Justin Feigelman

Network plasticity of pluripotency transcription factors in embryonic stem cells

Transcription factor (TF) networks are thought to regulate embryonic stem cell (ESC) pluripotency. However, TF expression dynamics and regulatory mechanisms are poorly understood. We use reporter mouse ESC lines allowing non-invasive quantification of Nanog or Oct4 protein levels and continuous long-term single-cell tracking and quantification over many generations to reveal diverse TF protein expression dynamics. For cells with low Nanog expression, we identified two distinct colony types: one re-expressed Nanog in a mosaic pattern, and the other did not re-express Nanog over many generations. Although both expressed pluripotency markers, they exhibited differences in their TF protein correlation networks and differentiation propensities. Sister cell analysis revealed that differences in Nanog levels are not necessarily accompanied by differences in the expression of other pluripotency factors. Thus, regulatory interactions of pluripotency TFs are less stringently implemented in individual self-renewing ESCs than assumed at present.

Software tools for single-cell tracking and quantification of cellular and molecular properties

Software tools for single-cell tracking and quantification of cellular and molecular properties

MCA: Multiresolution Correlation Analysis, a graphical tool for subpopulation identification in single-cell gene expression data.

BackgroundBiological data often originate from samples containing mixtures of subpopulations, corresponding e.g. to distinct cellular phenotypes. However, identification of distinct subpopulations may be difficult if biological measurements yield distributions that are not easily separable.ResultsWe present Multiresolution Correlation Analysis (MCA), a method for visually identifying subpopulations based on the local pairwise correlation between covariates, without needing to define an a priori interaction scale. We demonstrate that MCA facilitates the identification of differentially regulated subpopulations in simulated data from a small gene regulatory network, followed by application to previously published single-cell qPCR data from mouse embryonic stem cells. We show that MCA recovers previously identified subpopulations, provides additional insight into the underlying correlation structure, reveals potentially spurious compartmentalizations, and provides insight into novel subpopulations.ConclusionsMCA is a useful method for the identification of subpopulations in low-dimensional expression data, as emerging from qPCR or FACS measurements. With MCA it is possible to investigate the robustness of covariate correlations with respect subpopulations, graphically identify outliers, and identify factors contributing to differential regulation between pairs of covariates. MCA thus provides a framework for investigation of expression correlations for genes of interests and biological hypothesis generation.

Graph-based decomposition of biochemical reaction networks into monotone subsystems

Large-scale model development for biochemical reaction networks of living cells is currently possible through qualitative model classes such as graphs, Boolean logic, or Petri nets. However, when it is important to understand quantitative dynamic features of a system, uncertainty about the networks often limits large-scale model development. Recent results, especially from monotone systems theory, suggest that structural network constraints can allow consistent system decompositions, and thus modular solutions to the scaling problem. Here, we propose an algorithm for the decomposition of large networks into monotone subsystems, which is a computationally hard problem. In contrast to prior methods, it employs graph mapping and iterative, randomized refinement of modules to approximate a globally optimal decomposition with homogeneous modules and minimal interfaces between them. Application to a medium-scale model for signaling pathways in yeast demonstrates that our algorithm yields efficient and biologically interpretable modularizations; both aspects are critical for extending the scope of (quantitative) cellular network analysis.

Multiscale stochastic simulations of chemical reactions with regulated scale separation

We present a coupling of multiscale frameworks with accelerated stochastic simulation algorithms for systems of chemical reactions with disparate propensities. The algorithms regulate the propensities of the fast and slow reactions of the system, using alternating micro and macro sub-steps simulated with accelerated algorithms such as @t and R-leaping. The proposed algorithms are shown to provide significant speedups in simulations of stiff systems of chemical reactions with a trade-off in accuracy as controlled by a regulating parameter. More importantly, the error of the methods exhibits a cutoff phenomenon that allows for optimal parameter choices. Numerical experiments demonstrate that hybrid algorithms involving accelerated stochastic simulations can be, in certain cases, more accurate while faster, than their corresponding stochastic simulation algorithm counterparts.

Inference of spatiotemporal effects on cellular state transitions from time-lapse microscopy

BackgroundTime-lapse microscopy allows to monitor cell state transitions in a spatiotemporal context. Combined with single cell tracking and appropriate cell state markers, transition events can be observed within the genealogical relationship of a proliferating population. However, to infer the correlations between the spatiotemporal context and cell state transitions, statistical analysis with an appropriately large number of samples is required.ResultsHere, we present a method to infer spatiotemporal features predictive of the state transition events observed in time-lapse microscopy data. We first formulate a generative model, simulate different scenarios, such as time-dependent or local cell density-dependent transitions, and illustrate how to estimate univariate transition rates. Second, we formulate the problem in a machine-learning language using regularized linear models. This allows for a multivariate analysis and to disentangle indirect dependencies via feature selection. We find that our method can accurately recover the relevant features and reconstruct the underlying interaction kernels if a critical number of samples is available. Finally, we explicitly use the tree structure of the data to validate if the estimated model is sufficient to explain correlated transition events of sister cells.ConclusionsUsing synthetic cellular genealogies, we prove that our method is able to correctly identify features predictive of state transitions and we moreover validate the chosen model. Our approach allows to estimate the number of cellular genealogies required for the proposed spatiotemporal statistical analysis, and we thus provide an important tool for the experimental design of challenging single cell time-lapse microscopy assays.

Exact Bayesian lineage tree-based inference identifies Nanog negative autoregulation in mouse embryonic stem cells

The autoregulatory motif of Nanog, a heterogeneously expressed core pluripotency factor in mouse embryonic stem cells, remains debated. Although recent time-lapse microscopy data provide the unparalleled ability to monitor Nanog expression at the single-cell level, the extraction of mechanistic knowledge is precluded by the lack of inference techniques suitable for noisy, incomplete and heterogeneous data obtained from proliferating cell populations. This work identifies Nanog’s autoregulatory motif from quantified time-lapse fluorescence line-age trees with STILT (Stochastic Inference on Lineage Trees), a novel particle-filter based algorithm for exact Bayesian parameter inference and model selection of stochastic models. We first verify STILT’s ability to accurately infer parameters and select the correct autoregulatory motif from simulated data. We then apply STILT to time-lapse microscopy movies of a fluorescent Nanog fusion protein reporter and reject the possibility of positive autoregulation. Finally, we use STILT for experimental design, perform in silico overexpression simulations, and experimentally validate model predictions via exogenous Nanog overexpression. We finally conclude that the protein expression dynamics and overexpression experiments strongly suggest a weak negative feedback from the protein on the DNA activation rate. We find that a simple autoregulatory mechanism can explain the observed heterogeneous Nanog dynamics. This finding has implications on the understanding of the core pluripotency network, such as supporting the ability of mESC populations to diversify their proteomic profile to respond to a spectrum of differentiation cues. Beyond this application STILT constitutes a generally applicable fully Bayesian approach for model selection of gene regulatory models on the basis of time-lapse imaging data of proliferating cell populations. STILT is freely available at: http://www.imsb.ethz.ch/research/claassen/Software/stilt—stochastic-inference-on-lineage-trees.html