Recognition and reconstruction of cell differentiation patterns with deep learning.

Cell lineage decisions occur in three-dimensional spatial patterns that are difficult to identify by eye. There is an ongoing effort to replicate such patterns using mathematical modeling. One approach uses long ranging cell-cell communication to replicate common spatial arrangements like checkerboard and engulfing patterns. In this model, the cell-cell communication has been implemented as a signal that disperses throughout the tissue. On the other hand, machine learning models have been developed for pattern recognition and pattern reconstruction tasks. We combined synthetic data generated by the mathematical model with spatial summary statistics and deep learning algorithms to recognize and reconstruct cell fate patterns in organoids of mouse embryonic stem cells. Application of Moran's index and pair correlation functions for in vitro and synthetic data from the model showed local clustering and radial segregation. To assess the patterns as a whole, a graph neural network was developed and trained on synthetic data from the model. Application to in vitro data predicted a low signal dispersion value. To test this result, we implemented a multilayer perceptron for the prediction of a given cell fate based on the fates of the neighboring cells. The results show a 70% accuracy of cell fate imputation based on the nine nearest neighbors of a cell. Overall, our approach combines deep learning with mathematical modeling to link cell fate patterns with potential underlying mechanisms.

Adjusting the range of cell-cell communication enables fine-tuning of cell fate patterns from checkerboard to engulfing.

During development, spatio-temporal patterns ranging from checkerboard to engulfing occur with precise proportions of the respective cell fates. Key developmental regulators are intracellular transcriptional interactions and intercellular signaling. We present an analytically tractable mathematical model based on signaling that reliably generates different cell type patterns with specified proportions. Employing statistical mechanics, We derived a cell fate decision model for two cell types. A detailed steady state analysis on the resulting dynamical system yielded necessary conditions to generate spatially heterogeneous patterns. This allows the cell type proportions to be controlled by a single model parameter. Cell-cell communication is realized by local and global signaling mechanisms. These result in different cell type patterns. A nearest neighbor signal yields checkerboard patterns. Increasing the signal dispersion, cell fate clusters and an engulfing pattern can be generated. Altogether, the presented model allows us to reliably generate heterogeneous cell type patterns of different kinds as well as desired proportions.

Mouse ICM Organoids Reveal Three-Dimensional Cell Fate Clustering.

During mammalian preimplantation, cells of the inner cell mass (ICM) adopt either an embryonic or an extraembryonic fate. This process is tightly regulated in space and time and has been studied previously in mouse embryos and embryonic stem cell models. Current research suggests that cell fates are arranged in a salt-and-pepper pattern of random cell positioning or a spatially alternating pattern. However, the details of the three-dimensional patterns of cell fate specification have not been investigated in the embryo nor in in vitro systems. We developed ICM organoids as a, to our knowledge, novel three-dimensional in vitro stem cell system to model mechanisms of fate decisions that occur in the ICM. ICM organoids show similarities to the in vivo system that arise regardless of the differences in geometry and total cell number. Inspecting ICM organoids and mouse embryos, we describe a so far unknown local clustering of cells with identical fates in both systems. These findings are based on the three-dimensional quantitative analysis of spatiotemporal patterns of NANOG and GATA6 expression in combination with computational rule-based modeling. The pattern identified by our analysis is distinct from the current view of a salt-and-pepper pattern. Our investigation of the spatial distributions both in vivo and in vitro dissects the contributions of the different parts of the embryo to cell fate specifications. In perspective, our combination of quantitative in vivo and in vitro analyses can be extended to other mammalian organisms and thus creates a powerful approach to study embryogenesis.

Tissues as networks of cells: towards generative rules of complex organ development.

Network analysis is a well-known and powerful tool in molecular biology. More recently, it has been introduced in developmental biology. Tissues can be readily translated into spatial networks such that cells are represented by nodes and intercellular connections by edges. This discretization of cellular organization enables mathematical approaches rooted in network science to be applied towards the understanding of tissue structure and function. Here, we describe how such tissue abstractions can enable the principles that underpin tissue formation and function to be uncovered. We provide an introduction into biologically relevant network measures, then present an overview of different areas of developmental biology where these approaches have been applied. We then summarize the general developmental rules underpinning tissue topology generation. Finally, we discuss how generative models can help to link the developmental rule back to the tissue topologies. Our collection of results points at general mechanisms as to how local developmental rules can give rise to observed topological properties in multicellular systems.

The salt-and-pepper pattern in mouse blastocysts is compatible with signaling beyond the nearest neighbors.

Embryos develop in a concerted sequence of spatiotemporal arrangements of cells. In the preimplantation mouse embryo, the distribution of the cells in the inner cell mass evolves from a salt-and-pepper pattern to spatial segregation of two distinct cell types. The exact properties of the salt-and-pepper pattern have not been analyzed so far. We investigate the spatiotemporal distribution of NANOG- and GATA6-expressing cells in the ICM of the mouse blastocysts with quantitative three-dimensional single-cell-based neighborhood analyses. A combination of spatial statistics and agent-based modeling reveals that the cell fate distribution follows a local clustering pattern. Using ordinary differential equations modeling, we show that this pattern can be established by a distance-based signaling mechanism enabling cells to integrate information from the whole inner cell mass into their cell fate decision. Our work highlights the importance of longer-range signaling to ensure coordinated decisions in groups of cells to successfully build embryos.

Interactive, Visual Simulation of a Spatio-Temporal Model of Gas Exchange in the Human Alveolus.

In interdisciplinary fields such as systems biology, good communication between experimentalists and theorists is crucial for the success of a project. Theoretical modeling in physiology usually describes complex systems with many interdependencies. On one hand, these models have to be grounded on experimental data. On the other hand, experimenters must be able to understand the interdependent complexities of the theoretical model in order to interpret the model's results in the physiological context. We promote interactive, visual simulations as an engaging way to present theoretical models in physiology and to make complex processes tangible. Based on a requirements analysis, we developed a new model for gas exchange in the human alveolus in combination with an interactive simulation software named Alvin. Alvin exceeds the current standard with its spatio-temporal resolution and a combination of visual and quantitative feedback. In Alvin, the course of the simulation can be traced in a three-dimensional rendering of an alveolus and dynamic plots. The user can interact by configuring essential model parameters. Alvin allows to run and compare multiple simulation instances simultaneously. We exemplified the use of Alvin for research by identifying unknown dependencies in published experimental data. Employing a detailed questionnaire, we showed the benefits of Alvin for education. We postulate that interactive, visual simulation of theoretical models, as we have implemented with Alvin on respiratory processes in the alveolus, can be of great help for communication between specialists and thereby advancing research.

Cell fate clusters in ICM organoids arise from cell fate heredity and division: a modelling approach.

During the mammalian preimplantation phase, cells undergo two subsequent cell fate decisions. During the first decision, the trophectoderm and the inner cell mass are formed. Subsequently, the inner cell mass segregates into the epiblast and the primitive endoderm. Inner cell mass organoids represent an experimental model system, mimicking the second cell fate decision. It has been shown that cells of the same fate tend to cluster stronger than expected for random cell fate decisions. Three major processes are hypothesised to contribute to the cell fate arrangements: (1) chemical signalling; (2) cell sorting; and (3) cell proliferation. In order to quantify the influence of cell proliferation on the observed cell lineage type clustering, we developed an agent-based model accounting for mechanical cell-cell interaction, i.e. adhesion and repulsion, cell division, stochastic cell fate decision and cell fate heredity. The model supports the hypothesis that initial cell fate acquisition is a stochastically driven process, taking place in the early development of inner cell mass organoids. Further, we show that the observed neighbourhood structures can emerge solely due to cell fate heredity during cell division.

The transition from local to global patterns governs the differentiation of mouse blastocysts.

During mammalian blastocyst development, inner cell mass (ICM) cells differentiate into epiblast (Epi) or primitive endoderm (PrE). These two fates are characterized by the expression of the transcription factors NANOG and GATA6, respectively. Here, we investigate the spatio-temporal distribution of NANOG and GATA6 expressing cells in the ICM of the mouse blastocysts with quantitative three-dimensional single cell-based neighbourhood analyses. We define the cell neighbourhood by local features, which include the expression levels of both fate markers expressed in each cell and its neighbours, and the number of neighbouring cells. We further include the position of a cell relative to the centre of the ICM as a global positional feature. Our analyses reveal a local three-dimensional pattern that is already present in early blastocysts: 1) Cells expressing the highest NANOG levels are surrounded by approximately nine neighbours, while 2) cells expressing GATA6 cluster according to their GATA6 levels. This local pattern evolves into a global pattern in the ICM that starts to emerge in mid blastocysts. We show that FGF/MAPK signalling is involved in the three-dimensional distribution of the cells and, using a mutant background, we further show that the GATA6 neighbourhood is regulated by NANOG. Our quantitative study suggests that the three-dimensional cell neighbourhood plays a role in Epi and PrE precursor specification. Our results highlight the importance of analysing the three-dimensional cell neighbourhood while investigating cell fate decisions during early mouse embryonic development.

A GABAergic and peptidergic sleep neuron as a locomotion stop neuron with compartmentalized Ca2+ dynamics.

Animals must slow or halt locomotion to integrate sensory inputs or to change direction. In Caenorhabditis elegans, the GABAergic and peptidergic neuron RIS mediates developmentally timed quiescence. Here, we show RIS functions additionally as a locomotion stop neuron. RIS optogenetic stimulation caused acute and persistent inhibition of locomotion and pharyngeal pumping, phenotypes requiring FLP-11 neuropeptides and GABA. RIS photoactivation allows the animal to maintain its body posture by sustaining muscle tone, yet inactivating motor neuron oscillatory activity. During locomotion, RIS axonal Ca2+ signals revealed functional compartmentalization: Activity in the nerve ring process correlated with locomotion stop, while activity in a branch correlated with induced reversals. GABA was required to induce, and FLP-11 neuropeptides were required to sustain locomotion stop. RIS attenuates neuronal activity and inhibits movement, possibly enabling sensory integration and decision making, and exemplifies dual use of one cell across development in a compact nervous system.

The molecular recognition of phosphatidic acid by an amphipathic helix in Opi1.

A key event in cellular physiology is the decision between membrane biogenesis and fat storage. Phosphatidic acid (PA) is an important intermediate at the branch point of these pathways and is continuously monitored by the transcriptional repressor Opi1 to orchestrate lipid metabolism. In this study, we report on the mechanism of membrane recognition by Opi1 and identify an amphipathic helix (AH) for selective binding of PA over phosphatidylserine (PS). The insertion of the AH into the membrane core renders Opi1 sensitive to the lipid acyl chain composition and provides a means to adjust membrane biogenesis. By rational design of the AH, we tune the membrane-binding properties of Opi1 and control its responsiveness in vivo. Using extensive molecular dynamics simulations, we identify two PA-selective three-finger grips that tightly bind the PA phosphate headgroup while interacting less intimately with PS. This work establishes lipid headgroup selectivity as a new feature in the family of AH-containing membrane property sensors.

Multiscale image analysis reveals structural heterogeneity of the cell microenvironment in homotypic spheroids.

Three-dimensional multicellular aggregates such as spheroids provide reliable in vitro substitutes for tissues. Quantitative characterization of spheroids at the cellular level is fundamental. We present the first pipeline that provides three-dimensional, high-quality images of intact spheroids at cellular resolution and a comprehensive image analysis that completes traditional image segmentation by algorithms from other fields. The pipeline combines light sheet-based fluorescence microscopy of optically cleared spheroids with automated nuclei segmentation (F score: 0.88) and concepts from graph analysis and computational topology. Incorporating cell graphs and alpha shapes provided more than 30 features of individual nuclei, the cellular neighborhood and the spheroid morphology. The application of our pipeline to a set of breast carcinoma spheroids revealed two concentric layers of different cell density for more than 30,000 cells. The thickness of the outer cell layer depends on a spheroid's size and varies between 50% and 75% of its radius. In differently-sized spheroids, we detected patches of different cell densities ranging from 5 × 105 to 1 × 106 cells/mm3. Since cell density affects cell behavior in tissues, structural heterogeneities need to be incorporated into existing models. Our image analysis pipeline provides a multiscale approach to obtain the relevant data for a system-level understanding of tissue architecture.

Contractile and mechanical properties of epithelia with perturbed actomyosin dynamics.

Mechanics has an important role during morphogenesis, both in the generation of forces driving cell shape changes and in determining the effective material properties of cells and tissues. Drosophila dorsal closure has emerged as a reference model system for investigating the interplay between tissue mechanics and cellular activity. During dorsal closure, the amnioserosa generates one of the major forces that drive closure through the apical contraction of its constituent cells. We combined quantitation of live data, genetic and mechanical perturbation and cell biology, to investigate how mechanical properties and contraction rate emerge from cytoskeletal activity. We found that a decrease in Myosin phosphorylation induces a fluidization of amnioserosa cells which become more compliant. Conversely, an increase in Myosin phosphorylation and an increase in actin linear polymerization induce a solidification of cells. Contrary to expectation, these two perturbations have an opposite effect on the strain rate of cells during DC. While an increase in actin polymerization increases the contraction rate of amnioserosa cells, an increase in Myosin phosphorylation gives rise to cells that contract very slowly. The quantification of how the perturbation induced by laser ablation decays throughout the tissue revealed that the tissue in these two mutant backgrounds reacts very differently. We suggest that the differences in the strain rate of cells in situations where Myosin activity or actin polymerization is increased arise from changes in how the contractile forces are transmitted and coordinated across the tissue through ECadherin-mediated adhesion. Altogether, our results show that there is an optimal level of Myosin activity to generate efficient contraction and suggest that the architecture of the actin cytoskeleton and the dynamics of adhesion complexes are important parameters for the emergence of coordinated activity throughout the tissue.