Highlights d Cardiac fibroblasts and endothelial cells induce hiPSCcardiomyocyte maturation d CX43 gap junctions form between cardiac fibroblasts and cardiomyocytes d cAMP-pathway activation contributes to hiPSCcardiomyocyte maturation d Patient-derived hiPSC-cardiac fibroblasts cause arrhythmia in microtissues

Single-cell RNA sequencing (scRNA-seq) has enabled researchers to study gene expression at a cellular resolution. However, noise due to amplification and dropout may obstruct analyses, so scalable denoising methods for increasingly large but sparse scRNAseq data are needed. We propose a deep count autoencoder network (DCA) to denoise scRNA-seq datasets. DCA takes the count distribution, overdispersion and sparsity of the data into account using a zero-inflated negative binomial noise model, and nonlinear gene-gene or gene-dispersion interactions are captured. Our method scales linearly with the number of cells and can therefore be applied to datasets of millions of cells. We demonstrate that DCA denoising improves a diverse set of typical scRNA-seq data analyses using simulated and real datasets. DCA outperforms existing methods for data imputation in quality and speed, enhancing biological discovery.

Single-cell RNA sequencing (scRNA-seq) has enabled researchers to study gene expression at a cellular resolution. However, noise due to amplification and dropout may obstruct analyses, so scalable denoising methods for increasingly large but sparse scRNA-seq data are needed. We propose a deep count autoencoder network (DCA) to denoise scRNA-seq datasets. DCA takes the count distribution, overdispersion and sparsity of the data into account using a negative binomial noise model with or without zero-inflation, and nonlinear gene-gene dependencies are captured. Our method scales linearly with the number of cells and can, therefore, be applied to datasets of millions of cells. We demonstrate that DCA denoising improves a diverse set of typical scRNA-seq data analyses using simulated and real datasets. DCA outperforms existing methods for data imputation in quality and speed, enhancing biological discovery.

Recapitulating mammalian embryonic development in vitro is a major challenge in 1 biology. It has been shown that gastruloids 1-5 and ETX embryos 6 can display hallmarks 2 of gastrulation in vitro. However, these models fail to progress beyond spatially 3 segregated, yet amorphous cellular assemblies. Systems such as organoids 7 do show tissue 4 stratification and organogenesis, but require adult stem cells or exogeneous induction of 5 specific cell fates, and hence do not reflect the emergent organization of embryonic 6 development. Notably, gastruloids are derived exclusively from embryonic stem cells 7 (ESCs), whereas, in vivo, crucial patterning cues are provided by extraembryonic cells 8 . 8Here, we show that assemblies of mouse ESCs (mESCs) and extraembryonic endoderm 9 (XEN) cells can develop beyond gastrulation and produce a central hallmark of 10 organogenesis: stratified neural epithelia resembling a neural tube, which can be further 11 differentiated to cerebral cortex-like tissue. By single-cell RNA-seq, we show that our 12 2 model has a larger cell type diversity than existing models, and that mESCs and XEN 13 cells impact each other's differentiation. XEN cells promote neural tube formation 14 through local inhibition of primitive streak formation. In turn, the presence of mESCs 15 drives XEN cells to resemble visceral endoderm, which envelops the embryo in vivo. This 16 study provides a model system to investigate neurulation and extraembryonic endoderm 17 development, and may serve as a starting point to generate embryo models that advance 18 further toward the formation of the vasculature, nervous system, and digestive tube. 19 20We first implemented the original mouse gastruloid protocol 1 in which mESCs are aggregated 21 in N2B27 media and exposed to a pulse of WNT signaling for 24 h. After 96 h, this protocol 22 resulted in elongated gastruloids. As reported before 1-3 , gastruloids contained localized 23 primitive streak-and neural progenitor-like compartments, marked by Brachyury (T) and 24 SOX2, respectively (Fig. 1b, inset). We then adapted the gastruloid protocol by co-aggregating 25 XEN cells with mESCs, keeping all other conditions the same (Fig. 1a). After 96 h, the 26 resulting aggregates again showed T-positive and SOX2 positive compartments (Fig. 1b). 27However, in striking contrast with standard gastruloids, SOX2-positive cells were now 28 organized in stratified epithelia surrounding one or multiple lumina. The frequency of these 29 tubular structures depended on the fraction of XEN cells (Fig. 1c, Extended Data Fig. 1a). At 30 a XEN:mESC ratio of 1:3 we observed the concurrence of SOX2-positive tubes and T-positive 31 cells in the majority of aggregates. Since the canonical pluripotency marker OCT4 was not 32 expressed (Extended Data Fig. 1b), we hypothesized that the observed structures resemble 33 neural tubes. The presence of N-cadherin and absence of E-cadherin in the tubes (Fig. 1d) is 34 consistent with the known switch from E-to N-cadherin during neural differentiation in ...

Stem-cell derived in vitro systems, such as organoids or embryoids, hold great potential for modeling in vivo development. Full control over their initial composition, scalability, and easily measurable dynamics make those systems useful for studying specific developmental processes in isolation. Here we report the formation of gastruloids consisting of mouse embryonic stem cells (mESCs) and extraembryonic endoderm (XEN) cells. These XEN-enhanced gastruloids (XEGs) exhibit the formation of neural epithelia, which are absent in gastruloids derived from mESCs only. By single-cell RNA-seq, imaging, and differentiation experiments, we demonstrate the neural characteristics of the epithelial tissue. We further show that the mESCs induce the differentiation of the XEN cells to a visceral endoderm-like state. Finally, we demonstrate that local inhibition of WNT signaling and production of a basement membrane by the XEN cells underlie the formation of the neuroepithelial tissue. In summary, we establish XEGs to explore heterotypic cellular interactions and their developmental consequences in vitro.

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. Here, we present phiclust (ϕclust), a clusterability measure derived from random matrix theory that can be used to identify cell clusters with non-random substructure, testably leading to the discovery of previously overlooked phenotypes.

On its path from a fertilized egg to one of the many cell types in a multicellular organism, a cell turns the blank canvas of its early embryonic state into a molecular profile fine-tuned to achieve a vital organismal function. This remarkable transformation emerges from the interplay between dynamically changing external signals, the cell's internal, variable state, and tremendously complex molecular machinery; we are only beginning to understand. Recently developed single-cell omics techniques have started to provide an unprecedented, comprehensive view of the molecular changes during cell-type specification and promise to reveal the underlying gene regulatory mechanism. The exponentially increasing amount of quantitative molecular data being created at the moment is slated to inform predictive, mathematical models. Such models can suggest novel ways to manipulate cell types experimentally, which has important biomedical applications. This review is meant to give the reader a starting point to participate in this exciting phase of molecular developmental biology. We first introduce some of the principal molecular players involved in cell-type specification and discuss the important organizing ability of biomolecular condensates, which has been discovered recently. We then review some of the most important single-cell omics methods and relevant findings they produced. We devote special attention to the dynamics of the molecular changes and discuss methods to measure them, most importantly lineage tracing. Finally, we introduce a conceptual framework that connects all molecular agents in a mathematical model and helps us make sense of the experimental data.

The ability to differentiate human induced pluripotent stem cells (hiPSCs) efficiently into defined cardiac lineages, such as cardiomyocytes and cardiac endothelial cells, is crucial to study human heart development and model cardiovascular diseases in vitro. The mechanisms underlying the specification of these cell types during human development are not well-understood which limits fine-tuning and broader application of cardiac model systems. Here, we used the expression of ETV2, a master regulator of hematoendothelial specification in mice, to identify functionally distinct subpopulations during the co-differentiation of endothelial cells and cardiomyocytes from hiPSCs. Targeted analysis of single-cell RNA sequencing data revealed differential ETV2 dynamics in the two lineages. A newly created fluorescent reporter line allowed us to identify early lineage-predisposed states and show that a transient ETV2-high state initiates the specification of endothelial cells. We further demonstrated, unexpectedly, that functional cardiomyocytes can originate from progenitors expressing ETV2 at a low level. Our study thus sheds light on the in vitro differentiation dynamics of two important cardiac lineages.