H+-induced membrane insertion of influenza virus hemagglutinin involves the HA2 amino-terminal fusion peptide but not the coiled coil region.

Fusion of influenza virus with target membranes is induced by acid and involves complex changes in the viral envelope protein hemagglutinin (HA). In a first, kinetically distinct step, the HA polypeptide chain 2 (HA2) is inserted into the target membrane bilayer. Using hydrophobic photolabeling with the phospholipid analogue 1-O-hexadecanoyl-2-O-[9-[[[2-[125I]iodo-4(trifluoromethyl-3H-diazirin-3-yl)benzyl]oxy]carbonyl]nonanoyl]-sn-glycero-3-phosphocholine, we identified the segment within HA2 that interacts with the membrane. The sole part of the HA2 ectodomain that was labeled with the membrane-restricted reagent is the NH2-terminal fusion peptide (residues 1-22). No labeling occurred within the long coiled coil region generated during the acid-induced conformational transition (Bullough, P. A., Hughson, F. M., Skehel, J. J., and Wiley, D. C. (1994) Nature 371, 37-43). These data strongly suggest that the coiled coil region of HA2 does not insert into the lipid bilayer. This conclusion is at variance with the recent suggestion (Yu, Y. G., King, D. S., and Shin, Y.-K. (1994) Science 266, 274-276) that the coiled coil of HA may splay apart and insert into the target membrane, providing a mechanism by which the viral and the target membrane may come in close apposition.


Fusion of influenza virus with target membranes is induced by acid and involves complex changes in the viral envelope protein hemagglutinin (HA). In a first, kinetically distinct step, the HA polypeptide chain 2 (HA2) is inserted into the target membrane bilayer. Using hydrophobic photolabeling with the phospholipid analogue 1-O-hexadecanoyl-2-O-[9-[[[2-[ 125 I]iodo-4-(trifluoromethyl-3H-diazirin-3-yl)benzyl]oxy]carbonyl-]nonanoyl]-sn-glycero-3-phosphocholine, we identified the segment within HA2 that interacts with the membrane. The sole part of the HA2 ectodomain that was labeled with the membrane-restricted reagent is the NH 2 -terminal fusion peptide (residues 1-22). No labeling occurred within the long coiled coil region generated during the acid-induced conformational transition (Bul
Enveloped viruses enter cells by a process that involves fusion of the viral envelope with a membrane of the host cell (for recent reviews, see Refs. [1][2][3][4][5]. Fusion is catalyzed by viral envelope glycoproteins, among which hemagglutinin (HA), 1 the fusion protein of influenza virus, is best characterized and has provided the paradigm for viral fusion proteins. Following binding of influenza virus to sialylated surface receptors, a step also mediated by HA, the virus is internalized. In the acidic environment of the endosomal compartment, the HA protein switch from a inactive to a fusion-active state, promoting fusion of the viral envelope with the endosomal membrane.
Fifteen years ago, the atomic structure of BHA, the bromelain-solubilized form of HA, has been determined (6). Perhaps the most amazing aspect of this structure is that the HA2 NH 2 -terminal region, the so-called fusion peptide, is buried inside the homotrimeric HA globular structure. At the pH of fusion this peptide becomes exposed and can insert into the target membrane bilayer, as suggested by photolabeling studies (7,8). How this may happen can be envisioned from the pH 5 structure of TBHA2, a proteolytic fragment lacking most of HA1 as well as the fusion peptide itself (9). A 28-residue loop region that connects a short helix to a long helical stem is recruited to extend the triple-stranded ␣-helical coiled coil toward the amino terminus of HA2. In this way the fusion peptide moves by at least 100 Å to one tip of the molecule and can project toward the target membrane above the globular head domains.
However, this structural reorganization does not explain how the two membranes are brought together. On the basis of their finding, that synthetic 40-residue peptides corresponding to the loop region of native HA inserts into model lipid membranes at the pH of fusion, Yu et al. (10) proposed that, after initial attachment of the fusion peptide the coiled coil splays apart and inserts into the lipid bilayer of the target membrane, providing the mechanism for membrane merging.
In extending previous investigations (7,8,(11)(12)(13)(14), we now identify the polypeptide segment within HA2 that inserts into the membrane bilayer. Using the photoreactive lipid 125 I-TID-PC/16 (15) as the labeling reagent and new protocols for HA2 fragmentation, we show that the sole part of HA that becomes labeled in a pH-dependent manner is the NH 2 -terminal fusion peptide. The same is the case for BHA, the soluble form of HA, which inserts into membranes solely through its NH 2 -terminal fusion peptide. In contrast, synthetic loop-40 peptide in the presence of liposomes bearing 125 I-TID-PC/16, becomes labeled in a pH-dependent manner.

MATERIALS AND METHODS
Chemicals and Reagents-Egg PC (grade I) was from Lipid Products (South Nutfield, United Kingdom), POPC from Bachem Feinchemikalien AG (Bubendorf, Switzerland) and POPG from Avanti Polar Lipids, Inc. (Birmingham, AL). 125 I-TID-PC/16 and 125 I-TID-BE were prepared from the corresponding tin precursors as described (15,16). The product was stored as a solution in toluene/ethanol (2:1, v/v). Thermolysin was from Sigma and BNPS-skatole from Fluka (Buchs, Switzerland). Prior to use, BNPS-skatole was crystallized from acetone.
Viruses-The X-31 recombinant strain of influenza virus was propagated in the allantoic cavity of embryonated eggs and purified as * This work was supported by a grant from the Swiss National Science Foundation, Berne (to J. B.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Preparation of LUVs-LUVs were prepared by the extrusion technique of Hope et al. (18) using polycarbonate filters with a 0.2-m pore size.
Photolabeling-The Eppendorf tube containing the liquid sample was placed in a Pyrex glass vessel mounted approximately 10 cm from a SUSS LH 1000 lamphouse (KARL SUSS, Waterbury Center, VT) equipped with an Osram HBO 350 watt short-arc high pressure mercury lamp. Photolysis was performed for 30 s at a light intensity of 30 milliwatts⅐cm Ϫ2 . Under these conditions more than 90% of the reagent is photolyzed.
Isolation of 125 I-TID-PC/16-labeled HA Polypeptides-To the photolabeled viruses (0.5 ml) were added 3 volumes of chloroform/methanol (1:2, v/v). After 1 h at room temperature, the precipitated protein was collected by sedimentation (10 min at 14,000 rpm in an Eppendorf centrifuge). The protein was dissolved in sample buffer (95°C for 3 min) and subjected to SDS-PAGE (nonreducing conditions) using a Tris/ Tricine-buffered system (19). The HA band, visualized by brief staining with Coomassie Brilliant Blue R-250, was excised and the protein electroeluted at 50 mA for 4 h using a homemade apparatus. The elution buffer contained Tris (25 mM), glycine (0.18 M) and 0.1% SDS. The protein solution (1-2 ml) was concentrated by a Centricon 30 microconcentrator (Amicon) and re-electrophoresed under reducing conditions. After staining, the HA1 and HA2 bands were excised and the gel slice containing HA1 was processed for protein quantification using the Coomassie Brilliant Blue R-250 staining/extraction procedure of Ball (20). Stain was compared with that of reference samples determined by a modified Lowry procedure (21) with bovine serum albumin as a standard. HA2 was electroeluted, concentrated in Centricon 10 microconcentrator, precipitated with 9 volumes of acetone (this step served to remove the Coomassie Blue stain), and dried in vacuo.
BNPS-Skatole Cleavage of (B)HA2 and Separation of Fragments-HA2 or BHA2 was electroeluted from the SDS-polyacrylamide gel and the protein solution was concentrated using a Centricon 10 microconcentrator (Amicon, Inc.). SDS was removed by ion pair extraction as described previously (22). BNPS-skatole cleavage was performed in 60% acetic acid (0.5 g of protein/l) in the presence of 1 part of tyrosine and 10 parts of BNPS-skatole (by weight) under nitrogen for 20 h in the dark. Subsequently, excess cleavage reagent and by-products were extracted with 3 volumes of ethyl acetate. The aqueous phase was dried in the Speed Vac, and the residue dissolved in sample buffer for subsequent SDS-PAGE (Tris/Tricine system; 16% polyacrylamide, 6 M urea). For detection and quantification of the 125 I-TID-PC/16 radioactivity, the gel was subjected to autoradiography using a PhosphorImager (Image-Quant software, version 3.3) from Applied Biosystems.
Thermolysin Cleavage of 125 I-TID-PC/16-labeled BHA-To the solution of photolabeled BHA (30 g of protein in 106 l of fusion buffer adjusted to pH 5.0) was added 0.5 l of 1 M CaCl 2 and 5 g of thermolysin (1 g/l fusion buffer, pH 5). After incubation for 6 h at 37°C, protein was precipitated by the addition of 3 volumes of methanol/ chloroform (2:1, v/v), collected by sedimentation, and subjected to SDS-PAGE using 16% polyacrylamide containing 2.5 M urea (23).
pH-dependent Labeling of Loop-40 and BHA-LUVs were prepared from a mixture of phospholipids (POPC/POPG; molar ratio 8:2) in buffer A (100 mM NaCl, 50 mM citric acid, and 50 mM sodium phosphate adjusted to pH 7.4 with 2 M NaOH). An ethanolic solution of 125 I-TID-BE was added (final concentration of ethanol Ͻ2%). Incubation mixtures were then prepared containing LUVs (330 l in buffer A), 30 Ci of 125 I-TID-BE, and BHA or loop-40. The final volume of each sample was 0.5 ml. The lipid concentration was 5 mM, that of BHA or loop-40 was 10 M. After incubation (3 min at 37°C), 47 l of 1 M HCl (pH 7 control samples: 47 l of buffer A) was added to adjust the pH to 4.6. After a further incubation at 37°C for 10 min, the samples were exposed to UV light (1 min). The BHA-containing samples were further processed as follows: after neutralization (addition of 47 l of 1 M NaOH), 3 volumes of methanol/chloroform (2:1, v/v) were added and the precipitated protein analyzed as described above. Labeled loop-40 was isolated as follows: the photolyzed samples were concentrated on a Speed Vac, and the residue dissolved in 1 ml of ethanol, 90% formic acid (2.8:1, v/v). This solution was applied to a Sephadex LH-60 column (1.5 ϫ 30 cm) which had been equilibrated and was eluted with the same solvent. Fractions (1 ml) were collected and radioactivity determined by ␥-counting. Fractions containing the peptide were concentrated (Speed Vac), and the residue dissolved in sample buffer and analyzed by SDS-PAGE (Tris/Tricine system: 16% polyacrylamide, 6 M urea).
Sequence Analyses-Peptides separated by SDS-PAGE were transferred electrophoretically onto a polyvinylidine difluoride membrane using the semi-dry blot technique (Sartoblot II-S apparatus) and the buffer system of Laurière (24). Following brief staining with Coomassie Brilliant Blue R-250 and destaining, the bands containing individual peptides were excised and subjected to Edman sequence analyses. The latter analyses were performed in the ETH protein chemical laboratory (directed by Dr. Peter James).
Under prefusion condition labeling ( Fig. 2A and B, lanes 1), most of the radioactivity originally associated with HA2 was recovered within Hsk-5. The slight shift of the radioactive band toward higher molecular weight was expected and is due to the covalently bound lipid residue. While Hsk-6 may also contain traces of label, Hsk-7 as well as Hsk-4 are clearly not labeled. These results demonstrate that labeling of HA2 is restricted to the NH 2 -terminal fusion peptide (residues 1-22). The two weakly labeled bands at higher molecular weight are due to some uncleaved or incompletely cleaved material (HA2 and Hsk-1) bearing the fusion peptide. As expected, postfusion labeling of HA gave rise to a more complex pattern of radioactivity, reflecting labeling of the fusion peptide and of the COOHterminal anchor segment. The latter is evident from the heavily labeled fragment, Hsk-3, as well as the appearance of low molecular weight bands, most likely derived from the COOHterminal part of HA2. Despite the generally higher background radioactivity, we again note that only the larger band of the characteristic triplet is labeled.
Labeling of BHA-Next we investigated the interaction at pH 5 of BHA with liposomes containing 125 I-TID-PC/16. This soluble form of HA would be expected to undergo membrane insertion with particular ease. Following labeling of BHA and BNPS-skatole fragmentation of the BHA2 polypeptide, the cleavage products were analyzed by SDS-PAGE and autora-diography (Fig. 3). Again, we find radioactivity mainly associated with Bsk-3. Neither Bsk-5 nor Bsk-2 are detectably labeled. The second radioactive band corresponds to uncleaved BHA2.
BHA labeled at pH 5 was also subjected to thermolysin cleavage. A single major fragment of molecular mass 19 kDa was generated (Fig. 4). As confirmed by its NH 2 -terminal amino acid sequence (LKSTQA), this fragment was produced by removal from BHA2 of the NH 2 -terminal 37 residues and, therefore, corresponds to TBHA2, but lacking the disulfide bonded, short HA1 segment (23). As evident from Fig. 4, removal of the NH 2 -terminal region from BHA2 resulted in a complete loss of radioactivity, a result fully consistent with the data from chemical cleavage.
pH-dependent Labeling of Loop-40 Peptide-The absence of label within the coiled coil region of HA or BHA prompted us to investigate the membrane-binding behavior of a synthetic peptide (loop-40) that corresponds to a segment of HA2 from residue 54 to 93. This segment comprises the loop region and the first part of the long ␣-helical stem. It is thus identical to those peptides investigated by Yu et al. (10), except that it lacks the amino acid substitutions that had been necessary in the former study to introduce the nitroxide spin label.
For these experiments 125 I-TID-BE was used as the labeling reagent. Owing to its smaller size as compared to 125 I-TID-PC/ 16, labeled peptide and lipid could be more easily separated by Sephadex LH-60 gel filtration. Fractions containing loop-40 were analyzed by SDS-PAGE and autoradiography (Fig. 5). As shown in panels A and C, loop-40 was eluted between fractions 22 and 31. As assessed by Coomassie Blue staining, more than 70% of the peptide applied to the column was recovered. The corresponding autoradiographs revealed clear differences in the extends of labeling of loop-40 at pH 7.4 (panel B) and 4.6 (panel D), respectively. Whereas at pH 7.4 loop-40 was labeled only very weakly (less than 0.005% of the original radioactivity), labeling at acidic pH lead to clear incorporation of radioactivity into loop-40 (0.025% of the original radioactivity). The heavy radioactivity seen near the bottom at the right-hand sides of the gels originates from 125 I-TID-BE-labeled phospholipid which was eluted after loop-40 from the column. Experiments were also performed to compare the extent of labeling of loop-40 with that of BHA. The large excess of lipid (5 mM) over loop-40 or BHA (each 10 M) provide conditions that allow for (nearly) complete binding of the peptide/protein to the liposomes (10). After labeling at pH 4.6 the specific labeling of loop-40 and that of the BHA2 subunit of BHA were compared. BHA2 was labeled somewhat stronger (0.039% of the original radioactivity) than loop-40 (0.025%). We also confirmed that labeling of BHA by 125 I-TID-BE was confined to the fusion peptide (data not shown).

DISCUSSION
In a first, kinetically distinct step in the fusion process, influenza virions adhere to the target membranes. This study now establishes that the sole part of HA2 that inserts into the target membrane is the NH 2 -terminal fusion peptide. How this hydrophobic peptide can move outwards to the distal tip of the HA2 molecule and become accessible to the target bilayer is implicated by the x-ray structure of TBHA2, an aspect proposed earlier by Carr and Kim (25).
While insertion of the fusion peptide into the target bilayer is a key step in fusion, it does not explain how the two membranes come close together. A possible mechanism has been proposed by Yu  comprising part of the HA2 polypeptide chain. Following initial attachment of the fusion peptide, some or all of the long coiled coil trimer of HA may insert into the target bilayer, dragging the two fusing membranes into proximity. The results reported here do not support this model. Both during initial attachment and later in fusion, the sole region of the HA2 ectodomain that inserts into the membrane bilayer is the fusion peptide. This LUVs (5 mM lipid) prepared from POPC/POPG (molar ratio 4:1) and 125 I-TID-BE were incubated at pH 7.4 or 4.6. After reagent activation, the photolysis mixtures were concentrated on a Speed Vac, the residues dissolved in ethanol, 90% aqueous formic acid (2.8:1, v/v), and the solutions subjected to gel filtration using Sephadex LH-60 (same solvent). Individual fractions containing loop-40 were concentrated (Speed Vac) and the residues analyzed by SDS-PAGE. After staining with Coomassie Brilliant Blue R-250 (panels A and C), the radioactivity distribution pattern was determined with the PhosphorImager (panels B and D). Panels A and B, loop-40 labeled at pH 7.4; panels C and D, loop-40 labeled at pH 4.6. Lanes a, protein standards (cyanogen bromide peptides derived from myoglobin); lanes b, reference samples of unlabeled loop-40 (A, 3 g; B, 5 g). conclusion is further substantiated by experiments with soluble BHA. Unlike the situation for intact viruses, where penetration of the HA2 into the target membrane may be impeded by repulsive forces between the target and the viral membrane, insertion of BHA into the membrane is not hindered. Therefore, BHA was considered particularly suited to investigate a possible membrane interaction of the coiled coil region. While, as expected, the fusion peptide of BHA was strongly labeled, no labeling could be detected beyond residue 23 as assessed by the analysis of BNPS-skatole fragments. Although thermolysin fragmentation gave particularly clean results, these must be interpreted with some caution. Any BHA molecule that might splay apart and penetrate the membrane would possibly be degraded by protease and hence escape the analysis. However, the results from chemical fragmentation of labeled BHA rule out this possibility. In conclusion, these data suggest that the observed interaction of loop-40 peptide with membranes is unlikely to be relevant to what happens with intact HA during fusion. Membrane insertion of loop-40 as demonstrated by spin labeling experiments (10) and photolabeling may thus be a result of the marginal stability of this relatively short peptide (26).
How then can the two membranes be brought into close proximity? A possible clue to an answer may have come from previous studies suggesting that the HA2 fusion peptide can also insert into the viral membrane (14,23). Such topological configuration of the HA2 polypeptide chain is not only attained during acid inactivation of virus, but is also a consequence of membrane fusion following initial insertion of the fusion peptide into the target bilayer (27). That HA2 can adopt two distinct topological configurations is not inconsistent with the x-ray structure of TBHA2. In addition to the rearrangement resulting in NH 2 -terminal extension of the coiled coil, other changes occur in the COOH-terminal part of the molecule which becomes disordered possibly reflecting increased flexibility. This could allow the HA2 coiled coil to adopt two reversed orientations, one in which the fusion peptide is inserted in the target membrane, the other with the fusion peptide in the viral/fused membrane. As discussed previously (14,27) and depicted schematically in Fig. 6, the reversal of the coiled coil may be directly coupled to membrane merging and fusion. In the absence of a target membrane bilayer, the fusion peptide, after its transient exposure, may insert directly into the viral membrane, representing a plausible mechanism for the inactivation of the virus' fusion capacity.