The tremendous regeneration ability of urodele amphibians that helps regenerate lost limbs has long fascinated scientists. Once a limb is amputated, undifferentiated cells called blastema cells arise at the amputation site. Although the origin of blastema cells remains unknown, dermal fibroblasts are believed to be one of the major sources (Gardiner et al., 1986; Muneoka et al., 1986; Kragl et al., 2009; Hirata et al., 2010). Dermis-derived blastema cells redifferentiate into various cell types during limb regeneration. Moreover, dermal fibroblasts regulate patterning of the regenerating limb (Tank, 1981; Maden and Mustafa, 1982; Rollman-Dinsmore and Bryant, 1982). The main role of dermal fibroblasts in limb regeneration justifies the study of their behavior and function during limb regeneration.
While the mechanism of blastema induction from mature dermal tissue remains unknown, the apical epithelial cap (AEC) has been suggested to play an important role in the induction (Wallace, 1981). AEC is an epithelial structure located in the distal region of the blastema and is functionally similar to the apical ectodermal ridge (AER), which appears during limb development in higher vertebrates (Saunders, 2002). Therefore, gene expression patterns and functions of AEC have been studied and compared to those of AER. Not only AEC but also blastema mesenchyme has been likened to the limb bud mesenchyme of higher vertebrates. In fact, an induced regenerating blastema has features similar to those of a developing limb bud (Muneoka and Bryant, 1982). Most limb regeneration processes are believed to repeat developmental processes to restore the original structure (Gardiner et al., 2002). Indeed, the later phase of limb regeneration is called the redevelopment phase. Unique and key events occur during the early phase of limb regeneration. In limb development, undifferentiated tissues, such as the lateral plate mesoderm (LPM), are located adjacent to the limb induction site. Once AER is induced, LPM becomes a source of undifferentiated cells for the limb bud (Fernandez-Teran and Ros, 2008). In contrast, in limb regeneration, a mature limb consists of mature tissues and does not have undifferentiated or embryonic tissues such as LPM. Blastema cells emerge from such differentiated tissues, and therefore, their induction appears to be a regeneration specific process as compared to limb development. To elucidate the superior regeneration ability of urodele amphibians, it is necessary to focus on the unique early phase of limb regeneration, including dedifferentiation (reprogramming) of dermal fibroblasts.
The accessory limb model (ALM) was developed by Endo et al. after reviewing classical studies to investigate the early events of limb regeneration (Bodemer, 1959; Lheureux, 1977; Reynolds et al., 1983; Maden and Holder, 1984; Egar, 1988; Endo et al., 2004). A skin wound, deviated nerve, and contralateral skin graft are sufficient to induce accessory limb formation (Fig. 1). Generally, a small square skin wound is created in an anterior upper limb (Fig. 1B). Special attention must be given to the orientation of the axolotl forelimb because it is twisted (Fig. 1A). A relatively larger artery runs along the anterior side (Fig. 1A, green arrow), providing a good landmark for the anterior side of the forelimb. Axons are cut at the elbow/knee level and dissected out from the ventral side (Fig. 1C). The free axon ends are then placed on the anterior wound (Fig. 1D,E); these are sufficient to induce a blastema, though the induced blastema regresses at the end, as mentioned later. To induce regeneration of a perfectly patterned limb, a small piece of skin from the contralateral side (posterior) should be grafted by the deviated nerve (Fig. 1F,G). If the skin graft is not prepared from the exact contralateral region, a limb is not induced or is malformed, usually with fewer digits than normal. A blastema starts forming on top of the deviated nerve and continues to grow as observed during normal limb regeneration (Fig. 1H). Although it remains difficult to obtain a perfectly patterned limb (one possessing all skeletal elements), regeneration of a limb structure can be induced at a high rate (>73%) by means of the skin graft (Fig. 1I,J; Endo et al., 2004). As mentioned above, the skin graft from the contralateral side is not necessary to induce an ectopic blastema. An ectopic blastema can be induced by skin wounding and nerve deviation without skin grafting (Endo et al., 2004; Satoh et al., 2007). Without skin grafting the blastema does not continue growing and, in most cases, it starts regressing (Endo et al., 2004). However, although the induced blastema cannot continue growing, it shows blastema features (Satoh et al., 2007). These blastema cells expressed blastema marker genes and showed cartilage differentiation ability (Satoh et al., 2007). Thus, skin wounding and nerve deviation are sufficient to induce blastema formation, and it is the presence of the nerve that distinguishes wound healing from blastema formation.
How does a nerve control blastema induction? Nerve functions have long been the focus of limb regeneration studies. Neurotrophic factors are believed to be involved in limb regeneration, and several factors have been isolated, including FGFs, GGF, transferrin, substance P, and anterior gradient (AG) (Brockes and Kintner, 1986; Mullen et al., 1996; Wang et al., 2000; Nye et al., 2003; Kumar et al., 2007; Satoh et al., 2008b, 2009a., 2010a, 2011). However, the detailed role of these factors in blastema induction remains unknown. ALM has revealed that inducing AEC is one of the first nerve functions to occur in limb regeneration (Satoh et al., 2008b). The nerve–AEC relationship had already been implicated in some studies (Singer, 1949; Satoh et al., 2008b; Thornton, 1954). ALM appears to be a suitable system for investigating early nerve function and AEC induction owing to the ease with which nerve participation can be surgically manipulated on ALM from the beginning to the end of limb regeneration.
Comparative study of wound healing and blastema formation has been undertaken to determine what distinguishes these two processes. Skin wounding without nerve deviation results in simple wound healing, whereas wounding with nerve deviation results in blastema induction (Endo et al., 2004). Skin wound healing and blastema formation can be compared easily using ALM. Amputated limbs have been used for the same purpose, but such a comparison is difficult for several reasons. First, it is difficult to induce wound healing without blastema formation. Denerved limbs, which cannot grow a blastema, have been used as the wound-healing control, but repeated denervation surgery is required to preserve the denerved state because axolotl axons regrow quickly (Schotte and Butler, 1941; Petrosky et al., 1980). Second, it is difficult to determine whether or not denervation is complete at the time samples are collected. The success of denervation can be determined only several days after surgery. Third, multiple tissues are present in a limb, and amputation damages all of them. Each tissue has its own healing process independent of epimorphic regeneration, leading to higher background in a comparative study. Fourth, limb regeneration induced by amputation is likely to be driven by two distinct regeneration systems (Satoh et al., 2010a). One is AEC-dependent and regenerates distal structure(s), while the other is AEC-independent and regenerates proximal structures (Bryant and Iten, 1977). Amphibian limb regeneration has generally been studied with respect to AEC formation, and in this approach the AEC-independent mechanism adds noise to the analysis. ALM is a superior model for focusing on the AEC-dependent mechanism because this is the only mechanism used in ectopic limb regeneration. Thus, ALM provides an ideal model for a study comparing wound healing with blastema formation and is an advanced model to investigate limb regeneration.
MATRIX METALLOPROTEINASE (MMP) ACTIVATION
The very beginning of limb regeneration shares a common mechanism with wound healing in which MMPs are involved (Gill and Parks, 2008; Parks, 2008; Satoh et al., 2011). MMPs are enzymes that digest the extracellular matrix (ECM) such as collagens. The number of MMP genes in an axolotl remains uncertain, but some MMPs appear to be upregulated from the early phase of limb regeneration (Yang and Bryant, 1994; Miyazaki et al., 1996; Yang et al., 1999; Vinarsky et al., 2005). GM6001, a broad MMP inhibitor, inhibits limb regeneration (Vinarsky et al., 2005); however, the mechanism remains unclear. Immediately after limb amputation, the surrounding epidermis starts migrating to cover the amputation surface. MMPs are expressed in migrating epidermis/epithelial cells (Mullen et al., 1996; Satoh et al., 2008b, 2009b) and MMP-9 is strongly expressed in the basal layer of these cells. Not only the migrating epidermis but also leucocytes appear to show high MMP activity by histological observation (Satoh et al., 2008b). Specific leucocyte types cannot be identified because no axolotl leucocyte antibodies are available. However, from our knowledge of higher vertebrates and of the histology of the regenerating axolotl limb, we expect that at least neutrophils and microphages accumulate around an amputation surface and express MMPs (Fishman and Hay, 1962; Tsonis, 1996; Gill and Parks, 2008). It is reasonable to expect an accumulation of leucocytes after amputation as an anti-inflammatory response. MMP activities increase in the damaged region because of the migrating epidermis and leucocytes and might be nerve-independent. However, it has been suggested that MMP activity is at least partially controlled by nerves during later phases of regeneration (Yang et al., 1999; Satoh et al., 2011). In higher vertebrates, digestion of ECM by MMPs triggers cellular activation, resulting in activation of Prrx1, a blastema cell marker gene expressed during amphibian limb regeneration (Jones et al., 1999, 2001; McKean et al., 2003; Suzuki et al., 2005; Satoh et al., 2007). In case of axolotls, digestion of ECM also activated Prrx1 in skin cultures (Satoh et al., 2011). This may be reasonable because fibroblasts adhere to ECM in a normal (unamputated) limb, but they must be free from ECM to migrate toward a damaged region during limb regeneration. Fibroblasts, which are free from ECM, may be called preblastema cells. Suzuki et al. (2005) originally proposed fibroblast activation before blastema cell induction. Induction of fibroblast activation occurs through the integrin/focal adhesion kinase (FAK) signaling pathway (McKean et al., 2003; Satoh et al., 2011). Although this activation mechanism is conserved among species, the induction of full blastema cells from activated fibroblasts has not been reported.
FIBROBLAST GROWTH FACTOR (FGF)-DOMINANT ENVIRONMENT CREATED BY AEC AND NERVES
FGFs have been suggested as the neurotrophic factors involved in limb regeneration. Damaged axons are speculated to release FGFs (Poulin et al., 1993; Satoh et al., 2008b, 2011). Furthermore, some FGFs are expressed in dorsal root ganglia, from which sensory nerves are projected to limbs (Li et al., 2002; Satoh et al., 2008b, 2011). Basic FGF (FGF-2), which is expressed in nerves, rescues the denervation effect during limb regeneration (Mullen et al., 1996). Studies with ALM have revealed that FGF signaling induces expression of Sp9, an AEC marker gene, in an overlying wound epithelium (Satoh et al., 2008b). The deviated nerves are covered by a migrating epidermis/epithelium called wound epithelium (WE) within several hours. Then direct interaction between the nerve and WE occurs. Sp9 is initially induced broadly throughout the layers of the overlying WE, after which expression begins to be restricted to the basal layer of WE. Within 5 days of ALM surgery, Sp9 is completely restricted to the basal layer, suggesting that AEC is induced. AEC also functions as an FGF source (Endo et al., 2000; Han et al., 2001). FGF-8 is expressed in AEC during amphibian limb regeneration (Endo et al., 2000; Han et al., 2001). The basal layer of blastema epithelium, where Sp9 is expressed, expresses FGF-8. Considering the FGF gene expression pattern in AER of higher vertebrates (Niswander et al., 1993; Savage et al., 1993; Mahmood et al., 1995; Moon et al., 2000; Niswander, 2003), it is probable that other FGF genes are expressed in AEC, in view of the apparent equivalence of the AEC and AER structures. AEC promotes mitotic activity of blastema cells (Boilly and Albert, 1990); this supports the idea that AEC is another FGF source because many FGFs show mitotic activity. Characteristically, FGF-8 can be detected in axolotl blastema mesenchyme and in AEC (Han et al., 2001), although FGF-8 is expressed specifically in AER in higher vertebrates (Mahmood et al., 1995). However, blastema cells appear after AEC induction. Therefore, the initial FGF source could be AEC and the nerve. Coincidentally, fibronectin, which is related to FGF signaling, is expressed in AEC and distal blastema mesenchyme (Nace and Tassava, 1995). It is likely that FGFs are enriched in the distal region of blastema because FGFs can bind fibronectin (Martino and Hubbell, 2010). Thus, FGFs secreted by nerves and AEC might be enriched by binding to fibronectin in the distal region of the blastema. Thus, FGF-dominant conditions are created in the distal region of the regenerating limb.
CELL ACCUMULATION AT THE SITE WHERE A BLASTEMA FORMS
In ALM, a blastema is induced on top of a deviated nerve, suggesting that the nerve attracts blastema cells from the surrounding tissue. Migration of dermal fibroblasts was observed previously in axolotl limb regeneration (Gardiner et al., 1986). Gardiner et al. used a diploid/triploid cell marker in the axolotl to investigate the movement of cells from the dermis into the early limb blastema. Cells of dermal origin began to migrate beneath the WE at about 5 days postamputation, and by 10 days they were widely distributed across the amputation surface. By 15 days, a dense accumulation of blastema cells was detected beneath the apical cap, and these cells were oriented preferentially in a circumferential direction. In ALM, dermal fibroblast migration from a dermis to a blastema was traced by injecting red fluorescent dye into a dermis. In ALM, as in regular limb regeneration, dermal fibroblasts migrated into a blastema (Endo et al., 2004). However, although FGF signaling may play a role in the migration mechanisms of fibroblasts toward a blastema, the mechanism remains unclear. FGF has been suggested as a fibroblast chemoattractant in a higher vertebrate (Li and Muneoka, 1999). FGF-4, which was expressed in an AER, showed chemotactic activity for limb bud cells. Although FGF-4 expression has not been observed in a regenerating blastema (Christensen et al., 2002), it has been reported that ectopic AEC transplantation resulted in additional blastema growth (Thornton, 1960). This implies that the ectopic AEC accumulates blastema cells and directs the growth of an additional limb. Given that FGFs showed mutual functional redundancy, other FGFs would function as chemoattractants. Elucidation of the migration mechanism is important for understanding cell movement in limb regeneration as well as chemotaxis; however, it has received little attention. To focus on this issue, cell migration was investigated focusing on FAK/Src signaling. A focus on integrin/FAK/Src signaling was suggested, since this signaling system regulates cell migration in fibroblasts of higher vertebrates and is activated in axolotl limb regeneration (Jones et al., 1999, 2001; Parsons and Parsons, 2004; Mitra et al., 2005; Satoh et al., 2011). The ECM digestion leads to an alternation of integrin gene expression. As mentioned above, ECM degradation can be expected in the early phase of limb regeneration. Indeed, integrin switching from integrin β1 to β3 occurs in this phase (Tsonis et al., 1997; Satoh et al., 2011). FAK phosphorylation follows integrin switching (Satoh et al., 2011). Src is reported as a downstream gene of FAK that controls cell migration in higher vertebrates (Sieg et al., 1999). Accordingly, we investigated FAK/Src signaling in cell migration. Axolotl dermal fibroblasts were isolated enzymatically and cultured in a plastic dish. A scratch assay was performed to investigate cell migration (Fig. 2). Axolotl dermal fibroblasts were confluently cultured on a plate and a gap was made with a pipette tip (Fig. 2A1). Cells migrated and filled the gap within 24 h (Fig. 2A2,A3). As expected, treatment with FAK and Src inhibitors delayed cell migration (Fig. 2B1–3,C1–3). This FAK/Src-dependent migration was also confirmed in vivo (Fig. 2D–I). Animals that underwent accessory limb surgery were administered inhibitors for 5 days. The blastemas were harvested and sectioned for in situ hybridization. In the control (no inhibitor), Prrx1-positive cells could be detected between the overlying epithelium and the deviated nerve (Fig. 2F,I, arrowheads). In contrast, Prrx1-positive cells were not observed at the site where a nerve was rerouted when a FAK or Src inhibitor was administered (Fig. 2D,E,G,H), suggesting that FAK/Src signaling plays a role in axolotl fibroblast migration similar to as in higher vertebrates. It is possible that the inhibitors interrupt cell proliferation because of which cell migration appeared slow. However, because cultured axolotl cells showed very slow cell proliferation (Boilly and Albert, 1988), it is unlikely that cell proliferation contributed appreciably to migration. Fibroblasts accumulating by FGF-dependent chemotaxis and a FAK/Src-dependent migration mechanism may thus be responsible for blastema formation on top of a deviated nerve.
STEPWISE LIMB REGENERATION REGULATION
ALM studies are based on the stepwise model proposed by Endo et al. (2004). The names of “ALM” and “stepwise model” would be a little confusing. But ALM is an experimental system to probe a new concept of limb regeneration named stepwise model. In the stepwise model, at least three steps and two determination factors are necessary for amphibian limb regeneration. In the first step, wound healing takes place after skin wounding. When nerves are supplied to the wound, instead of wound healing, a blastema grows at the wound site (second step). As mentioned previously, it is the availability of a nerve that distinguishes the first and second steps. The third step is limb patterning, which is determined presumably by mixed positional information. Interestingly, skin wounding and nerves are sufficient to induce a blastema, but the induced blastema regresses; such a regressing blastema develops a cartilage, which eventually disappears. The cartilage in a regressing blastema is usually round and single (Endo et al., 2004), suggesting that there is no developmental axis. This patternless structure has similarity to a regenerating Xenopus froglet limb (Satoh et al., 2005). Xenopus can regrow a structure distally after amputation, but the structure does not have a definite pattern and is called a “spike.” The cartilage in the spike is cone-shaped, but no patterning occurs along the anterior–posterior (AP) axis (Yakushiji et al., 2009). This defect is at least partially because of the loss of Shh expression (Yakushiji et al., 2009). Inhibiting Shh signaling results in the same patternless structure in axolotls (Roy and Gardiner, 2002). Thus, it would be worthwhile to investigate Shh expression in ALM blastema. Although Shh expression may underlie the second determination factor, the factor remains unknown. An accessory limb forms when contralateral skin is placed adjacent to a deviated nerve, suggesting that all positional information is present and necessary for a patterned limb (Fig. 3). When a blastema is created on the anterior side without the skin graft, presumably three of the four positional values—such as anterior, ventral, and dorsal—can be expected in the wound (Fig. 3, left). Anterior–dorsal and anterior–ventral dermal fibroblasts can accumulate in the nerve's deviated region (Fig. 3, red arrows). However, posterior fibroblasts cannot accumulate in the nerve's deviated region because they are located at a great distance for this region. A skin graft from the contralateral side provides the missing value (Fig. 3, right). The skin graft contains many posterior dermal fibroblasts placed adjacent to the deviated nerve (Fig. 1G). Therefore, posterior dermal fibroblasts can participate in a blastema as well as in dermal fibroblasts from other three regions (Fig. 3, right). These mixed positional values may constitute a force for limb patterning. However, the importance of positional values in amphibian limb development and regeneration is not yet well explained at the molecular level. There are at least three axes during limb development and regeneration, including the AP, dorsal–ventral (DV), and proximal–distal (PD). The PD axis has been most studied during limb regeneration. HoxA genes, which are regulated along the PD axis, have been reported (Savard et al., 1988; Simon and Tabin, 1993; Beauchemin et al., 1994; Gardiner et al., 1995; Endo et al., 2000; Christen et al., 2003; Ohgo et al., 2010). However, the AP and DV axes are not well studied at the molecular level, and only a few studies have been published on Xenopus (Takabatake et al., 2000; Matsuda et al., 2001). Even the gene expression patterns related to the DV and AP axes have not been reported in axolotls. To investigate whether the same position-specific genes as in higher vertebrates were expressed in axolotl blastema, we performed in situ hybridization (Fig. 4). Lmx1b is expressed in a dorsal region of a developing limb bud in higher vertebrates (Vogel et al., 1995; Loomis et al., 1998) and in the dorsal region of the axolotl blastema (Fig. 4A). Tbx3 is used as an AP marker gene in higher vertebrates because of its expression in the anterior and posterior edge of a limb bud (Bamshad et al., 1997). Tbx3 was similarly expressed in AP regions of the axolotl blastema (Fig. 4B). The expression pattern of these genes would be helpful for studying the role of positional values in limb regeneration, although the expression pattern of position-specific genes does not represent positional value itself. While we can analyze limb regeneration in a stepwise fashion with the ALM system, molecular data will be required for further progress.
POSSIBILITY OF APPLICATIONS TO HIGHER VERTEBRATES
Many processes are common to limb regeneration in urodele amphibians and limb development in higher vertebrates. The FGF-signaling pathway plays a role in inducing a limb bud from a LPM during limb development (Tanaka and Gann, 1995). Based on ALM studies, it is likely that FGF signaling also functions during the induction phase of limb regeneration (Satoh et al., 2008a, b, 2011). Moreover, AER or AER factor (FGF-2 or FGF-4) grafting onto an amputated limb bud in chick embryos induced a regeneration response (Hayamizu et al., 1994; Taylor et al., 1994; Kostakopoulou et al., 1996; Satoh et al., 2010b). Hence, FGF signaling likely plays a role in the early phase of limb development and regeneration. In addition, as mentioned above, Prrx1 activation through the integrin/FAK-signaling pathway also appears to be conserved in both human and axolotl fibroblasts (McKean et al., 2003; Satoh et al., 2011). Notably, Prrx1 is a direct upstream regulator of tenascin-C in human fibroblasts, and tenascin-C is a classical blastema marker gene (Onda et al., 1990). Thus, it is likely that higher vertebrates maintain a basic gene network(s) of regeneration even though some elements may be missing. Research into limb regeneration in higher vertebrates has already started and it is becoming possible to supply missing factor(s) (Han et al., 2003, 2005, 2008; Masaki and Ide, 2007; Muneoka et al., 2008; Satoh et al., 2010b; Yu et al., 2010; Fernando et al., 2011). These advances encourage further attempts to describe the extent of conservation among amphibians and higher vertebrates.
Finally, new experimental techniques employing the ALM afford new opportunities to understand blastema induction mechanisms in animals that can regenerate. ALM studies suggest that even the initiation mechanisms of limb regeneration, which has been considered to be specific to regenerative animals, are conserved in higher vertebrates. However, human beings cannot reactivate most of their developmental genetic networks after birth. The question why higher vertebrates cannot induce a blastema after fibroblast activation needs to be answered. We hope that a combination of ALM and recently developed high-throughput analysis will provide the answer.
MATERIALS AND METHODS
Axolotl dermal fibroblasts and blastema cells were prepared as described previously (Satoh et al., 2007). Basic culture medium contained 80% L15, 1% fetal calf serum (FCS), and 200 μg mL−1 gentamycin. Mouse FGF-2 (100 ng mL−1; R&D Systems) and human FGF-8 (50 ng mL−1; Peprotech) were supplied for experimental purposes. Axolotl dermal fibroblasts were cultured in a penicillin cup to obtain confluent cells for the scratch assay. After a 24-hr incubation, a scratch was made with a 2-μL pipette tip and then inhibitors were added. FAK inhibitor 14 (20 μM; Sigma) and Src inhibitor (3 μM, SU6656; Calbiochem) were used to study their effects on limb regeneration. Cell migration was observed under a microscope (Olympus, SZX16).
The axolotls were raised in tap water. We used 10–12-cm length axolotl. The ALM surgical procedure has been described previously (Endo et al., 2004). They were kept for several hours to heal the surgical damage and then immersed in tap water with or without inhibitors. Concentrations of inhibitors were the same as for the scratch assay. The inhibitor-containing water was refreshed at Days 0, 2, and 4. Samples were harvested 5 days after the ALM surgery.
The in situ hybridization procedure was the same as that used in a previous study (Satoh et al., 2007). Lmx1b and Tbx3 were isolated by RT-PCR as previously described (Satoh et al., 2007). Their gene sequences have been deposited in Sal-Site (http://www.ambystoma.org/).
The authors thank Dr. D. M. Gardiner and Dr. T. Endo for constructive comments.