Mammalian MemPrep establishes the lipid composition of ER membranes in HEK293T cells

  1. Aamna Jain
  2. Alexander von der Malsburg
  3. Claudia Götz
  4. Mohamed Elmofty
  5. John Reinhard
  6. Per Haberkant
  7. Volkhard Helms
  8. Joseph H. Lorent
  9. Robert Ernst

Reviewed by

Read the full article See related articles

Discuss this preprint

Start a discussion What are Sciety discussions?

Listed in

Log in to save this article

Abstract

The endoplasmic reticulum (ER) forms a dynamic network of sheets and tubules, whose molecular lipid composition remains incompletely defined. Using an optimized MemPrep workflow, we establish a high-confidence lipidome of the mammalian ER and selectively enrich membrane vesicles originating from its major structural subdomains. Quantitative lipidomics show that ER membranes are dominated by phosphatidylcholine and mono-unsaturated glycerophospholipids, consistent with a highly compressible bilayer. Although proteomics reveals a clear segregation of sheet-and tubule-associated proteins in line with a functional specialization, their lipidomes are nearly identical, indicating that these principal ER architectures share a common lipid environment. Integration of lipidomic data with bioinformatic analyses of transmembrane helices further demonstrates that the physicochemical features of ER lipids mirror those of ER-resident membrane proteins, including reduced hydrophobicity and increased polarity compared to plasma membrane proteins. These findings support a coordinated evolution of ER proteins and lipids based on shared biophysical constraints. Together, this work provides a definitive characterization of the mammalian ER lipidome and clarifies how conserved membrane properties are maintained across structurally distinct ER subdomains.

Article activity feed

  1. Review Commons

    Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

    Learn more at Review Commons

    Reply to the reviewers

    The authors adapt MemPrep, a protocol they originally developed to purify organelle membranes from yeast, for use in human cell lines. To this end, they established immuno-isolation strategies based on tagged versions of the ER sheet protein SEC61β and the ER tubular protein REEP5 in HEK293T cells. Their purification strategy allowed them to generate highly pure ER sheet- and tubule-enriched fractions, which were then subjected to quantitative lipidomic and proteomic analyses.

    Overall, this manuscript is well written and presents a careful interpretation of the data. It introduces MemPrep in mammalian cells as a method that will be useful for studying the membrane lipid and protein composition of organelles, with a particular focus on the ER. As such, the manuscript provides sufficient information and controls to assess the experiments in terms of reproducibility and clarity.

    We thank the reviewer for a positive, thorough assessment and for raising important points that helped us to improve the manuscript.

    Major comments:

    1. Based on the immunofluorescence images in Figure 1, it is not clear that the tagged and slightly overexpressed versions of SEC61β and REEP5 localize specifically to ER sheets and tubules, respectively, or that these proteins are enriched in these distinct ER subdomains. Perhaps reducing the fixation time, for example to a maximum of 2 minutes, or using PFA fixation, could help to better preserve ER sheet and tubular domains.

    To address the localization of the bait proteins in the ER membrane network, we added new co-localization microscopy data and quantifications to the revised manuscript (new Figure 1E,F; new Supplementary Figure S1C,D). Despite its low level of overexpression (new Figure 1C; new Suppl. Fig. S1A), SEC61β localizes to the entire ER membrane network including ER tubules and the nuclear envelope (new Fig. 1E,F).

    Considering the new data, we have carefully rephrased all sections regarding the subcellular localization of bait-SEC61β. In the revised manuscript, we use SEC61β as a general ER marker.

    Intriguingly, quantitative proteomics of the SEC61β MemPrep isolate demonstrates a selective enrichment of ER sheet-associated proteins compared to the REEP5 MemPrep, which selectively enriches proteins associated with ER tubules (Fig. 5). While we do not claim to 'isolate' ER subdomains, we enrich ER subdomains.

    We have performed additional microscopy experiments and adjusted our fixation protocol as suggested by the reviewer (Revision Fig. 1). Shortening the fixation time has no apparent impact on the ER structure, while any PFA fixation seems to largely disrupt the ER.

    Does expression of tagged SEC61β or REEP5 influence the ER sheet:tubule ratio? In addition, does expression of these constructs affect the lipidome or proteome of the cells?

    The reviewer raises an important point, which is experimentally not easy to address. Our imaging modality is not sufficient to make a firm statement about the sheet:tubule ratio in HEK293T cells. We are not aware of any study that firmly quantifies the relative content of sheets and tubules in HEK293T cells. Imaging the ER in HEK293T cells is challenging and most studies on the ER membrane networks use other cell types to study the impact of ER-shaping protein on the ER membrane network.

    In the revised manuscript we state: 'We found no evidence that the expression of the bait constructs disrupts the tubule-to-sheet ratio or other aspects of the ER architecture, but distinguishing ER sheets and ER tubules is challenging in HEK293T cells.'

    Furthermore, we have studied if the expression of the bait constructs affects the cellular proteome (new Suppl. Fig. S1A,B) and lipidome (new Suppl. Fig. S4A-H (previously Suppl. Fig. S3)). The expression of the bait constructs has no substantial impact of the cellular proteome. Most importantly, we find no evidence that proteins characteristic for ER sheets or ER tubules (other than the bait proteins) change their expression level (new Suppl. Fig. S1A,B). In the revised manuscript we state:

    ' We decided to go one step further and compared the proteomes of wildtype HEK293T cells with the two cell lines using TMT multiplexed, untargeted protein mass spectrometry (Suppl. Fig. S1A, B). This experiment revealed that bait proteins have only a minimal, neglectable impact on the cellular proteome (Suppl. Fig. S1A, B). We did not find evidence for a systematic deregulation of proteins known to localize exclusively to ER tubules or other ER subdomains. Furthermore, quantitative proteomics validated the results from immunoblotting (Fig. 1B, C): Expression of bait-SEC61β has barely any impact on the total cellular level of SEC61β (Suppl. Fig. S1A) while the expression of the REEP5-bait results in a 1.8-fold overabundance of REEP5 (Suppl. Fig. S1B).'

    Likewise, the expression of the bait constructs has little to no effect on the cellular lipidome as shown in Suppl. Fig. S4A-J. In the revised manuscript we state:

    'As a control, we also tested the impact of the bait constructs on the HEK293T whole cell lipidome (Suppl. Fig. 4A-J). Overall, the lipid composition of the virally transduced cells was indistinguishable from HEK293T cells with only minor impact on the level of CL and lysolipids (Suppl. Fig. 4A-J).'

    Apart from hypotonic swelling and douncing, could the authors use alternative methods for cell disruption to exclude the possibility that mechanical stress confounds the interpretation of the data?

    Thanks to the reviewer's comment, we became aware of a mistake. Our cell lysis buffer is hypertonic and not hypotonic (15% sucrose w/v, 10 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid)(HEPES) pH 7.4, 300 mM NaCl, 1 mM EDTA freshly supplemented with protease inhibitor cocktail from Roche). We have corrected all relevant sections in the revised manuscript.

    The reviewer is right that different means of mechanical lysis, and/or the incubation of the cells in hypo/hypertonic buffer are likely to have impact on the structure of the ER and to affect the isolation procedure. Changing such critical parameters will likely affect the purity of the preparation. Performing additional MemPrep isolations using different means of cell disruptions goes beyond the scope of this manuscript.

    Upon establishing the MemPrep protocol, we have explored various mechanical cell disruptions: Different cannula, Dounce homogenizers, and a ball-bearing device. We experimented with both hypo- and hypertonic buffers. Given the costs and work associated with lipidomic and proteomic analyses, we have tried to find a suitable conditions for cell disruption without performing a full analysis each time. Therefore, we performed differential centrifugations as exemplary shown in Fig. 2B of the manuscript. Critical factors for our decision whether to further persue a certain condition was 1) the depletion of the mitochondrial TOM22 marker, 2) the enrichment of the ER markers, and 3) the total protein yield in the P100,000 fraction.

    In the revised manuscript we state: 'Compared to the MemPrep procedure in yeast, we tested various means of cell disruption and optimized the differential centrifugation protocol.'

    and

    'Mild cell disruption by Dounce homogenization in a hypertonic buffer is crucial for cracking cells open, but these procedures can disrupt normal ER architecture and might facilitate the undesired mixing of previously well-defined ER subdomains. Despite these limitations, our data underscore the purity of our ER membrane preparations, demonstrate a differential enrichment of ER subdomains (Fig. 5), and establish the lipid composition of the ER membrane (Fig. 6)'.

    What is the total amount of lipids and proteins isolated with REEP5- or SEC61β-based MemPrep? Are there differences in the total lipid:protein ratio between these isolates, and could this reflect differences in the ER sheet:tubule ratio?

    In response to the reviewers' question, we have included a new Supplementary table 1 to the manuscript outlining the yield of total protein and total lipid of MemPrep.

    The mammlian MemPrep protocol is not yet optimized for determining the lipid:protein ratio in the membrane. At this moment, we do not want to make a statement about the protein-to-lipid ratio in the ER or its subdomains. The isolates still contain material originating from the ER lumen.

    The combined analysis of lipid and protein composition demonstrates the capacity of the method. To test that MemPrep can capture changes in ER membrane architecture, it would be useful to compare ER protein and lipid composition across different cellular states, such as stressed versus unstressed cells, or growing versus resting cells.

    We agree with the reviewer that a comparison of the ER under different conditions would be extremely interesting. Currently, we see it beyond the scope of this study.

    Minor comment:

    In line 335, the authors state: "To address this possibility, we performed a new round of REEP5 and SEC61β MemPreps for a direct comparison of the isolates (Fig. 5A, B)." It is unclear whether the MemPrep protocol was altered or whether this refers simply to an additional round of purification. Please clarify.

    Thank you. This point was also raised by reviewer 2 and 3. We have clarified our statements. In the revised manuscript we state:

    'Hence, we performed a new round of REEP5 and SEC61β MemPreps in triplicates for a direct comparison of the isolates (Fig. 5A, B) rather than comparing the changes in abundance relative to the respective cell lysates as performed in Figure 3. Knowing that non-ER proteins are less efficiently enriched by the MemPrep procedure than ER proteins (Fig. 3C, D) and that the sensitivity and comprehensiveness of mass spectrometry-based proteomics experiments are reduced with increasing sample complexity (Ting et al, 2011; Beck et al, 2011) , we were hoping to gain a better insight into the distribution of low abundant and challenging to quantify proteins in the two MemPrep isolates'.

    Reviewer #1 (Significance (Required)):

    General assessment:

    The manuscript establishes MemPrep for mammalian cells as an important discovery tool to investigate how cells coordinate membrane lipid composition with membrane protein composition, and vice versa. This is a rapidly growing research field, which attracts a lot of interest.

    MemPrep is based on an immuno-isolation strategy using tagged versions of the ER sheet protein SEC61β and the ER tubular protein REEP5 in HEK293T cells. The purification strategy allowed to generate highly pure ER sheet- and tubule-enriched fractions, which were then subjected to quantitative lipidomic and proteomic analyses.

    The results show that the protein composition differs between the SEC61β- and REEP5-enriched fractions. Yet the lipid composition of ER sheets and tubules is largely indistinguishable. Both fractions are dominated by PC alongside other monounsaturated GPL, and hydroxylated ceramides. These physicochemical properties of the ER lipid bilayer are matched by ER-resident membrane proteins.

    Thorough bioinformatic analysis of a subset of ER membrane proteins further revealed that their transmembrane domains have reduced hydrophobicity and increased polarity compared with those of plasma membrane proteins, matching the ER lipidome.

    Hence the combined analysis of lipid and protein composition demonstrates the capacity of the method. Many variations of this approach will be possible in the future to understand on the molecular level how cells assemble and control their membranes.

    Advance: Other immuno-isolation methods, or "organelle immunoprecipitation" approaches, have been established for lysosomes, the Golgi apparatus, and other organelles.

    MemPrep is an important and complementary addition to the technical toolbox for organelle isolation, with a particular focus on the analysis of membrane lipid and protein content.

    Audience: The manuscript will be of broad interest to researchers in basic biology as well as clinical and translational research.

    Reviewer's field of expertise:

    Molecular membrane biology.

    __Reviewer #2 __

    Jain and colleagues develop a biochemical fractionation procedure in which ER microsomes are enriched through small epitope tags. The manuscript is pitched around the concept that there are ER sheets and tubules and ER proteins differentially localise to them. The authors use REEP5 as a 'tubule' bait and SEC61beta as a 'sheet' bait. These baits are immuoisolated after a sensible membrane fractionation and ER membraned purified. There is a convincing ER proteome as a result, and this is used to compare the TMD properties of the organelles resident membrane proteins. The authors make the interesting observation that the transmembrane domains are more polar in the ER. They then compare the two sheet and tubule preparations and see a different in the proteome, before comparing the lipidome. There is no difference observed between the lipidome of the sheet and tubule preps, however they see a difference in the whole cell lysate and use that to compare the ER lipidome against the whole cell.

    Overall the manuscript has an interesting premise and the data is well presented, the experiments well performed and the interpretations appropriate. I think there are some issues with the mechanistic insight and novelty, and essentially although the premise is with regards to sheets and tubules there is limited progress in that direction in terms of results. I am reluctant to be to critical overall as there are certainly interesting observations that may be insightful for future studies in the field. I have some more specific comments below:

    We thank the reviewer for a thorough, constructive assessment and for highlighting important points that helped us improve the manuscript.

    1. The authors cite nixon-abell, but they do not mention the major point of that manuscript which is that the 'sheets' in the cellular periphery are instead dense tubular networks. I think this is quite an omission for the introduction, as it points to the premise not being as clear as stated.

    In the revised manuscript we refer to the Nixon-Abell study and two additional studies from the Jokitalo lab. Notably, the Nixon-Abell study does not rule out the existence of ER sheets.

    In the revised manuscript we state: ' *[...] *dense tubular networks in the cell periphery can appear like ER sheets in diffraction-limited microscopy (Nixon-Abell et al, 2016). Furthermore, the edges of ER sheets are populated by curvature-stabilizing proteins also found in ER tubules (Shibata et al, 2010; Shemesh et al, 2014), and ER sheets show different degrees of fenestration dependent on the cell type and the cell cycle phase (Puhka et al, 2007, 2012; Nixon-Abell et al, 2016). Consistent with our microscopic data (Fig. 1E, F) and because ER sheets may be biochemically inseparable from ER tubules, we use SEC61β as a general ER marker.'

    We performed additional co-localization studies of the bait proteins with RTN4 and CLIMP63 (new Fig. 1E,F) suggesting that SEC61B can localize across many ER subdomains including ER tubules and the nuclear envelope.

    We have carefully revised our manuscript accordingly and shifting the focus of our discussion away from a molecular description of discrete ER subdomains.

    1. The first section when the protocol is discussed essentially relies on looking at other papers to understand. As the manuscript is centrally about this protocol, I think a brief but clear description is more appropriate.

    We agree with the reviewer. We added a short section to the results section providing an overview over the MemPrep procedure. We now state:

    'To this end, we adapted the MemPrep procedure originally developed for the isolation of organelle membranes from Saccaromyces cerevisiae (S. cerevisiae) (Reinhard et al, 2023, 2024). Mammalian MemPrep relies on a gentle, detergent-free, mechanical lysis of the cells in a hypertonic buffer followed by differential centrifugation to separate ER-derived microsomes from mitochondria-derived membranes. Next, larger organelle fragments are disrupted by brief pulses of sonication, and the resulting vesicles are subjected to affinity purification using magnetic dynabead-coupled antibodies directed against the cleavable tag of the bait protein. Specifically bound, ER-derived membrane vesicles are washed with harsh, urea-containing buffers and selectively released by proteolytically cleaving the bait tag.'

    1. In figure 1C the two markers are supposed to localise to sheets and tubules differentially. To me they look very similar. This, of course, is a major concern. Have the authors co-expressed them (at the same levels in these lines) and seen that indeed they do differentially localise?

    The reviewer raises an important point regarding the localzation of the bait proteins. While we have not co-expressed the bait proteins in cells, we have performed additional co-localization experiments with RTN4 and CLIMP63 as markers for ER tubules and ER sheets, respectively (new Figure 1E,F; new Suppl. Fig. S1C,D). The implications of these data are discussed in the manuscript.

    In light of these new data, we do not refer to SEC61β as an ER sheet marker any longer, instead we refer to SEC61β as a general ER marker. We carefully revised our discussion of the data throughout the manuscript along the line suggested by the reviewer in point 8.

    1. I found the TMD polarity section very interesting, but it was not clear to me why they needed their proteomics for this? Could this not be done with annotated ER membrane proteins?

    The reviewer is correct. The same type of analysis could have been performed with an even bigger dataset of all ER annotated proteins. One of the co-authors, Joseph Lorent, has performed such analysis at this larger scale (PMID: 40326394). The study by Lorent et al. addressed TMH length and side chain bulkiness (PMID: 40326394) in the ER, Golgi apparatus, and the PM. This work is referenced in the manuscript.

    We focused our analysis on the smaller dataset of 83 single-pass proteins found in our proteomics experiments, because we initially planned to perform a comparative analysis of ER proteins in either of the two isolates.

    In line of the reviewers' suggestion, we validate our new finding on the TMH hydrophobicity in the ER using a larger dataset covering all single pass TMHs of ER proteins (215 instead of 83), Golgi apparatus proteins (260), and plasma membrane proteins (1322) (Suppl. Fig. S3D).

    1. It was not clear to me based on the results section text the difference between the figure 5 proteomics and the previous runs.

    This point was also raised by reviewer 1 and 3. We clarified our statement in the revised manuscript:

    'Hence, we performed a new round of REEP5 and SEC61β MemPreps in triplicates for a direct comparison of the isolates (Fig. 5A, B) rather than comparing the changes in abundance relative to the respective cell lysates as performed in Figure 3. Knowing that non-ER proteins are less efficiently enriched by the MemPrep procedure than ER proteins (Fig. 3C, D) and that the sensitivity and comprehensiveness of mass spectrometry-based proteomics experiments are reduced with increasing sample complexity (Ting et al, 2011; Beck et al, 2011) , we were hoping to gain a better insight into the distribution of low abundant and challenging to quantify proteins in the two MemPrep isolates.'

    1. Again in figure 5- are the authors sure that the difference was not due to the over-expression (albeit mild) of their protein.

    After performing an important control experiment, we are sure that the mild over-expression of the bait proteins has no impact.

    We have compared HEK293T WT cells with the bait protein expressing cell lines by quantitative proteomics (new Suppl. Fig. S1A,B). The bait proteins have no impact of the cellular proteome and do not affect the abundance of proteins known to be enriched in ER sheets or ER tubules. Hence, the enrichment of these proteins in our MemPrep isolates as shown in Fig. 5 suggests that some of the identity of ER sheets and ER tubules is maintained in our preparations even though they are not resolved by our microscopy experiments (Fig. 1). In the revised manuscript, we carefully discuss the implications of these findings.

    1. There were no differences in the ER lipidome between the two baits. This may be because there is no difference between the lipid profile of sheets and tubules, but it is very hard to conclude that.

    The reviewer has a point. Even though our findings suggest that we can differentially enrich for ER subdomains (the proteomics data in Fig. 5 on MemPrep isolates can be regarded as a golded standard for this statement), we do not have any knowledge about their biochemical purity. Hence, we have carefully toned down our statements on the basis of new imaging data (Fig. 1E,F; Suppl. Fig. S1C,D) and new proteomics data (Suppl. Fig. S1A,B).

    Along the reasoning of the reviewer, we also rephrased our statements on the difference/similarity of ER subdomains.

    1. I do not see it as my job as a reviewer to propose reorganisations and rewrites, so I encourage the authors to feel free to ignore this comment. To me the lipidome and TMD polar observations are the key manuscript findings, and there is very limited insight into the tubules and sheets line of inquiry. I wonder if it would be worth changing the focus of the manuscript overall to rather be about the ER, and not the tubules and sheets.

    Again, the reviewer raises an important point that we did not want 'to ignore'. We have carefully revised the manuscript and toned down our interpretations. In the revised manuscript we put more emphasis on the ER lipidome and less so on the composition of specific ER subdomains.

    __Reviewer #2 (Significance (Required)): __

    Overall the manuscript has an interesting premise and the data is well presented, the experiments well performed and the interpretations appropriate. I think there are some issues with the mechanistic insight and novelty, and essentially although the premise is with regards to sheets and tubules there is limited progress in that direction in terms of results. I am reluctant to be to critical overall as there are certainly interesting observations that may be insightful for future studies in the field.

    Reviewer #3

    Summary: Jain et al., provide a clear and thorough manuscript that extends their prior biochemical analysis of the yeast ER-lipidome (MEMPREP) to mammalian cells. They use detergent free lysis and differential speed centrifugation from 293T cells bearing reporters with affinity handles targeted to sheet-like or tubular-like subdomains of the ER and enrich membranes and membrane-embedded proteins from these sites. The lipidomics reveals a distinct ER-lipidome, heavily enriched in PC and PI, contains predominantly mono-unsaturated phospholipids and is surprisingly invariant across sheet-like and tubule-like domains. Additional hydrophobicity analysis suggests that ER-localised TMDs are more polar and shorter than PM-resident TMDs, and the authors speculate about co-evolution of the lipidome and proteome to ensure targeting.

    Major comments:

    I think the data are solid, clear and convincing. The similarity of the lipidomes from sheet and tubule regions of the ER give good indication of the robustness of the technique. Whilst the yield is low, the authors go to good lengths to demonstrate purity of ER capture and de-enrichment of other cellular membranes. There is good discussion of the limitations of the technique and good comparison to recent data from other labs, most notably, a recent preprint and I think the manuscripts support eachother well. There's a fair amount of speculation in the manuscript, e.g., about lipid headgroup charge density being inferred by the charge distribution on the -1 position, but the speculation is clearly acknowledged.

    I think that blotting for SEC61B would really help. A clear comparison to endogenous SEC61B would be helpful. I appreciate that the authors lacked an antibody here, but there are several on CiteAb that seem to detect endogenous protein.

    Following the reviewers' advice, we added new data using a commercial antibody directed against SEC61β (new Fig. 1C). We also added proteomics data comparing HEK293T WT cells with the bait expressing cell lines (new Suppl. Fig. S1A,B).

    We also characterized the commercial Proteintech (15087-1-AP) antibody to make sure it recognizes the same epitopes in the tagged and untagged variant of SEC61β.

    It's not brilliantly easy to see the 'sharp decline' in relative frequency of hydrophobic amino acids at 21 aa for ER and Golgi; whilst the individual amino acid information is interesting (and some comment could be made about the favouring of Leucines in ER and Golgi TMDs), would this be clearer if the relative frequencies were binned into hydrophobic/aromatic, polar, positive, negative?

    The reviewer is right. We have removed our statement regarding a 'sharp decline'. In fact, the decline is rather gradual for ER and Golgi TMHs, but more clear for PM TMHs. This is also reflected in the data shown in Suppl. Fig. S3D and discussed in the revised manuscript.

    We state: Confirming our expectations based on the predicted TMH length (Suppl. Fig. S3A), we observed a gradual decline in the relative frequency of hydrophobic and aromatic resides at about 21 amino acids for ER (Fig. 4E) and Golgi-associated TMHs (Fig. 4F). Such decline was more clearly defined for plasma membrane TMHs but only after 24 aa or more (Fig. 4G).'

    We also state: 'We therefore challenged our finding and performed an additional analysis using this larger dataset of all annotated human single-pass TMHs (Fig. S3D) and compared the hydrophobicity profiles of TMHs from the ER (215), the Golgi apparatus (260), and the PM (1322) (Lorent et al, 2025). This analysis further substantiated our finding that the ER and the Golgi apparatus host less hydrophobic TMHs compared to the plasma membrane. Furthermore, we observed that the ER and Golgi profiles display a conical shape with hydrophobic maxima at the center of the membrane's hydrophobic core, while the PM TMH's possess higher hydrophobicity in the cytoplasmic part of the membrane, compared to the exoplasmic part (Fig. S3D).'

    We decided to keep the Fig. 4 with its single amino acid 'resolution' was it was in the original manuscript, because we feel that this representation still has its value. It helps connecting physicochemical parameters of an average TMH in an organelle (Fig. 4A-D; Suppl. Fig. S3A-D) with the preferred amino acid composition and distribution (Fig. 4E-G). Nevertheless, some 'noise' in inherent to the data and we hope that the adaptations to the text avoids any possible confusion of the reader.

    The frequency of leucine residues in TMHs from the PM (24.5%) is comparable to the frequency of TMHs from the ER (24.1%) and from the Golgi apparatus (26.3%). Our attempts to identify an organelle-selective usage of certain amino acids did not yield robust and significant results.

    Related to this point, it's hard to correlate the degree of polar amino acid incorporation in the TMDs of Golgi, ER, PM proteins (which don't appear to vary in 4E, 4F and 4G) with the variance described in 4C. Is there a better way of displaying this data, or are the polarity measurements calculated by some other metric in 4C?

    The reviewer is right. Figure 4A-D and Figure 4E-G are based on different metrics. Figure 4A-D considers different physicochemical parameters of the amino acid sidechains (Fig. 4C: Kyte-Dolittle scale). Figure 4E-G only represents the relative frequencies. We believe that both representations can be useful.

    Notably, the relative incorporation of polar and apolar amino acids is significantly different between TMHs from the ER and the Golgi versus the TMHs from the PM (Suppl. Fig. S3B,C).

    In the revised manuscript we state: 'Our new finding that the TMHs of ER proteins are more polar than the TMHs in the plasma membrane (Fig. 4C) is also reflected by the significantly different number of apolar and polar residues in the TMHs from ER-, Golgi apparatus-, and PM-derived proteins (Suppl. S3B, C)'.

    Indeed, the polarity in Fig. 4A and Fig. 4C is calculated via the Kyte-Dolittle scale, while only the normalized frequency of the amino acid is color-coded in Fig. 4E-G.

    Minor comments:

    Panel 2D isn't labelled on the figure

    We represented both MemPreps in a single Panel 2C because we aimed to label in the immunoblots only a single time to avoid redundancies. We are open to change our strategy of panel labeling if our current representation is confusing.

    There is limited co-enrichment of non-ER proteins in the ER-affinity preps, and the authors have done well to deal with misannotated GO terms. It might be worthwhile adding to the discussion that all TMD proteins that localise at steady-state to post-ER compartments must necessarily pass through the ER during biosynthesis. As such, detection of non-ER proteins in ER fractions is not inherently unexpected.

    This is of course correct. In the revised manuscript we state: 'Finding non-ER proteins in an ER proteome is not surprising, because a very large number of proteins are first delivered to the ER, before they are sent to other cellular destinations.'

    I didn't understand the line on L377 about the new round of extraction featureing inherently less complex proteomes.

    This point was also raised by reviewer 1 and 2. We clarified our statement in the revised manuscript:

    'Hence, we performed a new round of REEP5 and SEC61β MemPreps in triplicates for a direct comparison of the isolates (Fig. 5A, B) rather than comparing the changes in abundance relative to the respective cell lysates as performed in Figure 3. Knowing that non-ER proteins are less efficiently enriched by the MemPrep procedure than ER proteins (Fig. 3C, D) and that the sensitivity and comprehensiveness of mass spectrometry-based proteomics experiments are reduced with increasing sample complexity (Ting et al, 2011; Beck et al, 2011) , we were hoping to gain a better insight into the distribution of low abundant and challenging to quantify proteins in the two MemPrep isolates.'

    For line L390-391, in the speculation about progressively more unsaturation as you move ER-Golgi-postGolgi, is there any (published) data from ER-FLIPPR that could inform about the degree of membrane fluidity/packing as you traverse the secretory pathway?

    We agree that mentioning evidence on the biophysical changes along the secretory pathway is helpful in this section. In the revised manuscript we state:

    'These changes of the lipid acyl chains are associated with biophysical changes of the membrane properties along the secretory pathway as observed by molecular probes reporting on lipid packing and membrane tension (Goujon et al, 2019; López-Andarias et al, 2021, 2022; Wong & Budin, 2024).'

    Reviewer #3 (Significance (Required)):

    The strengths of the study are the conceptual novelty and information provided - I think this is the first comprehensive reporting of the ER lipidome. This is a major organelle and I think as the lipid biology field develops, resources like this are really important. Moreover, the MEMPREP protocol is applicable for protein extraction from these domains, which will help with functional characterisation of ER subdomains and is a strong technical advance.

    Weaknesses relate to the single cell type and overexpression (albeit mild) methodologies. I'm not hugely fussed about this as this manuscript describes an important 1st step.

    I'm a cell biologist studying the ER

  2. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #3

    Evidence, reproducibility and clarity

    Summary:

    Jain et al., provide a clear and thorough manuscript that extends their prior biochemical analysis of the yeast ER-lipidome (MEMPREP) to mammalian cells. They use detergent free lysis and differential speed centrifugation from 293T cells bearing reporters with affinity handles targeted to sheet-like or tubular-like subdomains of the ER and enrich membranes and membrane-embedded proteins from these sites. The lipidomics reveals a distinct ER-lipidome, heavily enriched in PC and PI, contains predominantly mono-unsaturated phospholipids and is surprisingly invariant across sheet-like and tubule-like domains. Additional hydrophobicity analysis suggests that ER-localised TMDs are more polar and shorter than PM-resident TMDs, and the authors speculate about co-evolution of the lipidome and proteome to ensure targeting.

    Major comments:

    I think the data are solid, clear and convincing. The similarity of the lipidomes from sheet and tubule regions of the ER give good indication of the robustness of the technique. Whilst the yield is low, the authors go to good lengths to demonstrate purity of ER capture and de-enrichment of other cellular membranes. There is good discussion of the limitations of the technique and good comparison to recent data from other labs, most notably, a recent preprint and I think the manuscripts support eachother well. There's a fair amount of speculation in the manuscript, e.g., about lipid headgroup charge density being inferred by the charge distribution on the -1 position, but the speculation is clearly acknowledged.

    1. I think that blotting for SEC61B would really help. A clear comparison to endogenous SEC61B would be helpful. I appreciate that the authors lacked an antibody here, but there are several on CiteAb that seem to detect endogenous protein.
    2. It's not brilliantly easy to see the 'sharp decline' in relative frequency of hydrophobic amino acids at 21 aa for ER and Golgi; whilst the individual amino acid information is interesting (and some comment could be made about the favouring of Leucines in ER and Golgi TMDs), would this be clearer if the relative frequencies were binned into hydrophobic/aromatic, polar, positive, negative?
    3. Related to this point, it's hard to correlate the degree of polar amino acid incorporation in the TMDs of Golgi, ER, PM proteins (which don't appear to vary in 4E, 4F and 4G) with the variance described in 4C. Is there a better way of displaying this data, or are the polarity measurements calculated by some other metric in 4C?

    Minor comments:

    1. Panel 2D isn't labelled on the figure
    2. There is limited co-enrichment of non-ER proteins in the ER-affinity preps, and the authors have done well to deal with misannotated GO terms. It might be worthwhile adding to the discussion that all TMD proteins that localise at steady-state to post-ER compartments must necessarily pass through the ER during biosynthesis. As such, detection of non-ER proteins in ER fractions is not inherently unexpected.
    3. I didn't understand the line on L377 about the new round of extraction featureing inherently less complex proteomes.
    4. For line L390-391, in the speculation about progressively more unsaturation as you move ER-Golgi-postGolgi, is there any (published) data from ER-FLIPPR that could inform about the degree of membrane fluidity/packing as you traverse the secretory pathway?

    Significance

    The strengths of the study are the conceptual novelty and information provided - I think this is the first comprehensive reporting of the ER lipidome. This is a major organelle and I think as the lipid biology field develops, resources like this are really important. Moreover, the MEMPREP protocol is applicable for protein extraction from these domains, which will help with functional characterisation of ER subdomains and is a strong technical advance.

    Weaknesses relate to the single cell type and overexpression (albeit mild) methodologies. I'm not hugely fussed about this as this manuscript describes an important 1st step.

    I'm a cell biologist studying the ER

  3. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #2

    Evidence, reproducibility and clarity

    Jain and colleagues develop a biochemical fractionation procedure in which ER microsomes are enriched through small epitope tags. The manuscript is pitched around the concept that there are ER sheets and tubules and ER proteins differentially localise to them. The authors use REEP5 as a 'tubule' bait and SEC61beta as a 'sheet' bait. These baits are immuoisolated after a sensible membrane fractionation and ER membraned purified. There is a convincing ER proteome as a result, and this is used to compare the TMD properties of the organelles resident membrane proteins. The authors make the interesting observation that the transmembrane domains are more polar in the ER. They then compare the two sheet and tubule preparations and see a different in the proteome, before comparing the lipidome. There is no difference observed between the lipidome of the sheet and tubule preps, however they see a difference in the whole cell lysate and use that to compare the ER lipidome against the whole cell.

    Overall the manuscript has an interesting premise and the data is well presented, the experiments well performed and the interpretations appropriate. I think there are some issues with the mechanistic insight and novelty, and essentially although the premise is with regards to sheets and tubules there is limited progress in that direction in terms of results. I am reluctant to be to critical overall as there are certainly interesting observations that may be insightful for future studies in the field. I have some more specific comments below:

    1. The authors cite nixon-abell, but they do not mention the major point of that manuscript which is that the 'sheets' in the cellular periphery are instead dense tubular networks. I think this is quite an omission for the introduction, as it points to the premise not being as clear as stated.
    2. The first section when the protocol is discussed essentially relies on looking at other papers to understand. As the manuscript is centrally about this protocol, I think a brief but clear description is more appropriate.
    3. In figure 1C the two markers are supposed to localise to sheets and tubules differentially. To me they look very similar. This, of course, is a major concern. Have the authors co-expressed them (at the same levels in these lines) and seen that indeed they do differentially localise?
    4. I found the TMD polarity section very interesting, but it was not clear to me why they needed their proteomics for this? Could this not be done with annotated ER membrane proteins?
    5. It was not clear to me based on the results section text the difference between the figure 5 proteomics and the previous runs.
    6. Again in figure 5- are the authors sure that the difference was not due to the over-expression (albeit mild) of their protein.
    7. There were no differences in the ER lipidome between the two baits. This may be because there is no difference between the lipid profile of sheets and tubules, but it is very hard to conclude that.
    8. I do not see it as my job as a reviewer to propose reorganisations and rewrites, so I encourage the authors to feel free to ignore this comment. To me the lipidome and TMD polar observations are the key manuscript findings, and there is very limited insight into the tubules and sheets line of inquiry. I wonder if it would be worth changing the focus of the manuscript overall to rather be about the ER, and not the tubules and sheets.

    Significance

    Overall the manuscript has an interesting premise and the data is well presented, the experiments well performed and the interpretations appropriate. I think there are some issues with the mechanistic insight and novelty, and essentially although the premise is with regards to sheets and tubules there is limited progress in that direction in terms of results. I am reluctant to be to critical overall as there are certainly interesting observations that may be insightful for future studies in the field.

  4. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #1

    Evidence, reproducibility and clarity

    Summary:

    The authors adapt MemPrep, a protocol they originally developed to purify organelle membranes from yeast, for use in human cell lines. To this end, they established immuno-isolation strategies based on tagged versions of the ER sheet protein SEC61β and the ER tubular protein REEP5 in HEK293T cells. Their purification strategy allowed them to generate highly pure ER sheet- and tubule-enriched fractions, which were then subjected to quantitative lipidomic and proteomic analyses.

    Overall, this manuscript is well written and presents a careful interpretation of the data. It introduces MemPrep in mammalian cells as a method that will be useful for studying the membrane lipid and protein composition of organelles, with a particular focus on the ER. As such, the manuscript provides sufficient information and controls to assess the experiments in terms of reproducibility and clarity.

    Major comments:

    1. Based on the immunofluorescence images in Figure 1, it is not clear that the tagged and slightly overexpressed versions of SEC61β and REEP5 localize specifically to ER sheets and tubules, respectively, or that these proteins are enriched in these distinct ER subdomains. Perhaps reducing the fixation time, for example to a maximum of 2 minutes, or using PFA fixation, could help to better preserve ER sheet and tubular domains.
    2. Does expression of tagged SEC61β or REEP5 influence the ER sheet:tubule ratio? In addition, does expression of these constructs affect the lipidome or proteome of the cells?
    3. Apart from hypotonic swelling and douncing, could the authors use alternative methods for cell disruption to exclude the possibility that mechanical stress confounds the interpretation of the data?
    4. What is the total amount of lipids and proteins isolated with REEP5- or SEC61β-based MemPrep? Are there differences in the total lipid:protein ratio between these isolates, and could this reflect differences in the ER sheet:tubule ratio?
    5. The combined analysis of lipid and protein composition demonstrates the capacity of the method. To test that MemPrep can capture changes in ER membrane architecture, it would be useful to compare ER protein and lipid composition across different cellular states, such as stressed versus unstressed cells, or growing versus resting cells.

    Minor comment:

    1. In line 335, the authors state: "To address this possibility, we performed a new round of REEP5 and SEC61β MemPreps for a direct comparison of the isolates (Fig. 5A, B)." It is unclear whether the MemPrep protocol was altered or whether this refers simply to an additional round of purification. Please clarify.

    Significance

    General assessment:

    The manuscript establishes MemPrep for mammalian cells as an important discovery tool to investigate how cells coordinate membrane lipid composition with membrane protein composition, and vice versa. This is a rapidly growing research field, which attracts a lot of interest.

    MemPrep is based on an immuno-isolation strategy using tagged versions of the ER sheet protein SEC61β and the ER tubular protein REEP5 in HEK293T cells. The purification strategy allowed to generate highly pure ER sheet- and tubule-enriched fractions, which were then subjected to quantitative lipidomic and proteomic analyses.

    The results show that the protein composition differs between the SEC61β- and REEP5-enriched fractions. Yet the lipid composition of ER sheets and tubules is largely indistinguishable. Both fractions are dominated by PC alongside other monounsaturated GPL, and hydroxylated ceramides. These physicochemical properties of the ER lipid bilayer are matched by ER-resident membrane proteins.

    Thorough bioinformatic analysis of a subset of ER membrane proteins further revealed that their transmembrane domains have reduced hydrophobicity and increased polarity compared with those of plasma membrane proteins, matching the ER lipidome.

    Hence the combined analysis of lipid and protein composition demonstrates the capacity of the method. Many variations of this approach will be possible in the future to understand on the molecular level how cells assemble and control their membranes.

    Advance: Other immuno-isolation methods, or "organelle immunoprecipitation" approaches, have been established for lysosomes, the Golgi apparatus, and other organelles.

    MemPrep is an important and complementary addition to the technical toolbox for organelle isolation, with a particular focus on the analysis of membrane lipid and protein content.

    Audience:

    The manuscript will be of broad interest to researchers in basic biology as well as clinical and translational research.

    Reviewer's field of expertise:

    Molecular membrane biology.

  5. Version published to 10.64898/2026.03.20.713196 on bioRxiv