We used Wt1^GFPCre and the Cre-activated reporters Rosa26^fsLz15 and Z/Red ¹⁶ to analyze the fate of Wt1-expressing cells in the heart. Upon Cre-mediated recombination, these reporters heritably express ®-galactosidase (®-gal) or red fluorescent protein (RFP), respectively. Using two different reporters minimized potential artefacts related to unanticipated behavior of Cre-activated reporters, or to false-positive immunostaining. While Wt1 and GFPCre expression were confined to the epicardium, descendants of Wt1-expressing progenitors (hereafter called Wt1-derived cells), marked by β-gal, were found in a mosaic pattern throughout the myocardium (Fig. 2a). Consistent with prior reports showing that epicardially-derived mesenchyme predominately differentiates into SMCs in mammals, most Wt1-derived cells adopted a SMC fate, and a minority differentiated into ECs (Suppl. Fig. 3) ^8,9.

Remarkably, we found that some Wt1-derived cells differentiated into cardiomyocytes (CMs) during normal heart development, as demonstrated by co-expression of lineage tracers and CM markers cardiac troponin T (Tnnt2) and sarcomeric -actinin (Actn1) (Fig. 2b-d). The Wt1-derived CMs also expressed cardiac transcription factors Gata4 and Nkx2-5 (Suppl. Fig. 4). Wt1-derived CMs were located in the myocardium of all four cardiac chambers and in the interventricular septum, constituting 7-10% of CMs in ventricles and 18% in atria (Suppl. Fig. 5). We further confirmed co-expression of CM and lineage tracers in isolated cells by immunostaining of dissociated fetal heart cultures (Fig. 2e-f). 4% of CMs in dissociated fetal heart cultures were Wt1-derived (Suppl. Fig. 5), comparable to the frequency observed in tissue sections.

To determine if Wt1-derived CMs had functional properties of CMs, we analyzed dissociated cells from Wt1^GFPCre/+; Z/Red fetal hearts. A subset of red fluorescent cells exhibited spontaneous contractile activity (Fig. 2g; Suppl. Movie 1). In addition, these contracting RFP⁺ cells exhibited calcium oscillations with kinetics, amplitude, and frequency characteristic of CMs (Fig. 2h-j; Suppl. Movie 2). Also characteristic of CMs were calcium sparks preceding calcium waves (Fig. 2h1) and caffeine augmentation of calcium transient amplitude (Fig. 2j), consistent with calcium release from CM ryanodine receptors. Calcium transients of RFP⁺ cells were indistinguishable from and synchronous with adjacent RFP^- cells (Fig. 2i), suggesting electrical coupling between Wt1-derived and non-Wt1-derived CMs. Consistent with this finding, the gap junction protein connexin 43 (Cx43) localized to the membrane between Wt1-derived and neighboring CMs (Suppl. Fig. 6a). A similar pattern of Cx43 expression was observed in Wt1-derived CMs in tissue sections of E15.5 hearts (Suppl. Fig. 6b), suggesting that Wt1-derived CMs are also electrically coupled to other CMs in vivo. Collectively, these data indicate that during heart development a subset of Wt1-expressing cells differentiate into CMs.

To further characterize the Wt1-expressing CM precursors, we utilized several independent methods to control the temporal and spatial window during which these precursors were labeled. To temporally regulate Cre-labeling, we knocked a cDNA encoding a Cre-modified estrogen ligand binding domain (CreERT2) into the Wt1 locus (Suppl. Fig. 7). CreERT2 fusion protein recombinase activity requires tamoxifen ¹⁷. Maternal injection of tamoxifen at E10.5 and E11.5 induced Cre activity and resulted in ®-gal expression within Wt1^CreERT2; Rosa26^fsLz myocardium (Fig. 3a-e), while Wt1^CreERT2 did not activate Rosa26^fsLz in the absence of tamoxifen (Fig. 3b and Suppl. Fig. 7). The frequency of Wt1^CreERT2-labeled cells in epicardium and myocardium (Suppl. Fig. 7-8) was reduced compared to constitutive labeling by Wt1^GFPCre, likely due to inefficient CreERT2 activation by tamoxifen concentrations compatible with maintenance of pregnancy. Co-staining for differentiation markers showed that β-gal⁺ cells differentiated into CM, EC, and SMC lineages (Fig. 3c-e). The distribution of labeled cells between these lineages was comparable between pulse and constitutive labeling approaches. We verified co-expression of CM markers and pulse-labeled lineage tracers at the single cell level by staining cardiomyocytes dissociated from E16.5 Wt1^CreERT2/+; Z/Red hearts, pulsed with tamoxifen at E10.5 and E11.5 (Fig. 3f-h). We consistently observed these pulse-labeled CMs, although the frequency (0.02 ± 0.01%) was notably less than with constitutive labeling with Wt1^GFPCre. Within the window of the tamoxifen pulse, cardiac Wt1 expression was confined to the epicardium (Fig. 1). Based on these data, we conclude that a subset of epicardial cells expressing Wt1 differentiate into CMs.

To further delimit the location of Wt1-expressing cells that differentiate into CMs, we microdissected E11.5 Wt1^GFPCre/+; Z/Red fetal hearts. Serial enzymatic digestion of intact hearts yielded epicardial cells preferentially in the early fractions, due to their location on the exterior of the heart. Early (epicardial) and late (negative control) digestion fractions were sorted for GFP fluorescence, yielding a population enriched for active GFP expression (Fig. 3i). These GFP⁺ cells were plated on either mitotically inactivated cardiac feeders or untreated tissue culture dishes (Fig. 3j-k). In both conditions a subset of the sorted Wt1^GFPCre/+; Z/Red cells differentiated into CMs, identified by co-expression of the RFP lineage tracer and CM markers (Fig. 3j-k). These data provide further evidence that a subset of heart cells actively expressing Wt1, confined within epicardium at E10-11.5 (Fig. 1), differentiated into cardiomyocytes.

We obtained additional independent evidence that epicardial cells differentiate into cardiomyocytes by selective dye labeling of epicardium in E11.5 explanted hearts (Fig. 3l). Explanted hearts were briefly incubated in culture media containing the dye CMFDA and then place in culture media without dye. This resulted in selective labeling of epicardium, as confirmed in sections of hearts fixed immediately after CMFDA incubation (Fig. 3l, 0 hours). After two days of explant culture, labeled epicardial cells were found within myocardium, and a subset expressed the CM markers Nkx2-5 (Fig. 3l, 48 hours). Presence of dye and the CM markers Nkx2-5, Tnnt2, and Actn1 within the same cell was further demonstrated in single cells isolated by dissociating heart explants two days after labeling (Fig. 3m). Collectively, these data indicate that precursors actively expressing Wt1 within E10.5-E11.5 epicardium differentiate into CMs.

Reported cardiac precursors derive from multipotent Isl1⁺/Nkx2-5⁺ progenitors ^1-4. We used Cre-based lineage tracing to ask if Wt1⁺ PE cells are related to these progenitors, or represent a different cardiogenic lineage. Using an Nkx2-5^IRES-Cre knockin allele ¹⁸, we found that Nkx2-5-driven Cre activated Rosa26^fsLz in a subset of PE cells (Fig. 4a-b), suggesting descent of Wt1⁺ PE cells from Nkx2-5-expressing cells.

We independently corroborated this result using a different Nkx2-5 knockin allele, Nkx2-5^{Cre 19}, and a novel Cre-activated reporter, Gata4^flap. Gata4 is expressed in CM, SMC, and EC compartments of the myocardium, as well as in PE (Suppl. Fig. 9) ^20-22. Therefore, within this domain endogenous Gata4 regulatory elements can be used to drive expression of a Cre-dependent reporter gene, alkaline phosphatase (AP). We generated such a reporter, Gata4^flap (Suppl. Fig. 10). In the absence of Cre, Gata4^flap did not express AP (Suppl. Fig. 10). In the presence of well-characterized Cre transgenes, Gata4^flap expressed AP in patterns consistent with the expected sites of Cre activity (cardiac troponin Cre (cTNTCre) and myosin heavy chain α-Cre (MHCαCre), myocardium; or Tie2Cre, endothelium; Suppl. Fig. 10). Quantitative analysis revealed that MHCαCre activated Gata4^flap in a greater percentage of cardiomyocytes than Rosa26^{fsLz (}Gata4^flap 93 ± 3%, versus Rosa26^fsLz 72 ± 6%, p < 0.005, n=4), suggesting that Gata4^flap has greater sensitivity to Cre recombination than Rosa26^fsLz. Therefore, we asked if Gata4^flap would show a greater contribution of Nkx2-5⁺ cells to PE than suggested by Nkx2-5^IRES-Cre; Rosa26^fsLz.

Gata4^flap demonstrated a robust contribution of Nkx2-5-expressing progenitors to PE (asterisk, Fig. 4c). The Nkx2-5^Cre-labeled PE cells expressed Wt1 (Fig. 4d), indicating that Wt1⁺ cells in PE are derived from Nkx2-5-expressing precursors. Gata4^flap also showed a robust contribution of Isl1-expressing precursors to the Wt1⁺ cells in PE (Suppl. Fig. 11b). Supporting this finding, at E8.0 Wt1 and Isl1 were expressed in adjacent regions, and a subset of cells were positive for both markers (Suppl. Fig. 12a-b).

While Wt1⁺ cells in PE were labeled by Nkx2-5-driven Cre, they did not actively co-express Nkx2-5 at E9.5 (Fig. 4e). In E8.0 embryos, Nkx2-5 and Wt1 were expressed in adjacent cells, but were not co-expressed (Suppl. Fig. 12c), suggesting that Nkx2-5 and Wt1 are expressed sequentially, or transiently co-expressed. To further investigate the relationship of Nkx2-5 and Wt1 expression, we studied the expression of Wt1 in Nkx2-5⁺ cells during embryoid body differentiation of embryonic stem (ES) cells. Using transgenic Nkx2-5^gfp ES cells ⁴ and FACS, we isolated Nkx2-5-expressing cells at several time points during embryoid body differentiation. Wt1 was transiently upregulated in Nkx2-5⁺ cells during ES cell differentiation (Fig. 4f). This result was specific, as we did not detect significant Wt1 expression in parallel experiments with Mef2c-AHF-GFP ES cells ²³ (data not shown). In embryos, the activity domain of the Mef2c-AHF enhancer (on in anterior heart field, off in PE) did not overlap with the Wt1 expression domain ²⁴ (yellow arrowhead, Fig. 4e). Collectively, these data suggest that Nkx2-5 and Wt1 are sequentially expressed, or transiently co-expressed, in a subset of PE precursors.

We have shown that Wt1⁺ PE/epicardial cells contribute to the CM lineage during normal heart development (Suppl. Fig. 1). Wt1⁺ cells located on the heart at E10.5-E11.5 differentiate into functional CMs. Although differentiation of PE cells into CMs was previously noted in vitro ²⁵, prior fate-mapping studies of PE cells, using retroviral labels in chick or transgene labels in mice, did not describe PE contribution to the CM lineage in vivo ^8-10,12,13. This may be attributable to differences in methodology, species, or domains of transgene activity. Consistent with the capacity of Wt1-expressing cells to differentiate into CMs, Wt1⁺ PE/epicardial cells are derived from progenitors that express Nkx2-5 and Isl1, suggesting that they share a common developmental origin with previously described multipotent cardiogenic progenitors ^1,2. These experiments identify a previously unrecognized CM progenitor population in the developing heart that may be of use for cardiac regeneration or repair.