|Home | About | Journals | Submit | Contact Us | Français|
Influenza virus enters host cells by endocytosis. The low pH of endosomes triggers conformational changes in hemagglutinin (HA) that mediate fusion of the viral and endosomal membranes. We have used cryo-electron tomography to visualize influenza A virus at pH 4.9, a condition known to induce fusogenicity. After 30 min, when all virions are in the postfusion state, dramatic changes in morphology are apparent: elongated particles are no longer observed, larger particles representing fused virions appear, the HA spikes become conspicuously disorganized, a layer of M1 matrix protein is no longer resolved on most virions, and the ribonucleoprotein complexes (RNPs) coagulate on the interior surface of the virion. To probe for intermediate states, preparations were imaged after 5 min at pH 4.9. These virions could be classified according to their glycoprotein arrays (organized or disorganized) and whether or not they have a resolved M1 layer. Employing subtomogram averaging, we found, in addition to the neutral-pH state of HA, two intermediate conformations that appear to reflect an outwards movement of the fusion peptide and rearrangement of the HA1 subunits, respectively. These changes are reversible. The tomograms also document pH-induced changes affecting the M1 layer that appear to render the envelope more pliable and hence conducive to fusion. However, it appears desirable for productive infection that fusion should proceed before the RNPs become coagulated with matrix protein, as eventually happens at low pH.
Influenza virus is an enveloped, single-stranded negative-sense RNA virus belonging to the Orthomyxoviridae family. Influenza A virions have three integral membrane proteins: an ion channel protein (M2) and two glycoproteins, hemagglutinin (HA), required for entry into host cells, and neuraminidase (NA), involved in the release of progeny virions from the host cell. Underneath the membrane, most virions have a layer of M1 matrix protein enclosing the viral ribonucleoprotein complexes (RNPs).
HA is a trimeric class I fusion protein synthesized as a precursor (HA0) that, to become fusion active, is cleaved by a cellular protease, yielding HA1 and HA2 subunits which remain disulfide linked (39). After the virus enters a host cell, HA is transformed by the low pH of the endosomes (4). The generally accepted model for these changes is based on X-ray crystallographic studies of the HA ectodomain at neutral (42) and low (2, 6, 10) pH. This model explains the low-pH-induced transition as an irreversible process (Fig. 1) in which (i) the HA1 membrane-distal domains dissociate (1, 15, 22), retaining their neutral pH structure; (ii) the fusion peptides are transferred to the membrane-distal region after a loop-to-helix transition in amino acids 55 to 76 of HA2 (known as the B loop); and (iii) there is a helix-to-loop transition in segment 106 to 112 of HA2 (the “kinked loop”) that positions the transmembrane domains and fusion peptides on the same side of the glycoprotein (19). The order in which these rearrangements take place, the existence of possible intermediate states, and the reversibility or otherwise of successive steps all remain to be established (11). However, mounting evidence points to the existence of reversible intermediate states, which would include an initial exposure of the fusion peptide and the kinked loop domain (25, 40) and a transition of the B loop to a loose conformation accompanied by a deformation of the HA1 subunit (43).
Previously, cryo-electron microscopy (cryo-EM) has been used to image influenza virions at neutral pH (3, 14, 44) and low pH (23, 32, 36, 37). However, the coprojection of structures from all levels in these ~100-nm-thick particles limited the interpretability of the resulting images. More recently, cryo-electron tomography (cryo-ET), which is capable of generating three-dimensional images of individual virions, has been used to compile a systematic account of the pleiomorphy exhibited in preparations of mostly spherical virions at neutral pH (17), to characterize filamentous viruses (7), and to investigate the initial steps of membrane fusion as mediated by influenza virus (24). In the present study, we used cryo-ET to characterize the structural changes undergone by virions of the X-31 strain of influenza A virus (serotype H3N2) when transferred to pH 4.9. Particular attention was paid to HA, whereby subtomogram averaging (12, 26, 41, 45, 47) was used to distinguish and characterize discrete conformational states of this fusogenic glycoprotein. We also observed major changes affecting the layer of matrix protein.
X-31 [A/Aichi/68 (H3N2)] influenza virus, grown in embryonated chicken eggs, purified according to standard procedures, and diluted to ~2 mg/ml protein in 140 mM NaCl, 10 mM HEPES, pH 7.4, was purchased from Charles River Laboratories (North Franklin, CT).
Grids of influenza virus at pH 7.4 were prepared essentially as described previously (17). In brief, virus was mixed 1:1 with a suspension of 10-nm bovine serum albumin (BSA)-coated colloidal gold particles (Aurion, Wageningen, The Netherlands) to serve as fiducial markers. Four-microliter drops were then applied to R2/2 holey carbon grids (Quantifoil; SPI, West Chester, PA), thinned by blotting, and vitrified by plunge-freezing in liquid ethane with a Vitrobot (FEI, Hillsboro, OR). To prepare the low-pH samples, Quantifoil grids were placed for 5 min on 10-μl drops of the virus suspension mixed 1:1 with gold particles. The pH was then lowered to 4.9 by adding 5 μl of 130 mM NaCl, 50 mM sodium acetate, pH 4.6. In experiments addressing the reversibility of the low-pH-induced changes, 1.25 μl of 0.5 M HEPES at pH 8.1 was added after 5 min of incubation at pH 4.9 in order to bring the final pH to 7.4. All incubations were done at room temperature (22°C) and in a humid chamber to avoid possible effects from buffer evaporation. After incubation, viruses were vitrified as described above. The resulting grids were then transferred to a cryogenic specimen holder (type 626; Gatan, Warrendale, PA) for data acquisition. Single-axis tilt series were recorded on a Tecnai-12 transmission electron microscope (TEM) (FEI) operating at 120 keV. The microscope was equipped with an energy filter (GIF 2002; Gatan) operating in zero-loss mode with an energy slit width of 20 eV. Images were acquired on a 2,048- by 2,048-pixel charge-coupled device (CCD) camera (Gatan), using SerialEM (29) for automated tilting, tracking, focusing, and image recording. In each tilt series, images were recorded in 2° increments over an angular range of approximately −66° to +66°, at an effective magnification of 38,500× (0.78-nm pixel size). Electron dose per image was ~1 e−/Å2, for a cumulative dose of ~70 e−/Å2 per tilt series. Image defocus was set at −4 μm, corresponding to a first contrast transfer function zero at (3.7 nm)−1.
Data were processed using the Bsoft package (21). Projections were aligned using 10-nm gold particles as fiducial markers, and tomograms were then reconstructed. The in-plane resolution of the tomograms was estimated as ~6.5 nm for full tomograms and 5.5 to 6.5 nm for individual virions as calculated by the NLOO-2D (noise-compensated leave-one-out in two dimensions) method (8) (Fig. 2A and B). Subtomograms containing individual virus particles were extracted and denoised by 20 iterations of anisotropic nonlinear diffusion (13) for further analysis.
Individual HA molecules were located on the denoised virion-containing subtomograms and extracted from the corresponding raw subtomograms in volumes of 50 by 50 by 50 voxels. The center of each virion was used as a reference point to determine an initial z-axis orientation for each glycoprotein. HAs from neutral-pH virions and from each low-pH group were aligned and averaged independently, using appropriate missing-wedge masks. The same synthetic model, consisting of a cylinder of approximately the same size as the neutral-pH HA spike and two sheets of density representing the viral membrane and matrix layers, was used as an initial reference in each case. Alignment procedures were performed with routines from Bsoft, modified as needed and wrapped into Python scripts. To improve the resulting density maps, original data were masked after a stable solution was obtained (usually, after 3 or 4 rounds of iteration). After seven rounds of refinement with the masked data, glycoproteins whose direction differed by more than 45° from the initial one (16 to 36% of the HAs, depending on the group) were discarded. Classification and averaging were then performed by maximum likelihood as implemented in the Xmipp package (35). The resolutions of the averages were 2.4 to 3.2 nm for different data sets, as calculated in terms of Fourier shell correlation coefficients (0.3 threshold) (Fig. 2C and D). All averages were low pass-filtered, and 3-fold symmetry was applied. The top 75% of subvolumes from each class, as ranked by correlation coefficients, were used to calculate the final density maps.
X-31 virions at neutral pH have been assigned to five morphological classes (17). The primary criteria were the presence or absence of a layer of the M1 matrix protein, underlying and resolved from the lipid bilayer of the viral envelope, and their internal contents, which consisted either of RNPs ~14 nm in diameter and of variable length or a single solenoid with a diameter of ~80 nm and a pitch of ~10 nm. Solenoid-containing virions are quite common in certain strains of influenza virus, such as A/USSR/50/79 or A/Texas/1/77 (30), but they are rare with X-31, accounting for <2% of the virions observed in the earlier study (17) and a negligibly small fraction of the current batch. Accordingly, as starting material for acidification experiments, we could focus on the three RNP-containing classes of virions: class I (spherical, with visible M1 matrix layer), class II (elongated, with visible M1 layer), and class III (spherical, lacking a visible M1 layer).
In these experiments, we distinguish particles according to two features: whether or not there is a resolved M1 layer (with this simplification, virions of classes I and II are grouped together [Table 1 and Fig. 3A to C]) and whether the array of glycoprotein spikes is “organized” or “disorganized.” In neutral-pH virions, the array is always organized in the sense that although the spikes do not form a crystalline lattice, they are close packed with an average center-to-center spacing of ~11 nm, and the distinctive “peanut” morphology of HA spikes and “mushroom” morphology of NA spikes are clearly defined. In disorganized arrays (of virions exposed to low pH), HA spikes, which form ~85% of the total, present mostly as blurred densities in which it is difficult to distinguish individual molecules.
To explore the effects of acid pH, we incubated the virus for 30 min at pH 4.9. Under these conditions, the virus is no longer fusogenic (23, 32) and HA should be in its postfusion state. In preparing these samples for microscopy, we encountered some aggregation, but many virions remained dispersed. Major changes were indeed observed in several respects (Fig. 3D to I and Table 1). (i) No elongated particles were seen, consistent with the work of Calder et al. (7). (ii) Approximately 20% of the particles were much larger, with diameters ranging from 145 to 200 nm (Fig. 4A to D), i.e., 1.5 to 2.0 times the average diameter for pH 7.4 virions (95 ± 14 nm) and corresponding to 2.2 to 4.0 times their average surface area. We conclude that they were produced by fusion of two or more virions (see also reference 7). (iii) All glycoprotein arrays were disorganized (Fig. 3D to I and and4A4A to D), leaving—in some virions—tracts of envelope seemingly devoid of spikes (arcs, Fig. 3E, H, and I and Fig. 4B to D). These bare patches were usually found at sites where there was no discernible M1 layer nor internal densities underlying the viral membrane (Fig. 3E and H and Fig. 4B and C; exceptions are shown in Fig. 3I and and4D).4D). These changes affect mainly the HA spikes, as normal NA spikes with their long-stemmed “mushroom” shape are readily apparent (Fig. 5). As at neutral pH (17), most NAs are in local clusters (Fig. 5H, J, and L), although single NAs can also be found (Fig. 5G, I, and K). As for HAs, their characteristic “peanut” shape assumed at pH 7.4 is seen only in a few patches (Fig. 5B to D). Not all HA spikes on a given virion have the same morphology (Fig. 3F and I), and many HAs are difficult to identify as individual molecules (Fig. 3D to I, I,4A4A to D, and and5E5E and F). (iv) An M1 layer is no longer seen in the majority (~60%) of virions and is complete in only 20% (e.g., Fig. 3D and and4A)4A) and partial in another 20% (e.g., Fig. 3E and F). (v) Most virions (~90%) contained, instead of dispersed RNPs, a single coagulate at the periphery of the internal cavity, leaving large voids that account for most of its volume (Fig. 3E to I and and4B4B to D). In many cases, this coagulate also contains M1 protein from the envelope-lining layer (see below). Here, we use the term “coagulate” instead of “aggregate,” because the observed densities were homogeneously dense, lacking the local gaps that we would expect to see if RNPs simply aggregated but retained their structures (see Fig. 7). Of note, dispersed RNPs, when observed, were found only in virions with a complete M1 layer (Fig. 3D and and44A).
Next, we sought to capture fusion intermediates. In order to identify conditions under which a substantial fraction of virions and HA molecules would have embarked on fusion but not yet reached the postfusion state, we referred to studies in which fluorescence dequenching measurements were used to show that the fusion activity of X-31 virions was partly but not largely reduced by preincubation for 5 min at pH 5.0 (23). Accordingly, we shortened the incubation time at pH 4.9 to 5 min. We also tried 15 min of incubation, but as the 5-min incubation appeared more promising, we focused our efforts on it. The outcome (see below) appears to justify this choice of experimental conditions.
Some of these virions retained neutral-pH-like morphologies (Fig. 6A to C), others resembled pH 4.9/30-min virions (Fig. 6J to L), and still others had differently altered morphologies (Fig. 6D to I). Whereas at neutral pH all virions had organized glycoprotein arrays and all pH 4.9/30-min virions had disorganized arrays, pH 4.9/5-min virion preparations had roughly equal numbers of virions of the two kinds (Table 1). Moreover, whereas a large majority (80 to 90%) of neutral-pH virions have visible M1 layers and many fewer (~20%) pH 4.9/30-min virions have this feature, pH 4.9/5-min virions are again intermediate in this regard (~55%). Finally, RNP morphologies in pH 4.9/5-min virions were also intermediate with respect to those from neutral-pH virions (which all had dispersed RNPs) and pH 4.9/30-min virions (~90% of which had coagulates), with about equal numbers of particles exhibiting dispersed RNPs (~47%) and coagulated RNPs (~53%).
Like neutral-pH virions, the populations observed after 5 min or 30 min of acidification are heterogeneous. Nevertheless, the data allowed us to draw the following conclusions as to the course of events. We assume that pH 4.9/5-min virions with a resolved M1 layer came from neutral-pH virions that also had this feature. Similarly, pH 4.9/5-min virions lacking an M1 layer but with organized spikes should come from their neutral-pH counterparts. On the other hand, the preponderance of matrix layer-negative virions after 5 min at pH 4.9, compared with those at neutral pH (45% versus 11%), argues that some of them originally had an M1 layer. This trend—a stripping or at least a major redistribution of the M1 layer—carries further in pH 4.9/30-min preparations (~80% versus 11%). As for pH 4.9/5-min virions lacking the M1 layer and with disorganized spikes, we propose that those with relatively little internal density (27% of these virions, e.g., in Fig. 6J) derive from virions that lacked an M1 layer, while the rest (73%, e.g., in Fig. 6K and L) derive from virions that formerly had an M1 protein layer and in which the RNPs and the M1 protein have coagulated together. According to these assignments, 78% of pH 4.9/5-min virions came from virions containing M1 layers, matching reasonably well the estimated ~90% in the initial population. The inferred courses of events for different kinds of virions, when transferred to pH 4.9, are sketched out in Fig. 10 and discussed further below.
As M1 is proposed, during an infection, to dissociate from the RNPs soon after they are released into the cytoplasm (20, 27, 28), we considered the possibility that M1 and RNPs might coagulate separately but rejected this idea because we did not observe virions to have two separate internal “lumps” (Fig. 7).
In order to probe for possible conformational changes in the HA spikes, we extracted sets of HA-containing subtomograms from the pH 4.9/5-min virions. These data were aligned, classified, and averaged independently within each class (Table 2). In Fig. 8A, sagittal sections through these density maps are compared with each other and with the average HA structure obtained from virions at pH 7.4 (38). All HAs from neutral-pH virions had the same conformation, i.e., they formed a single class. This structure shows distinguishable membrane-proximal (formed by HA2 and the N- and C-terminal segments of HA1) and membrane-distal (comprising residues 50 to 261 of HA1) domains, and it is ~15 nm long from the membrane surface to the outer tip and ~6 nm wide at the level of the membrane-distal domains, as measured from the averages. Some HAs (35%) on pH 4.9/5-min virions retained this conformation, but others exhibited two other structures that we call “state 1” (37%) and “state 2” (28%). In state 1, the membrane-proximal density of the peanut narrows and the “waist” between its two lobes is less evident, but the overall length of HA and the maximum width of the membrane-distal domain are very similar to those of the neutral-pH conformation. In state 2, however, the molecule is shorter (~13.5 nm) and wider (~7.5 nm) than in the neutral-pH structure and it is no longer bilobed. Density profiles along the axis of a state 1 HA and a state 2 HA (Fig. 8B) convey the difference in length between them. The percentage of HAs assigned to each subclass correlated with visual estimates of how each pH 4.9/5-min group was affected by the pH (Fig. 8).
Reversibility of the initial changes in HA at low pH has been shown by biochemical methods for HA expressed at the surface of eukaryotic cells (25) and by X-ray crystallographic analysis of purified HA (43). In order to test for reversibility in intact spikes on virions, we used cryo-ET to compare pH 4.9/5-min viruses with the same sample after a subsequent incubation for 45 min at pH 7.4. As illustrated in Fig. 9A, ~50% of the pH 4.9/5-min virions (from a total of 115) showed disorganized glycoproteins: this was as expected (see above). However, after the postincubation at pH 7.4, all virions (n = 150) exhibited an organized glycoprotein array (Fig. 9B), with most or all of the HAs reverting to their neutral-pH conformation (Fig. 9C to E), as confirmed by subtomogram averaging (inset in Fig. 9B).
The pleiomorphism of influenza virus long hampered detailed study by cryo-EM. Cryo-ET, however, has made possible three-dimensional (3D) structural analyses of individual virions (16). Here we employed this approach to investigate the structural changes undergone by influenza virions when switched to low pH. Although our initial focus was on the HA ectodomains, the observations also revealed major changes in virion size (from virion fusion) and shape (disappearance of elongated virions) and in the organization of internal material, i.e., coagulation of M1 protein and RNPs. It is noteworthy that—even within the same morphological class—all virions did not respond in the same way to the acidification treatments. The reason for this is unclear, although it could be due to differing amounts of the M2 ion channel protein in the viral membrane, which could lead to slower or faster decrease of the pH inside a virion. (Measurements of the average content of M2 per virion by two independent biochemical methods yielded values ranging from 14 to 68 molecules  but no information as to whether the M2 complement is fixed or variable on a given virion.)
We performed a morphological classification of pH 4.9/5-min virions and correlated these classes with those seen in the starting material at neutral pH and with pH 4.9/30-min virions. A summary of inferred transition pathways is given in Fig. 10. In virions that initially have a resolved M1 layer, the low pH first affects the HA ectodomains, which become disorganized (Fig. 10A, step 1); then the RNPs condense onto the M1 layer (step 2); and finally an M1 layer is no longer seen (step 3). The internal material finishes up as a dense coagulate, with most of the space inside the viral envelope void of protein density. We infer that the coagulate usually consists of RNPs plus M1 protein. This inference is consistent with reports that RNPs are still associated with some M1 protein just after being released into the cytoplasm (5, 28). It is not clear whether the RNPs break down or retain their structures, but the continuous nature of the coagulated density suggests the former, i.e., we would expect to see gaps between RNPs that were simply aggregated. In the minority of virions that initially lack a matrix layer, the RNPs lose their neutral pH organization faster, as they condense onto the inner surface of the envelope (Fig. 10B, step 1) before finally coagulating (step 2).
M1 protein has been proposed to be a determinant in the formation of filamentous virions (33, 34). The shape of elongated or filamentous virions is presumably imposed by the membrane-lining lattice of M1 protein. On the other hand, the low-energy state of membrane vesicles corresponds to spherical morphology. It appears likely that the disappearance of elongated virions at low pH corresponds to their converting to spherical shape. In the same vein, the sphericity of the giant particles produced by virion-virion fusion (Fig. 4) presumably results from remodeling from the hourglass morphology that would initially be realized. In both instances, spherical morphology is likely to result from relaxation of M1-envelope interactions as well as M1-M1 interactions (18) in response to low pH.
Since the characterization by X-ray crystallography of the neutral- and low-pH forms of HA (2, 6, 42), there has been progress—albeit still far from complete—toward understanding the complex process of virally induced membrane fusion. Here, subtomogram averaging has allowed us to address the effects of low pH on HA and thus to detect two intermediate conformations on the fusion pathway (Fig. 11). Both changes are reversible. Although the resolution is limited to 2.5 to 3.0 nm, the changes are large enough to be clearly demonstrated. The limited resolution also constrains the detail in which they may be interpreted. Thus, the reversible movement of the B loop to a relaxed form and the deformation of the HA1 subunit described by Xu and Wilson (43) that probably precede the state 1 and state 2 conformations are too subtle to be detected. With that caveat, we suggest the following: upon exposure to low pH, the fusion peptide and part of HA2 corresponding to approximately residues 24 to 54 (shown in yellow and green, respectively) move outwards to reach the state 1 conformation (Fig. 11B and E). This would expose both the fusion peptide and the kinked loop (residues 106 to 112 of HA2, in purple) (25). Next, the HA1 domains (blue) move downwards and outwards to reach the state 2 conformation (Fig. 11C and F) without “unclamping” the B loop. The existence of reversible intermediate conformations of HA could increase the success rate of the fusion process by allowing HA to have multiple attempts to reach the cellular membrane.
The rearrangement of HA1 proposed here differs from that which takes place as the molecule transitions, irreversibly, into the postfusion conformation (9, 15). After state 2, although HA is presumably still present in the viral envelope, it is hard to make out individual molecules in the tomograms: this appears to reflect the dissociation of HA1 membrane-distal domains and their dispersal, albeit still connected by flexible linkers to the core of the trimeric molecule.
These observations bear on the events that take place after a virion has entered an endosome, leading to release of viral RNPs into the cytoplasm. We envisage that, upon exposure to low pH, HA undergoes the reversible conformation changes described above. As the pH in the endosome drops, it is accompanied by an M2-mediated pumping of protons into the virion (31), eliciting a change in intermolecular interactions in the M1 layer and weakening its binding to the viral membrane. These changes render the envelope more pliable and hence more fusion-compatible (as hypothesized by Lee ). At the same time, disruption of M1-RNP interactions would free the RNPs to be discharged into the host cell cytoplasm after membrane fusion (20). The dense nature of coagulates seen after 30 min at pH 4.9 suggests that the RNPs trapped in them may no longer be infectious. Accordingly, we propose that soon after the M1 layer has been disrupted, HAs should complete their conformational changes and induce fusion in order to avoid coagulation of the RNPs.
This research was supported by the Intramural Research Program of the National Institute of Arthritis and Musculoskeletal and Skin Diseases of the National Institutes of Health.
Published ahead of print 18 January 2012