Single cell multi-omic analysis identifies a Tbx1-dependent multilineage primed population in the murine cardiopharyngeal mesoderm

The poles of the heart and branchiomeric muscles of the face and neck are formed from the cardiopharyngeal mesoderm (CPM) within the pharyngeal apparatus. The formation of the cardiac outflow tract and branchiomeric muscles are disrupted in patients with 22q11.2 deletion syndrome (22q11.2DS), due to haploinsufficiency of TBX1, encoding a T-box transcription factor. Here, using single cell RNA-sequencing, we identified a multilineage primed population (MLP) within the CPM, marked by the Tbx1 lineage, which has bipotent properties to form cardiac and skeletal muscle cells. The MLPs are localized within the nascent mesoderm of the caudal lateral pharyngeal apparatus and provide a continuous source of progenitors that undergo TBX1-dependent progression towards maturation. Tbx1 also regulates the balance between MLP maintenance and maturation while restricting ectopic non-mesodermal gene expression. We further show that TBX1 confers this balance by direct regulation of MLP enriched genes and downstream pathways, partly through altering chromatin accessibility. Our study thus uncovers a new cell population and reveals novel mechanisms by which Tbx1 directs the development of the pharyngeal apparatus, which is profoundly altered in 22q11.2DS.


Introduction
The heart develops from two successive waves of mesodermal progenitor cells during early embryogenesis. The first heart field (FHF) constitutes the first wave of mesodermal derived cardiac progenitors and results in the primitive beating heart, while the second heart field (SHF) forms the second wave that builds upon the two poles of the heart 1,2 . The SHF can be anatomically partitioned to the anterior SHF (aSHF; 3,4 ) and posterior SHF (pSHF; 4-6 ), whose cells migrate to the heart via the outflow tract or inflow tract, respectively. Expression of Mesp1 at gastrulation marks the earliest mesodermal cells that will form the heart 7 . Using single cell RNA-sequencing (scRNA-seq) of the Mesp1 lineage, it was discovered that the FHF, aSHF and pSHF are specified at gastrulation 8 .
Retrospective clonal analysis 9,10 and lineage tracing studies 11 revealed that the branchiomeric skeletal muscles (BrM) of the craniofacial region and neck share a clonal relationship with the SHF. The bipotent nature of these cardiac and skeletal muscle progenitor cells is supported by studies of the ascidian, Ciona, an invertebrate chordate, in which single cells gives rise to both cardiac and skeletal muscle cells 12 . When taken together, a new term, the cardiopharyngeal mesoderm (CPM), was introduced to clearly include both SHF cardiac and skeletal muscle progenitor populations 2 . A cartoon of these populations is shown in Fig. 1a. The Tbx1 gene, encoding a T-box transcription factor, and gene haploinsufficient in 22q11.2 deletion syndrome (22q11.2DS), is expressed in the CPM and is required for cardiac outflow tract and BrM development 2 , implicating its essential roles in the CPM.
A total of 60-75% of patients with 22q11.2DS have cardiac outflow tract defects, which often require life-saving surgery during the neonatal period 13 . Additionally, most individuals with this condition have speech, feeding and swallowing difficulties in infancy, due in part to BrM hypotonia 14 . Further, heterozygous mutations in rare, non-deleted individuals, phenocopy the symptoms of the deletion 15 . Inactivation of Tbx1 in the mouse results in a persistent truncus arteriosus (PTA) [16][17][18] and significant failure to form the BrMs 19 . Although there are many studies of Tbx1, we do not yet understand its functions on a single cell level, which is needed to elucidate the true molecular pathogenesis of 22q11.2DS.
The CPM is distributed throughout the embryonic pharyngeal apparatus during early gestation. The pharyngeal apparatus consists of individual bulges of cells termed arches that form in a rostral to caudal manner from mouse embryonic days (E)8-10.5. The cellular and molecular mechanisms of how CPM cells in the pharyngeal apparatus both are maintained in a progenitor state and are allocated to form the heart and BrMs in mammals, are unknown.
To fill these gaps, we performed scRNA-seq of the CPM at multiple stages during embryogenesis. We discovered a multilineage primed progenitor (MLP) population within the CPM, which is maintained from E8-10.5 and has differentiation branches toward cardiac and skeletal muscle fates, serving as common lineage progenitors. The MLP cells are localized to the nascent lateral mesoderm of the pharyngeal apparatus, deploying cells to the heart and BrMs. We found that the Tbx1 cell lineage marks the MLPs and TBX1 activity is critical for their function. Inactivation of Tbx1 disrupts MLP lineage progression and results in ectopic expression of non-mesoderm genes. We further identify the gene regulatory network downstream of Tbx1 in the MLPs providing insights into the molecular mechanism of mammalian CPM function, essential for understanding the etiology of 22q11.2DS.

Identification of common progenitor cells in the CPM
To identify the various populations that constitute the CPM (Fig. 1a), we performed droplet- Both aSHF and pSHF contribute to the cardiac outflow tract, while only pSHF cells form the inflow tract 22,23 . Marker genes for the CPM include Tbx1, Isl1 and Tcf21, among other genes 1,24 ( Fig. 1f, g). Cluster C9 contains BrM progenitor cells identified by expression of Tcf21, Lhx2 and Myf5 24 ( Fig. 1f and Extended Data Fig. 1e). Clusters C1 and C18 contains pSHF populations as identified by expression of Hoxb1 5 , Tbx5, Foxf1, and Wnt2 25,26 (Fig. 1f and Extended Data Fig. 1e). Many of the pSHF cells are located more medially and caudally in the embryo and also contribute to posterior organ development, such as the formation of the lung 25,26 . The cells in C3 express Nkx2-5 and Mef2c 27 and cardiac structural protein genes (Tnnt2, Tnni1, Myl4, ( Fig. 1f; Extended Data Fig. 1e). In addition, subdomains express either FHF genes (Tbx5 but not Isl1 or Tcf21) or aSHF genes (Fgf10, Isl1, Tcf21 but not Tbx5 (Fig.   1h). We discovered that cluster C15 expresses genes shared by the CPM clusters, including Tbx1, Isl1, Mef2c, Tcf21 and Foxf1 (Fig. 1f). Based upon this, we refer C15 as the common multilineage progenitors within the CPM. This shows for the first time that CPM progenitors can be distinguished from more mature CPM states by their multilineage primed gene signatures.

Multilineage progenitors of the CPM differentiate to cardiac and skeletal muscles
We next investigated the relationship between the common lineage progenitors and more differentiated CPM cells using partition-based graph abstraction (PAGA) 28 . Several clusters that are not part of the CPM and were already well separated in the above cluster analysis (C4, C5, C7, C13 and C17), were excluded from PAGA analysis (Extended Data Fig. 1d). The PAGA analysis partitioned the CPM cells into six branches (Fig. 2a), connecting all the mesodermal cell populations (Fig. 1c). Convergent results from pseudotime analysis ( Next, the gene expression in CPM progenitor cells (C15) were examined from E8-E10.5 ( Fig. 2e, Extended Data Fig 2c). We searched for marker genes that are enriched in expression in MLPs, and we identified two newly appreciated genes, Aplnr (Apelin receptor) and Nrg1 (Neuregulin 1) (Fig. 2f, 2g, Extended Data Fig 2d). Aplnr is expressed in the CPM 29 but not known for MLPs, while Nrg1 is not known to be a CPM gene, and it is required in the embryonic heart for the development of the chamber myocardium 30 . We examined the co-expression of genes in the cells in CPM populations. The heatmap in Fig. 2g (Extended Data Fig. 2e, f) shows expression of genes enriched in C15, with the same genes also expressed in more differentiated CPM populations, indicating that they are multilineage primed. We found that even at E10.5, these cells are still present and retain a multilineage state. Taken together, we refer to the cells belonging to C15 as the multilineage progenitors (MLPs).

The MLPs are bilaterally localized to the caudal pharyngeal apparatus
To elucidate whether MLPs are dispersed or localized within a defined embryonic region in the pharyngeal apparatus, we performed RNAscope in situ hybridization analysis using probes enriched in MLPs including Tbx1, Isl1, Aplnr and Nrg1 with the Mesp1+ lineage marked by GFP expression (Fig. 2h-m). The pharyngeal arches form in a rostral to caudal manner in which the most caudal and lateral mesoderm is the least differentiated, while the rostral mesoderm has already migrated to the core of the arch to form BrM progenitor cells or towards the poles of the heart 31,32 . In both E8.5 and E9.5 embryos, Nrg1 and Aplnr co-expressing cells were found bilaterally in the lateral part of the caudal pharyngeal apparatus containing nascent mesoderm that is not yet differentiated to cardiac or skeletal muscle (Fig. 2f, h, Extended Data Fig. 2g, i).
At E8.5, Nrg1 and Aplnr were expressed in these regions within the forming second arch (Fig.   2i) and at E9.5, by the forming fourth arch (Fig. 2j), both overlapping with Isl1 expression. The Mesp1+ lineage is marked with GFP expression in the second arch at E8.5 and the fourth arch at E9.5 (Fig. 2k, Extended Data Fig 2h, j). Tbx1 and Isl1 were also expressed in those regions ( Fig. 2l, m). We suggest that the MLPs remain in the same region of the caudal pharyngeal apparatus, while they deploy cells rostrally, medially and dorsally thereby explaining in part the mechanism for the extension of the pharyngeal apparatus caudally (Fig. 2n).

MLPs dynamically transition over time
An important question is whether MLPs as CPM progenitors, maintain the same state based upon gene expression over time. To address this, we examined differentially expressed genes in MLPs from E8-10.5. We identified core CPM genes that are expressed similarly at all time points, including Isl1, Mef2c and Nkx2-5 (Fig. 3a, d). However, we also found that early expressing genes such as Aplnr, Nrg1, Irx1-5, Fgf8/10 and Tbx1 (Fig. 3b, e) are reduced over time, with increasing expression of cardiac developmental genes such as Hand2, Gata3/5/6, Bmp4 and Sema3c (Fig 3c, f). This is consistent with the model that the MLPs continuously allocate progenitor cells to BrMs and CMs, while showing some maturation themselves (Fig.   3g). We next tested if Tbx1 has a specific role in MLPs.

Loss of Tbx1 increases the proportion of MLPs to those of more differentiated states
Tbx1 is not one of the highest expressed genes in MLPs (Figs. 1e, 2g and 3b). Yet, based upon its function in embryogenesis, we suspect it has a critical role in MLPs because its inactivation results in failed heart and BrM development. We therefore examined the Mesp1 versus Tbx1 lineages in control embryos to understand how the CPM lineages compare. We tested whether the Tbx1 lineage includes the MLPs and how inactivation would affect MLP function. To address these questions, we compared embryos that were Mesp1 Cre ;Tbx1 +/+ (Mesp1 Cre ;Tbx1 Ctrl) vs Mesp1 Cre ;Tbx1 flox/flox (Mesp1 Cre ;Tbx1 cKO) at E9.5 (Fig. 4a, Table 1) and Tbx1 Cre/+ (Tbx1 Cre ;Tbx1 Ctrl) vs Tbx1 Cre/flox (Tbx1 Cre ;Tbx1 cKO) at E8.5 and E9.5 (Fig. 4b, Table   1). In the SwissWebster background, Tbx1 Cre/+ and Tbx1 +/heterozygous mice have no heart or aortic arch defects 33 and thus serve as controls.  Table 3

). Although
Tbx1 is strongly expressed in the CPM, it is not expressed in the heart, neither in the FHF nor the caudal and medial pSHF at the timepoints analyzed 23,37,38 . Therefore, compared to the Both Mesp1 Cre and Tbx1 Cre mediated Tbx1 conditional null embryos, referred together as Tbx1 cKO embryos, at E9.5, exhibited similar phenotypes including hypoplasia of the caudal pharyngeal apparatus comprising pharyngeal arches 3-6 36 . After cell clustering of the integrated datasets from control and Tbx1 cKO embryos (Mesp1 and Tbx1 lineages), we analyzed the proportion of cell numbers in each cluster. We found a significant increase in the proportion of MLPs in Tbx1 cKO embryos at E9.5 (Fig. 4d, e, h, i), but not in Tbx1 Cre ;Tbx1 cKO embryos at E8.5 (Fig. 4h, i). In Tbx1 cKO embryos at E9.5, the BrM and CM/OFT populations were significantly smaller than in controls. We noted that the pSHF population was increased in proportion in the Mesp1 Cre ;Tbx1 cKO dataset but not in the Tbx1 Cre ;Tbx1 cKO dataset. These results suggest that there are changes in the proportion of particular CPM and BrM cell populations upon inactivation of Tbx1. To investigate this further, we next tested whether there is also a change in gene expression within these CPM populations.

Tbx1 provides a balance of gene expression in the MLPs
To understand how Tbx1 affects gene expression within the MLPs and derivative cell types, we analyzed differentially expressed genes (DEGs) with the scRNA-seq datasets. We analyzed DEGs in each cluster in control and Tbx1 cKO embryos ( Table 8). Genes affected are involved in cell differentiation or cell signaling (e.g., Nkx2-5, Tnnt2, Bmp7, Fgf10; Fig. 5c). Taken together, these results suggest that Tbx1 regulates expression of MLP genes required for lineage differentiation. Tbx1 is also expressed in the BrM progenitor cells (C9; Fig. 1d, e, f). Some genes were specific for and downregulated only in the BrM population, including Lhx2 and Myf5 (Supplementary Table 7).  Further, genes normally expressed in non-mesoderm cells were also expressed, such as Pax8 41 (Supplementary Table 8). Therefore, inactivation of Tbx1 results in both reduced expression of MLP genes for cell transitions and in increased/ectopic expression of non-mesodermal lineages, suggesting potential linage misspecification. We suggest that Tbx1 provides a balance of specific gene expression required for MLP function.
To confirm expression changes from the scRNA-seq experiments when Tbx1 is inactivated, we checked the expression pattern of Aplnr and Pax8 in vivo using RNAscope analysis. These Overall, Tbx1 provides a balance that promotes maturation but restricts ectopic expression of non-mesodermal genes in MLPs.

TBX1 defines a gene regulatory network in the MLPs for cardiac and BrM formation
To better understand how TBX1 regulates the expression of genes in the CPM at the chromatin level, we performed ATAC-seq of control versus Tbx1 mutant embryos (Fig. 6a The remaining CARs and DARs are thus referred to as "CARs-Mesp1" and "DARs-Mesp1" (Fig.   6c, d). A total of 81.4% of CARs-Mesp1 were in promoter regions, while 35.8% of DARs-Mesp1 were in promoter regions and 34.9% were in distal intergenic regions (Extended Data Fig. 6e, f), suggesting that Tbx1 inactivation has a large effect on regulation of genes at a distance from their promoters.
We then focused upon MLP genes within the CPM. A total of 80 of the 262 downregulated genes in the MLPs in Tbx1 cKO embryos were associated with DARs-Mesp1 (Fig. 6f). In DARs-Mesp1, several transcription factors binding motifs related to heart or BrM differentiation including the T-box motif were identified, and many of these are likely MLP relevant transcription factors (Fig. 6e). We also used GREAT 42 to identify genes that harbored DARs-Mesp1 and their enriched functions. GO analysis of the 80 genes indicated that these genes were involved in skeletal muscle cell differentiation, pattern specification process and cardiac muscle tissue development, important for the function of Tbx1 (Fig. 6g). On the other hand, those without DARs-Mesp1 were not specific to the CPM populations (Fig. 6h). Despite the fact that more genes are upregulated when Tbx1 is inactivated, they contained only 13 DARs that increased in accessibility, and none were associated with the DARs-Mesp1. Therefore, Tbx1 might indirectly affect these genes in the MLPs that were not detected by ATAC-seq analysis.
To determine which genes with DARs could be direct target genes of TBX1, we performed ChIP-seq with an Avi-tagged Tbx1 mouse line that we generated, and created double  Fig. 7f). Of the 255 ChIP-seq peaks, 151 (59%) did not show significant chromatin accessibility changes (i.e., overlapping with DARs), which include some TBX1 binding sites that were located in closed regions that did not change in accessibility in the mutant data (Fig. 7c), suggesting a diverse role of TBX1 in promoting chromatin remodeling.
We intersected the genes with DARs-Mesp1 and the 262 genes reduced in the MLPs in Tbx1 KO embryos to identify MLP specific gene regulation. When we further intersected the ATAC-seq and TBX1 ChIP-seq, we found eight genes ( Fig. 7b), including Aplnr and Nrg1, which are MLP enriched genes that are TBX1 direct transcriptional targets (Fig. 7d). In the Nrg1 locus, the TBX1 binding regions were closed in Tbx1 cKO embryos, while in the Aplnr gene region, the TBX1 binding site was not in a DAR (Fig 7d), indicating that multiple mechanisms of regulation occur.
Taking the results from the three types of functional genomic data in this report, we can generate a putative gene regulatory network for TBX1 function in the MLPs as summarized in Figure 7e. Here, we distinguish four categories of genes regulated by TBX1: 1) Direct target genes with or 2) without chromatin changes, and indirect target genes 3) with chromatin changes, that contain transcription factor binding sites and 4) without chromatin changes. Note that some DEGs changed expression in more than one cell type or population as indicated (gray lines). Overall, we suggest that TBX1 with Isl1, Fox, Six, Pitx and E-box proteins such as Tcf21, (Fig. 6e, 7e), act together to regulate progression of MLPs to more differentiated states in the CPM.

Discussion
Single cell RNA-seq is uncovering MLPs in different developmental contexts, these cells serving as progenitor states in organogenesis. In this report, we uncovered the cellular and molecular processes by which a specific MLP population functions to both maintain a progenitor state and allocate cells to derivative cell types. We discovered that MLPs are localized bilaterally within the posterior nascent mesoderm of the pharyngeal apparatus. The MLPs are needed for deployment of progenitor cells to derivative tissues during the time when individual arches are forming in a rostral to caudal manner, thereby elongating the pharyngeal apparatus.
We focused on function of Tbx1, the gene for 22q11.2DS, which marks the MLP lineage in the CPM. In Figure 7F, we show a model for Tbx1 function. Inactivation of Tbx1 results in dysregulation of gene expression in MLPs, and reduced progression to more mature states. In turn, this causes an accumulation of MLPs and reduction of more differentiated CPM cells, the latter resulting in fewer cardiomyocytes leading to a shortened outflow tract. To better understand TBX1 function we performed ATAC-seq and TBX1 ChIP-seq, and we defined a gene regulatory network in the CPM in which TBX1 plays a key role, thereby helping to explain the cause of the cardiac and BrM phenotype in 22q11.2DS patients. Loss of Tbx1 results in intermittent failure of BrMs from forming 19 and acts upstream of many of these important transcription factors 24 . This suggests that the Tbx1 gene regulatory network is central in MLPs and, independently to BrM specification.

Molecular mechanism for Tbx1 function
In this report, we identify a novel function of Tbx1, in a newly recognized CPM subpopulation termed MLPs. We found that Tbx1 is required for the MLPs to progress, particularly for the aSHF, CM and BrM fates. Therefore, one of the main functions of the MLPs is to allocate CPM cells to more differentiated states. Second, we found that Tbx1 contribution is for the aSHF, CM and BrM cells and less so for the Tbx5 expressing pSHF cells 25 , consistent with the expression pattern of Tbx1. To understand how the MLPs transition and with respect to Tbx1, we defined a gene regulatory network which provides a molecular mechanism that includes new genes for CPM function during the acquisition of aSHF, CM and BrM fates. By performing multi-omic studies, we identified Aplnr and Nrg1 among the genes enriched in expression in MLPs. Aplnr encodes the APJ/Apelin G-protein coupled receptor that binds Apelin or Elabela/Toddler peptide ligands that have many embryonic and adult functions 52 . In zebrafish, knockdown of Aplnr disrupts normal migration of cells during gastrulation including that of cardiac progenitors resulting in severe defects 53,54 . Unexpectedly, their role in early embryogenesis is not recapitulated fully in mouse models, implicating perhaps functional redundancy with other G-protein coupled receptors or ligands 55 . Aplnr is expressed in the CPM, and from stem cell studies, has a role in cardiomyocyte development 29 . Nrg1 is also of particular interest. In contrast to Aplnr, Nrg1 is not expressed in the DPW. Nrg1 encodes an EGF family ligand that binds to ErbB receptor tyrosine kinases and has multiple roles in cardiac development and function 30,56 . Interestingly, both Aplnr and Nrg1 are direct target genes of TBX1 based on our ATAC-seq and ChIP-seq results, suggesting that these genes are mediators of TBX1 function in MLPs.
Some of the genes that were differentially expressed and differentially accessible in This is particularly true for the BrM progenitor cells, where Tbx1 is actively expressed. Note that our finding of chromatin status changes in Tbx1 mutant embryo is different as compared to analysis of cell culture 66 , suggesting differences between cell culture versus the in vivo context or differentiation stages.

Conclusions
In this study, we identified MLPs as a continuous but evolving source of CPM cells that is maintained during development of the pharyngeal apparatus. Tbx1 marks the MLP cell population in embryogenesis. Further, Tbx1 is required in MLPs to promote their maturation to more differentiated states by direct and indirect regulation of chromatin accessibility and transcriptional regulation.

Mice
All experiments using mice were carried out according to regulatory standards defined by the National Institutes of Health and the Institute for Animal Studies, Albert Einstein College of Medicine (https://www.einstein.yu.edu/administration/animal-studies/), IACUC protocol is #0000-1034.
The following mouse mutant alleles used in this study have been previously described:  Table   1.

ATAC-seq
The ATAC-seq method has been previously described 70

Sequencing
The DNA libraries were sequenced using a Illumina HiSeq2500 system (at Einstein Epigenomics Core Facility), Illumina HiSeq4000 system (at Genewiz, South Plainfield, NJ) or NovaSeq6000 system (at Novogene, Sacramento, CA), with paired-end, 100 bp read length.
Incubation with xylene (Thermo Fisher scientific, Cat# X3S-4) was done at RT, one hour twice followed by incubation of paraffin (Thermo Fisher Scientific, Cat# T555) at 65˚C, one hour twice.
Embryos were embedded and stored at 4˚C.

scRNA-seq data analysis
We utilized Cell Ranger (v 3.1.0, from 10x Genomics) to align reads of scRNA-seq data to the mouse reference genome (mm10). All the samples passed quality control measures for Cell Ranger (Table 1), and the filtered gene-barcode matrices were used for the following analyses. For Mesp1 Cre four time point dataset analyses, Scran v1.10.2 was used to normalize the individual data sets by "computeSumFactors" function with the deconvolution method for scaling normalization 71 . The cells were clustered by densityClust v0.3 with the density peak clustering algorithm 21 . We found batch effects exist in the scRNA-seq data sets from different time points (Mesp1 Cre data) or experimental perturbations (Mesp1 Cre Tbx1 or Tbx1 Cre Tbx1 Ctrl vs cKO data). Therefore, we performed batch corrections before we comprehensively analyzed gene expression values across these scRNA-seq data sets. We employed the MNN (mutual nearest neighbors) method to identify shared cell types across data sets, and corrected The scRNA-seq data can be viewed at this URL: https://scviewer.shinyapps.io/heartMLP

ATAC-seq data analysis
ATAC-seq analysis pipeline has been described previously 66 .
We removed Nextera Transposase Sequence primers 77

ChIP-seq data analysis
We removed adapters using cutadapt 78 with -a option and a set of adapters detected with FASTQC 86 . Sequences were then trimmed using TrimGalore with option -length 0. Then sequences were aligned to the mouse genome (mm9) using Bowtie2 2.3.4. with default parameters. Only uniquely mappable reads were retained. A customized R script was used to remove reads with mates mapping to different chromosomes, or with discordant pairs orientation, or with a mate-pair distance >2 kb, or PCR duplicates (defined as when both mates are aligned to the same genomic coordinate). ChIP-seq peaks in each sample were identified using MACS2 2.1.2.1 with default parameters. Then, a consensus list of enriched regions was obtained using the intersectBed function from the BedTools 2.29 with the default minimum overlap and retaining only the peak regions common to at least two out of the three replicates.
Peaks were filtered by removing those overlapping with blacklist regions (Encode mm9 black regions Version 2) using findOverlapsOfPeaks of ChIPpeakAnno. The transcription factor binding motifs were obtained using the findMotifsGenome program with -size given parameter of the HOMER suite. For peak annotation as cis-regulatory regions, GREAT was used with the default settings. The comparison of the gene lists of DAR, DEG and ChIP regions was performed using standard R-scripts. IGV 2.4.8 85 was used for peak visualization. Coverage heat-maps and average enrichment profiles (TSS +/-10Kb) in each experimental condition were obtained using ngs.plot or deepTools2. Significance of the overlap between two list of peaks was evaluated using the ChIPseeker enrichPeakOverlap using mm9 annotation.

Data and Code Availability
The datasets generated during this study are available at GEO repository under the accession number GSE158941. The scRNA-seq data can be viewed at this URL: https://scviewer.shinyapps.io/heartMLP

Competing Interests statement
The authors have declared that no competing interests exist.               Figure 2A.
B. Single-cell embedding graph separated by stages, colored by pseudotime. The color spectrum from red/orange is the early pseudotime point to blue/purple is late pseudotime point.

C.
Single-cell embedding graph of the CPM lineages separated by stages, colored by clusters.
Cluster colors are consistent with those in Figure 2D.

D.
Single-cell embedding graph colored by expression level. The color spectrum from blue, then green to yellow indicates expression levels from low to high. Grey indicates no expression.
The genes are consistent with Figure 2F.