Fgf signalling controls diverse aspects of fin regeneration

The regenerative abilities of tissues in higher vertebrates, including mammals, are limited, but several vertebrate species, such as urodele amphibians and teleost fish, retain remarkable capacity for tissue regeneration, which is known as epimorphic regeneration. Numerous morphological and histological studies have shown that epimorphic regeneration of appendages such as limbs, tails and fins proceeds through several distinct steps: wound closure and healing, formation of a thick epidermis termed ‘wound epidermis’, formation of proliferating mesenchymal cells termed ‘blastema’, and tissue outgrowth (Akimenko et al., 2003; Poss et al., 2003; Kawakami, 2010). In the case of zebrafish fin regeneration, recruited blastemal cells are detected at 18 h post-amputation (hpa; Poss et al., 2000), but apparent blastema cell proliferation and outgrowth is observed after 2 days post-amputation (dpa) and thereafter.

Whereas the blastema plays a key role in regeneration by supplying new cells within the regenerated tissue, significance of the wound epidermis has been less clear. A number of studies have suggested that regeneration is impaired or delayed when normal formation of the wound epidermis is disturbed by insertion of the amputated limb into the body cavity, by skin grafting or by surgical removal of the wound epidermis (Goss, 1956; Thornton, 1957; Mescher, 1976; Tassava and Garling, 1979; Campbell and Crews, 2008). Thus, it is conceivable that the wound epidermis plays an essential role during regeneration, but the exact function of the wound epidermis remains unknown.

During the last decade, studies have been focused on identifying the molecular basis of regeneration. In particular, studies on zebrafish have pointed to the involvement of a variety of signalling molecules such as insulin-like growth factor (Chablais and Jazwinska, 2010), Notch (Grotek et al., 2013; Munch et al., 2013), bone morphogenetic protein (Smith et al., 2006), retinoic acid (Blum and Begemann, 2012), Wnt (Stoick-Cooper et al., 2007; Wehner et al., 2014), activin βA (Jazwinska et al., 2007), Sonic hedgehog (Quint et al., 2002), stromal cell-derived factor 1 (Bouzaffour et al., 2009), laminin (Chen et al., 2015), Kit ligand (Tryon and Johnson, 2014), adenosine (Rampon et al., 2014) and hydrogen peroxide (Gauron et al., 2013). Notably, several studies have suggested that Fgf signalling is necessary for fin regeneration (Poss et al., 2000; Lee et al., 2005, 2009; Wills et al., 2008). Furthermore, a genetic screen by Whitehead et al. (2005) showed that the temperature-sensitive mutant of fgf20a does not promote formation of the blastema in zebrafish at a non-permissive temperature, indicating that Fgf20a is at least an essential Fgf ligand during fin regeneration. In spite of the crucial role of Fgf signalling in regeneration, the exact mechanism of Fgf action, including the responsible Fgf ligands and their functions, is poorly understood.

In this study, we performed comprehensive expression analyses of Fgf ligands and showed that fgf20a is expressed in the wound epidermis from an early stage of blastema formation when the expression of msxb (msx1b) and msxc (msx3) genes and the blastema cell proliferation are not activated, whereas fgf3 and fgf10a are differentially activated in the blastema at the following stage concurrently with blastema cell proliferation. Using a cell-transplantation experiment of mesenchymal cells with inducible expression of fgfr1, we showed that Fgf signalling is directly required for the fin ray mesenchyme to form the blastema at the early pre-blastema stage and to activate the regenerative cell proliferation at a later post-blastema stage. Furthermore, we explored the specific functions of Fgf ligands by gain-of-function analyses and demonstrated that Fgf20a does indeed induce the distal blastema marker junbl (junbb) (Yoshinari et al., 2009; Ishida et al., 2010), and that Fgf3 promotes cell proliferation in the blastema explants. Our study reveals the diverse roles of Fgf signals during fin regeneration in zebrafish and how Fgfs in the wound epidermis and blastema cooperatively regulate regeneration.

To identify the Fgf ligands that are involved in fin regeneration, we conducted comprehensive expression analysis by means of reverse-transcription polymerase chain reaction (RT-PCR; Fig. S1A). A low level of expression was detected for some Fgf genes in uncut fins, which may represent a basal expression for normal tissue growth and/or maintenance. However, genes of several Fgf ligands – fgf3, fgf7, fgf10a, fgf10b, fgf12a, fgf12b, fgf13a, fgf20a, fgf20b and fgf24 – were apparently upregulated in response to tissue amputation. Notably, only expression of fgf20a and fgf20b was detectable after 1 dpa (Fig. S1A; Fig. 1A), a stage when the blastema is forming, but cell proliferation in the blastema has not yet begun (Fig. S2).

We further determined the tissue localisation of expression of Fgf genes by tissue-specific RT-PCR analysis using the RNAs isolated from the blastema and wound epidermis (Fig. 1B; Fig. S1B). The specificity of RNA samples was confirmed by the expression of epidermal and blastemal markers, p63 (tp63) and msxc, respectively (Akimenko et al., 1995; Fig. S3). Although cross-contamination with cells between the above samples is inevitable, this approach enabled us to analyse quickly the epidermal or blastemal expression of the Fgf genes in question. Among the regeneration-associated Fgfs, a group of ligands – fgf3, fgf7, fgf10a, fgf10b and fgf12b – was principally induced in the blastema, whereas another group, including fgf12a, fgf13a, fgf20a, fgf20b and fgf24, was principally induced in the epidermis.

The RT-PCR analyses indicated that fgf20a and fgf20b were the earliest Fgf genes activated in the wound epidermis. Other studies have suggested that fgf20a expression is localised to the blastema and/or wound epidermis (Whitehead et al., 2005; Lee et al., 2009). The cause of this ambiguity is thought to be the low resolution of in situ hybridisation (ISH) signals. To confirm the tissue localisation of fgf20a expression, we utilised a transgenic (Tg) zebrafish line with an enhancer trap (Nagayoshi et al., 2008), in which an enhanced green fluorescent protein (EGFP) cassette was inserted in the vicinity of the fgf20a gene (Fig. S4A). In the Tg line, EGFP expression was detected in response to amputation in cells around the amputation site as early as 12 hpa, a stage when the blastema formation is still not evident. EGFP expression became localised to the basal layer of the wound epidermis after 2 dpa and eventually was restricted to cells at the posterior margin of the fin after regeneration (Fig. 1C). Although a few EGFP-positive cells were also seen within fin rays at 12 hpa, it was not clear whether or not this reflects the endogenous fgf20a expression. Such cells were not observed at later stages (Fig. 1C). The principal EGFP expression was colocalised with the epidermal marker p63 (Fig. 1D), and such tissue localisation is in agreement with the results of the RT-PCR analysis (Fig. 1B). We also repeated the ISH analysis and confirmed that the endogenous fgf20a expression coincided with EGFP expression (Fig. S4B). Furthermore, EGFP expression in rhombomere neurons at the prim-15 stage (Fig. S4C) is also consistent with the previously reported pattern of fgf20a expression (Terriente et al., 2012). Taken together, these data supported the notion that fgf20a expression during fin regeneration is recapitulated by the enhancer trap zebrafish line, and that Fgf20a is an Fgf ligand that is locally expressed in the basal layer of the wound epidermis.

The blastema is composed of several distinct cell types, such as non-proliferating or slowly proliferating distal blastema cells, highly proliferative proximal blastema cells, osteogenic cells including the osteoblast, etc. (Fig. S5). To reveal the cellular localisation of expression of the blastema Fgf genes, we performed an ISH analysis of fgf3, fgf7, fgf10a, fgf10b and fgf12b. Many of these did not yield a reliable ISH signal, but only fgf3 and fgf10a were clearly detectable in the blastema (Fig. 1E; Fig. S6). The expression of fgf3 and fgf10a was the strongest in the distal blastema.

Because of the loss-of-blastema phenotype of the fgf20a mutant (Whitehead et al., 2005) and localisation of fgf20a expression to the wound epidermis, it may be hypothesised that the wound epidermis regulates blastema formation via an early Fgf20a signal. To verify the role of early Fgf signalling, we re-evaluated the effects of inhibition of Fgf signalling in a Tg zebrafish that overexpresses dominant-negative Fgf receptor 1 (dnfgfr1) under control of the heat shock protein 70 (hsp70l) promoter (Lee et al., 2005). When Fgf signalling was blocked starting at 0 dpa, the fins actually showed a loss-of-blastema phenotype at 2 dpa, as in the fgf20a mutant (Fig. S7A-D). Before the onset of regenerative cell proliferation (Fig. S2), a basal level of cell proliferation at 1 dpa detected by 5-ethynyl-2′-deoxyuridine (EdU) incorporation was not affected by the inhibition of Fgf signalling (Fig. S7E,F). We also examined apoptosis in the Fgf signalling-inhibited fins, but no significant increase in apoptosis was observed (Fig. S7G,H). These observations support the notion that the early Fgf signalling before 2 dpa, which is mainly mediated by Fgf20a, is required for blastema formation. More intriguingly, fgf20a expression in the wound epidermis was reduced by inhibition of Fgf signalling (Fig. S7D), suggesting that Fgf signalling also participates in maintenance of fgf20a expression in the wound epidermis.

Given that early Fgf signalling is required for blastema formation, we next investigated whether it acts on the fin ray mesenchymal cells to induce blastema formation. To this end, we performed a cell-transplantation experiment, in which donor blastema cells carrying the transgenes hsp70l:dnfgfr1 and Olactb:loxP-dsR2-loxP-EGFP (Yoshinari et al., 2012), which drives the DsRed2 expression in ubiquitous cell types including the osteoblasts (Fig. S8), were transplanted into the host blastema (Fig. 2A). After the transplantation and regeneration, progenies of the transplanted cells mostly contributed to the regenerated mesenchymal cells (Fig. 2B), but only a few cells (11 out of 226) to the Zns5-positive osteogenic cells including osteoblasts (Fig. 2C). This situation enabled us to inhibit Fgf signalling in the genetically labelled mesenchymal cells at a desired time point and to track their behaviour.

The mosaic fins were allowed to regenerate until 7 dpa when the progenies of the transplanted cells were stably integrated and were re-amputated to initiate regeneration in the presence or absence of heat shock. With normal Fgf signalling (no heat shock), the labelled mesenchymal cells distally migrated to initiate the blastema formation (Fig. 3A, −HS; Fig. S9). By contrast, Fgf non-responsive cells were rarely observed within the blastema, which consisted of wild-type cells (Fig. 3A-C, +HS). Besides the reduced contribution of Fgf non-responsive cells to the blastema, overall regeneration, including the expression of a blastema marker, msxc, and the number and distribution of the Zns5-positive osteogenic cells were not changed (Fig. 3D-G). These observations suggest that the early Fgf signalling before 2 dpa is directly required for the fin ray mesenchymal cells to migrate and form blastema.

The later activation of several Fgf genes including fgf3 and fgf10a in the blastema (Fig. S1; Fig. S6) raised a possibility that these Fgfs may perform additional functions during regeneration. We evaluated the overall function of later Fgf signalling after blastema formation by a stage-specific inhibition of Fgf signalling using Tg(hsp70l:dnfgfr1). When inhibition of Fgf signalling started at 3 dpa, a stage after the blastema formation, regeneration was immediately suppressed accompanying the decrease of blastema and the genes that mark the distal blastema (Fig. S10A-D). No significant increase in apoptosis was observed by the later suppression of Fgf signalling (Fig. S10E,F). However, cell proliferation in the blastema was severely decreased by the inhibition of Fgf signalling (Fig. S10G,H), indicating that the later Fgf signalling after 2 dpa is required for cell proliferation in the blastema. This function of later Fgf signalling is in contrast to that of early Fgf signalling, which did not affect basal cell proliferation in the epidermis and mesenchyme (Fig. S7E,F). In addition, the later blockage of Fgf signalling downregulated the expression of epidermal fgf20a as in Fig. S7D as well as the blastemal fgf3 and fgf10a, suggesting that Fgf signalling is also required for maintenance of the expression of epidermal and blastemal Fgf ligands.

Given the later role of Fgf signalling in blastema cell proliferation, we further confirmed whether a direct Fgf signal to the fin ray mesenchyme is required for it. To do this, we again performed the transplantation experiment of hsp70l:dnfgfr1 cells. However, the inhibition of Fgf signalling was started at 2 dpa when the expression of msxc and blastema Fgfs and the blastema cell proliferation have initiated. Without heat shock (Fgf signalling is normal), progenies of the transplanted cells contributed to formation of the entire region of regenerated tissue including the blastema (Fig. 4A, −HS). By contrast, after heat shock the number of dnfgfr1-expressing cells rapidly diminished in the blastema (Fig. 4A, +HS). EdU labelling confirmed that most of the Fgf non-responsive cells (Fig. 4B, green staining) were outside of the proliferation zone (Fig. 4B, brackets) and thus less proliferative (Fig. 4C). Except for the reduced contribution of Fgf non-responsive cells to the newly formed regenerate, overall cell proliferation within regenerates was not changed in the mosaic fins with inhibition of Fgf signalling (Fig. 4D). When the fin ray harboured many dnfgfr1-expressing cells, the length of the regenerate was slightly reduced in the operated fin rays (Fig. 4A, −HS). Altogether, these observations confirmed that the reception of Fgf signal by the fin ray mesenchyme is necessary for maintenance of its regenerative cell proliferation.

The early onset of fgf20a expression, its epidermal localisation during regeneration (Fig. 1; Fig. S1), and the inability to form the blastema in fgf20a mutant (Whitehead et al., 2005) strongly suggest that epidermal Fgf20a could be responsible for the blastema induction and maintenance. To prove the specific function of Fgf20a, we conducted gain-of-function experiments using a Tg zebrafish that overexpresses fgf20a under control of the hsp70l promoter. The Tg ubiquitously expressed fgf20a in both epidermal and blastemal compartments after heat shock (Fig. S11A). However, no drastic differences in length of the regenerating fin were observed between the Tg and wild-type fish that were given daily heat shocks (data not shown).

To analyse the role of Fgf20a in blastema induction, we performed an ISH analysis of msxc and junbl, genes that are expressed in the distal blastema. In comparison with the distally restricted expression of msxc in wild-type fins, the fgf20a-overexpressing fins showed a broader msxc expression (Fig. 5A,B). Significantly, the expression of junbl, an earlier blastema marker (Ishida et al., 2010), was strongly and ectopically upregulated in and around the blastema (Fig. 5C,D), suggesting that fgf20a overexpression causes mesenchymal cells to acquire the distal blastema identity. This situation is complementary to the no-blastema phenotype of the fgf20a mutant (Whitehead et al., 2005) and to the early blockage of Fgf signalling (Fig. 3A; Fig. S7).

Considering the expression of fgf3 and fgf10a in the blastema and the role of later Fgf signalling for blastema cell proliferation, it is thought that the blastema Fgfs including Fgf3 and Fgf10a could be responsible for cell proliferation. To test this possibility, we took advantage of a zebrafish Tg(hsp70l:fgf3) and overexpressed fgf3 during regeneration (Fig. S11B). As in the case of fgf20a overexpression, when fgf3 was overexpressed, we did not observe a drastic difference in length of the regenerating fin in the Tg overexpressing fgf3 (data not shown).

To further examine the effect of fgf3 overexpression on cell proliferation, we assayed cell proliferation by EdU incorporation, but no apparent increase of EdU-positive cells in the amputated fish fins was detected (data not shown). It was thought that cell proliferation in the regenerates may be saturated even in the absence of fgf3 overexpression. Therefore, we developed a blastema cultivation assay, in which the blastemas were separated from the wound epidermis and cultured in a collagen gel (Fig. 6A). In the cultured blastema without the epidermis, cell proliferation rapidly decreased within a few hours accompanying a downregulation of the distal blastema marker junbl (Fig. S12).

When the blastema explants, which were cultured for 6 h at 28°C, were then incubated at 37.5°C to induce fgf3 expression, an apparent increase of the EdU incorporation was detected (Fig. 6B,C). The fgf3 overexpression did not affect cell death in the explants (Fig. S13), therefore the increase of EdU-positive cells was not due to the survival of apoptotic cells by Fgf3. This result demonstrated that Fgf3 does indeed have a mitogenic effect on the blastema mesenchymal cells.

We also tested the effect of fgf20a overexpression on cell proliferation. Significantly, fgf20a overexpression also induced cell proliferation in the blastema explants (Fig. 6B,C). Because fgf20a overexpression induced the distal blastema marker junbl and fgf3 (Fig. 5; Fig. 6D), we suggest that fgf20a overexpression activates cell proliferation via the induction of fgf3. Importantly, despite the similar phenotype of fgf3 and fgf20a overexpression in cell proliferation, fgf3 overexpression did not induce expression of the blastema markers junbl or msxc (Fig. 6E). This indicates that Fgf3, unlike Fgf20a, does not have a blastema-inducing function.

Despite the increasing number of signals that are shown to be involved in regeneration, the mechanisms by which such signals regulate appendage regeneration are largely unknown. In this study, we aimed to determine the exact mechanism of Fgf signalling during zebrafish fin regeneration and revealed that Fgf20a in the wound epidermis and Fgf3 in the blastema regulate distinct processes of regeneration and cooperatively ensure appendage regeneration (Fig. 7).

The blastema and the wound epidermis are the principal components of epimorphic regeneration in urodeles and teleost fish. The importance of the wound epidermis has been suggested in many ways (Goss, 1956; Thornton, 1957; Mescher, 1976; Tassava and Garling, 1979), but its exact role has long been unclear. In this study, we showed that fgf20a expression is localised to the wound epidermis from the early stage of regeneration before the onset of blastema cell recruitment. We further showed that Fgf signalling at this stage is necessary for induction of the blastema, and that Fgf20a induces the distal blastema gene junbl. Taken together, these data strongly suggested that epidermal Fgf20a is an Fgf ligand that acts on the fin ray mesenchyme to promote formation of the blastema.

Fgf20 is a recently identified subfamily of Fgfs that is expressed in the brain (Ohmachi et al., 2000). Recently, it has been demonstrated that Fgf20 knockout mice do not show dermal condensation, which is required for hair follicle formation, and thus the defect results in the loss of guard hair (Huh et al., 2013). Furthermore, it has been reported that the Scaleless chicken mutant, which lacks feathers and limb scales, harbours a nonsense mutation in the Fgf20 gene (Wells et al., 2012). Moreover, the temperature-sensitive fgf20a zebrafish mutant shows gradual fin degeneration (Wills et al., 2008) and malformation of bony tissues such as the skull and scales (Cooper et al., 2013). These phenotypes across species are consistent with the notion that Fgf20 plays a conserved role as an essential epidermal signal during appendage development and regeneration in vertebrates.

In addition to the wound epidermis and blastema, recent studies have highlighted the role of nerves in regeneration. Meda et al. (2016) have used a denervation strategy and demonstrated that sensory neurons are responsible for controlling H₂O₂ levels. Furthermore, Simões et al. (2014) have shown that the expression of genes, including fgf20a, was altered upon denervation. These data suggest that redox regulation by nerves may play an important role for tissue regeneration through regulation of regeneration-associated genes such as fgf20a.

In contrast to the epidermal fgf20a expression, expression of fgf3 and fgf10a was detected in the distal blastema. A number of studies have suggested that Fgfs, including Fgf3 and Fgf10a, have mitogenic activity (Konishi et al., 2006; Shin et al., 2006; Wang et al., 2007; Itoh and Ornitz, 2008); however, their functions during fin regeneration have not yet been explored. In this study, we performed a gain-of-function analysis of fgf3 in the cultured blastema explants and demonstrated that Fgf3 promotes cell proliferation. Intriguingly, Fgf3 overexpression did not show an increase of cell proliferation in uninjured fins (Fig. S14). It seems that only the blastema cells have a competence to Fgf signalling probably due to the presence of Fgf receptor expression (Poss et al., 2000). Although we did not assess the mitogenic activity of Fgf10a, it is reasonable to assume that this protein could possess a function overlapping with that of Fgf3 (Konishi et al., 2006; Shin et al., 2006).

It has been reported that a mutation in the mouse Fgf10 gene causes a complete loss of fore- and hindlimbs, and that a mutation in zebrafish fgf10a displays a loss of pectoral fin (Min et al., 1998; Sekine et al., 1999; Norton et al., 2005). In the case of regeneration, application of exogenous Fgf10 protein to non-regenerative Xenopus limb stumps successfully stimulates regeneration of well-patterned limb structures (Yokoyama et al., 2001). These studies suggest that Fgf10 plays an essential role in the development and regeneration of vertebrate appendages.

Studies in chicks and mice have also shown that Fgf10 expression is activated by epidermal Fgfs in the developing limb, and that it reciprocally signals to the overlying ectoderm to maintain expression of epidermal Fgfs (Ohuchi et al., 1997; Xu et al., 1998). Intriguingly, we also observed that blockage of Fgf signalling results in downregulation of epidermal fgf20a and blastemal fgf10a and fgf3 (Fig. S7D; Fig. S10D), suggesting that a similar positive-feedback mechanism involving Fgf10a may operate during fin regeneration.

In addition to Fgf20a, Fgf3 and Fgf10a, other Fgfs may have additional roles in tissue homeostasis. In salamanders, it has been shown that the juxtaposition of anterior and posterior limb tissue plus innervation is necessary and sufficient to induce complete limb regeneration (Nacu and Tanaka, 2011). Nacu et al. (2016) have recently shown that FGF8 and sonic hedgehog provide complementary cross-inductive signals that are required for limb outgrowth. Though we did not detect fgf8 as one of the regeneration-induced Fgf ligands, fgf8a expressed in the uninjured epidermis (Fig. S1) could provide a positional cue together with other signals.

In this study, we used a novel approach for dissecting the target cells and analysing the functions of Fgf signalling during regeneration. According to a procedure described previously (Yoshinari et al., 2012), we transplanted hsp70l:dnfgfr1 blastema cells that ubiquitously express the β-actin1:dsRed2 transgene into a wild-type blastema region. Most of the progeny cells of the transplanted blastema gave rise to fin ray mesenchymal cells (but not osteogenic cells) after regeneration (Fig. 2C). Therefore, by re-amputating fins, we were able to specifically inhibit the action of Fgf signals to the fin ray mesenchymal cells at a desired time point and examine the effect by live cell imaging. Using this mosaic knockdown approach, we successfully demonstrated a direct role of Fgf signals on the mesenchymal cells and observed the effect on their behaviour under live conditions.

Cell transplantation has generally been performed during the early embryonic stage, but targeting the transplanted cells to specific tissues has been difficult. Blastema transplantation is expected to be an efficient method for manipulating a specific signal in labelled mesenchymal cells. Using a similar approach with other signalling molecules in future studies, we will be able to further dissect the cellular behaviour and responses to these signals and elucidate how the signals in question regulate appendage regeneration.

In this study, we revealed that Fgf20a is a signal from the wound epidermis for initiating distal blastema formation. We also showed that blastemal Fgf3 serves as a signal that promotes regenerative cell proliferation of blastema cells. Though it has been known that Fgf signalling is necessary for regeneration, our study firstly elucidated the responsible Fgf ligands and the mechanisms by which fin regeneration is regulated. Given the importance of epidermal Fgf20a for initiation of blastema formation and regeneration, questions remain regarding how fgf20a expression is regulated during regeneration. Identification of upstream regulators of fgf20a and characterisation of the cross-talk with other signalling pathways will further advance our understanding of the regeneration process.

Zebrafish of a wild-type line, which has been maintained in our facility for more than 10 years by inbred breeding, were kept in a recirculating system in a 14 h day/10 h night cycle at 28.5°C in accordance with the Animal Research Guideline of Tokyo Institute of Technology. The Tg(hsp70l:dnfgfr1-EGFP)^pd1 line was obtained from the Zebrafish International Resource Center (ZIRC). The Tg(hsp70l:fgf3)^x27 line was a generous gift from Dr Bruce B. Riley. Experiments with adult fin regeneration were performed on 3- to 12-month-old zebrafish. The fish were anaesthetised with tricaine (3-aminobenzoic acid ethyl ester; Sigma) and their caudal fins were cut with a scalpel. For reproducibility, fins were amputated in a straight line in the middle of the central fin ray. The regenerating fins were collected at appropriate time points for further analysis. Heat-shock induction of dnfgfr1, fgf3 or fgf20a was performed at 38°C for 1 h in a small water bath. Heat-shock treatment of fish was performed at 6 h before amputation and every 24 h subsequently. For quantitative analysis of cell proliferation, the third to fifth fin rays from the dorsal or ventral side were used for making tissue sections and scoring.

Total RNA samples were isolated from regenerating fin tissues (three to six fins) using TRIzol (Gibco BRL). For isolation of the wound epidermis and blastema, the regenerating fins at 2-3 dpa were incubated with 25 mg/ml pancreatin (WAKO) in Hank's solution and mechanically split into the epidermal and mesenchymal parts with fine needles and/or forceps. cDNAs were synthesised using the Thermoscript RT-PCR Kit (Invitrogen) with random hexamer as primers. The synthesised cDNAs were diluted to 100 μl with 0.1× TE buffer (1 mM Tris-HCl pH 8.0, 0.1 mM EDTA) and stored at −30°C. PCR was performed with rTaq polymerase and primers listed in Table S1. Forward and reverse primers were designed on separate exons so that only PCR products from cDNAs were amplified. We performed the PCR more than four times using 30, 33, 35 and 37 cycles and obtained similar results.

Whole-mount ISH was performed according to a previously described protocol (Thisse and Thisse, 2008). After the colour reaction, the samples were mounted with 80% glycerol. For RNA probe synthesis, the Fgf coding sequences were amplified by PCR using KOD polymerase (Toyobo) and cloned into the Zero Blunt TOPO vector (Clontech). Primers for cloning of the Fgf genes are listed in Table S2.

Cell proliferation was detected by labelling with Edu using the Click-iT EdU Imaging Kit (Life Technologies). Briefly, zebrafish were incubated in a solution containing 50 μM EdU in fish water (0.3% artificial sea salt, 0.0001% Methylene Blue) for 12 or 6 h at 28.5°C. EdU was detected according to the manufacturer's instructions. The samples were mounted with 80% glycerol containing 25 mg/ml triethylenediamine (DABCO, Nacalai Tesque) as an anti-fading reagent. The labelled cells were counted using the acquired confocal z-stack images.

The blastema transplantation was performed as described previously (Yoshinari et al., 2012). The zebrafish lines Tg(Olactb:loxP-dsR2-loxP-EGFP) (Yoshinari et al., 2012) and Tg(hsp70l:dnfgfr1) (Lee et al., 2005) were crossed to generate the double-Tg line. The caudal fin of the double-Tg line was amputated and allowed to regenerate until 2 dpa. The blastema and wound epidermis were manually separated with 30-gauge needles in Hank's solution and cut into small pieces. The donor blastema was transplanted into the host blastema region by means of a transplantation glass needle. The operated fins were allowed to regenerate for 5 days and re-amputated for analysis of the behaviour of DsRed2-labelled mesenchymal cells.

Tissues were fixed with 4% PFA (overnight, 4°C) and then briefly washed twice with PBS containing 0.1% Triton X-100 (PBTx), dehydrated using methanol, and stored at −30°C. Immunostaining was performed on tissue sections according to a standard protocol (Ishida et al., 2010). Polyclonal anti-DsRed antibody (632393, Invitrogen) and anti-p63 antibody (ab97865, Abcam) were used at 1:1000. Anti-Zns5 antibody (Developmental Studies Hybridoma Bank) was used at 1:200. The sections were counterstained with 4′,6-diamidino-2-phenylindole (DAPI; 0.1 μg/ml in PBTx; Invitrogen) and mounted with 80% glycerol containing DABCO. Images were taken using a confocal microscope (Fluoview FV1000, Olympus).

The Tg zebrafish line that expresses fgf20a under control of the hsp70l promoter was generated by injecting a plasmid, pTol2 (hsp70l:mCherry-2a-fgf20a), into fertilised eggs. The construct was made by replacing the creERt2 sequence of the pT2(hsp70l:mCherry-2a-creERt2) (Yoshinari et al., 2012) with the fgf20a coding sequence. The Tg line, which is simply referred to as Tg(hsp70l:fgf20a) in this paper, was identified by the expression of mCherry after heat shock and was used for analyses after two generations of out-crosses with wild-type fish.

Regenerating tissues were collected from the Tg(hsp70l:fgf3), Tg(hsp70l:fgf20a) and wild-type fish at 3 dpa. After removal of the wound epidermis, the blastema was embedded into a collagen gel (Cellmatrix type I-A, Nitta Gelatin). The embedded explants were placed in L15 medium containing 100 U/ml penicillin G (Nacalai Tesque) and 100 μg/ml streptomycin  (Nacalai Tesque) and incubated at 28.5°C for 6 h. For EdU labelling, the tissues were cultured in the presence of 10 μM EdU at 37.5°C for 4 h. The EdU-labelled cells were counted using the acquired confocal z-stack images.

We thank the ZIRC for providing the Tg(hsp70l:dnfgfr1) ^pd1 and Dr Bruce B. Riley (Texas A&M University) for providing the Tg(hsp70l: fgf3)^x27.