Nf1 limits epicardial derivative expansion by regulating epithelial to mesenchymal transition and proliferation

The epicardium, which comprises the outer epithelial layer of the heart, is a cell population that undergoes epithelial to mesenchymal transition (EMT) during development (Lie-Venema et al., 2007). Around embryonic day (E) 13.5, a subset of epicardial cells lose their epithelial characteristics, gain mesenchymal properties and migrate into the heart to differentiate into coronary vascular smooth muscle cells (cVSMCs) and cardiac fibroblasts (Dettman et al., 1998; Manner et al., 2001; Mikawa and Gourdie, 1996). Several growth factors, including transforming growth factor β (TGFβ) (Mercado-Pimentel and Runyan, 2007; Xu et al., 2009b) and fibroblast growth factor (FGF) (Pennisi and Mikawa, 2009) have been implicated in the EMT process of epicardial cells during heart development, but little is understood about signals that limit the EMT process. Identification of such pathways will provide insights into the complex regulation of EMT during heart development. Because many of these same signaling pathways have also been suggested to play a key role in cardiac fibrosis, this might also provide insights into pathological EMT.

Recent findings show that disruption of Ras-mitogen activated protein kinase (MAPK) signaling results in several syndromes that exhibit congenital heart defects, including the Costello (Aoki et al., 2005), LEOPARD (Kontaridis et al., 2006), cardio-facio-cutaneous (Niihori et al., 2006) and Noonan (Schubbert et al., 2006) syndromes. Loss of neurofibromin 1 (Nf1; also known as neurofibromatosis-related protein NF-1), a Ras-GTPase activating protein (GAP), leads to hyperactivation of Ras and its downstream components (Cichowski and Jacks, 2001; Martin et al., 1990; Xu et al., 1990). Although mutations in NF1 are best known for causing neurofibromatosis type 1 tumors of the skin and nervous system (Lynch and Gutmann, 2002), patients with NF1 mutations also have an increased risk for cardiovascular disorders (Lin et al., 2000).

Studies in mice have significantly advanced our understanding of Nf1 function during heart development. Inactivation of Nf1 causes lethality at mid-gestation, with severe heart defects including malformation of the outflow tract, a thinned myocardium, a ventricular septal defect and enlarged endocardial cushions (Brannan et al., 1994; Jacks et al., 1994). Loss of Nf1 in vascular smooth muscle cells (VSMCs) leads to an abnormal proliferative injury response (Xu et al., 2007), and cardiomyocyte-specific inactivation of Nf1 results in pathological hypertrophy and heart failure in adult mice (Xu et al., 2009a). Nf1-null endocardial cushion cells exhibit abnormal EMT (Lakkis and Epstein, 1998), and endothelial-specific deletion of Nf1 recapitulates many of the cardiovascular defects of the Nf1-null mouse, suggesting an indispensable role of Nf1 in endothelial cells during EMT (Gitler et al., 2003).

We have investigated the function of Nf1 in epicardial development using Cre/loxP technology to inactivate Nf1 in the mouse epicardium. We found that loss of Nf1 results in increased EMT and epicardial-derived cell (EPDC) proliferation, leading to a substantial expansion of this cell population that includes cardiac fibroblasts and cVSMCs. Our data point to a regulatory role for Nf1 in the process of EMT and suggest the possibility that patients with disruption of NF1 might be more prone to cardiac complications such as fibrosis and coronary artery disease.

Mice were maintained on a mixed C57/BL6 × 129SV background. Mice with the Gata5-Cre transgene (Merki et al., 2005) or Wt1^GFPCre allele (Zhou et al., 2008) were crossed with mice with the Nf1 floxed (Nf1^fl) allele (Zhu et al., 2001) to generate Nf1^fl/fl;Gata5-Cre^Tg (designated Nf1^G5KO) and Nf1^fl/fl;Wt1^GFPCre, respectively. Controls were Cre-negative littermates (Nf1^fl/fl or Nf1^fl/+) unless otherwise indicated. For epicardial tracing experiments, male Nf1^fl/+;Wt1^CreERT2/+ mice were crossed with Nf1^fl/fl or Nf1^fl/+ female mice with ROSA26R^lacZ (designated R26R^lacZ) (Soriano, 1999) or ROSA26R^tdTomato (designated R26R^T) (Madisen et al., 2010) reporter alleles to generate Nf1^fl/fl;Wt1^CreERT2/+ (designated Nf1^WTiKO). Wt1^CreERT2 was induced by oral administration of tamoxifen (MP Biomedicals, 02156738) to pregnant females at the indicated embryonic stages and induction efficiency was traced by R26R^lacZ or R26R^T reporter gene expression. Tamoxifen was dissolved in sunflower seed oil (Sigma) at 20 mg/ml and administrated at a final concentration of 0.1 mg per gram body weight. Other strains in these experiments include Pdgfra^fl (Tallquist et al., 2003) and K-Ras(G12D)^fl mice (JAX stock number 008180) (Jackson et al., 2001). All animal protocols were approved by the Institutional Animal Care and Use Committee of the University of Texas Southwestern Medical Center and conform to NIH guidelines for care and use of laboratory animals. Nf1^fl (Zhu et al., 2001) and Gata5-Cre^Tg (Merki et al., 2005) mice were kindly provided by Dr Luis Parada (University of Texas Southwestern, TX, USA) and Dr Pilar Ruiz-Lozano (Sanford-Burnham Institute, CA, USA), respectively. Wt1^GFPCre and Wt1^CreERT2 mice (Zhou et al., 2008) were kindly provided by Dr William Pu (Harvard, MA, USA) and X-LacZ4^Tg mice (Tidhar et al., 2001) were kindly provided by Dr Moshe Shani (Volcani Center, Israel).

In vitro culture of epicardial cells was described previously (Mellgren et al., 2008). Ventricles from E12.5 hearts were isolated and cultured in 1:1 DMEM:M199 with 15% FBS supplemented with glutamate, antibiotics and basic fibroblast growth factor (2 ng/ml, Sigma). After 3 days, heart explants were removed and cells were cultured for 2 additional days in media with reduced serum (10%). The epicardial nature of the culture was confirmed by the expression of epicardial gene reporters: Tcf21^lacZ (Lu et al., 1998) or Pdgfra^GFP (Hamilton et al., 2003; Smith et al., 2011) (data not shown). Cre-expressing adenovirus, kindly provided by Dr Robert Gerard, University of Texas Southwestern, TX, USA), was added to the culture as indicated. For RNA extraction followed by quantitative real-time PCR (qRT-PCR) analysis, hearts were cultured in 24-well plates. Loss of Nf1 in Nf1^G5KO cultures was confirmed by qRT-PCR. For immunostaining, hearts were placed on glass coverslips coated with collagen type IV (5 μg/cm², R&D Systems). Collagen-coated coverslips were prepared according to the manufacturer’s protocol. For the in vitro differentiation assay, epicardial cells were cultured for a total of 6 days followed by immunostaining.

Tissues and embryos were fixed in 4% paraformaldehyde (PFA) overnight at 4°C, washed in PBS, dehydrated and paraffin embedded. For histological analysis, tissues were sectioned to 8 μm, rehydrated and stained with Hematoxylin and Eosin (Sigma) as previously described (Mellgren et al., 2008). For immunostaining, antigens were retrieved in citrate buffer (pH 6.0) at 98°C for 15 minutes using a temperature-controlled microwave (BioGenex). For frozen embedding, hearts were fixed in 4% PFA for 1 hour at 4°C then embedded in OCT (Tissue-Tek). Hearts were sectioned at 10 μm, permeabilized with 0.1% Triton X-100 in PBS (PBT), blocked with 5% serum in PBT and stained with phalloidin (1:200; Invitrogen, A12379) or with antibodies against vimentin (1:500; Sigma, V6630), SM-MHC (1:250; Chemicon, MAB3572), phospho-histone H3 (1:200; Upstate, 06-570), collagen IV (1:250; Chemicon, AB748), β-catenin (1:500; BD Bioscience, 610153) and α-catenin (1:100; Abcam, AB51032). Immunostaining for Wt1 (1:50; DAKO, M3561) was performed with the Vectastain mouse ABC kit followed by detection by DAB (Vector Labs).

For the detection of β-galactosidase activity, hearts were fixed in 2% formaldehyde/0.2% glutaraldehyde in PBS for 15 minutes. Hearts were then washed with PBS and stained whole-mount with X-Gal (Gold Biotechnology, X4281C) or frozen embedded and sectioned, followed by staining as described previously (Acharya et al., 2011).

For whole-mount immunostaining, hearts were isolated and fixed in 4% PFA, then permeabilized with PBT for 30 minutes, blocked with CAS Block (Invitrogen, 008120) for 30 minutes, and immunostained for β-catenin (1:200). z-stack images (nine consecutive 4 μm images spanning 32 μm in depth) were taken from similar regions of the left ventricle of embryonic hearts starting from the epicardial surface using an LSM 510 META mounted on an Axiovert 200M fluorescence microscope (Zeiss). Three-dimensional images were reconstituted using ImageJ software, and the total number of cells expressing R26R^T within a 150 μm × 150 μm × 32 μm area were counted in the subepicardial space.

The collagen gel invasion assay was performed as described (Boyer et al., 1999; Potts et al., 1991) with modifications. E12.5 hearts were isolated, and ventricles were placed and cultured on 1.6% collagen (Roche) gels. Collagen gels were prepared according to the manufacturer’s instructions. After 3 days, heart explants were removed and cultured for 3 more days before fixing in 4% PFA for 10 minutes for analysis. To detect invasion, collagen gels were frozen embedded and sectioned, followed by staining with DAPI (Roche) and phalloidin. Invasion was identified by the presence of epicardial cells underneath the collagen gel surface. For quantification, invading cells were identified using a fluorescence microscope (Zeiss Axiovert 200M with a Hamamatsu ORCA-ER camera) and invasion was calculated by imaging from the center of each culture along a line to where cells had left the plane of the collagen gel (the invasion front). For quantification, the number of cells that had invaded the gel was divided by the total number of cells within the 20× field of view from three different images (Merki et al., 2005).

The ex vivo migration assay was performed as described (Mellgren et al., 2008). E12.5 hearts were isolated and incubated with adenovirus expressing GFP or Nf1 GAP-related domain (Miller et al., 2010), kindly provided by Dr Robert Gerard or Dr Nancy Ratner (Cincinnati Children’s Hospital, OH, USA), respectively. PDGF-BB (R&D Systems), imatinib mesylate (Sigma), AG1296 (Sigma) and U0126 (Sigma) were added to the cultures as indicated. After 2 days, hearts were fixed in 4% PFA and frozen embedded, sectioned and stained for DAPI. For quantification, GFP⁺ cells underneath the epicardium were counted in a 40× field of view from five nonconsecutive sections.

For mesenchymal index, primary cultured epicardial cells were stained for β-catenin and phalloidin as described above, and cells exhibiting a mesenchymal morphology were identified by the loss of adherens junctions and cortical actin and by the robust formation of actin stress fibers (Sridurongrit et al., 2008). Mesenchymal cells were counted and divided by the total number of cells in a 40× field of view from three different regions of the cultures. β-galactosidase staining was quantified as previously described (Morgan et al., 2008). Epicardial differentiation of smooth muscle cells was quantified following immunostaining against SM-MHC. The SM-MHC-positive area was measured and divided by the DAPI positive area in three 40× field-of-view images using ImageJ (NIH). All experiments used a minimum of two independent litters, and data were analyzed by Student’s t-test using Prism 5 (GraphPad Software).

Primary epicardial cells were collected as described above. RNA isolation and cDNA synthesis were as described previously (Mellgren et al., 2008) with slight modifications. Briefly, primary epicardial cultures from three hearts of each genotype were combined followed by RNA isolation using Trizol (Invitrogen). cDNA was synthesized using Superscript III reverse transcriptase (Invitrogen). Gene transcription was analyzed by standard qRT-PCR with iTAQ SYBR Green Master Mix (Bio-Rad) using the CFX96 real-time PCR detection system (Bio-Rad). Each sample was run in triplicate. Sequences of primers are as reported previously (Smith et al., 2011).

Section in situ hybridization was performed as described previously (Schaeren-Wiemers and Gerfin-Moser, 1993; Smith et al., 2011). Briefly, embryonic hearts were isolated, fixed in 4% PFA, frozen embedded and sectioned at 16 μm. Sections were digested with proteinase K (15 μg/ml; Fisher Scientific, BP1700-100) followed by brief fixation with 4% PFA and acetylation by acetic anhydride before hybridizing with digoxigenin-labeled RNA probes for Pdgfra (Bostrom et al., 1996), Col1a1, Col3a1 and Nf1. Sections were then immunostained for digoxigenin (1:2000; Roche, 11093274910) followed by development with BM purple (Roche, 11442074001). Plasmids for Col1a1 and Col3a1 probes were kindly provided by Benoit de Crombrugghe (MD Anderson Cancer Center, TX, USA). The plasmid for the Nf1 probe included a 338 bp fragment of the Nf1 3′ untranslated region corresponding to positions 9881-10,218 (GenBank L10370.1).

Loss of Nf1 in cardiomyocytes or the endocardial cushions results in heart abnormalities (Gitler et al., 2003; Xu et al., 2009a), but no reports have addressed the disrupted epicardium observed in Nf1-null hearts (Brannan et al., 1994). To determine whether Nf1 has a primary role in epicardial development, we performed in situ hybridization for Nf1 transcripts. Nf1 expression was detected in the epicardium, endocardium, endocardial cushions and myocardium at E11.5, but by E12.5 myocardial expression was decreased (supplementary material Fig. S1). Epicardial expression continued until E13.5, but by E14.5-15.5 was limited to a few cells in the epicardium (Fig. 1A; supplementary material Fig. S1).

To investigate Nf1 function in epicardial development, we initially used two mouse lines with constitutive expression of Cre in the epicardium: the Gata5-Cre^Tg (Merki et al., 2005) and the Wt1^GFPCre (Zhou et al., 2008) mouse lines. We monitored loss of Nf1 transcript by Gata5-Cre^Tg-driven recombination (referred to as Nf1^G5KO) and found little expression in the epicardium and endocardial cushions at E13.5 (Fig. 1B). Using Wt1 protein expression to track the epicardium and undifferentiated EPDCs (Moore et al., 1999) we found that, unlike control hearts in which Wt1⁺ cells were restricted to the epicardium at E12.5, Wt1⁺ cells in Nf1^G5KO hearts were detected in not only the epicardium but also the subepicardial zone (Fig. 1C). To determine whether these cells had adopted a mesenchymal phenotype, we stained for vimentin, a marker for mesenchymal cells that often indicates that a cell has undergone the process of EMT (Perez-Pomares et al., 1997). At E12.5, control hearts had few vimentin⁺ cells in the subepicardium (Fig. 1D), whereas in Nf1^G5KO hearts there were multiple patches of vimentin⁺ cells in the subepicardium. These patches were often concomitant with a disrupted basement membrane (collagen IV staining; Fig. 1D). Similar patches of vimentin⁺ cells were also found when Nf1 expression was disrupted using a Wt1^CreGFP allele for recombination (Fig. 1D).

Based on ROSA26 reporter activity, both of these epicardial Cre lines recombine in a significant number of cardiomyocytes and in the endocardial cushions (data not shown). Therefore, we performed the remaining in vivo experiments employing the inducible epicardial-specific Cre mouse Wt1^CreERT2 (Zhou et al., 2008). First, we confirmed the fidelity of recombination in this line to demonstrate epicardial-specific Cre activity with tamoxifen induction at E10.5 and E12.5. Single administration of tamoxifen at E10.5 resulted in reporter gene (R26R^T) expression in ∼95% of epicardial cells after 24 hours (data not shown). Lineage tracing and in situ hybridization for Nf1 demonstrated that in Nf1 conditional embryos transcripts were reduced in the epicardium just 24 hours after induction (Fig. 1B), that epicardial cells were exclusively tagged, and that lineage-tagged cells migrated into the heart ventricles as expected (supplementary material Fig. S2A). The only other lineage-tagged regions were the atrioventricular valves, where epicardial cell contribution has been reported previously (de Lange et al., 2004). By contrast, at these time points of induction no lineage-tagged cells were detected in the cardiomyocyte population, nor in the semilunar valves (supplementary material Fig. S2B,C).

To specifically examine the role of Nf1 in the epicardium, we generated Nf1^fl/fl;Wt1^CreERT2/+ mice (referred to as Nf1^WTiKO). We obtained the expected Mendelian ratios of animals and detected no overt phenotype in Nf1^WTiKO animals, suggesting that epicardial inactivation of Nf1 by Wt1^CreERT2 at E12.5 did not cause embryonic nor postnatal lethality (data not shown). Because we observed vimentin⁺ cells in the ventricles at time points earlier than expected (Fig. 1D), we examined whether premature EMT occurred upon loss of Nf1 by tracing the migration of epicardial cells labeled at E10.5. These hearts revealed that a substantial number of epicardial cells were present immediately below the basement membrane in Nf1^WTiKO hearts, whereas tagged cells remained in the epicardium in controls (Fig. 1E). These data suggested that, in the absence of Nf1, epicardial cells migrate into the heart earlier than expected in all three genotypes examined (Nf1^G5KO, Nf1^fl/fl;Wt1^CreGFP/+ and Nf1^WTiKO).

From the above data, we hypothesized that loss of Nf1 could result in accelerated EMT. First, we tested this possibility in vitro. We generated primary cultured epicardial cells from E12.5 hearts, which uniformly expressed the epicardial genes Tcf21 and Pdgfra (data not shown). After 3 days of culture, without any exogenous stimulus the control epicardial cells remained a cobblestone monolayer, whereas Nf1^G5KO epicardial cells exhibited a mesenchymal morphology (Fig. 2A). We investigated two hallmarks of EMT, namely the loss of cell-cell contacts and the formation of actin stress fibers, by localization of β-catenin and filamentous actin, respectively. Control epicardial cells maintained cell-cell contacts, had extensive cellular junctions and exhibited cortical actin. However, Nf1^G5KO epicardial cells formed extensive actin stress fibers and lost their junctions (Fig. 2B). These changes were inhibited by the Rho-associated protein kinase inhibitor Y27632 (Uehata et al., 1997), suggesting that known EMT signaling pathways were occurring in the Nf1 mutant cultures (data not shown). These results also demonstrated a cell-autonomous role for Nf1 in regulating EMT.

Although morphology is commonly used as a readout for EMT, changes in gene expression profiles from epithelial to mesenchymal can also be used to detect the transition. We measured the expression of epithelial markers, such as Krt14 (Chamulitrat et al., 2003; Ke et al., 2008) and Bves (Wada et al., 2001), by qRT-PCR (Fig. 2C). Consistent with a switch from epithelial to mesenchymal cell type, we found that epithelial gene expression was downregulated in Nf1^G5KO cultures. By contrast, mesenchymal gene expression, as indicated by Col7a1 (Vindevoghel et al., 1998), Mmp10 (Wilkins-Port and Higgins, 2007), Sox9 (Cheung et al., 2005; Sakai et al., 2006) and Opg (Tnfrsf11b – Mouse Genome Informatics) (Corallini et al., 2009; Sakata et al., 1999; Vidal et al., 1998), was upregulated (Fig. 2C). We also observed an increased level of mesenchymal gene expression in heterozygous cultures (data not shown).

An additional criterion for transition from an epithelial to a mesenchymal phenotype is invasion into a collagen gel (Thiery and Sleeman, 2006). In a collagen gel assay (Boyer et al., 1999; Potts et al., 1991), ∼12% of control epicardial cells invaded the collagen gel along the edge of the culture (Fig. 2D,F). In Nf1^G5KO epicardial cultures, ∼60% of the cells in the same perimeter of the epicardial culture invaded the collagen gel and formed actin stress fibers (Fig. 2D,F).

When epicardial cells undergo EMT they differentiate predominantly into two cell types: cVSMCs and cardiac fibroblasts (Mikawa and Gourdie, 1996; Vrancken Peeters et al., 1999). To determine whether the increased EMT led to an increase in differentiated cells in vitro, we examined the expression of SM-MHC, a smooth muscle cell marker. Epicardial cultures from Nf1^G5KO and to a lesser extent Nf1 heterozygous hearts exhibited an increased number of SM-MHC-expressing cells compared with control cultures (Fig. 2E,G).

In summary, Nf1 mutant epicardial cells spontaneously lost epithelial characteristics and adopted a mesenchymal phenotype, including an increase in mesenchymal gene expression, invasiveness and differentiation. It should be noted that loss of Nf1 resulted in EMT under basal culture conditions. Therefore, these cultured epicardial cells might be poised to undergo EMT, and signaling by Nf1 could be a key regulatory pathway inhibiting this process.

To determine whether the loss of Nf1 has a direct effect on EMT, we induced temporal deletions of Nf1 between stages E10.5 and E12.5. These embryos therefore had wild-type expression of Nf1 until just before the stage of EMT. This tracing resulted in efficient R26R^T reporter expression in a high percentage of the epicardium (Fig. 3A). We then quantified the number of EPDCs that had migrated into the heart ventricle using whole-mount confocal microscopy at each time point. Fig. 3 is representative of optical sections comparing control and Nf1^WTiKO hearts (see supplementary material Fig. S3 for images at other time points). At E11.5, very few R26R^T-positive cells were detected in the control myocardial compartment (Fig. 3A,B; supplementary material Fig. S3). Tagged epicardial cells were observed in the ventricle at E12.5 and further increased at E13.5, suggesting that EMT began at ∼E12.5 in control hearts (Fig. 3B; supplementary material Fig. S3). However, in Nf1^WTiKO hearts, at each time point examined a greater number of R26R^T-positive cells was detected within the myocardial compartment (Fig. 3A,B; supplementary material Fig. S3). The existence of migrated mutant epicardial cells at E11.5 suggested that epicardial EMT occurs earlier in the mutant hearts, and the increase in migrated epicardial cells demonstrated that, as in vitro, loss of Nf1 results in an increased number of cells undergoing EMT. Because confocal imaging only allowed us to examine a small region and time window of migrated epicardial cells, we also traced EPDCs in heart sections using two different reporters. A similar increase in R26R^T-positive or R26R^lacZ-positive EPDCs was detected at E13.5 and E14.5, respectively, upon loss of Nf1 at E12.5 in both the right and left ventricle (Fig. 4A-D).

To determine how loss of Nf1 impacted the proliferation and survival of epicardial cells and EPDCs at later stages of development, we inactivated Nf1 at E12.5 and quantified the number of proliferating (phospho-histone H3⁺) cells within the R26R^T-positive epicardial and EPDC population. At E13.5 we saw a modest increase in proliferation of Nf1-deficient epicardial cells (Fig. 4E). At the same stage, a similar number of the EPDCs in both control and mutant hearts were in mitosis (Fig. 4E). However, at later stages, Nf1-deficient EPDCs exhibited increased proliferation (Fig. 4E). Because alterations in cell survival have been reported in Nf1-deficient endocardial cushions (Lakkis and Epstein, 1998), we examined Nf1^WTiKO and Nf1^G5KO hearts for apoptosis, using an antibody for cleaved caspase 3, at various time points from E12.5 to P0. No differences in apoptotic cell numbers were observed between control and mutant hearts (data not shown).

Because enhanced epicardial cell EMT and EPDC proliferation were observed, we reasoned that there might be an expansion of EPDCs. As cardiac fibroblasts and cVSMCs are the predominant populations of cells derived from the epicardium, we determined how loss of Nf1 impacted these cells. Using in situ hybridization for three genes that identify cardiac fibroblasts, namely Col1a1, Col3a1 and Pdgfra (Smith et al., 2011), we found an increased number of cardiac fibroblasts in mutant hearts compared with controls (Fig. 5A,C). Collagen I is also expressed by some VSMCs (Ponticos et al., 2004) and therefore it is likely that this particular probe overestimated the number of fibroblasts, but the data clearly demonstrated an increase in non-vessel-associated Col1a1-expressing, as well as Col3a1- and Pdgfra-expressing, cells. This increase in cell numbers was not restricted to the cardiac fibroblast lineage. We utilized the X-LacZ4^Tg mouse (Tidhar et al., 2001) that expresses a nuclear-localized β-galactosidase in VSMCs and efficiently tags cVSMCs (Mellgren et al., 2008). We found that loss of Nf1 also resulted in an expansion of the VSMC lineage (Fig. 5B). Not only were more cVSMCs detected at E17.5, but the increase also appeared to lead to an extended and more highly branched cVSMC-coated network of coronary vasculature (Fig. 5D; data not shown).

It is established that loss of Nf1 leads to prolonged activation of the Ras-MAPK pathway in cardiomyocytes and VSMCs (Cichowski and Jacks, 2001; Xu et al., 2009a; Xu et al., 2007). To determine whether activation of Erk1/2 (Mapk3/1 – Mouse Genome Informatics) is responsible for EMT in Nf1-deficient epicardial cells, we inhibited the MAP kinase pathway and measured EMT by ex vivo migration assay (Mellgren et al., 2008). The epicardium of E12.5 hearts was labeled by adenoviral GFP transduction, and migration of GFP-expressing epicardial cells into the myocardium was quantified. In control hearts, GFP⁺ cells were restricted to the epicardium, whereas Nf1^G5KO hearts possessed an increased number of GFP⁺ cells within the myocardium, suggesting an enhanced ability of Nf1-null epicardial cells to leave the epicardial layer (Fig. 6A,B). The increased migration was abolished when hearts were cultured in the presence of U0126, an inhibitor of both Mek1 and Mek2 (Map2k1 and Map2k2 – Mouse Genome Informatics) (Fig. 6A,B). These data suggest that activation of ERK is responsible for EMT in Nf1-deficient epicardial cells.

Loss of Nf1 alone does not lead to extended activation of Ras. Upstream signals are required to initiate Ras signaling, then in the absence of Nf1, Ras remains in its active state (McCormick, 1995). We have recently reported that PDGF receptor signaling is an essential component of epicardial EMT. Pdgfrα and Pdgfrβ are expressed in the epicardium, and inactivation of these receptors in epicardial cells disrupts EMT (Mellgren et al., 2008; Smith et al., 2011). To determine whether PDGF signaling could be one pathway upstream of Ras-Nf1 signaling, we inhibited PDGF receptor tyrosine kinase activity in Nf1^G5KO epicardial cells. Imatinib mesylate, a potent inhibitor of both Pdgfrα and Pdgfrβ, inhibited the EMT phenotype caused by loss of Nf1 in the epicardial culture EMT assay (data not shown) and the ex vivo migration of epicardial cells (Fig. 6A,B). Similar results were obtained using AG1296, another inhibitor of the PDGF receptors (data not shown).

Next, we tested whether Nf1 Ras-GAP activity could negatively regulate PDGF-induced EMT in the ex vivo migration assay. Stimulation with recombinant PDGF-BB induced epicardial cell migration into the myocardium; however, adenoviral transduction of the Nf1 GAP-related domain (Nf1-GRD) (Hiatt et al., 2001; Miller et al., 2010) significantly reduced PDGF-BB-induced EMT (Fig. 6C,D). Conversely, we determined whether activation of Ras induced epicardial EMT using epicardial cultures from K-Ras(G12D)^fl/+ embryos. This transgene expresses a Cre-inducible oncogenic form of Kras (Jackson et al., 2001). Whereas control cultures had intact cellular junctions with cortical actin, when K-Ras(G12D) expression was induced the epicardial cultures formed actin stress fibers and lost cellular junctions, similar to Nf1^fl/fl cultures (Fig. 6E). Similarly, K-Ras^fl/+;Wt1^CreERT2/+ hearts had an increased number of Wt1⁺ cells in the myocardial compartment. This suggests that activation of Ras also results in increased epicardial EMT in vivo (Fig. 6F,G).

Finally, we determined whether loss of Pdgfrα signaling could partially rescue the excess EMT observed in the Nf1^WTiKO heart. As loss of Pdgfrα specifically affects only cardiac fibroblast progenitor EMT (Smith et al., 2011), we predicted that loss of Pdgfrα in an Nf1^WTiKO mutant background would lead to a reduction in EPDCs entering the epicardium as compared with an Nf1^WTiKO mutant that possessed Pdgfrα signaling. Consistent with our previous data, Nf1^WTiKO hearts had more Wt1⁺ cells within the myocardium than wild-type controls; however, simultaneous inactivation of both Pdgfra and Nf1 resulted in a significant reduction of migrated Wt1⁺ cells (Fig. 6F,G). One reason for the partial rescue of the Nf1^WTiKO EMT phenotype could be the presence of VSMC progenitors, which still express Pdgfrβ (Mellgren et al., 2008; Smith et al., 2011) and should continue to have excess Ras signaling due to loss of Nf1.

In conclusion, our results show that Nf1 is a key regulator of epicardial EMT and that this increased EMT as well as an increased rate of proliferation result in expansion of cardiac fibroblasts and VSMCs.

EMT is an essential process that plays a significant role in embryogenesis during gastrulation, heart development and neural crest cell formation (Thiery et al., 2009). There is now very compelling evidence that EMT is an essential component of tumor metastasis (Thiery et al., 2009; Yang and Weinberg, 2008). Because the best-known activity for Nf1 is its GAP activity, it is assumed that loss of Nf1 leads to abnormal Ras signaling. A further link with EMT can then be drawn because Ras signaling can induce EMT. One mechanism is by cooperating with TGFβ to promote Snail transcriptional activity (Horiguchi et al., 2009; Janda et al., 2002). Another is by activating MAPK and Rac, potentially leading to disruption of epithelial junctions (Edme et al., 2002). In fact, many of the EMT-inducing abilities of epidermal growth factor and hepatocyte growth factor have been directly linked to Ras activity (Boyer et al., 1997; Herrera, 1998). Here, we provide evidence that loss of Nf1 increases EMT in mouse epicardial cells, suggesting a possible regulatory role for Nf1 in Ras-driven EMT.

Interestingly, loss of Nf1 does not lead to persistent EMT. Instead, the EMT we observed is only amplified by occurring earlier and more robustly, as might be expected by its downstream signaling role. Nf1 does not initiate signaling (McCormick, 1995). Other factors must lie upstream to activate the Ras-MAPK pathway. Epicardial EMT is distinct from neural crest cell EMT in that only a subset of cells becomes mesenchymal, suggesting that the inductive signal is regionally localized and controlled temporally, thus providing an explanation for why epicardial EMT in the absence of Nf1 is still partially restricted. Indeed, we have shown that by inhibiting one of these potential upstream growth factor pathways, i.e. PDGF (Mellgren et al., 2008; Smith et al., 2011), we can block the effects of loss of Nf1. Interestingly, increased neointima formation in Nf1 heterozygous mice can also be mitigated by imatinib treatment, suggesting a possible role for PDGF signaling in the exaggerated vascular injury response that occurs upon the loss of Nf1 (Lasater et al., 2008). Because inhibition of MAPK also blunted the effect of loss of Nf1 on EMT, it is likely that Nf1 attenuates these inductive signals only in the epicardial cells that have been stimulated to undergo EMT.

Cardiovascular disease is a frequent cause of death in patients with neurofibromatosis 1 who are less than 30 years old (Rasmussen et al., 2001). Although some of this lethality is attributed to congenital abnormalities (Friedman et al., 2002; Lin et al., 2000), our findings also point to the possibility that an increase in the proliferation of epicardial-derived noncardiomyocyte lineages might also contribute to some of the heart abnormalities. Loss of Nf1 did not lead to excessive overgrowth of these cells under normal circumstances. It is likely that local environmental cues ultimately determine the differentiation and survival of EPDCs. For example, endothelial cells, which secrete PDGF ligands, are important regulators of cVSMC migration and proliferation (Tomanek, 2005). Similarly, local limitations of growth factor production by these cells might account for the lack of excessive cVSMC proliferation. Therefore, loss of Nf1 results in a controlled expansion of EPDCs rather than massive hyperplasia. This phenomenon would be reminiscent of what occurs with Nf1-mediated tumorigenicity, in which disruptions in the microenvironment are necessary for tumor progression (Zhu et al., 2002). Nonetheless, because loss of Nf1 is often linked to increased proliferation (Lynch and Gutmann, 2002) during a pathological response to heart injury, Nf1-deficient cells might respond more robustly by enhanced proliferation and fibrotic activity. Most neurofibromatosis 1 patients would be haploinsufficient in somatic cells, but our data and those of others suggest that even cells with reduced Nf1 protein might have elevated levels of GTP-bound Ras, thus leading to increased signaling and downstream cellular events (Atit et al., 1999; Ingram et al., 2000).

In summary, we demonstrate that epicardial loss of Nf1 results in early and increased EMT, which leads to the expansion of cardiac fibroblasts and cVSMCs. We were able to mitigate the increased EMT by altering PDGF signaling, which has recently been implicated in epicardial EMT (Smith et al., 2011). Our work indicates that EPDCs, along with endocardial-derived valve cells and cardiomyocytes are sensitive to perturbations in Nf1 activity. Further investigations will be required to determine what the long-term outcomes of this EPDC expansion are for the physiology of the heart under pathological and non-pathological conditions.

We thank Ray Runyan for essential comments and help with the collagen gel invasion assay; Moshe Shani for providing the X-LacZ4^Tg mouse line; Nancy Ratner for providing Nf1-GRD adenovirus; Christopher Smith and other M.D.T. laboratory members for scientific discussion and critical reading of the manuscript; and Greg Urquhart, Banu Eskiocak and Emily Webster for technical assistance.