|Home | About | Journals | Submit | Contact Us | Français|
The M1 matrix protein of influenza A virus, which plays multiple roles in virion assembly and infection, underlies the viral envelope. However, previous studies have given differing accounts of the number of layers in the M1-envelope complex and their thicknesses and compositions. To resolve this issue, we performed cryo-electron microscopy and cryo-electron tomography on the self-same specimens. At neutral pH, there were two kinds of complexes, corresponding to a lipid bilayer with embedded glycoproteins, with and without a closely associated, 4-nm-thick sheet of M1 protein. The reported discrepancies arose from differences in imaging conditions, i.e., in defocus and in whole-particle projections versus thin tomographic slices. Exposure of virions to low pH (as in the endosome) promotes membrane fusion, and previous work has shown that the M2 ion channel causes the virion interior to acidify also. We found that after 5 min at pH 4.9, the proportion of virions lacking an M1 layer increased from 10% to 50%. In virions retaining an M1 layer under these conditions, the M1-envelope complex exhibited two states, viz, the original, neutral-pH state and one in which the M1 layer appeared thinner and/or closer to the membrane. These observations extend previous indications that acidic pH causes the M1 layer to dissociate, leaving the envelope more pliable and, consequently, fusion compatible, and they show that dissociation is preceded by a conformational change in M1.
The envelope of influenza A virus has three integral proteins (hemagglutinin [HA], neuraminidase [NA], and the M2 ion channel) embedded in a lipid bilayer derived from the plasma membrane of the host cell. (In the present context, we use the term “envelope” to refer to the viral membrane together with the viral transmembrane proteins and “M1-envelope complex” to refer to the envelope and the M1 matrix layer.) Underlying the viral envelope, the M1 protein forms a matrix layer that is thought to mediate interactions between the envelope and the viral ribonucleoproteins (RNPs). M1 lacks a membrane-targeting signal (1), but its interaction with the cytoplasmic tails of the transmembrane proteins recruits M1 to the plasma membrane (2, 3). Furthermore, M1 has an inherent membrane-binding ability which allows the protein to interact with membrane lipids, even when it is expressed in the absence of other viral proteins (4).
The M1 protein has been assigned roles in virion assembly (5), membrane fusion (6), and unpacking of the viral genome (7). Accordingly, it has been the object of several imaging studies by cryo-electron microscopy (cryo-EM), a technique in which native virion morphology is well preserved. However, the scope of these studies has been limited by virion pleomorphism, which rules out the enhancement through averaging that is achieved in reconstructions of icosahedral viruses, and direct interpretation of micrographs is subject to ambiguity arising from the fact that features at all levels in the ~100-nm-thick virions are coprojected. The latter problem is alleviated in cryo-electron tomography (cryo-ET), which yields three-dimensional density maps of individual virions. However, tomogram resolution is anisotropic and limited in the best-resolved slices of single virions to 4 to 5 nm (8).
Under these circumstances, diverging accounts have emerged with respect to the M1-envelope complex for virions at neutral pH: (i) cryo-EM studies with a high-resolution instrument (9) described the M1 layer as an ~7.2-nm-thick layer underlying a thin line of density which was interpreted as a single lipid leaflet (i.e., it was proposed that influenza virions containing a matrix layer do not have a lipid bilayer); (ii) virions imaged on a microscope equipped with a Zernike phase plate (10) also showed a thin dense line surrounding a thicker one (total thickness, ~12 nm), but these investigators proposed that the M1 layer underlies a lipid bilayer whose second leaflet is not resolved from the M1 layer; and (iii) cryo-ET studies (11–13) have described the M1-envelope complex as two well-resolved layers of similar thicknesses and densities, with a center-to-center spacing of ~5.5 nm. All these studies concurred that a subpopulation of virions (5 to 10%) appears to lack a discernible M1 layer.
Exposure of virions to low pH in the endosome leads to membrane fusion and release of the RNPs into the cytoplasm. One consequence of acidic conditions is to trigger major conformational changes in the fusogenic HA glycoprotein (14–16). However, the virion interior (inside the membrane) is also acidified by the passage of protons through the M2 ion channel, a process that can be blocked by the drug amantadine (7, 17, 18). Acidification has been correlated with release of the RNPs (19, 20). More recently, cryo-ET has been used to show another effect of acidification: detachment of the M1 layer from the viral membrane (11, 12). (In the same vein, earlier observations on the use of detergents to disrupt influenza virions, releasing M1, found this process to be promoted by low pH [21, 22].) Assuming that at neutral pH the M1 layer stabilizes the viral membrane, its detachment would be expected to render the membrane more pliable, and hence conducive to fusion. On the other hand, it has been suggested that M1 stabilizes the membrane to promote fusion, at least in the early stages (6).
The goals of the present study were (i) to clarify the conflicting accounts of the M1-envelope complex and (ii) to establish how many distinct states of this complex exist at both neutral and acidic (4.9) pHs. We hypothesized that at least some of the reported differences might arise from differing imaging conditions, in particular the degree of defocus. In the phase-contrast mode normally used, certain spatial frequencies are accentuated or repressed according to the degree of defocus (23). (With Zernike phase-plate imaging, this spectral bias is much reduced .) The defocus value typically used for cryo-EM, ~2 μm, is seldom used in cryo-ET, as it gives an adverse signal-to-noise ratio. The defocus value typically used in cryo-ET, ~5 μm or more, is seldom used in cryo-EM because it limits the resolution. We performed cryo-EM and cryo-ET under both conditions. To reduce the impact of virion heterogeneity on the comparison, the same fields of virions were imaged by both methods. First, a focal pair of cryo-micrographs was recorded; a tilt series was then recorded at the same defocus level, and a tomogram was calculated. This analysis was performed for virions at pH 7.4. To provide references of bona fide lipid bilayers, we mixed unilamellar liposomes of similar diameter with the virions. Subtomogram classification and averaging were subsequently performed on pH 7.4 virions and on virions incubated for 5 min at pH 4.9 (12, 25), to afford more-precise measurements of the thicknesses of the laminar components and the distances between them and to probe for alternative conformations of the envelope-matrix complex.
Influenza virus strain X-31 (A/Aichi/68 [H3N2]) was purchased from Charles River Laboratories (North Franklin, CT). To produce liposomes, lipid mixtures of phosphatidylcholine (Avanti), phosphatidylethanolamine (Avanti), sphingomyelin (Avanti), and cholesterol (Sigma), at a molar ratio of 1:1:1:1.5, were dried down with N2, lyophilized overnight, and resuspended in HM buffer (20 mM HEPES, 20 mM morpholineethanesulfonic acid [MES], 130 mM NaCl, pH 7.5) to a 20 mM lipid concentration (26). The liposome suspension was then freeze-thawed five times, extruded 30 times through a 100-nm filter, and stored at 4°C for less than 1 week before use.
Grids bearing thin films of influenza virion suspensions at neutral pH or after 5 min of incubation at pH 4.9 were vitrified as previously described (12). In experiments involving liposomes, vitrification was performed immediately after liposomes were mixed with the virions. Cryo-EM projections and single-axis tilt series were recorded on a Tecnai-12 transmission electron microscope (TEM; FEI) operating at 120 keV and equipped with an energy filter (GIF 2002; Gatan, Warrendale, PA). The filter was operated 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 (27). Cryo-micrographs were acquired at an effective magnification of ×53,600 (0.56 nm/pixel) or ×65,300 (0.46 nm/pixel), and single-axis tilt series were acquired at ×38,500 (0.78 nm/pixel) or ×53,600 (0.56 nm/pixel). When cryo-EM and cryo-ET were performed on the same field, the combined electron dose for cryo-EM was ~8 e−/Å2 (two projections were acquired, the first at a 2-μm nominal defocus level and the second at a 5-μm nominal defocus level), and the electron dose per tilt series was ~70 e−/Å2 (angular range of −66° to +66°; images were recorded each 2° and at ~1 e−/Å2 per image; the nominal defocus level was 2 or 5 μm). This kept the total electron dose below 80 e−/Å2. When no tilt series was acquired, the electron dose per projection was ~5 e−/Å2, keeping the total dose below 10 e−/Å2 (2- and 5-μm nominal defocus projections were acquired). According to the two-dimensional noise-compensated leave-one-out (NLOO-2D) method (28), the in-plane resolution of the tomograms was ~5.5 nm for individual virions.
The nominal defocus values, 2 μm and 5 μm, are quoted in Results and in figures. Actual defocus values were slightly off (e.g., the 5-μm nominal defocus values ranged from 4.3 to 5.2 μm), but not enough to affect the conclusions. These values, given in the figure legends, were determined using the Bsoft package (29, 30) by calculating the power spectrum of a given micrograph and fitting a theoretical contrast transfer function (CTF) to it. For a tilt series, the defocus value was estimated from the defocus of the zero-tilt projection, although there is a focal gradient across any tilted projection: for tilt series acquired at a 4-μm nominal defocus value, the calculated defocus value was ~5 μm; for tilt series acquired at a 2-μm nominal defocus value, the calculated defocus value was ~2.5 μm.
We recorded 6 sets of focal pairs, each followed by a tilt series, of virions mixed with liposomes at neutral pH (giving ~100 virions); 6 tilt series of virions at neutral pH (giving 44 virions with a resolved M1 layer and 11 without); 31 tilt series of virions at pH 4.9 for 5 min (179 virions, of which only those with organized glycoprotein arrays were analyzed, giving 61 virions with a resolved M1 layer and 25 virions without); and ~100 focal pairs of virions at neutral pH (>750 virions).
Tomogram reconstruction, averaging, and computational modeling were done using Bsoft (29, 30), unless stated otherwise. Subtomogram alignment used Bsoft routines, modified as needed and wrapped into Python scripts, as described previously (12). The model used as an initial reference consisted of two sheets of density representing the M1-envelope complex, with or without a cylinder representing the HA spike. After completing the iterations, particles (subtomograms, each containing an HA molecule and underlying density) whose orientation differed by more than 45° from the initial one were discarded, and the remaining particles were classified using maximum likelihood methods as implemented in Xmipp (31), to exclude outliers. Cylindrical symmetry was applied to the average of the remaining subtomograms (~50% of the initially selected particles). To exclude model bias, a second refinement was performed for each class, using the resulting average from the first refinement as an initial reference. Each refinement involved 5 to 7 iterations of alignment and averaging and a final classification of the subtomograms. In terms of the Fourier shell correlation coefficient (32), the resolution of the subtomogram average was ~3.0 nm (threshold of 0.3) or ~3.5 nm (threshold of 0.5). The number of particles was 2,099 for M1-containing virions at neutral pH, 545 for M1-lacking virions at neutral pH, 3,035 for M1-containing virions held at pH 4.9 for 5 min, and 1,086 for M1-lacking virions held at pH 4.9 for 5 min.
Simple 2D models were generated for cross-sections of the M1-envelope complex in the native and tight conformations. The membrane was represented as two lines of density 4 nm apart, and the M1 layer was represented as a single sheet of uniform density. There were only two free parameters: the thickness of the M1 layer and its offset from the membrane. Various thicknesses in the range of 2.5 to ~4.5 nm and offsets of 0 to ~3.5 nm were tried. Cryo-EM imaging conditions at 2-μm and 5-μm defocus values were simulated by adding random noise and by weighting the model's amplitudes to match the micrographs shown in Fig. 1A. Finally, averaged density profiles were calculated and compared with the experimental data.
Influenza virus strain X-31 virions are mostly spherical. Visualized by cryo-ET, ~90% of them have—apart from the protruding spikes—two well-resolved peripheral layers of approximately equal densities, while ~10% have one such layer (13). These layers have been interpreted as the viral membrane with and without a lining of M1 protein. If we consider a typical virion, of ~100 nm in diameter, to have approximately 300 HA trimers and 50 NA tetramers (13), with each subunit having a single transmembrane segment of ~1 nm in diameter and 10 M2 channels, each ~3 nm in diameter (33, 34), then these protein components should occupy ~6% of the hydrophobic phase of the membrane, with lipid hydrocarbon chains comprising the remaining 94%. Thus, the viral membrane should appear very much like a lipid bilayer. To test this prediction, we prepared unilamellar liposomes, mixed them with virions, and imaged the mixture by both cryo-EM and cryo-ET (Fig. 1). Specimens were vitrified immediately after mixing to minimize the possibility of interactions between the two components.
In cryo-micrographs recorded at a 2-μm defocus level, the liposome periphery is clearly resolved into two thin lines which are ~4 nm apart from center to center (Fig. 1A1, left panel). This is as expected for these imaging conditions, taking the lines to represent the head groups of the two leaflets. The membranes of virions putatively lacking M1 are indistinguishable from liposomes in this respect (Fig. 1A2, left panel). On the other hand, the peripheral structure of virions putatively having an M1 layer presents as one thin line of density, like the two seen on liposomes, overlying a thicker one (Fig. 1A3, left panel). The two layers have a combined thickness of ~10 nm (full width at half-maximum) and a center-to-center distance between them of ~7.5 nm. In second exposures of the same particles at a 5-μm defocus level, the two leaflets of the bilayer are no longer resolved (Fig. 1A1 and andA2,A2, middle panels). This is a consequence of the higher defocus level, not of incremental radiation damage, as resolved leaflets persisted until at least the fifth exposure if the defocus level was not changed (data not shown). On putatively M1-containing virions, the periphery still presents as two layers, but the difference in thickness between them is less evident (Fig. 1A3, middle panel).
After recording of the two micrographs, a tomographic tilt series was collected from the same field at a defocus level of ~5 μm per projection, and a tomogram was calculated. Thin (~0.6 nm) central slices of the three particles described above are shown in Fig. 1A1 to toA3,A3, right column. The membrane of the liposome and the envelope of the M1-lacking virion appear very much as in the more defocused micrographs (Fig. 1A1 and andA2,A2, middle column), although the latter are full projections. However, the internal structures (mainly RNPs), the protruding glycoprotein ectodomains, and a small liposome inside the larger one are more clearly visualized in the tomogram. The periphery of the M1-containing virion (Fig. 1A3, right panel) appears as two well-resolved and approximately equal layers of density, ~5.5 nm apart from center to center, as in previous studies (11–13).
A similar montage of three focal pairs and corresponding tomographic slices is presented in Fig. 1B1 to toB3.B3. However, in this experiment, the defocus level of the tomogram (Fig. 1B1 to toB3,B3, right column) was reduced to ~2 μm to match the first-exposure cryo-micrograph (Fig. 1B1 to toB3,B3, left column). In this case, the two leaflets of the bilayer are marginally resolved in places in the tomograms (marked with arrowheads in Fig. 1B1 and andB2,B2, right column), but the images are much noisier, with the glycoprotein spikes barely visible.
When envelopes of M1-containing virions are appraised in cryo-micrographs recorded at an ~2-μm defocus level, some of them appear to exhibit regions where the M1-envelope complex is resolved into three layers (Fig. 2D and andE).E). To reduce the noise level, we rotationally averaged images of 40 spherical virions (26 with the M1 layer and 14 without it). This made the three-layered structure more evident: indeed, ~40% of the M1-containing virions were found to present a three-layered envelope, with the two outer layers ~4 nm apart from center to center (Fig. 2D and andE).E). We concluded that M1-containing virions also have a lipid bilayer; however, the M1 layer is closely apposed, such that—depending on the imaging conditions (mainly, the defocus level)—the inner leaflet of the membrane is often not resolved from the M1 layer, and they appear together as a single thick layer.
To reduce the noise level for a closer comparison between virions that have the M1 layer and virions lacking it, we performed subtomogram classification and averaging (Fig. 3). In each such experiment, many small subvolumes were extracted from the tomograms, computationally aligned, classified, and averaged within each class. Each subvolume was centered on an HA spike, whose distinctive peanut shape is apparent (Fig. 1A2 and andA3,A3, right column), and data from virions with the M1 layer and from virions lacking it were processed separately. In the initial reference used for alignment, the HA spike was represented by a cylinder on top of a bilaminar envelope (Fig. 3A). The process converged after 5 to 7 iterations of alignment and averaging (depending on the experiment), and the subvolumes were classified. For both data sets (virions with and without an M1 layer), only one class was defined. To avoid possible bias from using the same model as an initial reference for both data sets, a second refinement was then performed, proceeding from the averages already obtained. Again, only one class per data set was found. In the averaged subvolume of M1-lined membranes, the inner layer emerged as appreciably thicker and denser than the outer layer (Fig. 3A). We call this neutral-pH conformation of the M1 layer the “native” conformation. Note that the density of the outer layer is markedly fainter than that of the single layer of the M1-lacking envelope (Fig. 3A and andB,B, cf. “Native” and “No M1” data), which, as shown above, is indistinguishable from a lipid bilayer (but is not resolved into two lines of density due to the high defocus level employed in acquiring the tomographic tilt series). We concluded that the outer density layer of the M1-envelope complex corresponds to the outer leaflet of the lipid bilayer and that the inner layer corresponds to the inner leaflet plus the layer of M1 protein.
In the procedure outlined above, the HA densities were well defined and could be used to calibrate the envelope-related densities consistently between classes. However, as an additional check against model bias, we repeated the experiment, but using another initial reference, lacking the HA cylinder (Fig. 3C). We reasoned that alignment of the subvolumes in the dimension normal to the membrane would then be dominated by the laminae and perhaps be more precise, allowing better measurements of layer thicknesses and distances between layers. However, the outcome was essentially unchanged (Fig. 3C and andD,D, “Native” and “No M1” data), except that, in this experiment, the HA densities were smeared out laterally in the class averages.
Previously, we used cryo-ET to observe the structural changes that X-31 virions undergo on being switched to pH 4.9, focusing mainly on the HA spikes (12). After 5 min at pH 4.9, there was also a marked increase, from ~10% to ~50%, in the fraction of viral particles lacking an M1 layer. This implies that almost half of the virions that had an M1 layer at neutral pH shed it. In the present study, we refined the subtomogram averaging procedure to include the envelope-matrix complex in the classification process in order to probe for multiple conformational states of the M1 layer. This analysis was confined to virions whose envelopes retained an organized glycoprotein array. (An array is considered “organized” [12, 35] when the spikes are clearly visualized as having the distinctive “peanut” [HA] and “mushroom” [NA] morphologies. In contrast, in “disorganized” arrays produced by extended exposure to low pH, the spikes are perceived as blurred densities in which it is difficult to distinguish individual molecules.)
As shown in Fig. 3, subvolumes of M1-lined envelopes of virions held at pH 4.9 for 5 min fell into two classes. One of them (“native” conformation; ~75% of the data) was indistinguishable from the corresponding average for neutral-pH virions. In the other (“tight” conformation; ~25% of the data), the M1-envelope complex was barely resolved into two layers, with an appreciably smaller combined thickness than in the native conformation (~8.5 nm versus ~10 nm). A closer association of M1 with the viral membrane is also apparent in some areas of tomographic sections of virions held at pH 4.9 for 5 min, without averaging (Fig. 4, cages).
In our earlier study with subtomogram averaging (12), the classification procedure was based on the HA spikes, and the tight conformation was not detected. However, when the weighted averages of the native and tight conformations (75% and 25%) are compared with the earlier average for virions held at pH 4.9 for 5 min, with an M1 layer and an organized glycoprotein array, the two density profiles are entirely consistent (data not shown).
Because the complex with the tight conformation is significantly thinner than that with the native conformation (~8.5 nm versus ~10 nm), it follows that the M1 layer has to be thinner and/or closer to the membrane, assuming that the membrane is not appreciably changed. To make this inference more precise, we computed simple models of the M1 layer as a sheet of uniform density underlying the bilayer, varying only two parameters—its thickness and its separation from the membrane. A grid of values in the range of 2.5 nm to 4.5 nm were tried for thickness (which should be about 4 nm ), with values of 0 nm to 3.5 nm for separation. The corresponding imaging conditions were applied, and the resulting models were compared with the experimental images. We concluded that the observed conformational change (native to tight) reflects the M1 layer becoming ~1 nm thinner or moving ~0.5 nm closer to the membrane, or partial contributions from both effects (Fig. 5).
As influenza virions bud from the plasma membrane of the infected cell, the M1 protein mediates interactions between the nascent envelope and the viral RNPs. As M1 is recruited to assembly sites, it packs into a membrane-lining layer (37). Our first goal in this study was to resolve inconsistencies in previous accounts of the M1-envelope complex. Our second goal was to investigate how this complex responds to low pH, a condition that infecting virions encounter in the endosome. Low pH induces fusogenic conformational changes in the HA glycoprotein, which have been studied extensively (14–16). It has also been shown that low pH acidifies the virion interior (7, 17, 18, 38, 39), facilitating fusion (38, 39), but the structural consequences of acidifying the virion interior have not been clear.
Appraisal of influenza virions by cryo-EM and cryo-ET is complicated not only by pleomorphy of their sizes and shapes but also by heterogeneity in their internal structures. Our previous observations (12, 13) and those of others (6, 9–11) indicated that some X-31 virions lack an M1 layer. However, the apparent thickness and number of peripheral layers can be altered substantially by different imaging conditions (see above). In this context, our experiments with liposome-virion mixtures provide conclusive evidence that there is indeed a subpopulation of virions that lack an M1 layer. They also confirm that the envelopes of most virions are lined with a layer of M1 protein. This layer is so thin (~4 nm) that given the molecular dimensions in the crystal structure (36), it should be a monolayer. The M1 layer is closely apposed to the membrane, such that, in cryo-electron micrographs (which are projection images), it usually merges with the inner leaflet into a single layer of density. Occasionally, under optimal conditions of defocus, ice thickness, etc., the complex is marginally resolved into three layers (outer leaflet, inner leaflet, and M1 layer), and this resolution is confirmed in rotationally averaged density profiles in which noise is suppressed (Fig. 2D and andE).E). In cryo-ET slices, the conspicuous separation of the envelope into two layers of density (outer leaflet and inner leaflet plus M1 layer) that are 5.5 nm apart is somewhat exaggerated by the high defocus values used (~5 μm), which give a CTF maximum at spacings of about 5.9 nm, accentuating features in that size range.
The M1 layer is attached to the membrane via multiple interactions with the endodomains of the transmembrane proteins (4, 40). Individually, these interactions may be weak, but collectively, they amount to a high-avidity association because the layer is also consolidated laterally by M1-M1 interactions. As yet, no periodicity has been detected in the M1 stratum of X-31 virions, but the packing arrangement is likely to be locally similar to that in the helical lattice that underlies the envelope in elongated virions of the Udorn strain (11).
At neutral pH, 90% of X-31 virions have an M1 layer and 10% lack this layer. After 5 min at pH 4.9, these numbers switch to 50% with an M1 layer and 50% without. Assuming that the populations are randomly sampled, it follows that two new kinds of particles are present. First, most (80%) of the virions lacking an M1 layer are particles whose M1 dissociated after acidification. As we did not see the sheets of density inside these particles that would be expected if the M1 layer were simply to peel off the membrane, it follows that the M1 molecules dissociate from each other as well as from the membrane. This inference is consistent with the finding that M1 oligomers dissociate into dimers at pH 5 in vitro (41). Second, in ~25% of the virions that retain an M1 layer, its structure is significantly altered in extended patches, where we observe the “tight” conformation. The observed change is a manifestation of a conformational change in M1 that appears to be a prelude to dissociation of the M1 layer (Fig. 6). Since the structures of the N-terminal domain are very similar at neutral and low pHs (36, 42, 43), the observed change might reflect a disorganization of the C-terminal domain, which is proposed to be distal to the viral membrane (36, 43), or changes in the M1-M1 stacking interface (36, 42). Functionally, the pH-dependent conformational change of M1 that we observed might facilitate virus fusion by rendering the influenza virus envelope more pliable, and hence conducive to fusion.
We thank Dennis Winkler for support with resources for cryo-electron tomography and for insightful discussions, J. Bernard Heymann for advice on data analysis, and Matthew Brecher for providing the liposomes used in this study.
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 6 March 2013