Session G04: Physics of Cell Fate Transitions II

Mapping Transcriptomic Vector Fields of Single Cells

Cells are complex dynamical systems, and a grand challenge is to reconstruct the governing dynamical equations. Single-cell (sc)-RNA-seq, together with RNA-velocity and metabolic labeling, reveals cellular states and transitions at unprecedented resolution. Fully exploiting these data, however, requires kinetic models capable of unveiling governing regulatory functions. Here, we introduce an analytical framework dynamo, that infers absolute RNA velocity, reconstructs continuous vector-field functions that predict future cell fates, employs differential geometry to extract underlying regulatory networks, and ultimately predicts optimal reprogramming paths and perturbation outcomes. We highlight dynamo’s power to enable accurate velocity estimations on a metabolically-labeled human hematopoiesis scRNA-seq dataset. Furthermore, differential geometry analyses reveal the mechanism driving early appearance of megakaryocytes and elucidate asymmetrical regulation within the PU.1–GATA1 circuit. Leveraging the Least-Action-Path method, dynamo accurately predicts specific drivers of numerous hematopoietic transitions. Finally, in silico perturbations predict cell-fate diversions induced by gene perturbations. Dynamo thus represents an important step inadvancing quantitative and predictive theories of cell-state transitions.

*   

Landscape-inspired order parameters for classifying cell fate using single-cell RNA-seq data

Recent advances in single cell RNA sequencing (scRNA-seq) have made it possible to measure the gene expression profiles of individual cells on an enormous scale. However, with thousands of genes per measurement and high rates of dropout, the question of how to analyze this high-dimensional noisy data to understand and identify cell fate remains a pressing problem.

We show that we can use a spin-glass inspired epigenetic landscape model of cell identity to classify cell type quickly and accurately from scRNA-seq measurements using “order parameters” derived from single cell expression atlases (for example, the Mouse Cell Atlas). We demonstrate the efficacy of our method by analyzing directed stem cell differentiation protocols designed to produce murine lung lineages.   

Features of Gene Regulatory Dynamics on Minimally Frustrated Topologies

The dynamics of biological networks exhibit many features that differentiate them from random networks: robustness to node and edge deletion, sloppy parameter sensitivities, and conserved functionality across scales. Understanding the inter-dependence of these behaviors could shed light on the evolution of biological networks that exhibit all of the above mentioned features. We have previously shown that biological regulatory networks have minimally frustrated topologies, a feature not seen in random networks. Here, we analyze the ODEs-based dynamics of gene regulation on networks with minimally frustrated topologies and find that they exhibit stable states that vary mostly along a single principal component, a property that is key to establishing and maintaining cell types. This behavior is robust to node and edge deletion, and is preserved under network coarse graining. We further show that selection for the minimally frustrated property is sufficient to evolve networks that exhibit many of the well-known features of biological dynamics. Our analysis provides useful insights into the design principles of biological networks regulating cell fate and the evolutionary pressures that have shaped them.   

Glassy fluctuations in gene regulatory networks

The self-organisation of cells into complex tissues relies on a tight regulation of cell behaviour. The interaction of genes in gene regulatory networks is considered to be a main layer of cell fate regulation. The concentration of gene products, mRNA molecules and proteins, has been shown to be subjected to strong fluctuations. Here, by combining a theoretical framework with single-cell sequencing data, we show that cells can reside in a state where gene expression noise exhibits glass-like properties. Specifically, we develop a theory that maps fluctuations in gene regulatory networks to bipartite asymmetric spin glasses. The dynamics on these networks may be characterized by multiple attractors in a rough landscape and the transition between these attractors are of particular interest in the context of cellular decision making. We show that biologically plausible parameters pose cells in the vicinity of a phase transition to a glass-like phase, where fluctuations are strongly correlated in time and between genes. By mapping these findings to single-cell RNA sequencing we find that stem and progenitor cells exhibit signatures of glassy fluctuations in neural tissues. Our work highlights the possibility that long-lived noise could be a carrier of biological information.   

Master regulators as order parameters of gene expression states

Cell type-specific gene expression patterns are represented as memory states of a Hopfield neural network model. It is shown that order parameters of this model can be interpreted as concentrations of master transcription regulators that form concurrent positive feedback loops with a large number of downstream regulated genes. The order parameter free energy then defines an epigenetic landscape in which local minima correspond to stable cell states. The model is applied to gene expression data in the context of hematopoiesis.   

Individual and Collective Events in Cell Competition Dynamics using a Cellular Potts Model

Cell competition is a quality control mechanism that results in the elimination of less fit cells from a tissue. Studies in recent years have revealed that cell competition can either be driven by long-range mechanical stresses in the tissue or by short-range cell contact-dependent biochemical signaling. While mechanical competition arises from a difference in homeostatic density leading to a pressure gradient that drives cell elimination, biochemical competition results in an increased rate of apoptosis at the interface between two cell types. How these two mechanisms cooperate or compete to regulate tissue homeostasis remains unknown. Here we develop a Cellular Potts model for cell competition using parameters determined from cultured epithelial cells. Using this model, we determine the biophysical parameters that control the three possible outcomes of competition: elimination of mutant cells, elimination of wild type cells, and the coexistence of mutants with the wild type cells. We further investigate how the interplay between cell division, cell extrusion, and programmed cell death regulates tissue homeostasis and collective population dynamics. We compare these findings with a population-scale mathematical model for cell competition.   

Growth Dynamics of Bacterial Cell Cycles

Although many mechanisms lead to slow growth in bacterial cells, the phenomenology of slow growth has not yet been systematically explored with single-cell resolution. Using time-lapse microscopy of Escherichia coli and a variety of mutants and drug treatments, we explore the relation between the reduced growth rate at a population level and the distribution of cell-cycle durations. In particular, we explore the consequences of cell cycle arrest in a subpopulation on the population growth rate. We test a novel model for predicting population-level growth dynamics from single-cell measurements and we search for universal properties of slow growth in bacterial cells.   

Utilizing massively parallel CRISPRi assays to investigate persistence during antibiotic exposure

Bacterial persisters are characterised by a subgroup of cells within a population that have significant tolerance to antibiotics. This tolerance is primarily achieved due to the reduction of cell growth and metabolic activity which allows the bacteria to wait out the stress that would otherwise kill the cell and enabling the population to regrow once the stress is removed. The process has previously been attributed, at least in part, to toxin-antitoxin systems that allow the cell to inactivate itself in a probabilistic manner, however it is still unclear as to the global mechanisms that cause this.

Using our previously developed massively parallel CRISPRi assay[1], we explore the entire Escherichia coli genome identifying genes or operons that both enhance or impair the persistence mechanism. We show that the CRISPRi can replace the function of the toxin-antitoxin systems, inhibiting the cells metabolism and improving survival under certain antibiotics. We follow several highlighted genes, both beneficial and detrimental, using single cell microscopy in a microfluidic device to quantiify the phenotypic changes during and after antibiotic exposure.

[1] https://doi.org/10.1073/pnas.1918685117   

Covariance structure in sizes and numbers shows that membrane-bound organelles grow in a more correlated manner when cellular growth rate is limited

Among the fundamental questions in system cell biology is how the cell coordinates synthesis of its components to grow and self-replicate, in correspondence with its environment. While microscopic responses to cellular growth rate have been well studied at genome-scale, and much progress has been made for prokaryotic cells, we still do not have a quantitative, system-level understanding of how the various layers of biological organizations interact to regulate eukaryotic cellular growth. Here we begin to tackle this challenge by quantifying the relationship between cellular growth and membrane-bound organelles.  We simultaneously visualized six organelles in Saccharomyces cerevisiae via confocal hyperspectral microscopy while changing cellular growth rate by applying nutrient shifts and tuning the activity of nutrient-sensing pathways. The covariance structure in the sizes and numbers of different organelles shows that organelles generally grow in a more correlated manner when the growth rate is limited, while the correlation is less visible when the growth rate is higher. Our results potentially suggest that subcellular construction projects are susceptible to a fundamental speed-accuracy tradeoff that could be common to the replication of complex systems more broadly.    

Reversibiliy and cell division dynamics of elongated Escherichia coli cells obtained at high pressure

We have studied the dynamics of cell division of a population of heterogeneous morphology of Escherichia coli cells obtained after the application of high hydrostatic pressure. We show that the elongated cells obtained at high pressure reversibly grow back to normal cells upon depressurization. We further show that the dynamics of cell division upon depressurization can not be described by the prevailing cell division models. To explain the division of cells, we develop a physically motivated model of cell division and solve it using both continuous time Markov chain process and individual-based simulation. We find that the model aptly reproduces experimental results for cell division of cells with heterogeneous morphology obtained at high pressure. Furthermore, we show that such a model also reproduces the dynamics of the division of cells growing under normal conditions, and therefore provides a universal picture of cell division of rod-shaped bacteria.   

Dynamics of colonization and phenotypic adaptations of commensal gut bacteria in larval zebrafish populations.

An organism's microbiome is assembled from the microbiotic environment that surrounds it. The likelihood of a microbe colonizing a host depends on factors such as microbial density, duration of exposure, and the microbe's ability to sense a potential host. A quantitative understanding of how these factors influence the likelihood of colonization is lacking, due in large part to the experimental challenges of controlling the source pool of colonizers and of detecting colonization state. We address this using larval zebrafish, an established model vertebrate system, implementing an imaging-based assay to study the dynamics of colonization by fluorescently labelled bacteria native to the zebrafish gut. We delineate the dependence of colonization probability on environmental bacterial concentration, and also compare colonization with the process of transmission of bacteria between host individuals. The latter reveals phenotypic adaptations that indicate 'memory' of gut residence that improves bacteria's ability to colonize a host.   

Stochastic induction dynamics of the lac operon

Gene regulation is paramount to bacterial survival in changing environment. Important aspects of gene regulation have been discovered through bulk experiments and allowed the development of predictive models. However, two aspects of gene regulation remain challenging to observe experimentally without well controlled growth conditions and single cell information: (i) the mechanism driving the stochastic triggering of rare molecular events and (ii) the dynamics of gene regulation in response to environmental fluctuations.

The Lac operon can be conveniently exploited to measure rare events. Indeed, the feedback loop is enabling experimental measurement in two ways: (i) the gene expression can be gradually tuned down using inducer molecule and (ii) rare gene expression event can be easily observed as they will trigger a feedback loop and irreversibly turn it ON [Novick & Wiener 1957].

By using microfluidic devices coupled with fluorescence microscopy we dynamically monitor the state of the feedback loop in controlled growth conditions. We can measure both the time required to initiate and the dynamics of the feedback loop activation process. We observe that the feedback loop activation is stochastic, and days can separate the activation of cells growing in separated chamber when gene expression is tuned down. This finding supports the idea that rare event triggering phenotypic change can be solely due to stochastic expression of a strongly repressed gene.   

Competition between two cell types under cell cycle regulation with apoptosis

Competition between different cell or tissue types is critical in fields as diverse as bacterial ecology, developmental biology, and tumor growth. We have recently developed a mechanical model that incorporates cell cycle regulation [Li J. et al. PRX (2021)]. This model involves a characteristic pressure at which the cells become quiescent over time. Here we report on extensions of this model to multiple cell types. We study how the combination of characteristic pressure and apoptosis rate impact the competition process. Using an analytical model and discrete simulation, we explore how cells coexist/outcompete each other under two initial conditions: a nearly planar interface between two cell types and one in which a small nucleus of one cells type is surrounded by the other.

We determine the coexistence line on which two colonies coexist. We find that not always does the colony with the higher characteristic pressure invade the other, but the apoptosis rate can also play a key role. Comparing the dynamics from the two initial conditions reveals the role of statistical noise and Laplace pressure. Our findings shed light on how cancer develops and draw attention to the interplay between characteristic pressure and apoptosis rate.