Transcriptomic Analysis of Adult Mouse Cardiac Stromal Cells Using Single-Cell qRT-PCR

Abstract

Fate-mapping studies have challenged the longstanding view of the adult mammalian heart as a post-mitotic organ, suggesting limited cardiomyocyte renewal. This has spurred efforts to determine whether selected cardiac stromal cells have regenerative potential; however, their contribution to cardiac regeneration has been found to be minimal compared with that of cardiomyocyte proliferation. Despite this, transplantation of some cardiac stromal cell populations has shown therapeutic potential through paracrine signalling. The identity of the paracrine-active stromal cell populations remains unclear due to overlapping characteristics with other cardiac stromal cell populations, such as fibroblasts, mesenchymal cells, and pericytes. This study sought to clarify the transcriptional identity and heterogeneity of adult mouse cardiac stromal cells by developing a cardiac collagenase–trypsin protocol and comparing it to the established method for isolating cardiosphere-derived cells (CDCs). This novel protocol resulted in a higher cell yield and shorter expansion time, and the resulting cells showed superior survival under serum starvation compared to commercially acquired cardiac fibroblasts (CFs). Single-cell qRT-PCR analysis revealed that collagenase–trypsin cells (CTs) and CDCs share similar gene expression profiles, distinct from those of CFs. Notably, CTs exhibited higher expression of Tcf21 and lower expression of Tbx5, suggesting an epicardial-derived fibroblast phenotype, whereas Tbx5 was enriched in CDCs and CFs, reflecting heterogeneity within the cardiac fibroblast compartment. This study offers insights into the complex identity of cardiac stromal cells and concludes that CTs closely resemble CDCs but can be generated more rapidly, making them a robust and efficient source of paracrine-active cardiac stromal cells.

1. Introduction

Evidence from fate-mapping studies of cardiomyocyte renewal in the adult mammalian heart has challenged the dogma that considers the heart a post-mitotic organ [1,2,3]. Therefore, multiple research groups have attempted to characterise endogenous cardiac stromal cell populations that may play a role in the regenerative process. Identifiers of the presumed cardiac progenitor cell (CPC) populations include the SP dye-efflux phenotype [4,5], SCA1 [6], KIT [7,8], ISL1 [9,10], cardiosphere- and colony-forming assays [11,12], and re-activation of the embryonic epicardial programme with WT1 upregulation [13] (reviewed in Alonaizan and Carr, 2022 [14]). More recent findings indicate that the mechanism behind cardiomyocyte renewal is cardiomyocyte proliferation, and that the contribution of the presumed CPC populations is minimal [1,2,3,14,15,16]. Despite their low cardiomyogenic potential after transplantation into the infarcted heart, these stromal cell populations have demonstrated therapeutic benefits in pre-clinical studies, primarily through paracrine signalling [6,12,13,17,18].

The identity of the paracrine-active stromal cell (PASC) populations remains unclear, with numerous studies questioning the distinctions between these cells and various other cardiac stromal cell populations, including mesenchymal cells, fibroblasts, and pericytes [19,20,21,22,23]. This uncertainty stems from the overlap in their morphology and gene expression profiles. For instance, cardiac fibroblasts, like PASCs, express core cardiac transcription factors that contribute to cardiac development and repair [24]. Both cardiac and tail fibroblasts share a molecular signature similar to that of mesenchymal stem cells (MSCs) [25]. Significant efforts have been made to distinguish between these cardiac stromal cell populations. To complicate matters further, single-cell transcriptional profiling of mouse ventricular non-myocytes has revealed subpopulation heterogeneity within the cardiac fibroblast population [26]. Additionally, markers commonly used to identify pericytes are relatively non-specific, although some genes are enriched in pericytes compared to other cell types [26].

We developed a cardiac collagenase–trypsin protocol adapted from studies by Gharaibeh et al. [27] and Okada et al. [28] for isolating slowly adhering cells (SACs) from skeletal muscle. These SACs demonstrated superior differentiation, survival, and therapeutic potential compared to rapidly adhering cells (RACs). Similarly, we showed that collagenase–trypsin cells (CTs) can be induced to differentiate into the cardiomyocyte lineage in vitro [29] using TGFβ1 stimulation [30,31]. We also showed that fatty acid supplementation triggered a metabolic switch via the PPARα pathway, as evidenced by upregulated oxidative metabolism and increased expression of cardiomyocyte markers in CTs [29]. This finding aligns with the metabolic switch from glycolysis to fatty acid oxidation observed during cardiomyocyte maturation in development [32] and in zebrafish heart regeneration [33]. In addition, CTs were found to secrete a range of growth factors, including IGF-1, IGFBP-2/3, M-CSF, and VEGF-A, and their conditioned media increased HUVEC proliferation and reduced apoptosis in HL-1 cardiomyocytes under simulated ischaemia/reperfusion. These effects were further enhanced by hypoxic conditioning of CTs through hypoxic culture and/or miRNA-210 overexpression [34]. In this study, we compare CTs with the more established cardiosphere-derived cells (CDCs) and with commercially acquired primary cardiac fibroblasts (CFs) using single-cell qRT-PCR and cell survival assays. Our aim is to determine whether CTs more closely resemble CDCs or cardiac fibroblasts, to dissect the molecular signatures of cardiac stromal populations obtained using different isolation methods, and to highlight their heterogeneity, which may reflect underlying functional heterogeneity.

2. Materials and Methods

2.1. Mice

Male C57BL/6 mice aged 8–12 weeks (Harlan, Oxon, UK) were kept under a 12 h light–dark cycle and controlled conditions of temperature and humidity, with free access to water and chow. All animal procedures were reviewed and approved by the University of Oxford Animal Welfare and Ethical Review Board and conformed to the Animals (Scientific Procedures) Act 1986, incorporating Directive 2010/63/EU of the European Parliament (PPL #3003322, approved 5 December 2017).

2.2. Isolation and Expansion of Mouse CTs and CDCs

Mice were terminally anaesthetised with isoflurane, and hearts were isolated and washed with Dulbecco’s phosphate-buffered saline (DPBS) containing 50 mg primocin (antimicrobial agent; InvivoGen, Toulouse, France). Mouse atrial appendages were dissected and mechanically minced into 1–2 mm³ pieces. Cells from a single heart were seeded into one well of a 12-well plate for each biological replicate.

For CT isolation, the tissue pieces were then transferred into a digestion mixture (0.1% trypsin and 0.1% collagenase II (Calbiochem, Merck KGaA, Darmstadt, Germany, 286 U/mg) in DPBS) and incubated in a water bath at 37 °C for a total of 40 min. Every 10 min, the digestion mixture was mechanically triturated by pipetting, left on ice for 1 min to settle, and the supernatant was collected. Fresh digestion mixture was added to the tissue pieces for a total of 4 digestions. During the final digestion step, sterile 19 G needles and 1 mL syringes were used to triturate the tissue. After each digestion, the supernatant was neutralised, and the cell suspension was resuspended in fresh complete explant medium (CEM; Iscove’s modified Dulbecco’s medium supplemented with 20% FBS, 100 U/mL penicillin, 100 μg/mL streptomycin, and 2 mM L-glutamine, ThermoFisher, Waltham, MA, USA) [12] and plated in a 12-well plate through a 40 µm cell strainer. The wells were pre-coated with 50 µL per cm² of fibronectin (2 µg/mL in DPBS; Sigma-Aldrich, St. Louis, MO, USA) and cells were allowed to attach for 3 days without medium changes.

For CDC isolation [12], the tissue pieces were digested in 0.05% trypsin-EDTA (ThermoFisher) for 4 min at 37 °C, then neutralised with CEM. Processed tissue pieces were plated on fibronectin-coated (3 μg/mL in DPBS) 6-well plates containing CEM, and explants were cultured for 25–30 days at 37 °C in 5% CO_2, with media changes every 2–3 days. When explant-derived cells (EDCs) reached 70–80% confluency, they were harvested by washing with 0.53 mM Versene (ThermoFisher) before trypsin digestion. EDCs were resuspended in Cardiosphere Growth Medium (CGM), comprising 65% Dulbecco’s modified Eagle medium (DMEM/F12), 35% IMDM, 7% foetal bovine serum (FBS; ThermoFisher), 2% B27 (ThermoFisher), 25 ng/mL cardiotrophin (Peprotech, Waltham, MA, USA), 10 ng/mL epidermal growth factor (EGF; Peprotech EC), 20 ng/mL basic fibroblast growth factor (FGF; Promega, Madison, WI, USA), and 5 units thrombin (Sigma-Aldrich), at a concentration of 2 × 10⁴ cells/mL and seeded as 25 µL droplets on an uncoated lid of a Petri dish. DPBS was added to the Petri dishes to maintain humidity throughout 3 days of culture, allowing spherical multicellular clusters, known as cardiospheres, to form. Cardiospheres were collected by elution and gentle pipetting with DPBS and plated in fibronectin-coated 12-well plates with CEM, allowing spontaneous release of CDCs.

2.3. Cell Culture

Primary CTs and CDCs were cultured in CEM and plated on fibronectin-coated flasks. HL-1 cardiomyocytes [35] were maintained in Claycomb medium (Sigma-Aldrich), supplemented with 100 U/mL penicillin, 100 μg/mL streptomycin, 2 mM L-glutamine (ThermoFisher), 100 µM norepinephrine (Sigma-Aldrich) in 30 mM L-ascorbic acid (Sigma-Aldrich), and 10% FBS and plated on flasks pre-coated with 0.02% (wt/vol) gelatin (Sigma-Aldrich) containing 5 µg/mL fibronectin. C57BL/6 adult mouse atrial primary cardiac fibroblasts, obtained from Cell Biologics (Chicago, IL, USA), were maintained in complete fibroblast medium (Cell Biologics) supplemented with the provided supplement kit: FGF, hydrocortisone, L-glutamine, Antibiotic-Antimycotic Solution, and 10% FBS, and plated on flasks pre-coated with 0.02% (wt/vol) gelatin. As CFs were purchased and not isolated in our laboratory, some information is lacking, e.g., the age of the mice, duration of cell attachment, and composition of the accompanying media (Figure 1A).

2.4. Serum Starvation

For serum starvation experiments, cells were switched to serum-free medium for 72 h. Results are presented as means ± standard deviation. Statistical analysis was performed using GraphPad Prism (version 10.3.0). Data were analysed using a two-way analysis of variance (ANOVA) with Tukey’s post hoc test. Sample sizes are provided in the figure legends. Statistical significance was defined as p < 0.05.

2.5. Single-Cell qRT-PCR

Single cells were sorted directly (FACSAria II) into 96-well plates containing the reaction mixture for pre-amplification using CellsDirect One-Step qRT–PCR Kits (ThermoFisher). Pre-amplification was performed in a Veriti Thermal Cycler (Applied Biosystems, ThermoFisher) for 22 cycles. As negative controls, at least 3–5 non-template samples were included in each run at the pre-amplification stage. Quantitative amplification was performed using Dynamic Array chips (Fluidigm, Eurofins, Ebersberg, Germany) for 48 assays × 48 samples, the BioMark HD system (Fluidigm), and TaqMan probes (ThermoFisher) (

Table 1. Taqman probes used for single-cell qRT-PCR.

Gene	Reference	Gene	Reference
Mesp1	Mm00801883_g1	Nkx2-5	Mm00657783_m1
Mef2c	Mm01340842_m1	Myh11	Mm00443013_m1
Mef2a	Mm01318990_m1	Cnn1	Mm00487032_m1
Hand1	Mm00433931_m1	αSMA (Acta2)	Mm01546133_m1
Islet1 (Isl1)	Mm00627860_m1	Vim	Mm01333430_m1
Tbx20	Mm00451515_m1	Cd44	Mm01277163_m1
Gata4	Mm00484689_m1	Pdgfrb	Mm00435546_m1
Hand2	Mm00439247_m1	Ng2	Mm01283063_m1
Tbx5	Mm00803518_m1	Ddr2	Mm00445615_m1
Ly6a (Sca1)	Mm00726565_s1	Medag	Mm00551008_m1
Wt1	Mm00460570_m1	Ms4a4d	Mm00656404_m1
Klf4	Mm00516104_m1	Col1a1	Mm00801666_g1
Pou5f1 (Oct4)	Mm03053917_g1	Prg4	Mm01284582_m1
Nanog	Mm02019550_s1	Wif1	Mm00442355_m1
Tert	Mm0352136_m1	Colec11	Mm01289834_m1
Kdr	Mm00440111_m1	Kcnj8	Mm00434620_m1
Cdh5	Mm00486938_m1	Mrc1	Mm01329362_m1
Vwf	Mm00550376_m1	Cx3cr1	Mm00438354_m1
Pdgfra	Mm01211685_m1	Csf1 r	Mm01266652_m1
Tcf21	Mm00448961_m1	Kcna1	Mm00439977_s1
Myl2	Mm00440384_m1	Plp1	Mm01297210_m1
Myh6	Mm00440359_m1	Ptprc (Cd45)	Mm01293577_m1
Nppa	Mm01255748_g1	Ubc	Mm01201237_m1
Hmbs	Mm01143545_m1

), according to the manufacturer’s instructions. Each sample was normalised to ΔCt using Ubc expression. As described in Noseda et al. 2015 [6], data were plotted as colour-coded heatmaps of inverted ΔCt values (blue indicates low or absent expression; red indicates high expression). Samples were ordered by cell type, and genes were grouped using a hierarchical clustering algorithm based on the underlying co-expression pattern. Differences between samples were investigated using PCA. PCA applies multiple linear transformations (singular value decomposition) to the expression profiles (standardised ΔCt values) of individual samples and identifies a series of PCs that elucidate the most distinguishing features among the samples. The linear projections (PC scores) attempt to maximise variation among the samples, whereas the coefficients of those projections (PC loadings) measure the importance of genes in defining the underlying variability associated with each component. Due to the limited number of biological replicates, nonparametric statistical analyses were applied with consideration of multiple testing. To identify differential gene expression across the four cell populations, the Kruskal–Wallis test was applied to the expression data. The resulting p-values were adjusted for multiple comparisons using Dunn’s correction. Spearman correlation coefficients were calculated for all gene pairs using –ΔCt values from the complete single-cell qRT-PCR dataset, including HL-1 cardiomyocytes and stromal populations. The correlations were visualised as a dot plot, in which dot size and colour represent the strength and direction of co-expression, respectively.

3. Results

3.1. The CT Protocol Produces Higher Cell Yield than CDCs but a Similar Survival Potential

CTs and CDCs were isolated from adult mouse atrial tissue and expanded for further analysis. The average time to reach 90% confluence at passage 0 and passage 1 was considerably shorter for CTs, taking approximately 10 and 20 days, respectively. In contrast, CDCs required around 40 days to reach passage 0 and 60 days to reach passage 1 (Figure 1B) due to the additional steps involved in explant culture and cardiosphere formation (Figure 1A).

After expansion to passage 3, we assessed the survival potential of CTs compared to CDCs and commercially acquired CFs to determine their robustness for transplantation. Serum in cell culture media provides amino acids and fatty acids, and its withdrawal is therefore considered a form of nutrient restriction and an effective means of inducing cell apoptosis [34,36]. Therefore, serum starvation was used to challenge the cells in culture for 3 and 10 days. CTs, CDCs, and CFs continued to grow following serum starvation, with no significant difference in cell numbers detected at day 3. At day 10, CT cell numbers were significantly higher than those of CFs but not CDCs, suggesting that CTs have a superior survival potential compared to CFs (Figure 1C).

3.2. Single-Cell Profiling of Cardiac Stromal Cells

CTs, CDCs, and CFs isolated from adult mouse atrial tissue were expanded to passage 3 for single-cell qRT-PCR analysis and compared with the HL-1 cardiomyocyte line, which was used as a control (Figure 2). The genes selected for this experiment (detailed in Appendix A,

Table A1. Description of the markers used in the single-cell qRT-PCR characterisation.

Gene	Full Name of Encoded Protein	Type	Relevant Expression Pattern	Relevant Function	References
Cd44	Cluster of differentiation 44	Transmembrane glycoprotein	Fibroblasts, macrophages, endothelial cells	Cellular adhesion; angiogenesis; cytokine release; regulates inflammation and fibrosis	[75,76,77,78]
Cdh5	Cadherin 5 or VE-cadherin	Member of the cadherin family of calcium-dependent glycoproteins	Endothelial cells	Calcium-dependent cell adhesion; cell proliferation and migration	[75,79]
Col1a1	Collagen, type I, alpha 1	Member of group I collagen	Fibroblasts, smooth muscle cells	Collagen type I synthesis	[75,80]
Colec11	Collectin subfamily member 11	Member of the collectin family of C-type lectins	Pericytes, endothelial cells	Collectins initiate the lectin complement pathway, which is integral to the innate immune response; embryogenesis	[26,75,81,82]
Csf1r	Colony-stimulating factor 1 receptor	Receptor tyrosine kinase	Macrophages (used as an M2 marker)	Growth and differentiation of macrophages; maintenance of the intestinal stem cell niche	[70,83]
Cx3cr1	CX3C chemokine receptor 1	Member of the 7-transmembrane G protein-coupled receptor family	Macrophages	Macrophage survival and function	[75,84,85]
Ddr2	Discoidin domain receptor tyrosine kinase 2 or CD167b	Receptor tyrosine kinase	Fibroblasts, smooth muscle cells, pericytes	Activated by collagen binding; cell proliferation; adhesion and migration; ECM degradation	[75,86,87]
Gata4	GATA binding factor 4	Transcription factor of the zinc-finger GATA family	Cardiomyocytes, cardiac fibroblasts	Cardiogenesis	[24,25,88,89,90]
Hand1	Heart- and neural crest derivatives expressed protein 1	Transcription factor of the basic helix-loop-helix family	FHF; reported to be undetectable in adult mouse hearts	Cardiogenesis; epicardial development and coronary vessel formation	[91,92,93,94]
Hand2	Heart- and neural crest derivatives expressed protein 2	Transcription factor of the basic helix-loop-helix family	SHF, cardiomyocytes, endothelial cells, fibroblasts	Cardiogenesis; epicardial development and coronary vessel formation	[6,25,91,92,94]
Isl1	Insulin gene enhancer protein 1	Transcription factor of the LIM/homeodomain family	CPCs, cardiac neural crest, endothelial cells	Cardiogenesis	[9,10,46,47,95]
Kcna1	Potassium voltage-gated channel subfamily A member 1	Member of the 6-TM family	Schwann cells	Regulates membrane excitability	[26,75,96]
Kcnj8	Potassium inwardly rectifying channel, subfamily J, member 8	The inward-rectifier potassium channel family (also known as 2-TM channels)	Pericytes	Regulates vascular tone and cardiac adaptive response to systemic metabolic stress	[26,75,97,98,99]
Kdr	Kinase insert domain receptor or foetal liver kinase 1 (FLK1); VEGFR2; CD309	Receptor tyrosine kinase	Endothelial cells, Kdr+ developmental cardiovascular progenitor	Transduction of VEGF-A effects	[48,49,75,100]
Klf4	Krüppel-like factor 4	Member of the Krüppel-like family of transcription factors	Stem cells, endothelial cells, fibroblasts	Pluripotency; angiogenesis; embryonic vasculogenesis; myofibroblast differentiation	[63,75,101,102,103]
Medag	Mesenteric oestrogen-dependent adipogenesis protein	Member of the sex-steroid-regulated mesenteric oestrogen-dependent adipogenesis proteins	Cardiac fibroblasts “1”	Adipogenesis; glucose uptake; interacts with the fatty acid transporter CD36, which is involved in the uptake of apoptotic material	[26,104,105]
Mef2a	Myocyte-specific enhancer factor 2A	Transcription factor of the MEF2 family	Cardiomyocytes, cardiac fibroblasts, endothelial cells, smooth muscle cells, macrophages, and pericytes	Cardiogenesis; vasculogenesis	[6,75,106,107]
Mef2c	Myocyte-specific enhancer factor 2C or MADS box transcription enhancer factor 2	Transcription factor of the MEF2 family	Cardiomyocytes, cardiac fibroblasts, endothelial cells, smooth muscle cells, macrophages, and pericytes	Cardiogenesis; vasculogenesis	[6,24,25,75,90,108,109]
Mesp1	Mesoderm Posterior Transcription Factor 1	Transcription factor of the basic helix-loop-helix family	Mesoderm	Development of cardiac mesoderm	[110]
Mrc1	Mannose receptor C-type 1 or CD206	Transmembrane glycoprotein	Macrophages (used as an M2 marker)	Mediates the endocytic functions of macrophages and the DNA replication checkpoint	[26,75,111,112,113]
Ms4a4d	Membrane-spanning 4-domains subfamily A member 4D	Member of the membrane-spanning 4A (MS4A) family	Cardiac fibroblasts “1”, reported to be expressed in M2 macrophages	May play a role as an adaptor protein to facilitate intracellular protein–protein interactions. Its functionality is still not fully understood	[26,114,115]
Myh6	Myosin heavy chain 6 or αMHC	Hexameric ATPase cellular motor protein	Predominantly atrial cardiomyocytes	Cardiogenesis; force generation and muscle contraction	[116]
Myl2	Myosin light chain 2	Hexameric ATPase cellular motor protein	Predominantly ventricular cardiomyocytes	Cardiogenesis; force generation and muscle contraction	[116]
Nanog	Nanog Homeobox	Transcription factor of the Nanog homeobox family	Stem cells	Pluripotency; tumourigenicity	[64,65,117]
Ng2	Neuron-glial antigen 2 or chondroitin sulphate proteoglycan 4	Transmembrane proteoglycan	Pericytes, smooth muscle cells	Co-receptor in conjunction with PDGFRα; angiogenesis; cell survival and migration	[75,118,119,120]
Nkx2-5	NK2 homeobox 5	Transcription factor of the NK-2 homeobox family	Cardiomyocytes	Cardiogenesis	[75,121]
Nppa	Natriuretic Peptide A	The natriuretic peptide family	Predominantly atrial cardiomyocytes	Cardiac hormone involved in regulating blood volume and pressure and cardiovascular homeostasis	[122]
Pou5f1	Octamer-binding transcription factor 4 (Oct4) or POU class 5 homeobox 1	Transcription factor of the homeodomain POU (Pit-Oct-Unc) family.	Stem cells	Pluripotency; tumourigenicity	[123,124]
Pdgfra	Platelet-derived growth factor receptor α	Receptor tyrosine kinase	Fibroblasts, defines the SCA1+ SP+ population (CPCs)	Development of mesenchymal cell types; cardiogenesis	[6,75,125,126,127]
Pdgfrb	Platelet-derived growth factor receptor β	Receptor tyrosine kinase	Pericytes, fibroblasts, endothelial cells, smooth muscle cells	Development of mesenchymal cell types; regulation of autophagic processes; cell proliferation and survival; adult mammalian cardiomyocyte proliferation; cardiogenesis	[75,127,128,129,130,131,132,133]
Plp1	Proteolipid protein 1	Myelin proteolipid protein	Schwann cells, endothelial cells, smooth muscle cells, pericytes and macrophages	Myelination	[134]
Prg4	Proteoglycan 4	Mucinous glycoprotein	Cardiac fibroblasts “2”	Ligand for CD44 receptor, ECM constituent conferring compression resistance	[26,135,136]
Ptprc (Cd45)	Protein tyrosine phosphatase, receptor type, C	Transmembrane glycoprotein	Haematopoietic cells (except erythrocytes and plasma cells)	Haematopoietic cell activation and differentiation, and immune function	[137,138,139]
Ly6a	Stem cell antigen 1 (SCA1) or lymphocyte antigen 6 complex	Glycosyl phosphatidylinositol-anchored cell surface protein of the LY6 gene family	HSCs, fibroblasts, endothelial cells, CPCs	Tumourigenicity; cell migration; HSC self-renewal and fate determination	[6,75,140,141,142,143]
Tbx20	T-box 20	Transcription factor of the T-box family	Cardiomyocytes, cardiac pericytes and fibroblasts	Cardiogenesis	[6,75]
Tbx5	T-box 5	Transcription factor of the T-box family	Cardiomyocytes, cardiac fibroblasts	Cardiogenesis	[24,25,90]
Tcf21	Transcription factor 21	Transcription factor of the basic helix-loop-helix family	Epicardium, Cardiac fibroblasts “1”	Initiation of epicardial EMT for fibroblast development; proepicardial cell specification	[26,54,144]
Tert	Telomerase reverse transcriptase	Member of the reverse transcriptase family	Stem and progenitor cells	Cellular immortalisation tumourigenicity; cell proliferation	[145,146]
Vim	Vimentin or fibroblast intermediate filament	Type III intermediate filament protein	Fibroblasts, endothelial cells, smooth muscle cells, Schwann cells	Maintains cellular integrity; EMT; protective role against nuclear rupture and DNA damage during cell migration.	[42,75]
Vwf	von Willebrand factor	Adhesive plasma glycoprotein	Endothelial cells	Platelet adhesion and aggregation	[147]
Wif1	WNT inhibitory factor 1	Lipid-binding protein	Cardiac fibroblasts “2”	Mesoderm segmentation; cardiogenesis	[26,148]
Wt1	Wilms tumour 1	Transcription factor with proline/glutamine-rich DNA-binding domain	Epicardium, fibroblasts, endothelial cells, macrophages	Epicardial EMT; vasculogenesis	[24,59,60,149,150]
Acta2	Alpha smooth muscle actin (αSMA) or Actin alpha 2	Member of the globular Actin protein family	Smooth muscle cells, myofibroblasts, pericytes	Cell motility and contractility	[41,151]

) included those encoding core cardiac transcription factors, stem cell-related markers, fibroblast markers, cardiomyocyte markers, and other genes that were differentially expressed in the SCA1 subpopulations according to the study by Noseda et al. [6]. Moreover, we attempted to detect distinct cardiac subpopulations, including pericytes, endothelial cells, and macrophages, within the cardiac stromal cell populations. The heatmap (Figure 2A) shows that the HL-1 cardiomyocytes were distinct from the cardiac stromal cells; expression of each gene across the different populations is shown in Figure 2B. The cardiac stromal populations expressed the fibroblast-associated genes Cd44, Ddr2, Pdgfrb, Acta2 (encodes αSMA), and Vim, although CFs showed less consistent expression of Pdgfrb and Acta2. Interestingly, Ddr2, Acta2, and Vim were also expressed by HL-1 cardiomyocytes. Additionally, on rare occasions, CFs (n = 2/29) and CTs (n = 3/55) co-expressing Wt1 and Tcf21 were detected, perhaps suggesting an epicardial-derived fibroblast identity. Ly6a (encoding SCA1) and Pdgfra were expressed in all three populations. The three cardiac stromal populations expressed little or no Prg4 and Wif1 but displayed expression of Medag and Col1a1. Ms4a4d expression was detected only in CTs and in some CFs. Regarding pericyte-associated markers, Kcnj8 showed little or no expression in all populations, Colec11 showed sporadic expression, whereas Ng2 showed high expression in all three cardiac stromal populations.

In CTs and CDCs, the most prevalent cardiac transcription factors were Gata4, Tbx20, Hand2, and Mef2a, with little or no expression of Hand1, Isl1, or Nkx2-5. In contrast, CFs showed less uniform expression of Gata4, Tbx20, and Hand2. Unlike CDCs, both CTs and CFs expressed Mef2c heterogeneously. The expression of Tbx5 in CFs followed a similar pattern to that in CDCs but was more sporadic in CTs. Interestingly, only CFs expressed Isl1, which has a very limited expression profile in the adult heart [9,37]. The three cardiac stromal populations showed little or no expression of markers of cardiomyocytes (Nppa, Myh6, and Myl2), haematopoietic cells (Ptprc), or endothelial cells (Cdh5 and Vwf), except for Kdr expression, which was detected in CTs, CFs, and HL-1 cardiomyocytes. Expression of stem cell-related markers (Pou5f1, Nanog, Klf4, and Tert) was observed in the three cardiac stromal populations. Pou5f1, Nanog, and Klf4, but not Tert, were also expressed in HL-1 cardiomyocytes. As expected, HL-1 cardiomyocytes had a distinctly different expression profile and displayed expression of most core cardiac transcription factors (Gata4, Hand2, Mef2a/c, Nkx2-5, and Tbx5/20) and cardiomyocyte markers (Nppa, Myh6, and Myl2).

Expression of Schwann cell-related markers (Plp1 and Kcna1) was detected sporadically in all populations, except HL-1 cardiomyocytes, in which only Plp1, but not Kcna1, was expressed. We also assessed the expression of the macrophage-associated genes Mrc1, Csf1r, and Cx3cr1. Cx3cr1 showed no expression in any population, and Mrc1 was expressed only in CTs. In contrast, Csf1r was expressed in all three stromal populations but showed a more uniform pattern in CTs.

3.3. A Comparison of the Molecular Profiles of CTs, CDCs, and CFs

By principal component analysis (PCA), cardiac stromal cells and HL-1 cardiomyocytes were resolved as discrete groups, consistent with their distinct phenotypes. Gene loadings contributing to each dimension (Dim) suggest that a small subset of genes explains the cross-group variability captured by Dim.1 and Dim.2. The separation of HL-1 cardiomyocytes was attributable to Wif1, Nkx2-5, Mef2c, Nppa, Myl2, and Myh6 (Figure 3A). Some separation of CFs from CTs was also revealed, with the CDCs clustering between the other two populations. This is more clearly seen in the PCA of the stromal populations alone (Figure 3B). The separation of CTs was attributable to Tcf21, as evident in Figure 3B, while the separation of CDCs and CFs was mainly attributable to Tbx5. This Tcf21 versus Tbx5 polarity may reflect functional or origin heterogeneity within cardiac fibroblasts. CFs appeared more dispersed, suggesting higher heterogeneity, compared with CTs and CDCs, which again is more apparent in the PCA of stromal populations only.

Statistical analysis comparing the three cardiac stromal populations revealed that CTs were more similar to CDCs than CFs, with CTs showing significantly higher expression of Vim (p < 0.05), Kdr (p < 0.05), Mrc1 (p < 0.001), Tcf21 (p < 0.0001), Ms4a4d (p < 0.0001), Tbx20 (p < 0.0001), Csf1r (p < 0.0001), Mef2c (p < 0.0001), and Medag (p < 0.05) but lower expression of Tbx5 (p < 0.0001) and Acta2 (p < 0.01) compared to CDCs. In comparison to CFs, CTs showed significantly higher expression of Col1a1 (p < 0.0001), Pdgfrb (p < 0.0001), Hand2 (p < 0.0001), Pou5f1 (p < 0.05), Mrc1 (p < 0.001), Tcf21 (p < 0.0001), Tbx20 (p < 0.0001), Ng2 (p < 0.01), Mef2c (p < 0.05), Medag (p < 0.001), and Gata4 (p < 0.0001) but lower expression of Isl1 (p < 0.0001) and Tbx5 (p < 0.01) (Figure 4A).

Finally, we performed a co-expression correlation analysis using the complete single-cell qRT-PCR dataset to support the inferred biological functions of the stromal and cardiomyocyte populations and to identify co-expression relationships among the genes included in the study (Figure 4B). Correlation plot analysis showed a polarisation between the expression of some fibroblast-associated markers and core cardiac transcription factors, and a positive correlation between the expression of core cardiac transcription factors and cardiomyocyte markers. Tcf21 expression showed a positive correlation with the expression of Tbx20, Ms4a4d, and macrophage-associated genes Csf1r and Mrc1, reflecting their high co-expression in CTs. In addition, expression of the pericyte-associated gene Kcnj8 showed a positive correlation with the expression of Isl1 due to their co-expression pattern in a small number of CFs. Collectively, these data confirm the co-expression of core cardiac transcription factors, associated predominantly with CTs and CDCs, as distinct from the gene profile characterising CFs. Additionally, the enrichment of Tcf21 expression in CTs further delineates CTs from CFs.

4. Discussion

4.1. CTs Exhibit Shorter Expansion Time and Superior Survival Potential Following Serum Starvation

Low survival and retention of transplanted cells following ischaemic events are mainly due to the harsh conditions within the infarct region that include hypoxia, inflammation, and nutrient restriction. Serum in cell culture media provides amino acids and fatty acids, and its withdrawal is therefore considered a form of nutrient restriction; serum starvation is an effective means of inducing cell apoptosis [36]. Here, we found that all three cardiac stromal populations continued to grow following serum starvation, with no significant difference in cell numbers detected at day 3. However, CF cell numbers at day 10 were similar to initially seeded numbers, suggesting substantial cell death. At day 10, CT cell numbers were significantly higher than those of CFs but not CDCs, suggesting that CTs have superior survival potential compared with CFs. This is consistent with the single-cell qRT-PCR results showing that CTs and CDCs share a more similar gene expression profile compared with CFs. Tolerance to serum starvation has also been observed in MSCs under both normoxic and hypoxic culture conditions [38,39].

Reduced robustness was observed during routine culture of CFs compared with CTs and CDCs, as shown by increased time to reach confluency and higher numbers of floating dead cells in culture. CFs displayed reduced recovery following single-cell sorting for the single-cell qRT-PCR experiment compared with CTs and CDCs. It is difficult to pinpoint the cause of this fragility due to differences in gene expression, cell culture media, and coating substrate. Differential effects of various culture coating substrate types on cell proliferation have been shown [40]. However, standardising the culture conditions of all cell types would have been counterproductive, as our aim was to compare the populations as they would be cultured in other laboratories based on their recommended and published culture conditions. Ideally, the survival potential of these different stromal populations should be compared following transplantation into infarcted hearts or in in vitro conditions that more closely mimic the in vivo environment following MI. Overall, our data suggest that CTs and CDCs were more robust than CFs, both in long-term cell expansion and survival under serum starvation. The survival potential of cardiac stromal populations may differ in response to stress in vitro, which can be an indication of their in vivo behaviour following transplantation. Additionally, the reduced expansion time needed to generate CTs compared to CDCs indicates that a CDC-like population for therapeutic use could be achieved more efficiently.

4.2. CTs and CDCs Show a Fibroblast Phenotype with No Evidence of Endothelial Cells or Pericytes

A study by Skelly et al. (2018) examined the single-cell transcriptional profile of adult mouse ventricular nonmyocytes to identify distinct cell populations [26]. Subpopulation heterogeneity within the cardiac fibroblast population was described, where most fibroblasts (“fibroblasts 1”) were characterised by high expression of Medag, Ms4a4d, Pdgfra, and Tcf21, while a smaller, distinct cluster (“fibroblasts 2”) expressed higher levels of Tbx20, Wif1, and Prg4. Although we found that Tbx20 was highly expressed in both CTs and CDCs, Wif1 and Prg4 showed little to no expression in any stromal population. Moreover, Medag and Pdgfra expression was detected in all three populations, whereas Ms4a4d and Tcf21 were only found in CTs and CFs. However, CFs showed less consistent expression of Ms4a4d compared to CTs. This suggests that none of the three stromal populations fit the expression pattern of “fibroblasts 2” and that CTs were the best match to the expression pattern of “fibroblasts 1”. However, an important consideration is that the study by Skelly et al. examined ventricular rather than atrial nonmyocytes and used freshly isolated cells rather than passaged cells, which may undergo phenotypic drift in vitro. Additionally, a limitation of this study is the use of commercially sourced CFs as a reference population. While providing a standardised comparator, variables such as donor age, isolation methods, passage history, and batch variability are not fully defined and may influence gene expression profiles. These factors should therefore be considered when interpreting comparative results.

Additional mesenchymal- and fibroblast-associated genes were assessed, including Vim, Ddr2, Acta2, Cd44, and Col1a1. αSMA, encoded by Acta2, is a protein commonly used to identify smooth muscle cells, pericytes, and myofibroblasts, which acquire a contractile phenotype similar to that of smooth muscle cells upon differentiation [41]. Vim was the only gene expressed in all the screened cells of all populations. Vimentin is an intermediate filament protein that has been widely used as a reliable marker of mesenchymal cells, highly migratory cells derived by EMT, and it has a protective role against nuclear rupture and DNA damage during migration [42]. None of these markers showed specificity to a particular population. Vim, Ddr2, and Acta2 were also expressed in HL-1 cardiomyocytes, an adult immortalised atrial cell line that resembles a mitotic embryonic cardiomyocyte rather than a mature adult cell [35], as reflected by its immature energy metabolism [43]. The expression of mesenchymal- and fibroblast-associated genes in HL-1 cardiomyocytes may reflect the lack of specificity of these markers in in vitro cultures, the immature phenotype of HL-1 cardiomyocytes, and/or their dedifferentiation in culture. Genomic instability is a common setback associated with the use of cell lines in research [44]. Accordingly, HL-1 cells are used here as an in vitro reference for cardiomyocyte-associated gene expression rather than as a model of mature primary cardiomyocytes. Nonetheless, the data suggest a fibroblast-like phenotype for all stromal populations. A combination of Pdgfra, Col1a1, and Medag, expressed in stromal cells but not in HL-1 cells, might provide a more reliable set of markers for identifying in vitro fibroblast cultures.

Commonly used markers to identify pericytes (e.g., Pdgfrb and Ng2) are relatively non-specific. However, Kcnj8 and Colec11 have been identified as genes that show higher expression in pericytes relative to other cell types [26]. We found little or no expression of Kcnj8 in any population. Colec11 showed sporadic expression, whereas Ng2 and Pdgfrb showed high expression in all three cardiac stromal populations. Therefore, there is no conclusive evidence that these stromal populations have a pericyte phenotype. Our data also suggest that Kcnj8 and Colec11 may be more specific markers of pericytes than commonly used markers. Moreover, Kcnj8 expression was positively correlated with Isl1, attributable to co-expression in a small number of CFs. ISL1-mediated differentiation of coronary pericytes into coronary artery smooth muscle cells has been described in the embryonic heart [45]. Although Isl1 is detected in the developing heart and multipotent ISL1⁺ progenitors play an important role in giving rise to cardiac lineages during development [46], ISL1⁺ cells are very rare in adult human and murine hearts [10,47].

The three cardiac stromal populations showed very little or no expression of markers of cardiomyocytes (Nppa, Myh6, and Myl2), haematopoietic cells (Ptprc), or endothelial cells (Cdh5 and Vwf). Expression of Nppa and Myh6 was detected in a few CDCs (2 Nppa⁺ cells, n = 41) and CTs (4 Myh6⁺ cells, n = 55). Given that these cells were at passage 3, it is highly unlikely that the Nppa- or Myh6-expressing cells are surviving cardiomyocytes from the isolation protocol. This expression may result from spontaneous differentiation in culture or may reflect progenitor status. In addition, Kdr expression, a marker of endothelial cells, was detected in the CTs, CFs and HL-1 cardiomyocytes. During development, KDR⁺ progenitors derived from the MESP1⁺ mesodermal lineage possess cardiovascular tri-lineage potential [48,49]. Lineage tracing has shown that precursors expressing Gata5 and Wt1 give rise to the adult SCA1⁺ SP⁺ population (PDGFRα⁺/CD31⁻), whereas the adult SCA1⁺ SP⁻ population (PDGFRα⁻/CD31⁺) is mainly derived from precursors expressing Kdr [6]. However, CTs lack the expression of other endothelial markers observed in the SCA1⁺ SP⁻ population. Kdr expression has also been reported in coronary adventitial progenitors with cardiogenic potential [50] and in endothelial progenitors [51]. Therefore, Kdr expression may reflect some progenitor status rather than the presence of endothelial cells. CDCs have been examined at the single-cell transcriptomics level [52] and characterised as mesenchymal/stromal/fibroblast-like, with a small subset of endothelial-like cells. Notably, transplantation of SCA1⁺ CDCs, but not SCA1⁻ CDCs, enhanced cardiac function after MI, highlighting previously unappreciated functional differences between CDC subpopulations.

A few cells co-expressing Wt1 and Tcf21 were detected in both the CF and CT populations. In the developing heart, epicardium cells are identified through the expression of Wt1, Tbx18, Tcf21, and Raldh2 [53]. The majority of TCF21⁺ epicardium-derived cells (EPDCs) are committed to the cardiac fibroblast lineage, and Tcf21 expression persists in cardiac fibroblasts of the adult heart [54,55,56]. Although WT1 is downregulated following gestation [57,58], its expression is still detected in the epicardium and a subset of coronary endothelial cells in the healthy adult heart [55,59]. Enrichment of the epicardial markers Wt1 and Tcf21 has been shown in adult cardiac fibroblasts [24]. More recent studies have shown that Wt1 is expressed in more cell types than previously appreciated, including a Wt1b⁺ macrophage subpopulation essential for zebrafish heart and fin regeneration [60]. In our data, Wt1 expression did not correlate with Mrc1 or Kdr, but all cells expressing Wt1 were also positive for Tcf21, suggesting a fibroblast phenotype.

Expression of stem cell-related markers (Pou5f1, Nanog, Klf4, and Tert) was observed in all three cardiac stromal populations. In addition, Pou5f1, Nanog, and Klf4, but not Tert, were also expressed in HL-1 cardiomyocytes. The expression of these stem cell-associated genes (Pou5f1, Klf4, Nanog) has been observed in mouse cardiac stromal populations and rat CDCs [61]. A sporadic expression of these markers has also been noted in primary neonatal cardiomyocytes [6]. Other studies have also reported the expression of Pou5f1 [62], Klf4 [63], and Nanog [64,65] in adult differentiated cells. Furthermore, Tert is activated in human and mouse primary fibroblasts to protect against malignant transformation [66]. As stem cell-associated genes are implicated in tumorigenesis and proliferation [67,68], their expression in HL-1 cells and cardiac stromal populations is not unexpected.

4.3. Tcf21 Versus Tbx5 Polarity May Represent Functional Heterogeneity in Fibroblast Populations

Noseda et al. [6] revealed an enrichment of Tcf21 and Pdgfra; the core cardiac genes Gata4, Mef2a/c, Tbx5/20, and Hand2; and a lack of Cdh5 and Kdr in the SCA1⁺ SP⁺ population compared with total SCA1⁺ and SCA1⁺ SP⁻ populations. In our study, Cdh5 expression was absent in all cardiac stromal populations. Both CTs and CDCs showed high expression of the core cardiac transcription factors [69] Gata4, Tbx20, Hand2, and Mef2a, with CTs expressing Mef2c and CDCs expressing Tbx5. Interestingly, among these transcription factors, Tbx5 showed the lowest expression in the SCA1⁺ SP⁺ population. Moreover, these results are consistent with observations by Furtado et al., showing enrichment of core cardiac genes in cardiac fibroblasts [24]. In contrast, CFs showed less uniform expression of Gata4, Tbx20, and Hand2. PCA revealed an interesting polarisation between Tcf21 and Tbx5, which was also reflected in the correlation plot, showing a strong negative correlation between the two genes. Indeed, the clustering separation of CTs and CFs can be attributed to the expression of these two genes. Although the significance of this polarisation remains unclear, it is an interesting observation that warrants further investigation, as it may indicate functional heterogeneity among fibroblast populations.

In addition to Tcf21 enrichment in CTs relative to other stromal populations, the macrophage-associated gene Mrc1 was detected in a subset of CTs. This was reflected in the positive correlation between Mrc1 and Tcf21 and the negative correlation between Mrc1 and Tbx5. Mrc1 is reported to be more highly expressed in macrophages relative to other cardiac cell types [26]. Csf1r, enriched in M2-polarised macrophages [70], was expressed in all stromal populations, albeit at significantly higher levels in CTs. Csf1r expression has also been used as an indicator of embryonic myeloid lineage commitment [71]. However, cells expressing Mrc1 or Csf1r also co-expressed fibroblast markers and did not form distinct clusters. In a study using scRNA-seq data to explore the transcriptional profile of CSF1R+ monocytes, a small population of fibroblast-like cells expressing Pdgfrα, Pdgfrb, and Tcf21 was identified [72]. However, in a follow-up study, the authors analysed published data and concluded that fibroblasts do not express detectable levels of Csf1r or Cx3cr1 [73]. Hybrid mesenchymal phenotypes have been described previously, where fibroblast-associated genes are expressed in minor subpopulations of macrophages or endothelial cells [26]. However, because the hematopoietic marker Ptprc is absent in CTs, it is more likely that macrophage-associated genes were upregulated in a subset of fibroblasts rather than the reverse. This phenomenon has also been observed in in vitro fibroblast cultures [74] and may result from in vitro expansion.

5. Conclusions

Our results suggest that (1) CTs are similar to CDCs and differ from CFs in their expression profiles, enrichment of core cardiac factors, and robustness; (2) CTs, CDCs, and CFs resemble “fibroblasts 1”, the predominant fibroblast subtype in the adult murine heart; (3) CDCs cluster between CTs and CFs, which separate based on their expression of Tcf21 and Tbx5, respectively; (4) CTs are enriched for the macrophage-associated genes Mrc1 and Csf1r, likely reflecting a fibroblast phenotype in a subset of CTs; and (5) PCA reveals a transcriptional gradient, suggesting heterogeneity within fibroblast subtypes that may correspond to functional differences. Overall, our findings highlight the heterogeneity of the cardiac stromal compartment and how isolation protocols may enrich for specific cell subpopulations.