PDGFRα Defines Proepicardial and Epicardial Cells with CFU-F Ability

In situ hybridization (ISH) and immunostaining of embryos revealed dynamic and widespread expression of Pdgfra and PDGFRα protein (Figures 3B and 3C and data not shown). In hearts at 9.5 days postcoitum (dpc), however, high expression was seen only in proepicardium, the progenitor structure for the epicardium, and components of the coronary vasculature and interstitial fibroblasts, with the latter lineages formed from epicardium by epithelial-to-mesenchymal transition (EMT) (Carmona et al., 2010). In 12.5 dpc embryos, PDGFRα protein was evident in the epicardium, but not myocardium (Figure 3D), and at 14.5 dpc most cells expressing the highest levels of PDGFRα were seen in the subepicardium, with some isolated cells within the myocardial interstitium (Figure 3E, inset). We also analyzed GFP expression in a mouse knockin line in which a nuclear-localizing GFP cassette was inserted into the Pdgfra locus (Table S1 available online). FACS sorting for GFP fluorescence was equally efficacious compared to PDGFRα antibody in enriching for cCFU-F (Figure S1H). At 12.5 dpc, high GFP was seen in a mosaic pattern in epicardium (marked by Wilm's Tumor gene, WT1) and subepicardium, as well as endocardial cushions (Figure 3F). Perdurance of GFP allowed a surrogate fate tracking of the PDGFRα⁺ lineage. At 12.5 dpc, a few Pdgfra-GFP⁺ cells were present in the subepicardium of the atrioventricular and interventricular grooves (Figures 3F and 3G), while at 15.5 dpc many interstitial cells were evident, some at a considerable distance from the epicardium (Figure 3H arrows). They did not express the myocardial transcription factor NKX2-5 (Figure 3I) or the endothelial marker CD31 at this stage (except for some double-positive cells within endocardial cushions), although they appeared closely associated with CD31⁺ coronary vessels (Figure 3J). In coronary arterioles already invested with SM at 15.5 dpc, GFP^high cells were seen in a perivascular, largely adventitial, location, with GFP^weak cells located within the media and coexpressing the SM marker CALPONIN1 (Figure 3K). Likewise, in the fetal aorta, GFP^high cells localized to the perivascular adventitia, and GFP^weak cells, to the SM medial layer (Figure 3L). These localizations are consistent with the perivascular location of MSC-like cells from other tissues (da Silva Meirelles et al., 2008) and suggest that in development, PDGFRα⁺ epicardial cells undergo EMT and give rise to a minor subpopulation of cardiac interstitial cells that retain Pdgfra-GFP expression, the descendants of which include SM and other perivascular cells of the coronary vessels.

To explore this concept further, we confirmed the expression of Wt1 in epicardium and subepicardium at 15.5 dpc scoring GFP expression from Wt1^GFPCre/+ embryos (Table S1), and we confirmed that both Pdgfra and Wt1 transcripts were restricted to Wt1-GFP⁺ cells (Figures S3A and S3B). We then analyzed the lineage descendants of Wt1-expressing epicardial cells from 10.5 to 15.5 dpc. This was achieved by treating dams with embryos carrying both a sensitive CRE-dependent GFP reporter allele (Rosa26^mTmG/+) and a conditionally activated Wt1-Cre allele (Wt1^CreERT2/+) (Table S1) with tamoxifen at 10.5 dpc. At 15.5 dpc, in addition to epicardium and subepicardium, there was extensive GFP expression in interstitial lineages, including vessel-associated cells (Figures S3A and S3B), contrasting with the sparse pattern of Pdgfra-GFP⁺ cells at this stage (Figures 3H–3J). Pdgfra and Wt1 transcripts were again enriched in GFP⁺ cells, confirming the association between Pdgfra⁺ cells and the epicardial lineage tree.

Using FACS, we next purified a Pdgfra-GFP⁺ cell fraction from dissected and pooled proepicardia (9.5 dpc), the free ventricular walls of progressively older fetal hearts (devoid of endocardial cushions), and whole adult hearts. The majority of Pdgfra-GFP⁺ cells were positive for PDGFRα protein (Figure 3M and data not shown). Colony-forming ability was present in the proepicardium and at all stages in the ventricular free wall (Figure 3N). As in the adult, cCFU-Fs arose exclusively from the Pdgfra-GFP⁺ fraction (note that the cardiac mesenchymal fraction does not express SCA1 until midgestation). The number of cCFU-Fs rose approximately exponentially throughout fetal life and peaked in adulthood (Figure 3N), with 1705 ± 183 (SD) large cCFU-Fs/heart (n = 3).

Consistent with an important role for adult epicardium in heart regeneration in multiple species (Kikuchi et al., 2011; Smart et al., 2011), it was of interest to determine if cCFU-F persisted in the epicardial layer into adulthood. Immunofluorescence revealed that ~5% of total heart Pdgfrα-GFP⁺ cells were found in the epicardial layer, marked by expression of pan-CADHERIN (Figure S3D). However, when adult epicardial cells were isolated by FACS using E-CADHERIN antibody (E-CAD⁺S⁺P⁺), only small and micro colonies were detected (Figure S3D), suggesting that epicardium contains only committed progenitors and is devoid of cCFU-Fs of the highest rank.

Since both endocardial cushions and valve leaflets in the adult contained Pdgfra-GFP⁺ cells (Figure 3E), we also explored whether cushions and valves represent an alternative niche for cCFU-Fs, which is also of interest in light of studies showing that some valve cells have their origins in epicardium (Lie-Venema et al., 2008; Zhou et al., 2009). Dissected 13.5 dpc cushions yielded only few micro colonies (Figure S3E). S⁺P⁺ cells FACS-purified from adult mitral and tricuspid valve leaflets also yielded mostly micro colonies, although there were also some small and even a very few large colonies (Figure S3D), albeit with altered morphology (not shown). Large CFU-Fs within the valve leaflets represented only ~0.006% of all large CFU-Fs detected in the adult heart.

Finally, in the adult, Pdgfra-GFP^high cells did not express the pericyte and SM marker NG2 (Figure S3C). This is at odds with claims for the equivalence of CFU-Fs and perivascular cells or pericytes in BM and other organs (Armulik et al., 2011; Méndez-Ferrer et al., 2010) (see Discussion).

Adult BM Cannot Rescue cCFU-Fs after Irradiation Injury

Blood-born progenitors mobilized from the BM are a credible source of adult tissue stem cells, particularly after ischemic injury. To explore a possible BM origin for some or all cCFU-Fs, total BM cells from transgenic mice expressing GFP ubiquitously (Table S1) were engrafted into lethally irradiated mice. BM transplants were performed on 8 week C57BL/6 recipients, with subsequent aging from 2 to 8 months. We also induced MI by ligation of the anterior descending coronary artery in BM engraftment mice as a further stimulus for BM flux (Figure 4A). In engrafted mice, FACS confirmed the presence of GFP⁺ expressing cells in peripheral blood and BM at 4–6 weeks posttransplantation (86.4% ± 5% and 65.0% ± 15% of total cells, respectively; Figures S4A and S4B). However, in cCFU-F assays, no large colonies were seen in either the BM-derived (GFP⁺) or non-BM-derived (GFP^–) fractions (Figure 4B). Induction of MI had no effect on colony formation (analysis at 5 days post-MI). To investigate the immediate effects of irradiation, C57BL/6 mice were subjected to 9 Gy with cCFU-F assays performed 6 hr later. Again, no large colonies were formed (Figure S4E). We also assessed the effects of irradiation on BM CFU-Fs (bmCFU-Fs), which yielded only ~20% of the total colony numbers at both 3 months and 6 months (Figure 4C). Of these, ~70% were GFP⁺ (and therefore of nonirradiated, donor origin) at 3 months, diminishing to 50% at 6 months (Figure 4D). These data show that CFU-Fs residing in BM and heart are radiation sensitive. Importantly, engrafted total BM cells can reestablish the niche for bmCFU-F, but cannot rescue ablated cCFU-F in the irradiation injury setting, even after 8 months of aging and MI.

Open in a separate window

Figure 4

cCFU-Fs Are Lost after Irradiation and Cannot Be Rescued by Whole BM Transplantation

(A) Overview of BM transplant experimental design. All transplants were on mice 8 weeks of age.

(B) Large colony formation was completely lost in BM transplanted mice. Colony formation was not rescued after 3 months or 6 months aging, or after MI.

(C) Reduced BM CFU-F colonies from BM-GFP chimeric mice, which do not recover after 6 months.

(D) Majority of remaining BM CFU-F colonies are GFP⁺, indicating donor origin.

Error bars = SEM between independent experiments. See also Figure S4.

BM Does Not Contribute to cCFU-Fs after MI or HSC Mobilization

To avoid irradiation damage to the heart, we protected it by lead shielding (Figure 5A). Successful engraftment of GFP⁺ cells was again confirmed by FACS analysis of blood and BM (Figures S4C and S4D). GFP⁺ cells infiltrating the heart 5 days postsham or MI surgery in shielded (and unshielded) mice were all CD45⁺ (>99%), with majority subfractions positive for SCA1 and CD31 (Figure S5A). Within the S+P+ population from shielded mice, ~4% of cells were GFP⁺, although no cells were positive for the perivascular marker CD146. Shielded mice showed a decrease in cCFU-F to ~50% of nonirradiated controls (Figure 5A). After 3 months and 6 months, and at 5 days and 30 days post-MI (performed at 6 weeks posttransplantation), all large cCFU-Fs were derived from the non-BM, GFP^– fraction (Figures 5A–5C). Granulocyte colony stimulating factor (G-CSF) mobilizes HSCs and endothelial progenitor cells, and potentially also MSC-related cells (Lund et al., 2008). However, in mice that had received daily G-CSF administration for 5 days after MI, virtually all cCFU-F were from the non-BM-derived fraction, with only a single colony from donor cells (Figure 5A).

Open in a separate window

Figure 5

Large cCFU-Fs Are Not BM Derived

(A) Cardiac lead shielding during irradiation preserved large colony formation after BM transplantation. Large colony formation was from the non-BM-derived fraction after 3 and 6 months aging, MI (6 weeks posttransplant; cells isolated either 5 or 30 days after MI), or MI with G-CSF administration for 6 days.

(B and C) Crystal-violet-stained colonies from GFP^– and GFP⁺ fractions.

Error bars = SEM between independent experiments. See also Figure S5.

cCFU-Fs Are Distinct from cKIT⁺ CPCs

CPCs have been previously defined on the basis of expression of cKIT and a SCA1-like epitope (Linke et al., 2005). The majority of such cells are from BM and require cKIT function for mobilization and involvement in cardiac repair after MI (Fazel et al., 2006). We examined the relationship between cKIT⁺ cells and cCFU-F activity in the SCA1⁺ and S+P+ fractions in healthy, sham-operated and MI hearts (Figure S5B). There was a strong influx of cKIT⁺ cells after MI, the vast majority segregating to the SCA1⁺/PDGFRα^–/CD45⁺ population, meaning they were therefore of BM origin. cKIT⁺/CD45^– cells were absent or extremely rare within the SCA1⁺ and S+P+ fractions in healthy, sham-operated, and MI hearts, and neither cKIT⁺/CD45⁺ nor cKIT⁺/CD45^– fractions in MI hearts formed colonies. All large colonies came from the S+P+/cKIT^–/CD45^– population (quadrant 3; Figures S5B and S5C). However, within cCFU-F colonies expanded in vitro, some 3.3% of cells were cKIT⁺, and these cells were lost as cultures were induced to differentiate in low serum (Figure S5D). Therefore, while cCFU-Fs are distinct from the majority BM-derived cKIT⁺ population, it remains possible that there is some overlap between the cCFU-F and cKIT⁺ lineage trees in vivo, but only within a rare subset of the cKIT⁺ population.

Lineage Analysis of the Developmental Origins of cCFU-Fs

To investigate the embryonic origins of cCFU-Fs more formally, we used CRE-recombinase lineage tracing. A suite of lineage-specific CRE driver mice was crossed with the Z/EG transgenic reporter mouse that carries a ubiquitously expressed lacZ transgene (Table S1). After exposure to CRE, the lacZ cassette is lost, leading to expression from a GFP cassette. Lineage-CRE × Z/EG hearts were harvested at 8–12 weeks and FACS was used to isolate the cardiac S⁺P⁺ fraction. cCFU-F assays were performed with colonies scored at 12 days for both β-galactosidase (LACZ) and GFP (Figures 6A and 6B). In germ-line CMV-Cre × Z/EG progeny, 91.3% ± 1% of large colonies were GFP⁺/LACZ^–, the remainder being GFP^–/LACZ⁺, which is likely the result of insufficient CRE activity in rare cells (Figure 6C). Without CRE, 100% of the colonies were GFP^–/LACZ⁺, demonstrating the lack of ectopic GFP expression in this system (Figures 6B and 6C). Importantly, no GFP^–/LACZ^– colonies were observed in these or additional crosses, demonstrating a lack of transgene silencing.

Open in a separate window

Figure 6

Lineage Tracing Studies Suggest an Epicardial Origin for cCFU-Fs

(A) Overview of lineage tracing strategy.

(B) Expression of GFP and LACZ in colonies from control mice.

(C and D) Percentage of GFP⁺/LACZ^– or GFP^–/LACZ⁺ cCFU-F colonies from (C) adult whole hearts and (D) ventricles of 17.5 dpc lineage-CRE × Z/EG mice.

(E and F) Colonies derived from adult BM (E) or adult proximal aorta (F).

Error bars = SEM between independent experiments. See also Figure S6.

In the Mesp1-Cre line (Table S1), CRE is active in nascent mesoderm, but not in neuroectoderm, including neural crest or endoderm. In Mesp1-Cre × Z/EG progeny, most large colonies (81.3% ± 7%) were GFP⁺/LACZ^–, demonstrating a largely (if not exclusively) mesodermal origin (Figure 6C). Recent reports suggest that neural crest may give rise to some CPCs in the postnatal murine heart (Tomita et al., 2005) and may populate the proepicardium (Stottmann et al., 2004). We therefore used the Wnt1-Cre line in which CRE expression is neural crest specific (Table S1). However, all large colonies from Wnt1-Cre × Z/EG progeny were GFP^–/LACZ⁺ (Figure 6C).

Some adult CMs are capable of reentering the cell cycle and contributing to heart repair (Bersell et al., 2009). To investigate a CM origin for cCFU-Fs, we used Myl2-Cre knockin mice in which CRE expression occurs in ventricular myocytes from 8.5 dpc (Table S1). However, all colonies from Myl2-Cre × Z/EG progeny were GFP^–/LACZ⁺. Nkx2-5 is expressed from 7.5 dpc in CPCs, and subsequently, at the birth of all CMs, save for a minor population caudally (Christoffels et al., 2006). Virtually all colonies resulting from a cross with an Nkx2-5IRESCre knockin line (Table S1) were GFP^–/LACZ⁺ (89.3% ± 2%; Figure 6C), again arguing against a CM origin for cCFU-Fs. The few GFP⁺/LACZ^– colonies are discussed below.

Gata5-CRE Lineage Tracing Suggests an Epicardial Origin for cCFU-Fs

The Gata5-Cre transgenic line has been used previously to delete genes in the proepicardial/epicardial lineage (Martínez-Estrada et al., 2010; Mellgren et al., 2008; Merki et al., 2005; Zamora et al., 2007). In Gata5-Cre × Z/EG progeny, most large colonies (79.3% ± 5%) were GFP⁺/LACZ^– (Figure 6C). In our hands, using the Soriano R26R reporter (Table S1), CRE expression was first seen in proepicardium and in a few epicardial cells on the surface of the ventricles in 20 somite pair (sp) embryos (Figures S6A and S6B). Later (25sp–11.5 dpc), expression was seen more extensively in the epicardium, and also in substantial patches of myocardium and endocardium (Figures S6C–S6G). In the adult, expression was seen in myocardial patches, and in epicardium and its descendant lineages (Figures S6H–S6K).

As noted above, in Nkx2-5IRESCre × Z/EG progeny, virtually all colonies were GFP^–/LACZ⁺, consistent with a nonmyocyte origin for cCFU-Fs (Figure 6C). However, 10.7% ± 2% of colonies were GFP⁺/LACZ^–. This is most likely because Nkx2-5-driven CRE is also expressed transiently in a compartment of the early cardiac progenitor fields that encompasses precursors of the proepicardium (Ma et al., 2008; Zhou et al., 2008b).

The overall CRE lineage signature for cCFU-Fs seen above was also observed in isolated ventricles (devoid of endocardial cushion tissue) at 17.5 dpc (Figure 6D). We confirmed the lineage analyses using R26R as an independent reporter (Figure S6L) and also confirmed that Gata5-CRE was not activated as a result of in vitro culture (Figure S6M).

A Distinct CRE Lineage Signature for BM and Aortic CFU-Fs

CRE lineage signatures for bmCFU-F and aortic CFU-Fs (aCFU-Fs) were distinct from those of cCFU-Fs (Figures 6E and 6F). With CRE lines used here, all bmCFU-F colonies were GFP^–/LACZ⁺. aCFU-F colonies had a multi-germ layer origin, mostly derived from mesoderm (76.0% ± 4% from a Mesp1-Cre lineage), with one fifth from neural crest (22.0% ± 5% from a Wnt1-Cre lineage), consistent with the known contributions of both heart field mesoderm and neural crest to the mural compartment of branchial vessels.

Confirmation of an Epicardial Origin of cCFU-Fs using a Conditional Wt1-Cre Line

To confirm an epicardial origin for cCFU-Fs, CRE lineage tracing was performed using a tamoxifen-inducible CRE driver (Wt1^CreERT2) regulated by cis-elements of Wt1, expressed in proepicardium and epicardium (Zhou et al., 2008a). We used a tamoxifen regime (2 mg intraperitoneally daily from 9.5–11.5 dpc) gauged to maximize CRE-ERT2 activity in the forming epicardium, and the moderately sensitive CRE-dependent reporter R26R (Table S1). LACZ expression in tamoxifen-treated Wt1^CreERT2 × R26R progeny at 11.5–14.5 dpc was seen in epicardium, albeit in a patchy manner (Figures 7A–7E). Pericardium was also positive, as were interstitial cells including perivascular cells and fibroblasts (Figure 7E). However, in contrast to crosses with Gata5-Cre, endocardium, endocardial cushions, and myocardium were negative (Figures 7B–7E). LACZ⁺ cells were also seen at a number of extracardiac sites (Figures S7A–S7H), overlapping with the known expression sites of WT1 (Martínez-Estrada et al., 2010).

Open in a separate window

Figure 7

Conditional Lineage Tracing with WT1^CreERT2

(A–E) LacZ staining in 11.75 dpc embryonic hearts from Wt1^CreERT2 × R26R crosses shows patchy labeling of epicardium, interstitial and interventricular groove (ivg) cells, and pericardium (p) (arrowheads).

(F and G) Percentage of LACZ⁺ or LACZ^– cCFU-F large colonies from Wt1^CreERT2 × R26R progeny at 14.5 dpc (F) and adult whole hearts (G).

(H) Percentage of GFP⁺/LACZ^– cCFU-F large colonies from Gata5Cre × Z/EG adult mice was not significantly different 5 or 30 days post-MI compared to sham-operated controls.

Error bars = SEM between independent experiments. See also Figure S7.

In Wt1^CreERT2 crosses at 14.5 dpc, 19.3% ± 4% of cCFU-Fs were LACZ⁺, and these were tamoxifen-dependent (Figure 7F). Using our tamoxifen regime, fetuses were reabsorbed from 15.5 dpc onward, although coinjecting tamoxifen and progesterone (used to offset tamoxifen toxicity) yielded nine pups that survived to adulthood after Caesarian section and cofostering, five of which were of the relevant genotype. In the two tamoxifen-treated mice, 19.5% ± 6% of colonies were LACZ⁺ (Figure 7G). These data strongly support an epicardial origin for cCFU-Fs. While neither the Gata5-Cre nor Wt1^CreERT2 lineage tags were epicardial specific, only the epicardium, pericardium, and liver capsule and mesenchyme were uniquely at the intersection of their respective expression domains (Figure S7H).

Gene/Protein Expression

Probes and antibodies and methods are as described in Table S2.

Cell Isolation and Culture

Mononuclear cells were isolated from dissected hearts and aorta via mincing of tissue before digesting it in 0.1% Collagenase type II (Worthington) in PBS at 37°C for 30 min, and having it be triturated mechanically at 10 min intervals and with myocyte debris removed with a 40 μm filter. Dead cells were removed using MACS Dead Cell Removal kit (Miltenyibiotec) before incubation with fluorophore-conjugated antibodies (Table S1) and sorting using FACS Aria or FACS Canto (Becton and Dickinson) cytometers. For CFU-F assays, 5,000 S+P+ cells were plated on tissue culture plastic or Poly-L-lysine-coated coverslips and cultured in αMEM (GIBCO) plus 20% FCS for 12 days.

BM Transplantation and Irradiation

Mice were irradiated with 9 Gy using an X-RAD 320 (Precision Xray Inc, USA) irradiator. For BM reconstitution, 2.5 × 10⁷ fresh GFP-tagged BM cells from Bag1 donors (Table S1) were injected in an aseptic manner into the tail vein. After 4–6 weeks, reconstitution was assessed. Lead shielding was 3 mm.

We thank S. Ku, K. Evans, and M. Pinese for technical assistance, and Y. Saga, K. Chien, P. Soriano, P. Ruiz-Lozano, and C. Lobe for mice. Research was supported by the Victorian State Government Operational Infrastructure Support scheme and grants from the Australian Stem Cell Centre (PO82; S4M1), National Health and Medical Research Counsel Australia (IRIISS; 354400; 573705; 573707; and 535903), Australia-India Strategic Research Fund (BF020084), Johnson and Johnson (Focused Giving Program), Atlantic Philanthropies (Flagship Scheme; 19131), Cancer Institute New South Wales (09/CDF/2-40 and 10/CRF/1-01), Australian Cancer Research Foundation, Royal Australian College of Surgeons, The Avner Nahmani Pancreatic Cancer Foundation, and the R. T. Hall Trust. J.C. held a National Heart Foundation of Australia (NHF) scholarship, and O.W.J.P., an NHF Career Development Award (CR 08S3958).