The molecular basis of amphibian limb regeneration: integrating the old with the new

Is regeneration close to revealing its secrets? Rapid advances in technology and genomic information, coupled with several useful models to dissect regeneration, suggest that we soon may be in a position to encourage regeneration and enhanced repair processes in humans.


Introduction
The study of amphibian limb regeneration has a rich experimental history.After many decades of research at the tissue and cellular levels, much is known about the phenomenology and the basic organizing principles of regeneration.In recent years, molecular analyses have begun to provide insights into the mechanisms controlling regeneration.It is already clear that regeneration involves complex molecular interactions between multiple tissues, and thus we can expect that there will be many important genes involved, rather than just a small number of 'regeneration genes'.With the recent input from the axolotl EST project to the databases, nearly 500 non-redundant 'regeneration' genes have been cloned and identified, 1 a number that likely will increase dramatically over the next few years.Consequently, the availability of cloned regeneration genes will not be a limiting factor to progress in understanding limb regeneration.The challenge will entail drawing on the wealth of information from From the Department of Developmental and Cell Biology, Developmental Biology Center, University of California Irvine, Irvine, CA 92697, USA.
classical studies to guide the identification of the functions of this large set of genes.
The challenge of understanding the mechanisms controlling the biology of complex systems is hardly unique to regeneration biology.In recent years, techniques have become available to identify all the molecular components of a system, and to study the interactions between those components.Key to the success of such an approach is the ability to identify the molecules, while at the same time having an understanding of the cell and tissue level properties of the system.The goal of this review is to discuss key insights from the classical literature as well as more recent molecular findings.We focus on three critically important cell types: fibroblasts, epidermis and nerves.Each of these is necessary, and together they are sufficient for the regeneration of a limb.Although the final limb is composed of a variety of other cells and tissues, such as muscle, blood vessels and pigment cells, these other cell types do not appear to be necessary for the control of growth and pattern formation during regeneration, but rather they respond to signals from nerves, fibroblasts and the epidermis.Thus at this time, these three cell types pose both great challenges and promising opportunities for research directed toward developing an integrated view of the molecular interactions controlling limb regeneration.

Fibroblasts
During regeneration, growth and pattern formation are coordinately regulated by interactions between cells that are derived from fibroblasts of the connective tissues of the amputated stump (see References 2, 3).The function of connective tissue fibroblasts is demonstrated qualitatively by grafting studies to induce new or altered pattern and quantitatively by cell contribution studies.As a generalized conclusion, grafts of tissues that contain fibroblasts affect growth and pattern formation; whereas, grafts that do not contain fibroblasts do not. 4Among these various tissues, the dermis plays a particularly dominant role during regeneration.
A direct demonstration of the importance of the dermis comes from studies of regeneration in X-irradiated limbs.][7] The regenerated limbs are formed from graft-derived dermal fibroblasts, and have a normal pattern of skeletal and connective tissues, blood vessels and nerves, though they lack muscles.Since the stump muscles are irradiated, precluding muscle precursor cells from migrating distally into the regenerate, it follows that myogenic cells are not required to build a normal limb pattern during regeneration, in common with similar findings in developing limbs. 8We conclude that the fibroblast-derived mesenchymal blastema that eventually reforms the cartilaginous skeleton and associated connective tissues, forms a blueprint that guides the migration and growth of nerves, blood vessels and myogenic cells.
Descriptive histological studies suggested that each tissue of the mature limb contributed cells in proportion to its availability in the stump, 9,10 but subsequent lineage analysis does not support this interpretation. 11he progeny of dermal fibroblasts account for between 19 and 78% of the cells of the early blastema (42% on average), even though dermal fibroblasts represent less than 20% of the cells of the stump, suggesting that the fibroblast population is subject to selective expansion in the blastema.Since dermal fibroblasts account for about half of all fibroblasts in the limb, 12 it is possible that essentially all of the early blastemal cells are derived from fibroblasts, even though these cells account for less than half of the cells of the mature limb.At present it is unclear whether there is a population of quiescent stem cells in the dermis that are activated during regeneration, or whether dermal fibroblasts become dedifferentiated by losing their differentiated phenotype and reversing their cell fate to become stem cells.Nevertheless, an important consideration for future experimental work on the biology of stem cells for limb regeneration is that such a population of cells likely will be isolated from the dermis.

Epidermis
Although epidermal cells do not contribute directly to the blastema (see References 11, 13), a covering of specialized epidermis is critical for the success of regeneration.Epidermal cells function to enable outgrowth, and may also function in the control of pattern formation, as is the case during limb development.A specialized wound epidermis (WE) forms during the initial healing of the wound surface by the migration of an epidermal sheet of cells derived from the basal cells of the mature skin epidermis (see Reference 3).This epidermal layer subsequently thickens, as a result of continued cell migration, to form the apical epithelial cap (AEC), and acquires unique functions associated with regeneration.A new basal lamina is not reformed until relatively late during regeneration, which presumably is critical in allowing for epithelial-mesenchymal interactions (see Reference 3).
Treatments that affect the formation of the WE or the AEC alter the course of regeneration (see Reference 13).Formation of the WE is inhibited by a graft of mature skin over the amputation surface, in which case regeneration is inhibited.Removal of the WE or AEC after it has formed also inhibits regeneration.Conversely, experiments that relocate the position of the AEC induce limb outgrowth at the new position.

Nerves
The importance of nerves in regeneration has long been recognized, but the molecular mechanisms mediating neuronal influences are still largely unknown.Nerves are severed during amputation, begin to regenerate rapidly into the stump tissues at the amputation plane, and subsequently innervate the blastema and overlying epidermis (see Reference 3).If the limb is denervated during the early stages of regeneration, the limb fails to regenerate.In addition to axons, nerves contain connective tissue cells and Schwann cells.The function of nerves during limb regeneration is not considered to be associated with these non-neuronal cells since they are present whether or not the limb has been denervated.Consequently, the critical cell type is considered to be the neuron, which is hypothesized to produce a 'neurotrophic factor' required for the initiation and progression of the early stages of regeneration.Denervation of regenerating limbs at later stages inhibits further growth of the regenerate, but not redifferentiation of normally patterned limb tissues (see Reference 3).
The production of factors required for regeneration may be a normal function of nerves.Alternatively, this function of nerves may be acquired in response to amputation, as is the case for fibroblasts and the epidermis.Presumably this function would be induced and maintained by interactions with blastema cells and epidermal cells, and would be coupled with the interactions that stimulate growth and pattern formation.Since the behavior of nerves during regeneration has not been well characterized at the molecular level, the nature of such interactions cannot be evaluated at the present time.
Understanding the function of nerves is critical to devising strategies to induce regeneration in humans, since nerves are the source of factors that are required for the recruitment and proliferation of limb stem cells.Progress in this area of research has been limited by the lack of efficient and appropriate functional assays.Many experiments involve denervation of the limb, which inhibits regeneration, followed by application of test treatments for their ability to rescue regeneration.Rescue is generally limited in nature (see References 3, 14) and rarely results in normal regeneration (however, see Reference 15).Such experiments provide a negative assay for a positive acting factor, and have yet to identify the elusive neurotrophic factor.

Building on the past: model systems for understanding limb regeneration
Adult urodele amphibians are unique among vertebrates in their ability to regenerate their limbs perfectly.These organisms thus offer the unique opportunity to discover the events and processes that occur, and the genes that are expressed during successful regeneration.Ironically, it is challenging to test the ability of a gene to induce a regenerative response in urodeles since regeneration is the default response.Experiments designed to inhibit regeneration yield negative results, which are difficult, if not impossible to interpret.The alternative approach of first inhibiting regeneration (e.g. by denervating the limb) and then testing the ability of a gene or factor to rescue regeneration is again basically negative in design, and subject to experimental artifacts.Fortunately, as discussed next, there are a variety of experimental models that hold great potential to identify key signals controlling growth and pattern formation during regeneration.Some are based on relatively recent discoveries; whereas, others are based on classical studies that can be reinvestigated using modern techniques to analyze molecular function.

Molecular studies of normal regeneration in urodeles
Studies of gene expression in regenerating urodele limbs have led to the conclusion that there are at least three distinct phases of limb regeneration (Figure 1, Table 1).During phase I, the wound is healed, and patterns of gene expression are comparable for both amputation wounds and lateral skin wounds.7][18] Phase II is unique to limb regeneration and involves the process that generates the population of undifferentiated, proliferating blastema cells, either by 'dedifferentiation', and/or by activation of quiescent stem cells.The divergence of phase II limb regeneration from the wound-healing pathway is first evidenced by the reexpression of Hoxa-9/Hoxa-13, preceding by many days the morphologically observable changes in stump tissues that will eventually lead to the accumulation and proliferation of blastema cells. 19These genes function in patterning the proximal-distal limb axis and their early co-expression suggests that the earliest event in regeneration is the specification of the distal tip of the pattern (see Reference 2).1][22][23][24] Phase III of regeneration is characterized by growth and differentiation of the blastema.In this phase, studies of both gene expression and cell-cell interactions indicate that the blastema is functionally equivalent to the developing limb bud (see Reference 2), and thus phase III corresponds to the redevelopment of the limb.Each of the three phases has distinct characteristics that can be investigated individually in the accessory limb model discussed later.Most of the genes expressed during phases II and III of regeneration are also expressed during limb development.Although the spatial and temporal patterns of expression may vary, their functions are conserved during regeneration and development (see References 2, 25).One exception concerns the expression of Hoxc-10 in regenerating forelimbs of the axolotl.Hoxc-10 is expressed in developing hindlimbs and tails, but not in developing forelimbs.During limb regeneration, Hoxc-10 is reexpressed in hindlimbs and tails, and is also expressed in forelimbs. 26Genes within the HoxC complex are involved in the specification of positional identity along the rostral-caudal axis of vertebrate embryos, and expression during axolotl development is similar to that observed in other embryos. 26,27 lthough the function of Hoxc-10 in regenerating forelimbs is unclear at this time, its expression suggests the presence of signals in regeneration that are not present in development.Since Hoxc-10 gene expression is responsive to these signals, which are likely to influence expression of a suite of genes, analysis of the regulation of this gene provides an opportunity to isolate and identity regeneration-specific signals.

Regenerative decline in Xenopus as a model for inducing limb regeneration
Although the limbs of most adult vertebrates do not regenerate, in all examples studied, developing limb buds can regenerate.Regeneration of amputated limb buds has been most extensively investigated in anuran amphibians, in particular, Xenopus.Early stage limb buds can regenerate perfectly; whereas, by the end of limb development, all regenerative abilities are lost.Loss of regenerative ability occurs progressively, 28,29 and is associated with the progressive loss of the ability to reexpress genes involved in growth and pattern formation (e.g.shh 30 ).During this period, regenerative ability can be rescued to a limited extent by the application of exogenous FGF (FGF2 and FGF4 in chick and FGF10 in Xenopus (see References 31, 32).Exogenous FGFs presumably substitute for the normal function of the apical epidermis in the production of FGFs, an ability that is lost coincident with the loss of regenerative ability. 33,34 xperimental treatment of amputated Xenopus limb buds with FGF10, which is normally expressed in the mesenchyme, induces a number of genes, including fgf8, in the epidermis, and partially rescues regeneration of late stage limbs that normally would not regenerate. 32A comparable result is observed in amputated chick limb buds that are supplied with either FGF or a grafted apical epidermis (see Reference 31).Relatively little is known about the expression and function of FGFs in either developing or regenerating limbs of urodeles.fgf8 and fgf10 are expressed in both the apical epidermis and the mesenchyme of regenerating urodele limbs 35,36 and likely have equivalent functions to those in developing limbs.
The epidermis is also involved in dorsal-ventral pattern formation during limb development, in which expression of En-1 in the ventral epidermis restricts expression of Wnt-7a and Rfng to the dorsal epidermis.Wnt-7a in turns induces expression of Lmx-1 in the dorsal mesenchyme (see Reference 37).During regeneration of early stage limb buds in Xenopus, the epidermis similarly controls the dorsal-ventral expression pattern of Lmx-1 and the resultant DV limb pattern.Coincident with the ontogenetic loss of regenerative ability, the epidermis of later stage limb buds looses the ability to control Lmx-1 expression and dorsal-ventral pattern. 38It is unclear which genes expressed in the epidermis account for this function, since in Xenopus Wnt-7a and Rfng are uniformly expressed in both limb buds and regeneration blastemas. 39The role of the epidermis in pattern formation during urodele limb regeneration has yet to be investigated, though expression of Rfng in developing and regenerating newt limbs is reported to be uniform as in Xenopus. 40

Dedifferentiation of c2c12 myotubes in vitro
c2c12 cells behave as pluripotent mesenchymal precursor cells that can be induced to undergo myogenesis, adipiogenesis, chondrogenesis or osteogenesis in response to different growth and differentiation factors.Of particular significance to regeneration is the observation that when Msx-1 is expressed in c2c12 myotubes, these multinucleate cells fragment to give rise to mononucleate cells. 41A similar response is observed when myotubes derived from an urodele (newt) cell line is exposed to serum, 42 and both are being investigated as model systems for muscle 'dedifferentiation' and regeneration.Although the newt cells arrest in the cell cycle, the c2c12-derived mononucleate cells are proliferative and recapitulate the developmental potential of the parent cell line. 41[45][46][47] An extract of regenerating limb blastemas also induces formation of proliferative mononucleate cells from c2c12 myotubes in vitro, 48 indicating the presence of blastema factor(s) involved in the control of the differentiated state in a multipotent mesenchymal stem cell population.Given that Msx-1 is known to be causally linked to fragmentation, 41 this may indicate that there are factors in blastemas that regulate Msx expression.Several signaling molecules are known to have this ability, including members of the FGF, Wnt, and BMP signaling pathways, 32,[49][50][51][52][53][54] which are obvious candidates for further studies.

Induction of accessory limbs as a model to test for regeneration signals
6][57][58] In addition, each of the three phases of regeneration discussed above can be distinguished and studied independently (Figure 2).If a piece of skin (epidermis and dermis) is removed from the lateral surface of a limb, the wound heals and the skin is regenerated (phase I).If a nerve is deviated to the site of a lateral wound, a symmetrical outgrowth composed of undifferentiated cells is induced, but it does not continue to develop, and eventually regresses (phase II).Finally, if a piece of skin from the opposite side of the limb is grafted to a lateral wound along with a deviated nerve, a well-patterned accessory limb develops at the site of the wound (phase III).
Although the phenomenology of accessory limb induction has been well characterized, little is known about the cellular and molecular mechanisms controlling growth and pattern formation.Early genes (Msx-2 and Mmp-9) expressed in response to amputation (phase I) are also expressed in lateral wounds, 16,18 and thus are not dependent on a deviated nerve supply.Studies are in progress to study gene expression and the origin of the cells that form symmetrical 'bumps' in response to a deviated nerve (phase II).Fibroblasts from local connective tissues presumably are the source of cells, and likely are induced to proliferate by factors supplied by the deviated nerve.The final component required to form a limb is a source of fibroblasts with positional characteristics that are distinct from those of the fibroblasts at the host site (skin graft from the opposite side of the limb).Interactions between fibroblast-derived cells with disparate positional values stimulate continued growth and pattern formation, 20,21 leading to the development of the accessory limb.Phase III outgrowths express genes that are characteristic of late stage regenerating limbs (e.g.Dlx-3 and Hoxd-11).The accessory limb model offers the advantage of being able to test candidate signals in a positive regeneration response assay, and allows for the identification of the signals that are unique to each of the three phases of regeneration.

Future prospects for understanding regeneration
In nearly every review of amphibian limb regeneration, the authors seem compelled to speculate about the potential for inducing regeneration in humans.For many generations, regeneration biologists have expressed optimism for what has proven to be an illusive goal.It is possible that the opportunities for this generation of biologists are unique in that we are no longer limited to describing the phenomenology of regeneration, but can now alter the biological processes directly.In addition there are techniques to discover all the genes involved in the processes of regeneration (e.g.][61] Given these technological advances, combined with the rich experimental history and the availability of good experimental models, there is reason to be encouraged that we may yet realize the long dreamed of goal of stimulating regeneration in humans.

Figure 1 .
Figure 1.The converging pathways of limb regeneration and limb development.

Figure 2 .
Figure 2. The induction of lateral limbs as a model system to study the three phases of limb regeneration.(a) Three types of lateral wounds can be generated.A piece of skin can be removed (type I), a nerve can be deviated to the site of the wound (type II), and a piece of skin from the opposite side of the limb can be grafted to the site of the wound/deviated nerve (type III).(b) Type II wounds form a symmetrical outgrowth that eventually regresses.(c-e) Type III wounds form an outgrowth that becomes asymmetrical (c) and continues to grow to form a normally patterned accessory limb (arrow in (d)).Type III outgrowths express genes that are characteristic of regenerating limbs (e).

Table 1 .
The three phases of limb regeneration