Cellular Lineages and Development: from single cells to landscapes - Talk Abstracts
Exploratory Learning in Cell Biology
Learning entails the self-modification of a system under closed-loop dynamics with its environment. Not only the system's components may change, but also the way they interact with one another - like synapses during learning in the brain, that modify interactions between neurons. Such processes, however, are not limited to the brain but can be found also in other complex systems. I will describe a theoretical framework for a primitive form of learning that takes place within a single cell. Based on plasticity of gene regulatory networks, this type of learning is composed of random modifications guided by global feedback. The capacity to utilize exploratory dynamics, improvisational in nature, provides cells with the plasticity required to overcome extreme challenges and to develop novel phenotypes. I will present applications to microorganism adaptation and to cancer progression and suggest further potential applications to other biological contexts.
Learning from single cells with random matrix theory and physics-informed neural networks
Single-cell molecular profiling has produced a treasure trove of data on embryonic development and differentiation. While the fidelity and scale of single-cell assays is continuously expanding, there are still significant challenges with respect to data analysis. For example, the sparse and noisy nature of single-cell data makes it difficult to find interesting patterns, such as distinct cell types. Secondly, since many single-cell assays provide only snapshots of dynamic processes, it remains difficult to infer causal gene regulatory mechanisms. In this presentation, I will discuss our suggested solutions to these two problems. First, I will present phiclust (φ clust ), a clusterability measure derived from random matrix theory. The ability to discover new cell phenotypes by unsupervised clustering of single-cell transcriptomes has revolutionized biology. Currently, there is no principled way to decide whether a cluster of cells contains meaningful subpopulations that should be further resolved. In this presentation, I will show that phiclust can be used to identify cell clusters with non-random substructure, testably leading to the discovery of previously overlooked phenotypes.
I will also discuss unpublished work on how to infer the parameters of gene regulatory networks (GRNs) from single-cell data using physics-informed neural networks (PINNs). One of the main goals of developmental biology is to reveal GRNs underlying the robust differentiation of multipotent progenitors into precisely specified cell types. Most existing methods to infer GRNs from experimental data have limited predictive power as the inferred GRNs merely reflect gene expression similarity or correlation. In my presentation, I will demonstrate that PINNs can be used to infer the parameters of GRNs and outperform regular feed-forward neural networks on this task. I will focus specifically on GRNs that exhibit bifurcation behavior and can therefore model cell differentiation. In conclusion, I hope that the tools discussed in this presentation will help extract a reliable, mechanistic understanding of biological processes from single-cell data.
Inferring principles of cell-fate control from cancer cell lineages.
Rapid technological advances are now allowing measurements of various biological parameters at the single cell level. While such univariate distributions provide interesting insights into cell-to-cell variability and heterogeneity in cell states, much less explored and understood are the implications of correlations amongst single cells. Particularly in cellular populations that are growing and dividing, there is increasing realization that lineage correlations can provide key biological insights that cannot be gleaned simply from single-cell distributions alone.
In this talk, I will present two such cases where lineage correlations in (a) cell-fate after drug treatment and (b) in cell-cycle times, allow inference of underlying cell-fate control mechanisms in cancer cells. I will first demonstrate how the ubiquitous exponential growth model (and generalizations such as age-structured models), fail to predict population growth rates of drug-treated cancer cells from single cell measurements. I will argue that the key to resolving this paradox lies in accounting for cell-fate correlations, which indicate that fate decisions occur well before addition of the drug. Next, I will show how in the same cellular population, surprising correlation structures in cell-division times imply control of the cell cycle by the circadian clock. Using stochastic simulations of the underlying molecular networks, I will show that circadian control of the cell cycle can indeed produce the observed correlations, suggesting a novel, reporter-free approach to investigating "gating" of cell divisions by the clock.
Stochastic gene expression in cellular lineage trees
Stochasticity in gene expression is an elemental source of cell-to-cell variability (or noise) in clonal cell populations. So far, this phenomenon has been studied using the Gillespie Algorithm, or the Chemical Master Equation, which implicitly assumes that cells are independent and neither grow nor divide. This talk will discuss recent developments in modelling populations of growing and dividing cells through agent-based approaches. I will show how the lineage structure affects gene expression noise in time-lapse microscopy, which leads to a straightforward interpretation of cell-to-cell variability in population snapshots. I will also illustrate how cell cycle variability shapes extrinsic noise across lineage trees and discuss the consequences of gene expression noise on phenotypic selection. Finally, I outline how to construct effective chemical master equation models based on dilution reactions and extrinsic variability that provide surprisingly accurate approximations of the noise statistics across growing populations. The results highlight that it is crucial to consider cell growth and division when quantifying cellular noise.
Geometric models of cell fate specification
Cell fate decisions emerge as a consequence of a complex set of gene regulatory networks. Detailed models of these networks are known to suffer from over-parameterization. We will describe recent work formalizing an alternative approach first presented by Waddington, which likens differentiation of different cell types to flow through a landscape in which valleys represent alternative fates. This allows the construction of minimally parameterized models consistent with cell behaviour. We will describe how this construction leads to intuitive models that are well adapted to biological data. We will also describe how to think about models of spatial pattern formation in a geometric manner. This leads to a more unified description of cell fate specification and we will end with some remarks on differences and similarities with universality in physics.
Cellular lineages and dynamics of blood cancers
I will talk about some recent efforts in my lab to understand how the differentiation and proliferation dynamics of mutated blood stem cells deviate from that of healthy stem cells in certain types of blood cancers. We have been able to infer the history of expansion of cancer in individual patients by reconstructing the lineage tree of the cancer cells from the pattern of naturally-occurring somatic mutations in each cell's genome. I will also talk about how we can use synthetic biology to record each cell's lineage history in its own DNA, circumventing the need for naturally-occurring somatic mutations.
Regulation of cellular plasticity in the airways of the lung
Secretory Club cells of the lower airways (CCs) are an ensemble of highly plastic cells that self-renew and generate post-mitotic multiciliated cells. We are interested in the role of Notch signaling in the regulation of the CC fate and have utilized pharmacological, genetic and cell ablation approaches in the mouse model to investigate this. We report that CCs are maintained by canonical Notch signaling that is activated by multiciliated cells. Surprisingly, CC trajectories post Notch inhibition differ and the cells can be categorized as either bulk CCs, that transdifferentiate into multiciliated cells, or as variant-CCs (v-CCs), a rare subpopulation that corrals airway Neuroepithelial Bodies and lines Bronchioalveolar Duct Junctions (variant-CCs, v-CCs), that resist transdifferentiation and transition into lineage-ambiguous states instead. Lineage analysis post acute Notch inhibition shows that v-CCs proliferate and generate v-CCs in the microenvironment and bulk CCs and multiciliated cells further away. Taken together, our studies reveal that v-CC microenvironments comprise a transcriptomically distinct, reserve CC pool that can repopulate the airways with bulk CCs and multiciliated cells. I will discuss ongoing efforts to delineate the mechanisms by which bulk CCs/v-CCs make cell fate decisions.
Cell-division time statistics from stochastic exponential threshold-crossing
For cell division to take place, some specific proteins need to accumulate to a functional threshold. Most of these divisome proteins are highly abundant in the cell, and accumulate smoothly and approximately exponentially throughout the cell cycle. In this threshold-crossing process, stochastic components arise from variation from one cycle to the next of accumulation rate and division fraction, and from fluctuations of the threshold itself. How these combine to determine the statistical properties of division times is still not well understood. Here we formulate this stochastic process and calculate the statistical properties of cell division times by using first passage time techniques. We find that the distribution shape is not universal, determined by a ratio between two coefficients of variations, interpolating between Gaussian-like and long-tailed. The moments of division times are predicted to follow well-defined relationships with model parameters. Publicly available single-cell data span a broad range of values in parameter space; the measured distribution shape and moment scaling agree well with the theory over the entire range. Because of balanced biosynthesis, the accumulation statistics of any abundant protein -- as well as cell size -- predicts division time statistics equally well using our model. These results suggest that cell division is a multi-variable emergent process, which is nevertheless predictable by a single variable thanks to coupling and correlations inside the system.
Cellular compartmentalisation and receptor promiscuity as a strategy for accurate and robust inference of position during morphogenesis
Precise spatial patterning of cell fate during morphogenesis requires accurate inference of cellular position. In making such inferences from morphogen profiles, cells must contend with inherent stochasticity in morphogen production, transport, sensing and signalling. Motivated by the multitude of signalling mechanisms in various developmental contexts, we show how cells may utilise multiple tiers of processing (compartmentalisation) and parallel branches (multiple receptor types), together with feedback control, to bring about fidelity in morphogenetic decoding of their positions within a developing tissue. By simultaneously deploying specific and nonspecific receptors, cells achieve a more accurate and robust inference. We explore these ideas in the patterning of Drosophila melanogaster wing imaginal disc by Wingless morphogen signalling, where multiple endocytic pathways participate in decoding the morphogen gradient. This distributed information processing at the scale of the cell highlights how local cell autonomous control facilitates global tissue scale design.
Morphodynamics of pattern emergence and lineage specification in human cerebral organoids
Brain development involves a dynamic and complex integration of molecular, cellular, and mechanical changes that transition a 2D neural plate into a highly specialised 3D organ showing diverse cell types and tissue regions. Remarkably, this complexity can be modelled in vitro with human cerebral organoids that self-pattern and generate cell types specific to the forebrain, midbrain and hindbrain indicating that an axis is generated entirely through self-organisation. The tissue intrinsic mechanisms that lead to regionalisation, and emergence of heterogeneity starting with homogenous pluripotent stem cells to form multiple cell types and diverse brain regions have not been studied. We have established longterm lightsheet imaging to track the spatiotemporal morphodynamics that accompany lineage diversification in human cerebral organoids and are combing it with single-cell transcriptomics and spatial phenotyping to identify the mechanisms that create regional heterogeneity in this system. We also established a cellular barcoding system iTracer that combines reporter barcodes with CRISPR-Cas9 scarring and is compatible with single-cell and spatial transcriptomics. We provide an organoid protocol to enable single cell tracking of multiple subcellular features such as nuclei, actin and tubulin and have established an imaging setup to continuously monitor organoid development >3 weeks. We show that the progenitors exhibit commitment to distinct forebrain regions during this period, and the organoids undergo dynamic lumen morphogenesis events with transition of stem cells to neuroepithelium, interkinetic nuclear migrations, formation of radial glial cells and early born neurons. Combing iTracer with 4D lineage tracking in live imaging datasets, we show that cells undergo clonal proliferation and regionalisation during the first week of organoid development. This is followed by emergence of organising centres that which specify the rostro-caudal axis leading to emergence of morphological and molecular heterogeneity. Finally, we show that extracellular matrix (ECM) environment and composition impacts organoid morphogenesis and patterning. ECM mechanosensation and molecular composition both play a role in affecting the cell states and cell types that emerge leading to different brain regions formed within the organoids. Our work provides a comprehensive analysis of the early lineage commitment and patterning window dynamics in human brain organoids and sets up several technological advancements that would play an imperative role in future studies with disease modelling and development.
Rewinding the aging clock: Single cell sequencing reveals diverse cell states and trajectories that shape neuro- and gliogenesis
The African turquoise killifish combines a short lifespan with age-dependent loss of neuro-regenerative capacity, making it an attractive vertebrate model for studying brain repair mechanisms in the context of aging. The killifish telencephalon subdivided into the pallium and sub-pallium are homologous to the primary mammalian neurogenic niches SGZ and SVZ. To investigate the changes in neuro-regenerative cellular landscapes induced by aging, we performed single cell sequencing of the young and aged adult telencephali. Our analysis identifies 21 cell types including excitatory and inhibitory neurons, age-induced microglia states and progenitors of glial and non-glial nature. Iterative subclustering of progenitors revealed four unique RG types, including astroglia distinguishable by transcription factors and transitioning stages. Validation of our data in situ reveals a distinct spatial setting for defined RG subtypes, reflecting the distribution of morphologically and physiologically distinct populations. Lineage inference analysis of young vs aged telencephalon suggests neuroepithelial-like radial glia (NE-RG3) and non-glial progenitor (NGP) to be the start point and intercessor of neural development, respectively. We identified NGP as a hyper-proliferative mediator cell cluster, connecting different astroglial subtypes that also form distinct lineages. Overall, aging had the lasting impact on cell type-specific transcriptomic profiles. The complete catalogue of killifish telencephalic cell types is accessible via an online tool, providing a resource to understand adult neurogenesis in healthy brains and upon aging or disease.
Efficacy of information transmission in cellular communication
The ability of individual cells to communicate and correctly respond to any alteration in their environmental cues forms the basis of cellular decision-making. Such communication processes are carried out via diffusible molecules, whose transport is often aided by directional advection. How diffusion and advection together impact the accuracy of information transmission during signaling remains less studied. Here, we study this problem using a simplified model of signal transport in the presence and absence of crowding. Mutual information (MI), our measure of accuracy, shows three distinct regimes characterized by power law decay. I will discuss the possible physiological implications of our theoretical findings.
A matter of Proportions: The architecture of early embryonic development
Early embryos are deceptively simple entities, often described in textbooks as "a ball of cells". From this cell-ball, an organism with a three dimensional body plan emerges via evolutionarily conserved molecular interactions. However, the emergence of geometric complexity cannot be simply due to expression of genes in the time-space continuum of embryonic development. Our group explores the physical aspects of embryonic development, especially when collective cell migration during gastrulation shapes the zebrafish embryo body plan. Current knowledge of cell migration is based on differentiated cells in culture. A key difference between cells in an embryo vs cells in culture lies in cell size; differentiated cells maintain a homeostatic cell size whereas cell sizes in early embryos change constantly due to the inherent reductive nature of cell divisions. We explore the interdependence of cell size on the sizes of macromolecular assemblies such as the mitotic spindle and the nucleus during zebrafish development. This interdependence is especially interesting just after fertilization, when sizes of intracellular assemblies reach an upper size limit, beyond which sizes reduce and scale with respect to each other. I will discuss our recent efforts at understanding size scaling during embryonic development and theoretically modelling collective movements of gastrulation.
Tug of war in the gene regulatory networks underlying Epithelial-Mesenchymal Plasticity
HAS Shri Kishore
Epithelial Mesenchymal plasticity (EMP) is a crucial axis of cellular plasticity in developmental and cancerous contexts. Despite the heterogeneity across these systems, EMP has been observed to have a qualitatively consistent landscape: highly stable epithelial and mesenchymal phenotypes (deep valleys) and weakly stable hybrid phenotypes (shallow valleys). While the three-way decision of cell fate between these broad classes of phenotypes has been well studied, the mechanisms underlying the stability differences are poorly understood. Here, we show how coordinated “teams” can dictate the phenotypic landscape emergent from diverse complex gene regulatory networks. We investigate five different networks governing epithelial-mesenchymal plasticity that also show the presence of two “teams” of nodes engaging in a mutually inhibitory feedback loop (‘toggle switch’). These teams are specific to these networks and directly shape the underlying phenotypic landscape and consequently the stability and dynamical robustness of terminal phenotypes (epithelial, mesenchymal) vs. the intermediary/hybrid ones, as identified by many state stability metrics defined here. Our analysis reveals that network topology alone can contain information about phenotypic distributions it can lead to, thus obviating the need to simulate them. We propose “teams” of nodes as a network topology-based design hallmark that can drive canalization of cell-fates during diverse decision-making processes.
Self-organization and bifurcations during early embryonic development
Recent studies have shown that early embryonic development can be recapitulated in-vitro by three-dimensional stem cell aggregates, which lack initial asymmetries and inputs from extra-embryonic tissues. These data suggest that early embryonic development is a self-organizing process where embryonic stem cells can generate spatial differentiation patterns autonomously. The main goal of my lab is to identify the gene regulatory networks that drive early embryonic self-organization. In this talk, I will present two different projects that focus on the patterning of the main axis of all bilateral animals: the anterior-posterior axis of the embryo that extends from head to tail.
In part one, I will show that three-dimensional aggregates of mouse embryonic stem cells, known as embryoid bodies, can reproduce the symmetry breaking of the mouse epiblsat. Indeed, they can spontaneously form an organizer similar to the primitive streak of the mouse embryo that propagates from one side of the aggregate. This organizer defines the posterior side of the embryo and is crucial for gastrulation and germ layer specification. Combining mathematical modeling and experiments, I will propose that this symmetry-breaking event is controlled by a novel bi-stable Turing network that can generate a size-independent wavefront of Nodal signaling. Finally, I will show how we are characterizing this process using single-cell transcriptomics.
In the second part of my talk, I will discuss a new theoretical model of somitogenesis that explains anterior-posterior axis segmentation both in-vivo and in-vitro. The model is implemented by a minimal gene regulatory network between Wnt and Notch that generates a variety of self-organizing patterns by synchronized oscillations. When this network is modulated by anterior-posterior axis gradients, such as Fgf and RA, it can reproduce the most important aspects of somitogenesis along the developing tail of the embryo. Using complex systems theory, we analyzed the contribution of the self-organizing network and external gradients in pattern formation. In particular, we found that global gradients promote bifurcations of the Wnt/Notch network that lead to different diffusion-driven behaviors.
Lineage and Trajectory Inference using Single-cell omics Data
Reconstructing cell lineages that lead to the formation of tissues, organs, and complete organisms is of crucial importance in developmental biology. On the other hand, such dynamic cellular processes also involve cell-state transitions that are characterized by cascades of epigenetic and transcriptional changes. Recently emerged high-throughput single-cell RNA sequencing (scRNA-seq) techniques allow us to identify cellular identities at a single-cell resolution and thus can be utilized for elucidating the cellular heterogeneity of a dynamic cellular process and tracking cell fate decisions in normal as well as pathological development. These single-cell RNA sequencing techniques have also been coupled with CRISPR-Cas9 barcode editing for elucidating developmental lineages at the whole organism level. However, the inference of cell-state transitions or the cell lineage necessitates the development of novel computational approaches.
In this talk, I will introduce two computational methods, LinTIMaT and MARGARET for computational lineage tracing and trajectory inference respectively. LinTIMaT is a statistical method that reconstructs cell lineages using a maximum-likelihood framework by integrating mutation data from CRISPR-Cas9 barcodes and expression data from scRNA-seq. Our analysis shows that expression data helps resolve the ambiguities arising in when lineages are inferred based on mutations alone, while also enabling the integration of different individual lineages for the reconstruction of an invariant lineage tree. MARGARET provides an end-to-end framework that utilizes scRNA-seq data for inferring the cell state trajectory and dynamics of cell fate plasticity and thereby characterizes the differentiation landscape. When applied to real biological datasets representing human hematopoiesis, embryogenesis and colon differentiation, MARGARET accurately identified all major lineages along a pseudotemporal order that epitomized the expression trends of canonical cell-type markers in these processes. Using MARGARET, we also identified transitional progenitors associated with key branching events in hematopoiesis and characterized the lineage for BEST4/OTOP2 cells and the heterogeneity in goblet cell lineage in the colon under normal and inflamed ulcerative colitis conditions.
Generating neural diversity through the integration of spatial and temporal cues within stem cells
The nervous system consists of an incredible diversity of cell types. This is generated by a considerably small pool of neural stem cells (NSCs), which integrate two axes of information – spatial and temporal – to achieve this. In both vertebrates and invertebrates, NSCs experience unique molecular cues as they develop on the neuroectoderm. This imparts unique molecular identities to them and therefore the ability to generate unique neural linages. NCSs are also patterned in time – through sequentially expressed genes – that confer them with the ability to generate sequences of neural identities. Thus, while spatial patterning generates inter-lineage diversity, temporal patterning generates intra-lineage diversity, and the integration of the two within NSCs generates the diversity of cell types in the nervous system.
While much has been learned about temporal pattering of NSCs in the last decade, less is known about spatial patterning. In this talk, I will discuss how NSCs in the Drosophila brain acquire and maintain their unique spatial identities and the mechanism by which spatial patterning might mediate the integration of temporal information within the NSC to generate diversity in the brain.