Splashed E-box and AP-1 motifs cooperatively drive regeneration response and shape regeneration abilities

Multicellular organisms maintain tissue integrity by regenerating injured tissues. However, their abilities differ depending on species, tissue, and developmental stage (Kumar et al., 2007; Kawakami, 2010; Poss, 2010; Tanaka, 2016). Such heterogeneity is thought to be due to changes in the expression of key genes because this ability is not specific to certain evolutionary clades (Bely and Nyberg, 2010).

During regeneration, cells sense injuries and initiate the expression of related genes, where the enhancers, play a pivotal role (Visel et al., 2009). Several studies in zebrafish and killifish have identified regeneration-response enhancers (RREs) (Kang et al., 2016; Pfefferli and Jaźwińska, 2017; Begeman et al., 2020; Lee et al., 2020; Thompson et al., 2020; Wang et al., 2020; Goldman and Poss, 2020; Cao et al., 2022) and shown that the RRE response to injury is conserved in the fin, heart, and fingertips of neonatal mice.

Activator protein 1 (AP-1)-binding motifs have been suggested to be necessary for RRE activity (Kang et al., 2016; Begeman et al., 2020; Wang et al., 2020; Cao et al., 2022). AP-1, a heterodimeric protein consisting of Jun and Fos, is essential for several biological processes (Hess et al., 2004). However, whether only the AP-1 signal is sufficient for initiating a precisely regulated regeneration process and whether additional signals are also required remain unknown.

While recent studies have adopted high-throughput chromatin accessibility analyses to search for RREs in active chromatin regions, we identified RREs by comparing the surrounding genomic sequences of zebrafish regeneration-induced genes. Through a series of in vivo transgenic assays for amputation response and comparison of the identified RREs, we revealed that the combination of two transcription factor (TF)-binding motifs, E-box and AP-1, functions as an RRE. Our study highlighted that the fidelity of regeneration-induced gene transcription is ensured by merging two different signals activated by injuries.

Zebrafish fin folds, and fins have been used as powerful regeneration models to assess the process of tissue regeneration (Fig. 1A; Kawakami and Yoshinari, 2011). Our previous study identified genes that are induced during fin fold and fin regeneration (Yoshinari et al., 2009). When the genomic DNA sequence around zebrafish fn1b, a gene induced in the wound epidermis (WE), was compared with that around msxc, a gene induced in the blastema (Akimenko et al., 1995), we found highly conserved sequences of 200–500 bp, termed as E1/2 and E4 (Fig. S1A). Multiple copies of homologous sequences (> 60% nucleotide identity) were found around fn1b, msxc, and other regeneration-induced genes such as junbb, dlx5a, and junba (Fig. S1B). In addition to E1/2 and E4, other highly homologous sequences, E5 and E6, were found by comparing the surrounding sequences of junba, junbb, and dlx5a (Fig. S1A).

The database search revealed that the identified conserved sequences were transposons (TEs), such as short-interspersed elements (SINEs) and non-autonomous DNA transposons (MITEs) (Fig. 1B). These TEs have a large number of copies in the genome, 15,972 (hAT-N21B_DR, E1/2-like), 95,650 (HE1_DR1, E4-like), 16,907 (DNA-X-4_DR, E5-like), and 17,834 (TDR7, E6-like) (Fig. S2), many of which are estimated to be inserted into the genome more than ten million years ago based on the divergences from the TE consensus sequences.

In mammals, it has been shown that several TEs possess cis-regulatory functions and serve as tissue-specific enhancers (Chuong et al., 2016; Nishihara, 2019), though the function of TF motifs in the TE life cycle remains debatable (Hermant and Torres-Padilla, 2021). In this study, we aimed to examine whether the conserved TE sequences could direct transcriptional activation in response to tissue amputation. To test the RRE activity, we performed an enhanced green fluorescent protein (EGFP) reporter assay in the larval fin fold (Kawakami et al., 2004) and in the young adult fin of F0 animals (Fig. 1C). Positive responses in the F0 assay were validated in stable Tg lines (at least two lines).

We first tested Tg(4.5k:egfp), which has a −4.5 kb region of the fn1b containing E1, E4, E5, and E6, and successfully detected EGFP induction in response to fin fold and fin amputations (Fig. 1D; Fig. S3A). EGFP induction was mostly observed in the basal layer of WE and in a group of mesenchymal cells proximal to the amputation plane, and the expression was similar to that of Tg(fn1b:egfp) (Fig. S3B). The response was also detected in the −3.2 kb Tg, which excludes E4 and E1 sequences, and −1.8 kb Tg (Fig. 1E; Fig. S3C–E), but further truncated regions, −0.8 and −0.4 kb, did not display the RRE activity (Fig. S3C), indicating that at least one RRE exists in the region between −1.8 and −0.8 kb containing E5. Interestingly, while −4.5 and −3.2 kb promoters drove the EGFP expression in the WE, EGFP induction of −1.8 kb Tg shifted to the blastema and fin ray mesenchymes close to the amputation site (Fig. 1E), suggesting that a tissue-specific enhancer element that modifies the responding tissue may exist between −3.2 and −1.8 kb region.

To further test whether the RRE activity is contained within TE sequences, we firstly focused on E4 and E5 for their RRE activity. The construct E5-0.7k, in which E5 was placed upstream of the −0.7 kb promoter, displayed RRE activity in mesenchymal cells (Fig. 2A; Fig. S3F,G) as in the −1.8 kb Tg. However, the response was not detected in the construct where E5 was placed upstream of the short synthetic minimum promoter (miniP) (Fig. S3F), indicating that E5 functions as an RRE only when placed in combination with a helper element within the −0.7 kb promoter. We further tested the RRE activity of E4 and found that E4 also displayed RRE activity in combination with the −0.7 kb promoter (Fig. 2B; Fig. S3F,H).

In addition to RREs in E4 and E5 TEs, we also found an additional RRE in non-conserved genomic region. We initially sought to test hAT-N21B_DR and made an E2-containing construct, E2L, but a short flanking non-conserved sequence (E2S, ∼100 bp) was unintentionally cloned along with the highly conserved E2 sequence. Interestingly, E2L displayed RRE activity in combination with the miniP and drove EGFP expression in WE (Fig. 2C; Fig. S3F,I), indicating that E2L functions as a standalone RRE. However, when we tested the region of E2 homologous to hAT-N21B_DR, it did not display the RRE activity, instead the short non-TE element, E2S, which has no homologous sequence in the genome, functioned as an RRE (Fig. 2D; Fig. S3F,J). Though it is speculated that the TE sequence around E2S were lost during evolution or that E2S was formed by other mechanisms, the origin of the E2S RRE is unknown.

We next searched for TF-binding sites contained in RREs, E4, E5, and E2S and found that the E-box motif, 5′-CANNTG-3′, which is bound by the E2A subgroup of basic helix-loop-helix (bHLH) TF factors such as Tcf3/12, MyoD, Twist, and Myc (Jones, 2004; Cadigan and Waterman, 2012; Wang and Baker, 2015), is commonly found in E2S, E4, and E5 (Fig. 3A,C; Fig. S4A; Table S1), but not in E2 TE that did not display RRE activity (Table S2; Fig. S4B).

Furthermore, the comparison of E2S and fn1b −0.7 kb promoter showed that AP-1 motifs are common between E2S and E6 (Fig. 3B,C; Fig. S4A) but not E2 TE (Table S2; Fig. S4B). AP-1 is a leucine zipper TF that is composed of heterodimers of Fos, Jun, and Atf3 (Hess et al., 2004). Notably, the E-box and AP-1 motifs are also contained in many of the previously identified RREs, such as LEN (Kang et al., 2016), Z-IEN (Wang et al., 2020), careg (Pfefferli and Jaźwińska, 2017) (Fig. 3D), and many other RREs (Table S3).

To validate the roles of E-box and AP-1 motifs for the regenerative-response, we tested the removal of the E-box, AP-1, or both from the E2S construct. Amputation-induced EGFP expression was abrogated in all cases (Fig. S3F). Loss of RRE activity was also observed in the mutant versions of the constructs, in which the E-box and AP-1 motifs were replaced with a stretch of adenines (Fig. S3F). Furthermore, the RRE activity of E5-0.7k construct was lost by removing two E-box motifs, leaving the other motifs intact (Fig. S3F). Moreover, a copy of HE1_DR1, in which the E-box motif sequences were not conserved (Fig. S4B), did not exhibit a regenerative-response (Fig. S4C). These results support that both E-box and AP-1 motifs are required for the regenerative-response.

To further prove that these TF motifs are sufficient for the regenerative response, we created constructs containing tandem repeats of the E-box and/or AP-1 motif. The constructs that contained either E-box or AP-1 did not show the RRE response (Fig. S4D), whereas those containing both the E-box and AP-1 motifs, 6×E-0.7k and 6×E-6×A-miniP, displayed a response (Fig. 3E–H; Fig. S4E,F). Further, the response decreased in the construct with the mutated E-box sequence, 6×E(m)-0.7k (Fig. S4D). These results support that the combination of the E-box and AP-1 motifs is sufficient for a regenerative response.

More interestingly, as observed in the −3.2 kb and −1.8 kb Tgs, a difference in the responding cell types was also observed between 6×E-0.7k and 6×E-6×A-miniP. While the 6×E-0.7k Tg displayed the EGFP expression in both the mesenchymal cells and the distal part of the WE, the EGFP was only seen in the distal part of the WE in the 6×E-6×A-miniP Tg (Fig. 3F,H). It is speculated that a sequence variation of AP-1 motifs or any cis-regulatory elements in the −0.7 kb promoter could drive the mesenchymal response.

To estimate the conservation of RREs in E4- and E5-type TEs, we counted the numbers of HE1_DR1 and TDR-7 TEs with conserved E-box (11 bp) and AP-1 motifs (9 bp) and found that 37.1% of HE1_DR1 (35,469 out of 95,650) retain either of three E-box motifs in E4, and that 18.0% of TDR-7 (3218 out of 17,834) retain either of two AP-1 motifs in E6 (Table S4). Besides the motifs within these TEs, there are many additional conserved E-box (total 86,504 copies) and AP-1 motifs (total 46,699 copies) throughout the genome (Table S5). As the expected frequencies of 11 bp and 9 bp sequences in the zebrafish genome are 336 and 5378 copies, it is suggested that the zebrafish genome is highly enriched for E-box and AP-1 motifs. These observations suggest that TEs had a significant contribution, if not all, for spreading potent E-box and AP-1 motifs in the genome.

Thus, we found that E4, E5, and E2S display RRE activity; however, it should be noted that these RREs are not necessarily to be the enhancers for endogenous fn1b gene regulation. Due to the large number of conserved E-box and AP-1 motifs in the genome, these motifs are found in the vicinity of many genes including regeneration-induced genes, where they distribute both in the active and non-active chromatin regions (Fig. S5). If one sought to identify genuine enhancers for endogenous gene regulation, loss-of-function analyses like RRE knockout are essential. However, the identification of endogenously active enhancers would be difficult or impossible, because multiple RREs like E2S, E4, and E5 in the fn1b promoter could redundantly function as RREs (Thompson et al., 2020).

While E2S expresses EGFP in the WE, E4 drives the mesenchymal response despite having the same E-box sequence as E2. It is speculated that the difference of the responding cell types could be due to the combination of bHLH and AP-1 factors that preferentially bind to the RRE sequence variations. Alternatively, a cis-regulatory element such as that suggested to reside between −3.2 and −1.8 kb may alter the response from mesenchyme to epidermis (Fig. 1D,E). Although this region contains several E-box and AP-1 motifs (Table S6), the region did not display the RRE activity in the Tg assay (Fig. S3F). Thus, the region is unlikely to be an RRE, but further studies are required to reveal the mechanism of tissue-specific response.

It has been shown that a robust regenerative response is observed when the damage involves the regeneration of multiple tissues including fin ray bone and inter-ray tissues (Gauron et al., 2013; Wang et al., 2020). To validate whether the response by the E-box/AP-1 RRE is regeneration-specific, we used the Tg(E2L:egfp) (WE-response) and the Tg(E4-0.7k:egfp) (mesenchyme-response) and examined the response to ray injury and inter-ray incision. Both Tgs displayed prominent EGFP induction by ray injury but not by an inter-ray incision (Fig. 4A), confirming that the E-box/AP-1 RRE is activated by a regenerative signal.

We further confirmed that RRE responds to injuries in other tissues, such as pectoral fins (Fig. S6A) and scales (Fig. S6B), and larval trunk through the neural tube, somites, and notochord (Fig. S6C). Additionally, we examined responses during heart regeneration. Although fn1b expression is induced in epicardial cells (Fig. 4B) (Wang et al., 2013), the response of Tg(1.8k:egfp) was induced in cardiomyocytes around the amputation site.

Furthermore, we introduced 1.8k:egfp and 6xE-0.7k:egfp constructs into Xenopus laevis, an animal that displays incomplete regenerative abilities depending on the developmental stage (Phipps et al., 2020), and tested the response after limb amputation at regenerative stage. For both constructs, EGFP induction was detected in the amputated limbs (Fig. 4C). These results demonstrated that E-box/AP-1 also functions as an RRE in Xenopus limb buds. Because fn1 gene is known to be induced in various organs and species in response to tissue traumas and stresses (Gomes et al., 2021), it is speculated that equivalent enhancers and injury-response machinery also exist in amphibians and mammals.

We further investigated whether the signals through the E-box/AP-1 RRE are essential for the response and regeneration. It has been shown that Junba and Junbb, which plays an essential role in fin regeneration (Ishida et al., 2010), are phosphorylated by Jun N-terminal kinase (JNK). We tested the effect of SP600125, a JNK inhibitor, and observed that amputation-induced EGFP expression was downregulated by this inhibitor (Fig. 4D,E). Notably, EGFP induction in epidermis and mesenchymes proximal to the amputation site of the Tg(E2L:egfp) and Tg(E4-0.7k:egfp), respectively, were suppressed by the inhibitor, indicating that the decreased EGFP expression was not due to the absence of responding tissues but due to the decreased transcriptional response.

We further tested the knockdown of the E-box-mediated signal by expressing the dominant-negative form of Lef1 (dnLef1) (Fig. 4F). The canonical Wnt pathway, activated at the cell membrane, leads to the translocation of β-catenin into the nucleus, where it interacts with Tcf/Lef through its N-terminal domains. It has been suggested that this interaction converts Tcf/Lef proteins from repressors to activators (Vacik and Lemke, 2011). We verified that dnLef1 expression during embryonic development induced abnormalities that resembled those observed with decreased β-catenin signalling (Hao et al., 2013) (Fig. S7). As observed in the JNK inhibition, dnLef1 expression impaired fin regeneration and caused a decrease of EGFP expression in the epidermis and mesenchymes proximal to the amputation site in addition to that in the blastema and WE (Fig. 4G,H). Together, these data indicate that both Tcf/Lef and JNK/Jun signals through the E-box and AP-1 motifs, respectively, are required for the transcriptional response of the RRE.

Here, we newly identified several RREs and showed that they commonly contain two TF-binding motifs, E-box and AP-1, which suggested that signals through these motifs cooperatively activate gene expression (Fig. S8). Our study highlighted that RREs containing E-box and AP-1 mediate the regeneration-induced transcriptional response in a variety of tissues and in other species. In all Tg lines, EGFP expression was not observed during development and homeostasis, indicating that E-box/AP-1 RREs are specifically used for the regeneration response. In future studies, it would be interesting to further clarify the cooperation of two TF motifs such as spatial relationship, density, distance, and sequence variation between AP-1 and E-box motifs.

As discussed by Wang et al. (2020), the AP-1 complex has an ancient evolutionary origin and may have an original function such as injury response. It is speculated that additional signals mediated by the E-box have been acquired later in evolution to ensure correct regulation of the regenerative response.

More significantly, we showed that a large number of potential E-box and AP-1 motifs exist in the genome, a significant portion of which are thought to be splashed by TEs such as HE1_DR1 and TDR-7 (Nishihara, 2020). Huge reservoirs of E-box and AP-1 motifs in the genome may have helped accelerate the formation of regeneration gene networks.

Zebrafish were maintained in a recirculating water system under a 14 h day/10 h night photoperiod at 28.5°C. Zebrafish larvae [2–4 days post fertilisation (dpf) and adult fish (3–12 months old)] with similar sex ratios were analysed for all experiments, unless otherwise specified. The wild-type (WT) zebrafish strain used in this study was originally derived from the Tubingen strain and has been maintained in our facility for more than 10 years by inbreeding. Tg(fn1b:egfp) (Shibata et al., 2018) was used in this study in addition to the Tgs established in this study.

Animal experiments were performed in strict accordance with the recommendations of the Act on Welfare and Management of Animals in Japan and the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All animals were handled in accordance with the Animal Research Guidelines of the Tokyo Institute of Technology. The study protocol was approved by the Committee on the Ethics of Animal Experiments of the Tokyo Institute of Technology. All surgeries were performed under 0.002% tricaine (3-aminobenzoic acid ethyl ester, Sigma-Aldrich) anaesthesia, and every effort was made to minimise suffering.

The sequence homology search between BAC sequences CH73-4O17 (fn1b), CH211-217G15(msxc), CH73-348N11(junbb), CH211-57N23 (dlx5a), and CH211-145G9 (junba) was performed with the default settings using Genetyx software (Ver. 10.1.5). Harr-plot analysis was performed under the following conditions: unit size to compare=20 and dot plot matching number=15.

The TEs that were homologous to E1, E4, E5, and E6 were identified based on the repeat annotation of the zebrafish genome assembly GRCz11/danRer11 available in the UCSC Genome Browser. Consensus sequences of the four TEs were retrieved from RepBase (Bao et al., 2015). The number of copies of the four TE families was determined based on the same annotation data from UCSC Genome Browser. The hAT-N21B_DR repeats associated with E2 were counted as a subfamily of hAT-N21_DR.

To investigate the chromosomal distribution of the four TEs, a standalone RepeatMasker analysis was conducted with the rmblast search engine under the sensitive (-s) option (http://www.repeatmasker.org), using the zebrafish TE consensus sequence library (version 20181026) obtained from RepBase. The number of copies for each of the hAT-N21B_DR, HE1_DR1, DNA-X-4_DR, and TDR7 families in 1 Mbp windows of the zebrafish chromosomes were counted from the RepeatMasker output, and the TE densities were heat-mapped with the RIdeogram (Hao et al., 2020). For age distribution, the number of TE copies was counted for 1% bin of TE sequence divergence from the consensus sequences based on the RepeatMasker output.

Generally, TEs inserted at older ages represent higher divergences from the consensus sequence, because the consensus sequence is theoretically identical to the sequence at the time of insertion. In the human genome, 3%, 7%, 12%, and 18% divergences of TE copies indicate they have been inserted 25, 50, 75, and 100 million years ago, respectively (Lander et al., 2001). These ages correspond to the TE divergences of 9%, 18%, 27%, and 35% in mice, respectively (Mouse Genome Sequencing Consortium, 2002). However, such an estimation of TE insertion time has not been reported in fish. Therefore, we assumed that the evolutionary rate of zebrafish is equivalent to that of humans or mice, and the E1, E4, E5, and E6 sequences with 12–15% divergence from the consensus sequences were considered to have been inserted tens of millions of years ago.

To construct a series of fn1b promoter constructs, the respective regions of fn1b were amplified by polymerase chain reaction (PCR) using KOD Plus Neo DNA polymerase (Toyobo, Osaka, Japan) with primers containing XhoI and AgeI restriction enzyme (RE) sites. The amplified DNA was cloned into pCR Blunt II-TOPO (Thermo Fisher Scientific), excised with REs, and inserted into pT2KXIGDin (Urasaki et al., 2006). The 1.8k:egfp construct was constructed from 3.2k:egfp by removing the sequence between XhoI and PflmI. To facilitate the identification of transgenic fish, an egfp cassette under the control of the crystalline alpha A promoter (cryaa:egfp) was introduced at the EcoNI site upstream of the test DNA insertion site. The primers used for the PCR are listed in Table S5.

E4, E5, and E2L were cloned from the fn1b BAC clone by PCR using primers containing RE sites. The E4-, E5-, and E2L-miniP constructs were created by inserting cloned sequences in front of miniP and egfp (Shimizu et al., 2012). The cryaa:egfp cassette was introduced at the EcoNI site upstream of the test DNA. To construct the plasmids with the fn1b −0.7 kb promoter, miniP was replaced with a −0.7 kb promoter at SalI and BamHI. E2, E2S, and a series of constructs carrying the AP-1 and/or E-box deletion or mutation were created from E2L by removing the partial sequences through PCR-mediated mutagenesis (KOD Plus Mutagenesis Kit, Toyobo).

To create constructs with the tandem repeat of E-box (6xE), AP-1 (6xA), or E-box/AP-1 (6xE-6xA), synthetic oligonucleotides were annealed and inserted in front of the miniP or −0.7 kb promoters.

HE1_DR1(m) was cloned from the zebrafish genome and a randomly chosen clone was used to determine the sequence and create the constructs. All the constructs were confirmed by sequencing.

Tol2 transgenesis (Kawakami, 2007) was used to generate the transgenic lines. Plasmid DNAs at concentrations of 25–40 ng/μl and 25 ng/μl of transposase mRNA were injected into fertilised eggs at the one-cell stage.

Fin-fold amputation was performed using a scalpel at 3–4 dpf, as previously described (Kawakami et al., 2004). Fins of young (4–8 weeks post fertilisation; wpf) or mature zebrafish (3–12 months post fertilisation; mpf) were cut at approximately the middle of the fin. The regeneration response of constructs that were positive in the F0 assay was further confirmed in more than two stable transgenic lines to avoid a possible positional effect of the integrated genomic site. To isolate the F1 offspring, F0 embryos were raised to adulthood and then randomly outcrossed or inbred to obtain F1 carriers. In the case of cryaa:egfp cassette-containing constructs, the carriers were initially screened by lens EGFP fluorescence.

To induce scale regeneration, approximately ten scales were removed from the trunk region of each fish by using forceps. For the wound healing assay, the fin was punctured with a fine needle at the ray or inter-ray region.

The Jaspar 2022 open access database (9th release) (https://jaspar.genereg.net/) (Castro-Mondragon et al., 2022) was used to search for TF sites. The search was performed on the Mus musculus CORE collection (251 motifs) using the default settings. The potentially significant TF sites (highest Jaspar score>10.0) are shown in Fig. 3A–D and Fig. S3A,B and Tables S1–S4.

Heart resection was performed as previously described (Poss et al., 2002; Ito et al., 2014). Briefly, WT or Tg(1.8k:egfp) zebrafish at 6–12 months of age were anaesthetised with tricaine and placed ventral side up into a moist, slotted sponge. Using micro scissors, a small incision was made in the area ventral to the position where the heartbeat was visible. The chest of the fish was pressed with tweezers to completely expose the heart, and 20–30% of the lower part of the ventricle was removed with micro scissors. After surgery, the fish were returned to water and stimulated to breathe by sending fresh water over the gills with a pipette. Fish were further allowed to recover for 24 h in a tank with bubbling aeration. At seven days post-surgery, the hearts were dissected and fixed with 4% (w/v) paraformaldehyde (PFA) in phosphate-buffered saline (PBS).

The constructs, along with decondensed sperm nuclei and oocyte extracts, were injected into unfertilised eggs of X. laevis (Kroll and Amaya, 1996). GFP fluorescence from the alpha-crystallin promoter was used to identify tadpoles with a higher transgenic efficiency. Xenopus stages were identified according to the normal table of X. laevis (Nieuwkoop and Faber, 1994). Amputation of the limb buds of tadpoles under anaesthesia with tricaine was performed at St. 52–54 using micro scissors, according to a previously published procedure (Hayashi et al., 2015).

Wholemount in situ hybridisation (ISH) was performed according to a standard protocol (Thisse and Thisse, 2008). A region of the EGFP coding sequence was amplified by PCR and the product was used as a template to synthesise the RNA probe. The underlined part of the primer (Table S5) denotes the T7 promoter.

Zebrafish fins were fixed in PFA at 4°C overnight, subsequently dehydrated with methanol, and stored at −20°C. The stored samples were rehydrated with PBT (1× PBS, 0.1% Tween-20), equilibrated with 20% (w/v) sucrose, embedded in OCT compound (Tissue-Tek, Sakura Finetek), and stored at −20°C. Cryosections 16–20 µm in thickness were prepared for histological analysis.

For immunofluorescence staining, cryosections were washed twice with PBS and several times with PBT to remove any residual OCT compound. Antibody staining was performed as previously described (Shibata et al., 2016). GFP antibody (GF090R, Nacalai Tesque; #598, MBL) was used at 1:1000 dilution. The zns5 antibody was used at a 1:100 dilution in the hybridoma supernatant (Developmental Studies Hybridoma Bank). MF20 antibody (Invitrogen) was used at a 1:1000 dilution. The sections were counterstained with 4′,6-diamidino-2-phenylindole (DAPI; 0.1 μg/ml, Invitrogen) and mounted with 80% glycerol containing 25 mg/ml triethylenediamine (DABCO, Nacalai Tesque). Images were captured using a confocal microscope (FluoView FV1000; Olympus).

SP600125 (Selleck) was dissolved in dimethyl sulfoxide (DMSO) at 15 mM and stored at −80°C. The stock solution was diluted to a working concentration using E3 buffer. When testing the effect of SP600125 from 0 days post amputation (dpa), the compound was administered to zebrafish at least 6 h before the fin amputation. DMSO (0.1%) was used as the vehicle control. The chemical solution was changed every day.

The dominant-negative form of the lef1 plasmid, pTol2(hsp70l:mKO2-2a-dnlef1) (Akieda et al., 2019), was a generous gift from Toru Ishitani (Osaka University). A stable Tg line was generated at our facility. The line was established by screening the F1s for mKO2 fluorescence after a heat shock at 38°C for 2 h. When testing the effect of dnLef1 from 0 dpa, the first heat shock was performed 6 h before fin amputation.

The fluorescent images were captured by the DFC365Fx B/W camera and LAS X software (Ver. 3.7.0) on the FA6000 microscope (Leica) under the same non-saturated conditions for each experiment. The acquired fluorescent images were binarised at a fixed condition for each experiment using NIH ImageJ 1.49 to quantify the fluorescence. All statistical analyses were performed using Microsoft Excel 2013.

No statistical methods were used to determine the sample size. Sample sizes were chosen based on previous publications, and experiment types and are shown in figure legends. After selecting larvae or fish with WT morphology, the clutch-mates were randomised into different groups for each experiment. No animal or sample was excluded from the analysis, unless the animal died during the procedure. Most assessments of the phenotypes and expression patterns were repeated in at least three independent experiments. Whenever possible, blinding was performed during the data collection and analysis. In some experiments, when embryos had to undergo specific treatments, blinding was not possible, as the same investigator processed the samples and collected data. The sample sizes are indicated in the figures or legends.

For the F0 assay of the regenerative response, fin-fold or fin amputation was performed in animals with apparent EGFP expression in the lens, which reflects the rate of transgenesis. The numbers of larvae or fish with amputation induced EGFP expression were scored; however, the ratios of EGFP response does not represent the strength of the enhancers because the efficiency significantly varied in the respective injections. Throughout the Tg assays, all constructs that were judged as positive responses in the F0 assay were confirmed by their response in more than two Tg lines. When it was difficult to judge the response in the F0 assay, we repeated the injection and isolated the F1 carriers to confirm their responses.

We are grateful to Toru Ishitani for his generous gift of the plasmid pTol2(hsp70l:mKO2-2a-dnlef1) and Kohei Ito for teaching us the detailed procedure of heart resection. We thank the Open Facility Centre for Life Science and Technology at the Tokyo Institute of Technology for sequencing plasmid constructs and confocal microscopy services. E.S. was supported by a fellowship from the Education Academy of Computational Life Science of the Tokyo Institute of Technology. T.T. and T.Y. were supported by a Tsubame Scholarship for Doctoral Students from the Tokyo Institute of Technology.