Super-Resolution Imaging of Nuclear Pore Responses to Mechanical Stress and Energy Depletion
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThe manuscript by Torres-Rivera et al. presents an SMLM-based high-resolution microscopy study of nuclear pore complex structure in U2OS cells under various experimental conditions.
The perspective of this work is to investigate potential NPC alterations induced by the nuclear import of HIV-1 viral cores. To support their hypothesis the authors cite recent CLEM cryo-electron tomography and molecular simulation studies reporting NPC disruption induced by the nuclear import of HIV-1 viral cores.
The study is technically sound, and the authors provide high-quality reconstructed images demonstrating the eight-fold symmetry of Nup96 localization in the NPC, comparable to those published by the team of J. Ries. Presented imaging approach appears appropriate and I have no point to rise concerning the technical aspects of the NPC high resolution imaging and reconstruction. To validate the experimental model the authors show that the presence of HaloTag fused to NUP96 does not affect the infections kinetics, although it reduces the number of viral cores that enter in the cell nucleus .
In the Results section, the authors show that no changes in pore diameter are observed in ATP-depleted cells or in cells subjected to hypo or hyperosmotic shock. These observations provide new information about NPC plasticity under these conditions, however, their link to viral infection is not clearly established.
As this article is intended for a special issue focused on microscopy methods in virus research, I suggest adding a paragraph explaining the principle of SLML based microscopy and a simple schematic of the nuclear pore complex showing NUP96-Halo labeling. I also suggest commenting on key parameters for sample preparation and high quality image acquisition and reconstruction. In the perspective section, it could be relevant to mention HaloTag-DNA PAINT study (DOI: 10.1002/anie.201905685) that achieves superior spatial resolution.
In conclusion, this article presents the SMLM imaging of nuclear pores as a tool to investigate the impact of the infection on cellular structures and to elucidate the mechanism of nuclear entry of HIV-1 capsids. The SMLM based imaging of nuclear pores during the infection represents interesting alternative to electron microscopy and may help elucidate NPC modifications induced by the nuclear import of viral cores. Presented results showing that NPC structure is preserved under energy depletion and mechanical stress are original and worthy of publication.
Author Response
Rev 1:
These observations provide new information about NPC plasticity under these conditions, however, their link to viral infection is not clearly established.
A: We agree. Unfortunately, our attempts to image HIV cores docked at the nuclear pore were not successful, primarily because of difficulties with delivering a large number of fluorescent cores into intact cells. This is an active direction of research, but we currently have not progressed beyond a few apparent cases of HIV-1 core docking at the nuclear pore (see below).
As this article is intended for a special issue focused on microscopy methods in virus research, I suggest adding a paragraph explaining the principle of SLML based microscopy and a simple schematic of the nuclear pore complex showing NUP96-Halo labeling.
A: We thank the reviewer for this suggestion. The following paragraph introducing the principles of SMLM is added:
“In single molecule localization microscopy (SMLM), molecules transition between dark and emissive states which are localized in each time frame and fitted with a Point Spread Function model. The fluorescent events are spatially sparsely distributed. Thus, the position of a molecule can be pinpointed with much higher accuracy than the diffraction limit. From the estimated positions of many thousands of localizations, a super resolution image can be reconstructed (1, 2).”
We thank the reviewer for suggesting a cartoon of NUP96-SNAP labeling. The new Fig. 2A now shows the molecular model of the nuclear pore with the NUP96 positions marked.
I also suggest commenting on key parameters for sample preparation and high quality image acquisition and reconstruction.
A: We added a comment on sample preparation in Methods:
“A mapping calibration with less than 20 nm of Standard Deviation of Errors was accepted. After calibration, drift correction was tested by imaging coverslip-adhered TetraSpeck beads with 60 ms exposure time (16 frames per second) using 640 nm and 561 nm lasers, with 100 frames each and with a 45o (semiTIRF) illumination angle. Calibration/drift correction was confirmed by visual inspection of both colors colocalizing. After calibration, cells were imaged using 60 ms exposure time in a semiTIRF mode …”
For visualizing STORM images, drift correction, filtering and exporting of localization coordinates, we used NimOS software, and the cloud-based CODI (https://alto.codi.bio) program developed by ONi. Drift correction was performed using the built-in drift correction feature in NimOS. Localizations were filtered by removing localizations with less than 500 photons and with more than 5000 photons. In addition, events with localization precision worse than 20 nm were discarded. Averaging of NPCs were performed using a MATLAB script developed by the Rieger lab (3, 4). For this analysis, each NPC exhibiting symmetric circular appearance was manually picked from the field of view by cropping the point cloud with 180x180 nm boxes.”
In the perspective section, it could be relevant to mention HaloTag-DNA PAINT study (DOI: 10.1002/anie.201905685) that achieves superior spatial resolution.
A: The reference to the DNA-PAINT study is added. We also discuss the use of this technique as a future direction.
Reviewer 2 Report
Comments and Suggestions for AuthorsThe study by Torres-Rivera et al. uses 2D dSTORM to investigate the architecture of the NPC using available U2OS cells with endogenous Nup96-SNAP knock-in. These cells have previously been thoroughly characterized and serve as “performance” standard for super-resolution microscopy. The authors analyze the Nup96-SNAP ring radius by averaging hundreds of NPCs following 3 perturbations: energy depletion, osmotic swelling, and osmotic shrinking of the nucleus. The presented STORM data and performed analyses seem generally of high quality. However, the study lacks positive controls, that show that an expected change of a few nm (see the comment below) using the authors imaging system and analyses pipeline is able to differentiated between these states. Without these controls, the current data cannot draw any conclusions on NPC diameters.
The authors expect a ~ 20 nm change in diameter, or 10 nm in radius, which they claim would be resolvable with ~13 nm resolution based on FCC analysis. This 10 nm radius change was observed by Zimmerli et al. for the inner ring and cytoplasmic ring diameter in yeast using cyro-ET. However, Nup96 sits in the cytoplasmic and nuclear ring, which according to the (cited) cryo-ET study by Zimmerli et al. changes less upon energy depletion: about 10 nm radius change for the cytoplasmic ring, but only ~ 4 nm for the nuclear ring. Since the authors cannot distinguish between these rings in their 2D STORM data, a convolution of both with maybe ~ 7 nm radius change would be expected as maximum effect in yeast. Importantly, not all NPCs showed this response with significant heterogeneity, expecting the mean radius change to be much lower. Similarly, the authors should take this heterogeneity into consideration and differentiate between contracted and dilated NPCs. In summary, this effect is not resolvable with a claimed 13 nm resolution and therefore the data in this study cannot draw any conclusions on NPC architecture upon perturbation.
However, the averaging procedure should be able to distinguish mean structural changes with a higher resolution - akin to sub-tomogram averaging in ET data. Previously, it was possible to differentiate ~ 1 nm differences obtained in averaged structures in PMID 23845946 of yeast NPCs. To show this, the authors need to have positive controls that reproducibly changes the diameter, e.g. by knock-down or knock-out of NPC components, or by labeling another NPC component.
Furthermore, it is completely unclear what the (negative) results presented in this manuscript have to do with HIV infection. I would encourage the authors to include necessary controls, and either remove the HIV part with resubmission of the study to a more suitable journal, or perform the STORM analyses of NPCs under infection conditions with labelled IN. For these experiments positive controls will also be necessary.
Minor comments:
Line 27-28: What does the transport of transcription factors have to do with the study?
Line 54-56: The cited preprint by Hou et al. is now published PMID: 40093063. Perhaps some of the most convincing evidence for structurally intact/conical capsid structures deeper inside the nucleus away from the NPC – which suggest successful transport of at least some intact capsids – are described for cell lines in PMID 39623137 and 33904396 and for primary macrophages in PMID 41326693.
Line 78-80: The authors did not tag these cells but used a cell line created by colleagues. Please use precise language here.
Figure 2: What does raw storm images mean? Rendering of localization? What precision? Visualization of point list?
Line 154-156: The authors write that NE volume does not change upon ATP depletion. They show data for this in in 2 replicates (Fig. 2D and Fig S4F). No difference in Fig2D and a 25 % lower nuclear volume in Fig S4F for NaN3 treatment. The difference may be statistically significant with a p value below 0.05, but that does not mean its biologically relevant. It is better to state the magnitude of the effect instead of only using the word significant – e.g. the NE volume was ~ 25 % lower in one of two replicates. Please also use precise language here, and include the nuclear size analysis with the data of the 3rd replicate to make this statement more robust.
Figure 4 A: Please don’t show images with saturated pixels in the GFP channel. The quantification below even shows a ~40% lower intensity of GFP inside the nucleus as opposed to the NE. This should be reflected in the image presentation.
Author Response
Rev 2:
… the study lacks positive controls, that show that an expected change of a few nm (see the comment below) using the authors imaging system and analyses pipeline is able to differentiate between these states. Without these controls, the current data cannot draw any conclusions on NPC diameters. …The authors need to have positive controls that reproducibly changes the diameter, e.g. by knock-down or knock-out of NPC components, or by labeling another NPC component.
A: This is a great point. We have tried hard to get a positive control that shows a resolvable change in the NPC size. In fact, the ATP depletion and osmotic stressing experiments were attempted for exactly that purpose. Unfortunately, we could not detect significant changes in the NPC diameter after these interventions. We also failed to knock-in a SNAP tag into other NUPs in U2OS cells, despite trying two different CRISPR-Cas9-based approaches. Another strategy we have tried was to treat cells with PitStop2, which has been reported to disrupt the NPC structure (5–7). We ended up dropping this approach due to PitStop2 cytotoxicity. We agree that lack of a positive control is a problem but would argue that our results indicate an unexpected resilience of mammalian NPCs to different challenges.
…the authors should take this heterogeneity into consideration and differentiate between contracted and dilated NPCs.
A: Regarding the pore heterogeneity, we do not see any evidence in the data for two states. From Figures 2 and 3, the found radii of the pores are smooth Gaussian distribution with one peak independent of the condition. Therefore, a blind data driven classification of the pores into classes will not give anything. Only if we enforce two classes, will that produce two classes of pores, but that does not seem to be the correct way forward. In the past we have been investigating data driven heterogeneity in SMLM data (which is quite a bit different from EM where sorting into projection classes is done already) in the following publications, however, the current data do not suggest the existence of two classes.
[1] S. Haghparast, Y. Zhang, Q. Tao, S. Stallinga, and B. Rieger. Detecting continuous structural heterogeneity in single molecule localization microscopy data with a point cloud variational auto-encoder. Scientific Reports, 2026.
[2] S. Haghparast, S. Stallinga, and B. Rieger. Detecting continuous structural heterogeneity in single-molecule localization microscopy data. Scientific Reports, 13:19800, 2023.
[3] T.A.P.M. Huijben, H. Heydarian, A. Auer, F. Schueder, R. Jungmann, S. Stallinga, and B. Rieger. Detecting structural heterogeneity in single-molecule localization microscopy data. Nature Communications, 12:3791, 2021.
However, the averaging procedure should be able to distinguish mean structural changes with a higher resolution - akin to sub-tomogram averaging in ET data. Previously, it was possible to differentiate ~ 1 nm differences obtained in averaged structures in PMID 23845946 of yeast NPCs.
A: The reviewer is correct; the averaging procedure is able to detect changes in the average pore radii to a much higher degree than the localization uncertainty and the spread in the distribution. The mechanism is different from EM as in SMLM we have to deal with incomplete labelling and adding more pores does not just increase the resolution as in EM with the square root of the number of pores. The standard error of the mean of the radii estimations per condition on the average is and this means indeed we should be able to see small effects (see also the p-values). Given the data we think that there is no measurable effect, and we think that the algorithm is correct based on simulations and earlier work. At this point, we are unsure as to why no differences in the average NPC size were seen after ATP depletion or osmotic stressing. As the reviewer pointed out, the lack of positive control does not allow drawing definitive conclusions regarding our resolving power.
Furthermore, it is completely unclear what the (negative) results presented in this manuscript have to do with HIV infection. I would encourage the authors to include necessary controls, and either remove the HIV part with resubmission of the study to a more suitable journal, or perform the STORM analyses of NPCs under infection conditions with labelled IN. For these experiments positive controls will also be necessary.
A: That’s a good point (also made by Rev 1). Unfortunately, our attempts to image HIV cores docked at the nuclear pore were not successful, primarily because of difficulties with delivering a large number of fluorescent cores into intact cells. We are currently working on solving this problem but have not progressed beyond a few proof-of-principle 2-color STORM images of solitary HIV-1 cores in the vicinity of NPCs (see the figure included in the response to Rev. #1). We also agree that the lack of a positive control for NPC plasticity is a problem. The connection to HIV-1 infection is in the demonstration that SNAP tagging of NUP96 does not majorly inhibit HIV-1 infection, paving the way for future SMLM studies (perhaps using DNA-PAINT or MINFLUX) of HIV-1 entry using this labeling approach.
Minor comments:
Line 27-28: What does the transport of transcription factors have to do with the study?
A: We removed the reference to transcription factors.
Line 54-56: The cited preprint by Hou et al. is now published PMID: 40093063. Perhaps some of the most convincing evidence for structurally intact/conical capsid structures deeper inside the nucleus away from the NPC – which suggest successful transport of at least some intact capsids – are described for cell lines in PMID 39623137 and 33904396 and for primary macrophages in PMID 41326693.
A: These references are now included.
Line 78-80: The authors did not tag these cells but used a cell line created by colleagues. Please use precise language here.
A: In the original manuscript, we have acknowledged the source of these cells. We have changed the wording in the text to make clear that we have not made this cell line.
Figure 2: What does raw storm images mean? Rendering of localization? What precision? Visualization of point list?
A: STORM does not reconstruct images directly, but only a list of coordinates, i.e. the localizations of the fluorophores. Therefore, these localizations must be rendered in some fashion to represent an image. In the community 2D histogram binning has emerged as the way to present the SMLM in this fashion.
Line 154-156: The authors write that NE volume does not change upon ATP depletion. They show data for this in in 2 replicates (Fig. 2D and Fig S4F). No difference in Fig2D and a 25 % lower nuclear volume in Fig S4F for NaN3 treatment. The difference may be statistically significant with a p value below 0.05, but that does not mean its biologically relevant. It is better to state the magnitude of the effect instead of only using the word significant – e.g. the NE volume was ~ 25 % lower in one of two replicates. Please also use precise language here, and include the nuclear size analysis with the data of the 3rd replicate to make this statement more robust.
A: We now show the results from 3 biological replicates in the new Figure 2E.
Figure 4 A: Please don’t show images with saturated pixels in the GFP channel. The quantification below even shows a ~40% lower intensity of GFP inside the nucleus as opposed to the NE. This should be reflected in the image presentation.
A: We apologize for the oversight. The brightness of that image is now adjusted to avoid saturating the pixels.
References:
- Y.-L. Wu, A. Tschanz, L. Krupnik, J. Ries, Quantitative Data Analysis in Single-Molecule Localization Microscopy. Trends in Cell Biology 30, 837–851 (2020).
- M. Bates, S. A. Jones, X. Zhuang, Stochastic Optical Reconstruction Microscopy (STORM): A Method for Superresolution Fluorescence Imaging. Cold Spring Harb Protoc 2013, pdb.top075143 (2013).
- S. Haghparast, S. Stallinga, B. Rieger, Detecting continuous structural heterogeneity in single-molecule localization microscopy data. Sci Rep 13, 19800 (2023).
- W. Wang, H. Heydarian, T. A. P. M. Huijben, S. Stallinga, B. Rieger, Joint registration of multiple point clouds for fast particle fusion in localization microscopy. Bioinformatics 38, 3281–3287 (2022).
- I. Liashkovich, D. Pasrednik, V. Prystopiuk, G. Rosso, H. Oberleithner, V. Shahin, Clathrin inhibitor Pitstop-2 disrupts the nuclear pore complex permeability barrier. Sci Rep 5, 9994 (2015).
- I. Liashkovich, S. T. Stefanello, R. Vidyadharan, G. Haufe, A. Erofeev, P. V. Gorelkin, V. Kolmogorov, C. R. Mizdal, A. Dulebo, E. Bulk, I. U. Kouzel, V. Shahin, Pitstop‐2 and its novel derivative RVD ‐127 disrupt global cell dynamics and nuclear pores integrity by direct interaction with small GTPases. Bioengineering & Transla Med 8, e10425 (2023).
- A. Elosegui-Artola, I. Andreu, A. E. M. Beedle, A. Lezamiz, M. Uroz, A. J. Kosmalska, R. Oria, J. Z. Kechagia, P. Rico-Lastres, A.-L. Le Roux, C. M. Shanahan, X. Trepat, D. Navajas, S. Garcia-Manyes, P. Roca-Cusachs, Force Triggers YAP Nuclear Entry by Regulating Transport across Nuclear Pores. Cell 171, 1397-1410.e14 (2017).
Round 2
Reviewer 2 Report
Comments and Suggestions for AuthorsThe authors have not adressed the issue of a positive control showing that their imaging modality and analyses can differentiate between nm-scale changes. Hence, with the data presented, it is not possible to draw any conclusions. If none of the knock-in approaches worked, one could request a cell line from colleagues. If the authors aim to convincingly show that in mammalian cells ATP depletion etc. does not have an effect on pore size, ideally the authors would show with the same methodology that in yeast cells they can replicate the observed phenotype of the published Cryo-ET studies. The connection to HIV was also not improved.
