To examine the contribution made by differentiated cardiomyoctes to regenerated cardiac tissue, we performed a series of amputation experiments on adult zebrafish in which cardiomyocytes had been genetically labelled at 48 h post fertilization (hpf). Approximately 20% of the ventricle was removed and cardiac regeneration was subsequently assessed at 7, 14 and 30 days post amputation (dpa). At 7 dpa, the remaining cardiac tissue was uniformly GFP^pos (Fig. 1a, b) with much of the missing tissue now replaced by a fibrin/collagen clot (n=5 hearts) (Fig. 1c). Strikingly, at 14 dpa, when a significant amount of regeneration had occurred, cardiomyocytes within the regenerating tissue were uniformly GFP^pos (Fig. 1d, e) with only a small fibrin clot remaining (n=7 hearts) (Fig. 1f). These results suggest that the regenerated cardiomyocytes arise from differentiated GFP^pos cardiomyocytes. These findings were further substantiated at 30 dpa, when regeneration is nearly complete; all of the cardiomyocytes within the regenerated cardiac tissue were clearly GFP^pos (n=9 hearts) (Fig. 1g, h).

To determine the overall contribution made by GFP^pos cardiomyocytes to the regenerated heart, we immunostained 7, 14 and 30 dpa regenerating heart sections with anti-GFP and the anti-sarcomeric myosin antibody MF20 (Supplementary Fig. 9a–d). We were unable to detect any cardiomyocytes labelled with MF20 or GFP alone (n=4836). Together, our results demonstrate that after amputation, the vast majority, if not all, of newly formed cardiomyocytes arise from pre-existing cardiomyocytes.

We next sought to determine whether GFP^pos cardiomyocytes had re-entered the cell cycle. Adult GFP^pos transgenic zebrafish were treated with bromodeoxyuridine (BrdU) for 7 days following amputation (Fig. 2a–f). Subsequently, at 14 dpa, we found a significant increase in the number of BrdU^pos/GFP^pos cardiomyocytes in regenerating hearts versus non-amputated controls (Fig. 2g). We can conclude from this that differentiated GFP^pos cardiomyocytes have re-entered the cell cycle and engaged in DNA replication. We also analysed the position of BrdU-labelled GFP^pos cardiomyocytes within the regenerating heart (Fig. 2h and inset). While the majority of BrdU^pos/GFP^pos labelled cardiomyocytes are concentrated around the wound, a proportion can also be found in regions far from the site of amputation. This suggests that the response to the injury affects the heart in a global manner.

Recent studies have indicated that cardiomyocytes dedifferentiate in order to facilitate proliferation, yet it is unclear how far within the cardiac lineage they regress^7,8. An increase in the expression of the cardiac progenitor associated genes nkx2.5 and hand2 during zebrafish heart regeneration has been reported⁹. However, our own in situ hybridisation analyses failed to detect any significant upregulation of either transcript (data not shown), confirming previous results by our laboratory⁵. Furthermore, genome-wide transcriptome data^10,11 also failed to detect significant changes in the expression of either transcript during zebrafish heart regeneration. These results argue against an extensive dedifferentiation of cardiomyocytes as a pre-requisite for their proliferation in the context of heart regeneration.

To address the extent to which cardiomyocytes dedifferentiate during zebrafish heart regeneration, we analysed regenerating hearts using transmission electron microscopy (TEM). In uninjured hearts (Fig. 3a, b) cardiomyocytes displayed an ordered arrangement of sarcomeres and mitochondria with clearly defined z-lines (Fig. 3b). Following amputation, cardiomyocytes began to detach from one another, creating large intercellular spaces (Fig. 3c, d). The aligned array of actin and myosin filaments visible in uninjured hearts (Fig. 3b) became disorganised (Fig. 3d), associated with a loss of z-line structure. Later, at 7 dpa (Fig. 3e, f), intercellular spaces were also readily visualised as cardiomyocytes detached from one another (Fig. 3e). Furthermore, although sarcomeric filaments were visible, there was a lack of organisation leading to the presence of both transverse and longitudinal sarcomeric structures within individual cardiomyocytes (Fig. 3f). Using these criteria, we were able to count the number of cardiomyocytes undergoing phenotypic changes (Supplementary Fig. 10a–c). Following amputation, the proportion of these structurally/morphologically altered cardiomyocytes increased, peaking at around day 7 (Supplementary Fig. 10c). Furthermore, their distribution within the regenerating heart closely resembled that of BrdU-labelled cardiomyocytes (Supplementary Fig. 10d and inset). Despite these changes, cardiomyocytes appeared healthy and did not display any of the hallmarks associated with cell death^12,13 (Supplementary Fig. 11a–f). TUNEL labelling regenerating hearts further confirmed these observations (Supplementary Fig. 12a, b).

The sequence of events described thus far would indicate that during regeneration, cardiomyocytes detach from one another and disassemble their sarcomeric structure presumably to facilitate cell cycle re-entry. If this is correct, then cardiomyocytes in the process of cell division should not have any discernable sarcomeric structure. To test this hypothesis we labelled dividing cardiomyocytes with both phosphorylated histone 3 (PH3), a well established marker of cells undergoing mitosis, and either GFP or MF20. Despite the rarity of these events (an average of 3 cells per section, n=12 sections from 4 hearts) none of the PH3-labelled cells showed any discernable organisation of their sarcomeric structure (Supplementary Fig. 13a–d). Furthermore we also immunolabelled GFP^pos cardiomyocytes for proliferating cell nuclear antigen (PCNA)(Supplementary Fig. 14a–c). Similar to PH3-labelled cardiomyocytes, PCNA^pos/GFP^pos cardiomyocytes showed disorganized sarcomeric structure (Supplementary Fig. 14b, c) These results indicate that the sarcomere is disassembled in the process of cardiomyocyte divison, similar to the set of events described recently in the mouse⁸.

Next we sought to determine whether cardiomyocyte dedifferentiation during heart regeneration was associated with specific changes in gene expression. Previous reports have indicated an increase in expression of the mitotic checkpoint kinase mps1 in cardiomyocytes during regeneration and that perturbation of this effectively inhibits regeneration⁴. Thus, we examined the expression of polo like kinase 1(plk1), a regulator of cell cycle progression that was detected as upregulated in previous microarray analyses of zebrafish heart regeneration¹¹. plk1 transcripts were markedly upregulated in regenerating hearts at 1, 3, and 7 dpa, as detected by RT-PCR, closely resembling those of mps1 (Fig. 4a). In addition, in situ hybridisation analyses revealed that plk1 expression increased in cardiomyocytes during regeneration (Fig. 4b, c).

To further assess the role of plk1 during heart regeneration, we utilised a recently described embryonic model of heart regeneration¹⁴. Using this model system we were able to specifically ablate cardiomyocytes in 48-hpf embryos. These were subsequently washed and allowed to recover with or without the Plk1 inhibitor cyclapolin 9. In the absence of the inhibitor, 67% of the embryos were able to regenerate their heart. However, in the presence of the inhibitor, this number fell substantially to 17% (the treatments used in these experiments have no effect on wildtype embryos) (Supplementary Fig. 15a–d). We confirmed these results in the adult setting by inhibiting Plk1 activity with cyclapolin 9 in regenerating zebrafish throughout the 30-day recovery period. Inhibition of Plk1 drastically inhibited the regenerative process (Fig. 4d, e). TUNEL labelling showed that this inhibition was not due to an increase in cardiomyocyte apoptosis (on average 1 cell per section, n=3 sections from 3 different hearts) (Fig. 4f). Furthermore we found a significant decrease in the number of BrdU^pos/GFP^pos cardiomyocytes in regenerating animals treated with the inhibitor (Fig. 4g). These results indicate that plk1 is essential for heart regeneration to proceed.

Overall, our studies show that zebrafish heart regeneration is primarily driven by pre-existing cardiomyocytes, rather than by progenitor cells, as has been previously suggested⁹. The tightness and specificity of our cardiomyocyte lineage-tracing system (Supplementary Figs. 1–9, see also Methods for more details) allow us to demonstrate that the vast majority, if not all, of heart muscle cells formed during regeneration arise from pre-existing cardiomyocytes. Even though we cannot formally exclude the possibility that stem/progenitor cells may give rise to cardiomyocytes during this process, in light of our results we can conclude that their contribution to newly formed myocardium would only be marginal. To facilitate proliferation, we found that pre-existing cardiomyocytes undergo limited dedifferentiation. Soon after amputation, cardiomyocytes close to the wound start to disassemble their sarcomeric structure and detach from one another. Furthermore, reduced sarcomeric structure is also observed in zebrafish cardiomyocytes re-entering cell cycle. Microarray analysis also confirms these findings with many sarcomeric genes being downregulated after amputation¹¹. A comparable set of events has also been described in the newt. Here, the expression of cardiac sarcomeric genes is down regulated following amputation, then as regeneration proceeds, the expression returns to pre-amputation levels¹⁵. Similar structural changes are also associated with hibernating myocardium in humans following cardiac injury¹⁶. Hibernating cardiomyocytes typically show a depletion of sarcomeric structure and an expression pattern of structural proteins closely resembling foetal heart cells¹⁷. Although mammalian cardiomyocytes are unable to regenerate, it is tempting to speculate whether they can in fact complete one of the steps involved in this process. We finally show that, as a prelude to proliferation, zebrafish cardiomyocytes adjacent to the wound begin to express regulators of cell cycle progression, such as plk1 and mps1⁴ which are necessary for the regenerative process to proceed. The fact that zebrafish cardiomyocytes dedifferentiate to a limited extent during heart regeneration, rather than undergoing dramatic changes in gene expression, offers hope that heart regeneration can also be induced in mammals. In this respect, our studies provide mechanistic insight into previous findings in mice indicating that forced expression of cell cycle regulators can induce regeneration following cardiac injury¹⁸.

1