Molecular signature of progenitor cells isolated from young and adult human hearts

The loss of endogenous cardiac regenerative capacity within the first week of postnatal life has intensified clinical trials to induce cardiac regeneration in the adult mammalian heart using different progenitor cell types. We hypothesized that donor age-related phenotypic and functional characteristics of cardiac progenitor cells (CPC) account for mixed results of cell-based cardiac repair. We compared expression profiles and cell turnover rates of human heart-derived c-kitpos progenitors (c-kitpos CPC) and cardiosphere-derived cells (CDC) from young and adult donor origin and studied their in vitro angiogenic and cardiac differentiation potential, which can be relevant for cardiac repair. We report that 3-dimensional CDC expansion recapitulates a conducive environment for growth factor and cytokine release from adult donor cells (aCDC) that optimally supports vascular tube formation and vessel sprouting. Transdifferentiation capacity of c-kitpos CPCs and CDCs towards cardiomyocyte-like cells was modest, however, most notable in young c-kitpos cells and adult CDCs. Progenitors isolated with different methods thus show cell- and donor-specific characteristics that may account for variable contributions in functional myocardial recovery.

Gene expression profiling of CDCs and c-kit pos CPCs derived from adult and young donors. To identify biological differences between different progenitor cell populations, we performed RNA sequencing and pathway analysis of CDCs and c-kit pos CPCs derived from adult and young donors. The full list of differentially expressed transcripts and identified processes and pathways is available as an Online Supplement. Based on hierarchical clustering, the greatest contrast was observed between aCDCs and yCDCs and between aCDCs and c-kit pos CPCs derived from young and adult donors (Fig. 2a). In total, 2,886 transcripts showed a significant expression difference of at least 2-fold between aCDCs and yCDCs: 1,423 had higher expression levels and 1,463 had lower expression levels. String network analysis revealed clusters of transcripts related to cytokines and inflammation, growth factors, cell motility and extracellular matrix (Fig. 2b). A second independent approach using gene ontology analysis confirms significant enrichment of biological processes related to the inflammatory response, cell growth, angiogenesis and cell migration in aCDCs compared to c-kit pos aCPCs (Fig. 2c, blue) and enrichment for cell cycle processes, angiogenesis and cell proliferation when comparing aCDCs to yCDCs (Fig. 2d, blue). We then performed pathway analysis and quantified the activation of the identified canonical pathways. More specifically, this revealed significant activation of Toll-like Receptor, TREM1 and IL-8 signalling in aCDCs compared to c-kit pos aCPCs (Fig. 2c, green), and cell cycle related pathways in aCDCs compared to yCDCs (Fig. 2d, green). In addition, eNOS and glioma invasiveness signalling were more activated whereas LXR/ RXR and PPARα/RXRα activation, GnRH (Gonadotropin Releasing Hormone) signalling, α-adrenergic, and cardiac hypertrophy signalling were significantly deactivated in aCDCs ( Fig. 2c and Table 1).
Subsequently, we investigated the identified functional differences on a gene expression level for immunomodulation (Fig. 2e), cell cycle markers (Fig. 2f) and stem cell markers and growth factors (Fig. 2g). With respect to immunomodulation, many inflammatory genes including IL1B, IL8 (CXCL8) and TGFB1 were higher expressed in aCDCs compared to yCDCs, but also when compared to c-kit pos aCPCs (Fig. 2e). The difference between CDCs and c-kit-selected CPCs is predominantly driven by greater inflammatory signalling (Toll-like receptors TLR, TREM1, IL8) with the highest activation in aCDCs, an intermediate profile in yCDCs, and the lowest expression levels in c-kit pos CPCs irrelevant of donor age. Other transcripts including IL18, TGFB2 and TLR4, however, are lower expressed in aCDCs compared to yCDCs. The expression profile of cell cycle-related transcripts in aCDCs is consistent with greater proliferation potential when compared to yCDCs, while this difference appears to be less pronounced in c-kit-selected CPCs with a greater proliferation potential for the c-kit pos yCPCs (Fig. 2f). For the transcription factors, only MEF2C was higher expressed in aCDCs compared to yCDCs and c-kit pos CPCs isolated from adult and young donors. In contrast, the expression of MEST, a member of the hydrolase superfamily important in early mesoderm formation, is the lowest in aCDCs while TBX5 and GATA4 had comparable transcript levels in all progenitor cell populations. The expression of angiogenic growth factors CXCL12 (stromal-derived factor-1 alpha (SDF1-α)), HGF (hepatocyte growth factor), PlGF (placental growth factor), VEGFA (vascular endothelial growth factor A) and FGF2 (fibroblast growth factor 2, or basic fibroblast growth factor) is more highly expressed in aCDCs and to a lesser extent in yCDCs compared to c-kit pos aCPCs and c-kit pos yCPCs. Finally, the stem cell factors KIT and KLF4 are the lowest in aCDCs compared to yCDCs and c-kit pos CPCs derived from young and adult donors (Fig. 2g).
Taken together and when compared to c-kit pos aCPCs, aCDCs display a higher activation of cell cycle pathways, inflammatory pathways important for reparative responses (TLR, TREM1, Pattern Recognition Receptors), critical growth factors and eNOS signalling, all of which may contribute to a unique cardioprotective capacity during cytotoxic stress conditions. Gene signature and proliferation capacity of heart-derived c-kit pos CPCs and CDCs. To confirm RNA sequencing expression data, we performed qPCR analysis and measured significantly higher CD117 (c-kit) transcript levels in c-kit pos yCPCs and c-kit pos aCPCs compared to aCDCs (Fig. 3a) and within the CDC population no significant difference was found in CD117 levels. The fibroblast marker DDR2, was equally expressed in young and adult c-kit pos CPCs and CDCs. The expression of the proliferation-associated and hypoxia-inducible endoglin CD105 was prominent in all four cell populations but significantly higher in aCDCs compared to yCDCs and c-kit pos aCPCs (p < 0.05). Conversely, aCDCs showed significantly lower gene expression level of the mesenchymal marker CD90 compared to yCDCs and c-kit pos yCPCs and c-kit pos aCPCs (p < 0.05). To study lineage commitment of human progenitor cells, expression levels of the endothelial cell (EC) marker CD31, the early cardiac transcription factors MEF2c and GATA4 and the smooth muscle cell (SMC) marker MYH11, were measured. Interestingly, CD31 and MEF2c expression was significantly higher in aCDCs compared to yCDCs and c-kit pos CPCs derived from young and adult donors (p < 0.05). The same trend was observed for GATA4 expression however the expression was only significant compared to c-kit pos yCPC. The expression level of the SMC marker MYH11 was significantly higher in cells derived from young donors compared to cells derived from adult donors irrespective of the isolation method (p < 0.05).
Because heart-derived progenitor cells have important paracrine effector mechanisms, we compared gene expression of different candidate growth factors and cytokines. Transcript levels of FGF2 were significantly higher in yCDCs and aCDCs than in c-kit pos yCPCs (p < 0.05). In addition, the expression of HGF and PlGF, was To isolate c-kit pos cells, a human RA appendage biopsy is minced into small pieces and collagenase digested to culture a heterogeneous mixture of single cells for 2 passages prior to MACS-based selection. To obtain CDCs, a mild collagenase-based digestion is performed on small myocardial fragments to allow outgrowth of fibroblasts and phase-bright cells from the cardiac explants. The latter are transferred onto a poly-D-Lysine coated dish to form cardiospheres. Finally, the cardiospheres are collected and grown as monolayer cells called CDCs. (b) Phase bright-field pictures of c-kit pos CPCs and CDCs derived from young and adult donors. Scale bars represent 200 µm. CPC, cardiac progenitor cell; CDC, cardiosphere-derived cell; RA, right atrial. (a) Heat map of Spearman correlations with the greatest differences observed between aCDCs and c-kit pos CPCs and between aCDCs and yCDCs. (b) String network analysis using the top significant genes showed clusters of transcripts that are related to cytokines and inflammation, growth factors, cell motility and extracellular matrix. (c,d) Gene ontology results using DAVID software (blue) and pathway analysis (IPA) (green) comparing aCDCs and c-kit pos aCPCs (c) and yCDCs and aCDCs (d) confirming the string network analysis. Compared to c-kit pos aCPCs, aCDCs are more enriched in inflammatory processes (c), while cell cycle pathways are more prominent in aCDCs compared to yCDCs (d). significantly higher in aCDCs compared to c-kit pos yCPCs and c-kit pos aCPCs (p < 0.05 and p < 0.01, respectively). Finally, IGF-1 and insulin growth factor-receptor2 (IGF-2R) gene expression were significantly higher in aCDCs than in c-kit pos aCPCs (p < 0.05; Fig. S1). Also, the expression of inflammation markers IL1B, IL8 and TGFB1 were higher in aCDCs compared to yCDCs, which confirms the RNA sequencing data (Fig. S2). In addition, TGFB1 expression was highest in aCDCs compared to yCDCs, c-kit pos yCPCs and c-kit pos aCPCs (data not shown).
To evaluate proliferation capacity of the different progenitor cell types, we calculated population-doubling times (PDT) (Fig. 3c). Consistent with observed differences in transcript levels of cell-cycle regulated genes (Fig. 2f), c-kit pos yCPCs have a significantly lower PDT and thus higher proliferation rate compared to c-kit pos aCPCs (2.25 ± 0.10 days versus 3.62 ± 0.24 days, p < 0.02). In contrast, adult progenitor cells isolated according to the explant culture technique (aCDC) showed significantly lower PDTs and thus higher proliferation rates than yCDCs (1.78 ± 0.11 days versus 6.37 ± 0.66 days, p < 0.02). In addition, the average passage number at which proliferation of the cells stagnated or started to decline, was significantly lower in c-kit pos aCPCs and yCDCs (13.0 ± 0.7 and 12.1 ± 0.7 passages, respectively), the cell groups with the highest PDTs, compared to c-kit pos yCPCs and aCDCs (21.3 ± 3.0 and 22.0 ± 0.4 passages, respectively).
Vasculogenic and angiogenic potential of c-kit pos CPCs and CDCs derived from adult and young donors. Next, the vasculogenic potential of human progenitors was evaluated after 6 hours culture of the cells on Matrigel ® (Fig. 4a-c). To develop firmly established vascular networks, progenitor cells must be able to sprout, form branches and interconnect them into segments, bounded by a junction or node (Fig. 4a). The number of branches was similar between the 4 different cell populations (c-kit pos yCPC, 6.1 ± 1.4/mm 2 ; c-kit pos aCPC, 7.6 ± 0.8/mm 2 ; yCDC, 5.76 ± 0.17/mm 2 ; aCDC, 5.2 ± 0.5/mm 2 ). However, the ability to form vascular networks, as reflected by the number of nodes, was significantly lower for c-kit pos yCPCs (c-kit pos yCPC, 23.0 ± 5.6/mm 2 ) compared to progenitor cells isolated from adult donors (aCDC, 57.1 ± 8.1/mm 2 ). Of note, yCDCs show a trend for a higher tube formation potential when compared to c-kit pos yCPCs (Fig. 4a,b). These relative differences in vascular network formation may in part be attributable to variable CD90 mRNA and protein expression patterns (Fig. 3), which have been shown in previous studies to inversely correlate with endothelial transdifferentiation and regeneration capacity 15 . The CD90 expression in yCDCs varied between the different donors ( Fig. 3). To investigate if variation in CD90 expression is associated with differences in tube formation capacity, we measured the correlation between CD90 mRNA expression and the number of nodes. Here, we report an inverse correlation between the number of nodes and CD90 mRNA expression of aCDCs and yCDCs (Fig. 4c). Because the in vitro Matrigel ® assay does not allow evaluation of vascular sprouting and formation of lumen structures, a fundamental property of ECs, we complemented our analysis with an in vitro 3D spheroid assay in Matrigel ® to study sprouting angiogenesis (Fig. 4d). The additional 3D spheroid assay confirmed the superior potential of aCDCs (n = 2) to sprout and form a vascular network compared to yCDCs (n = 2), c-kit pos yCPCs (n = 2) and c-kit pos aCPCs (n = 2). Taken together, our data indicate that all human progenitor types have an intrinsic vasculogenic potential, however, progenitors isolated from young donors are possibly more immature, with reduced endothelial lineage commitment resulting in less developed vascular networks.
Cardiac differentiation potential of c-kit pos CPCs and CDCs derived from adult and young donors. To test the hypothesis that heart-derived progenitors from young donors have a greater cardiac differentiation potential than progenitors derived from adult donors, all cell populations were incubated with cardiac differentiation medium (DIFF) or regular expansion medium (CON). Immunofluorescence stainings showed the presence of the cardiac contractile protein α-sarcomeric actinin in CDCs and c-kit pos CPCs after 12 days incubation with DIFF medium (Fig. 5a,b). Contrary to our hypothesis, we observed a greater increase of α-sarcomeric actinin-positive cells (DIFF compared to CON) under these culture conditions in aCDCs compared to yCDCs. However, the opposite effect was observed in ckit pos CPCs, with a greater increase of α-sarcomeric actinin-positive cells in c-kit pos yCPCs compared to c-kit pos aCPCs (Fig. 5b). This suggests a greater cardiac differentiation potential of c-kit pos yCPCs compared to c-kit pos aCPCs, yCDCs and aCDCs. Of note, no contractile phenotype was observed after 12 days incubation with DIFF medium. In addition, expression of the cardiac lineage markers GATA4, MEF2c and NKX2.5 confirmed activation of the cardiac differentiation program in all cells treated with cardiac differentiation medium (DIFF) (Fig. 5c), which is consistent with the increased α-sarcomeric actinin-positive cells following incubation in DIFF medium (Fig. 5a,b).
Culture expansion conditions do not explain the observed differences in cell phenotype and functional activity between CDCs and c-kit pos CPCs. Finally, the cell phenotype and functional activity was investigated after culturing CDCs (yCDC, n = 2; aCDC, n = 2) in c-kit pos CPC expansion medium and c-kit pos CPCs (c-kit pos yCPC, n = 2; c-kit pos aCPC, n = 2) in CDC expansion medium on fibronectin-coated culture plates for at least 2 passages. The expression levels of CD117, CD105 and CD90 in these cells cultured in their counterpart medium (SWAP) was tested with qPCR and compared to the original transcriptome as reported in Fig. 3a  (Fig. S4A). We measured similar expression profiles for all 3 genes irrespective of the expansion medium with the highest expression of CD117 in c-kit pos yCPCs. CD105 was equally expressed in all cell groups and CD90 showed again the lowest expression in aCDCs. Next, the vasculogenic potential of the four cell groups was explored after culturing the progenitors with their counterpart medium (SWAP). Again, results showed that the ability to form vascular networks, as reflected by the number of nodes, was lower in progenitors isolated from young donors compared to adult donors (Fig. S4B). In addition and again in concert with our original observations (Fig. 4a,b), the number of junctions, meshes, segments and branches showed a comparable pattern with progenitors derived from adult donors forming more complex networks compared to progenitors derived from young donors. In conclusion, these data suggest that CDC and c-kit pos CPC culture conditions do not account for the phenotypic and functional differences between the four progenitor groups.

Discussion
Recent evidence indicates that the mammalian heart harbours multiple cardiogenic progenitor cell populations and stages a marked endogenous regenerative response after tissue damage in early postnatal life 18 . Because this capacity for scar-free healing through cardiomyogenesis is lost in the adult heart, intracardiac transfer of c-kit pos CPCs and a heterogeneous explant population of CDCs has been introduced in clinical trials in an effort to enhance cardiac repair after ischemic injury. In the present study, we aimed to bridge the gap in our understanding of the molecular signature of these 2 human heart-derived progenitor cell populations obtained at different stages in postnatal life. We compared in vitro the phenotypic, functional and genetic characteristics of heart-derived progenitor cells from neonates and young patients with congenital heart disease undergoing corrective surgery and from adult patients with advanced ischemic and valvular disease. We deliberately studied culture expanded cells after 4 to 5 passages as was peformed in pilot clinical studies (e.g. SCIPIO 12 and CADUCEUS 13,14 trials) and observed a distinct molecular signature between CDCs of different donor age, but not between c-kit-selected progenitor cells of young and adult patients. The molecular pathways with different states of activation or repression between adult and young CDCs play a pivotal role in cell survival, proliferation and endothelial cell (EC) or cardiac myocyte-specific functions. Importantly, adult heart-derived and culture expanded CDCs showed significantly higher activation of the EC-associated NOS3 gene, showed less CD90 surface gene expression and secreted high levels of pro-angiogenic cytokines and growth factors. These phenotypic characteristics conferred greater capacity for vascular tube formation and sprouting in a 3D spheroid assay. Transcriptional profiling also markedly contrasted aCDCs with c-kit pos CPCs from young and adult heart as evidenced by the consistent activation of cell cycle regulatory genes and of interest, a robust induction of immunomodulatory genes.
Transcriptome analysis of aCDCs revealed a baseline upregulation of myocyte-specific transcription factors (GATA4 and MEF2c) and the EC-specific marker (CD31), a constituent of the endothelial intercellular junction, also designated as PECAM-1 (platelet endothelial cell adhesion molecule-1). Adult CDCs also expressed significantly higher levels of endoglin or CD105, known to play a crucial role in angiogenesis 19 , but contrary to MSCs 20 co-expression of CD90 was significantly lower. CD90 expression is lower in aCDCs compared to yCDCs, c-kit pos yCPCs and c-kit pos aCPCs. In fact, it has previously been shown that injection of CDCs expressing high levels of CD90, is associated with a reduced reparative capacity and in contrast CD90 neg CDCs induce a more pronounced therapeutic effect 15 . Of interest, this inverse association is also clearly evidenced in our in vitro network formation assay where aCDCs induce markedly complex vascular network development. Taken together, transcriptional analysis suggests that heterogeneous aCDCs express early lineage-specific genes, some of which recapitulate an in vivo pattern in newly formed cells derived from activation and commitment of resident progenitor cells after injury 21 . Somewhat unexpectedly, aCDCs had also a significantly lower PDT and thus faster cell turnover compared to their counterparts from young hearts, but this observation was consistent with the significantly higher expression of cyclins and cell cycle regulatory genes. The faster cell turnover in aCDCs than in yCDCs was not recapitulated in c-kit pos CPCs, where cells from adult donors showed a significantly slower cell turnover than c-kit pos yCPCs. A recent study showed that human Isl1 pos c-kit pos CPC isolated from adult patients harbor an altered miRNA expression profile that correlated with a reduced proliferation capacity and abrogated response to growth factors compared to Isl1 pos c-kit pos CPC derived from neonates 22 . Another study, comparing CDCs derived from patients of various ages, showed limited differences in proliferation capacity, angiogenic potential and expression of growth factors between CDCs from young and elderly patients (≥65 years) 23 . However, in this study the youngest donor was 2 years old, the median age 72.5 years and 61.5% of them were 65 years or older which is in striking contrast with our study. The CDCs derived from young patients were probably underrepresented compared to the CDCs derived from elderly patients to see significant differences whereas in our study CDCs from much younger donors were studied. In contrast, S. Kaushal and co-workers compared CDCs and c-kit pos CPCs derived from donors with various ages, and reported a greater in vivo regenerative capacity and a higher secretion of angiogenic factors in CDCs derived from neonates (<30 days old) 24,25 . In our study, we only have 1 and 2 neonates included in yCDC and c-kit pos yCPC, respectively, and this underrepresentation might account for the different results. Also, CDCs from different young ages (neonates, infants and children) showed a comparable in vivo regenerative potential 26 . However, in contrast to our study, the angiogenic potential was not investigated in this study. Several studies showed that cell-based therapy improves cardiac function through paracrine mechanisms (mediation of angiogenesis, anti-apoptosis and recruitment of endogenous progenitors) since incorporation of injected cells is rare event 10,27 . In this study, aCDCs show a broad expression panel of angiogenic cytokines and potent cytoprotective growth factors, including IGF-1, HGF, bFGF, PlGF, VEGF, and SDF-1, all of which may mediate the paracrine effects of CDCs. Also, heart-derived CDCs displayed a strikingly distinct exosome release pattern when compared to CPCs. These exosomes gained a lot of interest in view of their important role in paracrine and autocrine signalling 28 . In addition, injection of cardiac progenitor-derived exosomes in an AMI rat model protected CM against apoptosis, showed enhanced angiogenesis and improved cardiac function 29 . However, follow-up studies need to examine to what extent exosome-mediated signalling can affect molecular and cellular pathways that are critically important for cardiac repair or protection. Also, a direct correlation of a particular secreted factor(s), the importance of exosomes as a mediator and in vitro tube formation was not performed in this current study.
Finally, we investigated the cardiogenic potential of CDCs and c-kit pos CPCs in an in vitro differentiation assay. C-kit pos CPCs derived from young donors and aCDCs showed a tendency for greater transdifferentiation capacity into cardiomyocyte-like cells than c-kit pos aCPCs and yCDCs. The greater commitment towards cardiac lineages for c-kit pos yCPCs and aCDCs and the marked immunomodulatory signature and neovascularization potential of aCDCs represent interesting characteristics for future cell transfer strategies. In chronic ischemic cardiomyopathy, the maladaptive immune activation that accompanies ischemic dysfunction would require a combined intervention that at the same time mitigates excessive immune activation while supporting neovascularization and cardiomyogenesis. Recent reports have called our attention for immunomodulatory properties of cardiac progenitors that are capable of generating an antioxidant, anti-fibrotic, and anti-inflammatory environment 30 . Additionally, de Couto et al. showed that injection of CDCs or CDC exosomes in a rat AMI model reduces the number of pro-inflammatory macrophages and simultaneously polarizes the macrophages into a cardioprotective phenotype resulting in less CM apoptosis and reduced infarct size 31,32 . We observed greater activation of some inflammatory transcripts in aCDCs compared to yCDCs and c-kit pos CPCs, while other transcripts including TLR4, implicated in post-infarction inflammation 33 , were less expressed. Pro-inflammatory cytokines IL1B and IL8 are both more highly expressed in aCDCs and promote inflammatory leukocyte recruitment and activation, while IL1 delays myofibroblast activation and the potent pro-angiogenic IL8 contributes to infarct neovascularization 34 . TGF-β is a key player in the transdifferentiation of fibroblasts into myofibroblasts which secrete ECM proteins to form fibrotic scar tissue 35 . We measured higher TGFB1 expression in aCDCs compared to yCDCs and c-kit pos CPCs, but lower TGFB2 levels. IL18 stimulates IFN-γ secretion by NK-cells and T-cells which leads to stimulation of pro-inflammatory macrophages 36 . Our data show that IL18 transcripts are lower in aCDCs compared to yCDCs and c-kit pos CPCs. Taken together, the biological significance of these variable inflammatory gene expression patterns remains incompletely understood and requires targeted in vivo studies in the context of cardiac dysfunction.
Several limitations need to be considered in our study. First, the limited amount of RNA available from human progenitor cells precludes transcriptional analysis of a wider panel of genes using qRT-PCR. Second, size constraints of myocardial biopsies precluded cell biology studies on freshly isolated and non-passaged cells and prompted standardized analysis on multi-passaged cells with similar expansion profiles and passage numbers as used in clinical trials. At the same time, limited tissue availability did not allow parallel isolation protocols for both cell types from identical tissue samples. Although ex vivo culture expansion may affect genetic signatures, the consistency of the data reported in our analysis irrespective of expansion medium likely reflects the imprinted phenotype of the cells as used in recent clinical trials. Third, the great age range present in the c-kit pos yCPC and yCDC groups and the underrepresentation of neonates may mask the potential of c-kit pos CPCs and CDCs derived from neonates, which should be addressed in further studies. Fourth, we acknowledge the importance of unraveling the factors responsible for the functional effects in our in vitro studies, which needs to be addressed in follow-up studies. Finally, additional mechanisms operate in vivo including stimulation of resident cardiac cells and cytoprotective effects on ischemic cells, which have not been tested directly in this in vitro study.
In conclusion, phenotypic analysis of aCDCs reveals paracrine and immunomodulatory effects that appear to be distinct from those in yCDCs. In addition, c-kit pos yCPCs and aCDCs comprise a greater differentiation potential into cardiomyocyte-like cells. Whether these in vitro cell-specific findings correlate with clinically significant changes and whether combination of different progenitor cell types may confer greater clinical benefit than either cell type alone, needs to be addressed in targeted studies. Isolation and culture of human progenitor cells. Cardiac tissue samples were obtained from the RA appendage that was routinely discarded from patients undergoing intra-cardiac surgeries. Patients undergoing extra-cardiac surgery were excluded. All adult patients were referred for valve surgery, assist device implantation or heart transplantation. Neonatal RA appendage specimens were collected during corrective surgery for congenital heart disease. No other specific inclusion criteria were used. Human c-kit pos (CD117) CPCs were obtained after collagenase-based digestion of a RA appendage tissue sample based on the methods used in the SCIPIO trial 12 . Briefly, adipose tissue was removed and tissue samples were minced into small pieces before collagenase-based (Collagenase NB4 Standard Grade from Clostridium histolyticum, Serva Electrophoresis) digestion in Ham's F12 medium containing 1-3 mg/mL collagenase. The supernatant containing single cells were collected and plated in Ham's F12 medium (Lonza) supplemented with 10% FBS (ThermoFisher Scientific), 1% penicillin/streptomycin (5000 units/mL penicillin, 5000 µg/mL streptomycin; ThermoFisher Scientific), 5 mU/mL human erythropoietin (EPO) (Sigma-Aldrich). When 80-90% confluency was reached, positive selection for c-kit by magnetic-activated cell sorting (MACS) (CD117 microbead kit, Miltenyi Biotec) was performed. Medium was changed twice a week and cells were passaged when 80-90% confluency was reached. Human CDCs were derived according to the protocol reported by Marbán et al. 10 with modifications. RA appendage tissue samples were minced and mildly digested with collagenase (Collagenase NB4 Standard Grade from Clostridium histolyticum, Serva Electrophoresis). Afterwards explants were generated and plated on fibronectin-coated (25 µg/mL) culture plates in cardiac explant medium (CEM) (Iscove's modified Dulbecco's medium (IMDM) supplemented with 20% FBS, 1% penicillin/ streptomycin (5000 units/mL penicillin, 5000 µg/mL streptomycin), 2 mM L-glutamine (200 mM) and 0.1 mM 2-mercaptoethanol (50 mM)). Tissue samples obtained from young donor patients were directly minced into explants without collagenase digestion. After 2 to 3 weeks, phase-bright cells growing on top of the cardiac fibroblast outgrowth, were collected and cultured at low density (15,000-30,000 cells/mL) on poly-D-Lysin-coated (Millipore) culture plates in cardiosphere-growing medium (CGM) (35% IMDM and 65% D-MEM-F12 supplemented with 10% FBS, 1% penicillin/streptomycin (5000 units/mL penicillin, 5000 µg/mL streptomycin), 2 mM L-glutamine (200 mM), 0.1 mM 2-mercaptoethanol (50 mM) and 1X B-27 supplement (50X)). After ca. 5 days, cardiospheres were collected and plated on fibronectin-coated (25 µg/mL) culture plates in CGM medium. Phase-bright cells were harvested from individual explant cultures every 3 to 5 days up to 4 consecutive times. Medium of CDCs was changed twice a week and cells were passaged when 90% confluency was reached.
String network analysis was performed on the top significant genes 37 . Clustering on the network analysis was performed using the MCL algorithm with inflation parameter set to 3. Gene ontology analysis to identify underlying biological processes was performed using DAVID Functional Annotation Tool 6.8 38 . All differentially expressed genes were submitted to Ingenuity Pathway Analysis (IPA). The top canonical pathways, significantly enriched by the differentially expressed genes, were selected by performing a Fisher Exact t-test (P < 0.05) on the ratio of enriched genes divided by the total number of genes in the pathway. A z-score statistic was calculated to predict the activation state of the pathway based on the fold change of the genes included in that pathway, with a cut-off z-score of >|1.5|. The sequencing data has been made available through the GEO website.

Real-time qRT-PCR. Total RNA from cultured cells was isolated using RNeasy Micro Kits (Qiagen) with
DNase treatment. The first strand reverse transcription was performed using GoScript Reverse Transcription System (Promega). Real-time qRT-PCR was performed using TaqMan Fast Universal PCR Master Mix and TaqMan Gene Expression Assays on a StepOnePlus System (Applied Biosystems). All reactions were run in duplicate and results were expressed as 2 −ΔΔCt . Relative gene expression was normalized to the housekeeping genes ACTB and GAPDH and an interplate calibrator was used to compare results between plates.

Flow cytometry analysis.
CDCs and c-kit pos CPCs were trypsinized at passage 5 to 7 and 50,000-100,000 cells were incubated with fluorochrome-conjugated antibodies (Table S1). Samples were evaluated on a FACSCanto II flow cytometer (BD Biosciences) with at least 10,000 recorded events. Cellular debris and aggregates were gated out based on forward and side scatter. Data were analysed with FACSDiVa software (BD Biosciences).
In vitro proliferation potential. Growth kinetics were determined by calculating cell numbers at every passage to derive expansion curves and PDT throughout the entire expansion process. PDT was computed by linear regression of log2 values of cell number. Only values in the exponential growth phase were used.
In vitro vasculogenic potential. Tube formation assay was performed as described previously 39 . Briefly, 65,000 c-kit pos CPCs or CDCs per well were plated on 24-well plates coated with Matrigel ® basement membrane matrix (BD Biosciences). Tube formation analysis was performed after 6 hours of incubation through phase-contrast mosaic images (Axiovert 200 M imaging microscope, Carl Zeiss). Angiogenesis analyser software (Gilles Carpentier) for ImageJ (National Institutes of Health) was used for morphometric analysis of the obtained images. Additionally, a 3D sprouting assay in Matrigel ® was performed using spheroids as described previously 40 .
Per group, cells of 2 different donors were used and for each donor, 4 different spheroids were imaged. C-kit pos CPC and CDC spheroids containing 5,000 cells were formed in culture medium supplemented with poly(vinyl alcohol) in V-shaped 96-well plates for 96 hours. Formed spheroids were transferred to glass bottom culture plates coated with Matrigel ® . A second layer of Matrigel ® was added to accomplish a 3D structure. The sprouting was observed after 48 hours incubation through phase-contrast images (Confocal laser scanning microscope LSM 800, Carl Zeiss). To visualize the 3D sprouting, a Z-stack was made.
Immunocytochemical analysis. For immunofluorescence staining, cells were fixed with 4% paraformaldehyde (PFA), permeabilized with 0.2% Triton-X100 and incubated with blocking solution (2% BSA) for one hour at RT. Next, cells were incubated overnight at 4 °C with primary antibodies, washed and incubated with secondary antibodies for 1 hour at RT. Secondary antibodies were conjugated with biotin to allow amplification of the signal (Table S2). For amplification, the tyramide signal amplification kit (TSA, Perkin Elmer) was used. Nuclei were stained with DAPI. Images were obtained using an Axiovert 200 M imaging microscope (Carl Zeiss).
Paracrine assay. To compare the production of growth factors and cytokines, CDCs and c-kit pos CPCs were seeded in 6-well culture plates in PBS with 2% FBS for 24 hours. Conditioned medium was collected and the concentrations of VEGF, bFGF, IGF-1, HGF, PlGF and SDF-1 were measured using human-specific ELISA kits (R&D Systems), according to the manufacturer's instructions. In addition, total protein levels were determined by a bicinchoninic acid assay (Pierce BCA Protein Assay Kit, Thermo Scientific), according to the manufacturer's instructions to normalize the production of growth factors and cytokines.
Conditioned medium and exosome purification. CDCs (yCDC, n = 2; aCDC, n = 2) and c-kit pos CPCs (c-kit pos yCPC, n = 3; c-kit pos aCPC, n = 3) were cultured for 48 hours in serum-free medium, before collection conditioned medium. Exosomes were isolated using the ultracentrifugation method. Briefly, cells were pelleted by low speed centrifugation (15 min. at 500 g). The supernatant, containing the exosomes and cell debris, were then centrifuged for 30 min. at 10,000 g to pellet the cell debris. The residual supernatant was then finally ultracentrifuged in 2 different steps (3 hours at 140,000 g, 70 min. at 140,000 g) to obtain the exosome fraction. The purified exosome pellet was dissolved in PBS for further analysis. All centrifugation and ultracentrifugation steps were performed at 4 °C. The nanoparticle analysis technology (Nanosight) was used to measure particle size and concentration. Six technical replicates per exosome sample were performed. The extraction of total RNA of the exosomes was performed using the total exosome RNA and protein isolation kit (Life technologies). The total exosome RNA yield and quality was measured by a spectrophotometer (NanoDrop2000, Thermo scientific). Normalized expression levels of miR-22-3p, miR-132-3p, miR-146a-3p and miR-210-3p were calculated using miR-16 as a housekeeping gene.
In vitro differentiation assays of mono-culture of human stem cells. To induce CM differentiation, c-kit pos CPCs and CDCs were seeded at densities of 5 × 10 3 cells/cm 2 . The c-kit expansion or CGM maintenance medium (CON) was replaced by cardiomyocyte differentiation medium (Millipore) (DIFF) for 12 days with fractional medium changes of 50% every 2-3 days. Cells were stained for sarcomeric alpha-actinin and F-actin to determine CM differentiation. Images were obtained using an Axiovert 200 M imaging microscope (Carl Zeiss) and quantification was done using ImageJ software (National Institutes of Health).
Statistical analysis. All data are presented as mean ± standard error of the mean (SEM) except for the RT-qPCR results, which are presented as geometric mean with 95% confidence interval (CI). Intergroup differences were analysed using a two-tailed unpaired t-tests or one-way ANOVA followed by a Bonferroni post-hoc test or a chi-square test for normally distributed data. Non-normally distributed data were compared using a non-parametric Mann-Whitney test or non-parametric Kruskal-Wallis test followed by a Dunn's post-hoc test.
Normality of the data was tested using a Shapiro-Wilk normality test. Spearman non-parametric correlation was used to analyse linear co-variation. The goodness of fit was expressed as the coefficient of determination R 2 . A probability value of P < 0.05 was considered statistically significant. All analyses were performed using Prism 6.0e software (GraphPad Software). Data availability. The sequencing data generated and analysed during the current study has been made available through the GEO website.