Principles of membrane protein interactions with ... - BioMedSearch

The EMBO Journal Peer Review Process File - EMBO-2010-73601

Manuscript EMBO-2010-73601

Principles of membrane protein interactions with annular lipids deduced from aquaporin-0 2D crystals Rich Hite, Zongli Li and Thomas Walz Corresponding author: Thomas Walz, HHMI & Harvard Medical School

Review timeline:

Submission date: Editorial Decision: Revision received: Editorial Decision: Revision received: Accepted:

07 January 2010 22 February 2010 08 March 2010 22 March 2010 23 March 2010 23 March 2010

Transaction Report: (Note: With the exception of the correction of typographical or spelling errors that could be a source of ambiguity, letters and reports are not edited. The original formatting of letters and referee reports may not be reflected in this compilation.)

1st Editorial Decision

22 February 2010

Dear Professor Walz, Thank you for submitting your manuscript for consideration by the EMBO Journal. Please let me first apologise again for the delay in getting back to you with a decision - as I told you, this was due to some difficulty in finding appropriate referees, and also due to a slight delay in receiving the third report. However, we have now received comments from all three referees, which are enclosed. As you will see, all three referees express interest in your work, and are in principle supportive of publication. However, referees 2 and 3 in particular raise a number of issues that would need to be addressed in a revised version of your manuscript before we could consider publication in the EMBO Journal. In the light of the referees' positive recommendations, I would therefore like to invite you to submit a revised version of the manuscript, addressing all the comments of all three reviewers. I should add that it is EMBO Journal policy to allow only a single round of revision. Acceptance of your manuscript will thus depend on the completeness of your responses included in the next, final version of the manuscript. When preparing your letter of response to the referees' comments, please bear in mind that this will form part of the Review Process File, and will therefore be available online to the community. For more details on our Transparent Editorial Process initiative, please visit our website: http://www.nature.com/emboj/about/process.html Thank you for the opportunity to consider your work for publication. I look forward to your revision. Yours sincerely,

© European Molecular Biology Organization

1


Editor The EMBO Journal REFEREE REPORTS Referee #1 (Remarks to the Author): This manuscript reports the high-resolution structure of AQP0 in two-dimensional (2D) crystals formed with E. coli polar lipids determined by electron crystallography. Based on this structure and the earlier structure in DMPC lipids, the authors report that the acyl chains of the lipids surrounding the protein adapt, in both cases, their conformations in a similar fashion to fit in into the core of the protein hydrophobic surface while the lipid head groups interact with the protein differently presumably to maximize their interactions in order to attain the lowest energy state. Although these general findings are expected, the structure provides excellent initial experimental evidence. This paper is clearly of significance. Given that the structures of AQP0 in two different lipids are now available, the authors should describe the factors that give AQP0 a strong tendency to form two dimensional arrays that is not shared with most other membrane proteins. Is the tendency to form 2D arrays governed primarily by the hydrophobic regions of the proteins or/and its hydrophilic regions? Could the adaptation of the length of the lipid acyl chains to the hydrophobic surface of the protein be the key factor that drives AQP0 to naturally form 2D arrays? Such information is crucial as to whether specific findings based on AQP0 can be readily extended to other membrane proteins; addressing this could help to strengthen the paper. As molecular dynamic simulations have become rather common, the authors may want to conduct simulations to gain further insights into the interactions of the lipids with AQP0. Minor point: How do the assumed membrane thickness used in data processing and the missing data normal to the membrane plane affect the observed thickness of the bilayers and the accuracy of model fitting normal to the membrane plane? Referee #2 (Remarks to the Author): The paper reports interesting studies of the interaction of lipids with aquaporin (AQP0), determined by crystallographic methods. It is, I believe, the first paper to study a protein in two different bilayers and to compare interactions in these two bilayers. As such I think the paper is well worth publishing. The paper is brief and to the point. In places however, I think that the discussion could be misleading to the general reader and that some rewriting would be helpful to avoid confusion. The first point concerns the question of lipid chain length, bilayer thickness and adaptation of AQP0 to the bilayer. The abstract stresses the point (line 6) 'AQP0 does not adapt to the length of the acyl chains of the surrounding lipids'. The question is, what would be expected? The lipids studied here are DMPC, a saturated lipid with C14 chains and an E.coli lipid mixture in which the average chain length will be about C17 and in which most of the lipids will be unsaturated. The hydrophobic thickness of a bilayer of saturated DMPC in the liquid crystalline phase is ca 25.4 Å, and that of a bilayer of unsaturated PCs with a chain length of C17 is 24.9 Å (Marsh, Biophys J. 94, 3996, 2008); so in this case the longer chain gives a slightly thinner bilayer because of the presence of the double bond in the longer chain. If bilayers of DMPC and E coli lipid have very similar thicknesses, then, of course, no adaptation of AQP0 would be expected. I think it is important therefore to discuss bilayer thickness rather than just chain length, and I think the experiments reported here tell us nothing about adaptation of membrane proteins to bilayer thickness. The issue of adaptation is


2


addressed in several places throughout the text, and the above comments apply to all these places. The hydrophobic thicknesses reported here (Page 7) in the presence of AQP0 are 31.2 Å for DMPC and 27.0 Å for E. coli lipids, but these are the thicknesses at very low temperatures (6 oK). In the absence of protein, DMPC would be in the gel phase at this temperature, with straight fatty acyl chains. It is clear from the crystal structure that this is not the case in the presence of AQP0 and the authors might like to comment on this. The idea that the 'annular' lipid is 'disordered' by the presence of the protein agrees with much previous data (see for example Lee, Biochim. Biophys. Acta 1612, 1 (2003)). Page 3. line 2. It is said that most of the available structural information on lipid-protein interactions comes from lipids in crystal structures. This is perhaps not a true reflection of the field - a great detail of information about lipid-protein interactions has come from EPR and fluorescence quenching studies. Page 3, line 3 from bottom. It is said that the data reported here on AQP0 is 'more representative of generic interactions between membrane proteins and annular lipids.' There is, however, a very important limitation to the AQP0 studies and that is that the lipid is trapped in between protein monomers in the tetrameric structure, and so is very unlike the lipid around a 'normal' membrane protein. This is discussed on page 10, but since it is such an important limitation I think it should be described in the introduction as well. Page 4 para 1. The absence of distinct interactions in the lipid headgroup region is described. This is, of course, very reminiscent of the situation with bacteriorhodopsin, established many years ago, and I think this should be referenced here. The fact that most lipids interact with a membrane protein in a non-specific fashion was also established many years ago by EPR and fluorescence quenching studies. Page 8. end of 1st para. It is said that 'these observations suggest that headgroups play a negligible role in the interaction of annular lipids with membrane proteins. ' I think I know what the authors mean, but the sentence as written could be misleading. The lipids studied here are mostly zwitterionic, with an anionic phosphate group and a positively charged amine group. It is clear from many lipid-protein studies that the existence of charge in the lipid headgroup is important - a diacylglycerol would not interact in the same way as a phospholipid. And so the lipid headgroup actually plays an essential role; it is just that the charged headgroups are interacting with a band of charged and polar residues flanking the hydrophobic surface of the protein, rather than binding to a distinct 'site' on the surface (see for example Lee, Biochim. Biophys. Acta 1612, 1 (2003)) Page 10 para 1 of discussion. It is probably too far to say that 'annular lipids have little influence on the structure of AQP0' - all that can be said is that those studied here have very little effect. I think the reference to Zeidel (1994) could be misleading - all these authors studied was the effect of cholesterol on AQP0 in bilayers of E. coli lipids and the effects of adding some (undefined) phosphatidylcholine and phosphatidylserine. It should be made clear what a small range of lipid effects were probed by Zeidel. Page 11, bottom. It is not clear why the reference to Jaud et al (2009) was chosen here. There are very many experiments showing that changing fatty acyl chain lengths leads to changes in protein function, and so must involve changes in protein structure. The paper by Jaud et al (2009) is also less clear cut than suggested here. For example, the glycine-flanked peptides showed CD spectra distinctly different from those of the Lys-flanked peptides so that not all C12 peptides behave in the same way - it is suggested in the paper that Lys-flanked peptides could act as if they were longer than C12, because of snorkelling by the Lys residues. Referee #3 (Remarks to the Author):


3


This interesting manuscript reports the structure of eye lens aquaporin 0 in E. coli polar lipids by electron crystallography, and compares it to a previous structure from the same laboratory of AQP0 in the synthetic lipid DMPC. As far as I am aware, this is the first time that the structure of the same membrane protein has been determined in two different lipids. The authors show conclusively that the lipid adapts to the protein, and not the other way round. This puts paid to some of the earlier literature, where much was made of "hydrophobic mismatch" between the lipid chain length and the hydrophobic surface of a membrane protein, which was thought to be a major determinant in membrane structure and assembly. By contrast, the present manuscript supports the alternative notion of the membrane lipid as a passive hydrophobic matrix for the protein. The most striking finding is that the DMPC bilayer is significantly thicker than the E. coli lipid bilayer, even though the C12 DMPC acyl chain is substantially shorter than the C16-C18 chains of EPL. This is the opposite of what one would expect, and goes completely against the "hydrophobic mismatch" theory, which is thus shown to be invalid. Perhaps the authors are too new to the field to be aware of this earlier literature. At least they do not mention it, although at least a brief discussion in the context of their new findings would be essential. Apart from that, there are a few points that need further thought, or a little more work. 1. The authors state that the structure of AQP0 in the two lipids is virtually identical. Yet the surface representations in Figs 2 and 3 suggest major differences between the two structures. How can this be? 2. On p 7 the authors state that all EPL head groups adopt different conformations. This is puzzling, since on the previous page they say that the head groups all looked the same and were therefore built as PE. Surely the differences between head groups are greater than between different conformations? Please explain. 3. Are the bends in the EPL acyl chains actually in the positions expected of cis double bonds? Is there anything special or systematic in the location of the bends, in terms of distance along the acyl chains, or surface structure of AQP0? Does the saturated PC bend in the same direction in these positions? This point deserves some more space, and perhaps a more detailed figure. 4. It is a pity that the alternative conformations of the acyl chains, evident in the form of branched densities in the original map, were refined away, as this removes interesting and potentially important information. If the data were "insufficient for the refinement of alternative conformations" (p6), then either more data need to be added, or the alternative positions should be refined separately and presented along with the main conformations.

1st Revision - Authors' Response

08 March 2010

Referee #1: This manuscript reports the high-resolution structure of AQP0 in two-dimensional (2D) crystals formed with E. coli polar lipids determined by electron crystallography. Based on this structure and the earlier structure in DMPC lipids, the authors report that the acyl chains of the lipids surrounding the protein adapt, in both cases, their conformations in a similar fashion to fit in into the core of the protein hydrophobic surface while the lipid head groups interact with the protein differently presumably to maximize their interactions in order to attain the lowest energy state. Although these general findings are expected, the structure provides excellent initial experimental © European Molecular Biology Organization

4


evidence. This paper is clearly of significance. We would like to thank the referee for appreciating the significance of our paper. Given that the structures of AQP0 in two different lipids are now available, the authors should describe the factors that give AQP0 a strong tendency to form two dimensional arrays that is not shared with most other membrane proteins. Is the tendency to form 2D arrays governed primarily by the hydrophobic regions of the proteins or/and its hydrophilic regions? The referee raises a very interesting point, which, unfortunately, cannot be addressed with our 2D crystals. We artificially induce the formation of 2D crystals in vitro by carefully adjusting the lipidto-protein ratio that is used for reconstitution. This is very different from how AQP0 forms 2D arrays in vivo. AQP0 arrays in lens membranes are surrounded by large non-crystalline membrane areas, so that a precise lipid-to-protein ratio cannot be the driving force. Thus, the lipid-protein interactions observed in our 2D crystals do not provide clues to what induces array formation in vivo. In lens membranes, AQP0 is enriched in detergent-resistant membrane (DRM) domains, which contain high concentrations of cholesterol and sphingomyelin (Tong et al., 2009, Biophys. J. 97: 2493-2502), lipids implicated in the formation of lipid rafts. Our current working hypothesis is that formation of AQP0 arrays in native lens membranes is based on similar principles as raft formation, which requires interactions with the specific raft lipids. We plan to address this question in the future by growing 2D crystals of AQP0 in the presence of cholesterol and/or sphingomyelin and compare the lipid-protein interactions with those seen in AQP0 2D crystals grown with conventional phospholipids. Could the adaptation of the length of the lipid acyl chains to the hydrophobic surface of the protein be the key factor that drives AQP0 to naturally form 2D arrays? We had the same idea as the referee, but decided that this is most likely not the case. The main reason is that AQP0 forms arrays in lens membranes, but a larger number of the proteins also exists as individual tetramers in these membranes. Since the annular lipids also have to adapt to the hydrophobic surface of individual AQP0 tetramers, this adaptation does not seem to be a sufficient force to drive AQP0 into arrays. We therefore favor the idea that array formation depends on interactions of AQP0 with specific raft lipids, such as cholesterol and sphingomyelin. Such information is crucial as to whether specific findings based on AQP0 can be readily extended to other membrane proteins; addressing this could help to strengthen the paper. As molecular dynamic simulations have become rather common, the authors may want to conduct simulations to gain further insights into the interactions of the lipids with AQP0. We fully agree with the referee that molecular dynamics simulations will provide additional and complementary insights into lipid-protein interactions and will allow it to generalize our findings for AQP0 to other membrane proteins. We have therefore already initiated a collaboration with Bert de Groot (MPI for Biophysical Chemistry, Goettingen, Germany) to perform such simulations to address a variety of questions. However, for de Groot’s group to perform the calculations, we first have to secure funding, for which we are currently applying. It is thus not feasible for us to include molecular dynamics simulations in the current manuscript.

Minor point: How do the assumed membrane thickness used in data processing and the missing data normal to


5


the membrane plane affect the observed thickness of the bilayers and the accuracy of model fitting normal to the membrane plane? Since 2D crystals are not periodic in the direction perpendicular to the membrane, the Fourier transform is continuous in this direction. Rather than discrete spots, as in the case of 3D crystals, the Fourier transform of a 2D crystal is thus a set of reciprocal lattice lines. An artificial thickness of the 2D crystal is assumed to sample the reciprocal lattice lines and to convert the reciprocal lattice lines into a set of discrete diffraction spots that can be processed with conventional X-ray crystallography software. As long as the assumed thickness of the unit cell is larger than the actual 2D crystal, the exact value will not influence the resulting density map. Because our data set was very large, we could assume a large unit cell in c direction of 200 Å and still have a high multiplicity (8.1 over the entire resolution range and 4.0 in the highest resolution bin). As a result of the larger c dimension we chose, the unit cell comprises empty volume that is not occupied by protein and lipid, which we could use for solvent flattening, a technique that helps reduce potential model bias. Missing data normal to the membrane plane, the so-called missing cone, lead to a reduction of the resolution in this direction. Because we included data collected from specimens tilted to an angle of 70º, the missing cone is very small in our data (we sampled 92.3% of all of reciprocal space and 83.5% of the highest resolution shell). The reduction of resolution in z-direction is therefore minimal. Furthermore, since the resolution of our structure was sufficient to perform crystallographic refinement, which takes advantage of chemical knowledge (lengths and angles of chemical bonds), providing very strong constraints for model building. The missing cone thus becomes a negligible concern. Indeed, our refined structure is very similar to the X-ray structure of AQP0 (Harries et al., 2004, PNAS 101: 14045-14050, PDB 1YMG) with an RMSD between the backbone atoms of only 0.61 Å. Referee #2: The paper reports interesting studies of the interaction of lipids with aquaporin (AQP0), determined by crystallographic methods. It is, I believe, the first paper to study a protein in two different bilayers and to compare interactions in these two bilayers. As such I think the paper is well worth publishing. The paper is brief and to the point. In places however, I think that the discussion could be misleading to the general reader and that some rewriting would be helpful to avoid confusion. We thank the referee for the positive evaluation of our manuscript. The first point concerns the question of lipid chain length, bilayer thickness and adaptation of AQP0 to the bilayer. The abstract stresses the point (line 6) 'AQP0 does not adapt to the length of the acyl chains of the surrounding lipids'. The question is, what would be expected? The lipids studied here are DMPC, a saturated lipid with C14 chains and an E.coli lipid mixture in which the average chain length will be about C17 and in which most of the lipids will be unsaturated. The hydrophobic thickness of a bilayer of saturated DMPC in the liquid crystalline phase is ca 25.4 Å, and that of a bilayer of unsaturated PCs with a chain length of C17 is 24.9 Å (Marsh, Biophys J. 94, 3996, 2008); so in this case the longer chain gives a slightly thinner bilayer because of the presence of the double bond in the longer chain. If bilayers of DMPC and E coli lipid have very similar thicknesses, then, of course, no adaptation of AQP0 would be expected. I think it is important therefore to discuss bilayer thickness rather than just chain length, and I think the experiments reported here tell us nothing about adaptation of membrane proteins to bilayer thickness. The issue of adaptation is addressed in several places throughout the text, and the above comments apply to all these places. We completely agree with the referee that according to the formula presented by Marsh, the thickness of a lipid bilayer formed by an unsaturated PC lipid with C17 acyl chains, 24.9 Å, would be similar to that formed by DMPC, 25.4 Å. However, about 45% of the acyl chains of EPLs are saturated (Lugtenberg & Peters,1976, Biochim. Biophys. Acta 441: 38-47), and according to Marsh’s formula, a saturated PC lipid with C17 acyl chains would form a bilayer with a hydrophobic © European Molecular Biology Organization

6


thickness of 32.2 Å. We do not know how to appropriately apply Marsh’s formula to complex lipid mixtures, but a rough estimation (0.45 x 32.2Å + 0.55 x 24.9Å) yields a hydrophobic thickness of a bilayer formed by this mixture of acyl chains of 28.2 Å, which is about 3 Å thicker than that of a DMPC bilayer. While our estimate is clearly just an approximation, it should be more realistic than assuming all EPL acyl chains to be unsaturated. We therefore believe that it is appropriate to assume that not only the chain length is longer in EPLs than in DMPC, but also that the hydrophobic thickness of the bilayer formed by EPLs is likely thicker than that of a bilayer formed by DMPC. However, we appreciate the referee’s concern and modified our manuscript in several places: Abstract: “Comparison of the two structures shows that AQP0 does not adapt to the different length of the acyl chains in EPLs and ...” Discussion: “Comparison of our new AQP0EPL structure with the previously determined AQP0 DMPC structure shows that the annular lipids studied here have very little influence on the structure of AQP0.” “In the two structures, it is exclusively the lipids that adapt to the protein, which bend and interdigitate to accommodate the longer acyl chains of the EPLs. A caveat of this observation is that the acyl chains of EPLs are not only longer than those of DMPC, but 55% of the acyl chains of EPLs are unsaturated (Lugtenberg & Peters, 1976), and double bonds reduce the hydrophobic thickness of the bilayer formed by unsaturated lipids (Marsh, 2008). Thus, while the AQP0DMPC and AQP0EPL structures suggest that it is exclusively the lipids that adapt to the protein, the hydrophobic thickness of the DMPC and EPL bilayers may be too similar to induce observable changes in the AQP0 structure. To conclusively rule out structural changes in AQP0 due to hydrophobic mismatch, it will be necessary to determine structures of AQP0, in which the protein is surrounded by lipids that form bilayers with a hydrophobic thickness that is significantly different from that of AQP0.” “While some membrane proteins can change their structure to adjust to bilayers with a significantly different hydrophobic thickness, it remains to be determined whether rigid membrane proteins, such as AQP0, can adapt in a similar manner.” The hydrophobic thicknesses reported here (Page 7) in the presence of AQP0 are 31.2 Å; for DMPC and 27.0 Å; for E. coli lipids, but these are the thicknesses at very low temperatures (6 oK). In the absence of protein, DMPC would be in the gel phase at this temperature, with straight fatty acyl chains. It is clear from the crystal structure that this is not the case in the presence of AQP0 and the authors might like to comment on this. The idea that the 'annular' lipid is 'disordered' by the presence of the protein agrees with much previous data (see for example Lee, Biochim. Biophys. Acta 1612, 1 (2003)). We fully agree with the referee, but despite the low temperature used for data collection, our structures are more likely to represent both lipids in the gel phase. Our AQP0 2D crystals are grown at 37ºC, which is above the phase transition temperature of both EPLs and DMPC. For electron microscopy, the crystals are quick-frozen by plunging them into liquid nitrogen. Although we cannot be absolutely sure, the cooling rate should be too fast to allow the lipids to transition from the gel phase to the liquid crystalline phase. Thus, while we have no doubts that the protein would prevent the lipid acyl chains from assuming straight conformations even below the phase transition temperature, we cannot comment on this, because we believe that the lipids are in the gel phase, in which the acyl chains are disordered anyway. Page 3. line 2. It is said that most of the available structural information on lipid-protein interactions comes from lipids in crystal structures. This is perhaps not a true reflection of the field a great detail of information about lipid-protein interactions has come from EPR and fluorescence quenching studies. We refer to structural information as atomic models of lipids interacting with membrane proteins,


7


but we certainly agree with the referee that spectroscopic studies have provided many insights into lipid-protein interactions. We have now modified the introduction to make this clear: “Spin labeling and fluorescence quenching studies have provided a thermodynamic understanding of lipid-protein interactions, but these methods do not allow a direct visualization of individual interactions between a protein and a lipid. Most of the available atomic resolution structural information on lipid-protein interactions comes from lipids in crystal structures of membrane proteins in detergent micelles.” Page 3, line 3 from bottom. It is said that the data reported here on AQP0 is 'more representative of generic interactions between membrane proteins and annular lipids.' There is, however, a very important limitation to the AQP0 studies and that is that the lipid is trapped in between protein monomers in the tetrameric structure, and so is very unlike the lipid around a 'normal' membrane protein. This is discussed on page 10, but since it is such an important limitation I think it should be described in the introduction as well. We now refer to this limitation also in the introduction: “The situation of the lipids in AQP0 2D crystals is, however, somewhat special, because they are sandwiched in between two AQP0 tetramers, which is not typically the case for lipids surrounding membrane proteins in biological membranes.” Page 4 para 1. The absence of distinct interactions in the lipid headgroup region is described. This is, of course, very reminiscent of the situation with bacteriorhodopsin, established many years ago, and I think this should be referenced here. The fact that most lipids interact with a membrane protein in a non-specific fashion was also established many years ago by EPR and fluorescence quenching studies. We have substantially reorganized the introduction and now describe these previous results: “Therefore, lipids in crystal structures must be strongly bound to the membrane proteins. These lipids are a special case of “annular” lipids, the lipids in direct contact with a membrane protein, because spin labeling and fluorescence quenching studies demonstrated that most annular lipids form only weak and non-specific interactions with membrane proteins (Lee, 2003).” “In contrast, density for the lipid headgroups was poor in electron and X-ray crystallographic density maps of bacteriorhodospsin (e.g., Grigorieff et al, 1996; Luecke et al, 1999), a light-driven proton pump that forms crystalline arrays in the membrane of Halobacterium salinarum, raising the question of what general principles govern the interactions of membrane proteins with their annular lipids.” Page 8. end of 1st para. It is said that 'these observations suggest that headgroups play a negligible role in the interaction of annular lipids with membrane proteins. ' I think I know what the authors mean, but the sentence as written could be misleading. The lipids studied here are mostly zwitterionic, with an anionic phosphate group and a positively charged amine group. It is clear from many lipid-protein studies that the existence of charge in the lipid headgroup is important - a diacylglycerol would not interact in the same way as a phospholipid. And so the lipid headgroup actually plays an essential role; it is just that the charged headgroups are interacting with a band of charged and polar residues flanking the hydrophobic surface of the protein, rather than binding to a distinct 'site' on the surface (see for example Lee, Biochim. Biophys. Acta 1612, 1 (2003)) We have modified this sentence in the revised manuscript: “Together, these observations suggest that the exact chemical identities of the phospholipid headgroups play a negligible role in the interaction of annular lipids with membrane proteins, corroborating previous results obtained by spin labeling studies (Lee, 2003).” Page 10 para 1 of discussion. It is probably too far to say that 'annular lipids have little influence


8


on the structure of AQP0' - all that can be said is that those studied here have very little effect. I think the reference to Zeidel (1994) could be misleading - all these authors studied was the effect of cholesterol on AQP0 in bilayers of E. coli lipids and the effects of adding some (undefined) phosphatidylcholine and phosphatidylserine. It should be made clear what a small range of lipid effects were probed by Zeidel. According to the referee’s comments, we have made the following modifications in the revised manuscript: “Comparison of our new AQP0EPL structure with the previously determined AQP0 DMPC structure shows that the annular lipids studied here have very little influence on the structure of AQP0.” “The virtually identical channel structure in AQP0 EPL and AQP0DMPC is consistent with a previous study that showed that the lipid environment did not affect water conduction by AQP1 (Zeidel et al, 1994), although only a very limited range of lipid compositions were tested in this study.” Page 11, bottom. It is not clear why the reference to Jaud et al (2009) was chosen here. There are very many experiments showing that changing fatty acyl chain lengths leads to changes in protein function, and so must involve changes in protein structure. The paper by Jaud et al (2009) is also less clear cut than suggested here. For example, the glycine-flanked peptides showed CD spectra distinctly different from those of the Lys-flanked peptides so that not all C12 peptides behave in the same way - it is suggested in the paper that Lys-flanked peptides could act as if they were longer than C12, because of snorkelling by the Lys residues. We agree with the referee and significantly rewrote this part of the discussion: “It has been shown for several membrane proteins that hydrophobic mismatch affects their activity (Lee, 2004), suggesting changes in protein structure. It is not clear, however, whether all membrane proteins have the flexibility to adjust to the hydrophobic thickness of the surrounding lipid bilayer. AQP0, which is a very stable and presumably a very rigid membrane protein, has the same structure in the EPL and DMPC bilayers (Figure 1D). In the two structures, it is exclusively the lipids that adapt to the protein, which bend and interdigitate to accommodate the longer acyl chains of the EPLs. A caveat of this observation is that the acyl chains of EPLs are not only longer than those of DMPC, but 55% of the acyl chains of EPLs are unsaturated (Lugtenberg & Peters, 1976), and double bonds reduce the hydrophobic thickness of the bilayer formed by unsaturated lipids (Marsh, 2008). Thus, while the AQP0 DMPC and AQP0EPL structures suggest that it is exclusively the lipids that adapt to the protein, the hydrophobic thickness of the DMPC and EPL bilayers may be too similar to induce observable changes in the AQP0 structure. To conclusively rule out structural changes in AQP0 due to hydrophobic mismatch, it will be necessary to determine structures of AQP0, in which the protein is surrounded by lipids that form bilayers with a hydrophobic thickness that is significantly different from that of AQP0.”

Referee #3: This interesting manuscript reports the structure of eye lens aquaporin 0 in E. coli polar lipids by electron crystallography, and compares it to a previous structure from the same laboratory of AQP0 in the synthetic lipid DMPC. As far as I am aware, this is the first time that the structure of the same membrane protein has been determined in two different lipids. The authors show conclusively that the lipid adapts to the protein, and not the other way round. This puts paid to some of the earlier literature, where much was made of "hydrophobic mismatch" between the lipid chain length and the hydrophobic surface of a membrane protein, which was thought to be a major determinant in membrane structure and assembly. By contrast, the present manuscript supports the alternative notion of the membrane lipid as a passive hydrophobic matrix for the protein.


9


We are delighted that the referee finds our results interesting, but we would like to state that our results do not discount hydrophobic mismatch at this point. We are attempting to determine more structures of AQP0 with different lipids to more conclusively address the question of hydrophobic mismatch. The most striking finding is that the DMPC bilayer is significantly thicker than the E. coli lipid bilayer, even though the C12 DMPC acyl chain is substantially shorter than the C16-C18 chains of EPL. This is the opposite of what one would expect, and goes completely against the "hydrophobic mismatch" theory, which is thus shown to be invalid. Perhaps the authors are too new to the field to be aware of this earlier literature. At least they do not mention it, although at least a brief discussion in the context of their new findings would be essential. We also believe that our structures do not support the hydrophobic mismatch theory, but they also do not render the theory invalid at this point. The issue is that we cannot be sure of the hydrophobic thickness of a bilayer formed by EPLs. While the acyl chains are longer in EPLs, they also partially unsaturated, which affects the thickness of the bilayer EPLs form. We still think that the EPL bilayer should be thicker than the DMPC bilayer, but we cannot be certain. Thus, to address the question of hydrophobic mismatch, we are currently trying to grow 2D crystals of AQP0 with different lipids with a well-defined acyl chain length to obtain a better understanding of the relationship between the acyl chain length and the thickness of the bilayer surrounding the AQP0 tetramers. In the revised manuscript, we now include, however, a discussion of our data in view of the hydrophobic mismatch theory: “The structures of AQP0 in two different lipid environments allow us to see how the protein and the lipid bilayer adapt to each other. To accommodate proteins in a lipid bilayer with a different hydrophobic thickness, a situation known as hydrophobic mismatch, it has been proposed that either the lipids adjust their length to match the protein or that a-helical membrane proteins may stretch or contract axially to adapt to the thickness of the surrounding lipid bilayer (Lee, 2004). It has been shown for several membrane proteins that hydrophobic mismatch affects their activity (Lee, 2004), suggesting changes in protein structure. It is not clear, however, whether all membrane proteins have the flexibility to adjust to the hydrophobic thickness of the surrounding lipid bilayer. AQP0, which is a very stable and presumably a very rigid membrane protein, has the same structure in the EPL and DMPC bilayers (Figure 1D). In the two structures, it is exclusively the lipids that adapt to the protein, which bend and interdigitate to accommodate the longer acyl chains of the EPLs. A caveat of this observation is that the acyl chains of EPLs are not only longer than those of DMPC, but 55% of the acyl chains of EPLs are unsaturated (Lugtenberg & Peters, 1976), and double bonds reduce the hydrophobic thickness of the bilayer formed by unsaturated lipids (Marsh, 2008). Thus, while the AQP0DMPC and AQP0EPL structures suggest that it is exclusively the lipids that adapt to the protein, the hydrophobic thickness of the DMPC and EPL bilayers may be too similar to induce observable changes in the AQP0 structure. To conclusively rule out structural changes in AQP0 due to hydrophobic mismatch, it will be necessary to determine structures of AQP0, in which the protein is surrounded by lipids that form bilayers with a hydrophobic thickness that is significantly different from that of AQP0.” Apart from that, there are a few points that need further thought, or a little more work. 1. The authors state that the structure of AQP0 in the two lipids is virtually identical. Yet the surface representations in Figs 2 and 3 suggest major differences between the two structures. How can this be? The major difference between the AQP0 structures in the two lipid bilayers the referee refers to is due to the C terminus, which was modeled in the DMPC structure but not in the EPL structure. We apologize for not pointing this out more clearly in the original manuscript and we now explicitly state this fact in the revised version: ”Residues 1 to 6 and residues 227 to 263, which include the Cterminal helix modeled in the previous structure of AQP0 in a DMPC bilayer (Gonen et al, 2005), did not show clear density in the map and were therefore excluded from the model of AQP0 in the © European Molecular Biology Organization

10


EPL bilayer.” 2. On p 7 the authors state that all EPL head groups adopt different conformations. This is puzzling, since on the previous page they say that the head groups all looked the same and were therefore built as PE. Surely the differences between head groups are greater than between different conformations? Please explain. A PE and a PG headgroup are actually not very different in terms of the density they produce in an EM map (see space filled models of the two lipids in Figure R1). Since there does not seem to be a preference for a particular lipid at a particular position, the same position is randomly occupied by a PE or a PG lipid (or even a cardiolipin), and the small differences between the headgroups are averaged out in the density map. This is described in the manuscript “In the absence of preferential binding sites, averaging of the three headgroups would likely result in all headgroups appearing as the predominant PE headgroup.” On the other hand, we saw sufficient density for the lipid headgroups to model the conformation, suggesting that regardless of the headgroup, PE or PG, the headgroup adopts the same conformation at a given lipid position, which we could thus model. Figure R1 – for referee only

3. Are the bends in the EPL acyl chains actually in the positions expected of cis double bonds? Is there anything special or systematic in the location of the bends, in terms of distance along the acyl chains, or surface structure of AQP0? Does the saturated PC bend in the same direction in these positions? This point deserves some more space, and perhaps a more detailed figure. We agree with the referee that correlating the position of the kinks observed in the acyl chains with the position of double bonds in their chemical structure would be very informative. Unfortunately, EPLs are a complex lipid mixture, which makes it impossible to be sure of the chemical structure of the acyl chains that are represented by a particular density in the map. However, precisely to address the point raised by the referee, we are currently attempting to grow 2D crystals of AQP0 with two defined lipids that differ only in that one of the lipids has an unsaturated acyl chain. Once we have these two structures, we should be able to make definitive statements about the relationship between observed kinks in the density and double bonds in the chemical structure. For our structure with EPLs, the positions of the kinks are only suggestive of double bonds. To address the referee’s comment, we added the following two statements in the revised manuscript: “Approximately 55% of the acyl chains of EPLs contain an unsaturated bond, with the two most abundant species being 16c1:9 and 18c1:11 (Lugtenberg & Peters, 1976). While some lipids showed kinks that could indicate the presence of a double bond, due to the heterogeneity of the acyl chains in EPLs, we modeled all acyl chains as being fully saturated.”


11


“Indeed, the positions of the kinks seen in the acyl chains of lipids PE3, PE4 and PE6 occur close to the positions of the double bonds in the most abundant unsaturated acyl chains of EPLs, 16c1:9 and 18c1:11, whereas the kink in the acyl chain of PE7 is located between C6 and C7.” 4. It is a pity that the alternative conformations of the acyl chains, evident in the form of branched densities in the original map, were refined away, as this removes interesting and potentially important information. If the data were "insufficient for the refinement of alternative conformations" (p6), then either more data need to be added, or the alternative positions should be refined separately and presented along with the main conformations. We agree with the referee that it is a pity that we cannot model the alternative conformations. It is, however, not possible to simply collect more data. We have already included many more diffraction patterns than what would be required (multiplicity of 8.1 over the entire resolution range and 4.0 in the highest resolution bin) and simply adding more diffraction patterns would not help. The additional data that would be required are additional reflections. The number of unique reflections that can be observed in diffraction patterns, or in other words the resolution, depends on the order of the crystals. To obtain more data, we would thus have to produce better-ordered crystals to be able to collect higher-resolution diffraction patterns, but we were unable to obtain better-ordered crystals of AQP0 with E. coli polar lipids (a resolution of 2.5 Å is in fact already exceptional for electron crystallography of 2D crystals). We used our data to refine the most populated lipid conformations, and it would also be technically inappropriate to try and use the same data to refine less populated lipid conformations. With our current crystals we are thus limited to modeling only the most highly populated lipid conformations.

2nd Editorial Decision

22 March 2010

Dear Dr. Walz, Many thanks for submitting the revised version of your manuscript EMBOJ-2010- 73601R. It has now been seen again by all three referees, whose comments are enclosed below. As you will see, all three find that you have responded well to the concerns raised in the previous round of review, and are supportive of publication. I am therefore pleased to be able to tell you that we will be able to accept your manuscript for publication in the EMBO Journal. However, there are just a couple of remaining points that need to be dealt with first. Referee 1 still feels that the molecular dynamics simulations would be valuable, but from an editorial standpoint, we accept your reasoning for not being able to include these at this point, and would therefore not insist on this point. Referee 3 raises a number of minor issues with the text, and I would like to ask you to modify the manuscript text to address these. I suggest that the easiest way forward would be for you to make these changes, and then e-mail a final version of the manuscript that we can replace in the system. I also notice that the PDB accession code has not yet been provided; please can you ensure to include this in the modified text? Once we have received this new version, we should then be able to accept your study without further delay. Many thanks and best wishes, Editor EMBO Journal


12


REFEREE REPORTS Referee 1: The authors have responded adequately to most of the comments raised by the reviewers. However, the authors may want to find a way to include the molecular dynamics simulations in the current paper, as performing such studies requires nominal expense. The de Groot group has conducted extensive simulation studies of water channels and it would only require modest efforts to modify their earlier simulation scripts for the suggested simulations. Availability of the molecular dynamics data may provide greater detail on protein-lipid interaction dynamics and can be expected to significantly enhance the importance of the paper. Referee 2: The authors have responded well to the points raised by the referees and have produced a well balanced and very clear paper. Referee 3: The authors have answered the queries of the three referees, and have improved the clarity of the manuscript in several respects. The manuscript should be published more or less as it stands, with the exception of the following minor points: 1. It is confusing (and potentially misleading) to say that the two structures are "virtually identical", if two lines above the authors state that there is in fact a significant difference, in that the C-terminal helix could not be modeled in the EPL map. This most likely indicates that it was disordered, in which case the two structures are not identical. 2. It does not really matter, but the second time round I was surprised to read that the electron diffraction data were recorded at 6K. I thought the Polara only went down to 8K or maybe even only 15K. Please check. Was the temperature actually measured? 3. The conclusion states that "While some membrane proteins can change their structure to adjust ..." I do not think this has been shown, certainly not in terms of experimentally determined 3D membrane protein structures. The corresponding statements in the results and discussion sections are much less apodictic. The authors may want to modify this statement, e.g. by replacing "can" with "may".

3rd Revision - Authors' Response

23 March 2010

Thank you very much for your email and the great news. I very much appreciate it that you understand why we cannot include molecular dynamics simulations in the current manuscript, although we certainly hope to do such studies soon. In response to Referee 3, we made the following changes: 1. It is confusing (and potentially misleading) to say that the two structures are "virtually identical", if two lines above the authors state that there is in fact a significant difference, in that the C-


13


terminal helix could not be modeled in the EPL map. This most likely indicates that it was disordered, in which case the two structures are not identical. Our understanding is that missing density is synonymous with disorder and we did not think that this could be misleading to readers. Nevertheless, to made this point absolutely clear and to avoid misleading any readers, we added the following text to the manuscript (change in bold): Other than the C-terminal helix, which was disordered and could therefore not be modeled, the structure of AQP0 in the EPL bilayer (AQP0EPL) is virtually identical with its structure in the DMPC bilayer (AQP0DMPC) (Fig. 1D). 2. It does not really matter, but the second time round I was surprised to read that the electron diffraction data were recorded at 6K. I thought the Polara only went down to 8K or maybe even only 15K. Please check. Was the temperature actually measured? We are perfectly certain that the temperature read out by our Polara is 6K. Of course, we cannot be certain that this is indeed the temperature of the specimen, but it seems reasonable to write in the manuscript the temperature that is provided by the instrument. Furthermore, as the referee states himself/herself, a difference in temperature from 6K to 15K would not affect our results in any way. 3. The conclusion states that "While some membrane proteins can change their structure to adjust ..." I do not think this has been shown, certainly not in terms of experimentally determined 3D membrane protein structures. The corresponding statements in the results and discussion sections are much less apodictic. The authors may want to modify this statement, e.g. by replacing "can" with "may". At least in the case of mechanosensitive potassium channels, there is good evidence that the protein changes its structure depending on the thickness of the lipid bilayer (Liu et al. (2009) "Structure of a tetrameric MscL in an expanded intermediate state". Nature 461: 47-49). However, because the lipids cannot be seen in this structure, it is not absolute proof that the protein structure changes in response to the structure of the lipid bilayer. We therefore made the requested change: While some membrane proteins may change their structure to adjust to bilayers with a significantly different hydrophobic thickness, it remains to be determined whether rigid membrane proteins, such as AQP0, can adapt in a similar manner. Finally, we have now included the PDB accession code in the manuscript (3M9I). Thank you very much for your help and I hope the attached manuscript is now acceptable for publication in the EMBO Journal. Best wishes, Tom


14