Mechanism of membrane pore formation by human gasdermin‐D

Abstract Gasdermin‐D (GSDMD), a member of the gasdermin protein family, mediates pyroptosis in human and murine cells. Cleaved by inflammatory caspases, GSDMD inserts its N‐terminal domain (GSDMDNterm) into cellular membranes and assembles large oligomeric complexes permeabilizing the membrane. So far, the mechanisms of GSDMDNterm insertion, oligomerization, and pore formation are poorly understood. Here, we apply high‐resolution (≤ 2 nm) atomic force microscopy (AFM) to describe how GSDMDNterm inserts and assembles in membranes. We observe GSDMDNterm inserting into a variety of lipid compositions, among which phosphatidylinositide (PI(4,5)P2) increases and cholesterol reduces insertion. Once inserted, GSDMDNterm assembles arc‐, slit‐, and ring‐shaped oligomers, each of which being able to form transmembrane pores. This assembly and pore formation process is independent on whether GSDMD has been cleaved by caspase‐1, caspase‐4, or caspase‐5. Using time‐lapse AFM, we monitor how GSDMDNterm assembles into arc‐shaped oligomers that can transform into larger slit‐shaped and finally into stable ring‐shaped oligomers. Our observations translate into a mechanistic model of GSDMDNterm transmembrane pore assembly, which is likely shared within the gasdermin protein family.

there are still some concerns that need to be addressed before the manuscript can be accepted for publication. 1) GSDMD-N domain has been shown to specifically bind to phosphoinositides and cardiolipin as well as phosphatidylserine. Phosphoinositides are known to be present in mammalian cell membrane, particularly the cytoplasmic leaflet of plasma membrane. Upon binding to phosphoinositides, GSDMD-N domain forms extensive pores on the membrane, disrupting the membrane integrity and eventually triggering cell pyroptosis. Compared with cardiolipin, phosphoinositides are the more physiological relevant targets of GSDMD as well as other Gasdermin-family members. Therefore, the authors should also exhibit the pore formation on membranes containing phosphoinositides. The proposed negative effect of cholesterol could also be tested on phosphoinositides-containing membranes. Additionally, phosphatidylserine-containing membrane can be assayed and the results will be of great interest to the field. These control assays will make the study more complete and the results obtained are more physiologically relevant.
2) In the time-lapse AFM analysis that tracks the process of GSDMD-N domain oligomerization and pore formation, the authors showed that the heights of oligomeric GSDMD-N domain protruding from the membrane surface did not change during pore formation. This observation indicates a "growing" process instead of involving a "prepore-to-pore transition" step seen with many other pore-forming proteins. This observation is interesting but needs to be strengthened by additional evidences. For example, the author could check whether a possible prepore of GSDMD-N domain could be induced in the solution state by adding phosphoinositide into caspase-cleaved GSDMD. Also a typical cholesterol-dependent cytolysin, which is known to form prepores in solution and undergo a vertical collapse on the membrane, could be included as a control.
3) Since other Gasdermin-family members also habor the pore-forming activity, it is important to known whether a similar dynamic process of Gasdermin N-domain assembling into transmembrane pores also applies to other gasdermins. This data will increase the impact of the current study.
Referee #2: Mulvihill and al present an AFM study to characterize the mechanism of pore formation by inflammatory Gasdermin-D in model membrane systems. They study the role of lipid composition on GSDMD binding to the membrane and pore formation and provide a time sequence for GSDMD assembly into rings. This work presents some new details about how lipids influence GSDMD binding and assembly in the membrane. Based on their data they conclude that cholesterol plays an inhibitory role in helping GSDMD binding to the membrane. Moreover, they provide high imaging resolution of GSDMD pores and some evidence about the kinetics of GSDMD pore formation, where GSDMD seems to pass through slits/arcs intermediates before forming rings that, according to the authors, represent a more thermodynamically stable assembling of oligomerized GSDMD. However, in my opinion this work represents an incremental advance on the previous work by the same authors in Sborgi et al Embo J. 2016, where many of the results and ideas presented in this AFM study have been previously proposed. That said, I would like to point out that the quality of the AFM imaging is impressive and offers the potential for a more in depth analysis of GSDMD pore formation beyond what is presented here and that would make a difference in terms of novelty of the findings. For example, the idea of a kinetic mechanism for GSDMD pore formation is intriguing; however the conclusions should be supported by more robust evidence. The authors should provide statistical numbers of how many slit and arcs shapes turn into rings, how many rings form pores over time, whether the pore size increases over time before reaching the final size, whether slit and arcs pores are smaller than ring pores and whether increasing concentration of GSDMD affect all these parameters. They could also analyze what is the smallest size of slit and arc that can form a pore, and more interestingly, given the high quality of the resolution imaging, what is the minimum number of GSDMD molecules that can be associated with a membrane pore. Based on the relatively wide distribution of the ring diameter, the authors claim that GSDMD pore formation is a "structurally rather flexible process" (pag 6). This can be a result of a dynamic assembling process where the pore grows over time before reaching a steady state. It would be interesting to explore this hypothesis by, again, looking at the pore size over time, also in the context of understanding the nature of GSDMD pores (only protein or lipid-protein based). Still, for a better The paragraph describing the sub-nanometer topographs of GSDMd is shallow. The authors refer to substructures, but it´s not clear to me what they are referring to. They should illustrate with arrows in the figures. What do they mean with lateral distance? Again an illustration would be useful.
Minor aspects: -For clarity, the authors should provide a zoom in picture of a slit, an arc and a ring -The title of the first paragraph in the results section is misleading because the authors said that GSDMD oligomerization depends on the lipid composition, however later on in the text they specify that GSDMD assembly is not sensitive to the lipid composition (pag 6). They should write again the title focusing more on the Cholesterol role rather than lipids in general.
-On page 5 the authors, based on the low height of the structures protruding the membrane, conclude that GSDMD monomers fully insert into the membrane. This is a quite strong statement not supported by high resolution structural data - Fig  Mulvihill study the insertion and pore-formation of GSDMD N-termini into membranes by AFM. They find that the structures formed are of different shapes and sizes. The AFM pictures are impressive and of very high quality and resolution. Taken that this is the strong part of the manuscript, the most should be made of this, e.g., the scale of height should be added and interpreted in all pictures. The main criticism is that it is difficult to judge how relevant the data is since a membrane on mica might not allow all conformations the protein might assume in a real membrane.
Some important points are listed below.
The dimensions of the pore should be quantified not only by diameter but also by the number of GSDMD subunits per pore, which seem possible, at least from the data in figure 4 and is the strong point of this work.
More detailed statistics on the shape and diameter of the pores should be presented. Is there a mostcommon diameter / number of subunits of the ring-shaped pore that is the most stable, or do the pores just keep growing?
We would have expected more orderly structures. Using this experimental system, do other, better studied pore-forming proteins also assume such divers pore shapes? How representative is this of the physiological structure?
Why do they only test the E. coli membrane with and without cholesterol, not also the synthetic membrane?
What happens if GSDMD and Casp1 without the membrane being around and later added to the membrane? Is it important that the GSDMD N-term is "fresh"? Figure 2: do the different caspases cleave GSDMD at different sites? If not, the result is highly expected. The authors provide no indication on why this was an interesting question for them to test.
The authors claim that "the relatively wide distribution of the diameter of ring-shaped oligomers and the coexistence of arc-, slit-, and ring-shaped oligomers indicates that the assembly of transmembrane pores by GSDMDNterm is a structurally rather flexible process." -How can the exclude that these phenomena are not a result of their experimental settings and that the natural pore form in a more orderly fashion?
The protrusion (Z-Axis) should be depicted separately and quantified not only in Figure 1.
The use of supported lipid membranes on mica (a solid carrier) could be a problem if the GSDMD pore involves GSDMD molecules protruding from the membrane on the outer side. The authors should comment how this is dealt with in the field for other pore-forming proteins In Figure 4, the central part in some pores is obviously protruding higher than the surrounding ring. Why is this not quantified? Could this be a consequence of a secondary structure trying to form (that might be able to stabilize the ring and stop its growth)? The presence of GSDMD in the middle of the pore is even depicted in Figure 5 (final stage), but not sufficiently discussed. However, the model shown does not properly represent the obtained data because the different height of the protein in the center of the pore in the terminal stage is not shown.
If the authors want to reach the broad readership of the EMBO Journal, they should explain why they think their synthetic membrane is physiologically relevant and how it is composed. The abbreviations are not explained anywhere.
Why do the authors keep repeating the composition of their membrane including the rations throughout the manuscript if it is always the same? It is enough the just say "SLM" after it was first mentioned, unless it differs.
Reviewer #1: 2) In the time--lapse AFM analysis that tracks the process of GSDMD--N domain oligomerization and pore formation, the authors showed that the heights of oligomeric GSDMD--N domain protruding from the membrane surface did not change during pore formation. This observation indicates a "growing" process instead of involving a "prepore--to--pore transition" step seen with many other pore--forming proteins. This observation is interesting but needs to be strengthened by additional evidences. For example, the author could check whether a possible prepore of GSDMD--N domain could be induced in the solution state by adding phosphoinositide into caspase--cleaved GSDMD. Also a typical cholesterol--dependent cytolysin, which is known to form prepores in solution and undergo a vertical collapse on the membrane, could be included as a control.

Authors:
The reviewer suggests to further strengthen our finding that GSDMD Nterm forms pores in a "growing" rather than in a "prepore--to--pore transition" step. We made the following controls, as suggested by the reviewer, to test and strengthen our finding: 1) The reviewer suggests to test whether a possible prepore of GSDMD Nterm could be induced in the solution state by adding phosphoinositide into caspase--cleaved GSDMD (GSDMD Nterm ). Unfortunately, phosphoinositide is not water--soluble and we couldn't find any commercially available lysophosphatidylinositide (which could possibly be slightly water--soluble). To test whether GSDMD Nterm can form prepores before inserting into the lipid membrane we incubated GSDMD with caspase in the absence of any lipids at 37°C for 60 min, which was the incubation time used in our AFM--assays, and for much extended time overnight. After this incubation, we imaged the sample by AFM ( Fig. R1.3B, included as new Appendix

Reviewer #1:
3) Since other Gasdermin--family members also harbor the pore-forming activity, it is important to known whether a similar dynamic process of Gasdermin N--domain assembling into transmembrane pores also applies to other gasdermins. This data will increase the impact of the current study.
Authors: To further increase the impact of the study the reviewer suggests to characterize whether other members of the Gasdermin family show a similar process of pore formation as observed for GSDMD. We have started to record some data of the Gasdermin family member Gasdermin--A3 ( Fig. R1.7). The data shows that Gasdermin--A3 similarly to GSDMD assembles slit--, arc--and ring--like oligomers. No vertical differences of the Gasdermin--A3 oligomers could be observed, thus indicating the absence of a vertical collapse. The data also shows that the diameter of ring--like Gasdermin--A3 oligomers is much more constrained compared to the ring--like GSDMD oligomers. However, as our manuscript focuses on the oligomeric assembly and pore formation mechanism of human GSDMD, we would like to keep this focus throughout the paper. Extending the submitted work to other Gasdermin family members, such as Gasdermin--A3, would be beyond the scope of our study. After having discussed this query with the editor handling our paper at EMBO J, we have thus decided not to include this additional data showing the assembly and pore--forming mechanisms of Gasdermin--A3 into our paper. Once our paper has been accepted for publication, we will address this issue raised by the reviewer and search for similarities and individualities of the oligomeric assembly and pore forming mechanisms of other Gasdermin family members (including Gasdermin--A3).  Authors: Thank you for your encouraging and critical comments, which guided us to revise our manuscript. The reviewer suggested to extend the analysis and to conduct a considerable amount of additional experiments, which the reviewer specified below. These experiments, together with the experiments suggested by the other reviewers, considerably advanced our study compared to previous works, strengthened the manuscript and our scientific findings. For a detailed explanation of how we addressed each of the specific concerns of the reviewer, we kindly refer to our point--by--point response.

Reviewer #2:
For example, the idea of a kinetic mechanism for GSDMD pore formation is intriguing; however the conclusions should be supported by more robust evidence. The authors should provide statistical numbers of how many slit and arcs shapes turn into rings, how many rings form pores over time, whether the pore size increases over time before reaching the final size, whether slit and arcs pores are smaller than ring pores and whether increasing concentration of GSDMD affect all these parameters. They could also analyze what is the smallest size of slit and arc that can form a pore, and more interestingly, given the high quality of the  The reviewer also asked to analyze the smallest size of slit and arc that can form a pore and more interestingly, given the high quality of the resolution imaging, what is the minimum number of GSDMD molecules that can be associated with a membrane pore. Thank you for bringing up this interesting question. The smallest pore size (area) can be extracted from the pore size distributions given in Fig. R2.3. It appears that the smallest pore size we could detect in our study was ≈ 20 nm 2 . We have also analyzed the minimum, maximum and average number of GSDMD Nterm molecules assembling the arc--, slit--and ring--shaped oligomers forming transmembrane pores (see revised Fig. 5). In average, arc--like oligomers were assembled from 16.1 ± 0.4 (mean ± SE, n =26) subunits, slit--like oligomers from 20.7 ± 0.4 (n = 46) subunits and ring--like oligomers from 30.2 ± 0.2 (n = 49) subunits (revised Fig. 5D). We have included this information in the manuscript (see revised Results, section 'Sub--nanometer topographs of GSDMD Nterm oligomers' and revised Fig. 5).

AB2:/79C!!9&$!/$<)$>$/!*3;;$*1$5!1,!*300,/1!1&$!%,(%A3*),(*!,-!,3/!0'0$/!.2!+,/$!
Reviewer #2: Based on the relatively wide distribution of the ring diameter, the authors claim that GSDMD pore formation is a "structurally rather flexible process" (pag 6). This can be a result of a dynamic assembling process where the pore grows over time before reaching a steady state. It would be interesting to explore this hypothesis by, again, looking at the pore size over time, also in the context of understanding the nature of GSDMD pores (only protein or lipid--protein based Authors: To better understand the dynamic assembly process of the GSDMD pores, the reviewer asks to look at the pore size over time. We kindly refer to our answer given above and in which we analyzed the size (area) of the transmembrane pore formed by GSDMD Nterm oligomers over time ( Fig. R2.2, included as new Appendix Fig. S9). The analysis and the discussion hereof have been included in the revised manuscript.
The reviewer also suggests to plot a graph showing the number of slits/arcs and rings changing over time. In our time--lapse AFM experiments we imaged the supported lipid membranes (SLMs) while incubating them with a limited amount of GSDMD and caspase--1 at 37°C. Throughout recording the time--lapse movie, new GSDMD Nterm could insert and assemble oligomers. The graph plotting the number of arcs, slits and rings over time show that the number and arcs and slits changes over time approaching a lower plateau whereas the number of rings increases with time approaching a higher plateau ( Fig. R2.1). The new figure has been included into the Appendix and the data highlighting the flexible process of the oligomeric assembly have been discussed in the revision (see revised Results, section 'Imaging GSDMD Nterm oligomerization and pore formation', new Appendix  Authors: The reviewer asks to analyze the GSDMD assembly and pore formation in a lipid composition mimicking the inner leaflet of the plasma membrane, which contains ≈ 30% cholesterol. We followed the advice of the reviewer and conducted additional experiments in lipid compositions more closely mimicking the plasma membrane inner leaflet. Since plasma membranes display an asymmetric lipid distribution with PS and PE enriched in the cytosolic leaflet (Devaux & Morris, 2004;van Meer et al, 2008), we first used a three--component mixture of POPS, DOPE and POPC (35:25:40 molar ratio) as a model of the inner leaflet of the plasma membrane and investigated GSDMD assembly and pore formation on SLMs made from this lipid composition (Fig. R2.5). After this, since phosphoinositides are known to be present in mammalian cell membranes, and since PI(4,5)P2 is a marker of the cytoplasmic leaflet of the plasma membrane and a physiological relevant target of GSDMD, we characterized the effect of PI(4,5)P2 on the binding and assembly of GSDMD Nterm . For comparison, we have also investigated possible effects of phosphoinositol (POPI). Therefore, we have assembled SLMS from POPS, DOPE and POPC (35:25:40 molar ratio), from POPS, DOPE and POPI (35:25:40 molar ratio) and from POPS, DOPE and PI(4,5)P2 (35:25:40 molar ratio). As described for the other experiments, each of the SLMs was incubated with 0.5 µM GSDMD and 0.1 µM caspase--1 for 60 min at 37°C. The AFM topographs showed that GSDMD Nterm also bound to SLMs made from POPS, DOPE and POPC, where it assembled arc-- slit-- and ring--like oligomers. Again, the oligomers could form transmembrane pores. However, we could not observe GSDMD Nterm binding to SLMs made from POPS, DOPE and POPI. Instead, GSDMD Nterm bound much more frequently to SLMs made from POPS, DOPE and PI(4,5)P2 compared to SLMs made from POPS, DOPE and POPC. This indicates that PI(4,5)P2 promotes the binding and oligomerization of GSDMD Nterm . The data showed that the presence of PI(4,5)P2 had no influence on the diameter of ring-like oligomers. However, because the occurrence of arc--and slit--like oligomers reduced and of ring--like oligomers increased one may thus speculate that PI(4,5)P2 promotes the assembly of ring--like oligomers. The new data has now been included in our revised manuscript and discussed (see new   Appendix Table S1.

Point--by--point response to the comments of reviewer #3
Reviewer #3: Mulvihill study the insertion and pore--formation of GSDMD N--termini into membranes by AFM. They find that the structures formed are of different shapes and sizes. The AFM pictures are impressive and of very high quality and resolution. Taken that this is the strong part of the manuscript, the most should be made of this, e.g., the scale of height should be added and interpreted in all pictures. The main criticism is that it is difficult to judge how relevant the data is since a membrane on mica might not allow all conformations the protein might assume in a real membrane.
Authors: We thank the reviewer for the encouraging and constructive comments which guided us to revise our manuscript. The reviewer asks to provide the scale of the height information to all AFM images. We have revised the figure legends to define the vertical scale (height scale) of every AFM topograph shown.
The reviewer further comments that is difficult to judge how relevant the data is since a membrane on mica might not allow all conformations a protein assumes in a real membrane. We agree that supporting a membrane could indeed restrict the conformations of a membrane protein. However, within the past 20 years we have characterized more than 20 different membrane proteins in native and reconstituted membranes adsorbed to mica and could never observe such an artifact (Bippes & Muller, 2011;Engel & Muller, 2000). In this study, we have used different supported lipid membranes as model systems to investigate the assembly and pore formation of GSDMD and how the lipid composition of the membrane itself can influence the pore formation. To revise our manuscript, we have applied TEM imaging to characterize whether the assembly of GSDMD into arc--, slit-- and ring--like oligomers is an artifact caused by the mica supporting the lipid membranes (Fig. R3.1). In the control experiments, liposomes suspended in buffer solution were first incubated overnight with GSDMD and catalytic amounts of caspase--1 at 37°C. Afterwards, the liposomes were imaged by TEM. The TEM images show the co--existence of arc--, slit--and ring--like oligomers and thus support our AFM results obtained using supported lipid membranes. The data has been included and discussed in our revision (see revised Results, section GSDMD Nterm oligomerization in liposomes, and new Appendix Fig. S5).

Reviewer #3:
We would have expected more orderly structures. Using this experimental system, do other, better studied pore--forming proteins also assume such divers pore shapes? How representative is this of the physiological structure?
Authors: The reviewer asks whether other pore--forming proteins also assume such diverse pore shapes. Yes, other pore--forming proteins including cholesterol-dependent cytolysins or membrane attack complex/perforin (MACPF) also assemble arc--, slit-- and ring--like oligomers all of which giving rise to diverse pore shapes (Hodel et al, 2016;Leung et al, 2014;Leung et al, 2017;Mulvihill et al, 2015;Sonnen et al, 2014;van Pee et al, 2016). How these proteins assemble and form transmembrane pores can differ substantially. The surprising finding of our study is that GSDMD Nterm oligomers show very similar shapes as other pore forming proteins/toxins produced by bacteria. However, the growing mechanism of GSDMD Nterm into diverse oligomeric and pore shapes appears to be less common.
The reviewer further asked how representative the observed variety of GSDMD Nterm oligomers is for the physiological structure. Recent publications on GSDMD Nterm oligomers have already shown their ring and pore size to vary (Ding et al, 2016;Sborgi et al, 2016). It is thus not surprising that we also observe such variety. However, the structural variation of the arc--, slit--and ring--like GSDMD Nterm oligomers, the dynamic fusion of arcs and slits into ring--like oligomers and the observation that each of the oligomeric forms can form transmembrane pores is new.
We have revised our manuscript to briefly elaborate on both issues (see revised Discussion).

Reviewer #3: Why do they only test the E. coli membrane with and without cholesterol, not also the synthetic membrane?
Authors: The composition of some lipid membranes used in this work mimic that of an E. coli membrane. However, in our revision we have additionally tested several other lipid mixtures, which more closely mimic those of mammalian cells (Figs. R3.4, R3.5 and R3.6, now included as new Appendix Figs. S6 and S11 and new Fig. 2). The new experiments particularly tested the role of phosphoinositides and cholesterol ( Fig. 2 and new Appendix Fig. S6). In summary, we now have tested eleven different lipid mixtures for GSDMD binding, oligomerization and pore formation (summarized in new Appendix Table S3).
Taken together, the experiments show that if GSDMD Nterm binds and inserts, it forms arc--, slit--and ring--like oligomeric structures and transmembrane pores. The In addition, as replied further above to the reviewer, recent publications on GSDMD Nterm oligomers showed their ring and pore size to vary (Ding et al, 2016;Sborgi et al, 2016). It is thus not surprising that we also observe such variety of the oligomeric forms. Taken together, we have now investigated the GSDMD Nterm oligomer formation on membranes assembled from eleven different lipid compositions (new Appendix Table S3). The experiments show that the composition of the lipid membrane influences on whether GSDMD Nterm can bind, insert and assemble oligomers. However, our experiments also show that the arc--, slit--and ring--like oligomers assembled by GSDMD Nterm appear to be independent of the lipid composition and thus appear to reflect an intrinsic property of GSDMD Nterm . We have revised our manuscript to include the additional experimental data and control experiments and to explain this issue in detail (see revised Results and Discussion, new Fig.  3, new Appendix Figs. S6, S11 and new Appendix Table S3).

Reviewer #3:
The protrusion (Z--Axis) should be depicted separately and quantified not only in Figure 1. Reviewer #3: If the authors want to reach the broad readership of the EMBO Journal, they should explain why they think their synthetic membrane is physiologically relevant and how it is composed. The abbreviations are not explained anywhere.
Authors: We thank the reviewer. In our paper, we have shown that GSDMD Nterm inserts and assembles pore--forming arc--, slit--and ring--shaped oligomers in different model membranes. It is standard to the field that the formation of oligomeric pores is structurally characterized using synthetic lipid membranes. Following the reviewer's suggestion and to further strengthen our manuscript, we have included additional experiments showing that GSDMD Nterm forms arc--, slit-and ring--like transmembrane pores in a broad variety of different lipid membranes (new Fig. 3 and new Appendix Figs. S6 and S11). All together we have now tested eleven different lipid membrane compositions (summarized in Appendix Table  S3). In our new experiments, we have for example investigated the effect on GSDMD Nterm binding and oligomerization of phosphatidylinositide and cholesterol in a lipid mixture mimicking the cytoplasmic leaflet of the plasma membrane, the physiological target of GSDMD. The reasons why we decided to prepare our own lipid mixtures instead of using commercially available lipid extracts from mammalian sources are the following: (i) Lipid extracts are often "dirty". They are chloroform extracts of the respective tissue and contain not only lipids but also several other hydrophobic molecules that like to partition in the organic phase. This dirt tends to contaminate the AFM tip affecting the imaging quality and the resolution of the AFM topographs.
(ii) Preparing our own lipid mixtures allows us to address the effect of individual lipids one at a time and to take into account the asymmetric distribution of lipids in biological membranes.
For all lipid membranes, we have observed that if GSDMD Nterm binds to the lipid membrane it can also assemble slit--, arc--, and ring--like oligomers and transmembrane pores. We can thus conclude that the binding and insertion of GSDMD Nterm into lipid membrane depends on the lipid composition and that the oligomeric assembly described in our paper is a general property of GSDMD. We have included the new data showing the generality for the GSDMD Nterm binding, assembly and pore--forming mechanism and discussed it accordingly in our revised manuscript (see revised Results, revised Discussion, new Fig. 3 and new Appendix Figs. S6 and S11).
Furthermore, we have carefully revised our manuscript to introduce all abbreviations used, including those of the lipids.

Reviewer #3: Why do the authors keep repeating the composition of their membrane including the ratios throughout the manuscript if it is always the same?
It is enough the just say "SLM" after it was first mentioned, unless it differs.

Authors:
We have reduced redundancy of the descriptions as far as possible. However, in particular in the revised version of the manuscript, where we now characterize a larger variety of eleven different lipid mixtures, we feel that it is essential to keep on specifying repetitively which lipid membrane composition was characterized in which experiment. The new Appendix Table S3 gives an overview of the lipid compositions characterized.

Reviewer #3:
In the first sentence, "been" is missing, or it should be "was" instead of "has".
Authors: Thank you. The sentence has been corrected.