Mouse HSA+ immature cardiomyocytes persist in the adult heart and expand after ischemic injury

The assessment of the regenerative capacity of the heart has been compromised by the lack of surface signatures to characterize cardiomyocytes (CMs). Here, combined multiparametric surface marker analysis with single-cell transcriptional profiling and in vivo transplantation identify the main mouse fetal cardiac populations and their progenitors (PRGs). We found that CMs at different stages of differentiation coexist during development. We identified a population of immature heat stable antigen (HSA)/ cluster of differentiation 24 (CD24)+ CMs that persists throughout life and that, unlike other CM subsets, actively proliferates up to 1 week of age and engrafts cardiac tissue upon transplantation. In the adult heart, a discrete population of HSA/CD24+ CMs appears as mononucleated cells that increase in frequency after infarction. Our work identified cell surface signatures that allow the prospective isolation of CMs at all developmental stages and the detection of a subset of immature CMs throughout life that, although at reduced frequencies, are poised for activation in response to ischemic stimuli. This work opens new perspectives in the understanding and treatment of heart pathologies.


Introduction
The cell types that form the mammalian heart have diverse developmental origins and temporal differentiation [1]. In the mouse, cardiomyocytes (CMs) are initially specified by embryonic day (E) 7.5 from the first set of cardiogenic progenitors (PRGs; first heart field) [2] followed by the incoming cells from the second heart field (SHF) [3]. At E 9.5 (looping-heart

Phenotype assignment to the cellular components in the fetal heart
To resolve the phenotype of the cellular components of the developing heart, we screened by flow cytometry cell suspensions isolated from the 3 heart regions (i.e., At, GV-AVJ-containing the connection of the 4 cavities, the great vessels and the valves-and Vt; Fig 1A) for the expression and relative abundance of 30 surface proteins. The analysis was performed in E 17.5 hearts that have similar structure and cellular components to those observed in the adult. We selected 11 antibodies recognizing surface proteins from which HSA (CD24) had not been previously associated with cardiac cells. Following a sequential gating strategy (Fig 1A and S1 Fig), we identified 13 distinct cell populations, after exclusion of hematopoietic (CD45 + Ter119 + ) cells. Multiparametric analysis of the flow cytometry data by nonlinear dimensionality reduction algorithms ( Fig 1B) created maps in which cells clustered together according to phenotype (tdistributed stochastic neighbor embedding [t-SNE], upper graphs) or were organized into hierarchies of related phenotypes (spanning tree, lower graphs). Intercellular adhesion molecule 1 (ICAM-1) + cells (olive arrow) were predominant in the At, Sca-1 + (blue arrow), and ALCAM + (cyan asterisk) in the GV-AVJ and Thy1 + (green arrow) in the Vt. The remaining subsets were represented in all 3 regions, although at different frequencies ( Fig 1B).
To determine the cell identity of the newly defined populations, we analyzed their transcriptional profiles and anatomical distribution. The expression levels of 31 transcripts affiliated to different cardiac lineages were analyzed in purified cells (20 cells) from each subset. Unsupervised hierarchical clustering and principal component analysis (PCA) grouped the subsets in 7 clusters (Fig 2A and 2B and Table 1). We observed a strong correlation between the clustering by the transcriptional profiles and the subset definition using the cell surface markers, highlighting the validity and the robustness of our approach. Cluster I encompassed CMs identified by the expression of Nkx2-5, Tnnt2, and Des. At CMs (Cluster Ia) expressed At-specific myosin, Myl7 and Myh6, and were phenotypically characterized by the coexpression of HSA, MCAM, and ALCAM. Vt CMs (Cluster Ib) expressing the Vt myosin Myl2 and Myh7, expressed HSA and MCAM but not ALCAM (Fig 2A and 2B). We also analyzed the expression and distribution of the identified proteins in situ ( Fig 3A). HSA + CMs identified by the characteristic striated actinin pattern were found in both chambers (more frequent in the At than in the Vt; Fig 3F). In Cluster II, cells expressed Acta2 together with Myh11-indicative of their SMC affiliation-and corresponded to ALCAM + GV-AVJ cells found in the wall of the great vessels, coexpressing smooth muscle actin (SMA) protein (Figs 2A, 2B and 3D). Cluster III cells expressed Kdr, Flk1, and Tek and comprised PECAM-1 + ECs and EndoCs (Fig 2A and  2B), delineating the blood vessels (Fig 3C and 3D) and the inner surface of the myocardium (EndoC). Cluster IV cells expressed Wt1, Tbx18, S100a4, and Gja1, were identified as ICAM-1 + (Fig 2A and 2B), and were observed in the subepicardial region of both chambers (Fig 3B  and 3E [inset #]), supporting their EPDC identity. Cluster V combined cells exhibiting a FB transcriptional profile (Col1a1, Col3a1, Dcn, Twist1, Ddr2, Vcan, and Fn1) with the expression of GV-AVJ-specific genes Tbx3 together with Isl1, revealing a Sca-1 or HSA phenotype (Fig  2A and 2B). Sca-1 + (PECAM-1 − ) cells were detected at the insertion of great vessels (Fig 3C, � ), whereas HSA + cells (actinin − ) were observed in the EndoC cushion mesenchyme (Fig 3C, � ). Cluster VI corresponded to MCAM + cells expressing the SMC-associated transcript Acta2 together with Tbx20, Snai2, Vim, Fn, and Fap (Fig 2A and 2B). Finally, in Cluster VII, glycoprotein 38 (Gp38) + , PDGFrα + , and Thy1 + cells were grouped by the expression of a different combination of FB-related transcripts (Ddr2, Col3a1, Dcn, Postn, Fn, Vim, Snai2, Tbx20; Fig  2A and 2B). PDGFrα expression was associated with FBs in the myocardium (Fig 3B), whereas Gp38 + cells were detected in the epicardium (Fig 3E, � ). The ensemble of these results validated the phenotyping strategy (Fig 1 and Table 1) that allowed the identification (Figs 2A, 2B and 3G) and the prospective isolation of the major cardiac cell subsets in the fetal heart.

Different stages of CM maturation coexist during development
A CM transcriptional profile (Nkx2-5, Tnnt2, and Des) was associated with membrane HSA expression in different, but related, cell populations (Fig 2) isolated from both heart chambers. To understand the kinetics of the newly defined populations along heart morphogenesis, we performed a similar analysis at earlier developmental stages (E 9.5 and E 13.5; S3A and S3D To investigate whether the phenotypically distinct CM subsets corresponded to different stages of maturation, we included Tnnt [23] and Cav3 [12] in our analysis. Surface Cav3 expression is detected in the adult but is undetectable in most immature CMs [12], whereas Tnnt is widely expressed in the CM lineage [23,24]. The transcriptional profile of the HSA + CMs along time (E 9.5, E 13.5, and E 17.5) revealed that Cav3 starts to be transcribed at E 9.5 (S2D Fig), and the first Cav3 + cells detected by protein staining still express high levels of HSA (Fig 4D), suggesting a lineage relationship between the 2 CM subsets. These results highlight the progressive loss of the surface markers ALCAM, MCAM, and HSA coinciding with the transcriptional expression of Cav3 at E 13.5, whereas at E 17.5 the majority of CMs (Tnnt + ) were HSA − MCAM − ALCAM − Cav3 + (Fig 4A). During maturation, CMs lose proliferative capacity and suffer morphologic alterations leading to increased sarcomere complexity, bigger cell size, and binucleation [9]. Cell-cycle analysis of E 17.5 CM subsets (HSA + and Cav3 + ) showed that HSA + cells were still proliferative (At: 24.4% and 20.5%; and Vt: 35.5% and 25.3%,  cells showing the expression of PECAM-1, ICAM-1, Sca-1, Gp38, Thy1, HSA, PDGFrα, ALCAM, and MCAM (n = 10). (B) t-SNE (upper graphs) and spanning tree (lower graphs) analyses applied to the populations defined in (A). Each point represents a cell, and colors represent surface signatures (see S1 Fig). Marker expression is represented by a color code: red arrows, HSA, MCAM, and ALCAM; cyan arrows, MCAM; green asterisks, PDGFrα; olive arrows, ICAM-1; green arrows, Thy1; blue asterisks, HSA; blue arrow, Sca-1; cyan asterisk, ALCAM; and black arrows, neg. Colors correspond to populations in (A). The number of dots denotes relative size of the populations. Multiple colors in the same node represent coexpression. ALCAM, activated leukocyte cell adhesion molecule; At, atria; CD45, cluster of differentiation 45; E, embryonic day; Gp38, glycoprotein 38; GV-AVJ, great vessels-atrioventricular junction; HSA, heat stable antigen; ICAM-1, intercellular adhesion molecule 1; MCAM, melanoma cell adhesion molecule; neg, negative for all analyzed markers; PDGFrα, platelet derived growth factor receptor alpha; PECAM-1, platelet/endothelial cell adhesion molecule 1; Sca-1, stem cells antigen 1; Ter119, lymphocyte antigen 76 clone TER-119; Thy1, thymus cell antigen 1; t-SNE, t-distributed stochastic neighbor embedding; Vt, ventricles. Our results demonstrate the coexistence of 4 stages of CM maturation along development (E 9.5-E 17.5). Immature HSA + MCAM + ALCAM + CMs are the dominant subset in the At. The expression of these markers is progressively lost such that at E 17.5 only a small fraction of Vt CMs is still HSA + MCAM + ALCAM − , whereas the majority of CMs are negative for all markers, display surface Cav3, and have initiated binucleation (Fig 4A, S3C Fig). At CMs retained the expression of the 3 markers until later embryonic stages than their Vt cardiomyocyte; E, embryonic day; EC, endothelial cell; EPDC, epicardial-derived cell; FB, fibroblast; Gp38, glycoprotein 38; GV-AVJ, great vessels and atrioventricular junction; HSA, heat stable antigen; ICAM-1, intercellular adhesion molecule 1; MCAM, melanoma cell adhesion molecule; PCA, principal component analysis; PDGFrα, platelet derived growth factor receptor alpha; PECAM-1, platelet/endothelial cell adhesion molecule 1; qRT-PCR, quantitative real time polymerase chain reaction; Sca-1, stem cells antigen 1; SMC, smooth muscle cell; Thy1, thymus cell antigen 1; Vt, ventricle.  Because HSA is the last surface protein to be lost before the acquisition of Cav3, we used these 2 markers to discriminate immature and mature CMs, respectively. Interestingly, HSA and Cav3 expression define 2 CM subsets with different proliferative capacity, even after birth. Three 5-ethynyl-2´-deoxyuridine (EdU) injections (P0, P1, and P2) in the neonates labeled 23% of HSA + CMs detected at P2 and 16% at P4, indicating that this CM subset maintains proliferative activity after birth (Fig 4C). By contrast, only 7% of Cav3 + CMs incorporated EdU, demonstrating a lower contribution of the mature CM subset to postnatal heart growth ( Fig  4C). Other non-CM (stromal) cells showed more than 50% EdU incorporation, compatible with their high proliferative activity at this stage ( Fig 4C). The reduced frequency of Cav3 + CMs at P7 detected by flow cytometry (S3D Fig, contour plots) reflects the sensitivity of postnatal CMs to enzymatic digestion. Compatible with this possibility, although most Cav3 + CMs did not incorporate DNA intercalating agents, such as propidium iodide (PI), used in the flow cytometry analysis, they also did not stain (>4%) with Hoechst 33342 that detects DNA in live cells ( Fig 4B). By contrast, more than 70% of HSA + CMs showed levels of DNA after Hoechst 33342 staining similar to live ECs (PECAM-1 + PI − ) or stromal cells (PDGFrα + PI − ). Consistently, we found that Hoechst positive and negative cells yielded a more than 30-fold difference in the amplified cDNA for Hprt, although with similar relative amounts of Tnnt2 transcripts (S4D Fig). These results indicate that different subsets of CM have different sensitivity to enzymatic digestion that bias the analysis of postnatal CMs in cell suspension.
To confirm that HSA discriminates immature CM, purified HSA + cells isolated from E 15.5, P2, and P4 hearts were cultured for up to 1 week. E 15.5 HSA + cells either divided (approximately 20%) or were contractile (approximately 75%) in culture, whereas no proliferation was observed in P2 or P4 cardiac cells, probably due to the rapid differentiation in culture ( Fig 4F, S6A Fig, MS1 and MS2). Seeded P2 and P4 cells adhered and were contractile at a frequency ranging from 1:30 to 1:40 (S6B Fig); by contrast, Cav3 + cells, irrespective of the stage at which they were isolated, did not adhere to gelatin-coated plates (more than 10,000 cells analyzed) and were not viable after 3 days in culture (S6B Fig). All P2 and P4 adherent cells showed contractility (S6B Fig, S3 and MS4). The nonorganized pattern of Tnnt observed on contractile HSA + CMs in culture reflects their immaturity ( Fig 4F) and is a feature also observed during development (Figs 3F and 4E, actinin staining).
To show that immature CMs have the capacity to integrate cardiac tissue in vivo, we used a previously described experimental model [26] Because most Cav3 + cells were not viable (H − ) and therefore did not express GFP in the Ub-GFP model, we isolated Cav3 + H + cells from wild-type (WT) mice and labeled with carboxyfluorescein succinimidyl ester (CFSE) dye that binds to cytoplasmic proteins, resists to fixation, and can be detected after several rounds of division [27]. Seven days after WT E 15.5 Vt tissue implantation, we transplanted either E 15.5 HSA + GFP + CMs from Ub-GFP mice (upper panels) or from WT mice E 15.5 or P1 HSA + and Cav3 + CMs and labeled with CFSE (middle and lower panels, Fig 4G). One week after cell transplantation, the implants injected with HSA + CMs displayed regions of GFP + or CFSE + cells with the characteristic striated pattern of myocytes that also coexpressed Tnnt ( Fig 4G). A quantification of the number of donor cells detected in the explants (cells per section) indicated HSA + CMs from P1 hearts integrate the grafted tissue more efficiently than P1 Cav3 + CMs and that no significant difference is found between E 15.5 and P1 HSA + CMs. Together, these results demonstrated the better efficiency of immature HSA + CMs compared with Cav3 + CMs to engraft cardiac tissue and maintain viability for up to 1 week (Fig 4G and  S8 Fig).
The transcriptional profile of P1 HSA + H + and Cav3 + H + CMs probed for over 40 transcripts (assembled from our previous panel and complemented with transcripts differentially expressed between E 14.5 and P0 [28]) indicated that most cells from both subsets exhibited a similar transcriptional profile (Fig 5A and 5B). This result is compatible with the observation that Cav3 + H + cells express higher levels of HSA than their H − counterparts, suggesting a recent transition to the Cav3 + compartment ( Fig 4D). Five major clusters were identified: (i) cluster 1 Immature cardiomyocytes are found in the adult heart is mostly composed of HSA + H + cells expressing high levels of Acta2, Vim, Tbx5, Nkx2.5, Arnt, Gjc1, Gja1, and Cdkn1a; (ii) cluster 2 has a stronger representation of Cav3 + H + and differed from cluster 1 by lower expression of Acta2 and absence of Cdkn1a expression; (iii) cluster 3 of Cav3 + H + cells is characterized by high levels of Col3a1, Actnb, Rgs4, and Vim and the absence of Myh7; (iv) clusters 4 and 5 are composed of HSA + H + cells characterized by low expression of Actnb, Arnt, Nkx2.5, Tbx5, and Vim and differ from each other by differential expression of Acta2, Myh7, Tbx20, and Egln1. Overall, HSA + CMs expressed significantly lower levels of transcripts for the structural cardiac proteins Gjc1, Vim, Actnb, and Casq1, indicative of their more immature stage (Fig 5C) [9]. No HSA + CMs were shown to express extracellular matrix (ECM) specific transcripts, suggesting that the recently described CM population [28] belongs to the Cav3 + subset (Fig 5A and 5B [arrow]). Remarkably, though, all cells isolated on the basis of the unique expression of HSA, and negative for CD45, Ter119, PECAM-1, and PDGFα, consistently expressed high levels of Tnnt2, Myh6, Gata4, and Des, demonstrating the relevance of this marker to identify immature CMs in the mouse (Fig 5A).
Overall, our results identify distinct stages of CM maturation along development based on surface-marker expression. We defined 2 major CM subsets: an immature subset of mononucleated cells with proliferative capacity (HSA + ) that can give rise to functional CMs both in vitro and in vivo and a mature fraction (Cav3 + ) with increased sarcomere complexity, binucleation, and decreased proliferative capacity.
Immature CMs persist in adulthood and increase after injury. To determine whether immature HSA + CMs persist in adulthood, we analyzed adult heart cell suspensions with the antibody panel defined above. In P21 heart cell suspensions, we found 1.76% HSA + CMs that expressed low levels of Cav3 and 30% of which are also H + , whereas only 0.8% of Cav3 + CMs are H + (S9B Fig). CM-specific transcripts (Tnnt2, Myl7, and Myh6; Fig 6B) were present in adult HSA + cells, more frequent in the At (At: 1.7% and Vt: 0.18%; Fig 6A), but also in HSA − cells expressing surface Cav3 (Fig 6A and S9C Fig). Imaging flow cytometry analysis further confirmed the presence of a discrete subset of HSA + Actinin + CMs in adult hearts (0.6% ± 0.33%, n = 3) restricted to the subset of mononucleated CMs smaller (in area and length) than HSA − CMs (Fig 6D). The HSA + subset was considered as the most immature adult CMs because they shared cytological and phenotypic properties with embryonic CMs, in culture adhered to gelatin-coated plates, survived for a few days, and expressed Tnnt but failed to divide or to contract (S7B Fig), and the majority also coexpressed the mature marker Cav3 (Fig 6A). Two stromal populations were also discriminated in the adult: (i) a PDGFrα + Sca-1 + Thy1 low subset of FBs compatible with a previously described stromal population [29] and (ii) ICAM-1 + Gp38 + Thy1 + cells, expressing Gata4, Tek, Dcn, Twist1, and Tbx18 and located in the subepicardial region (S9A, S9C and S9D Fig). c-kit expression previously associated with CM PRGs was exclusively found in adult PECAM-1 + ECs, and transcripts were also detected in ICAM-1 + sub-EpiCs (S9C Fig). The detection of an immature CM subset in the adult prompted us to investigate its frequency in the diseased heart. We found HSA expression largely circumscribed to the non-CM compartment in sham-operated hearts, although rare HSA + Actinin + CMs were also detected Immature cardiomyocytes are found in the adult heart (outlined by laminin expression, Fig 6C). Seven days after MI, and in spite of HSA expression associated to the upsurge of the hematopoietic and ECs (S9E Fig), we detected a 3-fold increase in the frequency of HSA + small-round and large-striated mononucleated CMs (# and � , respectively; 4.7% ± 3.6%) compared with sham-operated (1.7% ± 1.1%; Fig 6C). HSA + CMs were found in the peri-infarcted region as shown in the lower magnification image (Fig 6C). Moreover, a small percentage of HSA + CMs (i.e., approximately 1 per section) were in cycle (Ki67 expression) after MI (Fig 6C). A similar increase in HSA + CMs was evidenced after MI by imaging flow cytometry analysis of prefixed cells (1.8% ± 0.3%) compared with sham-operated hearts (0.6% ± 0.15%; Fig 6E).
Our results show that the cell surface signatures defined for the embryonic heart are suitable to isolate an affiliated adult population. Importantly, we identified a small subset of HSAexpressing adult CMs that, like their embryonic counterparts, are mononucleated, express low levels of Cav3, and exhibit a higher probability than Cav3 + cells to proliferate and increase in frequency after MI.

Discussion
The mouse heart is able to regenerate during the first days of life by proliferation of pre-existing CMs. This capacity is lost after 1 week of postnatal life, the time from which wounded tissue is replaced by a nonfunctional scar [30]. The loss of CM mitotic activity has been attributed to the binucleation process and the complete maturation of CMs that occur after birth [10,31]. However, recent reports showed that CM replacement and cell division could occur, although at low rate [21,22], raising the possibility that a subset of CMs in the adult might undergo mitosis. Specific surface markers need to be identified to allow the isolation and characterization of the low frequency of dividing CMs. We show here that, although HSA is not a marker for proliferating cells, alone it identifies a neonatal CM compartment that retains higher proliferative capacity, persists in the adult, and expands after MI. HSA has been associated with proliferating cells in several tissues (e.g., skeletal muscle PRG cells [32,33]), as well as in several cancer types [34]; however, its function is unknown. Although HSA is detected in all CMs in E 9.5 (S5C Fig), the constitutive inactivation of HSA does not appear to have a strong impact in heart development, because these mice are viable and fertile. It has, however, been reported that the fraction of born homozygous pups is lower than predicted, raising a possibility of a low penetrance lethal effect [32].
Expression of Isl1, Gata4, Mef2c, and Tbx20 is important for CM commitment and initial stages of differentiation [35][36][37], and accordingly they were used as indicative of the CM lineage. Consistent with previous reports [29], we found some of these transcripts also expressed during the development of stromal cells (SMCs and FBs), suggesting that, in the heart, they participate in the development of different lineages. These findings demonstrate that CMs can only be unambiguously identified by the combined expression of transcription factors, transcripts codifying structural and contractile proteins, and by the absence of stromal-or ECassociated markers.
Several studies identified CM PRGs based on the expression of Sca-1 and c-kit [14][15][16] cellsurface proteins, which in our work were not found within the CM compartment. c-kit was only expressed in ECs (PECAM-1 + ), as recently shown also by others [38], and in Thy1 + PDGFrα + FBs. Sca-1 expression was detected in a fraction of ECs and in a population of cells in the atrioventricular canal in the fetal heart and in the interstitium of adult myocardium [15]. Their spatial pattern and transcriptional profile are compatible with an FB lineage affiliation, supported by the description of Sca-1-expressing cells exhibiting a paracrine role in angiogenic stimulation after MI [15,39]. At E 9.5, we found an HSA, ALCAM, and PDGFrα expressing population that was highly proliferative and expressed Nkx2-5 and Tnnt2 together with Isl1 and Tbx5, suggesting they represent CM PRGs (S2D Fig) [40,41]. These cells sharply decrease in frequency, they are only detected in the At after E13.5 of development, and become undetectable after birth, a finding that is not compatible with the persistence of CM PRGs in the postnatal heart.
We identified HSA, so far never associated to heart development or maturation, as a transversal marker of immature CMs throughout life. Our analysis showed a continuum in CM maturation, which is an asynchronous and apparently stochastic process that starts during development and can be prospectively identified by the expression of distinct surface markers (Fig 7). Immature CMs can be identified by a unique phenotype, i.e., HSA + MCAM + ALCAM +-Cav3 − ; they progressively lose ALCAM and MCAM expression to become HSA + only. HSA + CMs decrease in frequency between E 17.5 and P7 but are the only CM subset that actively proliferates up to P7; they have spontaneously contractile properties in culture and are consistently mononucleated. Isolated from E 15 and P1 hearts, HSA + CM engrafted cardiac tissue transplanted in the ear pinna of adult mice with a better efficiency than that of Cav3 + CMs.
The first signs of CM binucleation are observed around E 17.5 of development in Cav3 + CMs, but never in immature CMs (Tnnt + Cav3 − ) that decrease in frequency, coinciding with an increase in noncycling and binucleated Cav3 + CMs that compose the myocardium in adulthood. These results are in agreement with alterations endured by CMs during the first week of life, which encompass a transition from hyperplasia to hypertrophy and terminal differentiation [9,10]. This is also consistent with the higher expression of transcripts specific for cardiac structural proteins found at the single-cell level of Cav3 + postnatal compared with HSA + CMs. By contrast, Col3a1 that characterizes a recently described population of ECM expressing CMs detected after E 14.5 [28] was found in the present study in Cav3 + cells that we likewise also first detect after E13.5.
Interestingly, a higher frequency of immature HSA + CMs (>70%) maintain cell integrity, after enzymatic digestion, compared with Cav3 + CMs (<4%). The observation that different subsets of postnatal CMs have different degrees of cell viability after enzymatic digestion is an important piece of information that impacts experiments performed with CM cell suspensions. Immature cardiomyocytes are found in the adult heart A subset (<1%) of adult CMs displayed HSA and remained mononucleated, thus resembling embryonic CM, despite expressing Cav3 at the cell surface. Postnatal and adult HSA + cells share the same morphological features of immature CMs (small, round shaped and mononucleated). However, adult HSA + CMs did not acquire spontaneous contractility in culture, indicating that they are less resistant than perinatal immature CMs to dissociation and thus more similar to Cav3 + CMs. Our findings are in line with previous reports showing low rates of cell division (0.76% [21] and 0.3% to 1% [42] per year) in small, mononucleated, and diploid CMs in the adult heart.
Foreseeing its therapeutic relevance, we tested whether HSA + immature CMs could respond to a pathological challenge. We observed an upsurge in the frequency of adult mononucleated HSA + CMs 7 days after MI (from 0.6% to 1.8%). This relative increase in HSA + CMs can be explained by proliferation, detected at very low frequency in our analysis, by increased resistance of HSA + CMs to hypoxia, by re-expression of HSA or by any combination of the above. In the developing heart, low oxygen tension is found in the compact myocardial layer [43] precisely where immature HSA + CMs were found in this study. An adult CM subset protected from oxidative stress in hypoxic niches and exhibiting low proliferative activity upon injury has been recently identified. Similar to HSA + CMs, these cells were mononucleated, small sized, and represented around 1% of the adult myocardium [42]. Although we have not found differences in the hypoxia regulated transcripts at steady state, we cannot rule out that both cell types might respond differently to ischemia.
Although HSA + immature CMs do not proliferate sufficiently to regenerate the myocardium, they might account for the low CM turnover rate previously described in the adult [21,22,44] and might be more amenable than binucleated CMs to respond to mitotic stimuli. Importantly, using the strategy herein described, CMs at different maturation stages can now be prospectively isolated as viable cells from the adult heart, enabling further mechanistic studies.

Ethics statement
All animal manipulations at i3S were approved by the Animal Ethics Committee and Direcção-Geral de Veterinária-DGAV; at Pasteur Institute, animal manipulations were approved by the Ethics Committee and by the French Agriculture ministry according to the Ethic Chart and the European Parliament Directive 2010/63/EU at both institutes.

Mice
C57BL/6 mice (Charles River) 6 to 8 weeks old or timed pregnant females were used. Timed pregnancies were generated after overnight mating. The following morning, females with vaginal plug were considered to be at E 0.5. Ub-GFP mice used for transplantation experiments were a kind gift from P. Bousso (Pasteur Institute) [45].

Mouse model of MI
MI was experimentally induced by permanent ligation of the left coronary artery as previously described by Nascimento and colleagues [46], and samples were analyzed 7 days after injury.

Transplantation of Ub-GFP or CFSE + cells in embryonic cardiac implants (ear-pinna model)
As shown in S8A Fig, E 15.5 cardiac Vt from WT embryos were dissected and grafted in the ear pinna of recipient adult WT mice, under anesthesia, as previously described by Ardehali and colleagues [26]. Seven days later, hearts from Ub-GFP (E 15.5) or from WT (E 15.5 and P1) mice were dissociated, and HSA + immature CMs, Cav3 + CMs, or PDGFrα + FBs (stromal cells) were sorted and directly injected into visible beating implants (10,000 cells per implant). In all experiments using P1 CMs, cells were additionally incubated for 30 minutes at 37˚C with Hoechst 33342 (0.2 μg/ml, Molecular Probes H3570) and sorted as positive for DNA content (H + ). After isolation, WT cells (E 15.5 or P1) were labeled with CFSE (Molecular Probes, C1157) as previously described by Tario and colleagues [27]. One week later, the implants were collected, and tissue was processed for immunofluorescence as described below.

Isolation of live cardiac cells
Embryonic hearts were collected under a stereomicroscope, and the 3 anatomic heart structures (At, GV-AVJ, and Vt) were microdissected. Heart tissue was minced into 1 mm 3 fragments and incubated for 15 minutes at 37˚C in the enzymatic solution: for E 13.5 and E 17.5 hearts, 0.2 mg/mL collagenase (Sigma-Aldrich) in Hank's Balanced Salt Solution with calcium and magnesium (HBSS +/+ , Invitrogen); for E 9.5 hearts, 0.1 mg/mL collagenase in HBSS +/+ ; for postnatal hearts 20 mM 2,3-Butanedione monoxime (BDM; Sigma-Aldrich) was added in all isolation steps to DPBS without Ca 2+ and Mg 2+ (DPBS −/− , Invitrogen); and for adult hearts, 0.2 mg/mL collagenase with 20 mM BDM (Sigma-Aldrich) and with 60 U/mL DNase I (Roche, Switzerland). At the end of each round of digestion, tissue fragments were resuspended using a P1000 pipette (approximately 20 times). The remaining tissue was allowed to sediment, and the supernatant was collected in a tube containing the same volume of 10% FCS (Life Technologies)-HBSS −/− and kept on ice while the digestion protocol continued. Digestion was repeated until no macroscopic tissue was detected. After digestion, cell suspensions were centrifuged 10 minutes, 290g at 4˚C, resuspended in 1% FCS HBSS −/− , and filtered with a 70 μm mesh strainer (Fisher Scientific).

Isolation of fixed CMs
Fixed CMs were isolated as described by Mollova and colleagues [47] with some alterations. E 13.5, E 17.5, P7, adult, and injured (MI or sham-operated) hearts were collected, washed in PBS (to remove blood, Invitrogen), minced into 2 mm 3 pieces, flash frozen in liquid nitrogen, and stored at −80˚C. For cell isolation, tissue pieces were fixed in 4% paraformaldehyde (Electron Microscopy Sciences) at room temperature for 2 hours, washed in PBS for 5 minutes, and digested with 3 mg/ml collagenase type II (Worthington) in HBSS on a rotator at 37˚C until no macroscopic tissue was detected. Enzyme activity was blocked with 10% FBS-HBSS (Life Technologies). Cell suspensions were filtered through a 100-μm cell strainer (Fisher Scientific).

Flow cytometry, cell sorting, and imaging flow cytometry
Heart-cell suspensions were stained with (conjugated or nonconjugated) antibodies (20 minutes, 4˚C in the dark) followed by incubation with conjugated streptavidin (10 minutes, 4˚C in the dark). Whenever using a nonconjugated antibody, a sequential incubation with a secondary antibody was performed for 15 minutes at 4˚C in the dark (see S1 Table for the antibodies list). PI (1μg/ml) was used to exclude dead cells. In designated experiments of postnatal hearts, cells were incubated with Hoechst 33342 (0.2 μg/ml, Molecular Probes H3570) in 2% FBS-HBSS for 30 minutes at 37˚C. Intracellular proteins (Ki67 and troponin) detection was performed after surface staining, and fixation and permeabilization with the Foxp3/Transcription Factor Staining Buffer Set (eBioscience, USA). DAPI (1/10000, Molecular Probes) was used to stain DNA in fixed cells (5 minutes at 4˚C). EdU staining was also performed after surface staining, fixation, and permeabilization with the Click-it EdU flow cytometry assay kit (Molecular Probes) following the manufacturer's procedures. Flow cytometry data were acquired in a BD FACSCanto II (BD Bioscience), BD LSRFortessa (BD Bioscience), and Sony SP6800 analyzer (Sony, Japan) and analyzed with the FlowJo version 10.0.8 (BD Bioscience), Kaluza 1.5 (Beckman Coulter), or R version 3.2.4 software (R foundation).
Cells were sorted in a BD FACSAria III directly into 96-well plates loaded with RT-STA Reaction mix (according to the manufacturer's procedures, CellsDirect One-Step qRT-PCR Kit, Invitrogen) and 0.2× specific TaqMan Assay mix (see S3 Table for  Fixed CMs were resuspended in BD Cytofix/Cytoperm Fixation/Permeabilization Kit (BD Biosciences), permeabilized in BD Perm/Wash buffer for 15 minutes, incubated with primary antibodies for 2 hours at 4˚C, and incubated with Alexa Fluor-conjugated secondary antibodies for 30 minutes at 4˚C. Prior to acquisition on ImageStream, nuclei were stained with 20 μM DRAQ5 (Biostatus, UK) and filtered with 100-μm cell strainer (Fisher Scientific). Data acquisition was performed using an Amnis ImageStreamX cytometer (Luminex). Files were collected with a cell classifier applied to the bright-field (BF) channel to capture events larger than 20 μm and included BF, FITC, PECy7, and DRAQ5 images. At least 30,000 cells were analyzed for each sample, and all images were captured with the 40× objective. Data analysis was performed with IDEAS software (version 6.0, Luminex). For each sample, only intact CMs, selected based on Actinin and DRAQ5 signal intensity, were considered for subsequent analysis. For the morphometric analysis, we applied a morphology mask to the BF channel, and for assessment of the number of nuclei per cell, we used the DRAQ5 images.
Culture and live-cell imaging E 15.5, P2, P4, and adult cardiac cells were isolated as above, and HSA + cells were sorted following the gating strategy in S1A Fig. For the neonatal and adult cells, 20 mM BDM (Sigma-Aldrich) was added throughout the isolation procedure [48]. HSA + cells were plated in 0.1% collagen (Life Technologies) for E 15.5 or in fibronectin/gelatin coated ibidi plates for postnatal cells, and cultured for 1 week in high-glucose Iscove's Modified Dulbecco's media (Life Technologies) supplemented with 20% FBS, 1× penicillin/streptomycin (Life Technologies, USA), 1× L-glutamine (Life Technologies), 50 μg/mL ascorbic acid, and 1.5 × 10 −4 M 1-thioglycerol (Sigma-Aldrich), as previously described by Wu and colleagues [4]. Adult cardiac cells were incubated at 37˚C in 3% O 2 . Live-cell imaging was performed on a temperature-controlled Zeiss Axiovert 200M microscope equipped with a CoolSnap HQ (Roper) camera (Zeiss, Germany). Sample position was controlled by an X-Y motorized stage, and images were acquired every 15 minutes using an A-Plan 20×/0.30 objective for 48 hours.

Histological processing and immunofluorescence staining
Embryonic and adult (MI and sham-operated) hearts were fixed in 0.2% paraformaldehyde (Merck) overnight at 4˚C, dehydrated in a sucrose gradient (4% followed of 15%), embedded in gelatin, and frozen. Tissue cryo-sections (4 μm thick) were blocked with either 4% FBS-1% BSA blocking solution or Vector M.O.M. basic kit (Vector Laboratories), depending on the specific conditions detailed in S2 Table. Tissue sections were incubated with primary antibodies overnight at 4˚C, followed by 1-hour incubation with Alexa Fluor-conjugated secondary antibodies (see S2 Table for the antibodies list; Invitrogen). Slides were mounted, and nuclei were counterstained with aqueous mounting medium with DAPI (Vector Laboratories). Representative high-resolution images were acquired for each heart structure (At, GV-AVJ, and Vt) at 40× magnification in a confocal microscope (Leica SP5II, Leica, Germany). Wholeheart acquisitions were obtained using the high-content imaging system (IN Cell Analyzer 2000, GE Healthcare).
Isolated fixed CMs were resuspended in 10% FBS-PBS and spun onto superfrost slides in a cytocentrifuge (ThermoFisher). Cytospins were incubated with primary antibodies overnight at 4˚C, followed by 1-hour incubation with Alexa Fluor-conjugated secondary antibodies. Acquired images were edited and quantified using the Image J version 1.51d software (NIH).

Gene expression analysis
Sorted cells in RT-STA Reaction mix from the CellsDirectTM One-Step qRT-PCR Kit (Life Technologies) were kept at −80˚C at least overnight before reverse transcription and specific target pre-amplification (20 cycles for single cells and 18 cycles for 20 cells). Pre-amplified samples were subjected to qRT-PCR (see S3 Table for Taqman assays list, Applied Biosystems) as previously described by Chea and colleagues [49].

Bioinformatic analysis
Flow cytometry data analysis was performed in FCS files of live CD45 − Ter119 − CD31 − cell fraction using R package flowCore from R version 3.2.4 revised (2016-03-16 r70336) and the interface R Studio version 0.99.467 (R foundation) [50]. Subsequently, gating, as described in Fig 1  and in S1 Fig, was used to define each population. Map clustering of the flow cytometry data was performed using custom R scripts from R package t-SNE to dimensionality reduction − t-SNE (R foundation) [51] and Bioconductor.org package flowSOM to visualize Self-Organizing and Minimal Spanning Trees (Spanning Trees, R foundation) [52,53].
Gene expression raw data (BioMark Fluidigm, Applied Biosystems) of sorted cells at the population level was normalized with HPRT, and data are presented in 2 −ΔCt . Single-cell gene expression analysis was performed in cells that expressed at least 1 of 3 housekeeping genes (Hprt, Gapdh, or Actb), and Ct values were used to the following analysis. A Ct value of 21 was the maximum value considered as expressed gene, and the background (i.e., nondetected) Ct value was 38. qRT-PCR data were processed with the QLUCORE (Qlucore AB 2008-2015, Sweden) software, and pheatmap package (version 1.0.10) R (R version 3.5.0, R foundation) was displayed in uncentered Pearson's correlation unsupervised hierarchical clustering and PCA either for surface phenotype or transcripts.

Statistical analysis
All results are shown as mean ± SD. Statistical significance was determined using the Student t test, except when comparing the frequency of HSA + CMs from E 13.5 to P7 (one-way ANOVA followed by Tukey test). The statistical analysis of the data was performed using Sig-maPlot software (p < 0.05 was considered statistically significant, R foundation) or QLUCORE (Qlucore AB 2008-2015, Sweden) software for the multidimensional analysis of multiplex qRT-PCR (two-way ANOVA, p = 0.007, q = 0.01).