INTRODUCTION

The fusion of the viral and the endosomal membranes subsequent to the endocytic uptake of influenza virus is mediated by the integral membrane protein hemagglutinin (HA).¹ HA is a homotrimeric glycoprotein that consists of the two subunits, HA1, bearing the receptor binding site, and HA2, anchored with its C terminus in the viral membrane. At the acidic pH of endosomes, a conformational change of the hemagglutinin ectodomain is triggered (1) that activates its fusion capacity (2, 3). The three-dimensional structure of the ectodomain of HA (X31) at neutral pH is known from x-ray crystallography with a resolution of 3 Å (4). Although the fusion active conformation is still unknown, one of its characteristics is, as has been shown, e.g. by anit-peptide-antibodies (1, 5), the exposure of the highly conserved N terminus of the HA2 subunit, the "fusion sequence," which is originally buried within the stem of the trimer (4). Based on the x-ray crystal structure of a fragment of the HA ectodomain from X31 in its low pH form, Wiley and co-workers (6) suggested that the fusion sequence moves toward the top of the ectodomain by the formation of a long -helix in the HA2 subunit at low pH, thereby extending the trimer stem straight up. Studies with synthetic peptides (7) predicted that this long -helix is formed by a transition of a loop region of HA into an -helix at low pH.

However, when incubated at low pH in the absence of target membranes, influenza virus can rapidly lose its fusogenic properties (8-11). The degree and kinetics of this inactivation is virus strain-specific (9, 12, 13). Clustering of conformationally altered HA trimers (14, 15) and association of the fusion sequence with the viral membrane (16-19) have been suggested as possible reasons for inactivation.

In the present study we have investigated whether acid-induced conformational changes of hemagglutinin leading either to the activation or inactivation of the fusogenic properties of HA, respectively, can be distinguished by CD spectroscopy in far and near UV. To this end HA trimers were purified from various influenza strains belonging to different subtypes (H1 (A/PR 8/34); H2 (A/Japan/305/57); H3 (X31)) by detergent extraction. Typically, in aqueous suspension HA trimers are organized in rosettes (20, 21). The near UV CD spectrum is determined by the three-dimensional arrangement and flexibility of aromatic amino acids, mainly tyrosine and tryptophan, and thus is sensitive to alterations of the spatial structure of the protein. HA of A/Japan contains 22 and 11, HA of X31 18 and 12, HA of A/PR 8/38 35 and 10 tyrosine and tryptophan residues, respectively. The aromatic amino acid residues of HA are almost exclusively located in the protein ectodomain. Thus, the near UV CD spectrum is predominantly determined by the higher order structure of the ectodomain. To resolve rapid alterations and short-lived intermediates of HA structure we have employed time-resolved CD spectroscopy in the near UV. We found transient rearrangements of the higher order structure of the HA ectodomain after acidification. Initial alterations occurred in less than 1 min at pH 5.0 and 37 °C. Significant differences between the various influenza subtypes were detected in the pH and temperature dependence of structural changes of HA. Comparing these results to the time dependence of viral fusion with erythrocyte membranes and of the low pH-mediated inactivation of fusion activity, we were able to relate distinct phases of the structural rearrangements detected by near UV CD to the formation of a fusion-competent state and of a fusion-inactivated form of HA, respectively. Our data support the hypothesis that the fusion-competent conformation of HA is a precursor of the fusion-inactivated structure of HA.

MATERIALS AND METHODS

Octadecylrhodamine B chloride (R18) was purchased from Molecular Probes (Eugene, OR). Fresh blood from healthy donors was obtained from the Blood Bank, Berlin-Lichtenberg, and was used within 3 days after sampling. Purified influenza virus X31 and A/Japan (A/Japan/305/57) were kindly provided by Dr. Robert Blumenthal (NIH, Bethesda, MD) and purified influenza virus A/PR 8/34 by Dr. Ulrike Gimsa (Federal Health Office, Institute of Veterinary Medicine, Berlin).

In dependence on the desired pH the following buffers were used (i) phosphate-buffered saline (PBS, 5.8 mmol/liter phosphate, 145 mmol/liter NaCl) (pH 7.4) or (ii) sodium acetate buffer (20 mmol/liter sodium acetate, 130 mmol/liter NaCl) (pH < 6.0 adjusted with citric acid).

After removal of buffy coat and plasma red blood cells were washed three times in PBS (pH 7.4). Unsealed erythrocyte ghosts were prepared according to Dodge et al. (22).

Influenza virus was grown for 48 h in the allantoic cavity of 11-day-old embryonated hen eggs. The allantoic fluid was collected, and cell debris was removed by a low speed spin (880 × g, 30 min). The virus was pelleted by spinning the allantoic fluid at 95,000 × g for 90 min. The pellet was resuspended in PBS and homogenized with a Teflon-coated homogenizer.

The viral membrane was solubilized in PBS containing 1.5% octylglycoside (Boehringer Mannheim GmbH, Germany) for 1 h at 4 °C. The insoluble material was removed by centrifugation at 100.000 × g for 60 min. For purification by affinity chromatography according to Doms et al. (23), the supernatant containing the hemagglutinin in detergent micelles was passed over an CL4B-column loaded with Ricin A (Sigma). HA was eluted by PBS containing 1% galactose in the presence of 1.5% octylglycoside. The detergent and galactose were removed by dialysis against PBS for about 14 h with one buffer change. By this procedure rosette-like structures were formed (20, 21) with about 5-9 trimers/rosette as visualized by cryoelectron microscopy (not shown). Purity of the hemagglutinin was checked by SDS-polyacrylamide gel electrophoresis with 12% gels under reducing conditions (not shown).

1.25 ml of a 2 mM stock solution of R18 in ethanol were added by rapid vortexing to 0.25 ml of influenza virus (1 mg of virus protein/ml). After incubation for 30 min at room temperature (in the dark) virus was washed by high speed centrifugation (45,000 × g) with ice-cold PBS, to remove unbound R18, and resuspended to a concentration of 1 mg of virus protein/ml (24, 25). The final concentration of added probe corresponds to approximately 2 mol % of total viral lipid. The protein concentration of viruses as well as of ghosts was determined using a total protein kit (Sigma-Aldrich Chemie GmbH, Deisenhofen, Germany).

Influenza virus 1 mg/ml was preincubated at low pH (see legends to figures) in the absence of the target membrane. After different time periods the virus suspension was neutralized (pH 7.4) and kept on ice. Subsequently virus was bound to cell membranes, and fusion was measured as described below.

Labeled virus (0.1 mg of protein) was incubated for 30 min on ice with 0.2 ml of erythrocyte ghost suspension (6-7 mg of protein/ml). Afterward, the suspension was washed in 10 to 15 volumes of ice-cold PBS, and the sediment was resuspended with PBS to a final concentration of 1 mg of virus protein/ml.

Fusion was measured by fluorescence dequenching of the lipid-like fluorophore R18 upon fusion of R18-labeled viruses with ghost membranes (26). Fusion was triggered by transferring 30 µl of ice-cold virus-ghost suspension to a quartz cuvette containing 1.8 ml of pretempered sodium acetate or PBS buffer of the respective pH (pH 4.5-7.4). The suspension was stirred continuously with a 2 × 8-mm Teflon-coated magnetic stir bar. Fusion was monitored continuously by measuring fluorescence dequenching (_ex, 560 nm; _em, 590 nm; cutoff filter, 570 nm) with a time resolution of 0.5 s. At the end of each experiment Triton X-100 (0.5% final concentration) was added to obtain maximum R18 fluorescence F(max). The percentage of fluorescence dequenching FDQ was calculated as described previously (27),

(Eq. 1)

with F(0) and F(t) corresponding to the fluorescence intensity before starting fusion and the fluorescence intensity at a given time t, respectively. pH-dependent alterations of the molecular properties and, thus, fluorescence of R18 have not been found (26).

CD spectra were recorded on a JASCO J-720 spectropolarimeter interfaced with an external water bath to maintain temperature control and connected to a PC for data recording and processing. Fused silica cells of 1 (far UV) or 10 mm (near UV) path length were used. For all measurements the HA concentration was adjusted to 0.5 mg/ml in sodium acetate buffer, pH 7.4. The pH was lowered by addition of microliter amounts of 250 mmol/liter citric acid. Spectra were recorded from 200 to 250 nm and 260 to 300 nm using a scan speed of 10 nm/min. The spectra correspond to the average of three scans. For data processing, the baseline was subtracted, and spectra were smoothed using the spectropolarimeter supplied software package. Kinetics of the changes in the CD signal at 283 nm were measured with a time resolution of 5 s. Due to the experimental setup, the first 30 s after lowering the pH of the suspension medium of HA could not be measured. Data are presented as the mean residue ellipticity based on a mean residue weight of 105.

HA sequences were obtained from the protein data base "SWISS-PROT."

RESULTS

The pH dependence of the secondary as well as the higher order structure of HA was assessed by CD in the far and near UV, respectively. As already reported for HA of X31 (28), we detected no significant changes of the far UV CD spectrum of HA from the various strains in the pH range from 4.6 to 7.4, neither at 37 °C nor at 20 °C (not shown). Thus, the relative contributions of various secondary structural elements of HA are independent of the pH in the investigated range.

Near UV CD spectra of HA from various strains measured at pH 7.4, 37 °C (not shown), were comparable to those reported previously for X31 by Wharton et al. (28). Negative bands of the spectra at about 283 and 290 nm correspond typically to tyrosine and tryptophan residues, respectively. The ratio between the intensities at 283 and 290 nm is not solely determined by the ratio between the numbers of these amino acids as deduced from a comparison of the spectra of HA from various strains but, too, by the intramolecular arrangement and molecular environment of the aromatic amino acids. While at pH 7.4 the near UV CD spectrum of HA remained unchanged within the experimental time course (up to 90 min), we observed time-dependent alterations of the spectra after acidification, indicating a rearrangement of the tertiary and/or quaternary structure. To resolve the kinetics of the low pH-triggered conformational change of HA, we have measured the time dependence of the CD intensity at a selected wavelength after acidification with a time resolution of 5 s.

At pH 5.0, 37 °C, we found for the virus strains X31 and A/Japan a rapid, transient increase of the negative CD intensity at 283 nm (Figs. 1A and 2A) followed by a slow, but continuous, decrease of the signal. For HA of A/PR 8/34 we observed only a drop of the CD signal after acidification (pH 5.0) but not an initial increase of the CD amplitude neither at 37 °C (not shown) nor at 20 °C (see below). Our data do not preclude the existence of such an initial phase for HA of A/PR 8/34 since its kinetics may be too fast to detect with our experimental setup (see "Materials and Methods").

When performing experiments at a slightly higher pH (pH 5.4, 37 °C), a different time-dependent behavior of the near UV CD signal emerged. For HA of X31 the initial increase of the absolute amplitude proceeded much slower (Fig. 1A), while for A/Japan again a rapid enhancement of the negative CD intensity upon acidification was measured (Fig. 2A). The subsequent decline of the CD intensity observed at pH 5.0 (37 °C) did not occur (X31) or was only very low (A/Japan) over the time course of the experiments.

To elucidate the influence of temperature we have measured the kinetics of the CD signal at 283 nm at 20 °C, pH 5.0. We found that the effect of a lowered temperature on the near UV CD intensity of HA from X31 and A/Japan, respectively, was comparable to that monitored by shifting the pH to 5.4 (37 °C, see above). For HA of X31, the initial phase, which is rapid at 37 °C, became significantly delayed at 20 °C (Fig. 3A). Only after about 80 min the intensity was like the value obtained at 37 °C immediately after acidification (pH 5.0). We did not observe a decline of the negative CD intensity at 20 °C over the time course of the experiment. For HA of A/Japan the initial enhancement of the CD intensity remained fast also at 20 °C (Fig. 4A). Subsequently, only a small decrease of the negative CD amplitude was observed at 20 °C, pH 5.0. 

For HA of A/PR 8/34 we observed only a rapid loss, but no initial enhancement of the negative CD signal at 283 nm at 20 °C, pH 5.0 (Fig. 5A). However, as already mentioned we cannot exclude that the latter alteration may exist but which is too fast to detect with our experimental.

For probing the reversibility of alterations of the higher order structure of HA at low pH, we have chosen HA from A/Japan due to its rather slow disappearance of the CD signal at pH 5.0, 37 °C. This enabled us to investigate the reversibility after different times of incubation of HA at pH 5.0, 37 °C, in a well defined manner by switching the pH back to 7.4 (Fig. 6). After 15 min of incubation at pH 5.0 the observed CD intensity was comparable to that found initially at pH 7.4 (at time t < 0, see Fig. 6A). This similar CD signal does not reflect the same structure of HA since a shift of the pH to 7.4 reestablished the negative CD intensity (Fig. 6A) measured after about 30 s of lowering the pH to 5.0. But, while the decline of the negative CD intensity was reversed by reneutralization after a short incubation (15 min), it was preserved after a prolonged period (60 min) at pH 5.0, 37 °C (Fig. 6B). These data suggest that the initial alteration of the tertiary and/or quaternary structure of HA, detected by near UV CD after acidification, is irreversible, while the reversibility of the second phase, the disappearance of the CD signal, becomes lost only upon longer incubation.

The fusion of influenza viruses with erythrocyte membranes as well as low pH-mediated inactivation of viruses (see below) were measured at conditions similar to that of CD measurements. Significant fusion with erythrocyte membranes probed by the well established R18-FDQ assay (26) was observed for all three virus strains at pH 5.0, 37 °C (Figs. 1B and 2B) in agreement with previous reports (11, 13, 25). Shifting of the pH to 5.4 (37 °C) did result in a decrease of the extent but not of the initial rate of fusion for A/Japan (Fig. 2B). For X31 we monitored a reduced initial rate but only a slight decrease of the extent of fusion at pH 5.4 in comparison to pH 5.0 (Fig. 1B). The data indicate that the pH dependence of the initial rate of fusion correlates with that of the initial rate of the increase of the negative CD intensity at 283 nm (see above). This is supported by experiments on the low pH mediated fusion of virus strains A/Japan (Fig. 4B) and X31 (Fig. 3B) with erythrocyte membranes at different temperatures, 37 and 20 °C, respectively (pH 5.0). The rapid fusion of influenza A/Japan seen at 37 °C is preserved at the lower temperature. This parallels with the preservation of the fast increase of the negative CD signal in the near UV at pH 5.0 upon lowering the temperature to 20 °C (see above). Such a coincidence between the kinetics of the near UV CD signal of HA and of the fusion activity became also visible in the case of X31. By shifting the temperature to 20 °C, the initial rates of both the fusion activity and the increase of the near UV CD signal of HA of X31 became reduced (see above).

We have measured the fusion activity of influenza viruses after preincubation for various times at different acidic pH. The fusion activity was assessed at pH 5.0, 37 °C, subsequently to binding of preincubated viruses to erythrocyte membranes at neutral pH (see "Materials and Methods"). Preincubation of influenza virus X31 at pH 5.0, 37 °C, was accompanied by a significant and fast drop of fusion capability (Fig. 1C). In agreement with previous results (12, 13), we found that the fusion activity of A/Japan (subtype H3) is less sensitive to low pH preincubation and declined much slower at pH 5.0 (Fig. 2C). However, we did not find any detectable inactivation of the fusogenic capacity for both strains when viruses were pretreated at pH 5.4, 37 °C (Figs. 1C and 2C) or at lower temperature (20 °C, pH 5.0) (Figs. 3C and 4C). We note that low pH preincubation of X31 and A/Japan at conditions preventing inactivation processes can lead to an enhancement of the initial rate of low pH-mediated fusion with erythrocyte ghosts.²

The results suggest a correlation between the stability of the negative CD signal of HA at 283 nm and the preservation of the fusion activity upon preincubation of influenza viruses at pH 5.4 (37 °C) or at 20 °C (pH 5.0). Presumably, the loss of the CD signal reflects the inactivation of HA. This conclusion is supported by our observations on A/PR 8/34. In contrast to X31 and A/Japan, the CD amplitude of HA from A/PR 8/34 at 283 nm declined rapidly at pH 5.0, 20 °C (Fig. 5A). Likewise, the fusion activity of A/PR 8/34 with erythrocyte membranes disappeared rapidly upon preincubation of viruses at those conditions (Fig. 5C).

DISCUSSION

We have studied the conformational transition of HA of various influenza virus strains at low pH continuously by time-resolved CD spectroscopy. Due to the almost exclusive localization of aromatic amino acid residues in the ectodomain of HA, alterations of the CD signal in the near UV reflects changes of the spatial arrangement and dynamics of this domain, but not of the transmembrane and intraviral HA domains. We could show that the kinetics of the CD signal of HA in the near UV upon lowering the pH are multiphasic. As far as we can deduce from CD data, the low pH-mediated conformational changes of influenza HA from various strains proceed presumably via similar stages but the kinetics of those stages may be strain-specific. We suspect that this observation reflects differences in the structure of HA based on the amino acid sequence of HA. Comparison with the kinetics of fusion between viruses and erythrocyte membranes enabled us to associate alterations of the CD signal with the formation of a fusion-competent intermediate of HA. By measuring the extent of the time-dependent low pH-mediated inactivation of fusion of influenza viruses, we could ascribe specific changes of the near UV CD to the formation of a fusion-inactivated structure of the HA ectodomain. Thus, we were able to distinguish between fusion-competent and fusion-inactivated stages. However, since near UV CD spectroscopy yields only qualitative indications for a rearrangement of the higher order, the tertiary and quaternary structure of a protein, we cannot ascribe alterations of the CD signal to distinct conformational changes of the HA ectodomain.

Typically we found a biphasic kinetics of the CD signal at 283 nm of HA from X31 and A/Japan at pH 5.0, 37 °C. Initially, a rapid raise of the negative CD amplitude was measured upon acidification. Subsequently to this phase a slow, but continuous, decline of the CD amplitude was observed. Variation of temperature and the acidic pH in a range where still significant fusion was observed did exhibit a different, virus strain-specific influence on both phases. The fast initial increase at pH 5.0, 37 °C, could not be well resolved by our experimental setup. We estimated that the half-time of this process is less than 30 s. Indeed, it has been shown by different approaches that structural alterations of the ectodomain upon acidification may occur in the order of minutes or even seconds. For example White and Wilson (1) have detected by binding of antibodies two subsequent conformational changes of the HA ectodomain of X31 with a half-time of 45 s and 4 min, respectively, at pH 5.0, 37 °C. Stegmann et al. (5) concluded that the exposure of hydrophobic sequences of the HA ectodomain (X31) at pH 5.1 detected by binding of X31 viruses to liposomes occurs in less than 15 s. Recently, we have shown by measuring the binding of the fluorescent probe bis-1-anilino-8-naphthalenesulfonate to the HA ectodomain (A/PR 8/34) (29) as well as the intrinsic tryptophan fluorescence of the ectodomain (X31) (30) that the low pH triggered conformational change is in the order of seconds or even less (pH 5.0, 37 °C).

How does the changes in the CD signal in the near UV correlate with the fusion activity of the various influenza strains? We found a close relation between the kinetics of fusion and of the initial increase of the negative CD amplitude at 283 nm. (i) For X31, at pH 5.0, 37 °C, both the increase of the CD signal as well as virus-ghost fusion proceeded rapidly. Shifting of the pH to slightly higher values (pH 5.4) or decreasing the temperature (20 °C) the initial rate of the CD signal and, in parallel, that of fusion were reduced. (ii) In the case of HA from A/Japan, the initial rapid increase of both the CD signal and the fusion, respectively, observed at pH 5.0 (37 °C) was not significantly affected by alterations of pH (5.4) or temperatures (20 °C). We suggest that the increase of the CD signal at 283 nm next to acidification is associated with the formation of a fusion-competent HA intermediate.

The subsequent decrease of the CD signal is closely associated with the formation of a fusion inactivated state of HA. This conclusion is based on our observation of a functional correlation between the disappearance of the near UV CD signal of HA and the inactivation of the fusogenic properties by low pH preincubation of viruses. For X31 and A/Japan, we found a significant decline of the negative CD amplitude and inactivation by preincubation at pH 5.0, 37 °C. At pH 5.4 (37 °C) or at pH 5.0 (20 °C), where the decay of the CD signal was significantly reduced or even abolished only a slight inactivation of the fusion activity of both viruses was observed. We note that the decline of the CD intensity and fusion inactivation correlates qualitatively, but not quantitatively. Both X31 and A/Japan have similar rates of CD signal loss at 283 nm, but different inactivation behaviors at pH 5.0, 37 °C. However, our conclusion that the loss of the CD signal in the near UV at low pH reflects the transformation to a fusion-inactivated stage of HA is supported by results from influenza A/PR 8/34. In contrast to X31 and A/Japan, we found a fast decrease of the CD signal of HA at pH 5.0 even at 20 °C, paralleled by a rapid loss of the fusion activity of A/PR 8/34.

Taken together, at least two subsequent stages of the conformational transition of HA at low pH can be distinguished by CD spectroscopy. The first stage upon acidification resembles a conformation associated with the fusion activity of HA, while the second stage reflects the formation of the inactivated structure of HA. The CD data suggest that the formation of the fusion-active HA conformation upon acidification is not characterized by a loss but by a rearrangement of the higher order structure of the HA ectodomain in agreement with current models (5, 6, 17, 31, 32). Likewise, and also in line with recent literature (17-19) (see below), our data do not point to a disappearance of the higher order structure and a rather disordered arrangement of the fusion-inactivated HA because no alteration of the secondary structure has been observed. The decrease of the CD intensity may suggest a structure of the inactivated ectodomain enabling a higher mobility of the aromatic side chains. It is already known that the intensity of the near UV CD spectrum is determined also by the rigidity of the protein (28, 33).

Two pieces of evidence suggest that the formation of the inactivated form of HA proceeds via intermediates of a fusion-competent state of HA. As we have shown for A/Japan, after a rather short incubation of HA at low pH (pH 5.0, 37 °C, 15 min) the loss of CD amplitude was reversible upon reneutralization, and the initial increase of the amplitude of the CD signal observed immediately upon acidification could be restored. The latter has been associated with the formation of a fusion-competent state that could be restored (see above). Support that indeed a fusion-competent structure could be restored came from the observation that we did not find a significant inactivation of fusion of A/Japan with ghosts after preincubation at pH 5.0, 37 °C, for 15 min. Therefore, the decline of the CD intensity of HA from A/Japan measured under those conditions is not associated with any irreversible loss of the fusion capacity of HA. However, a reduction of the fusion extent to about 25% of that of the control after low pH preincubation of viruses for 60 min correlates with the irreversible loss of the CD intensity. These data imply that the formation of an inactivated structure of HA may proceed via a fusion-competent conformation of HA. Such a hypothesis has been put forward by different studies (34, 35). Ramalho-Santos et al. (34) conclude from studies on fusion of influenza virus A/PR 8/34 with cultured cells that the rate-limiting step in HA-mediated fusion is similar to that involved in fusion inactivation. Recent investigations, using a photoactivatable cross-linking reagent (16) or performing electron microscopy (17-19), imply that the inactive conformation of HA is characterized by the association of the fusion sequence with the viral membrane. The latter can be caused by an inversion of the orientation of the long -helix subsequently to its formation at low pH (see the Introduction) toward the viral membrane (17-19). Thus, it is quite conceivable that the inactivated form of HA is the final state of successive conformational intermediates at low pH to which also the fusion-competent structure belongs (19).

Moreover, our results indicate that structural alterations of HA, which are subsequent to the fusion-competent intermediate and lead to the inactivated form, resemble a multistep process that in its early phase(s) is partly reversible. In the case of an one-step transition to the inactivated form we would not expect (i) the reversibility of the decline of the CD signal of HA upon reneutralization and (ii) a preservation of the fusion activity after an incubation for 15 min at pH 5.0, 37 °C, as shown for A/Japan (see above). Very recently, we have concluded from simulation of experimental data of virus fusion by a kinetic model that inactivation proceeds presumably via various intermediates but not in one step (35).

Finally, we would like to emphasize that our results may also have several implications for elucidating the fusion-active conformation of HA by appropriate methods. It is important for those investigations to choose conditions at which the fusion active conformation is stable and the inactivated structure is not formed in the time course of the experiment. To account for that, first, the choice of the virus strain is of importance. For example, as we have shown, the HA of A/PR 8/34 is not recommended for those investigations due to its rapid transformation into a fusion-inactive form, even at conditions suboptimal for fusion. Second, the choice of the acidic pH and the temperature may influence significantly the life time of the fusion-active conformation. It might be of value to choose those conditions which are suboptimal for fusion.