Visual interpretability of image-based classification models by generative latent space disentanglement applied to in vitro fertilization.

The success of deep learning in identifying complex patterns exceeding human intuition comes at the cost of interpretability. Non-linear entanglement of image features makes deep learning a "black box" lacking human meaningful explanations for the models' decision. We present DISCOVER, a generative model designed to discover the underlying visual properties driving image-based classification models. DISCOVER learns disentangled latent representations, where each latent feature encodes a unique classification-driving visual property. This design enables "human-in-the-loop" interpretation by generating disentangled exaggerated counterfactual explanations. We apply DISCOVER to interpret classification of in vitro fertilization embryo morphology quality. We quantitatively and systematically confirm the interpretation of known embryo properties, discover properties without previous explicit measurements, and quantitatively determine and empirically verify the classification decision of specific embryo instances. We show that DISCOVER provides human-interpretable understanding of "black box" classification models, proposes hypotheses to decipher underlying biomedical mechanisms, and provides transparency for the classification of individual predictions.

Anomaly detection for high-content image-based phenotypic cell profiling.

High-content image-based phenotypic profiling combines automated microscopy and analysis to identify phenotypic alterations in cell morphology and provide insight into the cell's physiological state. Classical representations of the phenotypic profile can not capture the full underlying complexity in cell organization, while recent weakly machine-learning based representation-learning methods are hard to biologically interpret. We used the abundance of control wells to learn the in-distribution of control experiments and use it to formulate a self-supervised reconstruction anomaly-based representation that encodes the intricate morphological inter-feature dependencies while preserving the representation interpretability. The performance of our anomaly-based representations was evaluated for downstream tasks with respect to two classical representations across four public Cell Painting datasets. Anomaly-based representations improved reproducibility, Mechanism of Action classification, and complemented classical representations. Unsupervised explainability of autoencoder-based anomalies identified specific inter-feature dependencies causing anomalies. The general concept of anomaly-based representations can be adapted to other applications in cell biology.

In silico labeling in cell biology: Potential and limitations.

In silico labeling is the computational cross-modality image translation where the output modality is a subcellular marker that is not specifically encoded in the input image, for example, in silico localization of organelles from transmitted light images. In principle, in silico labeling has the potential to facilitate rapid live imaging of multiple organelles with reduced photobleaching and phototoxicity, a technology enabling a major leap toward understanding the cell as an integrated complex system. However, five years have passed since feasibility was attained, without any demonstration of using in silico labeling to uncover new biological insight. In here, we discuss the current state of in silico labeling, the limitations preventing it from becoming a practical tool, and how we can overcome these limitations to reach its full potential.

Formation of recurring transient Ca2+-based intercellular communities during Drosophila hematopoiesis.

Tissue development occurs through a complex interplay between many individual cells. Yet, the fundamental question of how collective tissue behavior emerges from heterogeneous and noisy information processing and transfer at the single-cell level remains unknown. Here, we reveal that tissue scale signaling regulation can arise from local gap-junction mediated cell-cell signaling through the spatiotemporal establishment of an intermediate-scale of transient multicellular communication communities over the course of tissue development. We demonstrated this intermediate scale of emergent signaling using Ca2+ signaling in the intact, ex vivo cultured, live developing Drosophila hematopoietic organ, the lymph gland. Recurrent activation of these transient signaling communities defined self-organized signaling "hotspots" that gradually formed over the course of larva development. These hotspots receive and transmit information to facilitate repetitive interactions with nonhotspot neighbors. Overall, this work bridges the scales between single-cell and emergent group behavior providing key mechanistic insight into how cells establish tissue-scale communication networks.

Machine learning inference of continuous single-cell state transitions during myoblast differentiation and fusion.

Cells modify their internal organization during continuous state transitions, supporting functions from cell division to differentiation. However, tools to measure dynamic physiological states of individual transitioning cells are lacking. We combined live-cell imaging and machine learning to monitor ERK1/2-inhibited primary murine skeletal muscle precursor cells, that transition rapidly and robustly from proliferating myoblasts to post-mitotic myocytes and then fuse, forming multinucleated myotubes. Our models, trained using motility or actin intensity features from single-cell tracking data, effectively tracked real-time continuous differentiation, revealing that differentiation occurs 7.5-14.5 h post induction, followed by fusion ~3 h later. Co-inhibition of ERK1/2 and p38 led to differentiation without fusion. Our model inferred co-inhibition leads to terminal differentiation, indicating that p38 is specifically required for transitioning from terminal differentiation to fusion. Our model also predicted that co-inhibition leads to changes in actin dynamics. Mass spectrometry supported these in silico predictions and suggested novel fusion and maturation regulators downstream of differentiation. Collectively, this approach can be adapted to various biological processes to uncover novel links between dynamic single-cell states and their functional outcomes.

Inference of long-range cell-cell force transmission from ECM remodeling fluctuations.

Cells sense, manipulate and respond to their mechanical microenvironment in a plethora of physiological processes, yet the understanding of how cells transmit, receive and interpret environmental cues to communicate with distant cells is severely limited due to lack of tools to quantitatively infer the complex tangle of dynamic cell-cell interactions in complicated environments. We present a computational method to systematically infer and quantify long-range cell-cell force transmission through the extracellular matrix (cell-ECM-cell communication) by correlating ECM remodeling fluctuations in between communicating cells and demonstrating that these fluctuations contain sufficient information to define unique signatures that robustly distinguish between different pairs of communicating cells. We demonstrate our method with finite element simulations and live 3D imaging of fibroblasts and cancer cells embedded in fibrin gels. While previous studies relied on the formation of a visible fibrous 'band' extending between cells to inform on mechanical communication, our method detected mechanical propagation even in cases where visible bands never formed. We revealed that while contractility is required, band formation is not necessary, for cell-ECM-cell communication, and that mechanical signals propagate from one cell to another even upon massive reduction in their contractility. Our method sets the stage to measure the fundamental aspects of intercellular long-range mechanical communication in physiological contexts and may provide a new functional readout for high content 3D image-based screening. The ability to infer cell-ECM-cell communication using standard confocal microscopy holds the promise for wide use and democratizing the method.

Using Unlabeled Information of Embryo Siblings from the Same Cohort Cycle to Enhance In Vitro Fertilization Implantation Prediction.

High-content time-lapse embryo imaging assessed by machine learning is revolutionizing the field of in vitro fertilization (IVF). However, the vast majority of IVF embryos are not transferred to the uterus, and these masses of embryos with unknown implantation outcomes are ignored in current efforts that aim to predict implantation. Here, whether, and to what extent the information encoded within "sibling" embryos from the same IVF cohort contributes to the performance of machine learning-based implantation prediction is explored. First, it is shown that the implantation outcome is correlated with attributes derived from the cohort siblings. Second, it is demonstrated that this unlabeled data boosts implantation prediction performance. Third, the cohort properties driving embryo prediction, especially those that rescued erroneous predictions, are characterized. The results suggest that predictive models for embryo implantation can benefit from the overlooked, widely available unlabeled data of sibling embryos by reducing the inherent noise of the individual transferred embryo.

Emergence of synchronized multicellular mechanosensing from spatiotemporal integration of heterogeneous single-cell information transfer.

Multicellular synchronization is a ubiquitous phenomenon in living systems. However, how noisy and heterogeneous behaviors of individual cells are integrated across a population toward multicellular synchronization is unclear. Here, we study the process of multicellular calcium synchronization of the endothelial cell monolayer in response to mechanical stimuli. We applied information theory to quantify the asymmetric information transfer between pairs of cells and defined quantitative measures to how single cells receive or transmit information within a multicellular network. Our analysis revealed that multicellular synchronization was established by gradual enhancement of information spread from the single cell to the multicellular scale. Synchronization was associated with heterogeneity in the cells' communication properties, reinforcement of the cells' state, and information flow. Altogether, we suggest a phenomenological model where cells gradually learn their local environment, adjust, and reinforce their internal state to stabilize the multicellular network architecture to support information flow from local to global scales toward multicellular synchronization.

ERK1/2 inhibition promotes robust myotube growth via CaMKII activation resulting in myoblast-to-myotube fusion.

Myoblast fusion is essential for muscle development and regeneration. Yet, it remains poorly understood how mononucleated myoblasts fuse with preexisting fibers. We demonstrate that ERK1/2 inhibition (ERKi) induces robust differentiation and fusion of primary mouse myoblasts through a linear pathway involving RXR, ryanodine receptors, and calcium-dependent activation of CaMKII in nascent myotubes. CaMKII activation results in myotube growth via fusion with mononucleated myoblasts at a fusogenic synapse. Mechanistically, CaMKII interacts with and regulates MYMK and Rac1, and CaMKIIδ/γ knockout mice exhibit smaller regenerated myofibers following injury. In addition, the expression of a dominant negative CaMKII inhibits the formation of large multinucleated myotubes. Finally, we demonstrate the evolutionary conservation of the pathway in chicken myoblasts. We conclude that ERK1/2 represses a signaling cascade leading to CaMKII-mediated fusion of myoblasts to myotubes, providing an attractive target for the cultivated meat industry and regenerative medicine.

Interpretable deep learning uncovers cellular properties in label-free live cell images that are predictive of highly metastatic melanoma.

Deep learning has emerged as the technique of choice for identifying hidden patterns in cell imaging data but is often criticized as "black box." Here, we employ a generative neural network in combination with supervised machine learning to classify patient-derived melanoma xenografts as "efficient" or "inefficient" metastatic, validate predictions regarding melanoma cell lines with unknown metastatic efficiency in mouse xenografts, and use the network to generate in silico cell images that amplify the critical predictive cell properties. These exaggerated images unveiled pseudopodial extensions and increased light scattering as hallmark properties of metastatic cells. We validated this interpretation using live cells spontaneously transitioning between states indicative of low and high metastatic efficiency. This study illustrates how the application of artificial intelligence can support the identification of cellular properties that are predictive of complex phenotypes and integrated cell functions but are too subtle to be identified in the raw imagery by a human expert. A record of this paper's transparent peer review process is included in the supplemental information. VIDEO ABSTRACT.

Data science in cell imaging.

Cell imaging has entered the 'Big Data' era. New technologies in light microscopy and molecular biology have led to an explosion in high-content, dynamic and multidimensional imaging data. Similar to the 'omics' fields two decades ago, our current ability to process, visualize, integrate and mine this new generation of cell imaging data is becoming a critical bottleneck in advancing cell biology. Computation, traditionally used to quantitatively test specific hypotheses, must now also enable iterative hypothesis generation and testing by deciphering hidden biologically meaningful patterns in complex, dynamic or high-dimensional cell image data. Data science is uniquely positioned to aid in this process. In this Perspective, we survey the rapidly expanding new field of data science in cell imaging. Specifically, we highlight how data science tools are used within current image analysis pipelines, propose a computation-first approach to derive new hypotheses from cell image data, identify challenges and describe the next frontiers where we believe data science will make an impact. We also outline steps to ensure broad access to these powerful tools - democratizing infrastructure availability, developing sensitive, robust and usable tools, and promoting interdisciplinary training to both familiarize biologists with data science and expose data scientists to cell imaging.

Cytokinetic abscission is part of the midblastula transition in early zebrafish embryogenesis.

Animal cytokinesis ends with the formation of a thin intercellular membrane bridge that connects the two newly formed sibling cells, which is ultimately resolved by abscission. While mitosis is completed within 15 min, the intercellular bridge can persist for hours, maintaining a physical connection between sibling cells and allowing exchange of cytosolic components. Although cell-cell communication is fundamental for development, the role of intercellular bridges during embryogenesis has not been fully elucidated. In this work, we characterized the spatiotemporal characteristics of the intercellular bridge during early zebrafish development. We found that abscission is delayed during the rapid division cycles that occur in the early embryo, giving rise to the formation of interconnected cell clusters. Abscission was accelerated when the embryo entered the midblastula transition (MBT) phase. Components of the ESCRT machinery, which drives abscission, were enriched at intercellular bridges post-MBT and, interfering with ESCRT function, extended abscission beyond MBT. Hallmark features of MBT, including transcription onset and cell shape modulations, were more similar in interconnected sibling cells compared to other neighboring cells. Collectively, our findings suggest that delayed abscission in the early embryo allows clusters of cells to coordinate their behavior during embryonic development.

Ferroptosis occurs through an osmotic mechanism and propagates independently of cell rupture.

Ferroptosis is a regulated form of necrotic cell death that is caused by the accumulation of oxidized phospholipids, leading to membrane damage and cell lysis1,2. Although other types of necrotic death such as pyroptosis and necroptosis are mediated by active mechanisms of execution3-6, ferroptosis is thought to result from the accumulation of unrepaired cell damage1. Previous studies have suggested that ferroptosis has the ability to spread through cell populations in a wave-like manner, resulting in a distinct spatiotemporal pattern of cell death7,8. Here we investigate the mechanism of ferroptosis execution and discover that ferroptotic cell rupture is mediated by plasma membrane pores, similarly to cell lysis in pyroptosis and necroptosis3,4. We further find that intercellular propagation of death occurs following treatment with some ferroptosis-inducing agents, including erastin2,9 and C' dot nanoparticles8, but not upon direct inhibition of the ferroptosis-inhibiting enzyme glutathione peroxidase 4 (GPX4)10. Propagation of a ferroptosis-inducing signal occurs upstream of cell rupture and involves the spreading of a cell swelling effect through cell populations in a lipid peroxide- and iron-dependent manner.

Differential dynamics of early stages of platelet adhesion and spreading on collagen IV- and fibrinogen-coated surfaces.

Background: Upon wound formation, platelets adhere to the neighboring extracellular matrix and spread on it, a process which is critical for physiological wound healing. Multiple external factors, such as the molecular composition of the environment and its mechanical properties, play a key role in this process and direct its speed and outcome. Methods: We combined live cell imaging, quantitative interference reflection microscopy and cryo-electron tomography to characterize, at a single platelet level, the differential spatiotemporal dynamics of the adhesion process to fibrinogen- and collagen IV-functionalized surfaces. Results: Initially, platelets sense both substrates by transient rapid extensions of filopodia. On collagen IV, a short-term phase of filopodial extension is followed by lamellipodia-based spreading. This transition is preceded by the extension of a single or couple of microtubules into the platelet's periphery and their apparent insertion into the core of the filopodia. On fibrinogen surfaces, the filopodia-to-lamellipodia transition was partial and microtubule extension was not observed leading to limited spreading, which could be restored by manganese or thrombin. Conclusions: Based on these results, we propose that interaction with collagen IV stimulate platelets to extend microtubules to peripheral filopodia, which in turn, enhances filopodial-to-lamellipodial transition and overall lamellipodia-based spreading. Fibrinogen, on the other hand, fails to induce these early microtubule extensions, leading to full lamellipodia spreading in only a fraction of the seeded platelets. We further suggest that activation of integrin αIIbß3 is essential for filopodial-to-lamellipodial transition, based on the capacity of integrin activators to enhance lamellipodia spreading on fibrinogen.

Community standards for open cell migration data.

Cell migration research has become a high-content field. However, the quantitative information encapsulated in these complex and high-dimensional datasets is not fully exploited owing to the diversity of experimental protocols and non-standardized output formats. In addition, typically the datasets are not open for reuse. Making the data open and Findable, Accessible, Interoperable, and Reusable (FAIR) will enable meta-analysis, data integration, and data mining. Standardized data formats and controlled vocabularies are essential for building a suitable infrastructure for that purpose but are not available in the cell migration domain. We here present standardization efforts by the Cell Migration Standardisation Organisation (CMSO), an open community-driven organization to facilitate the development of standards for cell migration data. This work will foster the development of improved algorithms and tools and enable secondary analysis of public datasets, ultimately unlocking new knowledge of the complex biological process of cell migration.

Coalescing replication compartments provide the opportunity for recombination between coinfecting herpesviruses.

Homologous recombination (HR) is considered a major driving force of evolution because it generates and expands genetic diversity. Evidence of HR between coinfecting herpesvirus DNA genomes can be found frequently both in vitro and in clinical isolates. Each herpes simplex virus type 1 (HSV-1) replication compartment (RC) derives from a single incoming genome and maintains a specific territory within the nucleus. This raises intriguing questions about where and when coinfecting viral genomes interact. To study the spatiotemporal requirements for intergenomic recombination, we developed an assay with dual-color FISH that enables detection of HR between different pairs of coinfecting HSV-1 genomes. Our results revealed that HR increases intermingling of RCs derived from different genomes. Furthermore, inhibition of RC movement reduces the rate of HR events among coinfecting viruses. Finally, we observed correlation between nuclear size and the number of RCs per nucleus. Our findings suggest that both viral replication and recombination are subject to nuclear spatial constraints. Other DNA viruses and cellular DNA are likely to encounter similar restrictions.-Tomer, E., Cohen, E. M., Drayman, N., Afriat, A., Weitzman, M. D., Zaritsky, A., Kobiler, O. Coalescing replication compartments provide the opportunity for recombination between coinfecting herpesviruses.

Sharing and reusing cell image data.

The rapid growth in content and complexity of cell image data creates an opportunity for synergy between experimental and computational scientists. Sharing microscopy data enables computational scientists to develop algorithms and tools for data analysis, integration, and mining. These tools can be applied by experimentalists to promote hypothesis-generation and discovery. We are now at the dawn of this revolution: infrastructure is being developed for data standardization, deposition, sharing, and analysis; some journals and funding agencies mandate data deposition; data journals publish high-content microscopy data sets; quantification becomes standard in scientific publications; new analytic tools are being developed and dispatched to the community; and huge data sets are being generated by individual labs and philanthropic initiatives. In this Perspective, I reflect on sharing and reusing cell image data and the opportunities that will come along with it.