B4 - Subcellular Protrusions Encode Cell Function beyond Whole-Cell Morphology

Introduction: : Cells dynamically adapt their shape during development, differentiation, and in response to environmental cues or drug perturbations. Accordingly, cell geometry serves as a phenotypic marker of cell state and function, particularly in cancer diagnosis and prognosis. We recently developed a 3D analytical framework based on cell geometric features to quantify the plasticity responses of Ewing Sarcoma cells to the tissue microenvironment (PMID36155740), and macrophages in tumor environment in zebrafish (PMID39939720). While whole-cell morphology provides broad insight into cell states, it can be biased by cortical shape. Mechanistically, cellular interactions occur at subcellular scales, primarily at the membrane, which mediates signal transduction between the cell and its environments. At micron or sub-micron scales, the membrane is organized into distinct biophysical and biochemical domains that act as signaling hubs and regulate cellular functions. Micron-sized membrane protrusions such as blebs, lamellipodia, ruffles, and filopodia have been increasingly recognized as active regulators of signals. For instance, our lab has demonstrated that bleb formation promotes survival signaling in suspended melanoma cells (PMID:36859545), and that lamellipodia and ruffles activate proliferation pathways independently of canonical MAPK and adhesion signaling. Recognizing the functional importance of these structures, we conjecture that cell state and cell type are primarily encoded in the finer-grained surface morphology that is unfortunately ignored by most existing classifiers of cell morphology. Here we describe a framework to identify individual protrusions decorating the cell surface and establish a local feature space that is more representative of cellular behavior and cell signaling than the overall cell shape.

Materials and

Methods: : We build upon the framework from our previous study (PMID36155740) to characterize individual membrane-associated morphological motifs —blebs, lamellipodia, and filopodia. Expanding the whole-cell geometric feature set (including volume, surface area, solidity, sphericity, roughness, and various bounding sphere-based metrics), we develop a novel protrusion-specific feature set capturing salient geometric characteristics of individual structures. To visualize morphological diversity, we apply manifold learning to construct a 2D protrusion shape space using a high-resolution light-sheet microscopy dataset of three cell types (PMID31501548).
A major challenge in this analysis is the accurate and scalable segmentation of individual protrusions. We address this using two in-house tools. First, we generate high-resolution 3D cell surfaces using the u-Segment3D package (https://github.com/DanuserLab/u-segment3D), which reconstructs detailed 3D meshes as a consensus of 2D segmentations across orthogonal views. We then apply u-Unwrap3D (https://github.com/DanuserLab/u-unwrap3D) to detect all protrusions in an unsupervised approach. Surface tips with positive mean curvature are segmented and labelled using connected component analysis. These initial ‘seeds’ are expanded across the mesh using adaptive diffusion that accounts for local surface convexity and spatial continuity. The segmented protrusions initially exist as open 3D surfaces. To convert them into closed, volumetric structures, we first apply conformalized mean curvature flow to generate a smooth reference cortical surface devoid of protrusions. Then, we close each open surface by stitching it with the corresponding cortical surface patch and use harmonic smoothing to eliminate any residual spatial discontinuities. The volumized meshes were subsequently converted to voxelized 3D images to extract protrusion geometric features.

Results, Conclusions, and Discussions:: To demonstrate proof-of-principle potential of our pipeline, we segmented six MV3 melanoma cells, six dendritic cells and five HBEC cells in the dataset, three cell types of different origin and functionally distinct. As shown by the mean curvature the cell surfaces are decorated with a rich variety of blebs, lamellipodia and filopodia (Fig 1). Crucially, individual cell types have a wide spectrum of cortical shapes from round to branched. However, each cell type demonstrates primarily only one motif: MV3 cells predominantly exhibit hemispherical blebs, dendritic cells form thin, sheet-like lamellipodia, and HBEC cells display elongated, spike-shaped filopodia protrusions (Fig 1).
Global shape feature analysis characterizing the whole-cell geometry showed a subtle pattern without clear clustering (Fig 2A). There is a general separation of cells with higher sphericity (e.g., cell IDs 1, 7, and 13), regardless of their type, from cells with larger volumes and more elongated shapes (e.g., cell IDs 6, 12, and 2) locating to the left. This shows that although some global features—such as surface area and roughness—capture aspects of surface undulations, whole-cell morphology metrics alone are insufficient to clearly distinguish these cell types or their functional states.
In contrast, our new protrusion-specific feature analysis proved more effective. We measured 14 geometrical features for each isolated protrusion on each cell and mapped the features into a 2D principal component space. This protrusion-specific shape space, shown in Fig 2B, clearly separated cell types when we computed its mean PCA coordinate (indicated by a star) over all its protrusions in the reduced feature space, demonstrating the power of local morphological profiling.
These findings emphasize that microscale membrane protrusions carry distinct biophysical signatures that reflect cell signaling mechanisms related to function. By focusing on local geometry, our approach provides a more accurate and informative method for scientific discovery in cell phenotyping studies.

Acknowledgements and/or References (Optional):: We would like to thank Cyril Pillai for discussion and contribution. This work was supported by grant no. K99CA283246 to H.MF and grant nos. R35GM136428 and RM1GM145399 to G.D.