Membrane curvature sensing and symmetry breaking of the M2 proton channel from Influenza A

  1. James Lincoff
  2. Cole VM Helsell
  3. Frank V Marcoline
  4. Andrew M Natale
  5. Michael Grabe

  • Curated by eLife

    eLife logo

    Evaluation Summary:

    The authors combined atomistic simulations and continuum mechanics models to probe how structural features of the M2 channel impact the local membrane properties and stability of the channel in membranes of different curvatures. The insights gained in this work can potentially lead to novel strategies that screen for drug molecules that stabilize fission-incompetent conformations of the M2 channel. The multi-scale computational approach will find utility to many problems in membrane reshaping.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #2 and Reviewer #3 agreed to share their name with the authors.)

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

The M2 proton channel aids in the exit of mature influenza viral particles from the host plasma membrane through its ability to stabilize regions of high negative Gaussian curvature (NGC) that occur at the neck of budding virions. The channels are homo-tetramers that contain a cytoplasm-facing amphipathic helix (AH) that is necessary and sufficient for NGC generation; however, constructs containing the transmembrane spanning helix, which facilitates tetramerization, exhibit enhanced curvature generation. Here, we used all-atom molecular dynamics (MD) simulations to explore the conformational dynamics of M2 channels in lipid bilayers revealing that the AH is dynamic, quickly breaking the fourfold symmetry observed in most structures. Next, we carried out MD simulations with the protein restrained in four- and twofold symmetric conformations to determine the impact on the membrane shape. While each pattern was distinct, all configurations induced pronounced curvature in the outer leaflet, while conversely, the inner leaflets showed minimal curvature and significant lipid tilt around the AHs. The MD-generated profiles at the protein–membrane interface were then extracted and used as boundary conditions in a continuum elastic membrane model to calculate the membrane-bending energy of each conformation embedded in different membrane surfaces characteristic of a budding virus. The calculations show that all three M2 conformations are stabilized in inward-budding, concave spherical caps and destabilized in outward-budding, convex spherical caps, the latter reminiscent of a budding virus. One of the C2-broken symmetry conformations is stabilized by 4 kT in NGC surfaces with the minimum energy conformation occurring at a curvature corresponding to 33 nm radii. In total, our work provides atomistic insight into the curvature sensing capabilities of M2 channels and how enrichment in the nascent viral particle depends on protein shape and membrane geometry.

Article activity feed

  1. Version published to 10.7554/elife.81571 on eLife
  2. eLife

    Author response:

    Reviewer #2 (Public Review):

    I have two significant concerns that I believe can be resolved on the timescale of review.

    1. The work identifies substantial thinning in one leaflet. Lipids expand as they thin. Given this, are there too few lipids in this leaflet (which would also indicate thinning)? I would expect their deformations depend strongly on the number-balance of lipids in each leaflet. The authors should check if thinning, and the boundary, is sensitive to inter-leaflet-lipid imbalance.

    We thank Reviewer #2 for this insight, as it led us to evaluate the leaflet tensions in our restrained 2L0J simulation. We found there was an imbalance in the leaflet packing, which we addressed with an extensive set of new simulations and new analysis aimed at generating balanced leaflets.

    See Page 6-8, Appendix Section 1, Appendix – figures 1, 2. We discuss these findings in the new Results section “Protein footprint asymmetry can lead to differential leaflet stresses” and accompanying appendix. Many of the bilayer features in the repacked simulations are consistent with our original submission, but not all. For instance, while we continue to see large tilt immediately around the amphipathic helices in the lower leaflet and little in the upper leaflet, tilts in both leaflets decay to similar values at the box edge (Appendix - figure 2). The degree of membrane pinch along the membrane-protein contact boundaries are less sensitive to the leaflet packing, as demonstrated by the surface heights (Appendix - figure 1).

    Determining the proper change in leaflet count is quite difficult. We are actively extending our continuum model to address questions of differential leaflet strain and coupled lipid tilt, which may allow us to estimate changes in leaflet-count, but this is a significant undertaking beyond the scope of this resubmission.

    1. By constraining the pore to have 2-fold symmetry, the authors remove a large entropic penalty disfavoring such a conformation, and thus presumably disfavoring the negative- gaussian-curvature it induces. For example, if the free energy surface for the fluctuations were rather flat, and only 1% of the conformations were consistent with 2-fold symmetry, the coupling to NGC may be reduced by -kT log( 1 % ), neglecting enhancement by coupling to NGC. Therefore, I predict that the coupling to NGC would be reduced further were the constraint removed.

    We agree with the reviewer that if the 2-fold states are highly disfavored for entropic or enthalpic reasons, it would directly reduce the coupling to NGC. However, we don’t know the free energy difference between these states, and it is hard to calculate them from all-atom and beyond our current scope. While our unrestrained simulations are not converged, they demonstrate that there is a wide range of orientations for the amphipathic helices that are energetically accessible (see Figure 2, Appendix Section 1, and Appendix - figure 4). Still, the DEER data from the Howard lab (Kim et al., 2015) would be better described by further symmetry-broken states with greater inter-AH distances, suggesting that such conformations are not well represented in our equilibrium ensemble.

    Reviewer #3 (Public Review):

    Helsell et al. uses atomistic molecular dynamics simulations to characterize the structural dynamics of the M2 protein together with continuum elastic models to evaluate the energetic cost of the protein-induced bilayer deformations. Using unbiased simulations (without constraints on the protein) they show that the M2 structure is dynamic and that the AH helices are mobile (though they tend to retain their secondary structure), in agreement with experimental observations. Then, using simulations in which the peptide backbone was restrained to the starting structure, they were able to quantitatively characterize the protein- induced bilayer deformations as well as the acyl chain dynamics.

    Both the atomistic simulations and the continuum-based determinations of the bilayer deformation energies are of high quality. The authors are careful to note that their unbiased simulations do not reach equilibrium, and the authors' conclusions are well supported by their results, though some issues need to be clarified.

    1. P. 7: Choice of lipid composition: POPC:POPG:Cholesterol 0.56:0.14:0.3. This lipid composition (or POPC:POPG 0.8:0.2) has been used in a number of experimental studies that the authors use as reference. It differs, however, substantially from the lipid composition of the influenza membrane (Gerl et al., J Cell Biol, 2012; Ivanova et al., ACS Infect Dis, 2015), which is enriched in cholesterol, has a 2:1 ratio of phosphatidylethanolamine to phosphatidylcholine, and almost no PG. The choice of lipid composition is unlikely to impact the authors' major conclusions, but it should be discussed briefly. As noted by Ivanova et al., the lipids of the influenza membrane are enriched in fusogenic lipids. How will that impact the authors results.

    As noted by the Reviewer, the lipid composition we explored was based on DEER studies from Kathleen Howard. While there is a lot of cholesterol in our simulations, it is lower than the lipidomics papers suggest for the viral membrane (Gerl et al., 2012; Ivanova et al., 2015). We hypothesize that further increasing cholesterol would stiffen the membrane even more and cause the energy differences we report here to become even larger – accentuating our finding. We employ 14% POPG and the Simons lab finds about 14% PS. Chemically these headgroups are similar, but the size and spontaneous curvature difference could be a concern. This is the the different intrinsic curvatures of PE versus PC. However, we have not considered spontaneous curvature in our continuum calculations, so we cannot predict how this will influence our results.

    See Appendix - figure 6. We added a new panel to this figure with continuum parameters intended to mimic a high 50 % cholesterol membrane reported for viral coats, and we show that the curvature sensing of symmetry-broken states increases as the cholesterol content increases.

    See Page 25. We added text in the Discussion concerning the difference in lipids found in the virus versus those compositions employed in experiment and here.

    1. The definition of the lipid tilt needs to be revisited. On P. 13 (in the Pdf received for review, the authors do not provide page numbers), the tilt is defined/approximated as "the angle between the presumed membrane normal (aligned with the Z axis of the box) and the vector pointing from each phospholipid's phosphate to the midpoint between the last carbon atoms of the lipid tails." This (equating the normal to the interface with the Z axis of the simulation box) may be an acceptable approximation for the lower leaflet, which is approximately flat, but probably not for the upper leaflet where the interface is curved in the vicinity of the protein. The authors should, at least, discuss the implications of their approximation in terms of their conclusion that there is little lipid tilt in the upper leaflet.

    We agree that our lipid tilt calculations are approximate since we assume the membrane normal points along the z direction. We have now restated this assumption in the Results when we start to discuss tilt. Different models define lipid tilt in different ways, but the work of Deserno defines it with respect to the bilayer mid-plane which is a shared surface for the upper and lower leaflets. Thus, tilt would be moderately impacted in both leaflets. Examining the snapshots at the top of Figure 7, we surmise that the calculated tilts in both leaflets adjacent to the protein would be slightly reduced, leaving the values at the boundary unaffected. Thus, the upper leaflet likely experiences even less tilt than calculated.

    See Page 16. We have added the discussion above to the section on lipid tilt. Also, we have added page numbers to the resubmission.

    1. P. 14, last paragraph, Figure 5 and 6: The snapshots in Figure 5 are too small to see what the authors refer to when they write "tilt their lipid tails to wrap around the helices." The authors should consider citing the work of H W. Huang, e.g., Huang et al. (PRL, 2004), who introduced the notion of curvature stress induced by antimicrobial peptides, a concept similar to what the present authors propose.

    See Page 17. We have now drawn the connection between what our simulations are showing and the earlier work by Huey Huang on antimicrobial peptides.

    See Figure 7. To make the lipid deformations easier to see, we are attaching the full-size versions of each snapshot to the figure as supplemental data.

    1. P. 17-18, Figure 7: The authors introduce the bilayer midplane, which becomes important for the determination of the deformation energy in the (unnumbered) equation on P. 17, but do not specify how it is determined. This is a non-trivial undertaking, but critical for the evaluation of the deformation energy; please add the necessary details.

    See Pages 15 and 20. In the continuum model, we define CM (the compression surface) following the work of May and colleagues (and other groups) as the areal compression weighted mean of the upper and lower surface. In the MD simulation results in Figure 6, we define leaflet thickness as the absolute difference between the interpolated leaflet hydrophobic surface (calculated using the first carbon atoms of each POPC and POPG lipid tail) and the interpolated bilayer midplane surface (calculated as the average of the upper and lower leaflet tail surfaces, each interpolated based on the last carbon atoms of each POPC and POPG lipid tail for each leaflet, respectively). These two leaflet-based definitions are different, and a more sophisticated continuum model of the upper and lower leaflet coupling would require the incorporation of lipid tilt, which we do not currently have.

    1. P. 18-19, Figure 8: The comparison of the MD and continuum membrane deformations is very informative, but the authors should discuss the implications of the increased symmetry further in terms of the estimated deformation energies. (I do not believe the authors really mean that they predicted the energies, they estimated/approximated them.)

    The Reviewer is correct, we are not predicting the energies of the actual MD generated bilayers, but rather we are estimating the energies of these shapes using a continuum-based approximation. The good agreement between the MD generated surfaces and the continuum predicted surfaces suggested that the model is capturing the underlying physics. We argued that the increased symmetry of the continuum surfaces compared to the MD surfaces was due to incomplete sampling in the MD. We were right about that. Please see revised Figure 10 with new data and some longer simulations, where the symmetry in the MD is now apparent and the match between continuum and MD is even better. Frankly, we are very pleased with these new results.

    See Page 18 and Figure 10. We have changed language throughout moving away from “predicting” to “estimating”. The new MD generated data shows much greater symmetry reflected in the starting structures, and better agreement with model predictions.

    References

    Argudo, D., Bethel, N. P., Marcoline, F. V., Wolgemuth, C. W., & Grabe, M. (2017). New Continuum Approaches for Determining Protein-Induced Membrane Deformations. Biophys J, 112(10), 2159-2172. https://doi.org/10.1016/j.bpj.2017.03.040

    Bethel, N. P., & Grabe, M. (2016). Atomistic insight into lipid translocation by a TMEM16 scramblase. Proc Natl Acad Sci U S A, 113(49), 14049-14054. https://doi.org/10.1073/pnas.1607574113

    Drabik, D., Chodaczek, G., Kraszewski, S., & Langner, M. (2020). Mechanical Properties Determination of DMPC, DPPC, DSPC, and HSPC Solid-Ordered Bilayers. Langmuir, 36(14), 3826-3835. https://doi.org/10.1021/acs.langmuir.0c00475

    Ferreira, T. M., Coreta-Gomes, F., Ollila, O. H., Moreno, M. J., Vaz, W. L., & Topgaard, D. (2013). Cholesterol and POPC segmental order parameters in lipid membranes: solid state 1H-13C NMR and MD simulation studies. Phys Chem Chem Phys, 15(6), 1976- 1989. https://doi.org/10.1039/c2cp42738a

    Gerl, M. J., Sampaio, J. L., Urban, S., Kalvodova, L., Verbavatz, J. M., Binnington, B., Lindemann, D., Lingwood, C. A., Shevchenko, A., Schroeder, C., & Simons, K. (2012). Quantitative analysis of the lipidomes of the influenza virus envelope and MDCK cell apical membrane. J Cell Biol, 196(2), 213-221. https://doi.org/10.1083/jcb.201108175

    Henriksen, J., Rowat, A. C., Brief, E., Hsueh, Y. W., Thewalt, J. L., Zuckermann, M. J., & Ipsen, J. H. (2006). Universal behavior of membranes with sterols. Biophys J, 90(5), 1639- 1649. https://doi.org/10.1529/biophysj.105.067652

    Hossein, A., & Sodt, A. J. (2023). Membraneanalysis. jl: A Julia package for analyzing molecular dynamics simulations of lipid membranes. Journal of Open Source Software, 8(87), 5380.

    Hu, M., Briguglio, J. J., & Deserno, M. (2012). Determining the Gaussian curvature modulus of lipid membranes in simulations. Biophys J, 102(6), 1403-1410. https://doi.org/10.1016/j.bpj.2012.02.013

    Ivanova, P. T., Myers, D. S., Milne, S. B., McClaren, J. L., Thomas, P. G., & Brown, H. A. (2015). Lipid composition of viral envelope of three strains of influenza virus - not all viruses are created equal. ACS Infect Dis, 1(9), 399-452. https://doi.org/10.1021/acsinfecdis.5b00040

    Kim, S. S., Upshur, M. A., Saotome, K., Sahu, I. D., McCarrick, R. M., Feix, J. B., Lorigan, G. A., & Howard, K. P. (2015). Cholesterol-Dependent Conformational Exchange of the C- Terminal Domain of the Influenza A M2 Protein. Biochemistry, 54(49), 7157-7167. https://doi.org/10.1021/acs.biochem.5b01065

    Kučerka, N., Tristram-Nagle, S., & Nagle, J. F. (2006). Structure of fully hydrated fluid phase lipid bilayers with monounsaturated chains. J Membr Biol, 208(3), 193-202.

    Latorraca, N. R., Callenberg, K. M., Boyle, J. P., & Grabe, M. (2014). Continuum approaches to understanding ion and peptide interactions with the membrane. J Membr Biol, 247(5), 395-408. https://doi.org/10.1007/s00232-014-9646-z

    Liu, J., Kaksonen, M., Drubin, D. G., & Oster, G. (2006). Endocytic vesicle scission by lipid phase boundary forces. Proc Natl Acad Sci U S A, 103(27), 10277-10282. https://doi.org/10.1073/pnas.0601045103

    Pan, J., Tristram-Nagle, S., & Nagle, J. F. (2009). Effect of cholesterol on structural and mechanical properties of membranes depends on lipid chain saturation. Phys Rev E Stat Nonlin Soft Matter Phys, 80(2 Pt 1), 021931. https://doi.org/10.1103/PhysRevE.80.021931

    Rawicz, W., Olbrich, K. C., McIntosh, T., Needham, D., & Evans, E. (2000). Effect of chain length and unsaturation on elasticity of lipid bilayers. Biophys J, 79(1), 328-339. https://doi.org/10.1016/S0006-3495(00)76295-3

    Sun, D., Peyear, T. A., Bennett, W. F. D., Andersen, O. S., Lightstone, F. C., & Ingolfsson, H. I. (2019). Molecular Mechanism for Gramicidin Dimerization and Dissociation in Bilayers of Different Thickness. Biophys J, 117(10), 1831-1844. https://doi.org/10.1016/j.bpj.2019.09.044

    Tzlil, S., Deserno, M., Gelbart, W. M., & Ben-Shaul, A. (2004). A statistical-thermodynamic model of viral budding. Biophys J, 86(4), 2037-2048. https://doi.org/10.1016/S0006- 3495(04)74265-4

    Ursell, T. S., Klug, W. S., & Phillips, R. (2009). Morphology and interaction between lipid domains. Proc Natl Acad Sci U S A, 106(32), 13301-13306. https://doi.org/10.1073/pnas.0903825106

    Veatch, S. L., & Keller, S. L. (2003). Separation of liquid phases in giant vesicles of ternary mixtures of phospholipids and cholesterol. Biophys J, 85(5), 3074-3083. https://doi.org/10.1016/S0006-3495(03)74726-2

    Venable, R. M., Brown, F. L. H., & Pastor, R. W. (2015). Mechanical properties of lipid bilayers from molecular dynamics simulation. Chem Phys Lipids, 192, 60-74. https://doi.org/10.1016/j.chemphyslip.2015.07.014

  3. eLife

    Evaluation Summary:

    The authors combined atomistic simulations and continuum mechanics models to probe how structural features of the M2 channel impact the local membrane properties and stability of the channel in membranes of different curvatures. The insights gained in this work can potentially lead to novel strategies that screen for drug molecules that stabilize fission-incompetent conformations of the M2 channel. The multi-scale computational approach will find utility to many problems in membrane reshaping.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #2 and Reviewer #3 agreed to share their name with the authors.)

  4. eLife

    Reviewer #1 (Public Review):

    The authors probe the interaction between the influenza A M2 channel and lipid membrane using a multi-scale computational approach. Using extensive atomistic simulations and different starting structures, the authors are able to probe how various structural features of the M2 channel (e.g., symmetry and AH orientation) impact the properties of nearby lipids, such as bending and lipid tilt. The atomistic protein structures are then used together with a continuum mechanics model for the membrane to estimate the stability of the channel in membranes of different shape (e.g., positive or negative Gaussian curvature). While the energetic consequences were relatively modest, on the scale of a few kT, the results are consistent with experimental observation that M2 channel does not favor convex spherical caps. The symmetry-broken conformations were found to be stabilized by membrane of negative Gaussian curvature, which is important to the fission process. Therefore, the insights gleaned in this work can potentially lead to novel strategies that screen for drug molecules that stabilize fission-incompetent conformations of the M2 channel.

  5. eLife

    Reviewer #2 (Public Review):

    This paper by Helsell et al models the ability of influenza M2 proton channels to induce membrane bending, including the negative Gaussian curvature of budding viral necks. An extensive and clear introduction summarizes hypotheses for how M2 may favor pinching the viral neck, including as a linactant. Simulations first identify breakage of the four-fold crystallographic symmetry by M2 amphipathic helices. Following this, simulations of four-fold and two-fold restrained systems are generated. Four-fold symmetry is incompatible with negative Gaussian curvature, whereas two-fold has the possibility of coupling. To assess membrane reshaping on an energetic level, the protein boundary condition is extracted and used in a continuum model. The continuum model is able to determine the membrane deformation energy and thus how strongly a conformation prefers concave, convex, or saddle curvature. The authors find that the tetramers favor concave curvature (corresponding to inward invagination) and that the two-fold symmetric conformations favor negative Gaussian curvature.

    The paper is well and clearly written and introduces a promising method as it tackles an important problem. This is a paper I enjoyed reading very much.

    The approach here will find utility to many problems in membrane reshaping. All-atom simulations are ideal for determining the membrane deformation around proteins (the boundary condition) but are inadequate for assessing the energetics of the deformation in the broader context of a large membrane structure. Continuum modeling is perfect for this. Therefore, a method that couples the two is needed and provided here.

    I have two significant concerns that I believe can be resolved on the timescale of review.

    1. The work identifies substantial thinning in one leaflet. Lipids expand as they thin. Given this, are there too few lipids in this leaflet (which would also indicate thinning)? I would expect their deformations depend strongly on the number-balance of lipids in each leaflet. The authors should check if thinning, and the boundary, is sensitive to inter-leaflet-lipid imbalance.

    2. By constraining the pore to have 2-fold symmetry, the authors remove a large entropic penalty disfavoring such a conformation, and thus presumably disfavoring the negative-gaussian-curvature it induces. For example, if the free energy surface for the fluctuations were rather flat, and only 1% of the conformations were consistent with 2-fold symmetry, the coupling to NGC may be reduced by -kT log( 1 % ), neglecting enhancement by coupling to NGC. Therefore, I predict that the coupling to NGC would be reduced further were the constraint removed.

    Not only are these two points relevant to the M2 problem specifically, but to the method generally. If, in the future, boundary conditions are extracted from molecular simulations, the artificial constraints of such simulations must be adjusted for (e.g., periodic boundary conditions, lipid imbalance).

  6. eLife

    Reviewer #3 (Public Review):

    Helsell et al. uses atomistic molecular dynamics simulations to characterize the structural dynamics of the M2 protein together with continuum elastic models to evaluate the energetic cost of the protein-induced bilayer deformations. Using unbiased simulations (without constraints on the protein) they show that the M2 structure is dynamic and that the AH helices are mobile (though they tend to retain their secondary structure), in agreement with experimental observations. Then, using simulations in which the peptide backbone was restrained to the starting structure, they were able to quantitatively characterize the protein-induced bilayer deformations as well as the acyl chain dynamics.

    Both the atomistic simulations and the continuum-based determinations of the bilayer deformation energies are of high quality. The authors are careful to note that their unbiased simulations do not reach equilibrium, and the authors' conclusions are well supported by their results, though some issues need to be clarified.

    1. P. 7: Choice of lipid composition: POPC:POPG:Cholesterol 0.56:0.14:0.3. This lipid composition (or POPC:POPG 0.8:0.2) has been used in a number of experimental studies that the authors use as reference. It differs, however, substantially from the lipid composition of the influenza membrane (Gerl et al., J Cell Biol, 2012; Ivanova et al., ACS Infect Dis, 2015), which is enriched in cholesterol, has a 2:1 ratio of phosphatidylethanolamine to phosphatidylcholine, and almost no PG. The choice of lipid composition is unlikely to impact the authors' major conclusions, but it should be discussed briefly. As noted by Ivanova et al., the lipids of the influenza membrane are enriched in fusogenic lipids. How will that impact the authors results.

    2. The definition of the lipid tilt needs to be revisited. On P. 13 (in the Pdf received for review, the authors do not provide page numbers), the tilt is defined/approximated as "the angle between the presumed membrane normal (aligned with the Z axis of the box) and the vector pointing from each phospholipid's phosphate to the midpoint between the last carbon atoms of the lipid tails." This (equating the normal to the interface with the Z axis of the simulation box) may be an acceptable approximation for the lower leaflet, which is approximately flat, but probably not for the upper leaflet where the interface is curved in the vicinity of the protein. The authors should, at least, discuss the implications of their approximation in terms of their conclusion that there is little lipid tilt in the upper leaflet.

    3. P. 14, last paragraph, Figure 5 and 6: The snapshots in Figure 5 are too small to see what the authors refer to when they write "tilt their lipid tails to wrap around the helices." The authors should consider citing the work of H W. Huang, e.g., Huang et al. (PRL, 2004), who introduced the notion of curvature stress induced by antimicrobial peptides, a concept similar to what the present authors propose.

    4. P. 17-18, Figure 7: The authors introduce the bilayer midplane, which becomes important for the determination of the deformation energy in the (unnumbered) equation on P. 17, but do not specify how it is determined. This is a non-trivial undertaking, but critical for the evaluation of the deformation energy; please add the necessary details.

    5. P. 18-19, Figure 8: The comparison of the MD and continuum membrane deformations is very informative, but the authors should discuss the implications of the increased symmetry further in terms of the estimated deformation energies. (I do not believe the authors really mean that they predicted the energies, they estimated/approximated them.)

  7. Version published to 10.1101/2022.07.02.498578v1 on bioRxiv