Surface-guided computing to analyze subcellular morphology and membrane-associated signals in 3D
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (Arcadia Science)
Abstract
Signal transduction and cell function are governed by the spatiotemporal organization of membrane-associated molecules. Despite significant advances in visualizing molecular distributions by 3D light microscopy, cell biologists still have limited quantitative understanding of the processes implicated in the regulation of molecular signals at the whole cell scale. In particular, complex and transient cell surface morphologies challenge the complete sampling of cell geometry, membrane-associated molecular concentration and activity and the computing of meaningful parameters such as the cofluctuation between morphology and signals. Here, we introduce u-Unwrap3D, a framework to remap arbitrarily complex 3D cell surfaces and membrane-associated signals into equivalent lower dimensional representations. The mappings are bidirectional, allowing the application of image processing operations in the data representation best suited for the task and to subsequently present the results in any of the other representations, including the original 3D cell surface. Leveraging this surface-guided computing paradigm, we track segmented surface motifs in 2D to quantify the recruitment of Septin polymers by blebbing events; we quantify actin enrichment in peripheral ruffles; and we measure the speed of ruffle movement along topographically complex cell surfaces. Thus, u-Unwrap3D provides access to spatiotemporal analyses of cell biological parameters on unconstrained 3D surface geometries and signals.
Article activity feed
-
non-watertight surface meshes – surfaces that are not closed and have no clearly defined inside volume48, 61 possessing potentially complex internal volumetric structures that violate the assumptions of standard 3D mesh processing algorithms
these could be very interesting though!
-
A particular technical challenge that arises when adapting techniques from computer graphics with applications to cell biology is the non-convexity, irregularity and high curvature of surface protrusions on most cell shapes. Very few methods have been proposed to accurately follow such geometries over time and have largely been demonstrated on well-defined shapes such as human pose29 or hands30, 31. Generally, these methods track by matching meshes from consecutive timepoints. To match meshes, methods attempt to assign a unique signature per vertex or face to establish a matching between vertices and faces by minimizing a loss metric32, 33. However, this approach is inherently sensitive to mesh quality, uniqueness of the signature, optimizer convergence and is difficult to generalize when tracking surfaces over many timepoints. …
A particular technical challenge that arises when adapting techniques from computer graphics with applications to cell biology is the non-convexity, irregularity and high curvature of surface protrusions on most cell shapes. Very few methods have been proposed to accurately follow such geometries over time and have largely been demonstrated on well-defined shapes such as human pose29 or hands30, 31. Generally, these methods track by matching meshes from consecutive timepoints. To match meshes, methods attempt to assign a unique signature per vertex or face to establish a matching between vertices and faces by minimizing a loss metric32, 33. However, this approach is inherently sensitive to mesh quality, uniqueness of the signature, optimizer convergence and is difficult to generalize when tracking surfaces over many timepoints. Crucially, meshes segmented from two different timepoints have different numbers of vertices and faces and the lack of the exact same surface features poses ambiguity in matching.
I spent a lot of time re-reading this section because it seems very important. As a biologist without much experience in this area, I'm not entirely sure I understand it correctly. My overarching interpretation is that you have two distinct challenges that intersect to compound the complexity of image processing. Namely, the ability to track meshes across different planes or geometries. Second, the ability to follow these over time. Both are challenging, but together, really make it hard to do the types of surface analyses that could be useful here. Am I understanding correctly? Sorry if this is totally inaccurate, but I got a bit list in the problem statement about convexity, immediately followed by statements about timepoints. I would appreciate some clarification if possible!
-
With u-Shape3D we introduced a multi-class morphological motif detection by partitioning the 3D surface into convex patches and applying support vector machines trained with expert annotation to classify the patches into pre-specified motif types45.
Curious how this approach coupled with some simple cellular staining techniques (dyes that selectively stain certain cellular features such as cellular ends, blebs, etc) could maximize the power of this? I understand that takes away from some of the impact to be able to examine without such markers, but maybe it can be incorporated to guide some customized decision-making for subcellular surface compartments that may have distinct underlying intrinsic properties worthy of different optimizations.
-
s the cortical cell body exhibits little temporal variation and blebs protrude normally to the surface, the temporal mean cell surface,
is a good proxy of the cell cortex.I know this is not the punchline, but I found this exciting because I think we could be much more creative in biology about finding useful proxies
-
Importantly, the bijectivity of the mappings guarantees that for any point on any of the surface or volume representations matching points exist on any of the other surfaces or volumes. Moreover, the bijectivity guarantees preservation of the point topology, i.e. a series of points ordered in clockwise fashion on one surface representation maps to a series of points ordered in the same way on any of the other surface representations and preserves the local neighbourhood relationships.
This is very cool
-
D.
Curious if you saw any interesting bleb "fingerprint" differences in terms of the range distribution for short- to long-lived blebs at any given point in time as opposed to on- vs off-? or did the distribution of those mostly stay constant, but the average bleb time moved?
-
The mappings rely on two critical insights: i) the engineered surface deformation of S(x, y, z) to generate a genus-0 Sref(x, y, z) for which a 3D spherical parameterization exists; ii) a novel, efficient algorithm to relax geometric distortion on the 3D sphere in a bijective and tunable manner.
I found this easy to understand! Now I went back and can better absorb the intro materials. Thank you!
-
The resources and validation provided by this work will aid the cell biology community to generate testable hypotheses of the spatiotemporal organization and regulation of subcellular geometry and molecular activity.
Agree. Would love to brainstorm ways to this could be leveraged for cell bio work at Arcadia in the future. Thank you.
-
-
u-Unwrap3D is a useful new computational to map 3D biological surfaces onto 2D for further analysis. The software is an exciting development, and it's awesome that the code has been made available on Github for use by a broader community of scientists. The ability to correlate signals from specific proteins (Septins) with particular cell surface curvatures is impressive. I don't have any feedback about the actual tool, except to say that I want to try it out! However, the results presented in the manuscript can be simplified in a way that helps the reader understand the utility of u-Unwrap3D. In the figures, some of utility of this new and exciting tool is buried among a lot of distracting detail. For example, the interesting Septin data is not immediately clear from studying Figure 4 without a detailed reading of the accompanying text. …
u-Unwrap3D is a useful new computational to map 3D biological surfaces onto 2D for further analysis. The software is an exciting development, and it's awesome that the code has been made available on Github for use by a broader community of scientists. The ability to correlate signals from specific proteins (Septins) with particular cell surface curvatures is impressive. I don't have any feedback about the actual tool, except to say that I want to try it out! However, the results presented in the manuscript can be simplified in a way that helps the reader understand the utility of u-Unwrap3D. In the figures, some of utility of this new and exciting tool is buried among a lot of distracting detail. For example, the interesting Septin data is not immediately clear from studying Figure 4 without a detailed reading of the accompanying text. Could the authors come up with a way to more obviously connect the results in Figure 4B-4F to the mapping visualized in Figure 4A? Sometimes showing less can be more constructive for helping the reader understand the content.
-