Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • Matters Arising
  • Published:

Assessment of 3D MINFLUX data for quantitative structural biology in cells

Matters Arising to this article was published on 15 December 2022

The Original Article was published on 13 January 2020

This is a preview of subscription content, access via your institution

Access options

Rent or buy this article

Prices vary by article type

from$1.95

to$39.95

Prices may be subject to local taxes which are calculated during checkout

Fig. 1: Visualization of individual nuclear pores.
Fig. 2: Relative position distributions and model analysis.

Data availability

The original MINFLUX data2 was made available by S. Hell. All the reanalyzed data has been deposited to Zenodo at https://doi.org/10.5281/zenodo.5214631. Source data are provided with this paper.

Code availability

Plots for Fig. 2 and Extended Data Fig. 4 were generated using PERPL7 0.12m, available at https://bitbucket.org/apcurd/perpl-python3/commits/tag/0.12m.

References

  1. Balzarotti, F. et al. Nanometer resolution imaging and tracking of fluorescent molecules with minimal photon fluxes. Science 355, 606–612 (2017).

    Article  CAS  Google Scholar 

  2. Gwosch, K. C. et al. Minflux nanoscopy delivers 3D multicolor nanometer resolution in cells. Nat. Methods 17, 217–224 (2020).

    Article  CAS  Google Scholar 

  3. Thevathasan, J. V. et al. Nuclear pores as versatile reference standards for quantitative superresolution microscopy. Nat. Methods 16, 1045–1053 (2019).

    Article  CAS  Google Scholar 

  4. Löschberger, A. et al. Super-resolution imaging visualizes the eightfold symmetry of gp210 proteins around the nuclear pore complex and resolves the central channel with nanometer resolution. J. Cell Sci. 125, 570–575 (2012).

    Article  Google Scholar 

  5. Von Appen, A. et al. In situ structural analysis of the human nuclear pore complex. Nature 526, 140–143 (2015).

    Article  Google Scholar 

  6. Gwosch, K. et al. Assessment of 3D MINFLUX data for quantitative structural biology in cells revisited. Preprint at bioRxiv https://doi.org/10.1101/2022.05.13.491065 (2022).

  7. Curd, A. P. et al. Nanoscale pattern extraction from relative positions of sparse 3D localizations. Nano Lett. 21, 1213–1220 (2021).

    Article  CAS  Google Scholar 

  8. Prakash, K. At the molecular resolution with MINFLUX? Philos. Trans. R. Soc. A 380, 20200145 (2022).

    Article  CAS  Google Scholar 

  9. Virtanen, P. et al. SciPy 1.0: fundamental algorithms for scientific computing in Python. Nat. Methods 17, 261–272 (2020).

    Article  CAS  Google Scholar 

  10. D’Agostino, R. B. An omnibus test of normality for moderate and large size samples. Biometrika 58, 341–348 (1971).

    Article  Google Scholar 

  11. D’Agostino, R. & Pearson, E. S. Tests for departure from normality. Empirical results for the distributions of b2 and √b1. Biometrika 60, 613–622 (1973).

    Google Scholar 

  12. Burnham, K. P. & Anderson, D. R. Model Selection and Inference: A Practical Information-Theoretic Approach 75–117 (Springer, New York, 1998).

  13. Heydarian, H. et al. 3D particle averaging and detection of macromolecular symmetry in localization microscopy. Nat. Commun. 12, 2847 (2021).

    Article  CAS  Google Scholar 

Download references

Acknowledgements

All analysis was done on MINFLUX localization data provided by the Hell lab in tabular format (Extended Data Table 2). We would like to thank L. Schermelleh, J. Hohlbein, M. Peckham and P. Knight for helpful discussions. A.P.C. gratefully acknowledges funding by the UK Biotechnology and Biological Sciences Research Council (BB/S015787/1) and Wellcome Trust (204825/Z/16/Z).

Author information

Authors and Affiliations

Authors

Contributions

K.P. conceived the project. K.P. and A.P.C. designed the project, performed data analysis, interpreted results and wrote the manuscript.

Corresponding authors

Correspondence to Kirti Prakash or Alistair P. Curd.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Methods thanks Bernd Rieger and the other, anonymous, reviewers for their contribution to the peer review of this work. Primary Handling editor: Rita Strack, in collaboration with the Nature Methods team.

Extended data

Extended Data Fig. 1 Nuclear pores across different imaging modalities.

(a) A schematic of the Nup96 complex, taken from Thevathasan et al.3. 3D MINFLUX rendered data presented for comparison from Gwosch et al.2 (colormap removed for comparison). Note the uneven distribution in xz, compared with the EM model (Von Appen et al.5 and dSTORM data in Thevathasan et al.3 fig. 2h. (b) Membrane protein gp210 from amphibian oocytes imaged with dSTORM (Alexa Fluor 647). The 8-fold symmetry and circular structure of NPCs are generally seen. The diameter of gp210 is 164 ± 7 nm. Image adapted from Loeschberger et al.4. (c) Nup96 endogenously labeled with SNAP-tag-Alexa Fluor 647 in U2OS cell lines. 8- and 7-component pores are more commonly observed. The effective labeling efficiency for SNAP-Alexa Fluor 647 was ~60%. Image adapted from Thevathasan et al.3. (d) MINFLUX imaging of U2OS cell expressing Nup96–SNAP labeled with Alexa Fluor 647. In this dataset (Gwosch et al.2 fig. 2a), 6- and 7- component nuclear pores are more prominent throughout the FOV, raising a question about detection efficiency. Localizations were rendered with a Gaussian kernel, σ = 2 nm, to visualize the 4 individual copies of Nup96 per NPC subunit (15 sub-clusters per subunit apparent here). In multicolor MINFLUX imaging, the localizations in a subunit appear as a larger, undefined cluster (h). Cell line and labeling strategy as in Thevathasan et al.3. Image adapted from Gwosch et al.2. (e) Average images of gp210 (outer ring, N = 426) and WGA (central channel, N = 621) of the NPC. The outer ring (gp210) has an average diameter of ~164 nm. The diameter of the inner ring (WGA) is ~40 nm. Image adapted from Loeschberger et al.4. Scale bar: 100 nm. (f) dSTORM images of WGA labeled with ATTO 520 (green) and gp210 labeled with Alexa Fluor 647 (pink) in amphibian oocytes. Both the outer ring and inner channel are visible (Loeschberger et al.4). (g) Two-color SMLM image of Nup96–SNAP-Alexa Fluor 647 (red) and WGA-CF680 (cyan) in U2OS cell lines. The outer ring is clearly visible and the inner ring is also visible in most cases. Image adapted from Thevathasan et al.3. (h) Two-color MINFLUX imaging of U2OS cell expressing Nup96–SNAP labeled with Alexa Fluor 647 and WGA conjugated to CF680. The subunits of the outer ring, which each have 4 copies of Nup96, now appear as single clusters (comparing with c). The inner ring (WGA) also appears as undefined aggregations of the signal. Image adapted from Gwosch et al.2.

Extended Data Fig. 2 Visualization of the inner ring of nuclear pores (wheat germ agglutinin (WGA).

Scatter plots showing localizations from 10 segmented WGA complexes from the 3D, 2-color MINFLUX dataset. The segmented complexes did not contain rings of localizations as found by Thevathasan et al.3 and Loeschberger et al.4 Scale bar: 20 nm.

Extended Data Fig. 3 MINFLUX localization filtering.

Scatter plots for 2D, 1-color (a); 3D, 1-color (b); 3D, 2-color (c); and 2D, live (d) unfiltered and filtered MINFLUX datasets. The raw MINFLUX data comes in a tabular format, with a Boolean flag indicating that localization was assigned as either a background event or a true molecular event (Gwosch et al.2, Extended Data Table 2). For the final data, we give the density of true localizations over the FOV defined by the total molecular emission events before filtering. Scale bar: 500 nm (a), 200 nm (b), 500 nm (c), 50 nm (d) in the published data column.

Extended Data Fig. 4 NPC tilt and z-distances.

A 2D projection model of the two-layer NPC (a) requires its axis to be tilted (θ) by 36° for an inter-layer distance (ILD) of 50 nm between localizations to result in an average measurement in the z-direction (<ILDz>) of 40.5 nm. Considering z-measurements only from the center point of the lower layer to all points on the upper layer, ILDz has a range (Range(ILDz)center) of 63 nm, when the diameter (D) of the projected NPC is 107 nm. Including z-measurements from all points in the lower layer to all points on the upper layer doubles the total range of ILDz to 125 nm, which is large compared with <ILDz>. Therefore if localizations with ILD of 50 nm were measured to have <ILDz> of 40.5 nm, we would expect the NPCs to be tilted to ~36° and also expect z-measurements between the layers to have a broad spread, before considering additional spread owing to a distribution of NPC tilt angles. Distributions of 𝚫z between localizations in a 3D NPC model (b) show a similar pattern. The NPC model used localizations with a two-layer, 8-fold radially symmetric structure, with an inter-layer distance of 50 nm and diameter of 107 nm. At each value of θ (tilt of the model axis from the z-direction), the model was also rotated about its axis by angles 𝜑 from 0 to 44° in 1° increments. 𝚫z values were found between all localizations at each rotation angle 𝜑 and aggregated over 𝜑 to result in an averaged distribution at each tilt angle θ. In agreement with the 2D model (a), the inter-layer distance peak moved to shorter distances for higher tilt angles θ, approaching 40 nm between θ = 30° and θ = 40°. The inter-layer distance peak was still close to 50 nm at θ = 10°. In further agreement with a, and as may be intuitively expected, broadening of the 𝚫z distribution with increasing θ significantly reduced the contrast of the inter-layer distance peak for θ ≥ 20°. Therefore, it is highly unlikely that the high-contrast peak at 40.5 nm (Fig. 2g,h) of the experimental 𝚫z distribution of the data of Gwosch et al.2 fig. 3f would be generated by two layers of localizations with an inter-layer distance of 50 nm, tilted at 30–40°. Rather, we support the statement of Gwosch et al. that the layers of the NPCs in Gwosch et al.2 were typically parallel to the focal plane. We also support this statement as a reasonable approximation in a 3D scatter plot of the localizations (c, Supplementary Video 1) of Gwosch et al.2 fig. 3f. In this plot, NPC layers, when discernible, appear generally to be roughly parallel to a fitted surface representing nuclear envelope curvature (quadratic in x and y, fitted to the localization coordinates). The mean inclination of this surface at the localization coordinates is 5° (maximum: 9°). At θ = 5°, an ILD of 50 nm would result in <ILDz> of 49.8 nm (a), or a fractional difference of 0.4% between ILD and <ILDz>. A narrow distribution of local NPC tilts with a peak at this angle may be expected (for example s.d. 12° in Heydarian et al.13). Our result in Fig. 2g,h, therefore, reflects the inter-layer distance of the acquired localizations, not a projection of a two-layer structure at large tilt angles. Furthermore, a similar 3D plot and fit (d, Supplementary Video 2) of the Nup96 localizations of Gwosch et al.2 fig. 5c shows a similar (but denser) distribution of NPCs. In this case, the mean fitted nuclear envelope inclination was 2° (maximum: 5°), and we also expect NPCs to have a similar distribution of local tilt angles centered on this angle13. In this case, we found an inter-layer distance of 50.9 nm (Fig. 2i,j), which is in fact greater than the calculation of Gwosch et al.2 at ~46 nm, despite the similar NPC tilts between the two datasets. From these considerations (ad), the difference between our inter-layer distance results of 40.5 nm and 50.9 nm for the two datasets are not explained by a difference in NPC tilts.

Extended Data Fig. 5 dSTORM localization density from Thevathasan et al.3.

From publicly available data (https://www.ebi.ac.uk/biostudies/BioImages/studies/S-BIAD8), localization densities were calculated over the nuclear regions shown. Compared to 2D, 1-color MINFLUX data (Extended Data Fig. 3a) which has an average localization density of 435 µm−2, 2D, 1-color dSTORM has a ~6x greater average localization density of 2739 µm−2. Scale bar: 3 µm in (1) and 1 µm in (2), (3), (4).

Extended Data Table 1 Corrected Akaike Information Criterion values (AICc) and relative likelihoods for different symmetric models fitted to the Δxy distributions of Fig. 2
Extended Data Table 2 MINFLUX dataset (localization positions and filtering data) as provided by the authors (Gwosch et al.2)

Supplementary information

Reporting Summary

41592_2022_1694_MOESM2_ESM.avi

Supplementary Video 1: Animated 3D scatterplot and fitted surface for the data of Gwosch et al.2 fig. 3f.

41592_2022_1694_MOESM3_ESM.avi

Supplementary Video 2: Animated 3D scatterplot and fitted surface for the data of Gwosch et al.2 fig. 5c.

Source data

Source Data Fig. 1

Statistical source data.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Prakash, K., Curd, A.P. Assessment of 3D MINFLUX data for quantitative structural biology in cells. Nat Methods 20, 48–51 (2023). https://doi.org/10.1038/s41592-022-01694-x

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1038/s41592-022-01694-x

This article is cited by

Search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing