Adult mammalian heart tissue does not regenerate after injury, which is relevant to human heart disease, ischemic myocardial infarction, and scar formation (Ma et al., 2014). By contrast, zebrafish have the ability for scarless heart regeneration (Poss et al., 2002; Raya et al., 2003), and this regenerative ability relies mainly on the proliferation of existing cardiomyocytes (Jopling et al., 2010; Kikuchi et al., 2010; Gemberling et al., 2013). Along with an infiltration of the proliferating cardiomyocytes, the wound is highly vascularized with capillaries of coronary vessels (Lepilina et al., 2006; Kim et al., 2010). Regarding the molecular events in zebrafish heart regeneration, injury to the heart initiates an organ-wide reaction detectable as the induced expression of raldh2 in the endocardium as early as 3 hr post-injury. This expression in the endocardium adjacent to regenerating cardiomyocytes remains active in the area of injury for several days. Because raldh2 is not expressed in endocardium of infarcted hearts in mice, endocardial raldh2 expression in zebrafish is suggested to be the essential response for heart regeneration (Kikuchi et al., 2011b).
Recently, in zebrafish heart regeneration, the importance of extracellular matrix (ECM) molecules such as fibronectin and tenascin-C has been pointed out; these molecules are induced by TGF-β (Mercer et al., 2013; Wang et al., 2013). In the case of healing in the mammalian heart, another ECM molecule, periostin, newly appears and functions to induce the migration of cardiac fibroblasts and fibrillogenesis for scar formation together with fibronectin and tenascin-C (Horiuchi et al., 1999; Kii et al., 2010; Kudo, 2011). In humans, periostin is abundantly expressed in the infarct border after a myocardial infarction; and in mice, its expression is induced by TGF-β following inflammation at the infarct border (Shimazaki et al., 2008). Periostin functions in the migration of cardiac fibroblasts through engaging integrin αvβ3, which induces phosphorylation of the downstream kinase FAK (focal adhesion kinase) in these cells, leading to the production of type I collagen. Thereafter, periostin acts to promote collagen cross-linking in the ECM. In the later chronic stage of a myocardial infarction, fibrillogenesis is accelerated by periostin and proceeds to generate a tight scar, in a similar fashion as seen in cardiac hypertrophy (Oka et al., 2007).
In spite of the growing interest in using fish models for probing heart regeneration, the difference in cardiac regenerative ability among fish species is unknown. Medaka is another representative fish animal model in addition to the zebrafish; however, no reports on heart regeneration in medaka have been published, though the medaka fin does regenerate (Katogi et al., 2004).
Here, we examined the ability of medaka to regenerate heart tissue after ventricular resection. Unexpectedly, we observed persistence of collagen accumulation, mild proliferation of cardiomyocytes, lack of vascularization, and molecular marker patterns different from those of the zebrafish.
Results and Discussion
Medaka Heart Repair after Ventricular Resection
To examine the injury response of the medaka (Oryzias Latipes) heart, we firstly established a method of operation to injure the medaka ventriculum. We adopted a ventricular apex resection model for injury of the myocardium. After the operation, the medaka appeared to be less active and more died than in the case of zebrafish. Therefore, we reduced the ratio of the excision area to total ventricular area so that a greater number of medaka would survive. The hearts collected from surviving medaka were observed at different time points by acid fuchsin orange G (AFOG) staining, to which collagen is sensitive (Fig. 1F–L, magnified views in bottom row). We also observed zebrafish (Danio Rerio) samples as a technical control (Fig. 1A–E, magnified views in bottom row). The ratio of excision area to total ventricular area was reduced to the same as for the medaka.
In the zebrafish, fibrin formation and collagen accumulation were observed at the wound site from 1 week to 2 weeks post-amputation (Fig. 1B and C). The wound was filled with regenerated tissue, and little collagen remained by 30 days post-amputation (dpa, Fig. 1D). At 60 dpa, zebrafish heart regeneration was nearly complete, and only faint staining of collagen remained (Fig. 1E).
In medaka, at 1 dpa, the injury area was filled with a blood clot (Fig. 1G), and then the clot was replaced with fibrin by 4 dpa (Fig. 1H). From 1 week to 2 weeks' post-amputation, collagen accumulation was observed in the injured area (Fig. 1I and J). At 4 dpa, α-smooth muscle actin-positive putative myofibroblasts emerged in the wound area (Fig. 1K). Thus, the medaka heart showed a response similar to that of the zebrafish at these early time points. However, the wound in medaka heart was occupied by dense collagen by 30 dpa (Fig. 1L). Furthermore, in medaka, the collagen scar remained at 60 dpa (Fig. 1M). At 14, 30, and 60 dpa, the collagen area gradually expanded rather than being reduced by absorption (Fig. 1N).
To investigate the cardiomyocyte contribution to the wound in medaka, we performed co-immunostaining for cardiomyocytes and collagen at 30 dpa. Consistent with the AFOG staining results, collagen was detected; but no cardiomyocytes were observed in the injured area (Fig. 1O).
These results reveal that the medaka heart responded to an injury in a different manner than the zebrafish heart. The differential responses to injury in medaka and zebrafish may be explained by two possibilities. One is that an excessive fibrotic response occurred in the medaka heart. This fibrotic response may be attributed to an excessive proliferation of and collagen production by fibroblastic cells such as myofibroblasts. Alternatively, matrix-degrading proteinases such as matrix metalloproteinases (MMPs), which are highly expressed in regenerating heart in zebrafish or newts (Lien et al., 2006; Mercer et al., 2013), may not have been sufficiently activated in the injured medaka heart. A reduced collagen deposition is reported to facilitate myocyte progenitor engraftment to an infarcted heart (Dai et al., 2011); therefore, in medaka it is likely that the excessive collagen deposition disturbed the infiltration of cardiomyocytes into the wound and that repression of collagen deposition would have enhanced the infiltration by cardiomyocytes. The other possibility is that existing cardiomyocytes or cardiac progenitor cells in medaka remained dormant or that perhaps the activation and proliferation of these cells was limited to a smaller population than in zebrafish, leading to the domination of fibrotic repair.
Cardiomyocyte Proliferation in the Medaka Heart
To examine the cardiomyocyte proliferation in the medaka heart after the injury, we investigated DNA synthesis in medaka cardiomyocytes by use of EdU incorporation.
We focused our examination of the proliferation on the time period of 1–2 weeks post-amputation, as this is the period of the peak proliferation of zebrafish cardiomyocytes following cardiac injury (Poss et al., 2002). EdU was successfully incorporated into cardiac cells (Fig. 2A), and EdU+/ MF20+ (MF20 is an antibody that marks cardiomyocytes) putative proliferating cardiomyocytes were observed (Fig. 2B). However, there was no significant detectable difference between the numbers of double-positive cells (Fig. 2C, 13 dpa). Taking the small injury size in this study into account, it is possible that the peak of proliferation would come earlier than 1 week post amputation. Therefore, we also quantified the proliferation at an earlier time span; however, no significant increase in cardiomyocyte proliferation could be detected at the end of the first week after amputation (Fig. 2C, 7 dpa).
Since EdU-incorporating cardiomyocytes were seen in the uncut heart, it is possible that renewal of cardiomyocytes may occur in it. However, there was no numerical difference between the uncut and injured hearts with respect to EdU-incorporating cardiomyocytes, suggesting that the proliferation of medaka cardiomyocytes/progenitor cells remained at the base level or that their time of peak proliferation might be later than that of the zebrafish.
Expression Analysis of fli1 After Ventricular Resection
Neovascularization seems to be essential for zebrafish heart regeneration (Lepilina et al., 2006; Poss, 2007). To examine vascular dynamics after injury to the medaka heart, we investigated fli1-expressing cells by using a medaka fli1-GFP transgenic line. Fli1 is the gene encoding the ets family transcription factor Friend leukemia integration 1, a known marker for the vascular endothelium and endocardium (Brown et al., 2000; Cha and Weinstein, 2007). We focused on 2 weeks post-amputation, the time at which a highly vascularized wound is seen during zebrafish heart regeneration (Lepilina et al., 2006; Kim et al., 2010).
The endocardial and vascular endothelial cells in the heart of the fli1-GFP transgenic line showed the green fluorescent signals from anti-GFP antibody staining (Fig. 3A–D). GFP+ cells were not observed in the 4 and 7 dpa wound area (Fig. 3B, C, wo). Furthermore, no vascular cells were observed in the 14-dpa wound area, either (Fig. 3D).
In heart development, epicardial-mesenchymal transition (EMT) of the epicardium is important for formation of the coronary vasculature (Olivey et al., 2004; Gittenberger-de Groot et al., 2012). Neovascularization after an injury is also likely to be mediated by epicardium-derived cells through the EMT under the regulation of a growth factor such as FGF or PDGF in zebrafish (Lepilina et al., 2006; Kim et al., 2010; Kikuchi et al., 2011a). Our results may imply that the epicardium-derived cells failed to penetrate the wound and to form new vessel components in medaka. Furthermore, medaka does not have distinct coronary vessel structures in their heart, unlike the zebrafish (Lemanski et al., 1975; Hu et al., 2001). Therefore, the medaka heart may not have the program to vascularize a heart injury de novo by epicardium-derived or other cell populations.
Raldh2 Analysis After Ventricular Resection
To examine the dynamics of the medaka endocardium after an injury, we investigated the localization of raldh2 in zebrafish and medaka by immunostaining with an anti-raldh2 antibody. Raldh2 is an enzyme responsible for synthesis of the morphogenetic factor retinoic acid (RA), which is essential for normal cardiac development and regeneration in the zebrafish heart (Hoover et al., 2008; Kikuchi et al., 2011b)
Because the anti-raldh2 antibody was not active for immunostaining in the medaka heart, we used the medaka raldh2-GFP transgenic line instead.
In zebrafish, raldh2 protein was observed in the epicardium but not detected in the endocardium in the uncut heart (Fig. 4A). The endocardial raldh2 was detected at 3 hr post-amputation (hpa, Fig. 4B), as previously reported (Kikuchi et al., 2011b), and remained detectable at 2 weeks' post-amputation (Fig. 4C–D, arrowhead).
In the medaka raldh2-GFP transgenic line, the epicardium and valves gave the green fluorescence of GFP, which was not seen in the endocardium, in the uncut heart (Fig. 4E). The GFP expression was not observed in the endocardium of the cut heart at 3 hpa (Fig. 4F). Furthermore, from 4 to 14 dpa, although a few fibroblast-like GFP-positive cells were detected within the myocardial tissues in medaka (Fig. 4G–H, arrowheads), there were no endocardial GFP-positive cells.
The expression pattern obtained from the analysis of the medaka raldh2-GFP transgenic line resembled that of the raldh2 gene in the infarcted mouse heart, which lacks endocardial expression (Kikuchi et al., 2011b). Therefore, our results may imply that the medaka endocardium remained in an inactive state. Moreover, given that retinoic acid signaling is reported to be related to the proliferation of cardiomyocytes during zebrafish heart regeneration (Kikuchi et al., 2011b), the dormant state of the endocardium in the medaka may imply the inability of cardiomyocyte proliferation because of the absence of retinoic acid signaling after injury.
Localization of Periostin After Ventricular Resection
To examine the dynamics of ECM molecules after injury in zebrafish and medaka, we focused on the localization of periostin protein by immunostaining with anti-periostin antibody. For antibody production, we cloned two types of medaka periostin cDNAs and sequenced them (Fig. 5B). We named them periostin-a (AB931173) and periostin-b (AB931174), and zebrafish periostin-b was found to be relatively highly conserved with medaka periostin-b (Fig. 5C).
Zebrafish periostin-a localization was observed at the atrio-ventricular valves and ventriculo-bulbal valves in the uncut heart (Fig. 6A). Although from 7 to 30 dpa, this periostin type was not observed at the injured area consistently (Fig. 6B–E), zebrafish periostin-b was observed in the epicardium (Fig. 6F, arrowhead). In addition, it was detected within the blastema at 7 dpa (Fig. 6G); and furthermore, endocardial localization was also seen from 14 to 21 dpa (Fig. 6H–I, arrowhead). By 30 dpa, zebrafish periostin-b localization was limited to the epicardium and not seen within the wound (Fig. 6J).
Medaka periostin-a was detected in the epicardium and valves of the uncut heart (Fig. 6K, arrowhead). By resection of the ventricular apex, epicardial localization of periostin-a was removed in the injured area (Fig. 6L). However, the epicardial localization revived gradually (Fig. 6M, arrowhead), and covered the wound from 21 to 30 dpa (Fig. 6N,O, arrowhead).
On the other hand, medaka periostin-b was localized at the atrium (Fig. 6P, arrowhead) and valves (data not shown) in the uncut heart. After resection, this periostin was detected in the border area of the wound (Fig. 6Q, arrowhead). Thereafter, the localization was seen in the wound (Fig. 6R,S, arrowhead), and remained within the wound at 30 dpa (Fig. 6T, arrowhead).
The zebrafish periostin-b and the medaka periostin-b were localized at the wound site, in which the collagen deposition occurred. This coincidence of localization implies that zebrafish periostin-b and medaka periostin-b may similarly function in fibrillogenesis, which was previously reported as one of the functions of periostin (Kudo, 2011). Together with fibronectin and tenascin-C, periostin behaves like a scaffold protein to generate the ECM structure (Kudo, 2011). If that is the case, it is possible that repair of the medaka heart was prolonged through fibrillogenesis promoted by this ECM structure, because medaka periostin-b remained localized at the wound area at 30 dpa, whereas zebrafish periostin-b was no longer localized there but was detected in the epicardium at 30 dpa.
In the light of the data taken together, compared with zebrafish heart regeneration, the data on the medaka heart suggest the possibility that the medaka could hardly regenerate heart tissues or that these tissue phenotypes for heart regeneration showed a delay. Considering the transient cardiac regenerative capacity of neonatal mice (Porrello et al., 2011), the medaka may also have a potential for heart regeneration in their early life. Therefore, a further study of the injury response at an earlier time should clarify the regenerative potential of the medaka heart. Crucial factors and events are left to be investigated regarding the medaka heart, and further studies to find other markers should answer these questions. Our study may contribute to a better understanding of the specificity of the regenerative ability among fish species and to an unraveling of the mechanisms involved in regeneration of the fish heart.
Fish Strains, Maintenance, and Surgery
The Cab medaka strain (Loosli et al., 2000) and the Ekkwill zebrafish strain (EkkWill Waterlife Resources) were used for this study. The latter was kindly donated by Dr. Kazu Kikuchi (Victor Chang Cardiac Research Institute). Adult fish were maintained at 28°C in a re-circulating system. Fish at >3 months of age were used in all experiments. Although ventricular apex amputation for both species was performed as previously described (Poss et al., 2002), the ratio of excision area to total ventricular area was reduced for medaka survival. For medaka surgery, a sufficient supply of oxygen was needed for revival from the anesthesia. Experiments were carried out in accordance with the animal use guidelines at Tokyo Institute of Technology.
For optimization of staining conditions, the cryosection method was used with 14-µm sections for AFOG staining, fli1 analyses, raldh2 analyses, and periostin immunostaining of zebrafish; whereas for the medaka samples, the paraffin section method was used with 4-µm sections for the EdU proliferation assay and 14-µm ones for immunostaining. The AFOG staining was conducted as described previously (Poss et al., 2002). Bright-field images were collected by an upright microscope (Carl Zeiss, Thornwood, NY; Axioplan2 imaging). Quantification of collagen-positive or ventricular area was performed by using Image J software. The ventricular area contained both ventricular myocardium and lumen.
For immunohistochemistry, the following primary antibodies were used: MF20 at 1:100 (Developmental Studies Hybridoma Bank [DSHB, Indianapolis, IN]), rabbit anti-collagen1 at 1:100 (AbD Serotec, 2150-1410), mouse anti-smooth muscle actin at 1:100 (Dako, Carpinteria, CA; M0851), rabbit anti-GFP at 1:500 (MBL, No. 598), rabbit anti-raldh2 at 1:500 (GeneTex, Irvine, CA; GTX124302), rabbit anti-zebrafish periostin-a at 1:100, rabbit anti-zebrafish periostin-b at 1:500, rabbit anti-medaka periostin-a at 1:100, and rabbit anti-medaka periostin-b at 1:100. The secondary antibodies used were goat anti-rabbit AlexaFluor 488 and goat anti-rabbit AlexaFluor 568, both at 1:500 (Molecular Probes, Eugene, OR).
The sections were reacted with primary antibody overnight at 4°C after preventing non-specific binding with Blocking One reagent (Nacalai Tesque), and then incubated with secondary antibodies 30 min at room temperature. For immunostaining with MF20 antibody, the antigen unmasking technique with citrate buffer was performed. The fluorescent images were collected by a laser-scanning confocal microscope (Olympus, Center Valley, PA; FV1000).
EdU Incorporation and Quantification
EdU treatment and staining were performed by using a Click-iT® EdU Imaging Kit (Molecular Probes). EdU at 25 µM in breeding water was incorporated into adult medaka over a 7-day period. The staining procedure conformed to the manufacturer's protocol. The fish were sacrificed after the last day of the treatment. Quantification of EdU+/MF20+ cells was carried out by using Image J software (NIH). For the area quantification, the outline of the ventricle was traced from the differential interference-contrast (DIC) image. As described above, the ventricular area contained both the ventricular myocardium and lumen. By utilizing the scale information, the area was determined. The mean cell counts and area quantification were performed by using 3 arbitrarily chosen sections per sample.
Generation of Medaka Transgenic Line
A fosmid clone (GOLWF no 683_j24) including the raldh2 gene, which was identified from the Medaka ensemble database, was used for homologous recombination. The homologous recombination was performed according to a previous report (Nakamura et al., 2008). An EGFP-kanamycin cassette was amplified by using appropriate primers (5′-ATCAACAAAGCCATGACCGTCTCC ACAGCCATGCAGGCCGGCACTGTCTGATGAGCCATATTCAACGG GA-3′ [forward] and 5′-CTGGAAAATACACAGAGCATCGATC ACTCTCGGTCATTAGACTCCTGCTCTTAGAAAAACTCATCGAGCA -3′ [reverse]) and recombined into the exon 1 of the medaka raldh2 gene. The recombined fosmid clone was injected into the cytoplasm of one-cell stage embryos. For establishment of raldh2-GFP transgenic medaka, we used the fertilized eggs from the medaka Cab line.
The fli1-GFP transgenic line was kindly provided by Dr. M. Furutani-Seiki at the University of Bath, UK (Moriyama et al., 2010).
Preparation of Rabbit Polyclonal Anti-Periostin Antibody
Polyclonal rabbit anti-medaka periostin-a/periostin-b antibodies were raised by immunization of rabbits with keyhole limpet hemocyanin (KLH)-conjugated peptides DLLDPNEKLKLAENEN (representing amino acids 143–158 of medaka periostin-a) and ELLDEDVRNALVSNVN (representing amino acids 141–156 of medaka periostin-b). The antibodies were affinity purified with a minicolumn of HiTrap Protein G HP (GE Healthcare Biosciences, Waukesha, WI). Polyclonal rabbit anti-zebrafish periostin-a antibodies were raised by immunization of rabbits with KLH-conjugated peptide EELDPASKAAVISRGN (representing amino acids 143–158 of zebrafish periostin-a) and affinity purified by using a minicolumn of HiTrap NHS-activated HP (GE Healthcare Bio-sciences). The anti-zebrafish periostin polyclonal antibodies previously reported (Kudo et al., 2004) were used as anti-zebrafish periostin-b antibodies.
Cloning of the Medaka Periostin Gene
As a template for cloning PCR, a mixed cDNA synthesized by RNA from whole ventricle was used. RNA extraction was performed by using ISOGEN (NIPPON GENE), and the cDNA synthesis was carried out with a PrimeScript®II 1st strand cDNA synthesis Kit (Takara, Shiga, Japan). PCR products were cloned into pCR®4-TOPO vector and sequenced.
Specific primers designed for medaka periostin cloning were as follow: 5'-GCAGAGTGTGTTAGTGACCTT-3' (forward) and 5'-TACCATGCACTTTGTCTACAG-3' (reverse) for periostin-a; 5'-CTCTGGAACTGGAAAGGCTTG-3' (forward) and 5'-AGCAGGTAGCATGCAGTGAAG-3' (reverse) for periostin-b.
Protein sequence alignment was performed by using GENETYX software ver.10 (GENETYX).
We thank Masanobu Nishidate for initiating the studies on medaka periostin. This work was supported by grants-in-aid for scientific research from the Japan Ministry of Education, Culture, Sports, and Technology of Japan.