Murine models demonstrate distinct vasculogenic and cardiomyogenic cKit⁺ lineages in the heart

Abstract

Following two recent genetic studies in mice addressing the developmental origins¹ and regenerative activity of cardiac cKit⁺ cells², two additional reports by Sultana et al.³ and Liu et al.⁴ provide further information on the expression of cKit in the embryonic and adult heart. Here we synthesize the findings from the four distinct c-kit models to gain insights into the biology of this important cell type.

Introduction

In 1998, Anversa and colleagues provided the first evidence that the mammalian heart is not a terminally differentiated organ, demonstrating the production of new cardiomyocytes in adulthood^{5, 6}. Although intense debate still exists on the exact rate at which this occurs^7–11, this paradigm changing observation has been subsequently documented with sophisticated animal models^{8, 12, 13} and with ¹⁴C dating experiments in humans^{8, 9}. Following this initial experiment, evidence emerged that extra-cardiac cells, some of which express cKit, could contribute to cardiac cell renewal^13–17.

In 2003, cKit⁺ cells in the heart were identified as resident cardiac stem cells, capable of generating new coronary vessels and cardiomyocytes in response to injury¹⁸. These cells were described to reside in cardiac stem cell niches¹⁹; were capable of ex-vivo propagation and clonal expansion; exhibited capacity for in-vitro and in-vivo differentiation into cardiomyocytes, endothelial, and smooth muscle cells^{20, 21}; participated in the in-vitro formation of cardiospheres similar to neural stem cells^22–24; and were negative for hematopoietic lineage markers^{18, 19}. Several groups demonstrated ex-vivo propagation of cardiac stem cells which were subsequently transplanted in injured hearts leading to enhanced regeneration via mechanisms involving their direct differentiation into new cardiomyocytes and vessels^{18, 21, 23}. These observations warranted the approval of clinical trials from the FDA in order to test the safety and efficacy of resident cardiac stem/progenitor cells for the treatment of human heart disease (NCT00474461; NCT00893360)^{25, 26}.

In 2014, a study by van Berlo and colleagues reported that cKit cells in the heart have minimal cardiomyogenic ability, hence their therapeutic importance may be marginal². Two mouse lines, carrying 2 different versions of Cre recombinase [either a tamoxifen inducible MerCreMer or a non-inducible Cre-IRES-eGFPnls (nls stands for “nuclear localization signal”)] under the control of the endogenous cKit promoter, and Cre-reporter genes, were engineered to track the fate of cKit-expressing cells in the heart (Figure 1). Expression of reporters was analyzed either before or in response to myocardial infarction injury. In both cases, van Berlo et al found that the vast majority of the cardiac cells that had undergone Cre-mediated recombination were endothelial cells and only rarely cardiomyocytes². This finding led the authors to conclude that cKit⁺ cardiac cells “may be dedicated vascular endothelial progenitor cells…[not] capable of regenerating the heart with new contracting myocardium to a physiologically meaningful extent”¹¹. However, since cKit expression occurs in several cardiac and extra-cardiac cell lineages²⁷, the role and identity of cardiac cKit⁺ cells remained elusive. Do the endothelial cells and cardiomyocytes arise from the same cKit-expressing lineage, or are there multiple cKit⁺ lineages in the heart?²⁷ Why do cKit⁺ cells differentiate into far fewer cardiomyocytes than endothelium? Are they bona fide cardiovascular progenitors or do they represent non-cardiomyogenic cells that stochastically (trans)differentiate into myocardium?

Figure 1. The cKit receptor is expressed on Nkx2.5-negative vasculogenic cells and Nkx2.5⁺ cardiomyogenic lineage of neural crest origin.

Three genetic lineage fate-mapping studies demonstrate cardiac cKit⁺ cells forming cardiomyocytes in the mouse heart, at a low level. The three different murine studies are not in full agreement on endothelial differentiation. Contrary to van Berlo et al, Hatzistergos et al. showed that cKit⁺ cardiac cells are not primed toward endothelial cell differentiation. Sultana et al also show that the coronary endothelium does not descend from cKit⁺ cardiac progenitors, but ~43% of coronary endothelial cells express cKit (possibly at low levels). Both Hatzistergos et al. and Sultana et al demonstrate that cKit⁺/Nkx2.5⁺ cardiomyogenic progenitors are not present before E12.5, and both of these studies document unequivocal evidence of cKit⁺/Nkx2.5⁺ cardiomyogenic progenitors after this time. The nature of these cells is further elucidated by experiments showing their generation in-vitro from mouse iPSCs. *An important limitation between all three studies is the utilization of knock-in mouse models which render the gene-targeted cKit allele non-functional, which may cause abnormalities in cKit⁺ cardiac cells in these mice (e.g. proliferation, migration and differentiation).

Our group performed a high resolution genetic lineage-tracing study using a different mouse line which expresses the CreER^T2 version of Cre recombinase from the endogenous cKit locus and a loxP-STOP-loxP flanked Cre-reporter allele (Figure 1). In addition to addressing some of the issues raised by van Berlo et al, we sought to not only genetically track the fate of cKit⁺ cells, but also to identify their lineage and role in the heart¹. Importantly, we showed that cKit marks rare, bona fide cardiomyogenic progenitors in the neural crest lineage (as shown by intersectional genetic fate-mapping with a Wnt1 reported mouse) with unequivocal capacity to produce new cardiomyocytes, but not coronary endothelial cells. Further, cardiomyocyte differentiation is governed by spatiotemporal changes in the activity of bone morphogenetic protein (BMP), a signaling pathway which subsides during cKit⁺ neural crest cells migration to the heart²⁸, reducing an important activator of cardiomyocyte differentiation. Thus, the limited contribution of cKit⁺ cells in cardiomyocytes is not a function of deficient cardiomyogenic capacity, but rather reflects a non-permissive milieu¹.

The recent paper by Sultana et al.³ further clarifies the identity of the cKit⁺ cardiac lineage(s). The authors developed three new cKit knock-in mouse lines to address the developmental origin of cardiac cKit⁺ cells; the identity of cKit⁺ cardiac cells; and whether cKit⁺ cells are useful for cardiomyocyte regeneration (Figure 1).

Overview of the cKit knock-in alleles

To monitor real-time expression of cKit, the authors developed two new reporter mouse lines (Figure 1). In the first mouse line, transcription of cKit results in nuclear expression of the red fluorescent protein tdTomato (c-Kit^H2B-Tomato/+). The second is a dual reporter mouse line, where transcription of cKit results in nuclear expression of beta-galactosidase, which may be irreversibly substituted by nuclear expression of green fluorescent protein upon Cre-mediated recombination (c-Kit^{nLacZ-H2B-GFP/+}). Importantly, once a cell no longer expresses cKit (i.e. differentiates into a cKit-negative derivative), expression of the reporter genes (either TdTomato, GFP or nLacZ) should no longer be detectable.

Both cKit reporter mouse lines exhibited broad reporter gene expression in the heart, during both development and adulthood³. The degree of cKit reporter gene expression exceeds essentially all reports of cKit expression and/or fate-mapping studies, in rodents^{1, 2, 13, 18, 21, 29–34}, large mammals^{18, 35, 36}, and humans²⁰. For example, although van Berlo et al. showed that reporter gene expression under their Kit-Cre-IRES-eGFPnls allele “…was always coincident” with endogenous cKit protein expression in-vivo, they report that eGFPnls is not broadly expressed in the heart, but marks a rare mononuclear cell population². Similarly, databases such as the Human Protein Atlas (http://www.proteinatlas.org/ENSG00000157404-KIT/tissue/primary+data) and the EMAGE in situ mouse gene expression data base (http://www.emouseatlas.org/emagewebapp/pages/emage_gene_browse.jsf) do not show widespread expression of cKit in the heart.

Insight into this discrepancy derives from RNA in-situ hybridization data which overlapped with in-situ cKit transcriptional activity and reporter gene expression, suggesting that the two new alleles mark cells with levels of cKit transcriptional activity that do not result in protein expression sufficient for conventional immunodetection.

The identity of cKit⁺ cell lineage(s) in the heart

A point of controversy between the van Berlo et al and Hatzistergos et al studies is whether cKit⁺ cardiac cells are vascular endothelial progenitors¹¹. Although van Berlo et al. described an extensive contribution of cKit⁺ cells to endothelial cells², Hatzistergos et al¹ showed that coronary endothelial cells were not a product of cKit⁺ cardiac progenitor cell differentiation. This inconsistency has been shown in previous studies, which variably support ^{18, 21, 29} or refute endothelial cell^{16, 35} differentiation capacity of cKit cells. Similarly, most studies document that coronary endothelium-producing cells do not express cKit^37–39.

Elucidating the developmental origin(s) and identity of the cKit⁺ cardiac cells contribute substantially to resolving this issue. Both Wu et al.¹⁶ and Hatzistergos et al.¹ show that cardiomyogenic cKit⁺ progenitors express Nkx2.5 and do not contribute to coronary endothelium. Notably, although it was initially thought that the cKit⁺/Nkx2.5⁺ cardiac progenitors represent a subpopulation of Nkx2.5⁺ mesodermal progenitors^{16, 40–42} which are present in the developing mouse heart between ~E7.5–E9.0, Hatzistergos et al. clarified a later appearance of these cells in development, supporting the idea that they are likely of cardiac neural crest and not mesodermal origin¹. This was further substantiated with an intersectional fate mapping approach in which the cKit^CreERT2/+ mouse was crossed to mice carrying a Wnt1::Flpe and dual recombinase-responsive indicator alleles. Moreover, the majority of these cells were found within the interventricular septal wall¹, and not the endocardium as was thought⁴⁰.

Following these findings, the question that remained unanswered was whether a relationship existed between the cKit⁺/Nkx2.5⁺ cells^{1, 16} and the vasculogenic cKit⁺ cells described by van Berlo and others^{2, 18, 21, 29}. Sultana et al now provide further evidence that the cKit⁺/Nkx2.5⁺ cardiomyogenic progenitors are distinct from the cKit⁺ vasculogenic cardiac cells³. Unlike the cKit⁺/Nkx2.5⁺ lineage, the cKit⁺ vasculogenic cells first appear in the heart during ~E8.5 and do not express Nkx2.5³. These Nkx2.5-negative cKit⁺ cells, which likely include the cKit⁺ endocardial cells⁴⁰, are actually Pecam1⁺ endothelial cells and not progenitors³ (Figure 1).

Compared to the study by Molkentin and colleagues², Sultana et al show that the majority of cKit⁺ cells in the heart are mature endothelial cells of a Tie2 lineage (by crossing c-Kit^{nLacZ-H2B-GFP/+} to Tie2-Cre mice), and not endothelial progenitors. In fact, they report that ~43% of all endothelial cells in a 4-month old adult mouse heart co-express a cKit^H2B-GFP/+ reporter gene, meaning that cardiac endothelial cells express cKit³. This is a rather surprising finding, considering that with a similar approach, Molkentin and colleagues did not find eGFPnls expression in either cardiomyocytes or endothelial cells of Kit^+/Cre mice². Moreover, Fioret et al., recently reported that virtually all of the Tie1-Cre derived coronary endothelial cells do not express cKit³⁷. These discrepant findings likely represent differences in detection sensitivity between the Molkentin and Cai cKit reporter alleles.

cKit⁺ cardiac cells and cardiomyocyte regeneration

A consistent finding between the 4 independent genetic lineage fate-mapping studies of cKit in the heart, is that the degree to which cKit⁺ cells generate new cardiomyocytes in-vivo during embryogenesis, aging and myocardial injury is lower than expected, based on previous reports^{13, 18, 21}. However, there are important differences in the analysis and interpretation of the genetic findings.

Both genetic fate-mapping studies by van Berlo et al ² and Sultana et al³ interpret the minimal degree of cardiomyogenesis from cKit⁺ cells as a justification for discounting cKit⁺ cardiac progenitors as a physiologically relevant precursor for cardiac formation, therefore questioning efforts to use culture expanded cKit⁺ cells as a therapeutic strategy for myocardial regeneration. However, a recombinase-based genetic fate-mapping study provides a binary method for tracking the activity of a given cell via the activity of a single gene, and does not measure whether a cell holds an inferior or superior ability to differentiate. Thus, the assumption that the limited detection of cKit reporter genes in cardiomyocytes translates into limited cardiomyocyte differentiation capacity of cKit⁺ progenitors is not being tested under these 2 experimental studies, and should not therefore inform conclusions regarding cardiomyogenic ability and therapeutic importance of cKit⁺ cells.

Another key example where a genetic fate-map does not reflect therapeutic importance comes from studies in cardiomyocyte proliferation. Genetic lineage-tracing studies of proliferating cardiomyocytes illustrate that, similar to cKit⁺ cardiac progenitors, the degree of myocardial regeneration from pre-existing cardiomyocytes is functionally insignificant (Table 1)^{12, 13, 43, 44}. Intriguingly, here, the findings are not interpreted as therapeutically irrelevant, but as a justification for targeting cardiomyocyte replication as a meaningful therapeutic strategy for myocardial regeneration^{11, 12}. We would use the same logic to continue to support the use of cKit⁺ cells in therapeutic trials.

Table 1.

Comparison of cardiomyocyte regeneration from pre-existing cardiomyocytes and cKit⁺ cardiac cells.

Study	Regenerative cell source tested	Mouse model	Regenerated CMs post injury
Senyo et al12	Pre-existing CMs	MIMS in MerCreMer/ZEG (8 week chasing)	Analyzed n= 4 mouse hearts; Found 11 regenerated CMs (MIMS+/GFP+, diploid/mononucleated) in total.
Ali et al.44	Pre-existing CMs	Myh6CreER; MADMGT/TG (4 week chasing)	Analyzed n=5 mouse hearts; Found 476 regenerated CMs in total.
van Berlo et al.2	cKit+ cardiac cells	KitCre/+ and KitMCM/+ (4 week chasing)	Analyzed n=2 mouse hearts; Found 37 regenerated CMs in total.
Sultana et al.3	cKit+ cardiac cells	c-kitMerCreMer/+;TnnT2nlacZ-H2B-GFP/+ (~60-day chasing)	Analyzed n=3 mouse hearts; Found ~50 regenerated CMs per heart.

CM: cardiomyocyte; MIMS: multi-isotope imaging mass spectrometry.

The differentiation capacity of cKit⁺ cardiac progenitors is tested in the study by Hatzistergos et al¹. through the combined use of genetic lineage-tracing and iPSC modeling. These experiments define that cKit⁺ cardiac progenitors are fully capable of generating therapeutically meaningful numbers of bona fide cardiomyocytes, but other factors (such as the activity of BMP signaling pathway) limit their prevalence in in-vivo cardiomyogenic fate-mapping studies.

Last, the findings by Sultana et al, as well as a more recent (and fourth) report by Liu et al. with a new Kit-CreER^T2 allele⁴, suggest that some of the resident cKit⁺ cells in the heart are differentiated cardiomyocytes. Again, this finding is in contrast with previous reports⁴⁵, including van Berlo² and Hatzistergos et al.¹ both of which demonstrated that the original population of cells labeled with the Kit^+/Cre, Kit^+/MCM and cKit^CreERT2/+ alleles were not cardiomyocytes. However, since the Cai cKit reporter alleles appear to exhibit high sensitivity and label cells even with very low cKit transcriptional activity, the possibility that some of the cardiomyocytes detected by Sultana et al are pre-existing cardiomyocytes expressing low levels of cKit⁴⁶, or perhaps cKit-derived cTnT⁺ cells at early stages of cardiogenic specification with residual cKit (or cTnT) expression, cannot be excluded.

Discussion

The identification of bona fide cardiac progenitors in the postnatal heart is a major milestone in the field of cardiovascular research that would signify the departure from the longstanding view of human heart disease as incurable. In our view, this milestone has now been reached through the development of seven elegant research reagents, designed to address the cKit controversy.

Despite the heated debates and provocative headlines that are expected to characterize any type of research with such profound implications for human disease, all four independently-conducted genetic lineage fate-mapping studies of cKit with 7 different knock-in alleles, unambiguously record the presence of cKit⁺ cardiomyogenic progenitors in the embryonic and postnatal mouse heart. The physiological role of these cells in cardiomyogenesis may not be as extensive as was previously thought, but the fact that this unequivocal and evolutionarily conserved event is interpreted to mean that these cells are not cardiac progenitors, is at odds with the current experimental data^1–4.

The intriguing similarities and differences on cardiac cKit expression that have been presented by these mouse lines (Figure 1) underline the vastness under which the cardiomyogenic and vasculogenic programs are operated in mammals. The development of these new reagents offers a unique opportunity to comprehensively study the role of cKit in the heart and identify possible pathways for deploying cKit⁺ cells for therapeutic myocardial regeneration. For example, although pharmacologic or genetic modulation of the cardiac BMP pathway may be required for therapeutically enhancing the cardiomyogenic activity of the cKit⁺/Nkx2.5⁺ progenitor lineage¹, whether BMP pathway plays a role in vasculogenic cKit⁺ cells is unknown and needs to be explored since they are likely the most abundant cKit⁺ population in the heart. Similarly, we have previously shown that endogenous cKit⁺/Nkx2.5⁺ progenitors may be therapeutically stimulated with mesenchymal stem cell (MSCs)-based therapy^{35, 47, 48}. However, whether MSCs affect the cKit⁺ vasculogenic lineage is unknown. The combined use of more than one cKit knock-in alleles may help address these questions, and others.

On the other hand, these expression differences are also indicative of the inherent limitations of the Cre-loxP knock-in approach. The results from these studies may have been skewed by a number of possible factors such as abnormal behavior of cKit⁺ cells due to inactivation of the targeted allele; limited detection sensitivity of cells with physiologically low cKit expression, again due to allelic inactivation; improper expression of Cre from the cell of interest; improper expression of the reporter genes from the cell of interest; etc., all of which have been extensively discussed elsewhere⁴⁹. Thus, caution should always be exercised when interpreting the knock-in genetic fate-mapping findings.

In summary, all independently conducted Cre-loxP knock-in genetic fate-mapping studies of cKit identify a rare, cKit⁺ bona-fide cardiomyogenic cell lineage in the heart. These cells are likely of cardiac neural crest origin and are distinct from the mesodermal cKit⁺ vasculogenic lineage, which is abundantly present in the heart. More work is required to understand why these cells with clear-cut ability to generate cardiac myocytes enter the heart at a time when a required signal for their differentiation has waned.

Nonstandard Abbreviations and Acronyms

Cre

A recombinase enzyme derived from the P1 bacteriophage

LoxP

DNA sequences recognized by Cre recombinase (locus of X-over P1)

IRES

Internal Ribosome Entry Site

MerCreMer (or MCM)

a version of Cre, fused to two copies of a mutant murine estrogen-receptor ligand-binding domain (amino acids 281–599) with a G525R substitution

CreER^T2

a version of Cre, fused to a mutant human estrogen receptor ligand-binding domain (G400V/M543A/L544A triple mutation)

eGFP

enhanced green fluorescent protein

H2B

Histone H2B

nLacZ

nuclear beta-galactosidase

tdTomato

tandem dimer tomato red fluorescent protein

Flpe

The FLPe version (mutations P2S, L33S, Y108N, S294P) of flippase (FLP recombinase)

MADM^GT/TG

mice containing the Mosaic Analysis with Double Marker alleles

ZEG

A Cre reporter mouse line expressing a LacZ/EGFP double reporter allele

MIMS

multi-isotope imaging mass spectrometry