The Final Conformation of the Complete Ectodomain of the HA2 Subunit of Influenza Hemagglutinin Can by Itself Drive Low pH-dependent Fusion

Abstract

One of the best characterized fusion proteins, the influenza virus hemagglutinin (HA), mediates fusion between the viral envelope and the endosomal membrane during viral entry into the cell. In the initial conformation of HA, its fusogenic subunit, the transmembrane protein HA2, is locked in a metastable conformation by the receptor-binding HA1 subunit of HA. Acidification in the endosome triggers HA2 refolding toward the final lowest energy conformation. Is the fusion process driven by this final conformation or, as often suggested, by the energy released by protein restructuring? Here we explored structural properties as well as the fusogenic activity of the full sized trimeric HA2(1–185) (here called HA2*) that presents the final conformation of the HA2 ectodomain. We found HA2* to mediate fusion between lipid bilayers and between biological membranes in a low pH-dependent manner. Two mutations known to inhibit HA-mediated fusion strongly inhibited the fusogenic activity of HA2*. At surface densities similar to those of HA in the influenza virus particle, HA2* formed small fusion pores but did not expand them. Our results confirm that the HA1 subunit responsible for receptor binding as well as the transmembrane and cytosolic domains of HA2 is not required for fusion pore opening and substantiate the hypothesis that the final form of HA2 is more important for fusion than the conformational change that generates this form.

Introduction

Fusion mediated by the influenza virus hemagglutinin (HA) protein is often considered as a prototype of biological fusion reactions. This fusion process is utilized by the virus to deliver its RNA into the host cell by merging the viral envelope with the membrane of an acidified endosome of the host cell. Each monomer of homotrimeric HA consists of two disulfide-linked subunits: HA1, responsible for receptor binding, and HA2. In the native neutral pH conformation of the HA protein, the HA1 subunit confines the HA2 subunit in a metastable state that has been termed the “spring-loaded” state (1–3). However, evidence from calorimetric studies indicates that the compact folded structure of the influenza hemagglutinin protein is not a kinetically trapped metastable high energy form (4, 5). In the neutral pH structure of intact HA, the functionally important N-terminal amphiphilic region of HA2, referred to as the fusion peptide, is hidden within the HA molecule. The low pH-triggered restructuring of HA unlocks HA2 and allows refolding of the HA2 subunit toward its final, low energy conformation, which consists of a hairpin structure with the fusion peptide and the transmembrane domain at the same end of the rigid rod. Although it is commonly assumed that the energy released by this restructuring drives rearrangements of membrane bilayers (6–8), one may suggest an alternative hypothesis, i.e. that the final conformation itself is fusogenic.

In our earlier work, to test the fusogenic properties of the final conformation of the HA2 subunit in the absence of HA1, we studied a trimeric polypeptide, HA2(1–127), representing the first 127 amino acid residues of the 221 residues of the HA2 domain of influenza virus X-31 (9, 10) (see Fig. 1C). The conformation of HA2(1–127) both at neutral and at low pH shows many characteristic features of the lowest energy hairpin form of HA2 (11–13). HA2(1–127) did not mediate content mixing between bound cells and thus did not open fusion pores connecting the cells. However, HA2(1–127) induces lipid mixing between liposomes and between bound cells in a low pH-dependent manner. Although finding that a major portion of the final conformation of HA has fusogenic activity was intriguing, it had to be considered that HA2(1–127) represented only two-thirds of the HA2 ectodomain. Because a significant part of the central coiled coil of HA2(1–127) is not covered by the outer layer of antiparallel polypeptides of the C-terminal half of the HA2 ectodomain, one may hypothesize that hemifusion mediated by HA2(1–127) is not characteristic of the entire HA2 subunit. If this is the case, then a longer polypeptide that represents the entire HA2 ectodomain (3, 13) is expected to have no fusogenic activity. Up to the present time, the ability of the entire HA2 ectodomain to promote fusion had not yet been investigated.

FIGURE 1.

Schematic diagram of three different HA2 constructs of influenza virus hemagglutinin. A, HA2, corresponding to residues 1–221 of the full-length intact protein of the HA2 subunit containing the transmembrane (TM) and short cytoplasmic domains. B, HA2*, corresponding to residues 1–185 of the full-length ectodomain of the HA2 subunit. C, HA2(1–127), corresponding to residues 1–127 of the ectodomain of the HA2 subunit.

In the present work, we report the fusogenic activity of a 185-amino acid residue trimeric polypeptide, HA2*,⁵ that represents the entire ectodomain of the X-31 influenza virus HA2 subunit, i.e. almost the entire subunit, lacking only the transmembrane domain and the cytoplasmic segment (see Fig. 1B). HA2* mediated lipid mixing between liposomes and between cells in a low pH-dependent manner. Importantly, HA2* also mediated content mixing between the cells, and thus, in contrast to HA2(1–127) (10), it drove membrane rearrangements beyond hemifusion to the opening of a nascent fusion pore. The surface density of HA2* molecules required for fusion was close to the number of HA molecules required for HA-mediated fusion. Two point mutations in the HA2 sequence that are known to inhibit HA-mediated fusion also inhibited HA2*-mediated fusion. Thus, HA2*, which presents the final conformation of the HA2 subunit of HA, mediates a fusion process that shares important characteristics with the fusion mediated by full-sized HA. Our findings substantiate the hypothesis that neither the transmembrane domain nor the energy released in the conformational transition of the HA2 subunit from its initial metastable form in HA toward the final hairpin conformation of HA2 in the low pH form of HA is required for the early stages of fusion.

EXPERIMENTAL PROCEDURES

Expression and Purification of HA2*

The full-length ectodomain of the HA2 subunit of the hemagglutinin protein of influenza virus, HA2*, was subcloned into a pET-24b(+) plasmid vector (Novagen) with NdeI and XhoI/SalI restriction sites from the plasmid pHA containing full-length HA of the X-31 strain (Fig. 1). The strain used for expression of plasmids bearing HA2* or its site-specific mutant, G1E, was Escherichia coli Rosetta (DE3)pLysS (Novagen). The amino acid sequence of HA2* is Met-[HA2 amino acids 1–185]-Val-Glu-His₆ without the transmembrane domain and the short cytoplasmic segment on the C terminus. The C137S mutation was introduced for expression (14). Mutants were generated by QuikChange^TM site-directed mutagenesis (Stratagene) and confirmed by DNA sequencing. The cells were grown in 1.8 liter of LB medium with 2 g/liter glucose (with 25 μg/ml kanamycin and 50 μg/ml chloramphenicol) at 37 °C with inoculation of the overnight culture (1%, v/v). When A₆₀₀ reached 0.8–1.0, the cells were induced with 0.5 mm isopropyl β-d-galactopyranoside, the temperature was lowered from 37 to 22 °C, and the agitation speed was reduced from 200 to ∼100 rpm for 6 h. The cells were harvested by centrifugation at 6,000 × g for 10 min at 4 °C and stored at −80 °C. The cell pellet was resuspended in 20 mm imidazole buffer (50 mm NaH₂PO₄, 300 mm NaCl, pH 8.0, 0.2% (v/v) Triton X-100) with additional 0.3% (v/v) Triton X-100 (final concentration, 0.5%), 2 mm 4-(2-aminoethyl)benzenesulfonyl fluoride, 1 mm leupeptin, and 1 mm pepstatin A and broken on ice by sonication. The cell lysate was vortexed for 1 min and swirled at 4 °C for 30 min with 0.5% (w/v) n-lauroyl sarcosine. After centrifugation at 15,000 × g for 15 min at 4 °C, the supernatant was swirled at 4 °C for 1 h with 2 ml of nickel-nitrilotriacetic acid-agarose beads (Qiagen) equilibrated with 20 mm imidazole buffer. The column was washed thoroughly with 20 mm imidazole buffer and with 10 column volumes of 50 mm imidazole buffer. The proteins were eluted stepwise with 4 column volumes each of 100 mm imidazole, 150 mm imidazole, and 200 mm imidazole buffers containing 0.2% (v/v) reduced Triton X-100. The most concentrated protein fraction was eluted with 150 mm imidazole buffer with over 95% purity by 15% SDS-PAGE. HA2(1–127) was expressed and purified as described earlier (9, 10).

Preparation of Large Unilamellar Vesicles (LUVs)

All lipids, including the fluorescently labeled lipids, were purchased from Avanti Polar Lipids (Alabaster, AL). Lipids were dissolved in chloroform:methanol (2:1, v/v) at the desired molar ratio. The lipid was deposited as a film on the wall of a test tube by solvent evaporation with nitrogen. Final traces of solvent were removed for 2–3 h in a vacuum chamber attached to a liquid nitrogen trap. The lipid films were suspended in the appropriate buffer by vortexing at room temperature. The lipid suspensions were further processed with five cycles of freezing and thawing followed by 10 passes through two stacked 0.1-μm polycarbonate filters using an extruder made by the Lipex Corp. (Vancouver, British Columbia, Canada). The content of lipid phosphorous was determined by the method of Ames (15).

Lipid Mixing Assay for Liposome Fusion

The resonance energy transfer assay of Struck et al. (16) was used to monitor membrane fusion. Two populations of LUVs were prepared, one unlabeled and one labeled with 2 mol % each of N-(lissamine rhodamine B sulfonyl)phosphatidylethanolamine and N-(7-nitro-2,1,3-benzoxadiazol-4-yl)phosphatidylethanolamine. A 9:1 molar ratio of unlabeled to labeled liposomes was used in the assay. Fluorescence was recorded at excitation and emission wavelengths of 465 and 530 nm, respectively, using a 490-nm cutoff filter placed between the cuvette and the emission monochromator with 4-nm bandwidths using an SLM Aminco Bowman AB-2 spectrofluorometer. Measurements were made using siliconized glass cuvettes (1 cm²) with continuous stirring at constant temperature in a thermostated cuvette holder. Measurements were carried out using a buffer containing 5 mm Hepes, 5 mm Mes, 5 mm citric acid, 0.15 m NaCl, and 1 mm EDTA, pH 7.4. LUVs were added to 2 ml of buffer in the cuvette at 37 °C to give a final lipid concentration of 100 μm. No lipid mixing occurs at neutral pH in the absence of protein. A small aliquot of protein solution was then added. After monitoring the slow rate of lipid mixing at neutral pH, the solution was acidified to the desired pH by the addition of 1 mm citric acid. Fluorescence was recorded for several minutes, and then 20 μl of 10% Triton X-100 was added (final concentration, 0.1%). The initial residual fluorescence intensity prior to acidification, F_o, was taken as 0. The maximum fluorescence intensity, F_max, was obtained by dilution of the labeled lipids with 20 μl of 10% Triton X-100. The percentage of lipid mixing at time t is given by the following equation: ((F_t − F_o)/(F_max − F_o))100. All runs were done in duplicate and were found to be in close agreement. Acidification of the LUVs in the absence of protein was used as a negative control and showed little lipid mixing.

Circular Dichroism (CD)

The CD spectra were recorded using an AVIV series 215 CD instrument (AVIV Associates, Lakewood, NJ). The sample was contained in a 1-mm-path length cell that was maintained at 25 °C in a temperature-regulated cell holder. The CD data are expressed as the mean residue ellipticity. To ensure transparency in the UV range, proteins were dialyzed against phosphate buffer containing 0.1% reduced Triton X-100. All CD runs were made with aliquots of dialyzed protein diluted severalfold into 10 mm NaH₂PO₄, 0.15 m NaF, pH 7.4. The curves were analyzed for secondary structure content by the Contin and Selcon methods with the program CDPro (17). Protein concentrations were determined by the BCA assay (Pierce) using the same buffer as a blank.

Preparation of Cells and Cell Fusion Experiments

In our cell fusion experiments, we used two well characterized experimental systems: fusion between HAb2 cells and red blood cells (RBCs) and fusion between Sf9 cells. HAb2 cells express an uncleaved precursor form of HA, HA0, which is fusion-incompetent but mediates membrane binding (10, 18). Thus, in the case of fusion between HA0-expressing HAb2 cells (HA0 cells) and RBCs, HA2* fused the cells bound by physiologically relevant interactions between HA1 and sialic acid receptors. Experiments on Sf9 cells allowed us to evaluate the fusogenic activity of the polypeptide in the absence of any full sized HA, including fusion-incompetent HA0. RBCs labeled with the fluorescent lipid PKH26 (Sigma) and loaded with either 6-carboxyfluorescein (CF) or 10-kDa FITC-dextran (Invitrogen) and HA0 cells (HAb2 cells expressing the HA0 form of A/Japan/305/57 HA, subtype H2N2 (19)) were prepared for experiments as described (18). Spodoptera frugiperda (Sf9) cells were grown and labeled with l-α-phosphatidylethanolamine-N-(lissamine rhodamine B sulfonyl) (egg) (Avanti Polar Lipids) as described (20). In fusion assays, HA0 cells with 0–2 bound RBCs per cell or Sf9 cells were incubated for 5 min in PBS supplemented with the stated concentration of HA2*. The low pH medium (PBS titrated with citrate to the desired acidic pH) was applied at room temperature for a stated time. Low pH application was ended by replacement of the acidic solution with PBS. We assayed the final extents of lipid and content mixing in HA0 cell/RBC fusion by fluorescence microscopy as the ratio of dye-redistributed bound RBCs to the total number of bound RBCs (18). We should point out that this method of measurement of content mixing is independent of dye leakage because leakage would result in a in a huge (10 billion-fold) dilution of the probe that would not be observable because the fluorescence intensity would be too low.

Lipid mixing between Sf9 cells was quantified in a similar way as l-α-phosphatidylethanolamine-N-(lissamine rhodamine B sulfonyl) transfer from prelabeled to unlabeled Sf9 cells. We also used light microscopy to measure the percentage of Sf9 cells in syncytia (the ratio of nuclei within syncytia to the total number of cell nuclei in the same field) (20). Because fusion extents for the same concentration of HA2* and pH somewhat varied from day to day (apparently as a result of variation in the numbers of the cells and the properties of RBCs and RBC ghosts), we routinely started the experiments by adjusting conditions of the HA2* and low pH applications. Each set of experiments for each graph presented in the study was repeated on at least three occasions with similar results. Presented data were averaged from the same set of experiments.

Evaluation of Cell Surface Concentration of HA2*

Concentrations of cell-associated polypeptide sufficient to induce cell fusion were estimated as described (10). Approximately 10⁶ Sf9 cells with bound polypeptide were lifted from the plates with EDTA/EGTA (0.5 mg/ml each) in Ca⁺²- and Mg⁺²-free PBS, collected, and lysed in lysis buffer (50 mm Tris-HCl, pH 8.0, 150 mm NaCl, 5 mm EDTA, 1.5% Triton X-100, 10% glycerol, one proteinase inhibitor mixture tablet (Roche Applied Science)/50 ml of buffer). After lysis, samples were diluted 2-fold with denaturing and reducing SDS sample buffer (0.5 m Tris-HCl, pH 6.8, 10% SDS, 20% glycerol, 20 mm dithiothreitol, 0.1% bromphenol blue) and boiled for 3 min. The lysate and known dilutions of the same polypeptide were analyzed by 4–20% gradient SDS-PAGE. For Western blotting, we used rabbit polyclonal antiserum against HA2 LOOP-36 (21) (kindly provided by Dr. P. S. Kim) at 1:1,000 dilution and secondary antibodies (goat anti-rabbit IgG; 1:1,000) conjugated with horseradish peroxidase (Amersham Biosciences). To detect protein immunoreactivity, blots were incubated in the ECL reagent (Amersham Biosciences) followed by exposure to film. The amount of the bound polypeptide was found by comparison of the polypeptide band intensity with that of our standards.

RESULTS

Liposome Lipid Mixing

The HA2* exhibited considerable pH-dependent lipid mixing activity in liposomes composed of DOPC:DOPE:cholesterol (1:1:1) (Fig. 2A). There was virtually no lipid mixing activity at neutral pH, but after acidification to pH 5.0, the rate of lipid mixing was comparable with that observed with the shorter 127-amino acid construct, HA2(1–127) (9). HA2* had about a 2-fold greater activity than G1E-HA2* (Fig. 2A). This mutation in the first position of the HA2 is known to abolish HA-mediated fusion (22, 23). The difference between HA2* and G1E-HA2* became greater in the less fusogenic liposomes composed of DOPC:cholesterol (1:1) (Fig. 2B). The pH dependence of HA2*-mediated lipid mixing for DOPC:cholesterol (1:1) liposomes was shifted to more acidic pH values than that for DOPC:DOPE:cholesterol (1:1:1) (Fig. 2C). In the case of the G1E-HA2*, the pH dependence of the lipid mixing with LUVs of DOPC:DOPE:cholesterol (1:1:1) was different from that of HA2* with some lipid mixing observed at pH 7.4 with the mutant (Fig. 2D) but not with the wild type sequence (Fig. 2C). However, for both constructs, there was a marked increase in the rate at about pH 6.

FIGURE 2.

Promotion of lipid mixing in LUVs. A, 0.5 μm HA2* (curve 1) or G1E-HA2* (curve 2) with 100 μm LUVs of DOPC:DOPE:cholesterol (1:1:1) at pH 5.0 at 37 °C. The mixture at pH 7.4 was acidified to pH 5.0 at time 0. Curve 3 shows the spontaneous rate of lipid mixing in the absence of peptide. B, lipid mixing curves as in A except that the lipid composition of the LUVs is DOPC:cholesterol (1:1). C, pH dependence of lipid mixing. Lipid mixing was measured with HA2* at different pH values in 100 μm LUVs composed of DOPC:DOPE:cholesterol (1:1:1) (■) or DOPC:cholesterol (1:1) (□). The percentage of lipid mixing is expressed as the value measured at each pH at 300 s relative to that measured at pH 7.4. D, pH dependence of the effects on membrane fusion with 100 μm LUVs composed of DOPC:DOPE:cholesterol (1:1:1) as in C but with G1E-HA2*.

In brief, HA2*-induced lipid mixing between DOPC:cholesterol liposomes depended on low pH (similar to lipid mixing mediated by HA) and was strongly inhibited by the G1E mutation. The G1E mutation in intact HA also inhibits HA-mediated fusion (22, 23). Our results indicate that studies of membrane fusion activity using liposomes that are at the brink of spontaneous lipid mixing, such as those of DOPC:DOPE: cholesterol (1:1:1), are not as discriminatory of the fusogenic activity of different protein constructs with different intrinsic fusogenic activity. This is also indicated by the fact that there was some fusion activity of the G1E-HA2* with these liposomes, but no fusion was observed with the intact G1E mutant in the experiments on cell fusion (22).

CD

To explore the effects of the single residue mutation that distinguishes G1E-HA2* from HA2*, we compared the secondary structures of the fusogenic HA2* with the fusion-incompetent G1E-HA2* using far-UV CD (Fig. 3A). Secondary structure analysis is given in Table 1 and shows that HA2* has lower structure content than its G1E mutant.

FIGURE 3.

CD analysis of proteins. A, secondary structure. Shown is the CD spectrum at 25 °C in 10 mm phosphate buffer, 0.14 m NaF, pH 7.4. ■, 13.6 μm HA2*; □, 12.8 μm G1E-HA2*. Solutions contained 0.00033 mg/ml reduced Triton X-100. B–E, temperature dependence of the ellipticity at 222 nm. The bottom curve in each graph shows the heating scan, and the top curve is the cooling scan. Each data point was accumulated over a period of 20 s with a point taken every degree (deg) at a scan rate of 1.5 °C/min. B, 13.6 μm HA2* at pH 7.4 in phosphate buffer. C, 12.8 μm G1E-HA2* at pH 7.4 in phosphate buffer. D, 13.6 μm HA2* at pH 5 in phosphate buffer. E, HA2(1–127) at pH 7.4 in phosphate buffer.

TABLE 1.

Secondary structure estimation from CD measurements

Protein	α-Helix	β-Structure	Turns	Other
	%	%	%	%
HA2*	24	25	20	31
G1E-HA2*	40	18	19	23

The thermal denaturation of HA2* at neutral pH measured by CD exhibited a marked loss of secondary structure over a broad temperature range from about 55 to 90 °C (Fig. 3B). There are two components of the thermal transition: one around 65 °C and the other at about 85 °C. The lower temperature transition is similar to that seen with the intact protein (24, 25). The transitions are partially reversible on cooling. Similar behavior was observed with G1E-HA2* (Fig. 3C), although in this case, the two components were less distinct, and the reversibility was less marked.

The thermal denaturation of HA2* at pH 5 was irreversible and showed an initial loss of structure at 25 °C (Fig. 3D). The thermal denaturation of G1E-HA2* and HA2(1–127) at pH 5 could not be carried out using the buffer conditions required for CD due to precipitation of the proteins. However, at pH 7.4, the behavior of HA2(1–127) (Fig. 3E) was similar to that of the G1E-HA2* and HA2* proteins.

We also studied the unfolding of HA2* by differential scanning calorimetry. The major endotherm occurred at 86.8 °C, much higher than the transition observed by CD. We have no explanation for this discrepancy, but it should be noted that CD and differential scanning calorimetry measure different things, and the denaturation is irreversible and therefore influenced by kinetic factors. It is not possible to have identical heating protocols for both methods, but evaluation of this issue was not pursed. Our data indicate that replacing a single amino acid residue in G1E-HA2* results in a notable increase in the total α-helical content in the secondary structure of HA2* rather than just in a local change in the immediate vicinity of the fusion peptide.

Cell Fusion: HA2* Mediates Lipid and Content Mixing but Does Not Expand Fusion Pores

The ability of HA2* to fuse biological membranes was explored using two experimental systems. To test whether the HA2* is capable of mediating fusion of membranes bound by physiologically relevant interactions between HA1 and sialic acid receptors, we studied fusion of RBCs with HA0 cells. The HA0 form of HA is competent for receptor binding but incompetent for fusion if it is not activated by trypsin cleavage into the fusogenic HA1-HA2 form. In the second experimental system, to test whether the HA2* fuses cells in the absence of any wild type HA, we also used Sf9 cells, which spontaneously establish physiological contacts and thus do not require any artificial means of bringing cells into contact.

Application of HA2* to HA0 cells with bound RBCs followed by lowering the pH resulted in lipid mixing between the cell membranes detected as a rapid redistribution of the fluorescent lipid PKH26 from the RBCs to the HA0 cell membranes. The extent of lipid mixing depended upon HA2* concentration (Fig. 4A) and pH (Fig. 4B). Because HA2* was added to cells in 0.05% Triton X-100, we verified that low pH application to the cells in the presence of HA2*-free 0.05% Triton X-100 yielded no measurable lipid mixing (data not shown).

FIGURE 4.

Low pH-dependent HA2*-mediated fusion between HA0 cell and bound RBC. The points in A and B represent the final extents of lipid mixing (mean ± S.E., n = 3) assayed more than 20 min after low pH application. A, concentration dependence. Shown is lipid mixing between PKH26-labeled RBCs and HA0 cells after a 2-min application of pH 4.9 medium in the presence of different concentrations of HA2* (filled circles), G1E mutant HA2* (filled squares), or I173E mutant HA2* (open circles). B, pH dependence. Shown is lipid mixing between PKH26-labeled RBCs and HA0 cells after a 5-min application of low pH medium in the presence of 1.5 μm HA2* or I173E mutant HA2* (filled and open circles, respectively). C, HA2* induces lipid mixing and opens fusion pores that transfer CF but that are not large enough to pass 10-kDa dextran. Fusion between HA0 cells and bound RBCs was triggered by a 5-min application of pH 4.9 medium in the presence of 2.5 μm HA2* (bars 3, 4, and 5). In control experiments, we omitted application of either HA2* (bar 1) or low pH medium (bar 2). Fusion was assayed as lipid mixing by redistribution of the probe PKH26 (bars 1, 2, and 3), CF (bar 4), or 10-kDa dextran (bar 5). Data are presented as mean ± S.E. (n = 3). D, in contrast to HA2(1–127) that mediates only lipid mixing, HA2* also mediates content mixing. In side-by-side experiment, HA0 cells with bound PKH26-labeled and CF-loaded RBC ghosts were treated with a 5 μm concentration of either HA2(1–127) (bars 1 and 3) or HA2* (bars 2 and 4) for 5-min and then with pH 4.9 medium for 5 min at room temperature. Final extents of lipid mixing (bars 1 and 2) and content mixing (bars 3 and 4) were assayed by fluorescence microscopy and are presented as mean ± S.E. (n ≥ 3). Error bars show S.E.

In many fusion processes, the merger of contacting leaflets of the membranes that allows lipid mixing precedes the merger of the distal leaflets to form a fusion pore that allows content mixing. In contrast to HA2(1–127) that mediates only lipid mixing (10), HA2* formed fusion pores large enough to pass CF. These pores did not expand to sizes large enough to allow redistribution of a larger aqueous probe, 10-kDa FITC-dextran (Fig. 4C). Neither lipid mixing nor syncytium formation was observed in the control experiments where either low pH or HA2* application was omitted (not shown). Both the fusogenic activity of HA2* and its inability to form the expanding fusion pores that we observed using HA0 cell-RBC complexes were confirmed with Sf9 cells where application of HA2* resulted in lipid mixing between these cells without the formation of syncytia (not shown).

We verified the important difference between fusogenic properties of HA2* and HA2(1–127) in a side-by-side comparison (Fig. 4D). Although these two constructs induced similar extents of lipid mixing, the longer construct, HA2*, was much more effective in inducing content mixing.

The dependence on low pH is not the only feature of HA2*-mediated fusion that is in common with fusion mediated by full sized HA. We demonstrated that the same mutations that reduced the fusogenic activity of full sized HA also caused the loss of fusogenic activity of HA2* (Figs. 4, A and B, and 5A). A single amino acid substitution at position 173 (I173E), detrimental for HA-mediated fusion (26), significantly inhibited the fusogenic activity of HA2*. Similarly, a substitution of glutamic acid for the glycine in the N terminus (G1E) of HA2 that is known to inhibit HA-mediated fusion (22) resulted in an almost complete loss of the fusogenic activity of the HA2*. Even at concentrations more than 3-fold higher than the concentrations at which wild type HA2* mediated effective fusion, G1E-HA2* mediated no CF redistribution and produced almost no lipid mixing (Fig. 5A). Similar to fusion mediated by full size HA (27), fusion mediated by HA2* was blocked by lowering the temperature to 4 °C (Fig. 5B).

FIGURE 5.

Fusion mediated by HA2* reproduces several hallmarks of fusion mediated by full-sized HA. A, fusogenic activity of the HA2* is inhibited by mutations known to inhibit fusion mediated by full sized HA. Fusion between HA0 cells and bound RBCs was triggered by a 2-min application of pH 5.3 medium in the presence of 4 μm HA2*, 15 μm G1E mutant HA2*, or 5 μm I173E mutant HA2*. Fusion was assayed as redistribution of lipid probe PKH26 (gray bars) or the small aqueous probe CF (black bars). Data are presented as mean ± S.E. (n ≥ 3). B, HA2*-mediated fusion is inhibited at 4 °C. HA0 cells with bound CF-loaded RBC ghosts were treated with 3.2 μm HA2* for 5 min at 37 °C. The temperature was lowered to 4 °C, and pH 4.9 medium was applied for 10 min still at 4 °C. Then the cold low pH medium was replaced with warm (37 °C) medium of pH 7.4 (bar 2) or pH 4.9 (bar 3). Bar 1, HA0 cells with bound CF-loaded RBC ghosts were first treated with 3.2 μm HA2* for 5 min at 37 °C and then incubated at pH 4.9 for 5 min still at 37 °C. Final extents of content mixing were assayed by fluorescence microscopy and are presented as mean ± S.E. (n ≥ 3). C, HA2* fuses membranes when added to only one of the two fusing membranes and is inactivated if acidified in the absence of target membrane. Bar 1, HA0 cells with bound CF-loaded RBC ghosts were treated with 3.2 μm HA2* for 5 min at 22 °C and then for 5 min with pH 4.9 medium (22 °C). Bar 2, HA0 cells were treated with 3.2 μm HA2* for 5-min at 22 °C and washed with HA2*-free medium. Then we applied CF-loaded RBC ghosts, washed the cells to remove the unbound ghosts, and finally treated HA0 cell-RBC complexes with a 5-min pulse of pH 4.9 medium (22 °C). Bars 3 and 4, HA0 cells were treated with 3.2 μm HA2* for 5 min at 37 °C, washed with HA2*-free medium, and then treated for 5 min with pH 4.9 medium. We then applied CF-loaded RBC ghosts, washed the cells to remove the unbound ghosts, and either treated (bar 4) or did not treat (bar 3) HA0 cell-RBC complexes with a 5-min pulse of pH 4.9 medium. Final extents of content mixing were assayed by fluorescence microscopy and are presented as mean ± S.E. (n ≥ 3). Error bars show S.E.

Like many other viral fusogens, full size HA mediated fusion when expressed in only one of the two fusing membranes and was inactivated if treated with low pH in the absence of the target membrane. Likewise, we found HA2* to fuse membranes when added to only one of the two fusing membranes and to be inactivated if acidified prior to application of the target cells, RBCs (Fig. 5C).

Importantly, we found HA2* to fuse cells in the experiments where application of HA2* to HA0 cell-RBC complexes (Fig. 5C) or to Sf9 cells (not shown) was followed by washing the cells with HA2*-free medium prior to acidification. This tightness of the HA2* cell surface binding allowed us to estimate the surface density of HA2* required to observe HA2*-mediated fusion of Sf9 cells. We treated the cells with HA2*, washed unbound polypeptide away, and lifted and lysed the cells with bound polypeptide. We then used quantitative Western blotting to compare the HA2* band in a cell lysate with the bands obtained with known amounts of the free polypeptide. The total surface area of the plasma membrane of cells that is available for HA2* binding was evaluated as a product of the known number of the cells and the average area of cell surface, which was estimated on the basis of the capacitance of Sf9 cells reported in electrophysiological studies (∼16 picofarads (28)) using 1 microfarad/cm² as the specific capacitance of a membrane bilayer. According to this rough estimate, a 4 μm concentration of HA2* at which HA2* effectively fuses Sf9 cells corresponds to a surface concentration of ∼5,000 trimers/μm². This concentration is comparable with surface concentrations of full sized HA required for membrane fusion that vary from ∼2,000 in HA-expressing cells (29) to ∼15,000 trimers/μm² for the viral envelope (30). Thus, it appears that HA2* and full sized HA form fusion pores at comparable surface densities.

DISCUSSION

In this study, we explored the fusogenic activity of the complete 185-amino acid ectodomain of HA2 that represents the final conformation of the HA2 subunit of HA. As previously found with the shorter 127-amino acid construct, HA2(1–127) (9, 10), HA2* induced lipid mixing between liposomes (Fig. 2, see also Ref. 70) and between cells (Figs. 4 and 5). However, in contrast to HA2(1–127), HA2* drove fusion beyond hemifusion to fusion pore opening (Fig. 4, C and D). Thus, the addition of residues 128–185 allows the peptide to more closely resemble the full sized HA with regard to its ability to form a fusion pore. Like HA, both HA2(1–127) and HA2* mediate lipid and content mixing only after application of acidic pH. In the case of the intact HA, acidification causes dissociation of the HA1 and HA2 subunits, resulting in a large conformational change in HA2 (1, 12, 21). However, with HA2(1–127) and HA2*, there is no HA1 subunit, and there is little conformational change in these constructs upon acidification. Nevertheless, the fusion results with the HA2 constructs indicate that there must be functionally important low pH-dependent interactions between the final conformations of HA2 subunits and/or between these conformations and the membranes. It has been suggested that protonation of HA leads to a loosening of the structure of HA as a result of electrostatic repulsion (31, 32).

The nature of the low pH-dependent stage of HA2*-mediated fusion is not completely resolved. The pH dependence of fusion of the native virus is complex. The kinetics show a lag time as well as a subsequent rate of fusion that depends on the temperature, the pH, and the nature of the target membrane. In addition, the kinetics involve both a binding step and the fusion step (33). The rate of the fusion step is not directly given by the overall rate of fusion. Furthermore, additional effects that are subtle to detect may influence viral protein-promoted processes. For example, a small difference in the pH dependence of hemolysis between an intact influenza virus and the isolated hemagglutinin protein has been observed (34). The detailed study of the pH dependence of fusion by Remeta et al. (24) shows that at 35 °C the pH required for half-maximal activity for lipid mixing with erythrocyte ghosts was 5.50. The pH required for half-maximal promotion of heterokaryon formation with cells expressing the influenza virus X-31 hemagglutinin protein is 5.2 (35), whereas the pH required for half-maximal hemolytic activity is 5.75 (36). Our results for the pH dependence of the HA2* is in the range of those observed for other pH-dependent functions of the HA of the X-31 strain of influenza virus. Curiously, the pH dependence of HA2* more closely resembles that of the shorter 90-amino acid fragment of HA2 (37) than it does the longer 127-amino acid HA2(1–127) (9). This relative comparison also holds for the G1E mutation in that it is less inhibitory for both the 90- and 185-amino acid fragments than it is for the 127- amino acid fragment. This amino acid substitution in the intact hemagglutinin (22, 38) or in HA2(1–127) (37) results in virtually complete loss of fusion activity. There is also partial loss of lipid mixing activity with liposomes of the G1E mutant in the 90-amino acid fragment of HA2 (37) and in the 185-amino acid fragment (Fig. 2A). Fusogenic activity of the G1E mutant of HA2* detected in DOPC:DOPE:cholesterol (1:1:1) liposomes, even at neutral pH, suggests that for these liposomes protein-mediated fusion is overshadowed by spontaneous lipid mixing between very fusogenic bilayers (39). However, with the less fusogenic liposomes (DOPC:cholesterol, 1:1), the G1E mutant of HA2* had very little activity (Fig. 2B). These results indicate that the use of liposomes comprising very fusogenic lipids can give rise to non-physiological fusion.

There is a significantly greater helix content in the G1E-HA2* mutant protein compared with HA2* (Table 1). Some factors that would promote more helix in the G1E mutant are the fact that Glu is a better helix former than Gly; charge-dipole interactions between the negative charge on Glu and the helix dipole would stabilize the helix. In addition, there is a bend in the conformation of the fusion peptide when inserted into a membrane, bringing the N terminus close to the membrane interface (40, 41). The negative charge on Glu at the N terminus would facilitate forming this kinked structure and hence facilitate achieving the peptide conformation that promotes membrane fusion. Hence, the greater helicity that we observed with G1E-HA2* follows expectations. The quantitative extent of the increase in helicity is difficult to predict, but our results are in accord with theoretical calculations, indicating that the conformational change should not be local but more extensive (42).

Our findings argue against the hypothesis that the energy for membrane fusion is provided by the HA2 restructuring from its initial conformation to the final “spring-unloaded” form of HA2 (21, 43–45). Even at neutral pH, HA2* is already in the final low pH form equivalent to the HA2 subunit of the intact HA that is capable of fusion. Therefore, our results would indicate that at least the early stages of HA-mediated fusion, up to an opening of a small fusion pore, are not driven by the restructuring of an individual HA2 trimer toward its final conformation but rather by low pH-dependent interactions between this conformation and the membrane and by lateral interactions between adjacent low pH forms of the trimers and/or among the subunits of the trimer (Fig. 6).

FIGURE 6.

Schematic diagram depicting two mechanisms by which HA restructuring at low pH may be coupled to membrane fusion. HA1 subunits in HA homotrimers are shown as orange circles. HA2 subunits responsible for HA-mediated fusion are shown as dark blue shapes. Fusion peptides, amphiphilic N-terminal regions of HA2 exposed only in low pH forms of the protein, are shown as green shapes. For the sake of simplicity, one of the two fusing membranes is not shown. A and B, fusion can be driven by the conformational energy released in HA2 restructuring from the initial neutral pH form to the low pH hairpin form. In this scenario, the hairpin conformation represents the “discharged” postfusion form of HA2. A′, B′, and C′, low pH-triggered restructuring of HA2 produces the hairpin form of the protein, and this form mediates low pH-dependent fusion. In this hypothetical scenario, the energy for membrane remodeling may be provided by interactions between adjacent HA2 molecules and between HA2 and membranes. The finding that hairpin conformations of the HA2 ectodomain form fusion pores in a low pH-dependent manner substantiates the pathway A′–B′–C′.

Proteins that fuse membranes during enveloped virus entry (46, 47) as well as established intracellular and developmental fusogens (48–51) are anchored in the membranes by their transmembrane domains. Moreover, for several viral fusion reactions and for SNARE-dependent intracellular fusion, the sequence of the fusion protein transmembrane domain affects its fusogenic activity (52, 53). However, our finding that HA2* forms fusion pores argues against the hypothesis that only integral membrane proteins can serve as protein fusogens. Initial findings with the influenza HA protein indicated that when the transmembrane segment was replaced by a glycosylphosphatidylinositol anchor opening of fusion pores did not occur with membrane merger arrested at the hemifusion stage (54, 55). However, subsequent to those studies, several groups showed that lipid-anchored HA does form small, non-expanding fusion pores (56–58), indicating that transmembrane and cytosolic domains of HA are required for fusion pore expansion. Thus, the present study is an independent confirmation of these more recent studies demonstrating that neither the transmembrane domain nor the cytosolic domain is required for an opening of a fusion pore. This conclusion is also in agreement with reports that transmembrane proteins are not required for fusion in autophagy (59), in nuclear membrane fusion in yeasts (60), in membrane repair (61), and in fusion at the early stages of nuclear envelope assembly in Xenopus oocytes (62, 63). The fusogenic activity of the HA2 ectodomain is also consistent with the hypothesis that both fusion and fission of biological membranes are driven by membrane elastic stresses (64, 65). Note that many of the curvature-generating proteins involved in membrane remodeling lack transmembrane domains (66). The finding that HA2* does not produce fusion pores large enough to allow passage of 10-kDa fluorescent dextran (Stokes radius of 2.4 nm) and does not produce syncytia substantiates the hypothesis that the transmembrane domain (and perhaps its interactions with the fusion peptide) as well as the cytoplasmic domain of HA can be especially important for later fusion stages including fusion pore expansion (26, 57, 67, 68).

Specific mechanisms by which the complete ectodomain HA2 fuses both model and biological membranes as well as the role of the fusogenic properties of the final conformation of HA2 subunit in the context of fusion mediated by full sized HA remain to be established. However, our findings suggest that formation of the hairpin conformation of HA2, often referred to as a “postfusion” conformation, does not signify the end of functionally important HA and membrane restructuring but rather represents the beginning of the key fusion stage. Restructuring of protein fusogens into a hairpin-like final structure under fusion conditions is a strikingly conserved mechanistic motif shared by very diverse viruses (47, 69). Thus, the question whether these hairpin conformations represent discharged postfusion forms of the proteins or their functional fusogenic forms is not limited solely to fusion mediated by HA.