|Home | About | Journals | Submit | Contact Us | Français|
Recent studies indicate that mammals, including humans, maintain some capacity to renew cardiomyocytes throughout postnatal life1,2. Yet, there is little or no significant cardiac muscle regeneration after an injury like acute myocardial infarction (MI)3. By contrast, zebrafish efficiently regenerate lost cardiac muscle, providing a model for understanding how natural heart regeneration may be blocked or enhanced4,5. In the absence of lineage-tracing technology applicable to adult zebrafish, the cellular origins of newly regenerated cardiac muscle have remained unclear. Here, we used new genetic fate-mapping approaches to identify a population of cardiomyocytes that become activated after resection of the ventricular apex and contribute prominently to cardiac muscle regeneration. Through use of a transgenic reporter strain, we found that cardiomyocytes throughout the subepicardial ventricular layer trigger expression of the embryonic cardiogenesis gene gata4 within a week of trauma, before expression localizes to proliferating cardiomyocytes surrounding and within the injury site. Cre recombinase-based lineage-tracing of cells expressing gata4 before evident regeneration, or of cells expressing the contractile gene cmlc2 before injury, each labeled a majority of cardiac muscle in the ensuing regenerate. By optical voltage mapping of surface myocardium in whole ventricles, we found that electrical conduction is re-established between existing and regenerated cardiomyocytes between 2 and 4 weeks post-injury. After injury and prolonged Fgf receptor inhibition to arrest cardiac regeneration and enable scar formation, experimental release of the signaling block led to gata4 expression and morphological improvement of the injured ventricular wall without loss of scar tissue. Our results indicate that electrically coupled cardiac muscle regenerates after resection injury primarily through activation and expansion of cardiomyocyte populations, findings with implications for promoting regeneration of the injured human heart.
After removal of the apex of their cardiac ventricle, zebrafish replace the resected myocardium4,5. While these events involve cardiomyocyte hyperplasia, it is uncertain whether proliferating cardiomyocytes derive from existing myocytes or from non-myocyte sources like stem cells. Identifying cellular contributions to embryonic or regenerative organogenesis has typically involved genetic methods to irreversibly label different cell types and track their progeny 6–11.
While searching for molecular markers informative for regeneration, we identified a unique expression pattern driven by upstream regulatory sequences of gata4, a transcription factor gene expressed in the developing embryonic heart and essential for normal cardiac patterning and vascularization12–17. Through use of the Tg(gata4:EGFP)ae1 reporter line18, we found that EGFP fluorescence was largely absent in the uninjured adult ventricle. However, upon resection of the ventricular apex, gata4-driven EGFP was induced in a high percentage of cells throughout the outer compact layer of ventricular myocardium during a period from 3 to 7 days post-amputation (dpa) (Fig. 1a, d, f, g). gata4:EGFP was expressed in many cells surrounding and within the injury site by 2 weeks after injury (Fig. 1h), and was limited at all timepoints to cells positive for myocyte markers (Fig. 1b, c) and negative for markers of the epicardium, a vasculogenic mesothelial layer overlying the compact muscle (Fig. 1e). BrdU labeling studies demonstrated that many gata4:EGFP+ cells at the lateral edges of the wound at 7 dpa, and within the wound at 14 dpa, had recently undergone DNA synthesis (Fig. 1j, k). By 30 dpa, a substantial area of the regenerated ventricular wall remained labeled by gata4:EGFP fluorescence (Fig. 1i). At these stages, gata4:EGFP expression predominantly labeled compact cardiomyocytes, which normally occupy a greater portion of the regenerate than inner trabecular cardiomyocytes4. Quantification of fluorescent ventricular muscle at different stages of regeneration by histology indicated progressive redistribution of gata4-driven EGFP expression from wound edges into the injury site (Supplementary Fig. 2a, b).
To clarify the dynamics of these events, we generated new transgenic strains to facilitate inducible, Cre recombinase-based lineage-tracing from gata4+ cells. We created a line with a tamoxifen-inducible Cre driven by gata4 regulatory sequences, Tg(gata4:ERCreER)pd39, as well as an indicator line that would permit visualization of cardiomyocyte EGFP fluorescence after excision of loxp-flanked STOP sequences, Tg(β-actin2:loxP-DsRed-STOP-loxP-EGFP)s928 (β-act2:RSG; Supplementary Fig. 3). We injected 4-hydroxytamoxifen (4-HT) or vehicle once daily in gata4:ERCreER; β-act2:RSG animals from 5–7 dpa, a timepoint preceding detectable gata4-driven EGFP fluorescence in the injury site. Injection of gata4:ERCreER; β-act2:RSG animals with 4-HT, but not vehicle, labeled what presumably represented a subset of myocytes that fluoresced in the gata4:EGFP line, revealing a small number of EGFP+ cardiomyocytes bordering the wound by 9 dpa. Moreover, contiguous regions of EGFP+ cardiomyocytes could be detected in the injury site by 14 dpa (Fig. 2a), representing a quantifiably significant expansion in labeled cardiac muscle at 20 dpa (Supplementary Fig. 2a, c). These findings indicated a mechanism in which subepicardial cells throughout the ventricle respond to injury by inducing gata4 expression, with cells near the injury site proliferating and contributing a high proportion of new cardiomyocytes.
Although confocal imaging colocalized gata4-driven EGFP and muscle markers, it remained possible that non-myocytes induced gata4 after injury and rapidly differentiated into proliferative cardiomyocytes. To test the extent to which existing cardiomyocytes contribute to regeneration, we created a strain in which tamoxifen-inducible Cre is driven by regulatory sequences of the contractile gene cardiac myosin light chain 2 (cmlc2), Tg(cmlc2:CreER)pd10. Measurements of labeling efficiency indicated that our 4-HT injection protocol tagged ~95% of uninjured cmlc2:CreER; β-act2:RSG ventricular cardiomyocytes with EGFP fluorescence (Supplementary Fig. 4). Based on analysis of several different indicator lines, labeling by cmlc2-driven CreER was specific to cardiomyocytes (Supplementary Fig. 5), and was not instigated by injury or vehicle injection (Fig. 2b). Five days after labeling cardiomyocytes, we resected ventricular apices and allowed 30 days of regeneration. We found no significant difference in the proportion of EGFP+ cardiomyocytes in regenerated cmlc2:CreER; β-act2:RSG tissue compared to uninjured ventricles collected 5 or 35 days after injection, a result indicating that the vast majority of new cardiomyocytes derives from cells expressing cmlc2 prior to injury (Fig. 2b; Supplementary Fig. 4).
A critical aspect of successful regeneration is the functional incorporation of newly created cells into existing tissue. We labeled whole explanted hearts with the transmembrane potential-sensitive dye di-4-ANEPPS, and performed optical voltage mapping of surface level cardiomyocytes that include the compact layer and regenerate. At 7 dpa, and less-so at 14 dpa, there was an increased density of isochrones near the apex of the ventricle as compared to uninjured controls, indicating a marked slowing of conduction (Fig. 3a). In addition, electrical activity consistently failed to propagate into the regenerating apex at 7 dpa, while impulses again conducted throughout the ventricle by 14 dpa when substantial numbers of new gata4+ cardiomyocytes appear (Fig. 3a). By 30 dpa, isochrone densities at the apex appeared normal (Fig. 3a). These observations were confirmed by direct estimation of conduction velocity, which revealed slowing of distal ventricular conduction at 7 and 14 dpa, and normal conduction at 30 dpa (Fig. 3b). We also found evidence of a significant reduction in the maximum depolarization rate at 7 dpa, which was fully restored by 14 dpa, while action potential duration was the same in all groups (Fig. 3c; Supplementary Fig. 6). Thus, our imaging data indicate that electrical coupling of new apical cardiomyocytes begins to occur by ~2 weeks post-injury, with full coupling in the restored wall by 30 dpa.
The normal regenerative response of the zebrafish heart to injury is thought to deter or outcompete a secondary scarring response4. To test whether regeneration can occur after a scar is established, we injured Tg(hsp70:dn-fgfr1)pd1 ventricles and induced expression of a dominant-negative Fibroblast growth factor (Fgf) receptor by heat-shock for 30 days, causing regenerative arrest and scarring19 (Fig. 4a, b). Then, we enabled Fgf signaling by removal from heat-shocks for 14 or 60 days. Interestingly, 14 days of restored Fgf signaling increased myocardium in 73% of wounds (Supplementary Fig. 7), with hsp70:dn-fgfr1; gata4:EGFP apices containing areas of gata4:EGFP+ cardiomyocytes (Fig. 4c, d). When we restored Fgf signaling to scarred hearts for an extended period of 60 days, fibrin was cleared from the wounds, although we did not observe loss of scar tissue. Notably, 90% of injury sites showed histological improvement after extended restoration of Fgf signaling, including 60% that had formed a contiguous wall of muscle enveloping the scar (Fig. 4e–h). These findings suggest that regenerative signals are maintained in zebrafish hearts with established injury scars, an environment that in mammals presents a hurdle for cell-based MI therapies20,21. Accordingly, mechanisms that underlie zebrafish heart regeneration might be pertinent to human MI survivors harboring mature scar tissue and compromised ventricular walls.
Earlier studies assessed fast- (EGFP) and slow-folding (nuclear DsRed2) reporters in double transgenic Tg(cmlc2:EGFP)f1 or twu26; Tg(cmlc2nucDsRed2)f2 zebrafish to document evidence for fresh maturation of cmlc2− progenitor cells into proliferative cardiomyocytes. Specifically, many EGFP+nucDsRed2− cardiomyocytes were observed in the developing embryonic heart and the regenerating adult ventricle19,22. We re-explored this developmental timing assay by substituting a cytosolic DsRed2 reporter of cmlc2 which removes the element of nuclear localization that might reduce reporter sensitivity during dynamic developmental events (Tg(cmlc2:DsRed2)pd15). We observed EGFP+DsRed2− myocytes in embryos but not in 7 dpa adult regenerates, which instead contained EGFP+DsRed2+ myocytes with each cytosolic reporter fluorescing at lower intensities than in non-regenerating muscle (Supplementary Fig. 8). Together with cmlc2:CreER lineage-tracing data, these new results argue for modifying the previous interpretation, and indicate that cardiomyocytes participating in regeneration possess or acquire an immature phenotype with reduced cmlc2 expression. Such a phenotype is possibly reflected by ultrastructural features of cardiomyocytes in 14 dpa regenerates that we did not observe in subepicardial cardiomyocytes of uninjured ventricles, including reduced sarcomeric structure, dysmorphic mitochondria, and low mitochondrial density (Supplementary Fig. 9).
In conclusion, we have identified new mechanistic aspects of zebrafish heart regeneration germane to the origin, function, and capabilities of regenerated cardiomyocytes. Foremost, we found that a subpopulation of cardiomyocytes within the ventricular wall activates gata4 regulatory sequences, proliferates, and contributes substantially to local muscle regeneration. The extent to which other cell populations possibly supply the regenerate awaits further direct lineage-tracing experiments. Interestingly, the activation of gata4:EGFP expression in subepicardial cardiomyocytes parallels the rapid, chamber-wide injury response of the overlying epicardial cells before they also incorporate into the regenerating area19. The similar spatiotemporal dynamics of these muscularizing and vascularizing tissues suggests important interactions as they each become activated, proliferate, and integrate into the injury site.
Our findings are intriguing in light of recent reports describing factors that when introduced can increase proliferation of differentiated cardiomyocytes and improve function in the injured adult mammalian heart23–26. It is likely that the zebrafish heart provides an optimized injury environment that encourages activation and/or proliferation of cardiomyocyte subpopulations. Cre-based tools in zebrafish including those we have described here will enable precise experimental manipulation of gene expression or function in attempts to modify the injury environment or regenerative response. With knowledge of a key origin of new cardiomyocytes, the cellular and molecular regulatory interactions that mediate heart regeneration in zebrafish can now be more informatively pursued.
Outbred Ekkwill strain (EK) or EK/AB mixed background zebrafish 6–12 months of age were used for ventricular resection surgeries as described previously4. All transgenic strains were analyzed as hemizygotes; details of their construction are described in the separate Methods section. Animal density was maintained at ~4 fish/liter in all experiments. 4-hydroxytamoxifen (4-HT) (Sigma) dissolved with ethanol (5 mg/ml) was diluted in water to 0.5 mg/ml for intraperitoneal injections. 10% ethanol was used as a vehicle control. EGFP labeling quantification is described in the separate Methods section. Heat-shock experiments were performed as described previously27, using double transgenic hsp70:dnfgfr1; cmlc2:nucDsRed2 or hsp70:dnfgfr1; gata4:EGFP animals. For BrdU incorporation experiments, 2.5 mg/ml BrdU (Sigma) was injected intraperitoneally once daily for 3 days prior to collection. Immunofluorescence, in situ hybridization, and Acid Fuchsin Orange G stains (detecting fibrin and collagen) were performed as described previously4. Primary antibodies used in this study were: anti-Mef2 (rabbit; Santa Cruz Biotechnology), anti-Myosin heavy chain (F59, mouse; Developmental Studies Hybridoma Bank), anti-β-catenin (rabbit; Sigma), anti-zf Raldh2 (rabbit: Abmart), anti-BrdU (rat; Accurate), and anti-GFP (rabbit, used only for co-detection with BrdU; Invitrogen). Secondary antibodies (Invitrogen) used in this study were Alexa Fluor 594 goat anti-rabbit IgG (H+L) for anti-Mef2, Alexa Fluor 594 goat anti-mouse IgG (H+L) for F59, Alexa Fluor 594 goat anti-rat IgG (H+L) for anti-BrdU, and Alexa Fluor 488 goat anti-rabbit IgG (H+L) for anti-GFP. In situ hybridization and immunofluorescence images were taken using a Leica DM6000 microscope with a Retiga-EXi camera (Q-IMAGING), and confocal images were taken using a Leica SP2 or SP5 confocal microscope. Physiology methods are described in the separate Methods section.
ERT2-Cre-ERT2 cDNA from pCAG-ERT2CreERT228 was cloned downstream of the 14.8 kb gata4 promoter18. The entire construct was flanked with I-SceI sites for transgenesis using the meganuclease method29. The full name of this transgenic line is Tg(gata4:ERCreER)pd39.
Cre-ERT2 cDNA from pCre-ERT230 was cloned downstream of the 5.1 kb cmlc2 promoter31. A DsRed-Ex cassette controlled by the lens specific α-crystalline promoter was also included (α-cry:DsRed), which enables visual identification of transgenic animals by lens fluorescence32, and was subcloned upstream of the cmlc2:CreER sequences in the opposite orientation. The entire construct was flanked with two copies of the core element of the chicken β-globin insulator (2× core insulator elements) (provided by G. Felsenfeld), and then flanked with I-SceI sites. The full name of this transgenic line is Tg(cmlc2:CreER)pd10.
9.8 kb of genomic DNA immediately upstream of the β-actin2 trancriptional start site was subcloned into a modified pBSK vector with a multiple cloning site flanked by I-SceI restriction sites. A loxP-DsRed-STOP-loxP-EGFP cassette was then subcloned downstream of the β-actin2 promoter. The DsRed-STOP cassette serves as a marker for the transgene, and also prevents read-through translation of EGFP protein. This construct was injected into one-cell stage wild-type embryos using standard meganuclease transgenesis techniques29. One founder was isolated and propagated from the injected embryos. The full name of this transgenic line is Tg(β-actin2:loxP-DsRed-STOP-loxP-EGFP)s928.
The pBigT vector was used to make a cassette containing a floxed mCherry cassette with a transcriptional stop, upstream of an EGFP cassette tagged with three copies of nuclear localization signals (3 X NLS-EGFP) downstream of the second loxP site. The resulting cassette (loxP-mCherry-STOP-loxP-nucEGFP) was then subcloned downstream of a ~5.3 kb fragment of the ~10 kb zebrafish β-actin2 promoter, or a ~1.5 kb zebrafish hsp70 promoter. The constructs were flanked with two copies of the core element of the chicken globin insulator (2× core insulator elements) (provided by G. Felsenfeld), as well as two I-SceI sites. The full names of the transgenic lines are Tg(β-actin2:loxP-mCherry-STOP-loxP-nucEGFP)pd31 and Tg(hsp70l:loxP-mCherry-STOP-loxP-nucEGFP)pd30.
The translational start codon of gata5 in the BAC clone DKEYP-73A2 was replaced with the loxP-mCherry-STOP-loxP-nucEGFP (RnG) cassette by Red/ET recombineering technology (GeneBridges). The 5’ and 3’ homologous arms for recombination were a 716 bp fragment upstream of the start codon and a 517 bp fragment downstream, respectively, and were subcloned to flank the RnG cassette. To avoid potential mis-recombination between the RnG cassette and an endogenous loxP site in the BAC vector, we replaced the vector-derived loxP site with an I-SceI site using the same technology. The final BAC was purified with nucleobond BAC 100 kit (Clontech), and co-injected with meganuclease into one-cell stage zebrafish embryos. The full name of this transgenic line is Tg(gata5:loxP-mCherry-STOP-loxP-nucEGFP)pd40.
DsRed2 cDNA was cloned behind the 5.1 kb cmlc2 promoter, and the entire cassette was flanked with I-SceI sites. The full name of this transgenic line is Tg(cmlc2:DsRed2)pd15.
To determine the kinetics of labeling by 4-HT, we gave single 20 µl intraperitoneal injections of a low dose (2.5 µg/ml) of 4-HT to uninjured cmlc2:CreER; β-act2:RSG adults. We detected recombination by myocyte EGFP fluorescence in ventricular sections at 1 dpi, and at increased frequency (~20% of ventricular myocytes) and intensity by 2 dpi. Examination of ventricular sections at 3, 4, and 5 days after injection indicated EGFP labeling frequencies similar to 2 dpi. These findings provided evidence that intraperitoneally injected 4-HT is no longer able to stimulate new recombination events in ventricular cardiomyocytes of adult zebrafish after 2 days post-injection.
To quantify Cre-mediated release of EGFP expression in lineage tracing experiments with cmlc2:CreER; β-act2:RSG animals, 3 sections including the regenerate were selected from each heart. For uninjured control hearts, the 3 largest sections were selected from each heart as described previously27. We used the DsRed+ area to reflect total myocardial section surface area, as the background expression of DsRed persists in cardiac muscle after 4-HT injections in β-act2:RSG trangenic fish. This is likely due to multiple transgene copies, allowing escape of some floxed DsRed cassettes from Cre-mediated recombination (Supplementary Fig. 2). Images of single optical slices of the ventricular apex were taken at 40X (1024 X 1024 pixels) by adjusting gain to detect EGFP or DsRed signals above the background level. An area (504 X 504 pixels) at the injured apex was chosen to include the majority of the regenerate in the image, and cropped using Photoshop software. An identical area was cropped at the ventricular apex for uninjured animals. EGFP+ and DsRed+ areas were quantified in pixels by ImageJ software, and the percentage of EGFP+ versus DsRed+ areas was calculated.
To quantify EGFP+ areas, 3 sections including the wound/regenerate were selected from each heart. Images of the ventricular apex were taken at 20X using a Leica DM6000 microscope with a Retiga-EXi camera (Q-IMAGING). An area (109 X 109 pixels for gata4:EGFP animals and 73 X 73 pixels for gata4:ERCreER; β-act2:RSG animals) was designated at each of the lateral wound edges and cropped using Photoshop software (see Supplementary Fig. 2 for cartoon). An area (109 X 218 pixels for gata4:EGFP animals and 73 X 146 pixels for gata4:ERCreER; β-act2:RSG animals) non-overlapping with the lateral areas was designated and cropped at the injury site. EGFP+ areas were quantified in pixels by ImageJ software, and the number of pixels summed and averaged to yield an edge EGFP+ area and injury site EGFP+ area for each heart.
Injured hearts were fixed in 2.5% glutaraldehyde/2% formaldehyde in 0.1M sodium cacodylate buffer. Tissue was post-fixed in 1% OsO4, dehydrated, embedded in Spurr’s resin (EMS), and sectioned to 70–90 nm thickness using Ultracut or Ultracut E ultra-microtomes (Leica). The sections were counterstained with uranyl acetate and lead citrate. Microscopy was performed on a FEI Tecnai G2 Twin transmission electron microscope and images were obtained using a Hamamatsu ORCA-HR digital camera and AMT software.
Fish were anesthetized with 0.1% 3-aminobenzoic acid methyl ester, and hearts were isolated within 30 seconds and placed in Tyrode solution containing (in mM) Na+(136), K+(5.4), Mg2+(1.0), PO43−(0.3), Ca2+(1.8), GLUCOSE (5.0), HEPES (10.0) at pH 7.4. Hearts were loaded for 10 minutes with the transmembrane-potential sensitive dye di-4-ANEPPS (D-1199, Invitrogen), which was dissolved in Tyrode solution. After staining, the preparations were transferred into the recording chamber (Warner Instruments), which contained Tyrode solution supplemented with 30 µM of the excitation-contraction decoupler blebbistatin (EMD Chemicals) to inhibit contraction during optical measurements. The chamber was mounted onto the stage of an inverted microscope (TE-2000U, Nikon) equipped with a high-speed CCD camera (Cardio CCD-SMQ, Redshirt Imaging) with an 80 × 80 pixel frame. Hearts were point-stimulated at 60 beats-per-minute using a fine platinum electrode held in place near the base of the heart using a micro-manipulator. This electrode location eliminated artifactual increases in conduction velocity due to simultaneous local capture of multiple pixels proximate to the stimulus site (i.e. virtual electrode effects)33. Optical action potentials were recorded from the epicardial surface of the apex. Fluorescence was excited with a 120 W metal-halide arc lamp (X-Cite 122, Exfo) and filtered at 540 ± 25 nm. Fluorescence emission was passed through a long-pass emission filter (585 nm) before being focused onto the camera. Optical magnification was 2X resulting in 11 μm spatial resolution between recording pixels. The camera was operated at 2000 frames-per-second and signals were digitized with 14-bit precision. Signals were digitally filtered in the temporal (900 Hz cutoff) and spatial (4-pixel weighted average) domains to reduce noise. Action potentials recorded with this system depicted the time course of membrane potential change with fidelity comparable to action potentials recorded with patch-clamp technique34.
Acquired fluorescence data were analyzed using custom software written specifically for the analysis of action potentials recorded optically from the zebrafish heart (MatLab, Mathworks). Action potential duration (APD) was defined as the time interval between 20% depolarization and 80% return or repolarization to resting potential (APD80). Action potential upstroke velocity was derived from the maximum of the first time-derivative of the action potential. Activation time was defined as the time of 50% depolarization during the rising phase of the action potential. We have used this criterion of local activation because the time at half-maximum depolarization has been shown previously in computer simulations to correspond closely to the maximum cellular sodium influx35. Isochronal maps which display the position of the wavefront at constant time intervals (2 ms) were constructed from the activation times using the contour plotting functions provided by the Matlab software. Conduction velocity vector fields were estimated from the activation times using an established algorithm described by Bayly et al36. Briefly, local velocity vectors, which represent the magnitude and direction of the propagating depolarizing wave at each recording site, were calculated by fitting the depolarization time measured at each site to a parabolic 2-dimensional surface. The components of the velocity vector at each site were calculated from the components of the gradient on this surface. Conduction velocities were averaged across sites located within a 200 µm × 200 µm square near the apex of each heart (Supplementary Fig. 10).
We thank J. Burris and A. Eastes for zebrafish care, X. Meng (Abmart) and the Developmental Studies Hybridoma Bank for antibodies, M. Gignac for help with electron microscopy, lab members for comments on the manuscript, and G. Burns, P. Chambon, and G. Felsenfeld for plasmids. This work was supported by postdoctoral fellowships from AHA (K.K. and Y.F.), JDRF (R.M.A.), and JSPS (K.K.); NIH training grants HL007208 at Massachusetts General Hospital (A.A.W.) and HL007101 at Duke University Medical Center (G.F.E.); grants from NHLBI (HL064282 to T.E., HL054737 to D.Y.R.S., and HL081674 to K.D.P.), NIGMS (GM075846 to C.A.M.), and March of Dimes (C.A.M.); and grants from AHA, Pew Charitable Trusts, and Whitehead Foundation (K.D.P.).
Supplementary Information is linked to the online version of the paper at www.nature.com/nature. A figure summarizing the main result of this paper is also included as Supplementary Figure 1.
Author Contributions. K.K. and K.D.P. designed experimental strategy, analyzed data, and prepared manuscript. K.K., J.E.H., and Y.F. generated and characterized transgenic lines for lineage-tracing. R.M.A. and D.Y.R.S. provided unpublished reagents for lineage-tracing. K.K., J.E.H., and K.D.P. performed regeneration experiments. J.E.H. performed electron microscopy. A.A.W., G.F.E., and C.A.M. designed physiology experiments and interpreted data. A.A.W. performed optical mapping assays. T.E. helped design strategy and provided key reagents. All authors commented on the manuscript.
The authors declare no competing financial interests.