1. Cell Biology
Download icon

Single-cell transcriptomics defines heterogeneity of epicardial cells and fibroblasts within the infarcted murine heart

  1. Julia Hesse
  2. Christoph Owenier
  3. Tobias Lautwein
  4. Ria Zalfen
  5. Jonas F Weber
  6. Zhaoping Ding
  7. Christina Alter
  8. Alexander Lang
  9. Maria Grandoch
  10. Norbert Gerdes
  11. Jens W Fischer
  12. Gunnar W Klau
  13. Christoph Dieterich
  14. Karl Köhrer
  15. Jürgen Schrader  Is a corresponding author
  1. Department of Molecular Cardiology, Medical Faculty and University Hospital Düsseldorf, Heinrich-Heine-University Düsseldorf, Germany
  2. Biologisch-Medizinisches-Forschungszentrum (BMFZ), Genomics & Transcriptomics Laboratory, Heinrich-Heine-University Düsseldorf, Germany
  3. Algorithmic Bioinformatics, Heinrich-Heine-University Düsseldorf, Germany
  4. Division of Cardiology, Pulmonology and Vascular Medicine, Medical Faculty and University Hospital Düsseldorf, Heinrich-Heine-University Düsseldorf, Germany
  5. Institute of Pharmacology and Clinical Pharmacology, Medical Faculty and University Hospital Düsseldorf, Heinrich-Heine-University Düsseldorf, Germany
  6. Section of Bioinformatics and Systems Cardiology, Klaus Tschira Institute for Integrative Computational Cardiology and Department of Internal Medicine III, University Hospital Heidelberg, Germany
Research Article
  • Cited 1
  • Views 406
  • Annotations
Cite this article as: eLife 2021;10:e65921 doi: 10.7554/eLife.65921

Abstract

In the adult heart, the epicardium becomes activated after injury, contributing to cardiac healing by secretion of paracrine factors. Here, we analyzed by single-cell RNA sequencing combined with RNA in situ hybridization and lineage tracing of Wilms tumor protein 1-positive (WT1+) cells, the cellular composition, location, and hierarchy of epicardial stromal cells (EpiSC) in comparison to activated myocardial fibroblasts/stromal cells in infarcted mouse hearts. We identified 11 transcriptionally distinct EpiSC populations, which can be classified into three groups, each containing a cluster of proliferating cells. Two groups expressed cardiac specification markers and sarcomeric proteins suggestive of cardiomyogenic potential. Transcripts of hypoxia-inducible factor (HIF)-1α and HIF-responsive genes were enriched in EpiSC consistent with an epicardial hypoxic niche. Expression of paracrine factors was not limited to WT1+ cells but was a general feature of activated cardiac stromal cells. Our findings provide the cellular framework by which myocardial ischemia may trigger in EpiSC the formation of cardioprotective/regenerative responses.

Introduction

Myocardial infarction (MI), still the most frequent cause of death in western societies, is associated with massive activation of cardiac fibroblasts, ultimately resulting in excessive accumulation of extracellular matrix (ECM) components that finally impair cardiac function (Tallquist and Molkentin, 2017). During development, the majority of cardiac fibroblasts are derived from the epicardium, which forms the thin outermost epithelial layer of all vertebrate hearts and exhibits extensive developmental plasticity (Cao and Poss, 2018). A subset of epicardial cells undergoes epithelial to mesenchymal transition (EMT) and those epicardial progenitor cells can give rise to various cardiac cell types. In addition to cardiac fibroblasts, these include vascular smooth muscle cells and pericytes, which contribute to the coronary vasculature (Cao and Poss, 2018). Embryonic epicardial heterogeneity was recently studied at the single-cell level in the developing zebrafish heart and uncovered three epicardial populations, functionally related to cell adhesion, migration, and chemotaxis (Weinberger et al., 2020).

In the adult heart, the epicardium is a rather quiescent monolayer that becomes activated after MI by upregulating embryonic epicardial genes (Quijada et al., 2020; Masters and Riley, 2014). In the injured heart, epicardial cells form via EMT a multi-cell layer of epicardial stromal cells (EpiSC) at the heart surface that can reach a thickness of about 50–70 µm in mice (van Wijk et al., 2012). It is generally assumed that the activated epicardium recapitulates the embryonic program in generating mesenchymal progenitor cells, although there may be major molecular differences with respect to their embryonic counterpart (Bollini et al., 2014). On the functional side, adult EpiSC secrete paracrine factors that stimulate cardiomyocyte growth and angiogenesis (Zhou et al., 2011) and play a key role in post-MI adaptive immune regulation (Ramjee et al., 2017). When stimulated with thymosin β4, Wilms tumor protein 1-positive (WT1+) adult EpiSC can form cardiomyocytes; however, the rate of conversion is only small (Smart et al., 2011). Thus, the epicardium is a signaling center regulating cardiac wound healing and may have cardiogenic potential in the adult injured heart.

Despite its importance in cardiac repair, little is known on cell heterogeneity and molecular identifiers within the epicardial layer of the adult heart. Yet, this knowledge is essential to map the epicardial progeny and attribute meaningful functions. While the single-cell landscape of activated cardiac fibroblasts (activated cardiac stromal cells, aCSC) in the post-MI heart has been explored in detail (Farbehi et al., 2019; Forte et al., 2020), these studies did not assess epicardial heterogeneity because of a lack of specific identifiers. In a previous study, we have reported a novel perfusion-based technique (Owenier et al., 2020), which permitted the simultaneous isolation of EpiSC and aCSC with high yields and only minimal cell activation. The isolation of viable, purified preparations of aCSC and EpiSC from the infarcted heart permitted the first direct comparison of the two cardiac stromal cell fractions. We have combined single-cell RNA sequencing (scRNAseq) with lineage tracing of WT1-expressing cells and localization of cell populations by RNA in situ hybridization, characterized cellular hierarchy of adult EpiSC, defined similarities and differences to cardiac fibroblasts, and explored in detail the individual EpiSC populations in the activated epicardium.

Results

scRNAseq of post-MI stromal cells

EpiSC and aCSC were isolated from the same mouse hearts (n = 3) 5 days after MI (50 min ischemia followed by reperfusion) by combined collagenase perfusion and bathing (Figure 1A). This recently developed technique selectively removes the majority of EpiSC from the aCSC-containing myocardium by applying gentle shear forces to the cardiac surface and has been validated for cell yield, viability, and purity (Owenier et al., 2020). Day 5 was chosen because the epicardial cell layer had reached its maximal dimension (Ding et al., 2016) and fetal epicardial genes as a measure of epicardial activation were re-expressed (Zhou et al., 2011). CSC were isolated from uninjured hearts of sham-operated animals (n = 3) without separation of epicardial cells (see 'Materials and methods'). Cell preparations were depleted of cardiomyocytes, endothelial cells, and immune cells prior to scRNAseq using the 10x Genomics Chromium platform. Transcriptional profiles of 13,796 EpiSC, 24,470 aCSC, and 24,781 CSC were captured after quality control filtering. Unbiased clustering using the Seurat R package with visualization in uniform manifold approximation and projection (UMAP) dimension reduction plots was performed to identify cells with distinct lineage identities and transcriptional profiles. Average and significantly enriched RNA expression of EpiSC, aCSC, and CSC are listed in Supplementary files 1 and 2, respectively.

Figure 1 with 4 supplements see all
Cell populations in EpiSC from the infarcted heart.

(A) Schematic workflow. Epicardial stromal cells (EpiSC) and activated cardiac stromal cells (aCSC) were simultaneously collected from the surface and the myocardium of the isolated perfused heart by applying mild shear forces to the cardiac surface (Owenier et al., 2020) at 5 days post myocardial infarction (MI) (n = 3). CSC were purified from three non-infarcted control hearts 5 days after sham surgery. Mesh purification, low-speed centrifugation, and cell sorting by flow cytometry were performed to remove cardiomyocytes, CD31+ endothelial cells, CD45+ immune cells, and apoptotic or necrotic cells before analysis using the 10x Genomics Chromium platform. (B) UMAP plot of clustered single-cell RNA sequencing (scRNAseq) data of the pooled EpiSC fraction (n = 13,796 single cells). CM, cardiomyocytes; Ery, erythrocytes. (C) Cell count of EpiSC populations. (D) Dendrogram of EpiSC populations according to average RNA expression. (E) Dot plot of top five marker genes for each EpiSC population. (F) RNA velocity of EpiSC populations projected on the UMAP plot. Arrows show the local average velocity evaluated on a regular grid, indicating estimated future states. (G) Feature plot displaying the predicted classification in G1, G2/M, and S phase for each cell according to cell cycle gene expression scores. (H) Feature plot displaying the number of unique genes detected in each cell. (I) Networks visualizing the number of potential specific interactions between EpiSC populations as determined by CellPhoneDB.

Figure 1—source data 1

Source data for EpiSC population cell counts summarized in Figure 1C.

https://cdn.elifesciences.org/articles/65921/elife-65921-fig1-data1-v1.xlsx

Characterization of EpiSC populations

As shown in Figure 1B, we identified 11 transcriptionally different cell populations within the EpiSC fraction. Cell doublets with hybrid transcriptomes identified by DoubletFinder (McGinnis et al., 2019Figure 1—figure supplement 1A) and minor non-stromal cell populations such as cardiomyocytes and erythrocytes, identified according to cell type-specific marker expression (Figure 1—figure supplement 1B), were excluded from further analysis. EpiSC populations varied in size (Figure 1C), were hierarchically structured (Figure 1D), and the differentially expressed genes showed over-representations of distinct gene ontology (GO) biological process terms within each population (Figure 1—figure supplement 2). Differentially expressed population marker genes are summarized as top five dot plot in Figure 1E and top 10 heat map in Figure 1—figure supplement 3. Remarkably, transcripts of epithelial cell-associated genes and established epicardial genes (Bochmann et al., 2010) were enriched in both EpiSC-1 (Dmkn, Saa3) and EpiSC-7 (Msln, Krt7, Krt18, Krt19, Lgals7), while mesenchymal marker genes were primarily expressed in EpiSC-3 (Gsn, Pi16) and EpiSC-4 (Cd44, Ly6a). Genes coding for ECM proteins were highly expressed in EpiSC-2 (Spp1), EpiSC-3 (Smoc2, Sparcl1), EpiSC-5 (Clec3b), EpiSC-6 (Mfap4, Eln), and EpiSC-8 (Col5a3, Tnc). Genes encoding contractile proteins were preferentially expressed in EpiSC-2 (Acta2, Myl9, Tagln). Wnt pathway-associated gene transcripts were enriched in EpiSC-6 (Sfrp2) and EpiSC-10 (Wif1, Dkk3). Genes related to the cellular response to interferon characterized EpiSC-11 (Ifit1, Ifit3, Ifit3b, Cxcl10). Finally, expression of genes associated with high cell cycle activity and mitosis was a feature of EpiSC-9 (Pbk, Top2a, Prc1, Stmn1, Ube2c).

Hierarchy of EpiSC populations

To better define the cellular relationship between the identified EpiSC populations, we performed RNA velocity analysis (La Manno et al., 2018), which can predict and visualize future cell states based on the ratio of unspliced and spliced mRNA read counts. As shown in Figure 1F, this trajectory inference separated the EpiSC populations into three groups: EpiSC-7 and -1 (group I), EpiSC-2, -4, -8, -11 (group II), and EpiSC-6, -5, -3, -9, -10 (group III).

Cell cycle scoring showed that each of the three EpiSC groups identified above was equipped with a cell cluster expressing a number of G2/M and S phase genes (Figure 1G), indicating proliferating cells: within group I, a subpopulation of EpiSC-1, within group II, a subpopulation of EpiSC-2, and within group III, EpiSC-9. This location is very similar to the origin of velocity arrows (Figure 1F).

The number of expressed genes per cell was lower in group III than in groups I and II (Figure 1H). Since the expressed genes per cell correlate with developmental potential (Gulati et al., 2020), this may indicate that a large fraction of group III cells are terminally differentiated.

To study cell–cell communication mediated by ligand–receptor interactions between the EpiSC populations, we used CellPhoneDB (Efremova et al., 2020). As shown in Figure 1I, the highest numbers of interactions were predicted between groups I and II. Ligand–receptor pairs potentially involved in the interaction between the groups include IGF1, transforming growth factor beta-2 (TGF-β2), and Wnt4 signaling (Supplementary file 3, Figure 1—figure supplement 4).

Epicardial marker gene expression

The cellular distribution of well-established epicardial progenitor marker genes such as Wt1, Tbx18, Sema3d, Aldh1a2, Gata5, and Tcf21 (Cao and Poss, 2018) is shown in Figure 2A. Wt1, commonly used for lineage tracing studies (Quijada et al., 2020), was found to be highly enriched in group I EpiSC populations 1 and 7, as well as the adjacent population EpiSC-8. Sema3d, Aldh1a2, and Gata5 were also predominantly expressed in group I, with the highest expression levels in EpiSC-7. Tbx18, on the other hand, was broadly distributed in group I and II populations EpiSC-1, -2, -4, -7, -8, -9, and -11 (Figure 2A). In contrast, Tcf21 was strongly expressed in group III EpiSC-3, -5, -6, and -9 and showed no overlap with Wt1-expressing populations (Figure 2A).

Figure 2 with 3 supplements see all
Molecular characterization and location of EpiSC populations.

(A) Expression of epicardial progenitor cell markers proposed in the literature as visualized in feature and violin plots. (B) Upper panel: RNA in situ hybridization of epicardial stromal cell (EpiSC) population identifiers (red) in heart cryosections 5 days post myocardial infarction (MI). Representative images (n = 4 hearts) of the infarct border zone are shown. The dotted line marks the interface between myocardial/epicardial tissue according to cell morphology. Nuclei were stained with hematoxylin (blue). Scale bars, 50 µm. Lower panel: feature plots visualizing selected EpiSC population molecular identifiers and scheme of the post-MI heart with the infarct zone in grey. RV, right ventricle; LV, left ventricle. (C–E) Lineage tracing of Wt1-expressing cell populations post MI using Wt1CreERT2RosatdTomato mice. The experimental design is outlined in (C). UMAP plot of clustered scRNAseq data of the EpiSC fraction (n = 13,373 single cells) pooled from two hearts 5 days post MI is shown in (D). Expression of Wt1 and tdTomato is visualized in (E).

Since cells in EpiSC-1 showed substantial inhomogeneity of markers (Figure 2A), we carried out a separate clustering analysis for EpiSC-1, including the two adjacent Wt1-expressing populations EpiSC-7 and -8. We identified five subclusters in EpiSC-1 (EpiSC-1.1, -1.2, -1.3, -1.4, -1.5), with no subclustering in EpiSC-7 and -8 (Figure 2—figure supplement 1A–C). EpiSC1.4 is characterized by expression of cell cycle-associated genes (Rrm2, Pclaf, Hist1h2ap, Hmgb2, Ube2c, Top2a) (Figure 2—figure supplement 1D; Supplementary files 4 and 5) and correlates to the proliferative EpiSC-1 subpopulation identified by cell cycle scoring (Figure 1G). EpiSC1.5 showed enriched expression of genes encoding core ribosomal proteins such as Rps21 and Rpl37a (Figure 2—figure supplement 1D; Supplementary file 5), suggesting high protein synthesis activity.

We also searched for expression of Tgm2, Sema3f, and Cxcl12, which were recently found to mark three functional different epicardial populations in the developing zebrafish heart (Weinberger et al., 2020). We found Tgm2 and Sema3f preferentially expressed in group I populations EpiSC-1 and -7 (Figure 2—figure supplement 2A–B). However, Mylk, which is an additional marker of the Sema3f-expressing zebrafish epicardial population (Weinberger et al., 2020), was primarily expressed in EpiSC-8. Cxcl12, present only in a small cell population in the zebrafish (Weinberger et al., 2020), was rather broadly expressed with enrichment in EpiSC-2 and subcluster EpiSC-1.1 (Figure 2—figure supplement 2C–D). These findings not only demonstrate some degree of evolutionary preservation in epicardial populations from zebrafish to mice, but also reveal considerable differences in marker gene expression patterns and population sizes.

Spatial mRNA expression of EpiSC population identifiers

To determine the specific location of EpiSC populations in the infarcted heart, RNA in situ hybridization of gene transcripts for selected EpiSC populations was carried out (Figure 2B, upper panel), using Wt1 and other population-specific identifiers (Figure 2B, lower panel). Msln expression (EpiSC-7) was detected on the outmost layers of the epicardium, consistent with the epithelial signature of EpiSC-7. Expression of Wt1 (EpiSC-7, -1, -8) and Cd44 (EpiSC-4) similarly labeled cells that were localized in the outer part. In contrast, Ifit3 (EpiSC-11) expression was also found in deeper layers of the activated epicardium and was additionally detected in the myocardium, labeling aCSC. Expression of group III population markers Sfrp2 (EpiSC-6), Dkk3 (EpiSC-10), and Pcsk6 (EpiSC-3) was rather homogenously dispersed throughout the epicardium and the aCSC in the infarct border zone. Top2a, chosen as an identifier of the proliferative (sub)populations of all three groups (EpiSC-1.4, EpiSC-2 subpopulation, EpiSC-9), showed a similar expression pattern. This scattered distribution of proliferating cells in the post-MI heart and especially within the epicardial multi-cell layer has already been confirmed in BrdU incorporation experiments (van Wijk et al., 2012).

To explore whether the dispersed Wt1-negative group II and group III cells might derive from the Wt1-expressing populations located in the outmost layer of the activated epicardium, we combined scRNAseq of EpiSC with lineage tracing using tamoxifen-inducible Wt1-targeted reporter (Wt1CreERT2RosatdTomato) mice (Figure 2C). Transcriptional profiles of 13,373 cells from two mouse hearts 5 days after MI were assigned to previously defined EpiSC populations (Figure 2D, Figure 2—figure supplement 3). Expression of tdTomato fully overlapped with that of Wt1 (Figure 2E), suggesting that during the first days of MI-triggered epicardial activation and expansion, WT1+ cells did not contribute to cells of Wt1-negative EpiSC populations.

Expression of cardiomyogenesis-associated genes, HIF-1-responsive genes, and paracrine factors

Lineage tracing experiments have shown that cardiomyogenesis can be initiated in vivo in WT1+ epicardial cells when stimulated with thymosin β4 (10). We found thymosin β4 (Tmsb4x) highly expressed in Wt1-expressing EpiSC-8 (Figure 3A and B), together with BRG1 (Smarca4), a transcription activator for WT1 (Vieira et al., 2017). Interestingly, many key cardiogenic factors were also expressed within the Wt1-expressing EpiSC populations (Figure 3A, cardiogenic factors I). These include MESP1, which marks early cardiovascular progenitor specification (Chiapparo et al., 2016), as well as GATA4 (Figure 3B), which in combination with MEF2C and TBX5 has been shown to facilitate cardiomyogenesis (Ieda et al., 2010). Surprisingly, several Hox family members were expressed in EpiSC-8, for example, Hoxa5 (Figure 3A and B). HOX transcription factors are downstream effectors of retinoic acid signaling (Nolte et al., 2019), which is required for differentiation of cardiac progenitors during heart development (Roux and Zaffran, 2016). Intriguingly, a second set of cardiogenic factors was found in group III populations EpiSC-3, -5, -6, -9, and -10 (Figure 3A, cardiogenic factors II). This includes MEF2C (Figure 3B), Nkx-2.5, as well as BMP2 and BMP4, the latter of which are crucial in the regulation of Nkx-2.5 expression and specification of the cardiac lineage (Brown et al., 2004). In addition, we found low expression of genes encoding muscle structural proteins such as Myl2, -4, and -6, Tnnt2, Ttn, and Nebl, which appear to be present especially in EpiSC-7 and EpiSC-9 (Figure 2C, contractile proteins). The highest expression levels of Notch target genes were found in Wt1-negative EpiSC populations, especially EpiSC-10 (Figure 3A). This is remarkable, since Notch-activated epicardial-derived WT1+ cells were described as a multipotent cell population with the ability to express cardiac genes (Russell et al., 2011). The observation that EpiSC-7 and -9 express cardiac specification markers and sarcomere proteins is suggestive that these populations have cardiomyogenic potential.

Functional characterization of EpiSC populations.

(A, B) Heat maps (A) and violin plots (B) of selected genes showing the expression of cardiogenic factors and Notch target genes. (C, D) Heat maps (C) and violin plots (D) of genes associated with HIF-1 signaling. (E, F) Heat maps (E) and violin plots (F) of genes encoding chemokines and further secreted factors.

A hypoxia-responsive element in the Wt1 promoter was reported to bind HIF-1α, which is required for WT1 induction (Wagner et al., 2003). We found expression of Hif1a and HIF-1-responsive genes, particularly those associated with glycolysis, to be enriched in group I and group II EpiSC populations (Figure 3C). These include the lactate-generating enzyme lactate dehydrogenase (Ldha) and the lactate-exporting monocarboxylate transporter 4 (Slc16a3) (Figure 3C and D). Consistently, the genes with significantly higher expression in group I and II populations vs group III populations (Supplementary file 6) showed enrichment of several GO terms related to the hypoxia/HIF-mediated metabolic switch toward oxygen-saving glycolysis, such as glycolytic process (FDR q-value: 3.05E-02), glucose metabolic process (FDR q-value: 3.22E-02), and pyruvate biosynthetic process (FDR q-value: 2.66E-02). Due to the transient nature of the HIF-1 response that strongly depends on local oxygen levels, the enriched expression of HIF-1-responsive genes in group I and group II in comparison to group III suggests that the populations reside in microenvironments with different oxygen levels within the post-MI epicardium.

Epicardial cells have been reported to secrete numerous paracrine factors that can modulate myocardial injury in the mouse heart (Zhou et al., 2011). In line with this observation, we found enriched expression of various chemokines in group II EpiSC populations (Figure 3E). Several chemokines known to be involved in monocyte recruitment were primarily expressed in EpiSC-11, including MCP1 (Ccl2) and MCP3 (Ccl7) (Figure 3E and F). EpiSC-11 also showed strongly enriched expression of IP-10 (Cxcl10), which is involved in triggering anti-fibrotic effects after MI (Bujak et al., 2009). Expression of chemokines involved in neutrophil granulocyte recruitment (GRO-α, -β, -γ, Cxcl1, -2, -3; ENA-78, Cxcl5; NAP-2/CXCL-7, Ppbp) was preferentially found not only in EpiSC-2 and -4 but also in EpiSC-1 and -7. Expression of SDF-1 (Cxcl12), known to reduce scar size when administered to the damaged heart (Sundararaman et al., 2011), was enriched in EpiSC-2.

Expression of proangiogenic factors previously identified in the supernatant of WT1+ epicardial cells (Zhou et al., 2011), such as Angiopoietin-1 (Angpt1), FGF2, IL-6, VEGFA, and VEGFC, was not limited to the Wt1-expressing EpiSC populations but pertained to other identified EpiSC populations (Figure 3E). The most highly expressed paracrine factors were TGF-β1, which attenuates myocardial ischemia-reperfusion injury (Chen et al., 2003), IGF-1, which prevents long-term left ventricular remodeling after cardiac injury (Dai and Kloner, 2011Figure 3F), and MYDGF, which mediates ischemic tissue repair (Korf-Klingebiel et al., 2015Figure 3F).

Post-MI aCSC in comparison to EpiSC

Clustering analysis of aCSC revealed 11 transcriptionally different populations (Figure 4—figure supplement 1A–C). Again, cell doublets with hybrid transcriptomes identified by DoubletFinder (McGinnis et al., 2019Figure 4—figure supplement 2A) and minor non-stromal cell populations (Figure 4—figure supplement 2B) were excluded from further analysis. A dot plot of the top five and a heat map of the top 10 most differentially expressed genes in aCSC populations are displayed in Figure 4—figure supplement 1D and Figure 4—figure supplement 3, respectively. To underline the segregation of aCSC populations with partially overlapping marker genes (aCSC-4, -5, and aCSC-7, -10, respectively), the expression of selected population marker genes is also visualized in feature plots (Figure 4—figure supplement 1E). Genes encoding ECM proteins were associated with aCSC-1 (Eln, Wisp2), aCSC-2 (Thbs4, Cilp), and aCSC-4 (Smoc2, Sparcl1). Cell populations with similar gene signatures were recently termed activated fibroblasts (Farbehi et al., 2019) or late-response fibroblasts and matrifibrocytes (Forte et al., 2020). In these studies, a population similar to aCSC-5 was referred to as either Sca1-low or homeostatic epicardial-derived fibroblasts (Farbehi et al., 2019; Forte et al., 2020). aCSC-6 preferentially expressed genes characteristic of epithelial cells (Lgals7, Dmkn, Msln). Because this epithelial signature was comparable to that observed for EpiSC-7, it is very likely that cells of aCSC-6 are of epicardial origin and are included in the aCSC fraction due to an incomplete removal of the EpiSC layer from the myocardium as to be expected from the used cell isolation technique (Owenier et al., 2020). Consistent with this, in situ hybridization identified expression of the aCSC-6 and EpiSC-7 marker Msln exclusively within the epicardium (Figure 2B). Similar to so-called cycling fibroblasts/proliferating myofibroblasts (Farbehi et al., 2019; Forte et al., 2020), we found genes involved in cell cycle and mitosis preferentially in aCSC-7 (Cenpa, Stmn1, Ccnb2) and aCSC-10 (Top2a, Ube2c, Hist1h2ap, Pclaf). Again similar to data in the literature (Farbehi et al., 2019; Forte et al., 2020), we found genes related to the cellular response to interferon highly expressed in aCSC-8 (Ifit3, Isg15, Ifit1, Iigp1, Ifit3b), which were termed interferon-stimulated/interferon-responsive fibroblasts. aCSC-9 highly expressed Ly6a, encoding Sca1, and resembled the reported Sca1-high/progenitor-like fibroblasts (Farbehi et al., 2019; Forte et al., 2020). Genes encoding contractile proteins were preferentially expressed in aCSC-3 (Acta2, Tpm2), which were recently referred to as myofibroblasts (Forte et al., 2020). Expression of Wnt pathway-associated genes was enriched in aCSC-1 (Sfrp1, Sfrp2) and aCSC-11 (Wif1, Dkk3), the latter of which were referred to as Wnt-expressing/endocardial-derived fibroblasts (Farbehi et al., 2019; Forte et al., 2020). Taken together, all aCSC populations identified by us are consistent with previously identified cardiac fibroblast populations.

To directly compare the transcriptional profile of EpiSC to myocardial aCSC, we performed canonical correlation analysis (CCA) space alignment of these two post-MI scRNAseq data sets, together with CSC from uninjured hearts. Generated CCA clusters (Figure 4A, left panel) are also displayed according to original cell IDs (Figure 4A, right panels). Average RNA expression levels and differentially expressed marker genes of CCA clusters are listed in Supplementary files 7 and 8, respectively. The contribution of EpiSC, aCSC, and CSC fractions to CCA clusters is summarized in Figure 4B. As can be seen, CCA clusters A, D, and F were mainly composed of EpiSC, indicating that the transcriptional profile of EpiSC-1, -4, -7, and -8 is prevalent in the epicardium (Figure 4B). On the other hand, CCA clusters B, C, H, I, J, and O were dominated by aCSC. Individual assignments showed that Wt1-negative EpiSC-2, -6, and -9 carried to variable degree expression signatures of genuine aCSC-1, -2, -3, -4, -5, and -10. Despite these similarities, there were multiple significant differences in average gene expression levels when comparing Wt1-negative EpiSC and aCSC within individual CCA clusters (Supplementary file 7). Wt1-expressing CCA clusters A–D were excluded from this comparative analysis, since Wt1 expression was only found at the heart’s surface (Figure 2B), strongly suggesting that aCSC assigned to these clusters probably represent epicardial contamination in the aCSC fraction (see above). As shown in Figure 4C, expression of epithelial/epicardial genes was generally higher in EpiSC, while mesenchymal/fibroblast genes were more dominant in aCSC. A distinct fraction of cardiogenic factors set I (see Figure 3A) was preferentially expressed in EpiSC (Figure 4D) while the opposite was true for cardiogenic factors set II (Figure 4E). Notably, cardiac contractile proteins (Figure 4F) and HIF-1-responsive glycolytic enzymes (Figure 4G) were predominantly expressed within EpiSC. Among the paracrine factors (Figure 4H), we found multiple chemokines highly enriched in EpiSC.

Figure 4 with 7 supplements see all
Comparison of EpiSC, aCSC, and CSC.

(A) Canonical correlation analysis (CCA) space alignment of epicardial stromal cells (EpiSC), activated cardiac stromal cells (aCSC), and CSC single-cell RNA sequencing (scRNAseq) data in one UMAP plot (left) and split in one plot each (right). Cells are color-coded according to their assignment to CCA clusters (left) or previously identified populations (right). (B) Contribution of EpiSC, aCSC, and CSC fractions to CCA clusters. n.d., not detected. (C–H) Comparison of EpiSC and aCSC gene expression profiles in CCA clusters D-O regarding epicardial and stromal cell markers (C), cardiogenic factors (D and E), HOX transcription factors, contractile proteins and Notch target genes (F), genes associated with HIF-1 signaling (G) as well as chemokines and other secreted factors (H) as log2 fold change. n.d., not defined. (I–K) Gene regulatory network analysis in EpiSC (I), aCSC (J), and CSC (K) populations by SCENIC.

To further compare cell states in our stromal cell fractions, we used SCENIC (Aibar et al., 2017) for gene regulatory network reconstruction. This tool scores the activity of transcription factors by correlating their expression with the expression of their direct-binding target genes (Aibar et al., 2017). As shown in Figure 4I and J, EpiSC and aCSC showed distinct patterns of network activity, again emphasizing their different cellular identities and functions. As to be expected, both post-MI fractions showed major differences in network activity compared to CSC from uninjured hearts (Figure 4K).

Comparison to CSC from uninjured hearts

Clustering analysis of CSC revealed 12 transcriptionally different populations (Figure 4—figure supplement 4A–C) after cell doublets (Figure 4—figure supplement 5A) and minor non-stromal cell populations (Figure 4—figure supplement 5B) were excluded. Interestingly, the smallest CSC population (CSC-12) highly expressed epithelial/epicardial genes (Msln, Upk3b, Nkain4, Krt19) (Figure 4—figure supplement 4D, Figure 4—figure supplement 6) and most likely represents cells of the epicardial monolayer of the uninjured heart. Individual single-cell analysis of this monolayer is technically not feasible, because the number of cells liberated by shear forces from uninjured hearts is only rather small (estimated ~5600 cells/heart) and there is shear-independent release of cells (background) of unknown origin (Owenier et al., 2020). CSC-12 was the only Wt1-expressing population in the CSC fraction (Figure 4—figure supplement 4E).

As shown in Figure 4A–B, CCA space alignment of the CSC and EpiSC data sets revealed that the majority of CSC correlated with EpiSC group III populations EpiSC-3, -5, and -6 (CCA cluster K, L, M). Direct comparison of EpiSC and CSC within individual CCA clusters again showed that expression of epithelial/epicardial genes was enriched in EpiSC, while transcript levels of the conventional fibroblast marker Gsn were higher in CSC (Figure 4—figure supplement 7A). EpiSC showed a generally higher expression of HIF-1-responsive genes (Figure 4—figure supplement 7B) and proliferation-associated genes (Figure 4—figure supplement 7C), consistent with the notion of increased epicardial hypoxia and induction of epicardial cell proliferation in the infarcted heart. Indeed, direct comparison of the epithelial populations EpiSC-7 and CSC-12 in CCA cluster A revealed that HIF-1-responsive/glycolysis-associated genes (Eno1, Gale, Ldha, Slc16a3) and proliferation-associated genes (Calm, Cenpe, Ran, Ranbp1, Spc25, Ube2c) are upregulated in post-MI EpiSC-7 (Figure 4—figure supplement 7D).

Discussion

In this study, we provide a single-cell landscape of the post-MI epicardium with 11 transcriptionally different cell populations. This amazing degree of cellular heterogeneity is similar to that of myocardial cardiac fibroblasts (Farbehi et al., 2019; Forte et al., 2020). The widely used epicardial lineage marker gene Wt1 (Quijada et al., 2020) was selectively expressed in three populations (EpiSC-1, -7, -8), which were localized on the outer surface of the activated epicardium (Figure 2B). Other commonly used lineage markers and recently identified epicardial population markers of the developing zebrafish heart (Weinberger et al., 2020) were rather heterogeneously distributed (Figure 2A, Figure 2—figure supplement 2A–B), suggesting that they mark different stages of differentiation, commitment, or activity in the adult heart.

The quality of scRNAseq critically depends on the cell isolation technique, which ideally should preserve the native state of cells as close as possible. To minimally perturb the native expression profile of cardiac stromal cells, we used a recently elaborated perfusion protocol for the simultaneous isolation of EpiSC and aCSC from the same infarcted heart, which is short (8 min) and results in a high yield of viable cells (Owenier et al., 2020). The application of mild shear forces on the cardiac surface by a simple motor-driven device permitted the rather selective removal of the EpiSC fraction (Owenier et al., 2020). As compared to a commonly used mincing protocol (30–40 min), we found the yield of aCSC to be considerably higher with our technique, and this was associated with significant lower induction of immediate early response genes (Owenier et al., 2020). Because of these reasons, our data are difficult to compare with published studies at single-cell resolution of the post-MI heart, which relied on a mincing protocol. These studies have identified either no (Skelly et al., 2018; Gladka et al., 2018; Kretzschmar et al., 2018) or only one (Farbehi et al., 2019; Forte et al., 2020; Cui et al., 2019; Asp et al., 2019) epicardial cell population.

RNA velocity analysis (Figure 1F) and cell cycle scoring (Figure 1G) suggest that the EpiSC populations can be classified into three different independent population groups, each containing a cluster of proliferating cells. Marker genes of group I, comprising Wt1-expressing EpiSC-1 and -7, have been reported in healthy adult mouse epicardium obtained by laser capture (Bochmann et al., 2010) and have been exclusively used in previous single-cell studies to identify epicardial cells (Wagner et al., 2003; Bujak et al., 2009Figure 1F). Surprisingly, group I cells accounted for only 26% of the EpiSC fraction from the injured heart. Furthermore, the expression of several paracrine proangiogenic factors that were previously considered specifically derived from WT1+ epicardial cells (Zhou et al., 2011) was not limited to group I EpiSC, but was even higher in other EpiSC populations (Figure 3E). Thus, about two-third of all epicardial cells were Wt1-negative, but they are likely to be also involved in the secretion of paracrine factors. Furthermore, aCSC expressed numerous paracrine factors (Figure 4H). Therefore, secretion of paracrine factors appears to be a general feature of both epicardial and myocardial stromal cells, which all can contribute to cardioprotection. Group II, consisting of four cell populations, represented 40% of the EpiSC fraction and showed enriched expression of chemokines known to be involved in attraction of monocytes and neutrophils (Figure 3E and F). The expression of these chemokines was a general feature of EpiSC in comparison to aCSC (Figure 4H). This finding points to a role of epicardial cells in the modulation of the innate immune response post MI, as was already suggested with regard to adaptive immune regulation during the post-MI recovery phase (Ramjee et al., 2017).

Cells with the transcriptional profiles of group I and II populations were generally prevalent in EpiSC in comparison to aCSC (Figure 4B). Group I and II populations also showed the highest number of potential ligand–receptor interactions (Figure 1I). Interestingly, EpiSC-8, characterized by expression of Wt1 and high transcript levels of HOX transcription factors, also highly expressed thymosin β4 (Figure 3A and B), which has tissue-regenerating properties (Smart et al., 2011). The pronounced expression of cardiogenic factors and contractile proteins in group I (Figure 3A) and the parallel expression of WT1 and thymosin β4 suggest that this cellular network has cardiogenic potential and was involved in the previously reported formation of cardiomyocytes from WT1+ cells after thymosin β4 stimulation (Smart et al., 2011; Wang et al., 2021). Interestingly, group I and II also shared high expression of HIF-1-responsive glycolytic enzymes (Figure 3C), which again was a general feature of EpiSC compared to aCSC (Figure 4G). This is in support of the epicardium being a hypoxic niche, which was based on the immunostaining of dispersed HIF-1α-positive cells in the epicardium (Kimura and Sadek, 2012). Since WT1 expression is HIF-1-dependent (Wagner et al., 2003), this is consistent with the view that epicardial HIF-1 signaling is likely to be an important trigger in the ischemic heart to promote cardioprotection.

Group III comprised five cell populations accounting for 34% of all EpiSC. In contrast to group I, group III EpiSC were found localized throughout the activated epicardium (Figure 2B) and group III population identifiers also labeled stromal cells within the myocardium (Figure 2B). This finding is consistent with the transcriptional profile of group III EpiSC, which was quite similar to that of major aCSC and CSC populations (Figure 4A and B). This demonstrated that group III cells exhibit a fibroblast-like phenotype. In addition, group III EpiSC showed the lowest numbers of predicted cell–cell interactions (Figure 1I) and expressed a lower number of genes (Figure 1H), suggesting a less active cell state. Another remarkable feature of group III was the expression of a second set of cardiogenic factors (Figure 3A). However, these cardiogenic factors were prevalently expressed in aCSC (Figure 4E), which is consistent with the postulated cardiogenic potential of cardiac fibroblasts (Furtado et al., 2014).

That there are different and genuine cellular identities of EpiSC and aCSC is supported by CCA space alignment analysis and gene regulatory network analysis, which revealed distinct expression of epicardial vs mesenchymal/fibroblast genes (Figure 4C) and individual patterns in transcription factor activity (Figure 4I and J), respectively. That a direct comparison of cellular expression signatures between EpiSC and aCSC also revealed many transcriptional similarities was not surprising in view of their close developmental relationship (Doppler et al., 2017).

Histological observations by Zhou et al., 2011 and Quijada et al., 2020 found no evidence for migration of WT1+ cells into the infarcted myocardium even after longer periods of time. That WT1+ cells most likely do not migrate into the infarcted heart appears to be in contrast with a study using lentiviral labeling of epicardial cells after pericardial injection of virus expressing fluorescent proteins (Gittenberger-de Groot et al., 2010). Along the same line we have previously shown, that tracking of epicardial cells after labeling with fluorescently marked nanoparticles revealed migration into the injured heart (Ding et al., 2016). Since group II and III EpiSC constitute the majority of post-MI epicardial cells, it is well conceivable that populations of Wt1-negative cells were preferentially marked by the above-mentioned labeling techniques, and this may explain the reported migration into the injured myocardium. Our histological analysis of the epicardial post-MI multi-cell layer revealed that especially cells of group III populations are located in close proximity to the underlying myocardium (Figure 2B). In contrast, Wt1-expressing group I cells were located in the outmost part of the activated epicardium. Furthermore, our lineage tracing data suggest that until day 5 post MI, WT1+ cells have not contributed to Wt1-negative group II and group III populations (Figure 2E). Therefore, group I cells appeared to be spatially restricted to the surface layers, while the major epicardial expansion seemed to be driven by Wt1-negative EpiSC of groups II and III with their respective proliferative cell (sub)populations. It is also conceivable that the fibroblast-like cells in group III represent myocardial stromal cells that have migrated into the epicardial multi-cell layer. The notion of a compartmentalization of the epicardial multi-cell layer is supported by the differential expression of hypoxia/HIF-1-responsive glycolytic genes among EpiSC populations, which is suggestive of an exposition to different local oxygen levels.

In summary, our study explored post-MI epicardial cell heterogeneity in the context of all cardiac stromal cells at an unprecedented cellular resolution. Important epicardial properties in post-MI wound healing/regeneration can now be attributed to specific cell populations. A deeper understanding of adult epicardial hierarchy may help decipher the signaling mechanisms by which individual epicardial cell populations interact to specifically stimulate cardiac repair processes originating in the epicardium.

Materials and methods

Key resources table
Reagent type
(species) or resource
DesignationSource or referenceIdentifiersAdditional information
Strain, strain background (Mus musculus)C57BL/6JJanvier LabsStrain name: C57BL/6JRj
Strain, strain background (M. musculus)STOCK Wt1tm2(cre/ERT2)Wtp/JThe Jackson Laboratory;
PMID:18568026
Stock no. 010912; RRID:IMSR_JAX:010912
Strain, strain background (M. musculus)B6;129S6-Gt(ROSA)26Sortm14(CAG-tdTomato)Hze/JThe Jackson LaboratoryStock no. 007908; RRID:IMSR_JAX:007908
Chemical compound, drugTamoxifenSigma-AldrichT5648
Chemical compound, drug7AADBioLegend420404(1:100)
Chemical compound, drugFixable Viability Dye eFluor 780eBioscience65–0865(1:100)
OtherCollagenase CLS IIBiochromC2-22Enzyme blend; now available from Sigma-Aldrich, C2-BIOC
AntibodyAPC-conjugated rat monoclonal anti-CD31BioLegend102410, clone 390; RRID:AB_312905(1:100)
AntibodyPE-Cy7 rat monoclonal anti-CD45BD Biosciences552848, clone 30-F11; RRID:AB_394489(1:400)
Commercial assay or kitRNAscope 2.5 HD Reagent Kit-REDAdvanced Cell Diagnostics322350
Sequence-based reagentRNAscope Probe Mm-MslnAdvanced Cell Diagnostics443241
Sequence-based reagentRNAscope Probe Mm-Wt1Advanced Cell Diagnostics432711
Sequence-based reagentRNAscope Probe Mm-Cd44Advanced Cell Diagnostics479191
Sequence-based reagentRNAscope Probe Mm-Ifit3Advanced Cell Diagnostics508251
Sequence-based reagentRNAscope Probe Mm-Sfrp2Advanced Cell Diagnostics400381
Sequence-based reagentRNAscope Probe Mm-Pcsk6Advanced Cell Diagnostics593361
Sequence-based reagentRNAscope Probe Mm-Top2aAdvanced Cell Diagnostics491221
Sequence-based reagentRNAscope Probe Mm-Dkk3Advanced Cell Diagnostics400931
Software, algorithmCellRanger10x Genomics
Software, algorithmSeuratPMID:29608179RRID:SCR_007322
Software, algorithmDoubletFinderPMID:30954475RRID:SCR_018771
Software, algorithmScillushttps://github.com/xmc811/
Scillus Xu, 2021
Software, algorithmGOrillaPMID:19192299RRID:SCR_006848
Software, algorithmVelocyto.pyPMID:30089906RRID:SCR_018167
Software, algorithmCellPhoneDBPMID:32103204RRID:SCR_017054
Software, algorithmSCENICPMID:28991892RRID:SCR_017247

Mice

All animal experiments were performed in accordance with the institutional guidelines on animal care and approved by the Animal Experimental Committee of the local government ‘Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein-Westfalen’ (reference number 81–02.04.2019.A181). The animal procedures conformed to the guidelines from Directive 2010/63/EU of the European Parliament on the protection of animals used for scientific purposes.

For this study, male C57BL/6J (Janvier Labs, Le Genest-Saint-Isle, France) and male tamoxifen-inducible Wt1-targeted (Wt1CreERT2RosatdTomato) mice were used. Inducible Wt1CreERT2RosatdTomato mice were generated by crossing the STOCK Wt1tm2(cre/ERT2)Wtp/J strain (Zhou et al., 2008) (stock no. 010912; The Jackson Laboratory, Bar Harbor, USA) with the B6;129S6-Gt(ROSA)26Sortm14(CAG-tdTomato)Hze/J strain (stock no. 007908; The Jackson Laboratory, Bar Harbor, USA) followed by genotyping. Mice (body weight, 20–25 g; age, 8–12 weeks) used in this study were housed at the central animal facility of the Heinrich-Heine-Universität Düsseldorf (ZETT, Düsseldorf, Germany), fed with a standard chow diet, and provided tap water ad libitum.

Animal procedures

Request a detailed protocol

MI followed by reperfusion was performed as previously described (Bönner et al., 2012). In brief, mice were anesthetized (isoflurane 1.5%) and the left anterior descending coronary artery (LAD) was ligated for 50 min followed by reperfusion. LAD occlusion was controlled by ST-segment elevation in electrocardiography recordings. Sham control animals underwent the surgical procedure without LAD ligation.

Wt1CreERT2-mediated lineage tracing was performed as described previously (Smart et al., 2011). In brief, Wt1CreERT2RosatdTomato mice received tamoxifen injections (2 mg emulsified in sesame oil; i.p.) 5 and 3 days prior to MI to induce CreERT2 activity.

Isolation of EpiSC, aCSC, and CSC

Request a detailed protocol

EpiSC and aCSC at day 5 after MI and control CSC from uninjured hearts at day 5 after sham surgery were isolated as previously described (Owenier et al., 2020). In brief, mice were sacrificed by cervical dislocation and hearts were excised for preparation of the aortic trunk in ice-cold phosphate-buffered saline (PBS). Isolated hearts were immediately cannulated and perfused with PBS (3 min), followed by perfusion with collagenase solution (8 min; 1200 U/ml collagenase CLS II (Biochrom, Berlin, Germany)) in PBS at 37°C.

EpiSC were simultaneously isolated from post-MI hearts by bathing the heart in its collagenase-containing coronary effluat while applying mild shear forces to the cardiac surface. The effluat was collected, centrifuged (300 g, 7 min), and cells were resuspended in PBS/2% fetal calf serum (FCS)/1 mM ethylenediaminetetraacetic acid (EDTA). Application of shear force and effluat collection were omitted for uninjured sham control hearts due to the absence of an activated, expanded epicardial layer. aCSC and CSC were isolated by mechanical dissociation of the digested myocardial tissue, followed by resuspension in Dulbecco’s modified eagle medium (DMEM)/10% FCS. The cell suspensions were meshed through a 100 μm cell strainer and centrifuged at 55 ×g to separate cardiomyocytes from non-cardiomyocytes. The supernatants were again passed through a 40 µm cell strainer, centrifuged (7 min, 300 ×g), and cell pellets resuspended in PBS/2% FCS/1 mM EDTA. Cells were immediately stained for surface markers and applied to fluorescence-activated cell sorting.

Fluorescence-activated cell sorting

Request a detailed protocol

Cell fractions (EpiSC, aCSC, CSC) were isolated as described above and stained at 4°C (15 min) with fluorochrome-conjugated antibodies against surface markers of endothelial cells (APC-conjugated rat monoclonal anti-CD31 [BioLegend, San Diego, USA]) and myeloid cells (PE-Cy7-conjugated rat monoclonal anti-CD45 [BD Biosciences, Franklin Lakes, USA]) in the presence of 7AAD (viability marker [BioLegend]). Sorting was performed with a MoFlo XDP flow cytometer (Beckman Coulter, Brea, USA), where dead cells (7AAD+) and CD31+/CD45+ cells were excluded by gating on 7AAD-/CD31-/CD45- cells. Sorting of cells isolated from the Wt1CreERT2RosatdTomato mouse line followed the same gating strategy with one minor change. Fixable Viability Dye eFluor 780 (eBioscience, San Diego, USA) was used instead of 7AAD to avoid fluorescence spill-over.

scRNAseq

Request a detailed protocol

The sorted single-cell suspensions were directly used for the scRNAseq experiments. scRNAseq analysis was performed by using the 10x Genomics Chromium System (10x Genomics, San Francisco, USA). Cell viability and cell number analysis were performed via trypan blue staining in a Neubauer counting chamber. A total of 2000–20,000 cells, depending on cell availability, were used as input for the single-cell droplet library generation on the 10x Chromium Controller system utilizing the Chromium Single Cell 3’ Reagent Kit v2 according to manufacturer’s instructions. tdTomato lineage tracing experiments were conducted utilizing the Chromium Single Cell 3’ Reagent Kit v3 according to manufacturer’s instructions. Sequencing was carried out on a HiSeq 3000 system (Illumina, San Diego, USA) according to manufacturer’s instructions with a mean sequencing depth of ~90,000 reads/cell for EpiSC and ~70,000 reads/cell for aCSC. Differences in sequencing depth were necessary in order to achieve a similar sequencing saturation between all samples of ~70%.

Processing of scRNAseq data

Request a detailed protocol

Raw sequencing data were processed using the 10x Genomics CellRanger software (v3.0.2) provided by 10x Genomics. Raw BCL-files were demultiplexed and processed to Fastq-files using the CellRanger mkfastq pipeline. Alignment of reads to the mm10 genome and unique molecular identifier (UMI) counting were performed via the CellRanger count pipeline to generate a gene-barcode matrix.

The median of detected genes per cell was 3155 for EpiSC, 3265 for aCSC, and 2241 for CSC. The median of UMI counts per cell was 10,689 for EpiSC, 11,110 for aCSC, and 5879 for CSC. Mapping rates (reads mapped to the genome) were about 89% for EpiSC, 90.9% for aCSC, and 86.7% for CSC.

For tdTomato lineage tracing experiments, a custom reference, consisting of the mm10 genome and the full-length sequence of tdTomato, was generated via CellRanger mkref.

Filtering and clustering of scRNAseq data

Request a detailed protocol

Further analyses were carried out with the Seurat v3.0 R package (Butler et al., 2018). Initial quality control consisted of removal of cells with fewer than 200 detected genes as well as removal of genes expressed in less than three cells. Furthermore, cells with a disproportionately high mapping rate to the mitochondrial genome (mitochondrial read percentages > 5.0 for EpiSC and aCSC, >7.5 for CSC) have been removed, as they represent dead or damaged cells. Normalization has been carried out utilizing SCTransform. Biological replicates have been integrated into one data set by identifying pairwise anchors between data sets and using the anchors to harmonize the data sets. Dimensional reduction of the data set was achieved by Principal Component Analysis (PCA) based on identified variable genes and subsequent UMAP embedding. The number of meaningful Principal Components (PCs) was selected by ranking them according to the percentage of variance explained by each PC, plotting them in an ‘Elbow Plot’ and manually determining the number of PCs that represent the majority of variance in the data set. Cells were clustered using the graph-based clustering approach implemented in Seurat v3.0. Doublet identification was achieved by using the tool DoubletFinder (v2.0.2) (McGinnis et al., 2019) by the generation of artificial doublets, using the PC distance to find each cell's proportion of artificial k nearest neighbors (pANN) and ranking them according to the expected number of doublets. Heat maps were generated using Morpheus (https://software.broadinstitute.org/morpheus).

Gene ontology enrichment analysis

Request a detailed protocol

Over-represented GO terms in differentially expressed genes between populations or among population groups were identified by using the Seurat wrapper Scillus (v0.5.0, https://github.com/xmc811/Scillus; Xu, 2021) and the gene ontology enrichment analysis and visualization (GOrilla) tool (Eden et al., 2009), respectively.

RNA velocity analysis

Request a detailed protocol

The Python software velocyto.py (version 0.17) (La Manno et al., 2018) was run on the EpiSC count matrices and BAM files generated by CellRanger (see above) to predict and visualize future cell states based on the ratio of unspliced and spliced mRNA read counts. For visualization, cell clusters, PCA, and UMAP data were imported by the Seurat analysis (see above).

Cell cycle scoring

Request a detailed protocol

Scores for expression of cell cycle genes in EpiSC were assigned and classification of each cell in G1, G2/M, and S phase was predicted using the CellCycleScoring function implemented in Seurat.

Cell–cell communication analysis

Request a detailed protocol

Cell–cell communication mediated by ligand–receptor complexes between EpiSC populations was analyzed using the tool CellPhoneDB v.2.0 (19) after mapping mouse genes to human orthologs.

Canonical correlation analysis space alignment

Request a detailed protocol

A direct comparison of the EpiSC, aCSC, and CSC data sets was performed by Seurat’s CCA alignment procedure (v2.3.4). Briefly, the top 600 variable genes were identified for each data set and subjected to a CCA. Herein, canonical correlation vectors were identified and aligned across data sets with dynamic time warping. After alignment, a single integrated clustering was performed, allowing for comparative analysis of cell populations across both cell fractions.

Gene regulatory network analysis

Request a detailed protocol

Gene regulatory network reconstruction and cell-state identification in EpiSC, aCSC, and CSC data sets were performed using SCENIC (Aibar et al., 2017).

RNA in situ hybridization

Request a detailed protocol

In situ detection of selected marker gene expression was performed by RNA in situ hybridization using the RNAscope 2.5 HD Reagent Kit-RED (Advanced Cell Diagnostics, Hayward, USA) (Wang et al., 2012) with RNAscope Probes targeting Msln, Wt1, Cd44, Ifit3, Sfrp2, Pcsk6, Top2a, and Dkk3 mRNA sequences. Fresh frozen hearts (5 days after MI) were cut in 10 μm sections. Fixation and pretreatment of the cryosections were performed according to the manufacturer's instructions. The incubation time for the hybridization of RNAscope 2.5 AMP 5-RED was increased to 120 min to enhance the signal. For evaluation of the probe signal, the hybridized heart sections were examined under a standard bright-field microscope (BX61 Olympus, Hamburg, Germany) using a x20 objective. Images were processed for publication using ImageJ/Fiji (Schindelin et al., 2012).

Statistics

Markers defining each cluster, as well as differential gene expression between different clusters, were calculated using a two-sided Wilcoxon Rank-Sum test that is implemented in Seurat.

Data availability

ScRNAseq data have been deposited in the ArrayExpress database at EMBL-EBI (http://www.ebi.ac.uk/arrayexpress) under accession number E-MTAB-10035.

The following data sets were generated
    1. Lautwein T
    2. Schrader J
    (2021) ArrayExpress
    ID E-MTAB-10035. Single-cell sequencing of epicardial cells and fibroblasts within the infarcted heart.

References

Decision letter

  1. Matthias Barton
    Senior Editor; University of Zurich, Switzerland
  2. Hina Chaudhry
    Reviewing Editor; Icahn School of Medicine at Mount Sinai, United States
  3. Hina Chaudhry
    Reviewer; Icahn School of Medicine at Mount Sinai, United States

Our editorial process produces two outputs: i) public reviews designed to be posted alongside the preprint for the benefit of readers; ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Acceptance summary:

The article by Hesse et al. defines heterogeneity of epicardial cells and fibroblasts in a murine model of cardiac injury to analyze the resulting populations through single cell RNA sequencing. Spatial confirmation of associated markers is performed using in-situ RNA hybridization and discussed in light of other investigations. There is extensive data that provides new insights into the heterogeneity of epicardial stromal and activated cardiac stromal cells post-injury.

Decision letter after peer review:

Thank you for submitting your article "Single-cell transcriptomics defines heterogeneity of epicardial cells and fibroblasts within the infarcted heart" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, including Hina Chaudhry as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by a Senior Editor.

The reviewers have discussed their reviews with one another, this letter is to help you prepare a revised submission.

Essential revisions:

This manuscript is not suitable for publication in eLife in its current form. Essential revisions are required. All 3 reviewers have commented on the organization and flow of the manuscript which needs to be addressed.

In terms of title, the species needs to be identified according to the rules of eLife, thus the title should be changed to: Single-cell transcriptomics defines heterogeneity of epicardial cells and fibroblasts within the infarcted murine heart.

Other essential revisions include:

1) Please clarify your control and test conditions as Rev 2 requests. If the purity of the collected samples is described in a previous publication as noted by Rev 1, clarify that in the current manuscript.

2) Please address concerns regarding organization of the data and clarify cluster identities in Figure 1 as requested by Reviewer 3 and show heatmaps to show top 10 genes to see better segregation of genes across different clusters.

3) Please address the discrepancy brought up by Rev 3 between Figure 2 and 4: in Figure 2, the Hif1a/Hif1a-related and glycolysis related genes seem more highly expressed in 2 groups (WT1 and chemokine expressing cells). However, the feature plots and Dotplots supporting this need to be added to the figures. Show GO analysis with enrichment for Hif1a/glycolysis related genes in clusters vs. other clusters. It is shown that Hif1a related genes are expressed in clusters 1,7,8,4,2, and 11 but their lineage tracing analysis in Figure 4 suggests that Wt1+ cells (clusters 1,7) did not convert to any other cell type. Please address this discrepancy.

4) Please show experimental proof of dividing cells in Figure 2 with the use of (preferably) H3P to indicate mitosis or BrDU for DNA synthesis.

5) Please reorganize Results section: a lot of words are spent on cataloguing the genes identified in scRNA-seq instead of explaining their biological significance. Not until Figure 4 are the EpiSCs grouped into I, II, III based on their cell cycle active states where the whole manuscript begins to come into perspective. Then in Discussion, the authors continue to refer to groups I, II, III from Figures 1-3 where the 'grouping' concept was never introduced until Figure 4. Please implement the 'group' concept starting with Figure 1 so that it flows through the manuscript.

6) Provide further validations of genes defining cell clusters by in situ hybridization. It would be very interesting to understand more about where the different populations reside in the heart. Are they mixed? Are they at specific locations?

7) It would be helpful to see a comparison to EpiSc cells from the uninjured or sham control heart. If this is technically challenging, would it be feasible to pool 5-10 hearts to obtain adequate cell number?

8) It is surprising to see that TAM did not activate WT1+ cells at the chosen administration time window. Why was this time window chosen? Have the authors tested another time window as Day 5 is not really a 'regenerative' time window? If not, can you provide written justification? If 2/3 of all epicardial cells are WT1-negative, but are likely involved in paracrine signaling, is there another identifiable marker to label the WT1- cells?

Reviewer #1 (Recommendations for the authors):

In the manuscript "Single-cell transcriptomics defines heterogeneity of epicardial cells and fibroblasts within the infarcted heart", the authors isolated epicardial stromal cells (EpiSC) and cardiac interstitial/stromal cells (termed active CSCs) from the same I/R heart and identified transcriptionally distinct subpopulation of EpiSCs via 10x genomics technology. They also performed transcriptome profile comparison between EpiSCs and aCSCs. This manuscript shows rigorous scientific investigation. Their isolation protocol is supported by their previous publication. Method section documented in detail of step-wise QC process of bioinformatics analysis. In summary, the analysis identified 11 clusters of EpiSC, some of which overlap with the well-established epicardial marker WT1 with confirmed in situ anatomical localization. When compared to aCSC, the two groups showed clear different function/states as expected. In the lineage tracing model, RNA velocity predicts cell hierarchy, cell-cell communication between populations, as well as cell cycle activity. Overall this manuscript provides a ton of information that can be helpful to the field.

After I/R injury, the first 48-72hrs is the peak of inflammation followed by rapid cardiac function deterioration to one week after infarct. The chosen cell harvest time at 5 days post injury is in the middle of the deteriorating window, too early for any repair or regeneration to begin, especially without gene/cell intervention. My concern is that the EpiSC collected at this time is too early to be regenerative and therefore the manuscript missed peak of when epicardial cells are being most regenerative. This calls for a longitudinal comparison of EpiSCs post injury (eg. 5dpi, 10dpic, 30dpi).

In result section, authors spent a lot of words on cataloging the genes identified in scRNA-seq in stead of explaining the biological significance of them. It is until Figure 4 where the EpiSCs are grouped into I, II, III, based on their cell cycle active states where the whole manuscript began to come into perspective. Then in Discussion, the authors continued with group I, II, III terms talking about Figure 1-3 where the "grouping" concept was never introduced until Figure 4. It would be helpful if the "group" concept was implemented since Figure 1 and flows through the manuscript.

Cluster 3, being the 3rd largest EpiSC cluster, is shown to be "no GO enrichment found" in Figure 1E. Are these cells viable? please provide explanation of such observations.

As EpiSC can be a source of cell repair after injury, it'll be very interesting to see direct comparison of EpiSCs between MI and sham heart. RNA profiling of EpiSC in an injured heart may hold insights of their activity, migration, and paracrine signal during cardiac repair. Though the authors mentioned that isolating EpiSC from uninjured heart is technically challenging (very few cells), would it be feasible if pooling 5-10 hearts together to increase the cell number?

It is unexpected to see TAM did not activate WT1+ cells at the chosen administration time window. Why was such a plan chosen and have the authors tested other time window for TAM/harvest?

As mentioned in Discussion, about 2/3 of all epicardial cells were Wt1-negative but are likely to be involved in paracrine signaling. Is there an identifiable marker to label the 1/3 Wt1- cells?

Reviewer #2 (Recommendations for the authors):

1. It is important to explicitly define your control and test conditions. The schematic of Figure 1A needs to be explained in the text. What is the anticipated overlap in cell populations? What is the purity of isolation across these groups? This would need to be benchmarked. Does the isolation of aCSC have EpiSC contamination?

2. I would highly consider re-working the format of the manuscript. Namely, describing the changes across test conditions – which are unique/enriched to the injured heart, and then what changes are specifically enriched in EpiSC. For example, are EpiSC populations unique to the epidermal tissue but share transcriptomic similarity (i.e overlap of cells in Figure 3) or is there anticipated purity issues in isolation?

3. Please try to tease out differences between CSC and aCSC?

4. Walking a reader through the text using a cluster-based designation is difficult to follow because there is a need to constantly refer to a figure for orientation. I would suggest re-naming these with distinct population base names (e.g. Wt1-EpiSC7, Wt1-EpiSC1).

5. There is an attempt to perform EpiSC and aCSC comparison analysis in Figure 3, however without clarity on the anticipated purity, this is hard to interpret.

Reviewer #3 (Recommendations for the authors):

The following comments should allow the reader to better interpret or understand the significance of the work:

1) The current data is mostly based on scRNAseq analysis. There is confirmation by in situ hybridization analysis of only a few genes and information on the localization of cell populations within the heart superficially described and not being discussed.

2) In Figure 1, the authors name the clusters based on 2 genes and show Dotplots for 5 genes. It might help to check for more marker genes to confirm clusters identities, and show heatmaps (as supported by Seurat) to show top 10 genes and see better segregation of genes across different clusters.

3) The scRNAseq analysis broadly finds three groups of cells with 11 total clusters. It is possible that due to technical issues the resolution to cluster the cells may have been chosen such as to get 11 clusters. The three broad groups remain mostly conserved but 11 sub-groups might change if one changes minor parameters in the analysis.

4) In Figure 2, the Hif1a/Hif1a related and glycolysis related genes seem more expressed in 2 groups (WT1, and chemokine expressing cells). However, the feature plots and Dotplots supporting the same need to be added to the figures. It would be relevant to show GO analysis with enrichment for Hif1a/ glycolysis related genes in the clusters vs other clusters. The authors show that Hif1a related genes are expressed in clusters 1,7, 8, 4,2 and 11, but their lineage tracing analysis in Figure 4 suggests that Wt1+ cells (clusters 1,7) did not convert to any of the other cell type. It would be great to discuss this apparent discrepancy.

5) Groups defined as dividing cells were not corroborated experimentally by e.g. BrDu experiments (Figure 2).

6) In the cluster analysis for aCSC (Figure 3—figure supplement 1) similar to EpiSC, the cluster calling is based on only 2 genes. However, many of the clusters show overlapping gene expression patterns (cluster 4,5, clusters 7,10). The authors try to separate cluster 4, 5 based on differential expression patterns of Sca1, Smoc2, Sparcl1, but no-evidence for the same has been provided. Feature plots or Dot plots for such genes for Clusters 4,5 and 7,10 would improve cluster definition.

7) In Figure 3, authors club EpiSC and aCSC together, while in another analysis (Figure 3—figure supplement 5) they club EpiSC and CSC together. Such an approach would normalize the data for aCSC to EpiSC and CSC to EpiSC. However, this would not normalize the data for aCSC to CSC. The way the data is currently presented affects results interpretation. A better way to analyse this data would be to club all three datasets together and then perform the normalizations. If such an approach seems difficult it might be useful to perform clubbing of aCSC and CSC and analyse the levels of the genes in EpiSC in this condition.

8) The label transfer performed in Figure 4 (from analysis done in Figure 1) could also include the labels from the sub-clustering performed on Wt1+ EpiSC cells in Figure 2—figure supplement 1. This would help to understand better the RNA velocity data and help visualize and understand the lineage tracing to the more specific sub clusters- whether they originate from the more cell cycle associated what is the origin/fate of high protein expressing cells, etc.

9) The authors performed a ligand receptor analysis after the lineage tracing analysis, where they identify some pairs, which might be expressed in different groups. It would be useful to show at the analysis level the feature plot of the ligand receptors that indeed the expression levels of these pairs differs in the different groups. There is not confirmation by e.g. immunofluorescence that these proteins are expressed at different levels in different cells.

10) The manuscript seems to jump from EpiSC to comparisons of EpiSC with aCSC and again to analysis of EpiSC. The lineage tracing part of EpiSC of the manuscript could be clubbed after the clustering analysis of the EpiSCs. This would make all the EpiSC analysis stick together and the comparisons with aCSC at the end, making the flow of the story better.

https://doi.org/10.7554/eLife.65921.sa1

Author response

Essential revisions:

This manuscript is not suitable for publication in eLife in its current form. Essential revisions are required. All 3 reviewers have commented on the organization and flow of the manuscript which needs to be addressed.

In terms of title, the species needs to be identified according to the rules of eLife, thus the title should be changed to: Single-cell transcriptomics defines heterogeneity of epicardial cells and fibroblasts within the infarcted murine heart.

The title has been changed accordingly.

[…] Reviewer #1 (Recommendations for the authors):

In the manuscript "Single-cell transcriptomics defines heterogeneity of epicardial cells and fibroblasts within the infarcted heart", the authors isolated epicardial stromal cells (EpiSC) and cardiac interstitial/stromal cells (termed active CSCs) from the same I/R heart and identified transcriptionally distinct subpopulation of EpiSCs via 10x genomics technology. They also performed transcriptome profile comparison between EpiSCs and aCSCs. This manuscript shows rigorous scientific investigation. Their isolation protocol is supported by their previous publication. Method section documented in detail of step-wise QC process of bioinformatics analysis. In summary, the analysis identified 11 clusters of EpiSC, some of which overlap with the well-established epicardial marker WT1 with confirmed in situ anatomical localization. When compared to aCSC, the two groups showed clear different function/states as expected. In the lineage tracing model, RNA velocity predicts cell hierarchy, cell-cell communication between populations, as well as cell cycle activity. Overall this manuscript provides a ton of information that can be helpful to the field.

After I/R injury, the first 48-72hrs is the peak of inflammation followed by rapid cardiac function deterioration to one week after infarct. The chosen cell harvest time at 5 days post injury is in the middle of the deteriorating window, too early for any repair or regeneration to begin, especially without gene/cell intervention. My concern is that the EpiSC collected at this time is too early to be regenerative and therefore the manuscript missed peak of when epicardial cells are being most regenerative. This calls for a longitudinal comparison of EpiSCs post injury (eg. 5dpi, 10dpic, 30dpi).

We agree that the post MI inflammation is a well-orchestrated process over time. As to the dynamics of the multi-cell layer of EpiSC we know that it is formed within the first several days after I/R and does not further increase in thickness after 4 days post MI in the murine heart (1). Furthermore, the re-expression of fetal epicardial genes as measure of epicardial activation peaks 1 and 5 days after MI and declines afterwards (2). Because of these reasons, we chose day 5 post I/R for the scRNAseq analysis of EpiSC. The rational for selecting day 5 post MI is now better explained in the revised manuscript (page 4 line 87-89). We certainly agree that it also would be interesting to analyze the composition of EpiSC populations at later time points in future studies.

In result section, authors spent a lot of words on cataloging the genes identified in scRNA-seq in stead of explaining the biological significance of them. It is until Figure 4 where the EpiSCs are grouped into I, II, III, based on their cell cycle active states where the whole manuscript began to come into perspective. Then in Discussion, the authors continued with group I, II, III terms talking about Figure 1-3 where the "grouping" concept was never introduced until Figure 4. It would be helpful if the "group" concept was implemented since Figure 1 and flows through the manuscript.

Thank you for this very helpful suggestion. We re-structured the manuscript, the grouping concept is now introduced in Figure 1.

Cluster 3, being the 3rd largest EpiSC cluster, is shown to be "no GO enrichment found" in Figure 1E. Are these cells viable? please provide explanation of such observations.

EpiSC-3 has a distinct gene expression pattern with multiple differentially expressed genes and cells are certainly viable. EpiSC-3 marker genes show enrichment of GO terms for Molecular Function and Cellular Component, however, no GO term for Biological Process was found (which was displayed in Figure 1E of the original manuscript). This finding seems to be consistent with the observation of a reduced number of expressed genes per cell in EpiSC-3 (Figure 1G), suggesting a rather quiescent cell state. Triggered by your question we have removed the presentation of manually selected Biological Process terms in original Figure 1E and replaced this panel by an unbiased visualization of enriched GO terms using the Seurat wrapper Scillus. This latter information is now in Figure 1—figure supplement 2 of the revised manuscript.

As EpiSC can be a source of cell repair after injury, it'll be very interesting to see direct comparison of EpiSCs between MI and sham heart. RNA profiling of EpiSC in an injured heart may hold insights of their activity, migration, and paracrine signal during cardiac repair. Though the authors mentioned that isolating EpiSC from uninjured heart is technically challenging (very few cells), would it be feasible if pooling 5-10 hearts together to increase the cell number?

Since the epicardium of healthy hearts is a monolayer, the isolation of these cells by enzymatic procedures is accompanied by a significant proportion of non-epicardial cells. We think that this contamination is the most critical issue rather than the low number of cells. Unfortunately, this problem cannot be overcome by pooling heart samples. However, in our CSC data set from uninjured hearts, a small yet defined population of epithelial/epicardial cells are visible as CSC-12, likely representing the epicardial monolayer. In CCA space alignment CSC-12 clustered together with cells of post-MI EpiSC-7 in CCA cluster A (Figure 4A and B). The direct comparison of CSC-12 and EpiSC-7 within CCA cluster A revealed differential expression of proliferation-associated and hypoxia-responsive genes pre and post MI and is now included in the manuscript (Figure 4—figure supplement 7D with source data).

It is unexpected to see TAM did not activate WT1+ cells at the chosen administration time window. Why was such a plan chosen and have the authors tested other time window for TAM/harvest?

Tamoxifen injection 5 and 3 days prior MI did induce expression of tdTomato in Wt1-expressing cells as expected. The question which we intended to answer was, whether Wt1-expressing cells may contribute to the Wt1-negative populations of the thickened epicardium that is formed during the acute injury response. We have better clarified this in the text (Results section “Spatial mRNA expression of EpiSC population identifiers”).

As mentioned in Discussion, about 2/3 of all epicardial cells were Wt1-negative but are likely to be involved in paracrine signaling. Is there an identifiable marker to label the 1/3 Wt1- cells?

The 2/3 of Wt1-negative EpiSC display variable degrees of similarity to myocardial aCSC. While we observed distinct epicardial/myocardial signatures in relative gene expression (Figure 4 C-H), we did not identify any marker that clearly distinguishes Wt1-negative EpiSC and aCSC. Since the spatial distribution of Wt1-negative EpiSC in the epicardial post-MI multi-cell layer was verified by in situ hybridization, this most likely indicates that they are derived from cells already present in the epicardial monolayer of the uninjured heart, or might have migrated from the myocardium into the epicardial layer after injury.

Reviewer #2 (Recommendations for the authors):

1. It is important to explicitly define your control and test conditions. The schematic of Figure 1A needs to be explained in the text. What is the anticipated overlap in cell populations? What is the purity of isolation across these groups? This would need to be benchmarked. Does the isolation of aCSC have EpiSC contamination?

A detailed protocol and comprehensive validation data of our cell isolation procedure for EpiSC, aCSC and CSC have been previously published (4), including analyses of cell yield, viability, and purity. We have clarified this in the text and extended the explanation of the schematic in Figure 1A (Results section “ScRNAseq of post-MI stromal cells”).

The aCSC fraction contains a small fraction of EpiSC, as was observed by positive immunostaining of WT1 protein in a small quantity of short-term cultured aCSC before (4). In the present study, aCSC‑6 was identified as residual EpiSC according to high gene expression of WT1+ together with other epicardial/epithelial genes (Figure 4—figure supplement 1D). As is now better explained in the Results section “Post-MI aCSC in comparison to EpiSC”, the EpiSC layer is likely not completely removed from the myocardium during the cell isolation procedure. However, changing the protocol to extended digestion times would amplify the probability of non-epicardial cells in the EpiSC fraction.

In the revised manuscript we have rearranged data presentation – as requested by the other reviewers – and have carefully revised the text for clarity to avoid misunderstandings.

2. I would highly consider re-working the format of the manuscript. Namely, describing the changes across test conditions – which are unique/enriched to the injured heart, and then what changes are specifically enriched in EpiSC. For example, are EpiSC populations unique to the epidermal tissue but share transcriptomic similarity (i.e overlap of cells in Figure 3) or is there anticipated purity issues in isolation?

The epicardium of adult healthy hearts is a monolayer of epithelial cells. Myocardial ischemia is a prerequisite for generation of stromal cells in the epicardium, forming the EpiSC multilayer. As mentioned above in the response to reviewer 1, we identified epithelial cells of the epicardial monolayer in CSC-12 of our CSC data set from the uninjured heart. There was no other CSC population with expression of epicardial/epithelial markers (Figure 4—figure supplement 4D and E). In the CCA space alignment of cell fraction data sets, CSC-12 overlapped with the epithelial cells of the post-MI heart (EpiSC-7) in CCA cluster A. We have included a comparative analysis of CSC-12 and EpiSC-7 revealing the MI-induced transcriptional changes in these epithelial cells (Figure 4—figure supplement 7D with source data).

3. Please try to tease out differences between CSC and aCSC?

Since there are already elaborate scRNAseq studies specifically addressing the activation of CSC after MI published (5,6). We therefore refrained from comparing CSC and aCSC again and instead focused on the novel scRNAseq data of the post-MI epicardium.

4. Walking a reader through the text using a cluster-based designation is difficult to follow because there is a need to constantly refer to a figure for orientation. I would suggest re-naming these with distinct population base names (e.g. Wt1-EpiSC7, Wt1-EpiSC1).

Thank you for this very helpful comment. We have re-structured and revised the manuscript and now use the classification of EpiSC populations in group I, II, and III throughout the Results and Discussion sections as was also suggested by reviewer 1 (2.). We are confident that this has significantly improved the general readability of our text.

5. There is an attempt to perform EpiSC and aCSC comparison analysis in Figure 3, however without clarity on the anticipated purity, this is hard to interpret.

We have made an attempt to clarify the text in the Result sections “ScRNAseq of post-MI stromal cells” and “Post-MI aCSC in comparison to EpiSC” regarding the issue of purity (see also answer to 1.).

Reviewer #3 (Recommendations for the authors):

The following comments should allow the reader to better interpret or understand the significance of the work:

1) The current data is mostly based on scRNAseq analysis. There is confirmation by in situ hybridization analysis of only a few genes and information on the localization of cell populations within the heart superficially described and not being discussed.

We have now included a text passage specifically addressing the location of EpiSC populations in the post-MI epicardium in the Discussion section.

2) In Figure 1, the authors name the clusters based on 2 genes and show Dotplots for 5 genes. It might help to check for more marker genes to confirm clusters identities, and show heatmaps (as supported by Seurat) to show top 10 genes and see better segregation of genes across different clusters.

Heat maps visualizing the expression of top 10 marker genes are now included for EpiSC (Figure 1—figure supplement 3), aCSC (Figure 4—figure supplement 3), and CSC (Figure 1—figure supplement 6).

3) The scRNAseq analysis broadly finds three groups of cells with 11 total clusters. It is possible that due to technical issues the resolution to cluster the cells may have been chosen such as to get 11 clusters. The three broad groups remain mostly conserved but 11 sub-groups might change if one changes minor parameters in the analysis.

You are absolutely correct that changing the resolution parameter during cluster identification will lead to a different number of detected clusters. We have discussed this issue and finally decided for the chosen resolution because this resulted in the most biological meaningful results. We have tested different resolution parameters during data analysis, but this either leads to over-clustering, resulting in multiple biological non-meaningful clusters, or to under-clustering, leading to significant loss of cellular heterogeneity.

4) In Figure 2, the Hif1a/Hif1a related and glycolysis related genes seem more expressed in 2 groups (WT1, and chemokine expressing cells). However, the feature plots and Dotplots supporting the same need to be added to the figures. It would be relevant to show GO analysis with enrichment for Hif1a/ glycolysis related genes in the clusters vs other clusters. The authors show that Hif1a related genes are expressed in clusters 1,7, 8, 4,2 and 11, but their lineage tracing analysis in Figure 4 suggests that Wt1+ cells (clusters 1,7) did not convert to any of the other cell type. It would be great to discuss this apparent discrepancy.

To resolve this issue, we have added supporting violin plots to the Figure (now Figure 3). GO enrichment analysis of the genes with significantly higher expression in group I and II populations vs. group III populations (Supplementary file 6) indeed yielded terms associated with the hypoxia/HIF-mediated metabolic switch towards glycolysis, such as Glycolytic Process (FDR q-value: 3.05E-02), Glucose Metabolic Process (FDR q-value: 3.22E-02), and Pyruvate Biosynthetic Process (FDR q-value: 2.66E-02). This information is now included in the Results (section “Expression of cardiomyogenesis-associated genes, HIF-1-responsive genes, and paracrine factors”).

There is no discrepancy at all in the observations that these populations share enriched expression of HIF-1-responsive genes and that the WT1- among them do not derive from the Wt1+ ones. The highly conserved HIF-1 response is due to the tight regulation of HIF-1α protein stability only transient and strongly depends on environmental conditions (acute hypoxia). Therefore, if cells are located in a microenvironment with low oxygen levels, HIF-1 will be induced and its target genes will be expressed, but the cells don’t need to be of the same lineage. We have clarified this in the Results (section “Expression of cardiomyogenesis-associated genes, HIF-1-responsive genes, and paracrine factors”) and the Discussion.

5) Groups defined as dividing cells were not corroborated experimentally by e.g. BrDu experiments (Figure 2).

Van Wijk et al. have already nicely shown the presence of proliferating cells in the epicardial post-MI multi-cell layer by BrdU incorporation (7). We have now added this information to the text (Results section “Spatial mRNA expression of EpiSC population identifiers”).

6) In the cluster analysis for aCSC (Figure 3—figure supplement 1) similar to EpiSC, the cluster calling is based on only 2 genes. However, many of the clusters show overlapping gene expression patterns (cluster 4,5, clusters 7,10). The authors try to separate cluster 4, 5 based on differential expression patterns of Sca1, Smoc2, Sparcl1, but no-evidence for the same has been provided. Feature plots or Dot plots for such genes for Clusters 4,5 and 7,10 would improve cluster definition.

As already mentioned above (2.), we have now included a heat map of aCSC top 10 marker genes (Figure 4—figure supplement 3) that confirms cluster identities. Additionally, we included feature plots of selected marker genes visualizing the segregation of aCSC-4, 5 and aCSC-7,10, respectively (Figure 4—figure supplement 1E).

7) In Figure 3, authors club EpiSC and aCSC together, while in another analysis (Figure 3—figure supplement 5) they club EpiSC and CSC together. Such an approach would normalize the data for aCSC to EpiSC and CSC to EpiSC. However, this would not normalize the data for aCSC to CSC. The way the data is currently presented affects results interpretation. A better way to analyse this data would be to club all three datasets together and then perform the normalizations. If such an approach seems difficult it might be useful to perform clubbing of aCSC and CSC and analyse the levels of the genes in EpiSC in this condition.

Thank you for making us aware of this issue. To avoid potential misinterpretation due to separate normalization, we replaced the individual CAA space alignments (EpiSC and aCSC or EpiSC and CSC) by a combined analysis of all three fractions. This information is now displayed in Figure 4.

8) The label transfer performed in Figure 4 (from analysis done in Figure 1) could also include the labels from the sub-clustering performed on Wt1+ EpiSC cells in Figure 2—figure supplement 1. This would help to understand better the RNA velocity data and help visualize and understand the lineage tracing to the more specific sub clusters- whether they originate from the more cell cycle associated what is the origin/fate of high protein expressing cells, etc.

Thank you for this helpful suggestion. We have added the label transfer of the Wt1-expressing subclusters (Figure 2—figure supplement 3).

9) The authors performed a ligand receptor analysis after the lineage tracing analysis, where they identify some pairs, which might be expressed in different groups. It would be useful to show at the analysis level the feature plot of the ligand receptors that indeed the expression levels of these pairs differs in the different groups. There is not confirmation by e.g. immunofluorescence that these proteins are expressed at different levels in different cells.

Feature plots visualizing the expression levels of selected receptor-ligand pairs across different EpiSC populations are now included as Figure 1—figure supplement 4.

10) The manuscript seems to jump from EpiSC to comparisons of EpiSC with aCSC and again to analysis of EpiSC. The lineage tracing part of EpiSC of the manuscript could be clubbed after the clustering analysis of the EpiSCs. This would make all the EpiSC analysis stick together and the comparisons with aCSC at the end, making the flow of the story better.

Very good point, thank you! We have re-structured the manuscript (as was also suggested by reviewer 1), so that now the comparison of EpiSC, aCSC, and CSC is at the end in Figure 4 and the lineage tracing results are incorporated in Figure 2.

References:

1. Ding Z, Temme S, Quast C, Friebe D, Jacoby C, Zanger K, et al. Epicardium-Derived Cells Formed After Myocardial Injury Display Phagocytic Activity Permitting in vivo Labeling and Tracking. Stem Cells Transl Med. 2016 May;5(5):639–50.

2. Zhou B, Honor LB, He H, Ma Q, Oh J-H, Butterfield C, et al. Adult mouse epicardium modulates myocardial injury by secreting paracrine factors. J Clin Invest. 2011 May 2;121(5):1894–904.

3. Kocabas F, Mahmoud AI, Sosic D, Porrello ER, Chen R, Garcia JA, et al. The Hypoxic Epicardial and Subepicardial Microenvironment. J of Cardiovasc Trans Res. 2012 Oct 1;5(5):654–65.

4. Owenier C, Hesse J, Alter C, Ding Z, Marzoq A, Petzsch P, et al. Novel technique for the simultaneous isolation of cardiac fibroblasts and epicardial stromal cells from the infarcted murine heart. Cardiovasc Res. 2020 Apr 1;116(5):1047–58.

5. Farbehi N, Patrick R, Dorison A, Xaymardan M, Janbandhu V, Wystub-Lis K, et al. Single-cell expression profiling reveals dynamic flux of cardiac stromal, vascular and immune cells in health and injury. Morrisey E, Dietz HC, editors. eLife. 2019 Mar 26;8:e43882.

6. Forte E, Skelly DA, Chen M, Daigle S, Morelli KA, Hon O, et al. Dynamic Interstitial Cell Response during Myocardial Infarction Predicts Resilience to Rupture in Genetically Diverse Mice. Cell Reports. 2020 Mar 3;30(9):3149-3163.e6.

7. Wijk B van, Gunst QD, Moorman AFM, Hoff MJB van den. Cardiac Regeneration from Activated Epicardium. PLOS ONE. 2012 Sep 20;7(9):e44692.

https://doi.org/10.7554/eLife.65921.sa2

Article and author information

Author details

  1. Julia Hesse

    Department of Molecular Cardiology, Medical Faculty and University Hospital Düsseldorf, Heinrich-Heine-University Düsseldorf, Düsseldorf, Germany
    Contribution
    Conceptualization, Investigation, Visualization, Writing - original draft, Writing - review and editing
    Contributed equally with
    Christoph Owenier
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-8555-259X
  2. Christoph Owenier

    Department of Molecular Cardiology, Medical Faculty and University Hospital Düsseldorf, Heinrich-Heine-University Düsseldorf, Düsseldorf, Germany
    Contribution
    Conceptualization, Investigation, Methodology, Writing - original draft
    Contributed equally with
    Julia Hesse
    Competing interests
    No competing interests declared
  3. Tobias Lautwein

    Biologisch-Medizinisches-Forschungszentrum (BMFZ), Genomics & Transcriptomics Laboratory, Heinrich-Heine-University Düsseldorf, Düsseldorf, Germany
    Contribution
    Data curation, Formal analysis, Investigation, Visualization, Writing - original draft
    Competing interests
    No competing interests declared
  4. Ria Zalfen

    Department of Molecular Cardiology, Medical Faculty and University Hospital Düsseldorf, Heinrich-Heine-University Düsseldorf, Düsseldorf, Germany
    Contribution
    Investigation, Visualization
    Competing interests
    No competing interests declared
  5. Jonas F Weber

    Algorithmic Bioinformatics, Heinrich-Heine-University Düsseldorf, Düsseldorf, Germany
    Contribution
    Formal analysis, Visualization
    Competing interests
    No competing interests declared
  6. Zhaoping Ding

    Department of Molecular Cardiology, Medical Faculty and University Hospital Düsseldorf, Heinrich-Heine-University Düsseldorf, Düsseldorf, Germany
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  7. Christina Alter

    Department of Molecular Cardiology, Medical Faculty and University Hospital Düsseldorf, Heinrich-Heine-University Düsseldorf, Düsseldorf, Germany
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  8. Alexander Lang

    Division of Cardiology, Pulmonology and Vascular Medicine, Medical Faculty and University Hospital Düsseldorf, Heinrich-Heine-University Düsseldorf, Düsseldorf, Germany
    Contribution
    Conceptualization, Formal analysis, Writing - original draft
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-6006-7109
  9. Maria Grandoch

    Institute of Pharmacology and Clinical Pharmacology, Medical Faculty and University Hospital Düsseldorf, Heinrich-Heine-University Düsseldorf, Düsseldorf, Germany
    Contribution
    Conceptualization, Resources, Writing - original draft
    Competing interests
    No competing interests declared
  10. Norbert Gerdes

    Division of Cardiology, Pulmonology and Vascular Medicine, Medical Faculty and University Hospital Düsseldorf, Heinrich-Heine-University Düsseldorf, Düsseldorf, Germany
    Contribution
    Conceptualization, Resources, Writing - original draft
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4546-7208
  11. Jens W Fischer

    Institute of Pharmacology and Clinical Pharmacology, Medical Faculty and University Hospital Düsseldorf, Heinrich-Heine-University Düsseldorf, Düsseldorf, Germany
    Contribution
    Conceptualization, Resources, Funding acquisition
    Competing interests
    No competing interests declared
  12. Gunnar W Klau

    Algorithmic Bioinformatics, Heinrich-Heine-University Düsseldorf, Düsseldorf, Germany
    Contribution
    Resources, Formal analysis
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-6340-0090
  13. Christoph Dieterich

    Section of Bioinformatics and Systems Cardiology, Klaus Tschira Institute for Integrative Computational Cardiology and Department of Internal Medicine III, University Hospital Heidelberg, Heidelberg, Germany
    Contribution
    Resources, Formal analysis, Visualization
    Competing interests
    No competing interests declared
  14. Karl Köhrer

    Biologisch-Medizinisches-Forschungszentrum (BMFZ), Genomics & Transcriptomics Laboratory, Heinrich-Heine-University Düsseldorf, Düsseldorf, Germany
    Contribution
    Resources, Data curation, Investigation, Writing - original draft
    Competing interests
    No competing interests declared
  15. Jürgen Schrader

    Department of Molecular Cardiology, Medical Faculty and University Hospital Düsseldorf, Heinrich-Heine-University Düsseldorf, Düsseldorf, Germany
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Writing - original draft, Writing - review and editing
    For correspondence
    schrader@uni-duesseldorf.de
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7742-2768

Funding

Deutsche Forschungsgemeinschaft (236177352)

  • Maria Grandoch
  • Norbert Gerdes
  • Jens W Fischer
  • Jürgen Schrader

Deutsche Forschungsgemeinschaft (220652768)

  • Christoph Owenier

Medizinische Fakultät, Heinrich-Heine-Universität Düsseldorf (2018-12)

  • Julia Hesse

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank K Raba (Institute of Transplantation Diagnostics and Cell Therapeutics, Heinrich-Heine-University Düsseldorf) for technical assistance with cell sorting by flow cytometry and J Schulze (Department for Genomics and Immunoregulation, Limes, Bonn) for helpful advice and discussions.

JS, JF, MG, and NG were supported by a grant of the German Research Council (DFG, SFB1116, project identifier: 236177352). CO was supported by the DFG-funded International Research Training Group 1902 (project identifier: 220652768), and this work was part of his PhD thesis. JH was supported by the Research Committee of the Medical Faculty of the Heinrich-Heine-University Düsseldorf (project identifier: 2018–12).

Ethics

Animal experimentation: All animal experiments were performed in accordance with the institutional guidelines on animal care and approved by the Animal Experimental Committee of the local government "Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein-Westfalen" (reference number 81-02.04.2019.A181). The animal procedures conformed to the guidelines from Directive 2010/63/EU of the European Parliament on the protection of animals used for scientific purposes.

Senior Editor

  1. Matthias Barton, University of Zurich, Switzerland

Reviewing Editor

  1. Hina Chaudhry, Icahn School of Medicine at Mount Sinai, United States

Reviewer

  1. Hina Chaudhry, Icahn School of Medicine at Mount Sinai, United States

Publication history

  1. Received: December 18, 2020
  2. Accepted: June 8, 2021
  3. Version of Record published: June 21, 2021 (version 1)

Copyright

© 2021, Hesse et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 406
    Page views
  • 69
    Downloads
  • 1
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Cell Biology
    2. Neuroscience
    Georgina Kontou et al.
    Research Article Updated

    The spatiotemporal distribution of mitochondria is crucial for precise ATP provision and calcium buffering required to support neuronal signaling. Fast-spiking GABAergic interneurons expressing parvalbumin (PV+) have a high mitochondrial content reflecting their large energy utilization. The importance for correct trafficking and precise mitochondrial positioning remains poorly elucidated in inhibitory neurons. Miro1 is a Ca²+-sensing adaptor protein that links mitochondria to the trafficking apparatus, for their microtubule-dependent transport along axons and dendrites, in order to meet the metabolic and Ca2+-buffering requirements of the cell. Here, we explore the role of Miro1 in PV+ interneurons and how changes in mitochondrial trafficking could alter network activity in the mouse brain. By employing live and fixed imaging, we found that the impairments in Miro1-directed trafficking in PV+ interneurons altered their mitochondrial distribution and axonal arborization, while PV+ interneuron-mediated inhibition remained intact. These changes were accompanied by an increase in the ex vivo hippocampal γ-oscillation (30–80 Hz) frequency and promoted anxiolysis. Our findings show that precise regulation of mitochondrial dynamics in PV+ interneurons is crucial for proper neuronal signaling and network synchronization.

    1. Cell Biology
    2. Neuroscience
    Kirsty J McMillan et al.
    Research Article Updated

    The endosome-associated cargo adaptor sorting nexin-27 (SNX27) is linked to various neuropathologies through sorting of integral proteins to the synaptic surface, most notably AMPA receptors. To provide a broader view of SNX27-associated pathologies, we performed proteomics in rat primary neurons to identify SNX27-dependent cargoes, and identified proteins linked to excitotoxicity, epilepsy, intellectual disabilities, and working memory deficits. Focusing on the synaptic adhesion molecule LRFN2, we established that SNX27 binds to LRFN2 and regulates its endosomal sorting. Furthermore, LRFN2 associates with AMPA receptors and knockdown of LRFN2 results in decreased surface AMPA receptor expression, reduced synaptic activity, and attenuated hippocampal long-term potentiation. Overall, our study provides an additional mechanism by which SNX27 can control AMPA receptor-mediated synaptic transmission and plasticity indirectly through the sorting of LRFN2 and offers molecular insight into the perturbed function of SNX27 and LRFN2 in a range of neurological conditions.