While urodele amphibians (newts and salamanders) can regenerate limbs as adults, other tetrapods (reptiles, birds and mammals) cannot and just undergo wound healing. In adult mammals such as mice and humans, the wound heals and a scar is formed after injury, while wound healing is completed without scarring in an embryonic mouse. Completion of regeneration and wound healing takes a long time in regenerative and non-regenerative limbs, respectively. However, it is the early steps that are critical for determining the extent of regenerative response after limb amputation, ranging from wound healing with scar formation, scar-free wound healing, hypomorphic limb regeneration to complete limb regeneration. In addition to the accumulation of information on gene expression during limb regeneration, functional analysis of signaling molecules has recently shown important roles of fibroblast growth factor (FGF), Wnt/β-catenin and bone morphogenic protein (BMP)/Msx signaling. Here, the routine steps of wound healing/limb regeneration and signaling molecules specifically involved in limb regeneration are summarized. Regeneration of embryonic mouse digit tips and anuran amphibian (Xenopus) limbs shows intermediate regenerative responses between the two extremes, those of adult mammals (least regenerative) and urodele amphibians (more regenerative), providing a range of models to study the various abilities of limbs to regenerate.
Limb regeneration is one of the best examples of organ/appendage regeneration in vertebrates and has been called ‘epimorphosis’ since it requires blastema formation and proliferation (Brockes 1997; Suzuki et al. 2006) though there has been criticism of the classical definition of epimorphosis and morphallaxis (Agata et al. 2007). Among tetrapods, the cellular and molecular mechanisms involved in limb development are highly conserved; fully developed limbs share a common skeletal pattern (Fig. 1; Muneoka & Sassoon 1992). On the other hand, the regenerative responses of limbs after amputation differ from animal to animal among tetrapods (Fig. 1). Birds cannot regenerate limbs at any stage of development and, surprisingly, mammals have slightly better limb regenerative capacity than that of birds. Embryonic and neonatal mice can regenerate their digit tips if they are amputated through the distal phalanx (Borgens 1982; Reginelli et al. 1995), and similar digit tip regeneration occurs in humans (Douglas 1972; Illingworth 1974). After amputation at a more proximal level, a neonatal mouse cannot regenerate lost parts, and hypertrophy of amputated bones occurs (Masaki & Ide 2007). In contrast, amphibians have exceptionally high regenerative capacity for limb regeneration. Urodele amphibians such as newts and salamanders can regenerate their limbs following amputation any time during their life cycles (Fig. 1), although there is a non-regenerative mutant in axolotls (Sato & Chernoff 2007). Anuran amphibians such as Xenopus are intermediate between urodele amphibians and other vertebrates in terms of their regenerative capacity (Fig. 1), in that they can completely regenerate developing hindlimb buds prior to the onset of metamorphosis, but regenerative capacity declines gradually as metamorphosis proceeds (Dent 1962; Muneoka et al. 1986; Suzuki et al. 2006).
After amputation of most tetrapod limbs, seemingly similar gross morphological steps occur initially. If the limb is non-regenerative such as in adult mammals, the limb stump merely undergoes wound healing. If the limb is regenerative such as in salamanders, the limb stump initiates the wound healing process immediately after amputation but then initiates further processes of limb regeneration, resulting in successful regeneration. I will summarize the current knowledge about these sequential steps during limb regeneration, showing what steps are critical in determining limb regenerative capacity and how we can make further progress in revealing the molecular nature of these sequential steps during limb regeneration in the future.
Processes occurring after limb amputation: Wound healing versus limb regeneration
When a non-regenerative limb of tetrapods is amputated, wound healing without initiation of the limb regeneration process will occur, although hypertrophy of bones at the stump is observed in neonatal mice (Masaki & Ide 2007). Typical processes of such wound healing have mainly been studied in mammals, including mice and humans, in which limbs cannot be regenerated but entire damaged skin is healed (healing of other tissues such as bone, muscle and tendon also occurs, but the details of these processes are omitted in this review). Wound healing in the skin of adult mammals after injury is a multiple-step process (reviewed by Broughton et al. 2006; see also Stocum 1995a; Martin 1997). Three overlapping phases of the healing process of a skin wound can be distinguished: inflammation, proliferation and maturation (Broughton et al. 2006; Fig. 2A). The following description of the time courses of the three steps is based on results of a study by Broughton et al. (2006).
Wound healing (time course for human adults)
1Inflammation (immediately after wounding until days 4–6 involving clot formation, entry of neutrophils and macrophages, and initiation of re-epithelization): After injury, hemostasis serves as the initiating step and foundation for the healing process. Injured blood vessels vasoconstrict and the endothelium and nearby platelets activate the intrinsic part of the clotting cascade. Platelets release cytokines and growth factors that initiate the inflammatory response (Witte & Barbul 1997). Immediately after the clot has formed, it attracts neutrophils and then monocytes, which are transformed into macrophages (Broughton et al. 2006). Macrophages mediate angiogenesis, synthesize matrix metalloproteinases (MMPs) to degrade the collagens of damaged tissues and, most importantly, stimulate re-epithelization of the wound, resulting in transition to the proliferative phase (Stocum 1995a; Broughton et al. 2006).
2Proliferation (day 4 to day 14 involving continued re-epithelization, fibroblast migration/proliferation, change in matrix composition, and angiogenesis): Epithelial cells located at the edge of the wound begin proliferating and extending projections to form a barrier against fluid losses and further bacterial infections. Fibroblasts begin synthesizing collagen and proliferate to form granulation tissue. Transforming growth factor-β (TGF-β) induces fibroblasts to synthesize type I collagen and reduce production of MMPs (Goldman 2004). In the cases of healing of a larger wound, wound contracture and epithelization are induced, in part, by TGF-β (Yang et al. 1997).
3Maturation (day 8 until 1 year involving collagen remodeling): The main feature of this phase is the deposition of a well-organized network of extracellular matrix (ECM). Initially, the matrix is composed mainly of fibrin and fibronection (Kurkinen et al. 1980). The collagen is initially laid down as thinner collagen than that in uninjured skin and is oriented parallel to the skin. The initial collagen threads are reabsorbed, deposited thicker and organized along the stress lines over time. If there is excessive collagen synthesis, a hypertrophic scar or keloid is formed.
In adult mammals, wound healing often results in scar or keloid formation. In contrast, wounds in embryonic and fetal mammals heal without scar formation (Stocum 1995a; Martin 1997; Ferguson & O’Kane 2004). A remarkable difference between adult wound healing and embryonic/fetal wound healing is that the process of embryonic/fetal scar-free healing involves a smaller inflammatory response and expresses a lower level of TGF-β expression (Stocum 1995a; Martin 1997; Ferguson & O’Kane 2004). Therefore, it seems that there is a critical point in the early phases of healing (inflammatory and proliferation) when it is determined whether healing will result in scar formation or not.
Regeneration of a limb after amputation in urodeles usually involves the three steps of wound healing, dedifferentiation and redevelopment (Fig. 2B) as proposed by the Bryant-Gardiner lab (Bryant et al. 2002; Gardiner et al. 2002). The following description of the time course of the three steps is based on results of a study by Tank et al. (1976).
Limb regeneration (time course of forelimb regeneration in an adult axolotl)
1Wound healing (immediately after amputation until day 5 involving epithelial healing, start of MMP expression and histolysis): Within four to 12 hours after amputation, the amputation plane is covered with an epithelial cell layer called the ‘wound epithelium’, which is made up of cells that have migrated from the limb stump (Repesh & Oberpriller 1978; Carlson et al. 1998). This rapid healing of the wound minimizes tissue damage, infection and the inflammatory response (Han et al. 2005). Mesodermal tissues of the limb stump undergo histolysis involving remodeling of the ECM. This process is triggered by MMPs (Yang & Bryant 1994; Miyazaki et al. 1996). However, wound healing and histolysis are not specific to limb regeneration. These phenomena are also observed during the process of repair of damaged tissues in other vertebrates as mentioned above.
2Dedifferentiation (day 6 to day 20 involving dedifferentiation, blastmema formation and induction of regeneration-specific genes): The wound epithelium thickens and forms an apical epithelial cap (AEC). This structure functionally corresponds to the apical ectodermal ridge (AER), a specialized epithelial structure essential for outgrowth of developing limb buds of amniotes (Muneoka & Sassoon 1992). The secretion of MMPs and the formation of the AEC promote the generation of a population of undifferentiated mesenchymal cells called a ‘blastema’. This is most likely formed through a ‘dedifferentiation’ process from differentiated tissues. At the end of this process, a cone-shaped blastema is formed.
3Redevelopment (day 21 to day 30 or later involving growth of blastema, pattern formation and redifferentiation): Once a blastema is established, it can form an entire limb regenerate autonomously even if it is grafted to different locations of the body such as the eye chamber or dorsal fin (Pietsch & Webber 1965; Stocum 1968). The blastema continues to grow distally via cellular proliferation until the completion of regeneration. During this process, close interactions between the blastemal mesenchyme and the AEC are essential for elongation of the limb regenerate. These interactions are likely to be mediated by several growth factors, including fibroblast growth factors (FGFs), as during limb bud elongation in development. Simultaneous with blastema elongation, cells of the blastema direct redifferentiation and re-patterning to eventually form a completely regenerated limb.
While both wound healing in mammals and limb regeneration in amphibians take a long time to be completed, wound closure occurs much faster in amphibian limb regeneration. For example, in larval and adult urodele amphibians, wound closure is completed within 4 h and 12 h, respectively (Repesh & Oberpriller 1978; Carlson et al. 1998), while a similar-size mammalian wound caused by amputation of a mouse digit takes many days to close (Han et al. 2005). In urodele limb regeneration, the wound is re-epithelialized very quickly and the genes involved in wound healing e.g. Msx-2 and Mmp-9, are expressed early in response to amputation in a nerve-independent manner (Bryant et al. 2002; Gardiner et al. 2002). However, in the processes that follow wound healing such as blastema formation and induction of regeneration-specific gene expression, limb regeneration is distinct from a non-regenerative wound healing response and these processes depend on the nerve supply (Bryant et al. 2002; Gardiner et al. 2002).
As described above, Xenopus can regenerate limb buds after amputation until a certain point in development, after which limb amputation results in a non-regenerative wound healing response. Since it has been shown that the ontogenic decline of regenerative capacity is due to intrinsic change in the Xenopus limb bud itself (Sessions & Bryant 1988), the Xenopus provides an excellent model to investigate essential differences between regenerative limbs and non-regenerative ones within the same animal. Furthermore, there are several technical advantages in Xenopus, such as highly efficient generation of transgenic individuals and availability of information on the entire genome due to the completion of a genome project in Xenopus tropicalis (Blitz et al. 2006).
Signals involved in the initiation of limb regeneration
While it remains difficult to functionally analyze candidate genes for their involvement in the early stages of limb regeneration, much information on gene expression changes during this period has accumulated. Since many of these signaling molecules (e.g. FGF and Wnt) are used in other organ/appendage regeneration systems such as tail regeneration of amphibians, fin regeneration of teleosts and heart regeneration of teleosts (reviewed by Nakatani et al. 2007; Stoick-Cooper et al. 2007), molecular mechanisms of regeneration may be conserved among multiple organs/appendages to some extent.
Matrix metalloproteinases (MMPs) were discovered as proteases capable of digesting collagen in the remodeling tissues of metamorphosing tadpoles (Gross & Lapiere 1962; Fujimoto et al. 2006, 2007). MMPs are activated in regenerating limbs of newts and salamanders (Grillo et al. 1968; Yang & Bryant 1994; Park & Kim 1999) and are also activated during inflammation of wound healing and function to clear inflammatory debris in mammals (Parks 1999; Broughton et al. 2006). Recently, Vinarsky et al. (2005) reported that some MMPs are upregulated very early after amputation and that urodele limb regeneration can be partially inhibited by treatment with a synthetic MMP inhibitor. These findings suggest that MMPs are specifically required for limb regeneration, especially during initiation (wound epithelium formation/subsequent blastema formation). Since Vinarsky et al. (2005) treated regenerating newt limbs with the MMP inhibitor for 60 days from 1 h after amputation; it is possible that MMPs are required for limb regeneration not only during initiation but also throughout the entire process of regeneration. A temporally restricted application of the MMP inhibitor during limb regeneration might enable more precise determination of the role of MMPs during limb regeneration.
Developing limb buds in vertebrates are mainly composed of mesenchyme derived from the lateral plate mesoderm (LPM) and epithelium derived from the ectoderm. Epithelial–mesenchymal interactions are necessary for limb regeneration (Polezhaev & Faworina 1935; Goss 1956; Stocum & Dealove 1972; Mescher 1976) as well as for outgrowth of a developing limb bud (Saunders 1948; Zwilling 1956; Summerbell 1974). With regards to a developing limb bud, recent studies have revealed that these interactions are mediated by FGFs during embryogenesis (reviewed by Martin 1998; Mariani & Martin 2003). Specifically, fgf-10 and fgf-8 are expressed in the lateral plate mesoderm of the presumptive limb field and its overlying epithelium, respectively, during limb induction. FGF-10 in limb bud mesenchyme and FGF-8 in the apical epithelium of the limb bud actually constitute a positive feedback signaling loop essential for limb outgrowth in the amniote embryo (Ohuchi et al. 1997; Xu et al. 1998).
Several recent studies have shown that both fgf-10 and fgf-8 are expressed in axolotl limb blastemas (Christensen et al. 2001, 2002; Han et al. 2001). Similarly, in regenerating blastemas of Xenopus, fgf-10 and fgf-8 are expressed in the apical epithelium and in the mesenchyme, respectively, while neither fgf-10 nor fgf-8 are expressed after amputation of a non-regenerative limb (Christen & Slack 1997; Yokoyama et al. 2000). Reciprocal transplantation experiments have revealed that the regenerative capacity of a Xenopus limb depends on mesenchymal tissue, not epithelial tissue (Yokoyama et al. 2000). Furthermore, exogenously applied FGF-10 successfully stimulates regeneration in the stump of non-regenerative Xenopus limb buds (Yokoyama et al. 2001). These results suggest that a positive feedback signaling loop between FGF-10 and FGF-8 is also essential for successful limb regeneration as well as for limb development and that the reinforcement of this feedback loop may be a crucial key to enhance the regenerative response of non-regenerative vertebrate limbs.
The Wnt/β-catenin pathway is an evolutionarily conserved signaling pathway that is known to control cell proliferation and cell fate determination by regulating target gene expression (Miller et al. 1999; Huelsken & Behrens 2002). Wnt/β-catenin has been shown to be involved in the initiation of chick limb development and zebrafish pectoral fin formation, by inducing fgf-10 expression in the presumptive limb and fin region, respectively (Kawakami et al. 2001; Ng et al. 2002; reviewed by Yang 2003; Mercader 2007). Two recent independent loss-of-function studies have also indicated an essential role for Wnt/β-catenin during limb regeneration. Kawakami et al. (2006) inhibited Wnt/β-catenin signaling by infecting limb stumps with Axin1- or Dkk1-expressing viruses, resulting in hypomorphic regeneration or no regeneration. Conversely, by using a constitutively active form of β-catenin, Kawakami et al. (2006) were able to improve regenerative capacity in non-regenerative Xenopus limb buds at a stage just past when they lose their natural ability to regenerate limb buds (St. 53–54 [Nieuwkoop & Faber 1994]). Separately, my colleagues and I used transgenic Xenopus tadpoles that express Dkk1 under the control of a heat shock promoter (Fig. 3A), which allowed for heat shock-inducible expression of Dkk1 at specific time points (Fig. 3B,C). Heat shock immediately before limb amputation or during early blastema formation blocked limb regeneration completely with high efficiency (80%), while 70% of control tadpoles without Dkk1 expression regenerated their limbs completely (Fig. 3D,E; Yokoyama et al. 2007). After blastema formation, however, Dkk1 did not block regeneration completely but allowed a partial regenerative response (Yokoyama et al. 2007). Furthermore, inactivation of Wnt/β-catenin signaling by Dkk1 abolished fgf-8 but not fgf-10 expression (Yokoyama et al. 2007). These results indicate that Wnt/β-catenin signaling is required during the early stages of limb regeneration but is not absolutely essential after blastema formation.
BMPs, a subset of the TGF-β superfamily, have numerous roles during embryogenesis and organogenesis, regulating processes as diverse and fundamental as regional specification and cell proliferation, differentiation, survival and death (reviewed by Hogan 1996). Msx genes are often expressed in regions similar to those in which BMPs are present and mediate BMPs function. For example, in the dorsoventral axis formation of an embryo, Msx-1 mediates the ventralizing activity of BMP4 (Maeda et al. 1997). In early limb development, BMP signaling and its direct target, Msx-1, are involved in the induction of the AER (Pizette et al. 2001; Robert 2007). In regenerating urodele limbs, Msx-2 is rapidly induced in the healing epithelium and subjacent tissues following amputation or simple wounding, whereas Msx-1 expression is induced only in blastemal cells (Carlson et al. 1998; Koshiba et al. 1998). Msx-1 is also expressed in the blastemas of both tadpole hindlimb buds and froglet forelimbs (Endo et al. 2000). In amphibian limb regeneration, Beck et al. (2006) first demonstrated that BMP signaling is required for limb regeneration by using a stable transgenic Xenopus line in which expression of a BMP inhibitor, noggin, was induced under control of the heat shock-inducible hsp70 promoter. Interestingly, the period for BMP requirement during limb regeneration is later than that for Wnt/β-catenin requirement. While heat shock at 3–4 h prior to amputation sufficiently blocked limb regeneration in heat shock inducible Dkk-1 (hsDkk-1) tadpoles (Yokoyama et al. 2007), similar timing of a single heat shock (3 h prior to amputation) did not block limb regeneration in heat shock inducible noggin tadpoles (Beck et al. 2006). In the noggin tadpoles, at least two heat shocks at 24 and 48 h after amputation are required to sufficiently block limb regeneration (Beck et al. 2006). These findings indicate that both Wnt/β-catenin and BMP signaling have essential roles in the early stages of limb regeneration but that Wnt/β-catenin signaling is required at an earlier period than is BMP signaling during blastema formation.
BMP/Msx signaling is also involved in the regeneration of a mouse digit tip. As previously mentioned, fetal and neonatal mice can regenerate digit tips, and the regenerative capacity correlates with the expression domain of the Msx-1 gene (Reginelli et al. 1995). Furthermore, Han et al. (2003) showed that Msx-1 mutant mice display a digit tip regeneration defect. This defect can be rescued with BMP-4 and wild-type digit tip regeneration can be inhibited with Noggin in an organ culture regeneration model in vitro (Han et al. 2003). Moreover, Msx-1 and BMP-4 are both expressed in the regenerating digit tips and BMP-4 seems to be required for regeneration downstream from Msx-1 (Han et al. 2003). More recently, Allan et al. (2006) found by using a similar organ culture regeneration model that Msx-1 expression is also upregulated in human fetal digits after amputation during tip regrowth. It would be interesting to examine when BMP/Msx signaling is required for digit tip regeneration by applying signaling antagonists such as Noggin only at a specific time point.
Nerve signals and a stepwise model
It is well known that limb regeneration requires neuronal innervation in urodele amphibians (reviewed by Stocum 1995b). It has been suggested that axons secrete what have been called ‘neurotrophic factors’ into the amputated limb and that these factors stimulate mitotic activity of blastemal cells and upregulate genes important for the regenerative process. If a limb stump is denervated, it fails to regenerate. However, if denervation is done after some stage of blastema formation (medium bud stage), limb regeneration is not blocked (Singer 1952). In fact, blastemal cells in a denervated limb undergo apoptosis, suggesting that these ‘neurotrophic factors’ are survival factors during limb regeneration (reviewed by Bryant et al. 2002 and Mescher 1996). There are currently two major candidates for these so-called neurotrophic factors: GGF and FGF-2. Glial growth factor (GGF), a member of the neuregulin growth factor family, is present in regenerating blastema and GGF distribution is lost in the limb stump after denervation (Brockes & Kintner 1986). Furthermore, injection of GGF protein into the denervated blastema actually increases the rate of regeneration of blastemas (Wang et al. 2000). Similar to GGF, FGF-2 protein is distributed in nerves and the AEC in regenerating blastemas, and FGF-2-soaked beads can rescue regeneration of denervated blastemas (Mullen et al. 1996). Besides these two candidates, there are several other candidate molecules for ‘neurotrophic factors’. One example, an iron transport protein of the blood, transferrin, accumulates in peripheral nerves. The ‘neurotrophic factor’ -dependent proliferation of blastemal cells can be arrested by the removal of transferrin and can be restored by its re-addition in vitro (Mescher 1996).
Interestingly, ectopic nerve deviation to a wound on the side of a limb could occasionally induce blastema-like outgrowth (a bump) in urodeles (Bodemer 1958, 1959). Endo et al. (2004) refined this experiment and developed a stepwise system to induce ectopic limb formation by nerve deviation and skin grafting from the contralateral side of the limb. If a nerve is deviated to the wound site, a blastema-like bump is induced but it then regresses. However, if a skin flap is grafted from the opposite side of the limb concomitantly with nerve deviation, the ectopic blastema continues to grow and forms an ectopic limb (Endo et al. 2004). Based on these experiments, a stepwise model was proposed by the Bryant-Gardiner lab (Bryant et al. 2002; Gardiner et al. 2002): signals from nerves and then signals from fibroblasts are required for successful limb regeneration (Fig. 4A: Endo et al. 2004). While the molecular entities of these signals are still unknown, this model predicts when putative signals will be released from both nerves and fibroblasts and provides a valuable guideline for elucidating the initial steps of limb regeneration. When and at which point it diverges from successful limb regeneration to failure of limb regeneration is unknown at this point. Temporal-specific inhibition of signaling pathways by using transgenic animals with a heat shock-inducible system or by using drug inhibitors or activators could provide powerful tools to identify the molecular entities of the stepwise signals (Fig. 4B).
Both wound healing and limb regeneration take a long time to complete, but it is likely that the earliest processes of regeneration are critical in determining what the final outcome will be. While the molecular mechanisms required for the earliest processes that occur following amputation or wounding should be better characterized, attempts to improve the regenerative response of non-regenerative model systems by stimulating signaling molecules and/or expanding multipotent blastemal cells in limb stumps should be done as well. There exists a large gap between the process of wound healing with scar formation that occurs in adult mammals and the process of perfect limb regeneration of urodele amphibians (Fig. 5). However, there are excellent models of more ‘intermediate’ examples of regeneration/healing, including scar-free healing of the mouse embryo, digit tip regeneration of the mouse, and limb regeneration of anuran amphibians. Attempts to improve the regenerative responses of such models step-by-step by manipulating divergent points during the earliest processes of regenerative responses will hopefully lead in the not so distant future to an understanding about what needs to be done to enable non-regenerative mammals such as humans to regenerate their amputated limbs.
The author thanks Cristi Stoick-Cooper, Randall Moon, Koji Tamura and Hiroyuki Ide for critical reading of the manuscript and discussions. The author also thanks Tetsuya Endo and David Gardiner for permission to use the scheme of the stepwise model in this review. The author was supported by JSPS Research Fellowships for Young Scientists and JSPS Postdoctoral Fellowships for Research Abroad.