Membrane Fusion Activity of the Influenza Virus Hemagglutinin THE LOW pH-INDUCED CONFORMATIONAL CHANGE*

The hemagglutinin (HA) spike glycoprotein of influ- enza virus catalyzes a low pH-induced membrane fusion event which releases the viral genome into the host cell cytoplasm. To study the fusion mechanism in more detail, we have prepared the ectodornain of HA in water-soluble form by treating virus particles with bromelain. Under mildly acidic conditions (pH 5 5.8), the ectodomain undergoes a conformational change which we found to be biochemically and immunologi-cally equivalent to that in native viral HA. It became sensitive to proteinase K, it exposed new antigenic epitopes in its HA1 chain, and it acquired amphiphilic properties, notably the ability to bind to liposomes. The attachment to liposomes exhibited the same pH dependence and rapid kinetics as the conformational change and was mediated by HA2. The nature of the attach- ment resembled that of an integral membrane protein except that the bound HA was partially removed by base. As observed for virus fusion, attachment is in- dependent of divalent cations and lipid composition. Temperature was found to be a critical parameter only with dimyristoylphosphatidycholine vesicles where attachment was partially blocked below the major phase transition. These and other results obtained indicated that the low pH-induced conformational change in the isolated ectodomain is equivalent to that occurring in intact viral HA, and that its attachment to liposomes can serve as a model for the initial stages in the HA- induced membrane fusion reaction.

The hemagglutinin (HA) spike glycoprotein of influenza virus catalyzes a low pH-induced membrane fusion event which releases the viral genome into the host cell cytoplasm. To study the fusion mechanism in more detail, we have prepared the ectodornain of HA in water-soluble form by treating virus particles with bromelain. Under mildly acidic conditions (pH 5 5.8), the ectodomain undergoes a conformational change which we found to be biochemically and immunologically equivalent to that in native viral HA. It became sensitive to proteinase K, it exposed new antigenic epitopes in its HA1 chain, and it acquired amphiphilic properties, notably the ability to bind to liposomes. The attachment to liposomes exhibited the same pH dependence and rapid kinetics as the conformational change and was mediated by HA2. The nature of the attachment resembled that of an integral membrane protein except that the bound HA was partially removed by base. As observed for virus fusion, attachment is independent of divalent cations and lipid composition. Temperature was found to be a critical parameter only with dimyristoylphosphatidycholine vesicles where attachment was partially blocked below the major phase transition. These and other results obtained indicated that the low pH-induced conformational change in the isolated ectodomain is equivalent to that occurring in intact viral HA, and that its attachment to liposomes can serve as a model for the initial stages in the HAinduced membrane fusion reaction.
Influenza A viruses penetrate their host cells by membrane fusion. After binding to the cell surface, virus particles are internalized and transported to endosomes and lysosomes. The acidic environment in these organelles activates fusion between the viral and host cell membranes (1)(2)(3). The viral factor responsible for this fusion activity is the hemagglutinin (HA') (4-6), a spike glycoprotein that also binds the virus to * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisenent" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ Supported by a National Science Foundation predoctoral fellowship. neuraminic acid containing components on the plasma membrane (7). In this study, we have analyzed the acid-induced changes in the structure and properties of HA and its ectodomain in order to obtain a more detailed understanding of the fusion mechanism.
Structurally, the influenza HA is exceptionally well characterized. Each HA spike is composed of three identical subunits consisting of two disulfide-linked glycopolypeptides, HA1 and HA2 (8). The mature HA is generated in the infected cell by proteolytic cleavage of a precursor called HAO, in which the HA1 constitutes the N-terminal and HA2 the Cterminal portions (9,10). HA2 (Mr = 26,000) traverses the viral membrane once. The large N-terminal domain of HA2 resides outside the viral membrane and possesses a remarkably conserved hydrophobic amino-terminal sequence (11)(12)(13). HA1 (MI = 56,000) is located entirely outside the membrane and is connected to HA2 by a disulfide bond as well as by numerous noncovalent interactions. Exhaustive digestion with bromelain releases the water-soluble ectodomain of HA (called BHA) by detaching it from the hydrophobic membrane anchoring segment (14). The three-dimensional structure of the BHA from the X:31 strain (A/Hong Kong/1968), determined by Wilson et al. ( X ) , shows a roughly rectangular 13.5nm long trimer with globular head domains composed entirely of HA1 resting on top of a fibrous stem consisting chiefly of HA2. The cleavage point between HA1 and HA2 is in the stem region close to the viral membrane (15).
The site in the HA which mediates virus binding to host cell receptors is located in the globular head region of the trimer (16). The hydrophobic amino-terminal peptide of HA2, in the stem region of the molecule, is thought to be involved in HA's membrane fusion activity (17). However, the precise role of this peptide has not been defined and the fusion mechanism is far from understood. A recent study by Skehel et al. (18) demonstrated that a major conformational change takes place in BHA at acid pH. The BHA becomes amphiphilic, two previously hidden tryptic cleavage sites are exposed, and changes are observed in its spectral (18) and antigenic (19)(20)(21) properties. On the basis of these results it has been suggested that acid treatment exposes the hydrophobic N terminus of HA2, thereby imparting amphiphilic properties to the HA ectodomain (18).
Here, using biochemical, immunological, and morphological methods we have further analyzed the changes in BHA. We have determined the kinetics of the conformational change, studied the interaction of acid-treated BHA with liposomes of various compositions, and confirmed the role of HA2 as the principal amphiphilic subunit. Furthermore, we have demonstrated that the changes in BHA reflect the behavior of HA as it resides in the viral membrane. The results are discussed in terms of a model for HA's fusion activity.

Acid
Conversion of BHA and HA-Our experiments were aimed at two aspects of HA's response to acid pH, i.e. the conformational change in the protein and the nature of its subsequent interaction with added membranes and other amphiphiles. The studies on the conformational change were performed using both radioactively labeled and unlabeled BHA and intact X:31 virus in the absence of added membranes or detergents. To obtain a quantitative assay for the conformational change, the susceptibility of Iz5I-BHA to proteases was tested after incubation at different pH values. It was found that acid-treated BHA was susceptible to digestion with proteinase K, a nonspecific bacterial protease (Fig. 1).
Both of the glycopolypeptide chains, HA1 and HA2, were digested ( Fig. 1, inset), and greater than 85% of the radiolabel in '"I-BHA became trichloroacetic acid-soluble. Trichloroacetic acid precipitation and direct quantitation of the HA band after SDS-PAGE showed that the conversion into the proteinase K sensitive form began at pH 5.8, reached halfmaximal values at pH 5.3, and was essentially completed at pH 4.8.
When Iz5I-labeled BHA was treated at various pH and digested with proteinase K, the pH dependence of digestion was nearly indistinguishable from that observed for unlabeled BHA. Furthermore, little or no differences were observed between radioiodinated BHA, unlabeled BHA, and unlabeled viruses. In each case the conversion appeared to be irreversible judging by the fact that incubation at neutral pH for up to an hour after the initial low pH treatment did not affect the fraction of protease-sensitive BHA.
The kinetics of conversion was next determined at two fixed pH values, 5.0 and 5.3. A solution containing BHA was acidified at room temperature and at various times aliquots were removed, neutralized, digested with proteinase K, and precipitated with trichloroacetic acid. At pH 5.0 75-80% of the BHA converted within the first minute, with another 5-10% of the label becoming protease-sensitive over the next hour (Fig. 3). The rate of conversion of the isolated protein was thus quite similar to that previously observed for the kinetics of fusion between influenza viruses and liposomes (35).
Antigenic Change in BHA and HA-The differential sensitivity to proteinase K provided an easy method for discriminating between the neutral and the acid conformations of HA and BHA. Another method used to monitor the conformational change employed monoclonal antibodies. Mice were immunized with acid-treated and neutral BHA, and hybridomas were selected for the production of antibodies capable of immunoprecipitating either the acid or the neutral BHA (or both) under nondenaturing conditions. Whereas most of the monoclonals obtained were able to precipitate both forms of BHA (Table I), one monoclonal, designated H3C10, precipitated only the neutral form. Another, H2D10, specifically precipitated the acid form. To determine whether the latter recognized a determinant on HA1 or HA2, we took advantage * Portions of this paper (including "Materials and Methods," and under nonreducing conditions. Both HA1 and HA2 were digested as shown by the reduced gel in the inset.

TABLE I
Immunoprecipitation of I2'Z-BHA and "'I-X:31 HA 30,000 cpm of IZ5SI-BHA or '*'I-X:31 were immunoprecipitated with the indicated antibodies. Nonimmune rabbit serum used as a control precipitated 600 cpm of both BHA and X:31. This background value was subtracted from those below. Precipitations were carried out as described under "Materials and Methods." of a finding by Graves et al. (33) who reported that reduction of acid-treated Influenza A virus with dithiothreitol dissociates HA1 from HA2. We found that the BHA could also be dissociated in this way, and that immunoprecipitation with H2D10 only brought down the HA1 chain, as determined by SDS-PAGE (Fig. 7, lane 5, Miniprint). Therefore, the epitope recognized by this antibody resided somewhere in the HA1 chain, but was only accessible in the acid-treated form of BHA.

Monoclonal
The antibodies were next tested for their ability to interact with HA molecules of intact X:31 virus before and after the pH 5.0 treatment. The results in Table I indicate that the specificities were the same as observed for BHA. Thus, the antigenic changes occurring in BHA were comparable to those of HA in the virus itself. These results confirmed and extended the findings of Webster et al. (19) and Daniels et al. (20) who have shown that out of four well characterized epitopes on HA1, two are modified or lost after acid conversion, and that at least one new epitope is exposed. Our results indicated that a new epitope was exposed in HA1, and that BHA and HA displayed the same antigenic changes. The

Membrane Fusion
Activity of the Influenza Hemagglutinin 2975 antigenic change, and the identical proteinase K sensitivity of BHA and HA indicated that BHA responded to acid in a manner very similar to intact viral HA. pH Dependence of X:31 Membrane Fusion-Having shown that BHA and the HA in intact virus undergo a similar pHinduced change, we determined the pH dependence of X:31 virus-induced membrane fusion. The fusion activity of the virus was measured by quantitating the extent of cel1:cell fusion induced by added virus particles. After binding viruses at 0 "C, cell monolayers were warmed to 37 "C and acidified for 1 min. The average number of nuclei/cell was determined by microscopy (31). Fusion became detectable at pH 5.2 and was complete by p H 5.0 (Fig. 2). Half-maximal activity occurred between pH 5.0 and 5.1. The pH dependence of fusion was thus much steeper and shifted 0.25 pH units lower than the pH of conversion of the total BHA population to its proteinase K-sensitive form (Fig. 1). Half-maximal fusion was thus recorded at a p H where 80% of the viral HA was converted into the low p H conformation suggesting that the fusion activity may be a cooperative effect involving more than single HA molecules.
Association of BHA with Liposomes-One of the most dramatic consequences of the acid-induced conformational change in BHA is its change from a water-soluble protein to an amphiphilic molecule capable of attaching to liposomes, nonionic detergent micelles, or to itself in the form of oligomeric "protein-micelles'' (18). Since this change in BHA's solubility is likely to reflect the role of HA in fusion, we characterized the interaction of BHA with liposomes in some detail. The pH dependence, kinetics, morphology, and chemical nature of the interaction were determined as well as its dependence on lipid composition, divalent cations, and temperature.
The attachment of BHA to liposomes was assayed as fol- (1:1:1:1.5:0.3) with a trace amount of 32P-labeled phospholipid were mixed with '251-labeled BHA and the pH was adjusted by addition of acid. After 15 min at 37 "C, the mixture was neutralized, made 50% (w/v) with sucrose, placed in the bottom of a centrifuge tube, overlayed with a 0.5-ml sucrose step gradient, and subjected to ultracentrifugation. Owing to the high lipid-to-protein ratio used (>lO,OOO phospholipids/ BHA), the resulting BHA containing liposomes were of low density and floated to the top of the gradient together with BHA-free liposomes. Unbound BHA remained in the bottom fractions (Fig. 8, Miniprint).
The attachment of lZ51-BHA to liposomes was highly efficient and the pH dependence was almost identical to that observed for BHA's acid conversion (Fig.  2). Binding was apparently irreversible because return to neutrality overnight at room temperature did not reduce the amount of BHA attached to the liposomes. The kinetics of binding at various p H values (pH 5.0, 5.3, and 5.6) shown in Fig. 3 was also similar to that already observed for influenza fusion (35) and for BHA's conversion to the proteinase K-sensitive form (indicated for pH 5.0 in Fig. 3 by open circles). When BHA/ liposome mixtures treated at an intermediate pH (pH 5.3) were fractionated by flotation on sucrose gradients and subjected to proteinase K digestion, >90% of the BHA that had associated with the liposomes was digestable, compared to less than 10% of the non-lipid associated BHA (data not shown). Taken together, these results indicate that when converted to the low pH form, BHA rapidly and quantitatively associates with liposomes. For maximal binding, liposomes and BHA had to be present together at the time of acidification. If, in the standard assay (which contained only a trace amount of 'T-BHA) the liposomes were added 1 min after acidification to pH 5.0, 50% of t h e T -B H A bound; if they were added after 5 min, only 25% bound. Thus, the acid form of BHA retained its ability to bind to liposomes for a relatively short time. The rate of inactivation increased when increasing concentrations of unlabeled BHA were present. For example,  Fig. 8 (M). Additional aliquots were immediately digested with proteinase K as in Fig. 1. The fraction digested was determined by trichloroacetic acid precipitation addition of 5 pg of unlabeled BHA reduced the amount of bound BHA from 50 to 28% when liposomes were added 1 min after acidification. The inactivation was probably a consequence of the irreversible aggregation occurring between BHA molecules in the low pH form as reported by Skehel et al. (18). We confirmed that protein aggregation was occurring as judged by a higher sedimentation rate in sucrose velocity gradients and by negative staining, which revealed rosettelike protein complexes similar to those described by Skehel et al. (18).
Effect of Lipid Composition, Temperature, and Membrane Fluidity-The liposomes used up to this point consisted of a mixture of natural phospholipids and cholesterol (PE:PC: SPH:PA:CHOL in a molar ratio of 1:1:1:0.3:1.5). Liposomes with this composition were chosen because they are good target membranes for fusion with influenza virus (35). To determine if any of the lipids were crucial for BHA attachment, liposomes with different compositions were tested. As seen in Table 11, all liposomes tested, irrespective of composition, proved equally efficient as targets. Complete pH dependence curves were determined for liposomes lacking PA and for liposomes consisting of egg PC alone; they were identical to that observed for the standard liposomes (Fig. 2). Interestingly, the inclusion of gangliosides (17%) which can serve as receptors for HA did not influence the extent of binding at acid pH, although a slight increase in binding was observed at neutral pH. These findings are consistent with the observed fusion activity of an influenza A virus (fowl plague virus) which is also largely independent of the phospholipid head groups, the cholesterol content, and the presence of gangliosides (35). One interesting difference between viral-membrane fusion and the binding of BHA to liposomes was observed. Whereas fusion with the influenza virus is inhibited by 50% when PE is omitted (35), attachment of BHA to liposomes was unaffected by the PE content.
To determine whether temperature was an important factor  in the attachment of BHA to liposomes, BHA and liposomes of the standard composition were incubated for 5 min at pH 5.0 at a variety of temperatures. Under these conditions, BHA bound with equal efficiency over the entire 0-37 "C range. However, after only 2 min at pH 5.0, the binding was less efficient at temperatures lower than 37 "C, suggesting that binding occurred more slowly at lower temperatures. Since virtually all (30%) of the BHA is converted to its acid form after 2 min at pH 5.0 (see Fig. 3), this difference most likely reflects a slower rate of attachment of acid BHA to liposomes at low temperatures, as opposed to a slower rate of conversion. This observation is reminiscent of the slower rate of fusion observed between fowl plague virus and liposomes at reduced temperatures (35). The effects of membrane fluidity on BHA binding were determined after incubation at pH 5.0 for 30 min with egg PC or dimyristoylphosphatidylcholine (DMPC) liposomes. While egg PC, like the standard mixture, remains fluid throughout the 0-40 "C temperature range, DMPC bilayers are only fluid above the major transition temperature of 23 "C. Below 23 "C, the lipid molecules of the bilayer are in a crystalline array, with little lateral or rotational mobility (36). The results in Fig. 4 show that BHA binding to egg PC liposomes remained virtually unchanged over the entire temperature range whereas attachment to DMPC liposomes was gradually depressed at temperatures below 37 "C. Below the transition temperature, no further decrease in binding occurred. Thus, although low pH BHA binds to a greater extent to "fluid" bilayers, it can also bind to a significant degree (about 40%) to membranes where the lipids exist in a crystalline array.
Lipid:Protein Ratio-The effect of the lipidprotein ratio on BHA attachment to liposomes was studied by flotation of BHA-liposome complexes in continuous sucrose gradients. Variable amounts of 32P-labeled liposomes with the standard lipid composition were mixed with 5 pg of 'T-BHA, the pH was adjusted to 5.0 for 1 h, and the samples were subjected to flotation at neutral pH. The lipidprotein ratio and density of the proteoliposomes formed was found, not unexpectedly, to be highly dependent upon the amount of lipid added (Fig. 5). As long as the lipid was in considerable excess (>5000 lipid molecules/BHA trimer) nearly all the BHA present became liposome-associated. When less lipid was present, the fraction decreased. At a ratio of 2000 lipid molecules/BHA trimer, 65% of the BHA bound, and at a ratio of 1000, the amount bound was 40%. The lowest lipid-to-protein ratio recorded in   the resulting proteo-liposomes was approximately 700 phospholipids to every BHA trimer. This corresponds to approximately half of the spike density found in intact viral membranes (37-39). Owing to the high bouyant densities of these proteoliposomes, steeper sucrose gradients were required to separate them from unbound BHA. The gradients suggested that the bouyant density of the proteoliposomes became increasingly heterogeneous with decreasing 1ipid:protein ratios. Proteinase K digestion of the unbound BHA indicated that it had been converted to the acid form. Sedimentation and electron microscopic analysis showed that it was aggregated (not shown). Thus, the decreased availability of lipids resulted in the aggregation of BHA in the bulk solution, a situation similar to that observed with DMPC liposomes a t lower temperatures. Therefore, a reduction in the amount of available lipid, or obstruction of insertion into the bilayer, was found to favor aggregation of the BHA molecules. Nature of BHA's Interaction with Liposomes-The low pHinduced attachment of BHA to liposomes was not affected to any large extent by high or low ionic strength (1 or 0.001 M NaCI) or by metal chelators such as EDTA and EGTA (1 or 10 mM). Attempts to elute bound BHA from the liposomes using high concentrations of urea (6 M) or potassium iodide (1 M, a chaotropic agent) also failed, confirming that the BHA was quite firmly attached. The only condition which eluted the BHA (besides treatment with detergents) was elevation of the pH to 10.0-12.5 (which is below the pH required to hydrolyze peptide bonds or lipids (40)). Under these condi-tions, up to 75% of the bound BHA came off. When the pH was subsequently returned to neutrality, the eluted BHA quantitatively rebound to the lipids, and was susceptible to proteinase K digestion. The elution of BHA from liposomes by base proved to be independent of ionic strength.
Thus, in many respects the binding properties of BHA resembled those of integral membrane proteins hydrophobically anchored to lipid bilayers (41). The reversible elution by base suggested, however, that the BHA can undergo a second pH-dependent conformational change at high pH which allows it to detach. Such a phenomenon is not usually observed for integral membrane proteins (41), and it may suggest that the BHA is interacting with the bilayers more superficially.
Binding Occurs through the HA2 Chain-The observation of Graves et al. (33) that reduction of acid-treated influenza virus allows the dissociation of the HA1 and HA2 chains from each other was used to determine which of the subunits of BHA is responsible for the attachment to liposomes. Liposomes with bound '"I-BHA were incubated with dithiothreitol and subjected to separation in a sucrose flotation gradient. The material which remained in the sample zone and the material which bound to the liposomes was analyzed by SDS-PAGE and fluorography. As shown in Fig. 7 (lune 6), HA2 remained bound while HA1 had dissociated. The result was confirmed by detergent binding analysis. Bordier (34) has shown that Triton X-114-solubilized proteins, which have a hydrophobic intramembranous moiety, preferentially partition into the detergent phase formed above the detergent's cloudpoint, whereas soluble proteins and peripheral membrane proteins partition into the aqueous phase. As shown in Table 111, the neutral pH form of BHA behaved as a watersoluble protein. After acid treatment, it partitioned preferentially into the detergent phase confirming its conversion to an amphiphilic form. However, if the acid-treated BHA was reduced prior to partitioning, only a fourth of the lz5I activity was found in the detergent phase. SDS-PAGE of both phases showed that 93% of the HA1 was in the aqueous phase, compared to 33% of the HA2. The HA1 retained its antigenic reactivity with a variety of monoclonal antibodies including the antibody specific to the acid HA (Fig. 7). The remaining 7% of the HA1 and 67% of the HA2 were in the detergent phase. These results strongly suggested that the attachment of BHA to liposomes and detergent is through the HA2 chain.
Lastly, to test the possibility that BHA might have phospholipase activity, BHA was incubated with egg PC vesicles at pH 5.0 for 30 min at a protein:lipid ratio of 1:750. The sample was neutralized and dried under vacuum, and the lipid Partitioning of BHA, HAI, and HA2 in Triton X-114 '251-BHA was incubated at pH 7.0 or 5.0 (15 min), neutralized, and partitioned in Triton X-114 by the method of Bordier (34). Each phase was separated and the amount of '251-BHA in each was counted. In sample 3, the low pH-treated BHA was disassociated into separate HA1 and HA2 subunits by reduction with dithiothreitol (DTT) (33) prior to addition of detergent. After partitioning in Triton X-114, aliquots from each phase of sample 3 were subjected to SDS-PAGE and the radioactive HA1 and HA2 bands were excised and counted for the presence of '251-protein. was separated by thin layer chromatography (42). No decrease in the amount of PC was observed, and no lysophosphatidylcholine was detected. Morphological Analysis-Electron microscopy of Influenza virus after negative staining showed the HA molecules on the virus surface as densely packed rectangular projections perpendicular to the membrane (Fig. 6 A ) . Their apparent length (13 nm) and overall shape was consistent with the x-ray crystallographic structure of BHA (15). After acid treatment and reneutralization, the spikes on the virus surface became disorganized, and the individual HA molecules became difficult to distinguish as spikes (Fig. 6C). Isolated BHA molecules had a compact rectangular shape very similar to that of the HA spikes on the virus (Fig. 6 R ) . When mixed with liposomes a t neutral pH they showed no preferential attachment, but after a brief pH 5.0 incubation they were seen attached to the liposome surface. If the lipid-to-BHA ratio is low (2000:l) the aggregated fraction of BHA was seen in the background (Fig.  611). The proteinaceous layer on the liposomes extended 10-15 nm from the membrane surface, but individual BHA molecules were difficult to resolve because they seemed to have lost their compactness and defined shape. While thin, wiry connections were seen within the structure, the overall impression was that of a molecule which could assume many conformations.

DISCUSSION
Among the viral spike glycoproteins known to possess membrane fusion activity (l), the influenza HA is the best characterized. T o be active in fusion the HA must be in its mature (cleaved) form (6,9, lo), be anchored in one of the membranes to be fused (6), and be exposed to a pH value below a critical threshold which varies from 5.1 to 5.8 depending on the virus strain (43). In this study we have used quantitative assays to characterize the changes in the isolated ectodomain of the HA in response to acid treatment, specifically its conformational change and the concomitant attachment to liposomes and detergent micelles. We have validated that the behavior lJnbound BHA aggregates can he seen in the background. Magnification X 90.000. I of BHA a t low pH is a faithful reflection of the behavior of HA molecules in the viral membrane, characterized the conformational change in some detail, and probed the nature of the interaction between BHA and liposomes.
Our results indicate that the conformational change and acquired hydrophobicity of BHA reflect the response of HA in the viral membrane to acid pH. The changes in the protease susceptibility, morphology, and antigenic structure of BHA and viral HA are indistinguishable. Moreover, the acid-induced interaction of BHA with liposomes mirrors many of the functional characteristics previously observed for fusion between intact viruses and liposomes (35).
The Nature of the Conformationul Change-The differences in structure and properties of neutral and acid-treated BHA and HA have been studied by biochemical, immunochemical. morphological, and spectroscopic methods in several laboratories including our own. The results available so far can be summarized as follows. 1) Acid BHA possesses two tryptic cleavage sites (located in HA1, positions 27 and 224) which are not accessible in neutral BHA (18). 2) Both the HA1 and HA2 subunits of BHA and HA become fully susceptible to proteinase K upon acidification (Fig. 1). 3 ) Two antigenic epitopes in HA1 (called B and D, Ref. 44) are lost or modified after acid treatment. The acid form of the protein has at least one unique epitope (19-21) which we have located at the HA1 chain (Fig. 7). 4) The interchain disulfide bond located in the stem of the molecule close to the viral membrane is accessible to reduction only after conversion to the acid form (33). We found that the reduction resulted in the dissociation of HA1 and HA2 in both BHA and intact virus (Table 111) and that the HA2 was apparently responsible for the lipid-binding character of acid BHA. 5) Whereas the neutral HA and BHA have a compact rectangular morphology when viewed by electron microscopy, the acid forms have lost their compactness and defined shape (18) (Fig. 6). 6) While neutral BHA behaves like a soluble glycoprotein, the acid form is amphiphilic (18). 7) Dichroic spectra of neutral and acid BHA suggest differences in conformation (18). In addition we have recently found that polyclonal antipeptide antibodies against the C-terminal peptide of HA1 and the N-terminal peptide of HA2 bind specifically to the acid-treated form of BHA, suggesting a change in exposure and/or conformation of these moieties.' When projected against the known x-ray structure of neutral BHA (15). these observations indicate that the HA trimer is affected throughout its length, from the top of the HA1 domains to the very base where the interchain disulfide is located. At the same time, the available evidence suggests that the acid conversion does not lead to major denaturation; the sialic acid binding property of HA1 remains intact (18). most of the antigenic determinants on HA1 are retained, and most of the potential trypsin cleavage sites remain inaccessible (18). Our working hypothesis, based on these findings and the electron microscopic images, postulates that the trimer dissociates partially or completely along the trimer interface, thus exposing the surfaces between the three subunits. While retaining their overall tertiary structure, the top domains (HA1) may detach from each other and move away from the central axis of the molecule. Nevertheless they may remain flexibly attached to HA2 via the interchain disulfide bond. The stem and the hydrophobic N-terminal of the HA2 chains would thus be exposed and made accessible for attachment to a target membrane. In evaluating this and other models in terms of fusion activity one must, however, keep in mind that .J. White  the "acid BHA" analyzed by us and others has usually been reneutralized after acid treatment. Its structure may therefore differ in important respects from that of nascent fusion active molecules present shortly after acidification. BHA's Interaction with Liposomes-The exposure of a hydrophobic moiety in HA is probably a key element in the fusion reaction. We have previously suggested that attachment of the virus to the target membrane prior to fusion may depend on it, and that it may, in fact, help bring the two membranes close enough to allow direct interactions between the lipid bilayers (6). The nature of the HA-dependent attachment function has proved difficult to study with whole virus particles owing to their nonspecific binding to liposomes at neutral pH (45): The isolated BHA, however, has no tendency to associate with liposomes at neutral pH, though once converted to the acid form it attaches rapidly and irreversibly.
Our results indicate that the attachment of BHA to liposomes occurs via the HA2 glycopolypeptide chain and that the interaction is primarily hydrophobic in nature. It does not require the presence of charged phospholipids; it is not inhibited by high or low ionic strength, by metal chelators, or by low temperature; nor i s it reversed by chaotropic agents or protein denaturants, such as urea. It is, however, partially affected by the fluidity of the bilayers, as shown by the temperature dependence of binding to DMPC liposomes. The only general characteristic which differentiates the liposomebound BHA from most amphiphilic integral membrane proteins is the observed elution at elevated pH. It is likely that the hydrophobic determinant in HA2 corresponds to the Nterminal peptide because, apart from the membrane anchoring peptide, it is the only distinctly hydrophobic sequence in HA2 (23). In addition, it is highly conserved among HA molecules of different virus strains, and recent studies using site specific mutagenesis have shown that mutations in this region affect the fusion a~t i v i t y .~ T h e fact that BHA can be eluted off the liposomes by base suggests that it is less firmly attached than most integral membrane proteins. The relative shortness of the hydrophobic N-terminal peptide (10 residues) could mean that the protein only inserts into the outer leaflet of the bilayer, thus explaining, in part, its ability to be eluted at elevated pH.
H A S Role in Membrane Fusion-BHA binds avidly to membranes but it does not, itself, promote membrane fusion (6). Although informative in illuminating how acid BHA interacts with liposomes, our results do not explain how the membrane-bound, intact HA is able to elevate the otherwise extremely low fusion incidence between membranes. One major obstacle against fusion of membranes is the "hydration force," a strong repulsive force opposing close approach of two hydrated polar surfaces (46). Studies by Rand and coworkers have shown that the hydration force becomes an important factor when the distance between the bilayers is 2 nm or less (for review see Ref. 47). We have suggested (1) that one of the functions of HA (and other fusion proteins) may be to help overcome this energy barrier. The HA molecules, which traverse the viral membrane, may do this by integrating firmly into the target membrane via the hydrophobic N terminus of HA2. They may, in this way, mechanically bring the two membranes so close that direct contact between the lipid molecules becomes possible. The fact that PE enhances the fusion activity of both Influenza and Sindbis viruses (and to a lesser extent Semliki 48) is consistent with this notion. Being less hydrated than most other phospholipids, PE-containing membranes are known to exhibit significantly lower hydration forces (47). The demonstration that BHA-binding to liposomes does not require PE suggests that it is not the lipid-protein interaction per se, but rather the interaction of two membranes, which is PE-dependent.
The HA may also be instrumental in the second and third stages of fusion: the coalescence and subsequent separation of the membranes (49). For fusion to occur the lipids must be locally disorganized; this may also require the presence of HA. Our data suggest that HA penetrates into the hydrophobic interior of the target bilayer. Whether this can explain the drastic change in lipid organization remains to be seen. PE may also have a role at this stage of fusion since it is more likely to adopt nonlamellar structures within membranes than other phospholipids (50). The fact that BHA can be incorporated at high concentrations into liposomes will allow studies on the BHA-lipid interactions by biophysical techniques. Such studies will be performed to determine whether BHA does, indeed, modify the organization of the membrane lipids and induce the non-bilayer configuration(s) needed for fusion. the Influenza Hemagglutinin