were grown in Dulbecco's modified Eagle's medium (DMEM; Gibco BRL, Gaithersburg, MD) supplemented with 10% cosmic calf serum (HyClone Laboratories, Logan, UT) and 0.5 mg/ml of Geneticin (Gibco BRL). HEK 293T human kidney cells (referred to as 293T cells) were the effector (Ef) cells. They were grown in DMEM supplemented with 10% cosmic calf serum, 1% penicillin-streptomycin, 1% l-glutamine, and 100 μg/ml of Geneticin. Wild-type (WT) (HXB2) and mutant Envs were transiently expressed in the 293T cells, using a standard calcium phosphate transfection method with chloroquine treatment (24), and harvested 2 days after transfection. These cells were loaded with calcein AM (Invitrogen, Eugene, OR), and the Tg HeLaT4+ cells were loaded with 7-amino-4-chloromethylcoumarin (CMAC; Invitrogen) (35). This labeling allowed the Tg and Ef cells to be easily identified. For experiments in which the large fluorescent dye 5-chloromethylfluorescein diacetate (CMFDA; Invitrogen) was loaded into Tg cells, CMAC was omitted. Bovine serum albumin and chlorpromazine (CPZ) were obtained from Sigma Chemical Co. (St. Louis, MO).

Mutagenesis.

Mutant HIV envelopes were generated by double PCR methods. Forward and reverse primers, 2 μg each, were added to 96 μl of the PCR mixture containing 250 ng of plasmid DNA, 3 units of Pfu DNA polymerase, and 1 unit of Taq2000 DNA polymerase (Stratagene, La Jolla, CA). The forward primer carried the NheI site with sequence derived from positions 1610 to 1636 of E7-HXBc2(IIIexE7pA-Kpn2′); the reverse primer contained the mutated codon and had about 40 nucleotides flanking both sides of the codon. The first PCR product was used as the forward primer for the second PCR. The reverse primer of the second PCR carried the BamHI site with a sequence derived from positions 2810 to 2838 of E7-HXBc2(IIIexE7pA-Kpn2′). We used 33 cycles to obtain profiles for both the first and second PCR. For the first PCR, denaturation was performed at 95°C for 30 s, annealing at 55°C for 30 s, and extension at 72°C for 3 min. For the second PCR, the conditions were the same except that the extension time was 4 min. The PCR product with a mutant codon was ligated into the NheI and BamHI sites of E7-HXBc2(IIIexE7pA-Kpn2′).

Protein expression by flow cytometry.

After harvesting the transfected HEK 293T cells, they were incubated with 1:200 1B12 (NIH AIDS Research and Reference Program) for 4 h and then incubated with 1:500 fluorescein isothiocyanate-conjugated anti-human immunoglobulin G antibody (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) for 1 h. The distribution of HIV Env for these labeled cells was analyzed with a flow cytometer (Guava EastCyte; Guava Technologies, Hayward, CA).

Env cleavage from Western blots.

The membrane proteins of transfected HEK 293T cells were collected with a ReadyPrep protein extraction kit (Bio-Rad Laboratories, Hercules, CA). The extracted proteins were resolved by 16% sodium dodecyl sulfate-polyacrylamide gel electrophoresis, and the proteins were then transferred to a nitrocellulose membrane (46). Blotted proteins were identified by incubating with 2F5 monoclonal antibody (NIH AIDS Research and Reference Program) and then with 1:500 horseradish peroxidase-conjugated anti-human immunoglobulin antibody as the secondary antibody (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). The bands were chromogenically developed by adding a metal-enhanced DAB (3,3′-diaminodbenzidine) substrate (Thermo Scientific, Rockford, IL). The intensities of the gp41 bands were determined by using the Quantity One software (Bio-Rad Laboratories, Hercules, CA). All quantification was performed on original scans. However, to aid visualization of the bands shown in Fig. 4, overall contrast was digitally enhanced (Adobe Photoshop 6.0; Adobe Systems Inc., San Jose, CA).

FIG. 4.

HIV Env is cleaved into subunits. Membrane proteins from cells expressing HIV Env (or mutants) were resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and then labeled with an anti-HIV gp41 antibody (2F5). The bands shown were the most prominent on the gel; minor bands of dimerized gp41 and larger aggregates were also present. Bands are not present for the L663A and W666A mutants because the recognition site of 2F5 includes these positions (4). The contrast in the image has been uniformly heightened to aid visualization here. Quantification of all bands, however, was performed using the original images.

Monitoring fusion and pore growth.

For fusion experiments, previously published methods were followed (35). In brief, ∼10⁵ of Ef and Tg cells were mixed together in microcentrifuge tubes containing 0.2 ml of HEPES-buffered DMEM (pH 7.2) supplemented with 1 mg/ml bovine serum albumin and immediately transferred into the wells of an eight-chamber glass slide (Lab-Tek; Nunc) that had been pretreated with poly-l-lysine. Pairs of Ef and Tg cells that exhibited spread of both calcein and CMAC after 2 h at 37°C were visually identified and scored as fusion. The percentage of fusion was set equal to the number of fusion events normalized by the number of pairs of Ef and Tg cells that were in contact. A temperature-arrested state (TAS) of fusion was generated, as necessary, by incubating the Ef and Tg cells that were adhered to the coverslips at 23°C for 2.5 h. Hemifusion was assayed at TAS by the addition of 0.5 mM CPZ for 1 min at room temperature, followed by CPZ removal. Cell pairs whose aqueous dyes mixed upon the addition of CPZ were scored as hemifusion.

Experiments that monitor pore growth used techniques we had previously developed (28). Heat-absorbing glass formed the bottom of a specially designed chamber whose temperature was Peltier controlled (20/20 Technology, Wilmington, NC). Cells were brought to TAS on coverslips, and these coverslips were then placed in the chamber. The temperature-controlled chamber, containing a solution maintained at 4°C, was mounted in a fluorescence microscope (Axiovert 100A; Carl Zeiss, Thornwood, NY). For the pore growth experiments, Ef cells were loaded with calcein, but Tg cells were not labeled with any fluorescent dye. Bound Ef/Tg cells were selected and irradiated by an infrared diode laser. The heating of the absorbing glass increased the temperature of the solution, which was monitored by an immersed thermistor. The laser was immediately shut off by computer control once the temperature reached 37°C. The rise in temperature took less than 2 s; the fall back to 4°C took ∼10 s. As described previously in detail, the size of the pore is determined by monitoring the rate of calcein spread from an Ef cell to its bound Tg cell (28).

RESULTS

We mutated all residues at the “a” and “d” positions of the HR2 domain to alanine, one residue at a time, and measured fusion between Ef cells expressing the fusion protein and the Tg cells by monitoring transfer of calcein, a small aqueous dye (Fig. 2A). A clear overall pattern emerged. Mutations of residues closer to the N terminus were more deleterious to fusion than were mutations of residues closer to the C terminus. Although the pattern did not perfectly correlate with the position along the bundle, mutations at positions 649 and higher were of less consequence than were mutations at positions 645 and lower (Fig. 2A). All mutations resulted in reduced fusion, but for a few of the N-terminal mutations, the reductions were relatively small.

FIG. 2.

Extent of fusion before and after addition of CPZ. (A) Fusion extent for mutants of HR2 displayed according to increasing residue number. (B) Extent of fusion for mutants of HR1 shown in order of residue number. Black bars are percentages of bound Ef/Tg cells that exhibit transfer of calcein and CMAC. White bars show the percentages of bound cells exhibiting dye transfer after the addition of CPZ. Dye spread was monitored within 1 min after adding CPZ. Only mutations with open bars were tested with CPZ. Error bars are SEM.

Although fusion had not been achieved, it was possible that the upstream stage of hemifusion had been reached. The weak base CPZ is membrane permeable and disrupts hemifusion diaphragms; its addition to solution can therefore be used to test whether hemifusion has occurred (10, 34). For the mutants that did exhibit high extents of fusion, CPZ could not cause many more cell pairs to fuse because so many were already fused, but even for mutants that yielded low levels of fusion, the addition of CPZ did not cause an appreciable number of the unfused cell pairs to display calcein spread (Fig. 2A). The absence of effect upon CPZ addition indicates that the unfused cell pairs had not proceeded as far as restricted hemifusion, a state of hemifusion that does not permit transfer of lipids between the merged membranes.

The use of CPZ to identify the occurrence of hemifusion was originally established in a system of erythrocytes bound to nucleated cells expressing a viral fusion protein (34). It has become routine to use CPZ in the same manner to determine whether fusion has occurred between two nucleated cells (29, 42). As a confirmation that the use of CPZ can be safely extended to identify hemifusion between nucleated cells, we have performed a number of control experiments. We incubated Ef and Tg cells for 2 h, which led to a high extent of fusion; dye spread was not augmented by the addition of CPZ (Fig. 3, column 1). The addition of C34 and CPZ after the incubation was also without effect (column 2). The inclusion of 100 nM C34 in the solution for the 2-h incubation abolished fusion, and here, the addition of CPZ did not induce any dye spread (column 3). Clearly, CPZ-induced dye spread required that Env undergo conformational changes that included the packing of some HR2 segments into the grooves of the coiled coil. Taken together, these control experiments using WT Env show that the dye spread induced by the addition of CPZ requires prior Env-induced changes in the membranes; the addition of CPZ to intact cell membranes (i.e., that have not merged to bound membranes) does not induce dye spread. This conclusion is further buttressed by results with the G572A mutant; the increase in aqueous dye spread upon addition of CPZ exceeds the amount due to fusion naturally induced by the G572A mutant (column 4). Consequently, the CPZ-induced increase is easy to discern for this mutant. The simultaneous addition of C34 and CPZ after an ∼2-h incubation (column 5) induced the same increase in dye spread as did CPZ alone (column 4). This indicates that CPZ induces an increase in dye spread without affecting HIV Env. The conformational changes in the G572A mutant must occur prior to the addition of CPZ if CPZ is to have an effect (column 6). These control experiments indicate that the widespread use of CPZ in the identification of restricted hemifusion between nucleated cells is justified.

FIG. 3.

Promotion of aqueous dye spread by the addition of CPZ requires prior advancement of the fusion process. Addition of CPZ after a 2-h incubation of Ef and Tg cells did not induce further aqueous dye spread (column 1). Including 100 nM C34 during the 2-h incubation abolished the ability of CPZ to induce dye spread (column 2). The addition of CPZ after cell-cell incubation for 1 h and 50 min followed by addition of C34 did not lead to further dye spread (column 3). For Ef cells expressing the G572A mutant, the addition of CPZ after a 2-h cell-cell incubation led to a significant increase in aqueous dye spread (column 4). For the G572A mutant, the inclusion of C34 during cell-cell incubation abolished both fusion and the ability of CPZ to further promote fusion (column 5). Adding C34 after the cell-cell incubation did not affect the subsequent ability of CPZ to promote fusion (column 6).

The residues at the “e” and “g” positions of the N-terminal HR1 domains make the contacts with the residues of HR2 that fit into the grooves; the “e” position of HR1 contacts the “a” position of HR2, and the “g” position of HR1 directly interacts with the “d” position of HR2. We mutated all “e” and “g” positions of HR1 to alanine and measured cell-cell fusion. As was the case for the HR2 mutants, there was a clear positional break (with the exception of Q577A) in the effect of mutation on efficacy of fusion (Fig. 2B). At positions 565 and higher, the mutations inhibited fusion to a much greater extent than did the mutations at residues 563 and lower. This was true for both mutations at “e” (V549, Q563, and V570) and at “g” (Q551, L565, G572, and R579) positions. The break for the extent of fusion was sharp; fusion was low for mutations at position 565 and high at position 563. As controls, mutations that severely changed the physical properties of two residues that lie outside the grooves were tested and found to have no effect on the level of fusion (data not shown).

The high extent of fusion obtained for the Q577A mutation is a clear and definite exception to the overall pattern and is in agreement with a previous finding that this mutant was very effective in causing syncytium formation (23). The three-dimensional crystal structure of the 6HB (8, 45) shows that Q577 interacts predominantly with W628. Visual inspection of computer-generated graphical models indicates that replacing Gln with Ala at position 577 does not hinder the interactions with the Trp at position 628 and may even facilitate it.

Reductions in the extent of fusion were neither due to lower expression levels of the mutants or a consequence of reduced cleavage of Env into gp120-gp41 subunits. Fluorescence-activated cell sorter analysis showed that expression levels were comparable for WT Env and the mutants and, in fact, were generally somewhat higher for the mutants (Table 1). Western blot analysis against gp41 was performed for all mutants (Fig. 4), and densitometry showed that these mutants were as extensively cleaved as was WT (Table 1). Some of the mutants showed a higher percentage of cleavage than WT, while others showed less, but the relatively small differences in the amount of cleavage did not correlate with the amount of fusion. For example, some mutants that yielded a lower extent of fusion than WT were more extensively cleaved (e.g., Y628A, W631A, I642A, and L645A). 2F5, the antibody we used for Western blot analysis, includes positions L663 and W666 in its recognition site (4). Consequently, this antibody could not identify cleavage for mutants at these two positions. The W666A mutant yielded high extents of fusion, so its high level of cleavage is not in doubt. In the case of the L663A mutant, we cannot rule out the possibility that improper cleavage was the cause of its low fusion activity. But nothing apparent is unique at this site, and every other mutant was properly cleaved; we assume this was also the case for the L663A mutant. In general, mutations did not affect cleavage of gp160 into gp120/gp41 subunits, and so we conclude that changes in bundle formation are the direct cause of changes in fusion activity.

TABLE 1.

Mean expression levels and cleavage of HIV WT and mutant Envs^a

Env protein	% Mean Env expression	Western blot density (INT/mm2)	Relative Western blot density
WT	100	913	913
V549A mutant	99	584	590
Q551A mutant	114	658	577
L556A mutant	108	1,268	1,174
Q563A mutant	111	1,021	920
L565A mutant	136	574	422
V570A mutant	144	321	222
G572A mutant	132	499	378
Q577A mutant	147	928	631
R579A mutant	143	839	587
W628A mutant	125	1,327	1,062
W631A mutant	144	1,501	1,042
I635A mutant	149	1,082	726
Y638A mutant	149	1,115	748
I642A mutant	123	1,630	1,325
L645A mutant	112	2,026	1,808
S649A mutant	126	950	754
Q652A mutant	124	879	709
N656A mutant	120	952	793
E659A mutant	132	820	626
L663A mutant	130	78	N/A
W666A mutant	121	83	N/A
Mock		78	N/A

^aFor WT and all mutants, n = 3. Expression levels for WT were normalized to 100% for each experiment, so here, standard deviation is not applicable. The absolute densities of gp41 of the Western blot (Western blot density column) were normalized by the expression levels (percent mean Env expression column) to yield the relative densities (density/Env expression relative to WT), as listed in the relative Western blot density column. It provides a quantitative comparison of percent cleavage of each mutant. The L663A and W666A mutants could not be tested for extent of cleavage because the binding region of the 2F5 antibody includes these positions (4). N/A, not applicable.

We have previously shown that the 6HBs form late in the fusion reaction and that their formation is completed only after the creation of the initial pore (28, 35). It is not until all the bundles form that the fusion pore is stabilized to the point that it will no longer close (28). To assess the effect of mutation on late stages of fusion, we created a standard intermediate of HIV Env-mediated fusion for critical residues of the bundle. Cells were incubated at 23°C for 2.5 h, a treatment that yields a TAS of fusion in which the upstream, early steps of CD4 binding and engagement of chemokine receptors by HIV Env is completed (37). Measuring kinetics of fusion from TAS allows the effect of mutation on later, downstream steps of fusion to be assessed. So as to maximize the chance of recording fusion from TAS, we limited these measurements of fusion kinetics to mutations that yielded high extents of fusion (i.e., ≥50% of WT).

Fusion was triggered by quickly raising temperature of cells from 23°C to 37°C and monitoring the spread of calcein. The extent of fusion as a function of time was normalized to the full extent of fusion for that mutant so that kinetics could be compared between mutants. Fusion was reasonably fast from TAS, and for WT, it commenced within seconds of raising temperature (Fig. 5A and B, insets). Env that had been mutated at positions that led to the same extent of fusion as that for WT (i.e., 563 and 577) also yielded the same rates of fusion. These mutants and WT yielded the fastest fusion (Fig. 5B).

FIG. 5.

Kinetics of fusion from TAS. Extents of fusion for all mutants were normalized to 1. (A) HR2 mutants. (B) HR1 mutants. Insets within panels A and B show expanded views of fusion kinetics over the first minute.

The extent of fusion from TAS was the same as when fusion was allowed to proceed uninterrupted for 2 h at 37°C (data not shown). The time until half of the fusion events occurred was compared against the extent of fusion. Higher extents of fusion correlated well with faster kinetics for mutations within both HR2 (Fig. 6A) and HR1 (Fig. 6B). In other words, mutants that yielded the highest extents of fusion also exhibited the fastest kinetics. This finding indicates that one or more steps that lead to bundle formation are rate limiting for the creation of a pore once TAS has been achieved.

FIG. 6.

Correlation of extents and kinetics of fusion. Kinetics of fusion were measured from TAS and parameterized as the time for half the fusion events to occur. (A) HR2 mutants. (B) HR1 mutants.

While an open fusion pore remains relatively small, it can be prevented from further expanding by lowering the temperature. The pore remains open despite the low temperature, but addition of peptides that prevent bundle formation (e.g., C34 or N36) causes the pore to close (28). This shows that pore growth is intimately connected to bundle formation, and formation of 6HBs is not complete even after a fusion pore has opened. It follows that pore enlargement may also be sensitive to mutations that affect bundle formation. We tested this conjecture by measuring the rate of movement of calcein from Ef to Tg cells for WT and several of the HR2 mutations. This rate yields the time course of pore growth (28). Even for the same construct, growth was quite variable between pores (Fig. 7). We therefore measured growth for at least seven pores for each mutant. A significant fraction of the WT pores enlarged much more rapidly than did any of the mutant pores. WT pores did not, however, always grow rapidly, and some of its pores grew more slowly than the faster-expanding mutant pores. On average though, the early (within the first 4 s) growth of WT pores was significantly greater than growth of pores formed by the mutant Envs (Fig. 7, lower left panel). It is clear that reducing the interactions between HR1 and HR2 is deleterious to pore growth. The early growth of pores was comparable for mutations at the “a” (649 and 656) and “d” (652 and 666) positions. Although the extent of fusion for W666A was a full 60% that of WT, its rate of fusion pore growth was much lower. The fact that this mutation hindered pore enlargement to a much greater degree than it reduced extents or kinetics of fusion clearly demonstrates that 6HB formation is not complete prior to pore creation and that completion of the bundle is directly coupled to pore enlargement, as we argued previously (28). Moreover, fusion pore enlargement is not spontaneous; conformational changes of the fusion protein drive the enlargement.

FIG. 7.

Growth of fusion pores for mutations within HR2. Top panels show calcein fluorescence of the Tg cell as a function of time for all cell pairs tested for each mutant. Each curve tracks the results from a single cell pair. The rate of change in fluorescence is a measure of the rate of change in pore size. Lower left panel, average fluorescence of Tg cells versus time; lower right panel, the extent of cells for which the small dye calcein (black bars) and the large dye CMFDA (open bars) transferred from Ef to Tg cells. Error bars are the standard errors of the means. Measurements for each mutant were performed at least eight times.

Measuring the concentrations of small dyes as a function of time in fused Ef and Tg cells is ideal for determining the rate of pore growth in the early stages after pore formation. But over time, the small dye equilibrates between cells, and once it does, pore growth can no longer be monitored by dye spread. To assess the final extent of pore growth, we monitored transfer of a small (calcein) and large (CellTracker CMFDA) dye between cells (Fig. 7, lower right panel). The transfer of calcein reports the percentage of cells that fuse. CMFDA is a thiol-reactive dye that combines with glutathione and possibly with some cytosolic proteins. This creates a large, fluorescently tagged complex that can be used to score substantial pore enlargement. We found that after a 30-min incubation of cells, approximately three-fourth of the cell pairs that were induced to fuse by WT (i.e., cell pairs that transferred calcein) exhibited large pores. For cell pairs induced to fuse by the W666A mutant, approximately one-half exhibited large pores; for the other three mutants tested, an even smaller fraction of pores were enlarged. This order of final pore enlargement assessed with a large dye is the same as the order of rate of pore enlargement as measured by the rate of small dye transfer. In short, for each mutant, the extent and rate of fusion, as well as the growth of fusion pores, correlate quite well. That is, if one of these measures of fusion increased (or decreased), the other measures did so as well.

DISCUSSION

Crystal structures of three class I viral fusion proteins—influenza virus hemagglutinin (HA) (7, 52), simian virus 5 F (2, 54), and Ebola virus GP2 (22, 26)—are known in their initial and final states. The bundle is absent in the initial state of each of these proteins. Since these three proteins are contained by viruses of different families, it is thought that, in general, the 6HB is absent in the native, unactivated form of all class I fusion proteins (33). Although the structure of HIV gp41 is not known in the initial state, it is known that peptides which mimic HR2 do not bind to the initial structure but their presence prevents bundle formation, providing further evidence that the bundle is not present in the native, prefusion state (13). The absence of a 6HB in the initial structure of a class I fusion protein means that the bundle must form at some later point in the cascade of conformational changes that lead to the formation of a pore large enough to release the viral nucleocapsid into cytosol. When 6HB structures were first identified, it was commonly assumed that they formed relatively early in the fusion process and that this formation brought viral and target membranes into contact (8, 49). However, it was subsequently shown that for HIV Env, the formation of those 6HBs that participate in fusion occurs late in the process, and bundle formation is completed only after a pore has formed (28). In contrast to HIV Env, 6HBs may form prior to pore formation for Env of avian sarcoma and leukosis virus (27) and for HA of influenza virus (5, 39). For HIV Env, the sequence of amino acids intervening between the fusion peptide and HR2 is relatively short, whereas for avian sarcoma and leukosis virus Env and influenza HA, it is much longer. It appears that proximity between HR2 and fusion peptides causes tight linkage between bundle formation and pore formation/enlargement (11).

The 6HB contains a hairpin turn, residues 597 to 609, in the region between HR1 and HR2 of gp41. The results of the present study show that (except for Q577A) mutations of residues near this turn are quite deleterious to fusion, whereas mutations more removed from the turn are much less consequential in affecting the extent of fusion. This was the case for mutations within both HR1 and HR2 (Fig. 8). In fact, we found that the potency of mutation in HR1 and HR2 on fusion mapped to roughly the same cross section of the crystal structure bundle. It had previously been shown that a mutation of K574, also near the turn, eliminated fusion (15). This is in agreement with our present findings, although in this case, lysine's participation in a salt bridge, more than its topological location, may account for its importance in fusion (15). The overall pattern of greater extents of fusion for HR1 mutations at positions distal from the turn is in general agreement with fusion activity measured by a syncytium assay (23). However, the extent of syncytium formation was relatively low for the L556A mutant, whereas direct measurements of aqueous dye spread between cell pairs showed that fusion activity is high for this mutant, in accord with the overall pattern. Syncytium formation and aqueous dye spread measurements generally gave the same rank ordering for the HR2 mutants. The exceptions are the Y638A and L645A mutants, which exhibited high syncytium activity (47) but low direct fusion activity. Low fusion for these two mutants fits the overall pattern quite well.

FIG. 8.

Extent of fusion displayed after aligning N- and C-terminal residues according to a standard numbering system (19). The hairpin turn is comprised of residues 597 to 609. The mutated residues for which the extent of fusion is shown are underlined.

A simple interpretation of our data is that the bundle is initiated by binding of the HR2 residues closest to the hairpin into the portion of the HR1 grooves that are also adjacent to the turn (Fig. 8). This binding would nucleate packing of HR2 into HR1, allowing HR2 residues more removed from the turn to then bind to their HR1 positions within the bundle. Residues that directly participate in nucleation should be more important for bundle formation than residues that bind subsequent to nucleation. This interpretation has an appealing logic because residues on each side of the hairpin turn must be relatively close to each other, while more distal residues of HR1 and HR2 do not have to attain proximity. The proposed sequence of binding suggests a pathway for changes in the secondary structure of HR2 as it binds to HR1 grooves; peptides that mimic HR2 are unstructured in solution but are α-helical once they have bound to HR1 grooves (18, 31). Presumably, the same unstructured-to-structured transition holds for the HR2 region itself. We hypothesize that the unstructured-to-structured transition of HR2 is initiated as the region of HR2 closest to the hairpin turn binds to HR1; the transition of the initial bound portion into an α-helix allows the binding and transition of the remainder of HR2 into an α-helix to follow. However, the binding and structural transition should, a priori, depend on the primary sequence of the portions of the bundle that are distal to the hairpin turn. In fact, the primary sequences of these regions are highly conserved, suggesting that mutations in these regions are deleterious to viral fusion. Our data support these expectations. All aspects of fusion were adversely affected by mutation in the distal regions: extents of fusion were reduced, kinetics was slowed, and the rate of pore enlargement and final pore size were decreased. However, these effects were less pronounced in the distal region than in the proximal region (i.e., fusion activity was more sensitive to mutation in the regions proximal to the turn).

The consequences of the Q577A mutation provide the only prima facie evidence against our proposed model, but inspection of the three-dimensional structure of the 6HB indicates that Q577A could interact more favorably with HR2 (through interactions with W628) than does Q577. Since this position is close to the turn, a favorable mutation could even lead to greater fusion. The average amount of fusion was the same for WT and the mutant. The interaction between Q577 and W628 could also account, in part, for the smaller amount of fusion induced by the W628A mutant than WT (Fig. 2A); the hydrophobic Ala should interact less strongly with the polar Q577 than does the amphipathic Trp residue. Therefore, the W628A mutation should reduce fusion because of this weaker interaction as well as its position near the turn.

Structural and functional similarities between 6HBs of viral fusion proteins and intracellular SNARE complexes in eukaryotes have been discussed for many years (30, 44, 48). The pattern of folding we propose in the creation of a 6HB has a parallel in the folding of neuronal SNARE proteins. For neuronal SNAREs, the creation of tetrameric coiled-coil complexes that participate in fusion (i.e., the trans complexes) is initiated at the N termini of the monomers. The coiled coil is lengthened toward the transmembrane domains in a zippering process (9, 32) that is analogous to our conception of the process found in the 6HB of HIV Env. One cannot assume that because of similarities in class I viral fusion proteins, all 6HBs are likely to form by a zippering process, however. In contrast to neuronal SNAREs, it has been concluded that the coiled coil of SNARE complexes in Saccharomyces cerevisiae is not created by zippering but, rather, created over its entire length in a simultaneous, one-step reaction (56). Perhaps some class I viral fusion proteins also fold into a 6HB in a single step, rather than following the zippering action of HIV Env.

There is an appreciable delay between binding and the first event in fusion between Ef cells expressing HIV Env and Tg cells. This delay is due to the relatively slow early steps of CD4 binding and chemokine receptor engagement by Env (14, 40), steps that have been basically completed at the point of TAS (37). We previously concluded that the rate of rearrangement of Env toward the final, stable 6HB largely controls the kinetics of fusion after TAS has been achieved (28). Our present data provide additional support for this prior conclusion; point mutations of residues that directly participate in the HR1-HR2 interactions of the bundle reduce, in general, the kinetics of fusion from TAS. Although residues close to the hairpin turn are the most critical for fusion, residues removed from this region do have some effect on the fusion process, particularly on the rate of pore growth. This provides more evidence that bundle formation is completed only during pore growth.

In principle, mutations may inhibit bundle formation not only by reducing interactions between HR1 and HR2 but also by imposing unintended topological constraints. For example, the cytoplasmic tail (CT) of HIV Env is unusually long and, through its acylation (3), binds to the inner leaflet of the viral envelope, reducing possible movements. There is much evidence that the long CT hinders engagement of HIV Env to chemokine receptors (12, 53). The CT probably also limits the movement of the adjacent MSDs (1). Generally, the higher the rate of fusion, the less potent are the 6HB peptide inhibitors in blocking fusion (40). It is therefore possible that some point mutations were deleterious to fusion, not by directly inhibiting bundle formation, but by increasing topological constraints or by affecting nonbundle regions (55). But the findings that inhibition followed an orderly pattern in positional sequences and that mutations retarded pore growth, the final stage of fusion, argue that consequences of mutation were directly caused by reducing interactions between HR2 and the HR1 central coiled coil, rather than by secondary phenomena.

At the point an HIV Env-induced fusion pore forms, the extents of the rearrangements that still must occur in order for the 6HB to be completed are not known. One can imagine several possibilities. For example, one or two HR2 segments of Env may have packed into grooves of the coiled coil, while the third segment has not yet become bound. Alternatively, amino acids of HR2 that are proximal to the hairpin turn may be bound, but the distal residues have not yet done so. Our finding that the region of HR2 that is removed from the nucleation region can affect pore enlargement indicates strongly that for at least one monomer, this region has still not folded into a bundle. It may be that peptides that mimic HR2 (i.e., C peptides) inhibit pore growth (28) by binding to residues distal to the hairpin turn and the peptides then displace the native HR2 that is bound to residues nearer the turn. This zippering of inhibitory C peptides into grooves would occur in the reverse order of the natural zippering process. Reverse zippering in the artificial situation is consistent with the idea that the natural direction of binding between HR1 and HR2 is not conferred by the primary sequence but is due to the existence of the hairpin turn which would initiate closure from that direction.

Our proposed hypothesis for the nucleation of 6HBs is fundamentally a kinetic model; only the rate of bundle formation is of consequence. In the biological situation, bundle formation is irreversible. Therefore, the effects of mutation on fusion need not correlate with the effects on bundle stability. 6HBs created by N and C peptides are made reversible by the addition of high concentrations of denaturing agents such as guanididium (23). There is negligible correlation (not shown) between our extents of fusion and stability of bundles (as measured by melting temperatures in the presence of guanididium) created by N and C peptides with altered amino acids (23, 47). This noncorrelation between the experimental measures of bundle stability and fusion demonstrates that bundle stability per se is not a parameter relevant to the fusion process. It is the rate of bundle formation that is germane. Also, the hinge region of gp41 is absent (or is replaced by an unrelated very short sequence of amino acids) from a bundle created by C and N peptides, and so even the kinetics of bundle formation between peptides (31) does not necessarily correspond to that of native gp41.

It might have been thought that mutation of single residues could prevent bundle formation and therefore prevent fusion, thereby causing the process to be arrested at (or downstream of) an intermediate state of hemifusion, but the absence of additional dye spread upon CPZ addition indicates that mutation of single residues arrests fusion at states upstream of hemifusion (10, 34). Because hemifusion occurs prior to completion of 6HB formation, the complete 6HB is not required for hemifusion. We conclude that a portion of the 6HB must be created to achieve hemifusion (35). A large amount of free energy is released by transition from the initial state to the final state of HIV Env (16, 31). If the 6HB bundle forms in a stepwise process, a fraction of the total free energy would be released during each step. These dispensed packets of energy could be used to drive a series of energetically unfavorable lipid rearrangements that are necessary for fusion to occur.