The HA2 Subunit of Influenza Hemagglutinin Inserts into the Target Membrane Prior to Fusion*

The interaction between influenza virus and target membrane lipids during membrane fusion was studied with hydrophobic photoactivatable probes. Two probes, the newly synthesized bisphospholipid diphosphatidylethanolamine trifluoromethyl [3H]phenyl dia- zirine and the phospholipid analogue l-palmitoyl-2(11-(4-[3-(trifluoromethyl)diazirinyl]pheny1}-[2-3H]- undecanoyl)-sn-glycero-3-phosphocholine (Harter, C., Bachi, T., Semenza, G., and Brunner, J. (1988) Biochemistry 27, 1856-1864), were used. Both labeled the HA2 subunit of the virus at low pH. By measuring virus-liposome interactions at 0 “C, it could be demonstrated that HA2 was inserted into the target mem- brane prior to fusion. As we have recently demon- strated, at this temperature, exposure of the fusion peptide of HA2 takes place within 15 s after acidification, but fusion does not start for 4 min (Stegmann, T., EMBO 4231-4241). HA2 was labeled at least 2 min before fusion. No labeling of the HA1 subunit was seen. These data indicate that fusion is triggered by a direct

. Of these, the hemagglutinin of influenza virus (HA)' is the best-characterized. It is a homotrimeric integral membrane protein. Each monomer consists of two disulfide-bonded polypeptides, HA1 and HA2. As we have have recently shown (Stegmann et al., 1990), the low pHinduced fusion is mediated by HA and is triggered by a transient intermediate conformation of the protein. In this intermediate, the hydrophobic N termini of the HA2 polypeptides, the so-called "fusion peptides," are exposed, while the trimers remain largely intact. In the x-ray structure of the neutral p H form, these peptides are buried in the interface between the monomers of the trimer (Wilson et al., 1981).
While the models proposed for HA-mediated membrane fusion differ in several respects, it is generally assumed that a direct interaction between HA and the lipids of the target membrane is involved. The most convincing evidence for such an interaction was provided by Brunner and co-workers (Harter et al., 1988(Harter et al., , 1989. They incorporated a hydrophobic photoactivatable reagent, PTPC (a diazirine coupled to the acyl chain of a synthetic phospholipid), in the bilayer of liposomes. Since photolysis of the probe results in covalent modification of whatever is present in the membrane, they could demonstrate that the fusion peptides of bromelainsolubilized ectodomains of HA (BHA) insert into the outer leaflet of the liposomal bilayers. Their results suggested, moreover, that the peptides adopted an @-helical configuration (Brunner, 1989). Recently, these results were extended to intact influenza virus fusing with liposomes containing PTPC (Brunner et al., 1991).
We have further investigated the role of HA2 in triggering fusion. As was recently shown, studying fusion a t reduced temperatures allows analysis of the component steps in the fusion reaction (Stegmann et al., 1990). In this paper we demonstrate that, at pH 5.1 and 0 "C, HA2 is labeled both by PTPC and by the newly synthesized hydrophobic photoactivatable probe DIPETPD present in liposomal target membrane bilayers at least 2 min before fusion. The implications of these findings for the membrane fusion mechanism are discussed. TNBS, 2,4,6-trinitrobenzenesulfonic acid; MES, 4morpholineethanesulfonic acid; SDS-PAGE, sodium dodecyl sulfatepolyacrylamide gel electrophoresis.

MATERIALS AND METHODS
Photoactiuatable Reagents-The synthesis and properties of 'Hlabeled DIPETPD will be described in detail elsewhere.* Synthesis of a n analogue of DIPETPD lacking the photoactivatable group was reported by Delfino et al. (1987). The structure of DIPETPD is shown in Fig. 1. Specific radioactivity of the DIPETPD preparation used was 3.5 Ci/mmol.["H]PTPC (22.2 Ci/mmol) synthesized according to Harter et al. (1988) was the generous gift of Dr. J. Brunner. Both reagents were stored in the dark in toluene/ethanol 1:l a t -20 "C.
Virus, BHA, and Reconstituted Viral Membranes-The X-31 recombinant strain of influenza virus was propagated from plaquepurified virus (plaque C-22; Doms et al., 1986) in the allantoic cavity of embryonated eggs, purified, handled, and stored as described before (Stegmann et al., 1985). To measure the concentration of virus, viral phospholipid was extracted according to Folch et al. (1957) after which phosphate was determined according to Bottcher et al. (1961).
BHA was produced by digestion of virus (1 mg/ml) with bromelain (10 mg/ml) in the presence of 6-mercaptoethanol (50 mM) for 24 h at 37 "C (Brand and Skehel, 1972). The BHA was then purified by affinity chromatography on Sepharose-immobilized R. communis agglutinin.
Reconstituted viral membranes were produced essentially as in Stegmann et al. (1987), except that photoactivatable probes, dissolved in C12E8, were addeddirectly to the solubilized membranes in C12E8. The procedure was carried out under conditions of low illumination.
Liposomes-Large unilamellar liposomes were prepared by repeated low pressure extrusion of multilamellar liposomes (Mayer et al., 1986) through defined pore polycarbonate filters of decreasing pore size (0.4, 0.2, and 0.1 pm in diameter). An extruder with a 10-ml barrel from Lipex Biomembranes (Vancouver, Canada) was used. Multilamellar liposomes were produced by resuspending thin films of lipids, prepared by exhaustive evaporation of solvent (toluene/ethanol 1:l) in uacuo, into aqueous buffers. The suspensions were then frozen and thawed three to five times before extrusion. After extrusion, residual multilamellar liposomes were removed by pelleting at 16,000 X g for 20 min. For liposomes containing DIPETPD, the temperature during resuspension of the lipids and the subsequent extrusion was kept a t 60 "C. Liposomes were used within 2 days of their production. DIPETPD-containing liposomes were stored a t room temperature; PTPC-containing liposomes a t 4 "C. Phospholipid phosphate was determined according to Bottcher et al. (1961).
Fusion Experiments-To measure fusion by a resonance energy transfer based assay, liposomes were produced containing 0.6 mol% each of N-NBD-PE and N-Rh-PE (Struck et al., 1981). Measurements were performed under continuous stirring, in a thermostatted cuvette holder containing 2 ml of 135 mM NaCl, 15 mM sodium citrate, 10 mM MES, 5 mM HEPES set to various pH values by HC1 or NaOH. The increase in N-NBD-PE fluorescence emission, resulting from dilution of the fluorophores into the viral membrane upon fusion, was recorded continuously a t excitation and emission wavelengths of 465 and 530 nm, respectively. A 515-nm long-pass filter was placed between cuvette and emission monochromator (Stegmann et al., 1985). An SLM-8000 fluorometer was used for all measurements. The fluorescence scale was calibrated by setting the initial residual fluorescence of the liposomes to zero and the fluorescence a t infinite probe dilution to 100%. The latter value was obtained after addition of TX-100 (0.5% v/v) to the liposomes and correction of the fluorescence intensity for sample dilution and for the effect of TX-100 on the quantum yield of N-NBD-PE (Struck et al., 1981). For fusion drawings, curves were manually traced from the original stripchart recordings using a digitizer pad (Kurta, Phoenix, AR) and a drawing program (Canvas, Deneba Software, Miami, FL).
Photolysis-A cylindrical 450-watt Hanovia medium pressure mercury lamp was used for photolysis. The lamp was cooled by a circu-' J. M. Delfino and F. M. Richards, manuscript in preparation. lating cold water jacket. The light was filtered through a 1.5-cm thick saturated copper sulfate sleeve which cuts off all the UV emission below 320 nm. A shutter present between the light source and the sample holder was used to control the exposure time. Samples were placed a t 10 cm from the lamp in either 1-cm pathlength cuvettes fitting in a sample holder that was thermostatted by means of a circulating water bath or placed into larger thin (<1 cm) quartz vessels thermostatted by a heat gun. Just before photolysis, all buffers used for the samples were degassed and flushed with argon.
Gels were stained with Coomassie, destained, transferred to glacial acetic acid for 30 min, then immersed in 20% 2,5-diphenyloxazole in glacial acetic acid for 1 h, quenched with water, dried, and put down a t -70 "C with preflashed Kodak XAR-5 film for autoradiography. Bands were quantitated by densitometry with a digital gel scanner (Visage 2000). Care was taken to insure that only exposures with densities within the linear range of the film and the instrument were used for quantitation. Longer exposures were used for the figures in this paper.

Properties of Liposomes Containing PTPC or DIPETPD-
Large unilamellar liposomes were used as target membranes for fusion. They were composed of natural zwitterionic phospholipids (PC and PE) with added gangliosides. The gangliosides served as receptors for the viral HA, allowing virusliposome complexes to form at neutral pH (Stegmann et al., 1986(Stegmann et al., , 1989b. To produce liposomes containing PTPC or DIPETPD, trace amounts of these probes were added from stock solutions in toluene/ethanol (1:l) to PC/PE/ganglioside mixtures in the same solvent. After evaporation of solvents, multilamellar liposomes were produced from the dry lipid films by resuspension in an aqueous buffer. They were made unilamellar by repeated extrusion through polycarbonate filters (see "Materials and Methods"). The whole procedure was performed under conditions of low illumination to avoid photolysis of the probes. For DIPETPD-containing liposomes the temperature during resuspension of the lipids and extrusion was maintained a t 60 "C. Negative stain electron microscopy (not shown) confirmed that unilamellar liposomes with a diameter of about 0.1 Fm were formed. We found that up to 20% by weight of a nonradioactive analogue of DIPETPD could be incorporated into liposomal bilayers without affecting stability or tightness.3 As a bisphospholipid, DIPETPD ( Fig. 1) could potentially span the membrane of the liposomes from the outer leaflet of the bilayer to the inner leaflet and thus confine the photoactivatable group to the center of the bilayer. Alternatively, it could be present in a U-shaped configuration in either the outer or the inner leaflet of the bilayer with both headgroups present on one side of the bilayer. To determine the fraction of DIPETPD in these different configurations the membrane impermeant reactant TNBS was used, which reacts with the ethanolamine headgroup of the probe. When present in a liposome, transmembrane conformers will be labeled only on one end, the external-facing U-shaped molecules will be labeled on both ends, and the interior-facing ones will not be labeled. Liposomes were reacted with TNBS at pH 8.0 and the formation of a reaction product was determined spectrophotometrically. After a level was reached, the remaining unreacted TNBS was quenched with lysine, and lipids were extracted and separated by thin-layer chromatography. It was found that 50% of the probe was in a membrane-spanning configuration, and that 25% resided in each of the monolayers of the bilayer in a U-shape. The method and the results will be described in detail elsewhere?
Next, we investigated the ability of DIPETPD-containing liposomes to serve as targets for fusion with influenza virus. The fusion assay was based on resonance energy transfer between two fluorescent phospholipid analogues (N-NBD-PE and N-Rh-PE) when they are both present in the target liposomes (Struck et al., 1981). Fusion of a liposomal with a viral membrane resulted in dilution of the fluorescent lipids. This led to decreased energy transfer and an increase in energy donor (N-NBD-PE) fluorescence emission (Stegmann et al., 1986). Trace amounts of the fluorescent probes (0.6 mol% each) and DIPETPD were incorporated into liposomes. Virus was injected into a thermostatted fluorometer cuvette containing the liposomes in buffers at different pH and the change in fluorescence recorded.
A low-pH-dependent increase in fluorescence was observed both at 37 and at 0 "C ( Fig. 2, A and B ) . To confirm that this was caused by fusion and not by other modes of lipid transfer between membranes, virus alone was incubated at pH 5.1, 37 "C and then added to liposomes at pH 5.1. We have previously shown that after such treatment the virus will still bind to the ganglioside-containing liposomes but is unable to fuse (Stegmann et al., 1986). The lack of fluorescence increase In A: a, pH 5.1, 37 "C; b, pH 5.1, 37 "C after a preincubation of the virus alone at pH 5, 37 "C for 2 min; c, pH 7.4, 37 "C.
In R: pH 5.1, 0 "C. Liposome to virus ratio was 1:1.5 (12.5 pM of membrane phospholipid phosphorous products total). Liposomes consisted of egg PC:egg PE:bovine brain gang1iosides:DIPETPD (12:6:2:1), N-NBD-PE, and N-Rh-PE (0.6 mol% each). Fluorescence was measured as described under "Materials and Methods." in this control and the lack of an increase at neutral pH ( Fig.  2 A ) indicated that the increase in fluorescence measured at pH 5.1 represents bona fide membrane fusion. At 37 "C, fusion started immediately after addition of virus to liposomes, leveling off in about 2 min at a fluorescence increase of 20%. At 0 "C, fusion was preceded by a lag phase of about 4 min, after which the fluorescence increased slowly, reaching a level of about 25% after 2 h. The results showed that the characteristics of fusion of influenza virus with liposomes containing DIPETPD at either temperature were essentially identical to those with liposomes lacking the probes (Stegmann et al., 1990).
The HA2 Subunit of Intact Virus Is Labeled by Both Probes-Next, we investigated whether the two probes could be used to detect fusion of influenza virus strain X-31 with liposomes. Virus was incubated with liposomes at pH 5.1, 37 "C. Immediately after lowering the pH, samples were photolyzed. Fusion and photolysis were allowed to continue for 15 min, after which the samples were neutralized. In control experiments, samples were either kept at neutral pH during photolysis or not photolyzed during fusion at low pH. To determine whether covalent association of DIPETPD and PTPC with protein had taken place, samples were extensively delipidized by repeated chloroform/methanol precipitation of the protein before electrophoresis.
As shown in Fig. 3, HA2 was labeled by both probes at low pH. Since HA2 runs at the same position as does the viral matrix (M) protein, nonreducing gels were run to substantiate the identification of the labeled protein as HA2. No labeling was found in M (not shown). A second band, migrating just below the viral nucleoprotein, was labeled to some extent (1-4% of that found in HA2). Although invisible in the Coomassie-stained gel, this band could perhaps represent the viral neuraminidase, whose membrane anchor is present in the membrane of the fusion product. For either probe, the total amount of protein-associated label was around 0.03% of the input radioactivity (as determined by immunoprecipitation of the viral proteins with a polyclonal antibody against influenza virus). With PTPC, no labeling was seen at neutral pH, or in Autoradiograph of products separated by SDS-PAGE after incubation of liposomes containing PTPC (1 mol%) or DIPETPD (as in the legend to Fig. 2) with virus at 37 "C. Liposomes and virus were incubated in a small volume at neutral pH, 0 "C for 15 min. Aliquots of the mixture were transferred to a large volume of buffer (final concentrations 100 p M of liposomal and 150 pM of viral membrane phospholipid phosphorous, respectively) at the pH indicated (5.1 or 7.4) at 37 "C and immediately photolyzed for 15 min. Control samples were acidified but not photolyzed (-ph). Samples were then neutralized by the addition of 1 M HEPES, pH 7.6, and briefly stored on ice. The proteins were then repeatedly precipitated with methanol/chloroform (2:l) at 37 "C and prepared for electrophoresis. The noncovalently bound lipid ( L ) ran slightly behind the dye front.

HA2 Inserts before
Fusion 18407 nonphotolyzed controls after fusion. In similar controls with DIPETPD, a small amount (about 3% of the protein-associated radioactivity found at low pH), was found associated with HA2. These data indicate that both probes were able to label HA2 during or after fusion at 37 "C.

The Membrane Anchor of H A 2 Can be Labeled by PTPC
and DZPETPD-The radioactivity found associated with HA2 at low pH could arise from labeling of the fusion peptide, the membrane anchor of HA2, or both. To investigate whether the probes used in the current study could label the membrane anchor of HA2, influenza virus membranes were solubilized with the detergent C12E8 and added to PTPC or DIPETPD in an aqueous buffer containing C12E8, after which the membranes were reconstituted according to Stegmann et al. (1987). After photolysis, 90-98% of the label was found associated with HA2 (Fig. 4), and the rest in HA1. No radioactivity was detected in the region of the gel where the neuraminidase would be expected, but it is not known whether neuraminidase is present in reconstitutes made by this method. These data indicate that both probes can label the membrane anchor of HA2 when present in the same bilayer.

H A 2 Is Labeled before Fusion Takes Place-To determine
whether HA2 inserts into the target membrane during the lag phase that precedes fusion, liposomes were incubated with virus at pH 5.1, 0 "C. It has been shown that the fusion peptide is exposed within 15 s, and that its exposure coincides with hydrophobic association of the virus with liposomes. However, fusion does not take place until 4-8 min later (Stegmann et al., 1990, see also Fig. 2B). The lag time was measured to be 4.1 k 0.2 min (n = 5) for the virus preparation used in these studies. We photolyzed several different samples during 2-min periods at pH 5.1, 0 "C, both during the lag phase and during fusion. In our photolysis set-up this resulted in only half-maximal photolysis. Immediately after photolysis, samples were neutralized. As shown previously, samples neutralized during the lag phase will not go on to fuse at neutral pH. In control experiments, samples were either not photolyzed after fusion for 1 h, or they were kept at neutral pH for 1 h and then photolyzed for 2 min. The results obtained with probe PTPC are shown in Fig.   5A. HA2 was already labeled during the first 2 min after acidification, i.e., well before fusion began. Photolysis during the last 2 min of the lag phase resulted in labeling of an approximately equal amount of HA2 as during the first 2 min. After the onset of fusion, labeling increased with time. The HA2 band appeared as a doublet, probably because the Coomassie-stained M protein, which runs as a very narrow band at about the same position as HA2, acts as a filter during autoradiography for these very low quantities of radioactivity. HA did not appear as a doublet in nonreducing gels, and no labeling of M was found on those gels (not shown). Traces of radioactivity were found associated with HA2 at pH 7.4. Extensive delipidization by repeated chloroform/methanol precipitation of the samples did not remove this radioactivity (not shown). No labeling was seen in samples that were fused but not photolysed. Other faintly visible labeled bands were similar to those seen after fusion at 37 "C (Fig. 3). The total protein-associated radioactivity found after 1 h of fusion was about 0.01% of the added radioactivity. With DIPETPD, some labeling of HA2 was seen even at pH 7.4 or in the nonphotolyzed control (Fig. 5 B ) . Extensive delipidization by repeated chloroform/methanol precipitation of the protein removed about 95% of the input radioactivity, but did not lower the labeling in these controls appreciably (not shown). The radioactivity found associated with HA2 HA HA1 NP  Liposomes and virus were incubated in a small volume at neutral pH, 0 "C for 15 min. Aliquots of the mixture were then transferred to a large volume of buffer at 0 "C and photolyzed. Different samples were either photolyzed for 2-min intervals, 0-2, 2-4, 4-6, 15-17, or 60-62 min after lowering the pH, or kept at at low pH for 1 h without photolysis (-ph), or photolyzed for 2 min following a 1-h incubation at pH 7.4. After photolysis, samples were immediately neutralized, precipitated by trichloroacetic acid, washed twice with acetone, and analyzed by SDS-PAGE.
(which again appeared as a doublet) during the lag phase, e.g. that found 0-2 min after lowering the pH exceeded the level of radioactivity in the controls. After the onset of fusion, HA2-associated radioactivity increased. Other faintly visible bands were similar to those seen at 37 "C. The total proteinassociated radioactivity found after 1 h of fusion was about 0.01% of the added radioactivity.
Quantitation of the gels shown in Fig. 5, A and B , obtained by densitometry, is presented in Table I. The amount of labeling found at pH 7.4 was subtracted from that found in the other lanes. Expressed as a percentage of the labeling obtained after 1 h, total labeling during the lag phase amounted to 30% with PTPC and 17% with DIPETPD. Importantly, only a slight increase in labeling was found during the lag phase.
Taken together, these results indicate that HA2 entered the hydrophobic interior of the target membrane before fusion took place. No significant additional insertion occurred during the later part of the lag phase. HA1 did not come into contact with the hydrophobic interior of the bilayer during fusion.

BHA2 Is Labeled by Interaction with PTPC-but Not DI-PETPD-containing Liposomes at Low pH-To determine
whether the newly synthesized probe DIPETPD incorporated into liposomal membranes was able to label the fusion peptide of HA at low pH as does PTPC, the experiments of Brunner and coworkers (Harter et al., 1988) with the bromelain-solubilized ectodomain of HA (BHA) were repeated with DI-PETPD. PTPC-containing liposomes were used as a control. BHA was incubated with liposomes at pH 5.1, 37 "C. Immediately after lowering the pH, samples were photolyzed for 15 min and then neutralized and prepared for SDS-PAGE. Control experiments were carried out at neutral pH. The results are shown in Fig. 6. As expected, BHA2 was labeled by PTPC at low pH. Radioactivity was also found associated with a band migrating at the position of HA1 at low pH (approximately 30% of BHA2 labeling by densitometry). It has been reported by Harter et al. (1989) that at least part of this band is not HA1, but a dimer of BHA2 which comigrates with HA1 under these conditions. Only a small amount of labeling was seen at neutral pH (Fig. 6) and no labeling was found in nonphotolyzed controls (not shown). By comparison, Harter et al. (1988), working with the BHA of strain A/PR/8/34 found less labeling at neutral pH but a somewhat higher incorporation of label into BHA2 relative to HA1 at low pH. Strain differences or a difference in protein:lipid ratio (Harter et al., 1988) might account for these discrepancies.
With DIPETPD, very little labeling was observed at low or neutral pH (Fig. 6). Our interpretation of these results is as

TABLE I Quantitation of the label associated with HA2
before and during fusion at 0 "C Densitometric interpretation of the autoradiographs shown in Fig.  6, A and R. The optical density of the bands corresponding to HA2 was determined and integrated over the surface area of the band. The integrated intensity of the band formed after neutral p H incubation was subtracted. Intensities are expressed relative to the labeling obtained during the 2-min interval following 1 h of incubation a t p H 5.1, 0 "C. were incubated with BHA (4 pg/ml) in a small volume a t neutral pH, 0 "C for 15 min. Aliquots of the mixture were transferred to a large volume of buffer a t 37 "C and, at the pH indicated, immediately photolyzed for 15 min. Samples were then neutralized, precipitated by the addition of trichloroacetic acid (IO%), washed twice with acetone, dissolved in sample buffer, and subjected to SDS-PAGE using a Tris/tricine buffer system (10% T/3% C gels) (Schagger and von Jaggow, 1987). The noncovalently bound lipid ( L ) ran behind the dye front (RF approximately 0.8).

DIPETPD PTPC min
follows. The results of Brunner (1989) indicate that the fusion peptides of BHA are inserted as shallow a-helices into the outer leaflet of the liposomal bilayer. They are accessible to the photoactivatable group of PTPC, which probably requires transient movement of the diazirine toward the headgroup or transient dipping of the BHA helix further into the bilayer. The lack of reactivity with DIPETPD may suggest that the photoactivatable group of this probe, whether present in a transmembrane or a U-shaped configuration, is deeper within the bilayer and therefore not within reach of the inserted fusion peptide of BHA2. Even in the U-shaped configuration, the diazirine on DIPETPD is likely to be much more restricted near the middle of the bilayer than that on PTPC, because it is covalently attached to two fatty acid chains (cf. Fig. 1). Therefore, in all likelihood, DIPETPD does not label the fusion peptide of BHA because of its more defined location as compared to PTPC.

DISCUSSION
HA2 Is Inserted into the Target Membrane Bilayer before Fusion-In this study, it was demonstrated with two different photoactivatable probes that the HA2 subunit of influenza virus, strain X-31, inserts into a liposomal bilayer prior to fusion of the virus with liposomal membranes. We were able to measure this because at pH 5.1,O "C fusion is preceded by a 4-min lag phase (cf. Fig. 2B; Stegmann et al., 1990). As reported previously, the fusion peptide is already exposed within 15 s after acidification at this temperature and pH, and concomitantly virus interacts hydrophobically with target membranes. Harter et al. (1988;1989) have demonstrated that, with the bromelain-solubilized ectodomain of HA (BHA), PTPC labels only the fusion peptide of the BHA2 subunit at low pH. In the intact virus at 0 "C, the fusion peptide of HA2 is the only hydrophobic moiety known to be exposed early in the lag phase by the limited conformational change taking place at low pH (Stegmann et al., 1990). Therefore, it is most likely that the observed labeling of HA2 during the lag phase (Fig. 5, A and B ) is in the fusion peptide of HA2. A more direct way of identifying the fusion peptide would have been to sequence peptides of HA2, but we were unable to do so because of the low level of radioactive labeling obtained.
BHA was found to interact somewhat differently with target membranes than the the HA of intact virus. While labeled by PTPC as reported by Harter et al. (1988) it was not labeled by DIPETPD. The fusion peptide of BHA2 probably inserts into the bilayer as a shallow a-helix parallel to the plane of the membrane (Brunner, 1989). Such an orientation could put it out of reach of the photoactivatable group of DIPETPD which may be more deeply buried in the bilayer than that of PTPC. In contrast, the fusion peptides of viral HA could be inserted in a more perpendicular orientation, explaining their labeling with DIPETPD. Other differences between BHA and the viral HA were also observed. HA1 from BHA but not viral HA1 was labeled at low pH by PTPC. This probably indicates that hydrophobic domains not normally exposed on HA1 are accessible after removal of the membrane anchor and exposure of the fusion peptide. This may in part be due to the aberrant trimer dissociation which is observed with BHA but not with intact HA (Doms and Helenius, 1986). Moreover, both subunits of BHA were also labeled by PTPC to a significant extent at neutral pH, indicating that some hydrophobic moieties are exposed even before the pH is lowered. In a recent study, Brunner et al. (1991), working with intact virus, strain A/PR/8/34, found weak labeling of HA2 after incubation at pH 5, 0 "C with PTPC-containing liposomes. Although the virus bound to zwitterionic liposomes, fusion was not detected at this temperature. Unfortunately, little is known about the kinetics of fusion and conformational change for this strain. It is possible that strain differences resulted in the different threshold temperature for fusion. Importantly, significant labeling was found during the first part (10-15 s) of a biphasic labeling pattern, probably indicative of a lag phase, at 23 "C (Brunner et al., 1991).
Properties of DIPETPD-The results presented in this study demonstrate the usefulness of the newly synthesized hydrophobic photoactivatable reagent DIPETPD (Fig. 1). The molecule was designed to provide enhanced geometric resolution compared to existing hydrophobic photoactivatable reagents such as the phospholipid-like PTPC. In our hands, 50% of DIPETPD was present in a membrane-spanning configuration, which would most likely restrict the photoactivatable diazirine of DIPETPD to the middle of the membrane bilayer.* However, this remains to be demonstrated directly. Despite their differences in structure, PTPC and DIPETPD labeled HA with similar efficiency. Implicit in our interpretation of the results is the notion that the labeling seen with DIPETPD at 0 "C in the absence of photolysis or at neutral pH, can be subtracted from the labeling obtained at low pH. Labeling at neutral pH could be due to a population of fusion peptides that is already exposed at neutral pH. However, very little labeling was seen at 37 "C or with PTPC, indicating that such a population is probably small. Labeling without photolysis (Fig. 5B) can only occur if the association between protein and lipid is noncovalent or if a side reaction takes place. As the association withstood repeated washing with chloroform/methanol or acetone, boiling in the presence of SDS and separation on a gel it is probably covalent. Therefore, it was assumed that the labeling was due to a nonspecific side reaction. Considering the small percentage of specific labeling obtained with these probes, side reactions involving as little as 0.001% of label are readily problematic. We do not know why it is so much more extensive at 0 "C than at 37 "C. The amount depended somewhat on the age of the preparation and on the solvent in which DIPETPD was stored.
The Mechanism of Membrane Fusion-Several models for the role of HA in fusion have been presented. In one model the fusion peptide of HA2 is not inserted into either the viral or the target membrane, but it serves as a hydrophobic surface on the HA molecules which enables the flow of lipids during fusion (White, 1990). Our observation that HA2 was labeled by both probes before any lipid mixing takes place would appear to contradict this. In a second model, the fusion peptide of HA2 is thought to interact with the viral membrane,

HA2 Inserts before Fusion
or with other fusion peptides (Ruigrok et al., 1988). While we cannot exclude that this occurs for some of the fusion peptides, we found clear evidence for an interaction between HA2 and the target membrane.
We have recently proposed a model for membrane fusion in which there is a limited conformational change in HA at low pH (Stegmann et al., 1990). The model is summarized in Fig. 7. After the conformational change (Fig. 7 B ) , the tops of the trimers remain intact, while exposed fusion peptides insert into the target membrane (Fig. 7C). Considering the distance between the target membrane and the fusion peptide, it was necessary to assume that the proteins are able to bend sideways in order for the peptide insertion to take place, as schematically indicated in Fig. 7C. The present results are consistent with this hypothesis. They indicate that HA2 is an integral membrane protein in both the viral membrane and the target membrane just before fusion.
We hypothesized further that the lag phase which precedes fusion at 0 "C could represent the time needed for multiple spikes to form a fusion complex, for the lipids to be destabilized, or for further insertion of fusion peptides to reach a critical concentration that would trigger fusion (Stegmann et al., 1990). We found that the amount of labeled HA2 did not increase significantly during the lag phase (Table I). This indicates that little additional insertion of HA2 occurred during the later part of the lag phase. Thus, the insertion of HA2 molecules into the target membrane occurs almost immediately upon acidification. The lag phase then presumably represents the time needed to rearrange the bound trimers to assemble to a functional fusion complex (Fig. 70). The initial association could, for instance, involve attachment of spikes at the focal site of interaction in a random orientation. For the final fusion complex, we have hypothesized the formation of a rosette-like structure involving several HA trimers. After complex formation, merger of the membranes, the final event in membrane fusion (Fig. 7 E ) , takes place.