Regulated addition of new myocardial and epicardial cells fosters homeostatic cardiac growth and maintenance in adult zebrafish Free

Organ homeostasis mechanisms calibrate organ size and function in response to changing physiological conditions. Perhaps the clearest example of homeostasis occurs during animal growth, a process accompanied by organ growth to accommodate increasing demands. For full-sized adult mammals, homeostasis maintains the status quo, replacing damaged or senescent cells through direct structural cell proliferation, progenitor cell activity or hypertrophy of surrounding cells. Some organs are capable of rapid homeostatic adjustments to tissue loss or gain. For example, the liver will increase mass through compensatory cell proliferation after partial hepatectomy, or reduce mass through compensatory cell death after experimentally induced hyperplasia, to maintain the appropriate size (Conlon and Raff, 1999; Potter and Xu,2001; Taub, 2004). Understanding the cellular and molecular mechanisms of organ homeostasis is a major research goal, with potential to shed light on therapies for trauma,degenerative disease and aging.

Although cellular and molecular mechanisms of homeostatic regulation in most organs are incompletely understood, cardiac homeostasis is particularly mysterious. For many years, it was believed that the adult mammalian heart is a post-mitotic organ, and that all postnatal cardiac growth is achieved through hypertrophy of a fixed number of cardiomyocytes (CMs). Indeed,analyses of DNA synthesis in the adult murine heart estimated a very low percentage (a maximum of 0.0005%) of ventricular CMs entering the cell cycle each day (Soonpaa and Field,1997). Similar analyses during postnatal rodent growth indicated that CMs undergo a transition shortly after birth from hyperplastic to hypertrophic growth, associated with the production of large, multinucleated CMs through karyokinesis rather than with new, mononucleated CMs through cytokinesis (Li et al., 1996; Soonpaa et al., 1996). Furthermore, there is little evidence of consequential regeneration after cardiac injuries such as ischemic infarction, and the resulting hypertrophy of existing muscle that occurs concomitantly with scar formation is detrimental to cardiac function. The notion that the major structural cells within such a vigorous and vital organ survive 70 to 100 years in humans without support from new CMs is contested, especially given recent identification of cell populations that may possess progenitor activity in the postnatal mammalian heart (Beltrami et al., 2003; Laugwitz et al., 2005; Martin et al., 2004; Oh et al., 2003). However, a natural in vivo contribution of these progenitor cells to adult tissue homeostasis has not been experimentally demonstrated.

The mechanisms by which non-myocardial cardiac cell types are maintained and replaced are also poorly understood. Interestingly, recent studies have found that cells of the adult mammalian epicardium, a thin epithelial tissue surrounding the myocardium proper, can be experimentally stimulated to differentiate into smooth muscle and endothelial cells in vitro(Smart et al., 2007; van Tuyn et al., 2007). In this way, the in vitro activity of the adult epicardium appears to mimic the capacity of the embryonic epicardium, a structure that serves as the primary source of coronary vasculature during heart development(Reese et al., 2002). Thus, it is possible that adult epicardial cells also function as a progenitor tissue to maintain the vasculature or other cardiac cell populations.

Certain non-mammalian vertebrates such as amphibians and fish may have greater potential for hyperplastic cardiac homeostasis. By contrast with mammals, many of these species display indeterminate growth, and can rapidly increase adult mass in response to changes in population density and nutrition(Jordan, 1905), an increase that is typically accompanied by organ augmentation. Moreover, some of these species have demonstrated various degrees of regeneration after mechanical injury of the cardiac ventricle, a response that may be perceived as a homeostatic method to restore cardiac mass(Flink, 2002; Oberpriller and Oberpriller,1974; Poss et al.,2002). In particular, the zebrafish, which displays indeterminate growth (Tsai et al., 2007),has a strong cardiac regenerative response, and is amenable to molecular genetic approaches, represents a unique model system to visualize and dissect cardiac homeostasis.

Here, we show that adult zebrafish display dramatic, hyperplastic cardiac growth in response to aquarium conditions that stimulate rapid animal growth,whereas animals maintaining cardiac size show distinct but rare addition of new CMs. Additionally, we find that rapid growth conditions induce epicardial expression of embryonic markers such as raldh2 (also known as aldh1a2 - Zfin) and tbx18, and that the epicardium regularly contributes cells to the adult ventricular wall even in the absence of cardiac growth or myocardial injury. Inhibition of Fgf signaling, a pathway necessary for normal heart regeneration, disrupts epicardial cell supplementation and causes spontaneous ventricular scarring in uninjured adult fish. Our study exposes dynamic myocardial and epicardial mechanisms that mediate cardiac homeostasis.

Zebrafish of outbred Ekkwill or EK/^*AB mixed background strains were raised at a standard density of approx. five fish per liter to 8-10 weeks of age, at which point fish weighing 70-100 mg were selected for growth experiments. Fish were placed in conditions that stimulate rapid growth (RG; 3 fish/10 l tank) or returned to conditions that allow normal, slower growth(SG; 15 fish/3 l tank) for 10-14 days, at which point the animals were weighed and the hearts extracted. Adult fish, 6-months old, maintaining animal size(MS) were raised at 3-5 fish/l, the standard in our facility. Ventricular resection surgeries were performed as previously described(Poss et al., 2002). All experiments on animals were performed in accordance with institutional guidelines and regulations.

In situ hybridization on cryosections of paraformaldehyde (PFA)-fixed hearts was performed using digoxigenin-labeled cRNA probes as described previously (Poss et al.,2002). Acid fuchsin-Orange G staining was performed as described previously (Poss et al.,2002). Immunofluorescence was performed(Poss et al., 2002) using antibodies against BrdU (rat; Accurate), Mef2 (rabbit; Santa Cruz Biotechnology), PCNA (mouse; Sigma) and DsRed (rabbit; Clontech). Terminal deoxynucleotidyl transferase biotin-dUTP nick end-labeling (TUNEL) reactions were performed on 10 μm cryosections of PFA-fixed hearts with reagents from Invitrogen, and visualized with peroxidase substrate reagents from Vector Laboratories.

To calculate ventricular section surface area, hearts were sectioned longitudinally, stained with TRITC-phalloidin, and the three largest sections of the ventricle were imaged and measured for calculation of the area using Openlab software. The measurements were averaged to give one value for each heart.

For BrdU incorporation experiments, animals were injected once daily with 2.5 mg/ml BrdU for 3 days prior to collection. Hearts from RG (injected 7, 8 and 9 days after introducing density conditions), SG, and MS animals were cryosectioned and immunostained for Mef2 and BrdU. Images of the middle of the lateral ventricle wall (opposite the atrium) from the three largest sections of each ventricle were selected for analysis. Nuclei labeled with Mef2, a marker of CMs, BrdU, or both were counted by hand using Adobe Photoshop images. The region analyzed included the compact muscular wall and a region approximately six cells thick of trabecular muscle, so that both trabecular and compact muscle would be included. For these experiments, 150-300 Mef2^pos nuclei were counted per section, or 500-900 per animal.

To calculate CM density, Mef2-stained ventricular sections were used to produce similar images. Then, the area of the region was calculated using Openlab software, and the number of CM nuclei within the traced region counted by hand. About 100-300 nuclei were counted per section, or 400-900 per animal. To assess CM nucleation, ventricles were isolated from cmlc2:nRFP RG,SG and MS fish (see Results for description of transgenic zebrafish), and cells were dissociated as described previously(Warren et al., 2001). Isolated cells were collected in L-15 medium and live cells imaged to determine the number of nuclei.

For analysis of RFP^cyto cells (see Results for RFP^cyto description), cryosections from cmlc2:nRFPventricles were stained with anti-DsRed antibody and the three largest sections from each heart were used for imaging. For these experiments, images of the apex were selected for analysis. RFP^nuc and RFP^cyto cells were counted by hand from regions including the ventricular wall and approximately four cell layers of trabecular muscle. About 150-300 CMs per section were counted, or 600-900 per animal.

For analysis of epicardial-derived cells (EPDCs) within the ventricular wall, in situ hybridizations for tbx18 were performed on hearts from wild-type or hsp70:dn-fgfr1 animals under various growth conditions. Images from the middle of the lateral ventricle wall (opposite the atrium)were taken from the three largest sections of each heart. For each section,only tbx18-positive cells clearly separated from the epicardium and within the compact myocardial wall were counted (see Figs 5 and 6).

RNA was collected from 12-15 MS, RG, SG and regenerating [7 days post-amputation (dpa)] ventricles by extracting the heart in PBS on ice and mechanically removing the atrium and outflow tracts. Ventricles were transferred to TRI reagent (Sigma) and the RNA was isolated according to the manufacturer's instructions, before purification using Qiagen RNAeasy columns. RNA integrity was assessed by gel electrophoresis, and concentration determined by spectroscopy. cDNA was made from 200 ng RNA using oligo(dT)primer and SuperScript III reverse transcriptase (Invitrogen). Quantitative PCR (Q-PCR) was run using LightCycler FastStart DNA Master^PLUS SYBR Green I (Roche) on a LightCycler 2.0 machine (software version 4.05x). Dilutions of cDNA generated from pooled RNA of 20 hours and 56 hours post-fertilization embryos was used to determine the reaction efficiency (E)for each primer pair, and ΔC_T for each group was calculated using 7 dpa regenerating hearts as a standard. hand2 expression was normalized to gapdh for each sample using the following equation:X=((E)_gapdh)^-ΔCTgapdh)/((E)_hand2)^-ΔCThand2). All samples were run in triplicate, and the averages of two independent experiments are shown.

Adult zebrafish ∼6 months of age were placed into 1.5 l tanks (3 fish/l) to maintain a constant density during the experiment. Hearts from uninjured control animals, as well as manipulated animals, were fixed 3 days after manipulations. In one group of animals (Open), a single incision was made in the pericardial sac using iridectomy scissors. In saline-treated animals, ∼5 μl Hank's buffered saline was injected into the pericardial cavity from a 30 gauge needle until expansion of the sac was visible. For an additional group, a needle was similarly inserted into the pericardial sac,but no fluid was injected (Pierce). Only animals that showed no bleeding after any of these treatments were used for this study. As a control, iridectomy scissors were used to make a small incision on the dorsal side of the animal(Dorsal). For each group 9-11 animals were used in two experimental trials.

DiI labeling was performed by pipetting ∼0.5 μl of 1 mg/ml Cell Tracker CM-DiI (Molecular Probes) into the pericardial cavity of∼6-month-old zebrafish through a small incision in the pericardial sac. Smaller animals used for RG and SG conditions did not survive this procedure. Hearts were extracted and fixed at 1 hour, 3 days, 7 days and 14 days post-labeling. Labeled hearts were sectioned and imaged without coverslips, as mounting with a coverslip commonly caused diffusion of the dye. At least 15 animals were used for 1 hour, 3 day, and 7 day timepoints in three experimental trials.

For RG experiments, heterozygous hsp70:dn-fgfr1 transgenic and wild-type clutchmates (weighing 70-100 mg each) were selected at 8-10 weeks of age. Fish received a daily, transient increase in temperature from 26°C to 38°C as described previously (Lee et al., 2005). Animals remained at 38°C for ∼30 minutes each day. We found that this temperature increase could not be achieved in 10 l tanks using our system, so we instead placed one fish each in a 3 l tank, and performed analyses at 14 days. For adult maintenance experiments, uninjured 4-to 6-month-old hsp70:dn-fgfr1 and wild-type adult clutchmates received a daily temperature increase for 60-70 days.

We predicted that the vigor of cardiac homeostasis would be proportional to the rate of adult animal growth. Therefore, we developed three conditions based on aquarium density that we expected to stimulate rapid growth (RG),slow growth (SG) or size maintenance (MS). We found that when small adult fish(8- to 10-weeks old, 70-100 mg) were removed from standard aquarium density conditions (SG; 5 fish/l) and placed in low-density conditions (RG; 0.3 fish/l), body mass tripled after only 14 days (a ∼208% increase from day 0). Clutchmates maintained at SG displayed only a ∼13% increase in mass over this period. Fully mature, 6-month-old fish, considerably larger at the onset of experiments than RG and SG fish, maintained mass for at least 28 days when kept at standard aquarium density conditions(Fig. 1B). In general we found that starting size was a greater determinant of animal growth rate than age,consistent with other published studies(Tsai et al., 2007).

In these experiments, the large increases in body mass seen in RG fish were accompanied by dramatic increases in the size of both heart chambers(Fig. 1A). To quantify changes in ventricular size, we measured the surface area of longitudinal ventricular sections using digital imaging software. These analyses revealed that ventricular size more than tripled after only 14 days in RG conditions (a∼221% increase from day 0), a rate of cardiac growth approximately equivalent to that of animal growth under these conditions. Clutchmates in SG conditions exhibited a more modest increase in ventricular size over this period (∼48%). Because this increase was greater than the increase in animal size during this period, it is possible that the rate of cardiac growth may slightly diverge from the rate of animal growth in SG conditions. MS animals experienced no significant change in ventricular size over 14 days(Fig. 1A,C). Thus, by varying aquarium density and starting size of the animal, we closely controlled the rate of cardiac homeostasis. In particular, low aquarium density conditions induced remarkably rapid cardiogenesis in adult zebrafish.

To determine whether density-dependent control of ventricular size resulted from hyperplasia or hypertrophy, we administered three daily injections of BrdU prior to harvesting tissue after 9 days in RG, SG or MS conditions. CMs were specifically assayed for BrdU incorporation by costaining with an antibody against Mef2, a transcription factor that regulates myocardial differentiation (Molkentin and Markham,1993). We observed a high rate of BrdU incorporation in CMs within both trabecular and compact muscle compartments of RG fish, with 10.1±1.8% of CMs labeled by BrdU (mean ± s.e.m.; Fig. 2A,B). This labeling index was almost eight times that of SG animals (1.3±0.3%), and 25 times that of MS animals sustaining constant body mass and ventricular size(0.4±0.1%; Fig. 2A,B). We obtained similar results when identifying proliferating cells with an antibody against PCNA (see Fig. S1 in the supplementary material). We suspect that rare CM proliferation in MS animals serves to replace dying cells, as we observed occasional CM apoptosis events by TUNEL staining of ventricular sections (see Fig. S2 in the supplementary material).

To assess possible contributions of hypertrophy to growth, we measured the density of cells positive for the nuclear marker Mef2 in ventricular sections. There were no differences in CM nuclear density per myocardial area in RG, SG and MS fish, indicating a lack of CM hypertrophy(Fig. 2C). In addition, we dissociated ventricles and assessed the extent of binucleation, a consequence of karyokinesis without cytokinesis that is common in adult mammalian CMs. We found that the vast majority of ventricular CMs in rapidly growing fish were mononucleate (95.6%), as were CMs from fish kept at SG (97.9%) and MS conditions (95.1%), indicating derivation through cytokinesis(Fig. 2D,E). Thus, homeostatic cardiac growth and maintenance are primarily the result of bona fide CM hyperplasia, with the vigor of this hyperplasia dependent on the rate of animal growth.

Homeostatic CM generation must occur through the division of mature CMs, or the maturation of myocardial progenitor cells, or both. During regeneration,cells expressing markers of embryonic cardiac progenitor cells such as hand2 (Yelon et al.,2000) are established at the apical edge of the wounded muscle within 3-4 days post injury, and are maintained at that edge as regeneration progresses (Lepilina et al.,2006). The origin of these cells is unknown; they might exist as resident non-myocardial cells activated upon injury, or be created through reduction in contractile function of existing CMs, known as de-differentiation. Strong, discrete hand2 in situ hybridization signals analogous to that seen during myocardial regeneration were common in RG animals particularly within the compact myocardium, sparingly present in SG animals, and largely absent in MS adults(Fig. 3A). Real-time quantitative PCR experiments indicated that hand2 expression in RG ventricles was 1.4 fold that of SG clutchmates, and 2.5 fold that of MS ventricles (Fig. 3C).

hand2 is expressed in embryonic progenitors but is not a definitive marker of undifferentiated cells. To test the idea that progenitor cell populations give rise to new CMs during homeostasis, we assessed CM maturation in the cmlc2:nRFP transgenic line, a strain that reports a slow-folding, nuclear-localized DsRed2 protein under control of the cardiac myosin light chain 2 (myl7 - Zebrafish Information Network) promoter (Mably et al.,2003). Mature CMs in these transgenic fish show nuclear-localized DsRed fluorescence. However, an anti-DsRed antibody detects unfolded,cytosolic DsRed fluorescence in the new CMs of the 24-hour post-fertilization embryo and the regenerating edge of the injured adult ventricle, apparently before nuclear localization occurs and natural fluorescence is manifest. Thus,this marker (RFP^cyto) appears to serve as an indicator of developmental transition from undifferentiated progenitor cells to differentiated CMs (Lepilina et al.,2006). We found that 10.7±2.1% of CMs throughout the ventricle of RG animals expressed the RFP^cyto marker, a percentage over three times that of SG clutchmates (3.1±0.3%) and 36 times that of MS fish (0.3±0.1%; Fig. 3B,D). In ventricles of cmlc2:nRFP RG animals that had been injected with BrdU, both RFP^cyto and RFP^nuc CMs showed BrdU-labeled nuclei, suggesting that both the most recently maturing CMs and earlier-maturing CMs undergo proliferation(data not shown). In total, these results suggest that myocardial progenitor cell populations contribute at least in part to the vigorous generation of new CMs stimulated by rapid animal growth, and in rare events during homeostatic maintenance of heart size.

During zebrafish heart regeneration, local injury strongly activates expression of the embryonic epicardial markers raldh2 and tbx18 throughout the entire ventricular and atrial epicardium. This activation is accompanied by epicardial cell proliferation, expanding the epithelial cover of the ventricle to eventually cover the injury(Lepilina et al., 2006). raldh2 and tbx18 were expressed at low levels in a small number of epicardial cells in both SG and MS ventricles, indicating moderate activation. By contrast, these markers were robustly induced in contiguous stretches of RG atrial and ventricular epicardium(Fig. 4A, Fig. 5B). In addition,endocardial cells surrounding cardiac myofibers near the injury site have been shown to induce raldh2 during regeneration (R.J.M. and K.D.P.,unpublished) (Lepilina et al.,2006). We found that RG animals, but not SG or MS animals,displayed strong raldh2 expression in endocardial cells throughout the ventricle (arrows in Fig. 4A; A.A.W., J.E.H. and K.D.P., unpublished). Thus, embryonic gene expression within both the epicardium and endocardium is activated by rapid animal growth.

We suspected that, because of its location and exquisite sensitivity to growth and injury, the epicardium may act as a sensor to identify changes in animal growth through sensation of pressure changes in the extra-cardiac space. The pericardial sac is filled with a plasma ultrafiltrate that is believed to have a restrictive function during diastole. Therefore, opening the cavity might relieve pressure, increasing cardiac stretch. Conversely,saline injection is expected to increase pericardial pressure and limit the ability of the heart to expand (Farrell et al., 1988; Shabetai et al.,1985; Spodick,1997). To test whether these procedures affect epicardial gene expression, we examined raldh2 expression in several groups of experimentally manipulated 6-month-old MS animals. We found that ventricular raldh2 expression was induced to roughly the same level as in RG animals (but at levels lower than after partial ventricular resection) in experiments in which (1) the pericardial sac was surgically opened, or (2)saline was carefully injected into the pericardial cavity(Fig. 4B). Different manipulations that resulted in no effect on epicardial raldh2expression included: (1) piercing the pericardial sac without saline injection, (2) lesions to the dorsal region of the animal and (3) fin amputation. These results show that the epicardium is a dynamic tissue responsive to changes in the milieu surrounding the heart, and suggest that the effects of rapid animal growth on epicardial gene expression are due in part to an ability of the epicardial tissue to sense changes in resistance within the extra-cardiac space.

During embryonic heart development, a subset of epicardial cells undergoes epithelial-mesenchymal transition (EMT) and migrates into the underlying subepicardial space and myocardial wall as epicardial-derived cells (EPDCs). There, EPDCs contribute fibroblasts as well as smooth muscle and/or endothelial cells for building coronary vasculature(Olivey et al., 2004; Reese et al., 2002). To determine whether the adult zebrafish epicardium actively transfers cells subepicardially to the ventricular wall, we performed a pulse-chase experiment. We labeled the epicardium by filling the pericardial sac of MS zebrafish with the fluorescent lipophilic dye DiI, and then harvested hearts at different timepoints post-injection. Dye was limited to the epicardium at 1 hour post-injection; by contrast, we noticed a small number of labeled cells separated from the epicardium by 3 days post-injection. Occasionally, these labeled cells had a tubular morphology reminiscent of vascular tissue (inset, Fig. 5A). At 7 and 14 days post-injection, DiI was concentrated in cells that were embedded within the otherwise unlabeled ventricular wall inward from the now faintly labeled epicardium (Fig. 5A). Interestingly, labeled cells occasionally appeared to constitute a distinct layer of 15-20 μm at the junction of compact and trabecular myocardium. These surprising data indicate that the adult epicardium is a dynamic tissue that regularly contributes EPDCs to the ventricular wall, presumably through EMT.

Applying labeled dye to the pericardium elevated epicardial expression of raldh2 similar to saline injection (data not shown), implying that this treatment induces strong developmental activation of the epicardium and likely affected its contributions to the ventricular wall. To evaluate EPDC activity in unmanipulated animals, we utilized the embryonic epicardial marker tbx18. During regeneration, cells positive for the embryonic epicardial marker tbx18 emerge not only in the epicardium, but in subepicardial tissue within the wound and regenerating muscle(Lepilina et al., 2006). Interestingly, in uninjured ventricles, tbx18-expressing cells were also present both in the epicardium and several cell layers deep into the compact myocardium. Many tbx18-positive cells were located in what appeared as a distinct layer 15-20 μm subepicardially, a location similar to the EPDCs revealed by our pulse-chase experiments(Fig. 5B). For these reasons,we inferred that tbx18 is a robust marker for emergent EPDCs in the uninjured ventricular wall. Quantification of these internalized tbx18-positive EPDCs revealed that the density of EPDCs in RG animals was almost six times as high as in SG animals (27.4±1.6 vs 4.9±1.0 tbx18-positive cells/mm ventricular wall). Somewhat unexpectedly, MS animals had more than twice as many tbx18-positive EPDCs as SG animals (11.9±1.8; Fig. 5C). This may be a consequence of age differences between these groups, and/or the possibility that tbx18 is a cumulative marker of EPDCs that have emerged in the ventricular wall over weeks to months. The latter idea is supported by the observation that regenerated muscle possesses cells robustly expressing tbx18 for 30 days or more after injury(Lepilina et al., 2006). Given the role of epicardial tissue during embryonic heart growth and the appearance of tbx18-positive cells coincident with new vascular tissue during regeneration, we postulate that tbx18-positive EPDCs contribute to neovascularization of emergent myocardial tissue and/or vascular maintenance in uninjured animals. Indeed, we could observe in some cases cells positive for tbx18 that had the appearance of vascular tissue(Fig. 5B, red arrowhead and inset), similar to DiI-labeled cells in pulse-chase experiments. Thus, the adult epicardium supports cardiac homeostasis by regular contribution of new EPDCs to the ventricular wall.

Recent studies have revealed roles for Fgfs in epicardial EMT during both embryonic heart development and adult heart regeneration. EMT of cultured embryonic epicardial cells can be stimulated by Fgfs in vitro, and this pathway is necessary for normal coronary vascular development in mice(Lavine et al., 2006; Morabito et al., 2001). During zebrafish heart regeneration, the injured zebrafish myocardium increases expression of fgf17b, while the receptors fgfr2 and fgfr4 are expressed in EPDCs. Furthermore, abrogation of Fgf signaling disrupts EPDC supplementation of the wound and new muscle during regeneration, inhibiting neovascularization and myocardial renewal(Lepilina et al., 2006). We found that fgf17b, fgfr2 and fgfr4 were all present in SG and RG ventricles, although no differences in expression between the two groups were detectable by in situ hybridization (data not shown). Based on these findings, we suspected that Fgf signaling may also be important for EPDC supplementation of the ventricle during cardiac homeostasis.

To test this idea, we placed zebrafish transgenic for a heat-inducible dominant-negative Fgf receptor (hsp70:dn-fgfr1) and their wild-type clutchmates in RG conditions for 14 days, and applied a daily heat-shock(Lee et al., 2005). Transgenic animals appeared healthy during the experiment, although they grew less than wild-type animals under RG conditions (see Fig. S3 in the supplementary material). Although tbx18 expression was comparable between transgenics and wild types in the epicardium itself, the density of tbx18-positive EPDCs within the ventricular walls of hsp70:dn-fgfr1 animals was less than half that in wild-type animals(Fig. 6A,B; 24.6±2.1 EPDCs/mm in wild type, versus 11.5±2.1, in hsp70:dn-fgfr1). Fgfr inhibition also resulted in a slight reduction in ventricular growth under RG conditions, with transgenic animals exhibiting ∼31% less of a growth increase than wild-type clutchmates(Fig. 6C). Thus, normal homeostatic supplementation of the ventricular wall by EPDCs requires Fgf signaling. We suspect that the partial phenotype is due to the continued presence of other factors that act on EPDCs, incomplete inhibition of Fgf signaling, and/or the possibility mentioned above that tbx18expression reflects cumulative events of EPDC contribution.

As pulse-chase and tbx18 expression experiments indicated ongoing EPDC contribution even in the absence of significant cardiac growth, we tested the requirements for Fgf signaling in mature adult zebrafish. We delivered a long-term block of Fgf signaling by daily heat-shocks to hsp70:dn-fgfr1 transgenics for 60-70 days, with identical heat-shocks given to wild-type clutchmates. We found that ∼15% of hsp70:dn-fgfr1 fish developed extensive collagen deposition within the ventricular wall after this period, indicating the presence of scar tissue(Fig. 6D; n=39 fish). No wild-type animals displayed scarring (n=44 fish; Fisher-Irwin test, P<0.01). We suspect that these fibrotic events resulted from reduced efficiency of incremental EPDC supplementation over the 2-month period, although marker examination revealed no obvious defects (data not shown). These results demonstrate that Fgf signaling is essential for normal homeostatic growth and maintenance of cardiac tissue in the adult zebrafish.

We have discovered that new myocardial and epicardial-derived cells are created during homeostatic responses of the adult zebrafish heart, helping to couple cardiac and animal growth rate and maintain cardiac integrity. One molecular signaling pathway essential for these responses is that mediated by Fgfs, which ensures normal supplementation of the ventricular wall with EPDCs and helps to prevent scarring.

In contrast to what is known about postnatal ventricular growth in mammals,we have found that cardiac growth in zebrafish is a result of CM proliferation, not the hypertrophy of existing cells. Ventricular expression of the embryonic progenitor marker hand2, as well as a transitional marker for newly generated cardiomyocytes in cmlc2:nRFP transgenic fish, showed positive correlations with cardiac growth and CM proliferation. Thus, these results suggest that undifferentiated progenitor cells contribute to CM hyperplasia. Future experiments to address the extent to which this occurs must involve genetic fate-mapping strategies(Cai et al., 2003; Dor et al., 2004; Hsieh et al., 2007; Meilhac et al., 2003). Moreover, it will be critical to define homeostatic signals that trigger cardiomyogenesis and how these signals are distributed and calibrated in the adult animal.

As a thin epithelium covering the entire surface of the heart, the epicardium has an ideal architecture for relaying such homeostatic signals. Numerous studies support this idea, because: (1) the epicardium serves as a source of mitogens in the embryonic heart(Chen et al., 2002; Merki et al., 2005; Reese et al., 2002), (2)organ-wide developmental gene expression is activated within the adult zebrafish epicardium within hours of cardiac injury(Lepilina et al., 2006), and(3) during regenerative cardiogenesis as well as homeostatic cardiac growth,developmentally active epicardium is present at sites of cardiogenesis(Lepilina et al., 2006). Indeed, consistent with this idea are the findings here that manipulations of the pericardial environment that mimic growth-induced changes in pericardial space cause epicardial responses similar to those elicited by homeostatic growth. Because epicardial retinoic acid (RA) synthesis is regulated in each of these scenarios and is known to influence CM proliferation in embryos(Chen et al., 2002; Stuckmann et al., 2003), this molecule is a strong candidate for homeostatic regulation of cardiogenesis. We have found that the endocardium also increases RA synthesis during adult growth and regeneration. It is possible that endocardial cell activity regulates growth and/or remodeling of the inner trabecular muscle, which would appear to have reduced access to epicardial signaling molecules.

In addition to serving as a source of RA, we found that the epicardium also regularly contributes cells to the compact myocardial wall of the ventricle. Thus, the epicardium is by no means a static tissue in adult animals, but rather one that actively sustains the ventricle. This conclusion is bolstered by our finding that sustained inhibition of Fgf signaling, which disrupts EPDC supplementation in rapidly growing fish, led to spontaneous scar formation in animals maintaining their animal and organ size. Further exploration of the fate and function of these adult EPDCs, using tissue-specific fate-mapping and ectopic expression tools, will illuminate the roles of the epicardium in cardiac homeostasis. From what is known of embryonic heart development, it is likely that the adult zebrafish myocardial wall requires new EPDCs to replenish vasculature, or to neovascularize newly created muscle, or both. It will also be fascinating to learn to what extent the epicardium participates in homeostasis of the adult mammalian ventricular wall.

Our study exposes important similarities between cardiac homeostasis and cardiac regeneration. In each case, new myocardial and epicardial-derived tissues are generated in amounts appropriate to animal size, involving expression of a suite of cardiogenic developmental markers. Whereas regeneration is characterized by an intense focus of cardiogenesis at the injury site, homeostasis involves organ-wide cell addition in doses dependent on the vigor of animal growth. Thus, the adult zebrafish not only has the ability to generate new cardiac tissues, but can also control the intensity and the spatial distribution of these cardiogenic events to select replacement of a portion of tissue, rapid chamber-wide growth, or occasional exchange of a few cells (Fig. 7).

This relationship between cardiac regeneration, homeostatic growth and homeostatic tissue maintenance is likely to be important for explaining why zebrafish regenerate injured cardiac muscle. Both regeneration and homeostasis stand to benefit from sophisticated mechanisms to activate and deploy epicardial and myocardial cells for cardiogenesis. In some teleosts, adult growth rate is controlled by environmental and/or social factors and can influence mating preference (Borowsky,1973; Hofmann et al.,1999). Thus, one reason why cardiac regeneration persists in zebrafish, and probably other teleosts, may be the accessibility of injured tissue to cellular and molecular machinery that is normally employed for rapid indeterminate growth. This notion makes sense in light of extremely weak hyperplasia in the injured or uninjured hearts of adult mammals, species with determinate growth. Continued study of cardiac regeneration and homeostasis in zebrafish, including comparisons with mammalian cardiac dynamics, will expose critical mechanisms by which cardiac tissues are rejuvenated.

We thank A. Lepilina for TUNEL assays, C. Wheeler, O. Ighile, C. Parobek,J. Bumgardner and T. Yu for excellent fish care, A. Nechiporuk, S. Odelberg,and lab members for helpful comments on the manuscript. R.J.M. was supported by NIH training grant 2 T-32 HL007101-31. This work was supported by grants to K.D.P. from NHLBI, American Heart Association, Whitehead Foundation, and Pew Charitable Trusts.