Receptor-linked environment-sensitive probe monitors the local membrane environment surrounding the insulin receptor

  1. Miwa Umebayashi
  2. Satoko Takemoto
  3. Luc Reymond
  4. Mayya Sundukova
  5. Ruud Hovius
  6. Annalisa Bucci
  7. Paul A. Heppenstall
  8. Hideo Yokota
  9. Kai Johnsson
  10. Howard Riezman

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

Functional membrane proteins in the plasma membrane are suggested to have specific membrane environments that play important roles to maintain and regulate the function of proteins. However, the local membrane environments of membrane proteins remain largely unexplored due to the lack of techniques allowing to monitor them in living cells. We have developed a method to probe the local membrane environment surrounding a membrane protein in the plasma membrane by covalently tethering a solvatochromic, environment-sensitive dye, Nile red, to a membrane protein via a flexible linker. Our direct imaging reported on the spatio-temporal properties of membrane fluidity of the local environment surrounding the insulin receptor. The local environment was distinct from the average plasma membrane fluidity and was quite dynamic and heterogeneous. Upon addition of insulin, the local membrane environment surrounding the receptor increased in fluidity in an insulin receptor-kinase dependent manner. This new technology should allow researchers to examine changes in membrane properties caused by receptor activation and devise ways to address the role of these changes in physiological processes.

Article activity feed

  1. Review Commons

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

    Learn more at Review Commons

    Reply to the reviewers

    Reviewer #1 (Evidence, reproducibility and clarity (Required)): **Summary:** Techniques to probe the local environment of membrane proteins are sparse, although the influence of lipids on the membrane protein's function are known since many years. Therefore, the paper by Umebayashi et al. is important. The environment-sensitive dye Nile red (NR) coupled to a membrane protein is an appropriate sensor for monitoring the local membrane fluidity. Linking of Nile red to the receptor via a flexible tether was achieved with the acyl carrier protein (ACP)-tag method. Experiments showed that depending on the ACP site a certain linker length is required to have NR inserted in the membrane and thus be an effective sensor for lipid disorder. This technology could be of general usability to study the environment of membrane proteins in the context of their function. As an example, the technique allowed insulin induced membrane disorder in the close insulin receptor vicinity to be observed. Further, results suggested that tyrosine activity is required for this disorder to happen. The experimental results appear to be complete and controls were made.

    **Major comments:**

    1. Sometimes technical terms are used without explanation: What is the GP value? What is ACP-IR? The spectrum was measured in number of rois? The reader can find those abbreveations out, but it would be nice to have them defined.

    We have made a list of abbreviations.

    1. Fig. 1d) is confusing. The ACP-IR labelling is evident in 3 panels, but there is no difference in the color (emission spectra of 1992-ACP-IR vs 2031-ACP-IR should be visible??). The DAPI staining is very different. When doing the latter, how difficult is it to get the staining equal?

    The differences in spectra cannot be seen because we used pseudo colors for display of the DAPI and CoA-PEG-NR staining. The reviewer’s comments about the unequal DAPI staining is correct. The reason for this is most likely that the cell membrane is unequally permeabilized by PFA treatment. As the point of this figure is just to show that the plasma membrane is labeled, dependent upon the expression of the ACP-tagged insulin receptor, we don’t think that the variable intensities of the DAPI staining is important. DAPI is simply used to indicate the position of the cells.

    1. How can one interpret Fig. 4: a) Control goes over 4 frames, at 240" insulin is added, and 10 frames should show a fluctuation difference?

    We showed 4 frames after control treatment that showed no significant change was observed by control treatment. We expected that clear changes would be invoked by insulin treatment in GP images, however these changes, while visible in the GP images, are difficult to see for the untrained observer. This is the reason why we used the ZNCC method in the subsequent figures to better visualize the changes.

    1. b) A color shift from blue to green is visible after insulin addition. But it is faint - difficult to assess from the pseudo color scheme. What does 1000 pixel top/1000 pixel bottom mean in c). Is it an attempt to better visualize the fluctuation? It is difficult to recognize a difference before and after adding insulin. d) It seems that the kymograph set should show this. What is the color scale? Why is 3 so untypical, i.e., no change? Box 6 is also peculiar: the left side does not show a strong change upon insulin administration, the right side does. Why? We appreciate the helpful comments for improving our manuscript.

    As pointed out, the change of GP value is extremely small before and after insulin addition, so it is difficult to fully visualize the change with normal pseudo-color expression. To deal with this, we adopted the following two methods to visualize minute changes.

    1. Visualization of local changes of the statistical GP value showed by ZNCC throughout the time-lapse images (Fig. 6 and Fig. S2B).

    2. Visualization of the top/bottom 1000 pixels of the sorting ZNCC value in each image (Fig. 7 and Fig. S2C). The top 1000 pixels are the ones that showed the largest changes. The bottom 1000 pixels are the ones that showed the smallest changes.

    Owing to these expressions, we found out that the level of the response against the insulin signal was spatially and temporally heterogeneous in the membrane.

    As for the color scale, in order to clarify the meaning of the difference of color, we have added the description about the relationship between the color and the ZNCC value in the results section.

    1. How is the kymogram calculated? The legend says 'The horizontal dimension represents the averaged ZNCC inside the rectangular area, and the vertical dimension represents time'. The averaged ZNCC is a single value, so it is not clear why the kymogram shows a variation from left to right. May it be the ZNCC was averaged just vertically?

    We apologize that we did not provide information regarding making the kymograph.

    In the yellow rectangular area (Fig. 6B), the ZNCC values of the pixels with the same x coordinate value were vertically averaged, which were represented as the horizontal direction of the kymograph. That is, one horizontal line of the kymograph holds the spatial distribution of the ZNCC value along the horizontal direction of the membrane, and the vertical direction shows their time changes. To make it easier to understand, we refined the description about the kymograph in the legend of Fig. 6.

    1. When calculating cross-correlation values on images, they need to be aligned. What fraction of the total image does the selected 19x19 box represent? As described, I imagine that a rolling CC over 19x19 pixels is calculated over an image from the time lapse series comparing it with the reference Iave(x,y). Compared to the 3x3 median filtered CP image, the ZNCC image should then be much more blurred??

    Below we provide more information regarding the calculation of ZNCC.

    Each local window for ZNCC calculation is set to a 19x19 pixels centered on every single pixel excluding the edges of an image. The ZNCC value calculated in that window is set to a center pixel of that area. After that, a new window centered on the adjacent pixel is set and calculate the new ZNCC. That is, the calculation window is slid throughout the image. Also, the calculated ZNCC value is not set to all the pixels of the window, but is set to only the center pixel of the window, so there is no blur effect like median filtering.

    The figure below shows a schematic view of our ZNCC calculation.

    Schematic view of our ZNCC calculation

    **Minor comment:** On page 16 supplementary is not spelled properly.

    corrected

    Reviewer #1 (Significance (Required)):

    The key point of this paper is convincing and the new technology appears to have a lot of potential. It can be applied to study membrane protein function in the context of its environment, the lipid bilayer.

    Membrane fluidity measurements have been developed (e.g., using fluorescent probes like laurdan). However, the trick to link a probe like nile red by ACP technology to the insulin receptor and to observe its activity is quite new.

    A most recent description of such a technology is in TrAC Trends in Analytical Chemistry Volume 133, December 2020, 116092.

    This is an interesting review, but not directly impacting on our work.

    **Referees cross-commenting**

    All comments are constructive and important. The paper is important but needs to be amended as proposed.

    Reviewer #2 (Evidence, reproducibility and clarity (Required)): **Summary:** In this manuscript, authors generated an ACP-attached Nile Red probe in order to specifically label Insulin receptor in the membrane. Owing to this specificity, one can measure the lipid membrane properties around a specific protein in the membrane. **Major comments:**

    For the conclusions in the manuscript to be convincing, in my opinion, these additional data need to be added. Some of these are new experiments, and some are detailed analysis of existing data. The new experiments are not for new line of investigation, instead it is to confirm their statements and conclusions. The major point is the reliability of spectral shift. In usual environment sensitive probes, it is certain that they are in the membrane whatever is done to the membrane. However, when the probe is attached to a protein, it is not trivial to have the same confidence that the probe is always inside the membrane, and it is in the same plane of the membrane. 1992-ACP-IR is a good example; authors state that it binds to the protein outside the membrane, but when there is cholesterol addition and -maybe more interestingly- cholesterol removal, the dye still reacts and changes its emission (even PreCT changes its emission quite a bit at the 570 nm region). This is a clear indication of a change in localization of the probe upon some changes in the membrane. This implies that observed spectral shifts may not be due to lipid packing differences, but due to localization of the probes. For this reason, it is crucial to know where any environment sensitive probe localize in the membrane with respect to membrane normal, and this knowledge is more important for this probe. Related to this, the spectral difference upon insulin treatment and activation of insulin receptor could be due to changes in probe's localization in the membrane. Especially because authors show in Fig1e, the spectra can change depending on the probe localization. Relatedly, quantum yield of NR should be significantly different when it is inside vs outside membrane. Authors should show QY for 1992-ACP-NR and 2031-ACP-NR with different PEG lengths and upon insulin treatment.

    We understand the logic of the request to measure the QY, since the QY of Nile red is much higher in organic solvents than in aqueous solutions, so it might be predicted that the QY of Nile red is higher in a lipid bilayer than when covalently bound to the protein in an aqueous environment. However, this argument depends upon the mechanism for the increase in quantum yield when going from aqueous to a non-polar solution. One possible explanation is based on the intrinsic properties of the dye under the two conditions. The alternative explanation would be that the dye would aggregate (be insoluble) in aqueous solution and therefore either not fluoresce or self-quench. In this case, we believe that the latter is the explanation because we and others have previously shown the turn-on properties of the probe when binding to proteins (SNAP-tag and others). It is not simple to measure QY in the cell under a microscope, but we have done something similar shown in supplementary figure 4. We labeled the three ACP-receptor complexes with PEG11-Nile red and co-stained with antibody to the Insulin Receptor. We then calculated a relative quantum yield. There were very little differences at all between the relative quantum yields, so we conclude that it is not the environment of the probe, which affects the quantum yield under these conditions, but the fact that it is covalently attached to a protein and incapable of forming aggregates. What distinguishes these constructs is the emission spectrum, not the quantum yield. In supplementary Table 2 we also did QY measurements in vitro and we could reproduce the increase of quantum yield by association with liposomes or in organic solvents. We tested whether non-covalent association with a protein would increase the QY by incubation with the lipid binding protein, BSA, in PBS. This was not the case, strongly pointing to the conclusion that it is the covalent association with the protein that increases the QY, not association with a protein. We believe that our demonstration of changes in fluorescent spectra with changes in cholesterol, large changes in fluorescent spectra with linker length for the 1992 construct and voltage sensitivity using patch-clamp prove that the Nile red is reporting on the membrane environment under the conditions we propose.

    **Minor comments:**

    • Fig 1d requires quantification We do not agree on this. This is simply to show that the labeling is dependent upon expression of the relevant ACP-IR constructs. There is no detectable labeling of the control.

    • Voltage sensitivity of different PEG length of 2031-ACP probe should be added. We have added this data in figure 2 panel E.

    • Fig 3a graph should show all data points, not only bar graphs. Also, the band in 3a for +CoA-PEG-NR is dimmer than other bands, is it specific to this particular gel since quantification does not show any difference?

    There is no significant difference- Fig 4d, colour code is needed.

    Done

    • Fig 5b and Fig3d are basically the same experiments in terms of control measurement, why is the difference in 3b is 0.04 GP unit while it is 0.007 GP unit?

    We explain in the MS, but have improved the title of Y-axis in Fig.5 b graph so that the difference in what is plotted is clear.

    • Why is inhibitor data so noisy? We should discuss.

    We don’t know the exact reason why inhibitor data is noisy, but we speculate that the actin cytoskeleton and phosphoinositide-dependent signaling could affect the membrane stability, and the membrane environment would be fluctuated in the presence of latrunculin B or PI3K inhibitor.

    Reviewer #2 (Significance (Required)): Overall, this is a very useful approach, and this line of research will yield very useful tools to shed light on how lipids surrounding proteins can change their function. Major advance of the paper is the new chemical biology tool. There is also biological data on how insulin can change the insulin receptor's membrane environment which is contradictory to some old literature claiming that InsR becomes more "rafty" upon insulin treatment (e.g., PMID: 11751579).

    If this type of tagging proves robust and reproducible (limitations and concerns listed above and below), it could be used by other researchers to tag their protein of interest and investigate the lipid environment around those proteins.

    The downside of this method is that the probe requires ACP tag, a relatively less used tag than others in biology, therefore researchers interested in using this probe should have their proteins with ACP tag. Moreover, the linker length and ACP-tag position are quite crucial parameters (and probably should be optimized for each protein). Longer PEG lengths cannot report on changes efficiently (Fig3b), while shorter lengths are prone to artefacts as they can go out of membrane (Fig1 and Fig2). This might limit its widespread use.

    The reason for using the ACP tag is that neither the SNAP tap nor the HALO tag working. The tethered Nile Red preferred to bind to the tqg rather than inserting into the membrane.

    **Referees cross-commenting** I agree with all comments and concerns of other reviewers. I see the usability and potential of this new technology along with its limitations as all three reviewers pointed out.

    Reviewer #3 (Evidence, reproducibility and clarity (Required)): See below. No concerns on any of these issues.

    Reviewer #3 (Significance (Required)): **Critique:** This MS reports a proof-of-principle for using site-directed environmentally sensitive probe technology to assess the local membrane environment of a receptor tyrosine kinase (IR) upon activation. This technology addresses a major gap in our arsenal of tools to study the mechanisms of membrane signaling as the parameters of interest are biophysical parameters rather than purely biochemical ones. How to do this with spatial and temporal resolution is a major challenge. This study builds on previous work by the Riezman group that develops an extrinsic labeling system to tether Nile Red to specific sites on the ectodomain of a signaling receptor and then probe local membrane environments as a function of receptor activity. This is a carefully done study is well-controlled, is clever in design and is well-described. Although the major issues to which such a general technology could contribute involve intracellular (and not extracellular) event, the advances described will be of general interest -- particularly that local membrane order decreases when IR becomes activated. Specific comments for the authors' consideration follow:

    **Specific Comments:** (i) As a general comment, the authors are measuring extracellular plasma membrane leaflet properties that may or may not translate to what is happening in the local inner leaflet environment. A general reader may well miss the significance of this. This point needs to be more explicitly emphasized in the Discussion.

    This has been discussed in the revised version.

    (ii) Why not treat cells with a PLC inhibitor to block PIP2 hydrolysis and ask if that inhibits membrane disorder. It is PIP2 hydrolysis/resynthesis that regulates the actin cytoskeleton at signaling receptors and this seems an attractive candidate for study.

    There is a long list of attractive post-signaling events of the insulin receptor and how this works in different cell types that could be tested. We believe that this is beyond the scope of this study and we encourage others to do this.

    (iii) The data acquisition time is at least 4 min which is long enough for activated receptors to be recruited to sites of endocytosis. Can the authors exclude the possibility that what they are measuring isn't reflective of such spatial reorganization? Does a clathrin inhibitor block the observed change in local membrane order for activated IR? We determined localization to AP2 adaptor containing clathrin coated pits at the cell surface and showed that during the time-course of the experiment that there is no significant change in co-localization or evidence for endocytosis (new figure 9). Therefore, we decided not to do the clathrin inhibitor blocking experiment because we believe that it could only lead to indirect effects.

    (iv) Receptor activation is accompanied by other transitions such as dimerization, etc. Can the authors exclude the possibility that what they are measuring is related to changes in depth of insertion of the NR probe into the plasma membrane outer leaflet that is a consequence of IR conformational transitions associated with activation? This is highly unlikely given the fact that fluidification of the membrane environment is found with all length linkers. Given the intervals in increases in linker length on the 2031 construct, which is the closest to the membrane, it is very difficult to conceive that any of the ones larger than 5 PEGs restrict significantly the membrane insertion of the dye. **Referees cross-commenting**

    I think we have a consensus opinion

  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

    See below. No concerns on any of these issues.

    Significance

    Critique:

    This MS reports a proof-of-principle for using site-directed environmentally sensitive probe technology to assess the local membrane environment of a receptor tyrosine kinase (IR) upon activation. This technology addresses a major gap in our arsenal of tools to study the mechanisms of membrane signaling as the parameters of interest are biophysical parameters rather than purely biochemical ones. How to do this with spatial and temporal resolution is a major challenge. This study builds on previous work by the Riezman group that develops an extrinsic labeling system to tether Nile Red to specific sites on the ectodomain of a signaling receptor and then probe local membrane environments as a function of receptor activity.

    This is a carefully done study is well-controlled, is clever in design and is well-described. Although the major issues to which such a general technology could contribute involve intracellular (and not extracellular) event, the advances described will be of general interest -- particularly that local membrane order decreases when IR becomes activated. Specific comments for the authors' consideration follow:

    Specific Comments:

    (i) As a general comment, the authors are measuring extracellular plasma membrane leaflet properties that may or may not translate to what is happening in the local inner leaflet environment. A general reader may well miss the significance of this. This point needs to be more explicitly emphasized in the Discussion.

    (ii) Why not treat cells with a PLC inhibitor to block PIP2 hydrolysis and ask if that inhibits membrane disorder. It is PIP2 hydrolysis/resynthesis that regulates the actin cytoskeleton at signaling receptors and this seems an attractive candidate for study.

    (iii) The data acquisition time is at least 4 min which is long enough for activated receptors to be recruited to sites of endocytosis. Can the authors exclude the possibility that what they are measuring isn't reflective of such spatial reorganization? Does a clathrin inhibitor block the observed change in local membrane order for activated IR?

    (iv) Receptor activation is accompanied by other transitions such as dimerization, etc. Can the authors exclude the possibility that what they are measuring is related to changes in depth of insertion of the NR probe into the plasma membrane outer leaflet that is a consequence of IR conformational transitions associated with activation?

    Referees cross-commenting

    I think we have a consensus opinion

  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

    Summary:

    In this manuscript, authors generated an ACP-attached Nile Red probe in order to specifically label Insulin receptor in the membrane. Owing to this specificity, one can measure the lipid membrane properties around a specific protein in the membrane.

    Major comments:

    For the conclusions in the manuscript to be convincing, in my opinion, these additional data need to be added. Some of these are new experiments, and some are detailed analysis of existing data. The new experiments are not for new line of investigation, instead it is to confirm their statements and conclusions. The major point is the reliability of spectral shift. In usual environment sensitive probes, it is certain that they are in the membrane whatever is done to the membrane. However, when the probe is attached to a protein, it is not trivial to have the same confidence that the probe is always inside the membrane, and it is in the same plane of the membrane. 1992-ACP-IR is a good example; authors state that it binds to the protein outside the membrane, but when there is cholesterol addition and -maybe more interestingly- cholesterol removal, the dye still reacts and changes its emission (even PreCT changes its emission quite a bit at the 570 nm region). This is a clear indication of a change in localization of the probe upon some changes in the membrane. This implies that observed spectral shifts may not be due to lipid packing differences, but due to localization of the probes. For this reason, it is crucial to know where any environment sensitive probe localize in the membrane with respect to membrane normal, and this knowledge is more important for this probe. Related to this, the spectral difference upon insulin treatment and activation of insulin receptor could be due to changes in probe's localization in the membrane. Especially because authors show in Fig1e, the spectra can change depending on the probe localization. Relatedly, quantum yield of NR should be significantly different when it is inside vs outside membrane. Authors should show QY for 1992-ACP-NR and 2031-ACP-NR with different PEG lengths and upon insulin treatment.

    Minor comments:

    • Fig 1d requires quantification
    • Voltage sensitivity of different PEG length of 2031-ACP probe should be added.
    • Fig 3a graph should show all data points, not only bar graphs. Also, the band in 3a for +CoA-PEG-NR is dimmer than other bands, is it specific to this particular gel since quantification does not show any difference?
    • Fig 4d, colour code is needed.
    • Fig 5b and Fig3d are basically the same experiments in terms of control measurement, why is the difference in 3b is 0.04 GP unit while it is 0.007 GP unit?
    • Why is inhibitor data so noisy?

    Significance

    Overall, this is a very useful approach, and this line of research will yield very useful tools to shed light on how lipids surrounding proteins can change their function. Major advance of the paper is the new chemical biology tool. There is also biological data on how insulin can change the insulin receptor's membrane environment which is contradictory to some old literature claiming that InsR becomes more "rafty" upon insulin treatment (e.g., PMID: 11751579).

    If this type of tagging proves robust and reproducible (limitations and concerns listed above and below), it could be used by other researchers to tag their protein of interest and investigate the lipid environment around those proteins.

    The downside of this method is that the probe requires ACP tag, a relatively less used tag than others in biology, therefore researchers interested in using this probe should have their proteins with ACP tag. Moreover, the linker length and ACP-tag position are quite crucial parameters (and probably should be optimized for each protein). Longer PEG lengths cannot report on changes efficiently (Fig3b), while shorter lengths are prone to artefacts as they can go out of membrane (Fig1 and Fig2). This might limit its widespread use.

    Referees cross-commenting

    I agree with all comments and concerns of other reviewers. I see the usability and potential of this new technology along with its limitations as all three reviewers pointed out.

  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:

    Techniques to probe the local environment of membrane proteins are sparse, although the influence of lipids on the membrane protein's function are known since many years. Therefore, the paper by Umebayashi et al. is important. The environment-sensitive dye Nile red (NR) coupled to a membrane protein is an appropriate sensor for monitoring the local membrane fluidity. Linking of Nile red to the receptor via a flexible tether was achieved with the acyl carrier protein (ACP)-tag method. Experiments showed that depending on the ACP site a certain linker length is required to have NR inserted in the membrane and thus be an effective sensor for lipid disorder. This technology could be of general usability to study the environment of membrane proteins in the context of their function. As an example, the technique allowed insulin induced membrane disorder in the close insulin receptor vicinity to be observed. Further, results suggested that tyrosine activity is required for this disorder to happen. The experimental results appear to be complete and controls were made.

    Major comments:

    1. Sometimes technical terms are used without explanation: What is the GP value? What is ACP-IR? The spectrum was measured in number of rois? The reader can find those abbreveations out, but it would be nice to have them defined.

    2. Fig. 1d) is confusing. The ACP-IR labelling is evident in 3 panels, but there is no difference in the color (emission spectra of 1992-ACP-IR vs 2031-ACP-IR should be visible??). The DAPI staining is very different. When doing the latter, how difficult is it to get the staining equal?

    3. How can one interpret Fig. 4: a) Control goes over 4 frames, at 240" insulin is added, and 10 frames should show a fluctuation difference? b) A color shift from blue to green is visible after insulin addition. But it is faint - difficult to assess from the pseudo color scheme. What does 1000 pixel top/1000 pixel bottom mean in c). Is it an attempt to better visualize the fluctuation? It is difficult to recognize a difference before and after adding insulin. d) It seems that the kymograph set should show this. What is the color scale? Why is 3 so untypical, i.e., no change? Box 6 is also peculiar: the left side does not show a strong change upon insulin administration, the right side does. Why?

    4. How is the kymogram calculated? The legend says 'The horizontal dimension represents the averaged ZNCC inside the rectangular area, and the vertical dimension represents time'. The averaged ZNCC is a single value, so it is not clear why the kymogram shows a variation from left to right. May it be the ZNCC was averaged just vertically?

    5. When calculating cross-correlation values on images, they need to be aligned. What fraction of the total image does the selected 19x19 box represent? As described, I imagine that a rolling CC over 19x19 pixels is calculated over an image from the time lapse series comparing it with the reference Iave(x,y). Compared to the 3x3 median filtered CP image, the ZNCC image should then be much more blurred??

    Minor comment:

    On page 16 supplementary is not spelled properly.

    Significance

    The key point of this paper is convincing and the new technology appears to have a lot of potential. It can be applied to study membrane protein function in the context of its environment, the lipid bilayer.

    Membrane fluidity measurements have been developed (e.g., using fluorescent probes like laurdan). However, the trick to link a probe like nile red by ACP technology to the insulin receptor and to observe its activity is quite new.

    A most recent description of such a technology is in TrAC Trends in Analytical Chemistry Volume 133, December 2020, 116092.

    Referees cross-commenting

    All comments are constructive and important. The paper is important but needs to be amended as proposed.

  5. Version published to 10.1101/2020.12.23.424145v1 on bioRxiv