Regulation of multipotent cardiac progenitor cell (CPC) expansion and subsequent differentiation into cardiomyocytes, smooth muscle or endothelial cells is a fundamental aspect of basic cardiovascular biology and cardiac regenerative medicine. However, the mechanisms governing these decisions remain unclear. Here, we show that Wnt/β-catenin signalling, which promotes expansion of CPCs1,2,3, is negatively regulated by Notch1-mediated control of phosphorylated β-catenin accumulation within CPCs, and that Notch1 activity in CPCs is required for their differentiation. Notch1 positively, and β-catenin negatively, regulated expression of the cardiac transcription factors, Isl1, Myocd and Smyd1. Surprisingly, disruption of Isl1, normally expressed transiently in CPCs before their differentiation4, resulted in expansion of CPCs in vivo and in an embryonic stem (ES) cell system. Furthermore, Isl1 was required for CPC differentiation into cardiomyocyte and smooth muscle cells, but not endothelial cells. These findings reveal a regulatory network controlling CPC expansion and cell fate that involves unanticipated functions of β-catenin, Notch1 and Isl1 that may be leveraged for regenerative approaches involving CPCs.
Heart malformation is the most frequent form of birth defects in humans, and heart disease remains the leading cause of adult mortality in the developed world, largely because of the limited regenerative capacity of the heart. Recent advances have provided insights into potential therapies based on multipotent CPCs. These cells can be isolated from early embryos or ES cells and cultured to differentiate into numerous cardiac cell types4,5,6,7,8,9,10,11,12. For example, Nkx2.5+, Flk1+ or Isl1+ CPCs purified from embryoid bodies can each give rise to cardiomyocyte, endothelial and smooth muscle cell lineages7,8,10,12.
Nkx2.5 is an ancient cardiac gene activated in CPCs of early embryos13. Nkx2.5+ cells and their progeny populate the precardiac mesoderm located dorsal to the cardiac region and the developing heart tube in vivo14. Isolated Nkx2.5+ cells differentiate spontaneously into distinct cardiac cell lineages including cardiomyocytes, smooth muscle cells and endothelial cells in vitro7,12. These cardiac cell lineages can also be generated from cells expressing Flk1, a marker of the primitive streak in early embryogenesis10, or Isl1, a CPC marker8,15. All of these CPCs show overlapping expression patterns in precardiac mesodermal cells in vivo8 and have similar differentiation potential in vitro7,8,10,12 suggesting that they comprise a similar CPC population. Although these multipotent CPCs hold great potential for cardiac repair, the mechanisms that regulate their self-renewal, expansion and differentiation remain unknown.
We and others have reported that canonical Wnt signalling is an important regulator of Nkx2.5+ and Isl1+ CPCs and is responsible for their expansion in vivo and in vitro1,2,3. In precardiac mesoderm, inactivation of β-catenin, the transcriptional mediator of canonical Wnt signalling, resulted in nearly complete loss of Isl1 cells that contribute to the right ventricle2. Conversely, stabilization of β-catenin in the same cells led to expansion in the number of CPCs2 in vivo, whereas Wnt/β-catenin signalling promoted renewal of CPCs isolated from ES cells2,3. Notch signalling reciprocally affects Wnt signals in many contexts16 and is thought to inhibit cardiac differentiation17,18, although its function in CPCs in vivo is unknown. Ultimately, these and other early signals must be integrated with a network of transcriptional regulators that influence CPCs.
To examine the CPC-autonomous role of Notch1 signalling in vivo, we deleted Notch1 in precardiac mesodermal progenitors by crossing Notch1flox mice19 with mice containing Cre recombinase in the Isl1 locus (Isl1Cre)20, resulting in Cre-mediated recombination in early CPCs by embryonic (E) day 7.75. The resulting Notch1-null embryos failed to populate the developing right ventricle segment, which is derived from Isl1+ CPCs (Fig. 1a–c, g–i). Strikingly, the affected Isl1+ CPC pool dorsal to the developing heart was expanded with an increase in the percentage of proliferating cells, marked by an anti-phospho-histone H3 (PH3) antibody (Fig. 1d–f, j–m). Accumulation and proliferation of CPCs behind the developing heart was similar to the effect of stabilized β-catenin on CPCs2, although in the latter, CPCs also migrated into the heart.
The striking similarity between Notch1 loss-of-function and β-catenin gain-of-function mutants in CPCs led us to hypothesize that Notch and β-catenin signalling intersect during CPC fate or expansion decisions. No significant changes in the expression of genes involved in the Notch signal transduction pathway were observed in β-catenin-stabilized mice (data not shown), suggesting that it is unlikely that β-catenin regulates Notch signalling in CPCs. Using an ES cell line with a bacterial artificial chromosome (BAC) containing green fluorescent protein (GFP) in the Nkx2.5 locus21, we isolated Nkx2.5–GFP+ cells by fluorescence-activated cell sorting (FACS). The Nkx2.5–GFP+ cells expressed high levels of Isl1 (Supplementary Information, Fig S1a), consistent with these cells representing CPCs. We knocked down Notch1 with short interfering RNAs (siRNAs) in Nkx2.5–GFP+ CPCs cultured in a monolayer. Endogenous levels of Notch1 were reduced considerably by siRNA transfection, as shown by western blot analysis (Fig. 1n). Consistent with the in vivo data, Notch1 knockdown resulted in an increase in the number of CPCs (Fig. 1o). Notch1 knockdown did not affect the levels of total β-catenin in CPCs (Fig. 1n). However, the levels of dephosphorylated (free) β-catenin, a form required to mediate Wnt/β-catenin signalling, were considerably higher in the Notch1 knockdown CPCs (Fig. 1n). Consistent with this observation, Notch1 knockdown CPCs showed significantly increased levels of Topflash activity, a luciferase-based reporter system for Wnt/β-catenin signalling (Fig. 1p). Increased levels of nuclear β-catenin were also observed in the cardiac mesoderm of the Notch1 mutant embryo (Supplementary Information, Fig S1b). These findings suggest that Notch1 normally represses CPC expansion and negatively regulates the active form of β-catenin.
To search for genes responsible for CPC expansion in an unbiased manner, we performed gene expression analyses of β-catenin-stabilized CPCs in vivo. For this analysis, we generated RosaYFP; Isl1Cre; β-catenin(ex3)loxPloxP embryos that express yellow fluorescent protein (YFP) in descendants of Isl1+ progenitors in the cardiac region with stabilized β-catenin (Fig. 2a). YFP+ cells from E9.0 embryos, before cardiac dysfunction, were purified by FACS (Fig. 2b) and used for mRNA expression arrays.
Many known targets of canonical Wnt signalling, including components of the Wnt signalling pathway, were upregulated in mutants, confirming the quality of the data set (Supplementary Information, Fig S1c, Table S1). We found that expression of genes implicated in cell proliferation and differentiation (for example, Ndrg1, Bhlhb2 and Fgfs) was highly upregulated (4–11 fold) in mutants (Supplementary Information, Fig S1c, Table S1). Unexpectedly, several genes essential for CPC development, including Isl1, Myocd, Shh and Smyd1 were significantly downregulated in the mutants and this was validated by quantitative real-time PCR (qPCR) (Fig. 2c, d). Curiously, Isl1 was downregulated when β-catenin was stabilized. In agreement with the array analyses, Isl1 transcripts were barely detectable by in situ hybridization in CPCs of β-catenin-stabilized embryos (Fig. 2e, i; Supplementary Information, Fig. S1d). Smyd1 and Myocd transcripts were also significantly downregulated in β-catenin-stabilized embryos, whereas Bhlhb2 was upregulated specifically in the Isl1Cre domain (Fig. 2f–h, j–l; Supplementary Information, Fig. S1d). Consistent with the opposing functions of Notch1 and β-catenin described above, Isl1, Myocd, Shh and Smyd1 were significantly downregulated and Bhlhb2 was upregulated in Notch1 mutant embryos (Fig. 2d; Supplementary Information, Fig. S1e).
Isl1 is a homeodomain-containing transcription factor that is transiently expressed in CPCs before their migration into the heart tube, but is silenced as further migration and differentiation proceed4. Although on the basis of its temporal expression, Isl1 is intuitively thought to promote CPC expansion, we investigated whether Isl1 downregulation mediates expansion of CPCs observed in embryos with stabilized β-catenin. To test this possibility, we used the Isl1Cre line described above, which contains an IRES-Cre cassette inserted into the exon encoding the second LIM domain of Isl1, resulting in an Isl1-null allele20. Isl1Cre mice were bred with RosaYFP mice to generate Isl1Cre/Cre; RosaYFP embryos. We quantified the number of YFP+ cells at E8.0 (5 somite stage), before Isl1Cre expression is initiated in neural cells, by FACS. Surprisingly, Isl1-null embryos had a significantly higher percentage of YFP+ cells than control embryos (Fig. 3a, b). The results suggest that Isl1 negatively regulates the number of CPCs in vivo. The significant increase is unlikely to be attributable to higher Cre expression in Isl1-null embryos, as heterozygous Cre mice mediate recombination as efficiently as homozygous Cre mice.
To determine whether Isl1 also negatively regulates expansion of CPCs derived from pluripotent ES cells, we transiently knocked down Isl1 levels in the Nkx2.5–GFP ES cell line by introducing an Isl1 short hairpin RNA (shRNA) construct, which efficiently reduced Isl1 transcripts by about 75% (Supplementary Information, Fig. S1f). We then quantified the number of Nkx2.5–GFP+ CPCs in embryoid bodies from embryoid day (ED) 6, as cardiac progenitors begin to emerge and differentiate from primitive mesoderm7,8. Knockdown of Isl1 from ED0–3 did not change the number of Nkx2.5+ progenitors (data not shown). However, Isl1 knockdown from ED3–6, just after emergence of mesoderm, resulted in an increase in the CPC population at ED6–8 (Fig. 3c; Supplementary Information, Fig. S2a), consistent with our in vivo data.
These findings prompted us to test whether Isl1 downregulation was required for CPC expansion induced by β-catenin. We transfected Nkx2.5–GFP+, FACS-purified CPCs from day 5 embryoid bodies with a stabilized β-catenin expression construct22 with or without an Isl1 expression construct. As previously reported, increased CPC expansion was evident two days after transfection with stabilized β-catenin (Fig. 3d). However, co-transfection with Isl1 restored the number of CPCs to normal levels (Fig. 3d). This suggests that the decrease in Isl1 is necessary for Wnt/β-catenin signalling–mediated expansion of CPCs.
Because Isl1 seemed to be involved in repressing expansion of CPCs, we investigated whether Isl1 promotes differentiation in the ES cell system. We generated a stable Isl1 knockdown ES cell line by introducing an Isl1 shRNA construct into Nkx2.5–GFP ES cells and clonally isolating cells with effective (∼80%) Isl1 knockdown (Supplementary Information, Fig. S1f). Similarly to transient Isl1 knockdown, the number of Nkx2.5–GFP+ CPCs was significantly increased at ED6 (Supplementary Information, Fig. S2b). However, cells differentiated from the Isl1 knockdown ES cells showed severely reduced beating frequencies and compromised expression of cardiac sarcomeric genes (Myh6, Myh7, Mlc2a, Mlc2v) from ED9 (Fig. 3e, f). To determine the CPC-autonomous role of Isl1 during cardiac differentiation, we FACS-purified Nkx2.5–GFP+ CPCs from ED5 embryoid bodies and differentiated them by re-aggregation in suspension (Fig. 3g). Nkx2.5–GFP+ CPCs are multipotent and differentiate into myocardial, smooth muscle and endothelial cell lineages7,12. Normal levels of endothelial gene expression (CD31, Flk1) were observed in differentiating Isl1 knockdown CPCs (Fig. 3h). However, expression of cardiomyocyte and smooth muscle cell genes was markedly downregulated (Fig. 3h). This suggests that Isl1 not only represses expansion of CPCs, but is also necessary for proper differentiation of CPCs into myocardial and smooth muscle, but not endothelial, cell lineages.
Given that Isl1 loss-of-function suppressed cardiomyocyte differentiation, we sought to determine whether Isl1 has an instructive role in myocardial lineage formation. Isl1 expression levels were upregulated in embryoid bodies from ED4–5 (Supplementary Information, Fig. S3a). To prematurely increase Isl1 expression levels in a temporally and physiologically relevant way, we transiently transfected an Isl1 expression construct (30 ng per 105 cells) into dissociated ED2 embryoid body cells and re-aggregated them for further differentiation (Fig. 4a). This resulted in an increase of about twofold in Isl1 levels at ED6 (Fig. 4b). Myocardial differentiation was monitored by sarcomeric gene (for example, Myh7, Mlc2v, Actc1) expression over the course of embryoid body differentiation. Sarcomeric gene expression levels did not change during the early phase of CPC differentiation (data not shown). However, by ED8, Isl1-transfected embryoid bodies expressed higher levels of cardiac muscle genes than control embryoid bodies (Fig. 4c). To determine the effect of excess Isl1 on the number of cardiomyocytes, we used the Myh7–GFP ES cell line to quantify cardiomyocyte number. We observed a 25% increase in Myh7+ cells in Isl1-overexpressed embryoid bodies (Fig. 4d; Supplementary Information, Fig. S3b). This suggests that Isl1 can promote myocardial differentiation of CPCs in an instructive manner.
In addition to Isl1, Myocd and Smyd1 are important genes for cardiogenesis23,24,25,26,27 that were downregulated in CPCs with increased β-catenin (Fig. 2c–g, i–k). Myocd is a potent co-activator of serum response factor regulation of smooth muscle24 and cardiac gene expression27. Smyd1 is a muscle-restricted histone methyltransferase essential for cardiomyocyte differentiation in vivo23,25. To determine whether Isl1 regulates these genes in CPCs, we used Nkx2.5–GFP+ CPCs purified from the stable Isl1 knockdown ES cell line. Smyd1 levels did not change, but Myocd levels were significantly reduced in the Isl1 knockdown CPCs (Fig. 5a). To determine whether this is also the case in vivo, we performed in situ hybridization for Myocd transcripts in Isl1-null embryos. In agreement with in vitro data, Myocd levels were severely compromised in Isl1-null embryos, whereas Smyd1 levels did not change (Fig. 5b–i). This suggests that Isl1 is required for normal Myocd expression.
Through bioinformatic searches, we identified an Isl1 consensus site in an evolutionarily conserved island (555 bp) located in the first intron of the Myocd locus (Fig. 5j). We observed robust transactivation of luciferase when the element was linked to luciferase reporter and introduced into embryoid bodies at ED6–8 (when endogenous Isl1 is enriched and biologically functional) (Fig. 5k). However, luciferase activity was significantly reduced when the Isl1 site was mutated (Fig. 5k). In addition, excessive Isl1 further increased luciferase activity when the Isl1 site was intact but not when it was mutated (Fig. 5k). Chromatin immunoprecipitation (ChIP) with anti-Isl1 antibodies in ED8 embryoid bodies revealed that the site was associated with Isl1 protein (Fig. 5l). This association was further confirmed by electrophoretic mobility shift analyses, which showed specific binding of Isl1 to the site (Fig. 5m). Together, these data suggest that Isl1 may directly regulate Myocd expression.
As Isl1 did not affect Smyd1 expression, we hypothesized that β-catenin might activate a transcriptional repressor to downregulate Smyd1 expression. Among the transcriptional repressors affected by β-catenin in our array, Bhlhb2 was the most highly upregulated. Bhlhb2 is a basic helix-loop-helix (bHLH)-containing, DNA-binding repressor that is involved in many biological processes, including proliferation, differentiation and regulation of circadian rhythm28,29,30. Quantitative PCR confirmed that Bhlhb2 was highly upregulated in embryos with stabilized β-catenin (Fig. 5n), consistent with upregulation shown by in situ hybridization in the cardiac area and other domains of Isl1Cre activity (Fig. 2h, l). Overexpression of Bhlhb2 in Nkx2.5–GFP+ CPCs mimicked Smyd1 repression observed with β-catenin stabilization (Fig. 5o). Isl1 expression was not affected by Bhlhb2, providing an important control (Fig. 5o). We identified four conserved Lef/Tcf consensus sites in the 5´ and 3´ UTRs of Bhlhb2 (Fig. 5p) and tested whether any were directly bound by β-catenin. ChIP with anti-β-catenin antibodies in ED8 embryoid bodies revealed that two of the four sites (A and D) were indeed associated with β-catenin (Fig. 5q). To determine which site can mediate Wnt/β-catenin signalling, conserved elements encompassing the Lef/Tcf sites were individually inserted upstream of a luciferase reporter and luciferase activity was examined in ED8 embryoid bodies. We found that the construct-containing site D, but not A, resulted in a significant increase in luciferase activity on stimulation with β-catenin or 6-bromoindirubin-3´-oxime (BIO), a Wnt/β-catenin signalling activator (Fig. 5r). This increase was, however, not observed in cells transfected with a mutant construct (Fig. 5r). These data suggest that Bhlhb2 may be a direct target of the Wnt signal.
Through use of mouse genetics and the ES cell system, we have shown that Wnt/β-catenin signalling functions as a central regulator of CPCs by integrating signals from the Notch pathway and regulating a cascade of downstream transcriptional events involving Isl1, Myocd and Smyd1 (Fig. 5s). We found that Notch1 activity within CPCs was required for their exit from the expansive state into the differentiated state, providing the first evidence for Notch signalling requirement within multipotent CPCs in vivo. Consistent with negative regulation of active β-catenin by Notch1, Notch1 loss-of-function and β-catenin gain-of-function had similar effects on expression of the cardiac transcription factors, Isl1, Myocd, Smyd1 and Bhlhb2. Our finding that CPCs in vivo and in vitro had greater expansion on disruption of Isl1 and that Isl1 could promote differentiation suggests that despite its very transient expression, Isl1 triggers further development of CPCs into cardiac cells rather than promoting its renewal state. Strikingly, Isl1 downregulation induced by β-catenin was necessary for Wnt/β-catenin-induced expansion of CPCs. Manipulation of the cascade described here may be useful in guiding the expansion and directed differentiation of CPCs for regenerative therapies and other uses of stem cell-derived cardiomyocytes.
Note: Supplementary Information is available on the Nature Cell Biology website.
Mouse genetics and CPC and ES cell culture.
Control (RosaYFP/+; Isl1Cre/+) or mutant (RosaYFP/+; Isl1Cre/+; β-catenin(ex3)loxPloxP/+) embryos were obtained by crossing RosaYFP/+; β-catenin(ex3)loxPloxP/+ with Isl1Cre/+ mice20,31. YFP+ cells from the resulting embryos were purified by FACS and used for gene expression analyses. To quantify embryonic CPCs, RosaYFP/+; Isl1Cre/+ were crossed with Isl1Cre/+ mice, and YFP+ cells from the resulting embryos were counted by FACS. To generate Isl1Cre/+; Notch1loxP/loxP, Isl1Cre/+; Notch1loxP/+ mice were crossed with Notch1loxP/loxP mice19. Genotyping was performed as described previously2. To identify Isl1-het (Isl1Cre/+) or null (Isl1Cre/Cre) embryos, DNA was isolated from individual embryos, and qPCR was performed using SYBR Green (Applied Biosystems). Control Isl1 and Cre primers are shown in Supplementary Information, Table S2. ES cells and purified Nkx2.5–GFP+ CPCs were propagated undifferentiated or differentiated as described previously2. For CPC differentiation, FACS-purified GFP+ cells were re-aggregated in suspension (105 cells per well) in ultra-low-attachment 24-well plates (Corning).
Flow cytometry and gene expression analysis.
A Becton Dickinson FACS Diva flow cytometer and cell sorter were used for quantifying and purifying Nkx2.5–GFP+ or Myh7–GFP+ cells. For the microarray analysis and qPCR, total RNA was amplified with the WT-Ovation Pico RNA Amplification System, fragmented and labelled with the FL-Ovation cDNA Biotin Module V2 (Nugen). Hybridization, staining and scanning of the Affymetrix GeneChips were performed in the Gladstone Genomics Core Lab. Raw data generated from at least three independent experiments were further analysed by the group of Ru-Fang Yeh at the Center for Informatics and Molecular Biostatistics, UCSF. To quantify gene expression in Notch1 mutant embryos, total RNA was isolated from hearts and pharyngeal arches from E10.0 embryos. qPCR was performed with the ABI Prism system (7900HT, Applied Biosystems). TaqMan primers used in this study are listed in Supplementary Information, Table S2. All samples were run at least in triplicate. Real-time PCR data were normalized and standardized with SDS2.2 software.
Constructs, siRNA, transfection, EMSA and luciferase assays.
For Isl1 knockdown experiments, an Isl1 shRNA construct set (RMM4534-NM_021,459, Open Biosystems) was used to transiently transfect embryoid bodies and to generate stable knockdown ES cell lines. For Isl1 or Bhlhb2 overexpression studies, their full-length cDNAs (Open Biosystems) were amplified and cloned into the pEF-DEST51 vector (pDEST51-Isl1 or Bhlhb2) through the pENTR vector (pENTR-Isl1 or Bhlhb2) using the Gateway system (Invitrogen). pEF-lacZ (Invitrogen) was used as a control. For Notch1 knockdown studies, Block-iT Alexa Fluor Red (46-5,318, Invitrogen) or Notch1 siRNA (M-041,110-00-0005, Dharmacon) was used at concentration of 50 or 100 nM. Myocd–luc was generated by cloning their corresponding regions into the pGL3 luciferase vector (Promega). Myocd–lucmt was generated using QuikChange Site-Directed Mutagenesis Kit (Stratagene). For Bhlhb2D–luc and Bhlhb2D–lucmt generation, oligonucleotides containing the Tcf/Lef site were cloned into the pGL3 vector. All the oligonucleotide sets are listed in Supplementary Information, Table S2. Stabilized β-catenin and Top/Fop-flash luciferase constructs were provided by A. Barth (Stanford University, CA) and the laboratory of R. Moon (University of Washington, WA), respectively. ES cells, embryoid bodies or CPCs were transfected with the indicated constructs or siRNA using Lipofectamine 2000 (Invitrogen) after generating single-cell suspensions with Accutase (Chemicon). EMSAs and luciferase assays were performed as described previously32,33. For EMSAs, the pCITE–ISL134 construct containing the truncated Isl1 cDNA with the homeodomain was provided by B. Black (University of California, San Francisco) and used to generate Isl1 protein. All EMSA probes are listed in Supplementary Information, Table S2. For luciferase assays, Renilla was used as an internal normalization control.
In situ hybridization, immunostaining and western analysis.
Whole-mount in situ hybridization was performed as described previously, with designated antisense probes4,23,26. Bhlhb2 antisense riboprobe was synthesized and purified from pENTR–Bhlhb2. To detect proliferating cells in CPCs, embryo sections were stained with anti-PH3 (Upstate) and anti-Isl1 (DSHB). To visualize Isl1 protein in Notch1 mutant embryos, the TSA System (PerkinElmer) was used to amplify Isl1 signals. Nuclear β-catenin was detected with anti-PY489 antibody (DSHB). For western blotting, lysates from day 3 CPCs after transfection with indicated siRNAs were analysed using antibodies against Notch1 (DSHB), dephospho β-catenin (Calbiochem) and GAPDH (Santa Cruz Biotechnology).
Chromatin immunoprecipitation assays.
For chromatin immunoprecipitation (ChIP) assay, embryoid bodies were treated with BIO (2.5 uM) or transfected with Isl1 or β-catenin constructs22 (100 ng 10−5 cells) from ED5–7, and collected at ED8. Crosslinking of histones to DNA, chromatin extraction, immunoprecipitation and elution were performed using the ChIP assay kit (Upstate) with anti-IgG–HRP, Isl1 (Abcam) or β-catenin (Santa Cruz Biotechnology). PCR primer sets spanning the indicated Lef/Tcf binding sites in the Bhlhb2 locus are shown in Supplementary Information, Table S2.
The two-tailed Student's t-test, type II, was used for data analyses. P < 0.05 was considered significant.
The full microarray data performed in this study are available in NCBI Gene Expression Omnibus (GEO, accession number: GSE15232).
Gene Expression Omnibus
We thank R. Kopan (Washington Univeristy, St. Louis, MO) and M. M. Taketo (Kyoto University, Kyoto, Japan) for providing Notch1flox and β-catenin/loxP(ex3)loxP mice, respectively. The authors thank G. Howard and S. Ordway for editorial assistance, R.F. Yeh for statistical analyses, K. Cordes for graphical assistance, B. Taylor for manuscript and figure preparation and Srivastava lab members for helpful discussions. C.K. was supported by a fellowship from the American Heart Association (AHA) and California Institute for Regenerative Medicine (CIRM); D.S. was an Established Investigator of the AHA and was supported by grants from NHLBI/NIH and CIRM. This work was also supported by NIH/NCRR grant (C06 RR018928) to the Gladstone Institutes.
About this article
Scientific Reports (2017)