Susanne M. Rafelski

Cell states are increasingly conceptualized as attractors of high-dimensional dynamical systems, yet quantitative approaches for integrating phenotypic information into this framework remain limited. Here, we take an image-based approach that combines unsupervised machine learning (ML) with timelapse imaging to extract and characterize the temporal dynamics of morphological features. Using a cell line with endogenously tagged VE-cadherin, we acquired brightfield and fluorescence timelapse images of human induced pluripotent stem cell-derived endothelial cell (hiPSC-EC) monolayers, which adopt distinct phenotypes at two different magnitudes of shear stress in terms of their morphology, behavior, and VE-cadherin organization. To quantify these phenotypic cell states without segmentation, we trained a diffusion autoencoder to predict VE-cadherin signal from brightfield images. We identified interpretable ML-based features representing cell orientation, elongation, and local density. Treating these variables as dimensions of a morphological state space, we estimated a data-driven vector field and found that the two observed phenotypic cell states correspond to stable fixed points of the inferred dynamical system. Mapping measured cell migration coherence onto this space further distinguished the states. Imaging cells across intermediate shear stresses revealed a regime of bistability in which both states coexist, indicating that the shear-stress-dependent transition between endothelial cell states occurs as a bifurcation of the inferred dynamical system. Finally, we applied this method to study an N-terminal truncation of VE-cadherin, finding that mutated cells preserve alignment and coherent migration, but exhibit altered morphology and increased migration speed. This work demonstrates the applicability of a dynamical systems approach to quantitatively characterize morphological aspects of cell state from interpretable ML-based features.

The epithelial-to-mesenchymal transition (EMT) is a widely studied cell state change, yet differences in model design and measurement approaches limit comparison across studies. Addressing this challenge requires experimental model systems and analysis frameworks that support standardization across contexts. Here, we show that human induced pluripotent stem (hiPS) cells in defined cell culture geometries, two-dimensional colonies and three-dimensional lumenoids, enable multimodal measurements of EMT dynamics within a single experimental platform. Using fixed-cell and live-cell image-based assays, we quantify changes in cell migration, EMT-related molecular markers, cell–cell junction organization and interactions with the basement membrane, a specialized form of the extracellular matrix, during EMT induced in hiPS cells. We identify cell culture geometry-dependent differences in the timing of migration onset and show that basement membrane integrity can be quantitatively linked to these differences. Together, these results establish an imaging-based framework for analysis of cell state transitions and provide accessible datasets and tools.

A key challenge in understanding subcellular organization is quantifying interpretable measurements of intracellular structures with complex multi-piece morphologies in an objective, robust and generalizable manner. Here we introduce a morphology-appropriate representation learning framework that uses three-dimensional rotation-invariant autoencoders and point clouds. This framework is used to learn representations of complex shapes that are independent of orientation, compact and interpretable. We apply our framework to intracellular structures with punctate morphologies (for example, DNA replication foci) and polymorphic morphologies (for example, nucleoli). We explore the trade-offs in the performance of this framework compared to image-based autoencoders by performing multi-metric benchmarking across efficiency, generative capability and representation expressivity metrics. We find that the proposed framework, which embraces the underlying morphology of multi-piece structures, can facilitate the unsupervised discovery of subclusters for each structure. We show how this approach can also be applied to phenotypic profiling using a dataset of nucleolar images following drug perturbations.

Summary: The diversity of cellular and tissue structures can arise from a few basic cell shapes, which undergo various transformations based on biophysical constraints on cytoskeletal organization. While cellular geometry has been linked with selected biological processes such as polarity, signaling or morphogenesis, the orchestration of the whole proteome in association to cell shape is still poorly understood. In this study, using more than 1 million images of single cells stained for 11,998 proteins across 10 cell lines in the Human Protein Atlas database, we performed an integrated analysis of organelle, pathway and single protein levels in association to a 2D cellular shapespace. We found that cell and nuclear shapes across cell lines exist in a shared continuum. We also found that the subcellular organelle topology varies across cell lines, but remains robust within each cell line’s shapespace. At the single protein level, we found that cells of different shapes in the same cell cycle phase might be preparing for different fates, and that many non-cell cycle proteins expressed shape-based abundance variation. Using the same coordinate framework defined by shape, we could analyze the distribution shift of protein spatial localization under drug perturbation.

The epithelial to mesenchymal transition (EMT) is a widely studied but poorly defined state change due to the variety of ways in which it has been characterized in cells. There is a need for reproducible cell model systems that enable the integration and comparison of different types of measured observations of cells across many distinct cellular contexts. We present human induced pluripotent stem (hiPS) cells as such a model system by demonstrating its utility through a comparative analysis of hiPS cell-EMT in 2D and 3D cell culture geometries. We developed live-imaging-based assays to directly compare examples of changes in cell function (via migration timing), molecular components (via expression of marker proteins), organization (via reorganization of cell junctions), and environment (via dynamics of basement membrane) in the same experimental system. The EMT-related changes we measured occurred earlier in 2D colonies than in 3D lumenoids, likely due to differences in the basement membrane environments associated with 2D vs. 3D initial hiPS cell culture geometries. We have made the 449 60-hour-long 3D time-lapse movies and the associated tools used for analysis and visualization open-source and easily accessible as a resource for future work in this field.

A key challenge in understanding subcellular organization is quantifying interpretable measurements of intracellular structures with complex multi-piece morphologies in an objective, robust and generalizable manner. Here we introduce a morphology-appropriate representation learning framework that uses 3D rotation invariant autoencoders and point clouds. This framework is used to learn representations of complex multi-piece morphologies that are independent of orientation, compact, and easy to interpret. We apply our framework to intracellular structures with punctate morphologies (e.g. DNA replication foci) and polymorphic morphologies (e.g. nucleoli). We systematically compare our framework to image-based autoencoders across several intracellular structure datasets, including a synthetic dataset with pre-defined rules of organization. We explore the trade-offs in the performance of different models by performing multi-metric benchmarking across efficiency, generative capability, and representation expressivity metrics. We find that our framework, which embraces the underlying morphology of multi-piece structures, facilitates the unsupervised discovery of sub-clusters for each structure. We show how our approach can also be applied to phenotypic profiling using a dataset of nucleolar images following drug perturbations. We implement and provide all representation learning models using CytoDL, a python package for flexible and configurable deep learning experiments.

Summary: To investigate the fundamental question of how cellular variations arise across spatiotemporal scales in a population of identical healthy cells, we focused on nuclear growth in hiPS cell colonies as a model system. We generated a 3D timelapse dataset of thousands of nuclei over multiple days, and developed open-source tools for image and data analysis and an interactive timelapse viewer for exploring quantitative features of nuclear size and shape. We performed a data-driven analysis of nuclear growth variations across timescales. We found that individual nuclear volume growth trajectories arise from short timescale variations attributable to their spatiotemporal context within the colony. We identified a strikingly time-invariant volume compensation relationship between nuclear growth duration and starting volume across the population. Notably, we discovered that inheritance plays a crucial role in determining these two key nuclear growth features while other growth features are determined by their spatiotemporal context and are not inherited.

Validation metrics are key for tracking scientific progress and bridging the current chasm between artificial intelligence research and its translation into practice. However, increasing evidence shows that, particularly in image analysis, metrics are often chosen inadequately. Although taking into account the individual strengths, weaknesses and limitations of validation metrics is a critical prerequisite to making educated choices, the relevant knowledge is currently scattered and poorly accessible to individual researchers. Based on a multistage Delphi process conducted by a multidisciplinary expert consortium as well as extensive community feedback, the present work provides a reliable and comprehensive common point of access to information on pitfalls related to validation metrics in image analysis. Although focused on biomedical image analysis, the addressed pitfalls generalize across application domains and are categorized according to a newly created, domain-agnostic taxonomy. The work serves to enhance global comprehension of a key topic in image analysis validation.