Extracting nanoscale membrane morphology from single-molecule localizations

Abstract

Membrane surface reconstruction at the nanometer scale is required for understanding mechanisms of subcellular shape change. This historically has been the domain of electron microscopy, but extraction of surfaces from specific labels is a difficult task in this imaging modality. Existing methods for extracting surfaces from fluorescence microscopy have poor resolution or require high-quality super-resolution data that are manually cleaned and curated. Here, we present NanoWrap, a new method for extracting surfaces from generalized single-molecule localization microscopy data. This makes it possible to study the shape of specifically labeled membranous structures inside cells. We validate NanoWrap using simulations and demonstrate its reconstruction capabilities on single-molecule localization microscopy data of the endoplasmic reticulum and mitochondria. NanoWrap is implemented in the open-source Python Microscopy Environment.

Figure 1

A flow diagram representing the major steps of NanoWrap. Localization data are first approximated by a coarse, density-based isosurface. This surface is moved toward the localizations subject to a curvature force constraint. The pipeline runs iteratively until stopping criteria are met. The result is a membrane approximation of the underlying continuous structure sampled by the input localizations. To see this figure in color, go online.

Figure 2

A comparison of SPR and NanoWrap on simulated point clouds of a 3D figure-eight, which is approximately 800 $\text{nm}$ in diameter and 100 $\text{nm}$ thick. (A) Our simulation method of going from a signed distance function to a set of localizations to a fit surface. (B) Heatmap plots of screened Poisson reconstruction and NanoWrap as functions of simulated localization density and background. For reference, the experimental SMLM data shown in Figs. 3 and 4 had a median localization density of approximately $6\times {10}^{-2}{\text{nm}}^{-2}$ and background ration of $2\times {10}^{-1}$. To see this figure in color, go online.

Figure 3

(A) 4Pi fluorogenic DNA-PAINT data of overexpressed mCherry-Sec61β, an endoplasmic reticulum membrane protein. Top: point cloud data displayed as 10 $\text{nm}$ point sprites with alpha set to 0.5 and colored by each point’s position in z according to the lookup table described in (B and C). Middle: NanoWrap surface created based on the point cloud data colored by each vertex’s position in z according to the lookup table described in (B and C). Bottom: the same surface, but colored by the mean curvature at each vertex according to the lookup table described in (D). (B–D) ROI shown by the hashed cyan box in (A) displayed in the three different ways described in (A). (B and C) Lookup table (from bottom to top): 0 to 800 $\text{nm}$. (D) Lookup table: −0.01 to 0.01 ${\text{nm}}^{-1}$. (E–G) Cross section in x-z shown by the hashed magenta line in (A). (E) Point cloud displayed as 10-$\text{nm}$ green spheres. (F) Surface displayed in magenta. (G) Point cloud and surface displayed together. Scale bars, 1 μm (A–D) and 200 nm (E–G). To see this figure in color, go online.

Figure 4

Application of NanoWrap on data from varying SMLM imaging modes. (A–C) 4Pi fluorogenic DNA-PAINT localizations from outer mitochondrial membrane protein TOMM20 (green) and their resulting surface (magenta). (D–F) Two-color 4Pi dSTORM localizations of TOMM20 (green) and overexpressed endoplasmic reticulum membrane protein Sec61β (cyan) and their resulting surfaces (magenta and yellow, respectively). (G–I) Astigmatic 3D fluorogenic DNA-PAINT localizations from overexpressed Sec61β (cyan) and their resulting surface (yellow). (A, D, and G) x-y view of localized point cloud. (B, E, and H) x-y view of resulting surface. (C, F, and I) x-z view of surfaces and points overlaid. Scale bars, 1 μm (A, B, D, E, G, and H) and 200 $\text{nm}$ (C, F, and I). To see this figure in color, go online.