Complete regeneration of lost body parts and organs is commonly seen in invertebrate species such as hydra, planarians, and many of the arthropods (reviewed by Slack,2003). However, during evolution many vertebrate species, including mammals, have retained only a limited capacity for regeneration. In humans, regeneration capacity, particularly the ability of “epimorphic regeneration,” which means the reformation of body parts, is seen in the liver and digit tips (Illingworth,1974; Muller et al.,1999; Han et al.,2005; Carlson,2005; Hata et al.,2007; Yoshizato,2007). In contrast to mammals, urodele amphibians, such as newt and axolotl, and teleost fish have an exceptionally high capability for regeneration as adult animals (reviewed by Akimenko et al.,2003; Nye et al.,2003; Poss et al.,2003). In particular, teleost fish can regenerate complex tissues including many internal organs and spinal cord in addition to their scales and fins. Once the molecular and cellular bases of such regeneration are elucidated, we will be able to apply this knowledge to improve the regenerative capacity of humans.
Studies on regeneration in fish and urodeles have been pursued over a century, but the molecular mechanisms have not been extensively studied, despite the long history of research in this area. Only recently, advances in molecular genetics, genomic resources and the understanding of signal transduction pathways have greatly accelerated the molecular analysis, particularly in the case of the zebrafish. As compared with the 1–2 months required for regeneration of the urodele limb, regeneration of the fins of small fish species such as zebrafish and medaka occurs within only 1–2 weeks. From the study of these animals, an increasing body of cellular and molecular data has been accumulated over the past decade (reviewed by Nakatani et al.,2007).
As is the case for the urodele limbs, fish fins regenerate according to the following processes: wound healing, formation of wound epidermis and blastema, tissue outgrowth, and morphogenesis. Initially, the wound healing occurs by the rapid closure of the wound and epidermal cell migration, which is independent of cell proliferation (Poleo et al.,2001). After the wound healing, a thick epidermal cell layer, termed as the wound epidermis or the apical ectodermal cap (Becerra et al.,1996), is formed to cover the tissues exposed by the amputation. The wound epidermis is thought to play an essential role in the progression of regeneration and pattern formation (Stocum and Dearlove,1972). Recent studies have begun to reveal the molecules involved in the process. It has been shown that β-catenin, lef1, and several wnt ligands are induced in the basal wound epidermis and that Wnt-signaling is required for regeneration (Poss et al.,2000a; Stoick-Cooper et al.,2007). Moreover, the basal wound epidermis also expresses bmp2b, shh, patched1 (Laforest et al.,1998), homeobox transcription factors msxA and msxD (Akimenko et al.,1995), suggesting that Bmp and Hedgehog signaling also have functions during regeneration.
After the formation of the wound epidermis, a population of stem-like cells, the blastema, appears underneath the wound epidermis. These blastema cells are characterized by the expression of msx genes (msxB and msxC), transcriptional repressors that presumably maintain the dedifferentiated state of blastema cells (Akimenko et al.,1995; Odelberg et al.,2000). It has been shown that the induction of blastema cells requires Fgf signaling (Poss et al.,2000b). In addition to msx genes, several other molecules, such as fgf20, bmp4, hoxA11B, hoxA13B, retinoic acid receptor gamma, chemokine sdf-1, and so on are expressed in the blastema during fin regeneration in the zebrafish (White et al.,1994; Murciano et al.,2002; Geraudie and Birraux,2003; Whitehead et al.,2005; Dufourcq and Vriz,2006). Actual involvement of these molecules in regeneration has been suggested based on the results of several studies using pharmacological inhibitors (Poss et al.,2000b; Quint et al.,2002; Bai et al.,2005), transgenic fish lines that express antagonistic molecules against Wnt and Fgf signaling (Lee et al.,2005; Stoick-Cooper et al.,2007), and genetic approaches using temperature-sensitive mutations (Johnson and Weston,1995; Poss et al.,2002; Makino et al.,2005; Whitehead et al.,2005).
On another front, efforts have been made to identify additional molecules involved in the regeneration process. In zebrafish, differential display screening (Padhi et al.,2004) and microarray analysis of regenerating caudal fins have been performed (Schebesta et al.,2006), and databases including the respective 298 and 829 transcripts of differentially expressed transcripts were constructed. Comparison of these databases showed that 24 of the detected genes were common to both databases (Schebesta et al.,2006). Additional microarray analyses and comparison of them may help us to identify genes that directly and strongly respond to regeneration. Previously, we also reported the expressed sequence tag (EST) and microarray analyses during regeneration in medaka (Katogi et al.,2004), a species distantly diverged from zebrafish (Wittbrodt et al.,2002). Although the report contained overall EST analysis and in situ expression data on several transcripts, a comprehensive analysis of microarray data was not performed.
Here, we describe an analysis of medaka microarray data collected from six independent analyses, in which combinations of four different stages of regenerating and uncut fins were compared. We focused on clones that were reliably and reproducibly induced during regeneration and identified 140 up-regulated transcripts. To verify the efficacy of data set extraction, we performed in situ hybridization analysis and observed that the expression of 12 of 22 transcripts was detected either in the differentiating cartilage, blastema, or basal layer of the wound epidermis. Taken together, our results provide a useful additional database and suggest that the identified genes could be useful molecular markers for dissecting the regeneration process at a fine cellular resolution.
RESULTS AND DISCUSSION
Analysis of Microarray Data and Evaluation
The enriched custom microarray, which contains ∼3,000 EST clones from the cDNA libraries at 3 days postamputation (dpa) and 10 dpa of fin regeneration, was used in this study (Katogi et al.,2004). We collected the expression data from six independent microarray analyses, in which the expression was compared between four different time points of caudal fin regeneration and uncut fins: 3 dpa vs. 1 dpa (× 2 microarrays), 3 dpa vs. 2 dpa, 3 dpa vs. 10 dpa, 1 dpa vs. 10 dpa, and 3 dpa vs. uncut fins. From the whole data set, we tried to extract reproducible expression data. Among the transcripts that displayed over twofold induction in each comparison, we considered the transcripts that showed a low basal expression level in uncut fins and high expression levels during regeneration (also see the Experimental Procedures section). Generally, the microarray analysis may not provide strictly quantitative data, as was also suggested by Schebesta et al. (2006). To further extract the reproducible data set, we picked out the transcripts whose induction was repeatedly observed. Thus, we identified the transcripts of 140 genes that were reliably and reproducibly induced during fish-fin regeneration (Supplementary Table S1, which can be viewed at http://www.interscience.wiley.com/jpages/1058-8388/suppmat).
Next, we made a functional assignment of the respective transcripts. Because many EST sequences are only the 5′- or 3′-untranslated regions or regions nonconserved during evolution, we obtained longer sequence information by gene prediction from the medaka genome DNA (http://medaka.utgenome.org/) and/or by searching for overlapping EST sequences. Based on these longer or full-length amino acid sequences, we searched the protein homology and assigned their possible functions. The resultant proteins were classified into the following nine functional categories: transcription, signaling, metabolism and transport, immune response, extracellular matrix (ECM), cytoskeleton, cell cycle, stress response, and unknown functions.
The most abundant group encoded proteins having no sequence homology with other proteins (Unknown 33% in Fig. 1). This was due to the inevitable presence of the large number of transcripts that have diverged sequences in fish species, and also partly to the selection criterion for microarray clones that was biased to unknown genes (Katogi et al.,2004). However, the ratio of unknown genes was comparable to that of a preceding report in zebrafish by Schebesta et al. (2006). The next most abundant clones were the extracellular matrices, many of which were collagens. This group was followed by the proteins involved in metabolism and transport, which was followed by the signaling proteins and transcription factors (Fig. 1). The relative percentages of respective functional groups were comparable to those of preceding studies on zebrafish (Padhi et al.,2004; Schebesta et al.,2006). Significantly, our results from medaka included Catenin beta-1, Fgfr1, MsxC, and ApoEb (Supplementary Table S1), all of which are molecules that have been implicated in regeneration of the zebrafish fin (Akimenko et al.,1995; Monnot et al.,1999; Poss et al.,2000b). Furthermore, the adequacy of our data extraction is also supported by the fact that robust and localized in situ expression was detected for more than 50% of the clones examined (12 transcripts of 22 chosen from Supplementary Table S1; details in later sections), but no localized signal for four clones with lower induction ratios or nonreproducible clones (Table 1). Hence, we conclude that we could successfully extract reliable and reproducible clones and that a significant proportion of these were genes that may directly respond to regeneration cues.
Table 1. Activation Ratios of Genes Known to Be Involved in Fin Regenerationa
Ratios of expression values at respective days (d) after amputation. EST, expressed sequence tag.
Genes Up-Regulated During Cartilage Differentiation
To see if these transcripts were expressed in localized regions or cells during regeneration, we chose 22 transcripts (Table 1) and analyzed their in situ expression patterns. These clones were selected at random, excluding enzymes and structural proteins such as collagens. Of these, we detected apparent localized expression for 12 transcripts in regenerating tissues. For most of them, there are homologous genes in higher vertebrates; however, for 3 transcripts we could not find homologs in animals other than fish; therefore, we designated these as rug (regenerationup-regulatedgene). According to the regions of expression, the 12 transcripts were classified into either of two groups: one group was expressed in the region proximal to the blastema (Fig. 2); and the other, in the blastema and/or wound epidermis (Fig. 3).
By their expression in the region proximal to the blastema, the first group seems to be involved in cartilage differentiation. Indeed, this group contained osterix (sp7), a transcription factor that has a role in osteoblast differentiation (Nakashima et al.,2002). In addition to osterix, rug1, prohibitin, and ENTPD5 were also expressed in the region adjacent to the blastema at 3 dpa (asterisks, Fig. 2A–C). These genes continued to be expressed in possible osteoblast cells adjacent to the blastema at 5 dpa (distal region of regenerating fins; Fig. 2F–J) and at 10 dpa (data not shown). It has been suggested that Prohibitin has a role in the activation of the Ras-Raf signalling pathway and in modulating cell adhesion and migration (Rajalingam et al.,2005) and that ENTPD5 is involved in neoplastic development during carcinogenesis (Paez et al.,2001). Their roles in differentiating osteoblasts would be intriguing.
Three transcripts, prohibitin, ENTPD5, and biglycan, had a strong expression domain in the proximal region of the differentiating cartilage (Fig. 2B,C,E,G,H,J). Biglycan is a small leucine-rich proteoglycan (SLRP) abundant in the extracellular matrix of skeletal tissues and involved in the growth and differentiation of osteoblast precursor cells and regulation of collagen fibril formation (Hocking et al.,1998). The expression of prohibitin, ENTPD5, and biglycan in the proximal region suggests their involvement in cartilage differentiation.
We further investigated the expression of these genes during embryonic and larval development. Consistent with their roles in cartilage differentiation, the expression of osterix, prohibitin, ENTPD5, and rug1, but not that of biglycan, was observed in the developing cartilage of the branchial arches at stage 39, a stage immediately after hatching. The expression of the cartilage genes, except for that of biglycan, was also detected in the basal region of median fin fold (arrows; Fig. 2O–R). Although we did not detect biglycan expression in other stages of development, biglycan might be expressed at some later stages of cartilage development.
Genes Up-Regulated in the Blastema and Basal Wound Epidermis
The second group of transcripts was observed in either the blastema and/or wound epidermis (Fig. 3). According to the cells expressed, these transcripts were further classified into two subgroups: the “basal wound epidermis” genes including thrombospondin 2 (tsp2), signal peptide peptidase-like 2 (sppl2), and novel transcript rug2 (Fig. 3D–F), and the “blastema” genes including beta ig-h3, neurocan, angiopoietin-like protein 2 (angptl2), and novel transcript rug3 (Fig. 3N–Q). Members of the first subgroup were mainly expressed in the basal wound epidermis, although a weak expression was also seen in the peripheral region of blastema, as such an expression has been observed for lef1 (Poss et al.,2000a) and wnt5a and wnt5b (Stoick-Cooper et al.,2007). Studies have suggested that the extracellular matrix protein Tsp2 modulates cell–matrix interactions and may be involved in the wound healing process in vertebrates including mammals (reviewed by Bornstein et al.,2004). Relatively little is known about the function of Sppl2, although Spp is an unusual aspartyl protease that mediates clearance of signal peptides by proteolysis within the endoplasmic reticulum (ER). In zebrafish, the reduction of Sppl2b by antisense gripNA-mediated knockdown resulted in erythrocyte accumulation in an enlarged caudal vein (Krawitz et al.,2005). Like the function of Presenilin, another member of the signal peptidase family, in Notch signaling (reviewed by Parks and Curtis,2007), Sppl2 may have a role in activating a signal during cellular differentiation. All three transcripts of the first subgroup continued to be expressed in the epidermal region at 10 dpa (Fig. 3G–I).
In contrast to the basal wound epidermis genes, beta ig-h3 (reviewed by Litvin et al.,2005), rug3, neurocan (reviewed by Rauch et al.,2001), and angptl2 (reviewed by Morisada et al.,2006) were exclusively expressed in the blastema (Fig. 3N–Q). Whereas the expression of beta ig-h3 was detected throughout the blastema region, the expression of rug3, angptl2, and neurocan seemed to be localized in the peripheral or distal part of the blastema. As Nechiporuk and Keating (2002) suggested that the blastema consists of distal and proximal parts, such a differential gene expression within the blastema suggests that the blastema might contain several different cell populations in addition to the distal and proximal populations. Among these four blastema genes, only the beta ig-h3 expression was retained at 10 dpa (Fig. 3R–U), suggesting that the blastema cell subpopulations expressing rug3, neurocan, and angptl2 may exist during the early phase of regeneration. The expression of angptl2, which encodes a protein that induces angiogenesis, is also intriguing; because it has been suggested that angiogenesis may have a role during limb regeneration in the newt and axolotl (Smith and Wolpert,1975).
We next looked at the expression of these genes during embryonic and larval development (Fig. 4). Expression of the basal wound epidermis genes was detected in the developing median fin fold at stage 28 (Fig. 4A,D,G), and at stage 39 the expression extended to the whole median fin fold and the peripheral pectoral fins (Fig. 4B,C,E,F,H,I). The peripheral region of pectoral fins is thought to correspond to the apical ectodermal ridge in limb development and to function to mediate a morphogenetic signal to the underlying mesoderm (Capdevila and Izpisua Belmonte,2001). In contrast to the basal wound epidermis genes, we could not detect any localized expression of blastema genes in the developing fins (Fig. 4J–L). Neurocan expression was only seen in tissues of the nervous system (Fig. 4J), β-igh3 was expressed in the somites and ubiquitous cells in the embryo (Fig. 4K,L), and no rug3 was detected at stages 28 and 39. Such expressions in tissues other than fins suggest that they may have roles in the regeneration process itself, rather than in the fin formation.
Comparison of the Transcriptional Profiling in Medaka and Zebrafish
Thus, by transcriptional profiling we identified several molecules expressed during medaka fin regeneration. We next compared our results with the previous ones on zebrafish (Padhi et al.,2004; Schebesta et al.,2006). Among the 298 genes reported by Padhi et al. (2004), periostin and CCAAT/enhancer binding protein were detected in our array screening in addition to several collagens (Supplementary Table S1). They also identified the up-regulation of biglycan-like protein 3, tsp4, mmp2, which may have similar and/or overlapping functions with biglycan, tsp2, and mmp9, respectively, found in our medaka screening. The number of overlap was 13, including closely related family members. Importantly, they also identified two molecules: rug2 (reported as 2-F11) and rug3 (reported as 2-H06). The study by Schebesta et al. (2006) detected 563 up-regulated zebrafish genes by their microarray analysis. Among them were catenin beta-1, neurocan, mmp9, and rug3 (reported as a hypothetical protein, LOC566702) in addition to the tsp and mmp families of transcripts, that is, tsp3, mmp2, mmp13a, mmp13b, and mmp14. The number of overlap was 18, including closely related family members, and the total number of zebrafish genes that overlaps with the medaka analysis was 26 among 140 up-regulated genes. Interestingly, they also reported an angiopoietin-like molecule, angptl7; whereas we detected the expression of angptl2 in the medaka fish (Supplementary Table S1). Kubota et al. (2005) also showed that angptl2 expression was induced during zebrafish fin regeneration. In medaka, we observed the induction of angptl7 expression (Nakatani et al., manuscript in preparation) in addition to angptl2 in the blastema. Both of angptl7 and angptl2 may have a function for angiogenesis during regeneration.
In our microarray analysis we identified three uncharacterized proteins, that is, Rug1, Rug2, and Rug3. As the protein sequences of Rug1–3 had putative signal sequences at the N-terminus (Fig. 5), these proteins seem to be secreted proteins. A Blast search of Rug1 detected a low sequence homology with a novel zebrafish protein (Fig. 5A), and similarly Rug2 and Rug3 had zebrafish homologs. However, these proteins did not have apparent homologs in higher vertebrates, including mammals. It is possible that the homologs in nonfish species might have highly diverged protein sequences or that they could have been lost during evolution. Surprisingly, the amino acid sequences of Rug2 and Rug3 shared a high sequence homology with each other (43.3%), and it is interesting to note that all of the cysteine residues were perfectly conserved between Rug2 and Rug3 in medaka fish and zebrafish (Fig. 5B). Pairs of cysteine residues were regularly positioned with seven amino acid residues intervening between each two cysteines, but this pattern does not fit with any known protein motifs described so far. Functions of Rug2 and Rug3 are intriguing in view of their different expression during regeneration and development (Figs. 3F,O, 4G–I).
Comparing our data with those on zebrafish, we conclude that the screening for differentially expressed genes during regeneration in zebrafish and medaka fish have revealed several regeneration-related molecules. Using these molecules as markers, we will be able to dissect the regeneration process at a fine cellular resolution. At the same time, it will be intriguing to uncover the functions of respective proteins during regeneration as well.
Fish Maintenance and Fin Amputation
A medaka (Oryzias latipes) wild-type strain, the cab strain, was maintained in a re-circulating system with a 14 hr/day and 10 hr/night cycle at 28.5°C, and used for fin amputation. Fish were anesthetized in water containing the tricaine (Sigma), and the caudal fins were amputated with a sharp scalpel. These fish were returned to the aquarium and allowed to regenerate their fins for 1–10 days. The fins were collected at appropriate time points for further analysis. Embryos and larvae were staged according to the medaka developmental stages described by Iwamatsu (2004).
Microarray Data Analysis
The custom microarray, which included 2,900 random ESTs from libraries at 3 dpa and 10 dpa, was used in this study (Katogi et al.,2004). These EST clones have been clustered; however, it appeared that they still contain some extent of redundant clones. Using this enriched microarray, the expression data were collected from six independent analyses, in which the expression comparisons were performed: 3 dpa vs. 1 dpa (× 2 times), 3 dpa vs. 2 dpa, 3 dpa vs. 10 dpa, 1 dpa vs. 10 dpa, and 3 dpa vs. uncut, respectively. We adopted the following criteria for extracting a reliable and reproducible data set of up-regulated genes: (1) transcripts with greater than twofold up-regulation, (2) transcripts with lower basal expression (absolute expression values lower than 1,000 in uncut fins), (3) transcripts with enough expression during regeneration (absolute expression values higher than 1,000 at 1 dpa, and higher than 10,000 at 2 and 3 dpa), and (4) transcripts that repeatedly fulfilled the above criteria (more than two times).
According to these criteria, we could extract 140 transcripts. We defined these as the genes reliably and reproducibly up-regulated during regeneration (Supplementary Table S1).
For the functional assignment of the respective transcripts, the gene prediction (GenScan: http://genes.mit.edu/GENSCAN.html) from the genomic sequences (medaka genome database: http://medaka.utgenome.org/; http://www.ensembl.org/Oryzias_latipes/index.html) and/or the search for the EST sequences was performed, and the deduced or full-length protein sequences were used to search for the homologous proteins. The categories and functions of gene products (Supplementary Table S1) were based on the descriptions in the NCBI database of protein sequences. The EST sequences, except those reported by Katogi et al. (2004), were deposited in the GenBank (10,368 ESTs from 3 dpa and 13,056 ESTs from 10 dpa).
Whole-Mount In Situ Hybridization Analysis and Histology
Among the resulting 140 transcripts, we chose 22 genes for the in situ expression analysis. Digoxigenin-labeled RNA probes were synthesized from the respective EST clones. Whole-mount in situ hybridization was performed according to the method described earlier (Schier et al.,1997). After staining, the fins were fixed with 4% paraformaldehyde in phosphate-buffered saline for color preservation. Samples were equilibrated with 80% glycerol and mounted on slide glasses for taking photographs. For histological analysis, stained fins were embedded in Tissue-Tek compound (Miles) and cross-sectioned at 10–20 μm by use of a cryostat. 2
Table 2. Transcripts Whose in situ Expression Were Analysed in This Studya
in situ No.
Expression During Regeneration
Embryonic and Larval Expression
Underlined EST names are the early induced genes (1dpa signal > 9,000).