SEARCH

SEARCH BY CITATION

Keywords:

  • limb regeneration;
  • amphibian;
  • mechanisms of blastema formation and development

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. LOOKING PROXIMALLY: LIMB REGENERATION STUDIES FROM 1911–2011
  5. LOOKING DISTALLY: PERSPECTIVE FOR THE 21ST CENTURY
  6. NEW APPROACHES AND CHALLENGES
  7. TRANSLATION TO CLINICAL THERAPIES
  8. Acknowledgements
  9. REFERENCES

The experimental study of amphibian limb regeneration spans most of the 20th century and the first decade of the 21st century. We first review the major questions investigated over this time span: (1) the origin of regeneration blastema cells, the mechanism of tissue breakdown that liberates cells from their tissue organization to participate in blastema formation, (3) the mechanism of dedifferentiation of these cells, (4) how the blastema grows, (5) how the blastema is patterned to restore the missing limb structures, and (6) why adult anurans, birds and mammals do not have the regenerative powers of urodele salamanders. We then look forward in a perspective to discuss the many unanswered questions raised by investigations of the past century, what new approaches can be taken to answer them, and what the prospects are for translation of basic research on limb regeneration into clinical means to regenerate human appendages. Developmental Dynamics 240:943–968, 2011. © 2011 Wiley-Liss, Inc.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. LOOKING PROXIMALLY: LIMB REGENERATION STUDIES FROM 1911–2011
  5. LOOKING DISTALLY: PERSPECTIVE FOR THE 21ST CENTURY
  6. NEW APPROACHES AND CHALLENGES
  7. TRANSLATION TO CLINICAL THERAPIES
  8. Acknowledgements
  9. REFERENCES

The power to regenerate complex structures composed of multiple tissues is a long-standing dream of biomedical science. The ultimate goal is to chemically induce regeneration directly at the site of injury via combinations of soluble molecules and/or regeneration templates. One such complex structure that is a valuable and fascinating research model is the regenerating amphibian limb. Urodele amphibians, early anuran tadpoles, and some fish can regenerate their limbs and fins after amputation at any level (for review, see Tanaka, 2003; Brockes and Kumar, 2008). Limbs and fins are regenerated by the formation of a blastema resembling the early limb bud that grows and differentiates into a copy of the parts lost by amputation.

In urodeles, the cells that form the blastema appear to be derived primarily by a natural reprogramming of skeletal and muscle cells, Schwann cells, and fibroblasts to a less differentiated state (for review, see Stocum and Rao, 2010). The blastema cells first form an accumulation under the wound epidermis, which is thickened at its apex into a structure called the apical epidermal cap (AEC). The accumulation blastema then grows through a series of more or less distinct morphological stages (Iten and Bryant, 1973; Stocum, 1979), during which growth, patterning, and redifferentiation of the new structures takes place. The accumulation blastema (early bud) grows to a conical configuration called the medium bud (Fig. 1), which continues to elongate and change shape as the digits begin to form. Were we able to understand in detail the mechanisms of urodele limb regeneration, we might be able to replicate the process in mammalian appendages.

thumbnail image

Figure 1. Medium bud stage blastemas derived from (A) the distal third of the humerus of the spotted salamander larva, Ambystoma maculatum. B: The distal tip of the radius and ulna of the axolotl, Ambystoma mexicanum. Note the thickened apical wound epidermis (AEC) covering the tip of each blastema. The epidermis consists of keratinocytes and Leydig (gland) cells that are distinguished by their large size and lighter color. M, muscle. Photos by DL Stocum.

Download figure to PowerPoint

Regeneration of body parts has been studied for well over a century. Nineteenth century studies, summarized in T.H. Morgan's classic volume “Regeneration” (1901), were carried out primarily on invertebrates. Although Spallanzani first demonstrated the ability of newts to regenerate limbs in 1768, and anatomical and histological studies on urodele limb regeneration were carried out in the late 19th century, experimental studies of urodele limb regeneration have been underway only since about 1911; these studies are the focus of this review. The principal questions addressed over these past hundred years have been (1) the origin of blastema cells, (2) the mechanism of histolysis that liberates regenerative cells from their tissue organization, (3) the mechanism of dedifferentiation of these cells and their accumulation to form a blastema, (4) the mechanisms of blastema growth, (5) the models and mechanisms of blastema patterning, and (6) why adult anurans, birds, and mammals have lost the power to regenerate appendages. We first look proximally from the year 2011 to examine the major discoveries of the past century with regard to these questions, and then look distally for a perspective on the future that includes unresolved issues and how they might be resolved, new concepts and approaches, technical challenges, and the challenge of translating basic regenerative science to the clinic.

LOOKING PROXIMALLY: LIMB REGENERATION STUDIES FROM 1911–2011

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. LOOKING PROXIMALLY: LIMB REGENERATION STUDIES FROM 1911–2011
  5. LOOKING DISTALLY: PERSPECTIVE FOR THE 21ST CENTURY
  6. NEW APPROACHES AND CHALLENGES
  7. TRANSLATION TO CLINICAL THERAPIES
  8. Acknowledgements
  9. REFERENCES

Origin of Blastema Cells

In 1911, histological studies by Fritsch indicated that new tissues of a regenerating urodele limb were formed by a bud of undifferentiated cells, the blastema, rather than by a process of direct outgrowth of stump tissues as had been previously believed (Barfurth, 1891; Fraisse, 1911). The origin of these cells was controversial, some arguing that they were derived from limb tissues and others from the blood via the bone marrow. G. Hertwig (1927) settled this question in favor of a limb origin by grafting haploid T. taeniatus limb buds to diploid hosts, and amputating the developed limbs. The cells of the regenerates were always of donor ploidy, indicating their origin from limb tissues.

The next question was whether the blastema cells were derived from tissues local to the amputation surface, or from throughout the limb. To resolve this issue, Butler and O'Brien (1942) used the then new tool of X-irradiation to inhibit the ability of cells to divide. They irradiated a short segment of Ambystoma maculatum limb while shielding the rest of the animal, and vice versa. Amputation through the test limb segment resulted in regeneration when the segment was shielded, but not when unshielded, demonstrating that limb tissues local to the amputation site were the source of the blastema cells.

The third question was whether blastema cells were the progeny of reserve (stem) cells or of mature cells that underwent dedifferentiation. Careful histological studies by Thornton (1938) suggested that the blastema cells were derived by the dedifferentiation of cells released from their tissue organization by histolysis. In the late 1950s, the use of another new tool, the electron microscope, lent support to this view. Regenerating limbs sectioned at different time intervals after amputation appeared to show mononucleate muscle fragments separating from the cut ends of myofibers and the disappearance of contractile filaments in the fragments (Hay, 1959).

Still, there was the possibility that these mononucleate fragments were actually reserve cells being added to myofibers to regenerate the muscle. What was needed to strengthen the evidence for dedifferentiation was a way to follow marked cells after amputation. This was accomplished by grafting tissues marked by 3H-thymidine, ploidy, fluorescent dextran or dye, and transfected gene constructs to host limbs, followed by amputation (Table 1). Each of the mesodermal tissues of the limb contributed to the blastemal mesenchyme, indicating that cells of these tissues undergo dedifferentiation. By contrast, the epidermis did not contribute to the mesenchyme of the blastema (Riddiford, 1960).

Table 1. Urodele Limb Tissues Marked by Various Techniques, Grafted in Place of the Corresponding Unmarked Limb Tissues, and Tracked After Amputation to Determine Whether They Contribute to the Dedifferentiated Cell Population of the Regeneration Blastemaa
Grafted tissueMarkerForms blastema cells?Author
  • a

    All except epidermis contribute to the dedifferentiated cell population of the regeneration blastema. Cartilage, muscle, Schwann cells, and epidermis regenerate true to their parent tissue, but dermal fibroblasts have a limited plasticity. 3N, triploid; 1-nu, haploid; 3H-T, tritiated thymidine; RD, rhodamine dextran; PKH 26, red fluorescent lipophilic dye; ALP, alkaline phosphatase rertroviral gene construct; GFP, green fluorescent protein.

Cartilage3N+Steen (1968)
3N+Namenwirth (1974)
 GFP+Kragl et al. (2009)
Dermis3N+Dunis and Namenwirth (1977)
 GFP+Kragl et al. (2009)
Epidermis3H-TRiddiford (1960)
MyotubesRD+Lo et al. (1993)
 PKH26+Kumar et al. (2000)
 ALP+Kumar et al. (2000)
Myofibers (tail)RD+Echeverri et al. (2001)
Myofibers (limb)1-nu+Steen (1973)
 GFP+Kragl et al. (2009)
Schwann cellsGFP+Kragl et al. (2009)

An experiment by Cameron and Hinterberger (1984) had suggested that tissues of the limb stump give rise to the same tissues in the regenerate. They separated axolotl medium bud blastemas into peripheral, core, and distal regions, and cultured them in vitro. Core and distal blastema cells differentiated as chondrocytes, whereas blastema cells from the periphery differentiated as muscle. A major technical breakthrough in urodele cell marking was achieved with the production of axolotls transgenic for green fluorescent protein (GFP) (Sobkow et al., 2006). Kragl et al. (2009) grafted a variety of tissues from the limbs of GFP axolotls in place of their counterparts in white axolotls (Table 1). Their results showed conclusively that not only did cells expressing GFP contribute to the blastema, i.e., underwent dedifferentiation, but also that dedifferentiated cells retained a memory of their parent cell type and redifferentiated into like cells. The one exception was dermal fibroblasts, which also differentiated as cartilage, confirming previous observations (Steen, 1968; Dunis and Namenwirth, 1977). The cartilage of the regenerate is thus derived from two sources, chondrocytes and fibroblasts. In fact, triploid-marking experiments have shown that dermal fibroblasts contribute nearly half the blastema cells of the amputated axolotl limb, and contribute the majority of the regenerated chondrocytes (Muneoka et al., 1986a).

However, the possibility remained that stem cells also contributed to the regeneration blastema. In 1961, Mauro discovered the satellite cells of muscle in frogs. He proposed that these are stem cells that regenerate injured muscle, a proposition since verified many times over in multiple species. Satellite cells were reported in newt limb muscles over 20 years ago (Cameron et al., 1986). Subsequently, by using the cell-specific transcription factor Pax-7 as a marker, satellite cells were shown to contribute to the limb regeneration blastema and develop into new muscle (Morrison et al., 2006). Thus the muscle of the regenerate appears to originate from two sources in the parent muscle, dedifferentiated mononucleate cells derived by the cellularization of myofibers and satellite cells. Whether other stem cell populations of the limb such as the mesenchymal stem cells (MSCs) of the periosteum contribute to the blastema is unknown.

Mechanisms of Histolysis

The cells that form the blastema, whether they are stem cells or are derived by dedifferentiation, are released from their tissue organization by degradation of the extracellular matrix (ECM), a process called histolysis. All of the tissues subjacent to the wound epidermis undergo intense histolysis for a distance of 1–2 mm, resulting in the liberation of individual dermal fibroblasts, Schwann cells of the peripheral nerves, and skeletal cells from their ECM. Myofibers cellularize while simultaneously releasing satellite cells during this process. The liberated cells undergo dedifferentiation to mesenchyme-like cells with large nuclei and sparse cytoplasm that exhibit intense DNA, RNA, and protein synthesis (Bodemer, 1962; Hay and Fischman, 1961; Anton, 1965). Histolysis and dedifferentiation are visible histologically within 2–3 days post-amputation in larval urodeles and within 4–5 days in adults (Hay and Fischman, 1961; Thornton, 1968).

Degradation of tissue ECM is achieved by acid hydrolases and matrix metalloproteinases (MMPs). Acid hydrolases identified in regenerating urodele limbs include cathepsin D, acid phosphatase, β-glucuronidase, carboxyl ester hydrolases, and N-acetyl-glucoaminidase (Schmidt, 1966, 1968 for reviews; Rivera et al., 1981; Ju and Kim, 1998; Park and Kim, 1999). Osteoclasts degrade bone matrix via hydrochloric acid, acid hydrolases, and MMPs. Up-regulated MMP transcripts include MMP-2 and −9 (gelatinases), and MMP- 3/10a and b (stromelysins) (Grillo et al., 1968; Dresden and Gross, 1970; Yang and Bryant, 1994; Miyazaki et al., 1996; Ju and Kim, 1998; Yang et al., 1999; Park and Kim, 1999; Kato et al., 2003; Vinarsky et al., 2005). In the newt limb, the basal layer of the wound epidermis transcribes MMP3/10a and b, as well as a novel MMP with low homology to the others (Kato et al., 2003). Chondrocytes express MMP-2 and -9 transcripts in the newt limb, and these enzymes are proposed to diffuse outward from the degrading skeletal elements (Kato et al., 2003). An important function of the MMPs encoded by the basal layer of the wound epidermis is thought to be the prevention of basement membrane re-assembly beneath it, thus maintaining communication between the wound epidermis and subjacent blastema cells. Loss of such communication, either by removing the wound epidermis (Stocum and Dearlove, 1972; Mescher, 1976) or conditions under which a pad of connective tissue becomes prematurely interposed between wound epidermis and blastema cells (Stocum and Crawford, 1987), inhibits regeneration. These MMPs might also diffuse into the underlying tissues to participate in the degradation of other ECM components.

The expression patterns of MMP-2, 3, 8, 9, 10, and 13 proteins have been examined by antibody array at different time points during blastema formation in wild-type axolotls, short-toe axolotls (a regeneration-deficient mutant, see Sato and Chernoff, 2007), and Xenopus froglets (also regeneration-deficient; Dent, 1962). The regeneration-deficient limbs exhibit different enzyme levels and temporal patterns of enzyme expression than the regeneration-competent wild-type axolotl limb (Santosh et al., 2011), suggesting that these differences play a role in the abnormal histolysis and thus availability of cells for dedifferentiation noted in regeneration-deficient limbs (Wolfe et al., 2000). The importance of MMPs to regeneration was underscored by the failure of blastema formation in amputated newt limbs treated with the MMP inhibitor GM6001 (Vinarsky et al., 2005).

Histolysis continues through the medium bud (cone) stage of blastema growth, and then ceases due to the activity of tissue inhibitors of metalloproteinases (TIMPS) (Stevenson and Vinarsky, 2006). TIMP1 is up-regulated during histolysis when MMPs are at maximum levels, and exhibits spatial patterns of expression congruent with those of MMPs in the wound epidermis, proximal epidermis, and internal tissues undergoing disorganization.

Mechanisms of Dedifferentiation

Dedifferentiation involves a nuclear reprogramming of limb cells that alters their global pattern of transcriptional activity to effect a less differentiated state. The activity of differentiation genes is suppressed and genes associated with stemness are activated (Geraudie and Ferretti, 1998; Gardiner and Bryant, 1996, 2007; Endo et al., 2000). Carlson (1969) showed that inhibition of this transcriptional shift by actinomycin D does not affect histolysis, but does prevent or retard dedifferentiation, leading to regenerative failure or delay. This suggests that at least some of the protease expression involved in histolysis is not regulated at the transcriptional level, but that proteins effecting dedifferentiation are so regulated.

Stemness genes up-regulated during blastema formation are msx1 (Crews et al., 1995; Koshiba et al., 1998; Schnapp and Tanaka, 2005), nrad (Shimizu-Nishikawa et al., 2001), rfrng and notch (Cadinouche et al., 1999). Msx1 inhibits myogenesis (Song et al., 1992) and its forced expression in mouse C2C12 myotubes causes cellularization and reduced expression of muscle regulatory proteins (Odelberg et al., 2001). Inhibition of msx-1 expression in cultured larval axolotl myofibers by anti-msx-1 morpholinos prevents their cellularization (Kumar et al., 2004). However, Schnapp and Tanaka (2005) found no effect of anti-Msx-1 morpholinos on the number of myofibers that dedifferentiated in amputated tails of larval axolotls. Thus, the role of msx-1 in dedifferentiation remains unclear. Nrad expression is correlated with muscle dedifferentiation (Shimizu-Nishikawa et al.,, 2001), and Notch is a major mediator of stem cell self-renewal (Lundkvist and Lendahl, 2001). Dedifferentiated cells express a more limb bud–like ECM in which type II collagen synthesis and accumulation are reduced, collagen I synthesis is maintained at a steady level, and fibronectin, tenascin and hyaluronate accumulate (for reviews, see Linsenmayer and Smith, 1976; Gulati et al., 1983; Onda et al., 1991; Mescher and Cox, 1988; Stocum, 1995).

Mammalian adult fibroblasts have been reprogrammed to pluripotency (induced pluripotent stem cells, iPSCs) equivalent to that of embryonic stem cells (ESCs) by transfecting them with four of six transcription factor genes (Oct 4, Sox 2, c-myc, Klf-4; Takahashi et al., 2007) and Oct 4, Sox 2, Nanog, and Lin 28 (Yu et al., 2007). Three of these six genes (klf4, Sox2, c-myc), are up-regulated during blastema formation in regenerating newt limbs, and also during newt lens regeneration (Maki et al., 2009). Up-regulation of the Lin 28 protein was detected during blastema formation in regenerating axolotl limbs (Rao et al., 2009). Thus, some of the transcription factors that reprogram fibroblasts to iPSCs may be common effectors of the nuclear reprogramming that ensues during blastema formation. Other factors, however, must be in play to insure that dedifferentiated cells reverse their transcription programs only far enough to attain a state that is responsive to proliferation and patterning signals, while for the most part maintaining a memory of their parent cells. Reprogramming also involves changes in epigenetic marks such as methylation and acetylation, as well as micro RNAs. Studies on such changes have begun for newt lens regeneration (Maki et al., 2010; Nakamura et al., 2010) but have not yet been reported for limb regeneration beyond the observation that the long-range limb-specific shh enhancer, a conserved sequence called mammals-fishes-conserved-sequence 1 (MFCS1), which is located in a non-coding region of the LMBR1 gene (Sagai et al., 2004), is hypermethylated in Xenopus versus moderately methylated in the axolotl and newt (Yakushiji et al., 2007). This hypermethylation is associated with the lack of shh expression and formation of a cartilage rod by the fibroblastema of the amputated Xenopus limb (Endo et al., 2000; Yakushiji et al., 2007), in contrast to the expression of shh in the posterior tissue of the urodele limb regeneration blastema (Imokawa and Yoshizato, 1997, 1998; Endo et al., 1997; Torok et al., 1999).

The early wound epidermis has an important function in generating the initial signals for limb regeneration. Na+ influx in the amputated newt limb and H+ efflux in the amputated tail of Xenopus tadpoles generate ionic currents across the wound epidermis essential for regeneration. Na+ influx is via sodium channels (Borgens et al., 1977). H+ efflux in the amputated tail is driven by a plasma membrane ATPase in the epidermal cells (Adams et al., 2007), and is likely to be important for limb regeneration as well, given that a gene encoding a v-ATPase was the most abundant clone in a suppressive subtraction cDNA library made from dedifferentiating axolotl limb tissue (Gorsic et al., 2008). These ion movements are obligatory for regeneration, since drug-induced inhibition of either Na+ or H+ movements during the first 24 hr or so after amputation results in failure of blastema formation (Jenkins et al., 1996; Adams et al., 2007).

Two early regeneration signals that may be triggered by ion flux are nitric oxide (NO) and inositol triphosphate (IP3). The enzyme that catalyzes NO synthesis, nitric oxide synthase 1 (NOS1), is strongly upregulated in the wound epidermis of amputated axolotl limbs at 1 day post-amputation (Rao et al., 2009). NO has a wide variety of signaling functions (Lowenstein and Snyder, 1992). It is produced by macrophages and neutrophils as a bactericidal agent, and has a role in activating proteases known to be important effectors of histolysis in regenerating limbs. IP3 and diacylglycerol (DAG) are the products of PIP2, which in turn is derived from inositol. IP3 synthase, a key enzyme for the synthesis of inositol from glucose-6-phosphate, is upregulated during blastema formation in regenerating axolotl limbs (Rao et al., 2009). IP3 stimulates a rise in cytosolic Ca2+ that results in the localization of protein kinase C (PKC) to the plasma membrane, where it is activated by DAG and regulates transcription (Lodish et al., 2008). During blastema formation, there is a general down-regulation of proteins involved in Ca2+ homeostasis, which suggests that IP3 might signal a rise in cytosolic Ca2+ in regenerating limbs by this mechanism (Rao et al., 2009). Other studies have shown that IP3 is generated from PIP2 within 30 sec after amputation in newt limbs (Tsonis et al., 1991) and that PKC rises to a peak by the accumulation blastema stage (Oudkhir et al., 1989). These changes, and their relationship to histolysis and dedifferentiation, represent topics of interest for future investigation.

Mechanisms of Blastema Cell Accumulation

The wound epidermis thickens at the apex of the limb within a few days after amputation to form the AEC. The AEC directs the migration of blastema cells to form the accumulation blastema beneath it. This was shown by experiments in which shifting the position of the AEC laterally caused a corresponding shift in blastema cell accumulation (Thornton, 1960), and transplantation of an additional AEC to the base of the blastema resulted in supernumerary blastema formation (Thornton and Thornton, 1965). Nerves that innervate the AEC do not appear to be involved in physically guiding blastema cells, since similar experiments on aneurogenic limbs also resulted in eccentric blastema formation (Thornton and Steen, 1962). The redirected accumulation of blastema cells under an eccentric AEC may be due to the migration of the cells on repositioned adhesive substrates produced by the AEC. TGF-β1 is strongly upregulated during blastema formation in amputated axolotl limbs (Levesque et al., 2007). A target gene of TGF-β1 is fibronectin, a substrate molecule for cell migration that is highly expressed by basal cells of the wound epidermis during blastema formation (Christensen and Tassava, 2000; Rao et al., 2009). Inhibition of TGF-β1 expression by the inhibitor of SMAD phosphorylation SB-431542 reduces fibronectin expression and results in failure of blastema formation (Levesque et al., 2007), suggesting that fibronectin produced by the AEC provides directional guidance for blastema cells.

Blastema Cell Proliferation

The wound epidermis and regenerating nerves play crucial roles in blastema cell proliferation. The wound epidermis is invaded by sprouting sensory axons within 2–3 days after amputation, while motor axons make intimate contact with mesenchyme cells (presumably prospective muscle) as the blastema forms (Salpeter, 1965; Lentz, 1967). Blastema cells enter the cell cycle and synthesize DNA, but the level of mitosis during blastema formation is very low (Kelly and Tassava, 1973; Tassava et al., 1974; Mescher and Tassava, 1975; Tassava and Mescher, 1975; Tassava and McCullough, 1978; Maden, 1978a; Tassava and Garling, 1979).

The signals that drive re-entry into the cell cycle have been studied in vitro in myotubes derived from a newt limb regeneration blastema cell line (Tanaka et al., 1997, 1999; Straube and Tanaka, 2006). Cell cycle re-entry in cultured newt and mouse myoblasts and newt myotubes is promoted by a thrombin-activated factor present in the serum of all vertebrates tested thus far, which inactivates the retinoblastoma protein (pRb). Mouse C2C12 myotubes do not respond to this factor, even though it is present in mouse serum. Although the thrombin-activated protein is both necessary and sufficient to stimulate myonuclear entry into the cell cycle, it is not sufficient to drive them through mitosis and myonuclei arrest in G2 phase. Mitosis appears to require the cellularization of myotubes into mononucleate cells. Cell cycle re-entry is independent of myotube cellularization, since cell cycle–inhibited myotubes implanted into newt limb blastemas cellularize (Velloso et al., 2000). The mechanism of myotube or myofiber fragmentation into single cells is not known, nor is it known whether the thrombin-activated protein is also necessary to drive mononucleate cells such as chondrocytes and fibroblasts into the cell cycle as well, or whether this is a feature unique to myofibers.

Once the accumulation blastema has formed, mitosis increases 10-fold or more (Mescher and Tassava, 1975; Loyd and Tassava, 1980). Numerous experiments over many decades have shown that both wound epidermis and nerves are required for blastema cell mitosis. Denervation leads to cessation of blastema cell mitosis at any stage of blastema growth and differentiation (Schotte and Butler, 1944; Tassava and Bennett, 1974; Maden, 1978a; Goldhamer and Tassava, 1987). In a remarkable series of experiments, Singer showed that the neural requirement is quantitative, that is, dependent on axon number, not sensory or motor quality (summarized in Singer, 1952).

Blastema formation and growth are also inhibited in the absence of the AEC (Goss 1956a, b; Mescher, 1976; Tassava and Garling, 1979; Loyd and Tassava, 1980). Well-established blastemas of larval A. maculatum denuded of wound epidermis and grafted into dorsal fin tunnels (where they would presumably become innervated by axons of the fin connective tissue) form skeletal elements that are subnormal in size, suggesting that cell proliferation is reduced or halted in the absence of the AEC (Stocum and Dearlove, 1972). Furthermore, DNA synthesis and mitosis of epidermis-free Notophthalmus viridescens limb blastemas cultured in the presence of dorsal root ganglia are reduced 3–4-fold (Globus et al., 1980; Smith and Globus, 1989). A major difference in the results of denervation and epidermal deprivation in vivo, however, is that denervated blastemas form a complete set of limb parts (Singer and Craven, 1948; Powell, 1969), whereas blastemas grafted to fin tunnels are distally truncated, suggesting that the AEC has a role in proximodistal patterning in addition to proliferation (Stocum and Dearlove, 1972).

The DNA labeling index of denervated and epidermis-free limb stumps was the same as controls during formation of the accumulation blastema. The mitotic index of epidermis-free limbs was also the same as controls, though the mitotic index was very low. The mitotic index of denervated limbs, however, was zero. The difference in mitotic index between denervated and epidermis-free limbs was interpreted to mean that nerve and wound epidermis played different roles in the blastema cell cycle, with the epidermis maintaining the cells in a dedifferentiated state allowing DNA synthesis and the nerve providing factors for mitosis (Tassava and Mescher, 1975). However, more recent investigations into the molecular factors driving blastema growth suggest a different type of relationship between the wound epidermis and the nerve in which the epidermis supplies the blastema cells with mitogenic factors, but requires continual stimulation by nerve axons to maintain this function (Fig. 2). This relationship explains why a blastema fails to form properly if either the wound epidermis or the nerve is eliminated just after amputation, and why both nerve and AEC are required for mitosis throughout blastemal growth.

thumbnail image

Figure 2. Relationship between nerves (yellow), AEC (green) and blastema cells (red) in the growing blastema (based on data of Kumar et al., 2007). Nerve axons stimulate the AEC to express AGP, which diffuses into the blastema and promotes blastema cell proliferation by binding to its ligand, Prod1.

Download figure to PowerPoint

The crucial experiment leading to this idea was the demonstration that a single protein, the anterior gradient protein (AGP), can substitute for the nerves in regenerating newt limbs (Kumar et al., 2007). AGP is a ligand for the blastema cell surface protein Prod1, a member of the Ly6 family of three-finger proteins anchored to the cell surface by a glycosylphosphatidyl inositol (GPI) linkage (Morais da Silva et al., 2002; Brockes and Kumar, 2008; Garza-Garcia et al., 2009). AGP is strongly expressed in the distal-most Schwann cells of regenerating newt limbs at 5 and 8 dpa, when initial dedifferentiation is underway (Kumar et al., 2007). This expression is abolished by proximal nerve transection, indicating that it is induced in the Schwann cells by axons. By 10 days post-amputation, when the accumulation blastema is beginning to form, AGP expression shifts from Schwann cells to secretory sub-epidermal gland cells of the AEC (Kumar et al., 2007). The axolotl does not have sub-epidermal gland cells and here AGP expression is observed in the Leydig cells of the AEC. Both sets of gland cells appear to discharge secretions by a holocrine mechanism (Kumar et al., 2010). The expression of AGP by gland cells is nerve-dependent, and is abolished in denervated limbs. When electroporated into denervated newt limbs at 5 days post-amputation, the AGP gene supported regeneration to digit stages. Conditioned medium of Cos7 cells transfected with the AGP gene stimulated BrdU incorporation into cultured blastema cells, and antibodies to Prod1 blocked this incorporation (Kumar et al., 2007). These results indicate that nerve axons induce the AEC to express AGP, which then acts through Prod1 on subjacent blastema cells to stimulate their proliferation.

Other evidence supports this idea. Firstly, the epidermis of a wound made in the skin of an axolotl limb develops a thickening comparable to the AEC of a regenerating limb. This thickening subsequently regresses. However, if a nerve is deviated into the wound, the thickening is maintained and a blastema-like growth is formed (Endo et al., 2004). This result implies that the AEC can form independently of the nerve, but that maintenance of AEC structure and function is nerve-dependent, an impliction that fits the timing of AEC formation and regeneration of axons into the AEC. Secondly, aneurogenic limb buds are AEC-dependent, but nerve-independent for their regeneration (Yntema, 1959a, b). However, their regeneration becomes nerve-dependent when they are allowed to be re-innervated (Thornton and Thornton, 1970). A similar shift from nerve-independence to dependence occurs as nerves invade the differentiating limb bud (Fekete and Brockes, 1987). These shifts again suggest an interaction between nerves and epidermis that renders the epidermis competent to promote mitosis.

Factors other than AGP that promote blastema cell proliferation in vitro have been detected in the wound epidermis of the regenerating limb, including Fgf-1, 2, and 8 (Chew and Cameron, 1983; Boilly et al., 1991; Zenjari et al., 1997; Han et al., 2001; Christensen et al., 2001, 2002; Dungan et al., 2002; Giampaoli et al., 2003). Likewise, nerves produce other mitogens including transferrin, Fgf-2, Ggf-2 (neuregulin), and substance P (Munaim and Mescher, 1986; Mescher and Kiffmeyer, 1992; Mescher et al., 1997; Globus and Alles, 1990; Anand et al., 1987; Mullen et al., 1996; Wang et al., 2000). Blastema cells express Fgf-10, which is essential for maintaining Fgf-2 expression by the AEC in regenerating Xenopus limb buds (Yokoyama et al., 2000; 2001). These factors may be synergistic to AGP in its effect on the wound epidermis and thus essential to blastema growth as well.

Igf-1 has been implicated in formation of the accumulation blastema (Fahmy and Sicard, 1998). Intraperitoneal injections of Igf-1 shortened the time required to form the accumulation blastema by amputated newt limbs. The role played by Igf-1 is unknown, but insulin itself was shown to be crucial to 3H-thymidine incorporation and mitosis by cone stage newt limb blastemas cultured in chemically defined media supplemented with 2 or 10% fetal calf serum (Vethamany-Globus et al., 1978; Kesik et al., 1986). Thus Igf-1 might have a role in DNA synthesis during blastema formation, and in mitosis at stages of blastema growth.

Mechanisms of Blastema Patterning

From the 1960s through the 1980s, the mechanisms whereby the tissue patterns of the amputated limb segments are specified by the blastema have arguably been the subject of more research on limb regeneration than any other. These mechanisms have been investigated primarily by grafting experiments that alter spatial relationships of differentiated limb tissues followed by amputation, or that alter the relationship of the blastema with respect to the limb stump.

Structural discontinuities can be filled in by intercalary regeneration.

The limb can be viewed as a three-dimensional “normal-neighbor” map of cells, each with a unique positional identity that marks its location on the anteroposterior (AP), dorsoventral (DV), and proximodistal (PD) axes (Cartesian coordinates) or its PD, angular, and radial location (Mittenthal, 1981). Only those structures distal to the amputation plane are ever regenerated (the “rule of distal transformation”). Experiments in which discontinuities were created in regenerating limbs revealed that cells sense the discontinuity and proliferate to eliminate it, a process called intercalary regeneration. When undifferentiated distal blastemas were autografted to more proximal stump levels, further development of the blastema was delayed while cells at the host proximal level dedifferentiated, whereupon growth and differentiation of the composite blastema resumed (Iten and Bryant, 1975; Stocum, 1975). Grafts of triploid blastemas to diploid limb stumps showed that the graft developed according to its origin while the dedifferentiated cells derived from the host proximal level intercalated the intermediate structures between graft and host levels (Pescitelli and Stocum, 1980) (Fig. 3).

thumbnail image

Figure 3. Intercalary regeneration in the PD axis of the axolotl limb (after Pescitelli and Stocum, 1980). A: A triploid (3N) wrist (W) blastema is heterografted to a diploid (2N) upper arm stump (S, stylopodium), discarding the intermediate structures. B: After a delay during which the host limb tissues dedifferentiate and contribute to the blastema, a normal limb is regenerated by the intercalation of the distal stylopodium (bright blue) and the zeugopodium (lower arm, green). Counts of triploid vs. diploid cells show that the 3N graft develops into carpals (C) and digits, whereas the intercalated intermediate structures are the host 2N ploidy.

Download figure to PowerPoint

Several kinds of experiments have shown that blastema cells sense and fill in discontinuities in the AP and DV axes. Removal of half the zeugopodium of a newt limb followed by amputation resulted in the regeneration of a half limb from the remaining half (Goss, 1957). By contrast, if the muscle and bone was removed from one half of the zeugopodium, leaving the skin intact, the limb regenerated normally after amputation, demonstrating that radial intercalation of positional identity had taken place from the dermis of the operated half (Fig. 4a). Radial intercalation from the dermis was also demonstrated by replacing the skin of an irradiated limb with normal skin. The limbs regenerated a complete skeleton and dermis (but not muscle) after amputation (Dunis and Namenwirth, 1977). Reversal of either the AP or DV axis by grafting the blastema to the contralateral limb, or reversing both axes by rotating the blastema 180° on its limb stump, confronts opposite axial poles. The discontinuities between the poles can be filled in by radial intercalation to create the base of a supernumerary blastema that grows out under the influence of the wound epidermis to form a supernumerary limb (Bryant and Iten, 1976). Supernumeraries also arise following amputation of limbs in which AP and DV discontinuities have been created by rotation of skin cuffs or cross-transplantation of muscles (Carlson, 1974, 1975a).

thumbnail image

Figure 4. A: Upper panel: The ulna and associated muscle (orange hatch) of the posterior half of a newt lower arm was removed while leaving the skin (green) intact (after Goss, 1957). After amputation, radial intercalation (arrows) restored the positional identities of the posterior half. Lower Panel: A half limb regenerates when all the tissues of the posterior half are removed because radial intercalation is not possible. B: Lheureux' experiment (1975). Upper panel: An unirradiated dorsal longitudinal skin strip was turned 90° and wrapped around the circumference of an irradiated limb (I, lightning bolts) at the level of the upper arm. After amputation through the strip, all blastema cells have the same positional identity and sense no discontinuity. Lower panel: Control experiment grafting rotated longitudinal strips of skin from each quadrant of the limb. Positional discontinuity is now sensed and distal transformation takes place. R, radius; S, humerus; d, dorsal; a, anterior; v, ventral; p, posterior.

Download figure to PowerPoint

Other experiments have shown that confrontation of cells from separate quadrants of the limb cross-section is necessary for distal transformation. Lheureux (1975) demonstrated this by rotating a longitudinal strip of limb skin from one quadrant 90° and grafting it around the circumference of an irradiated limb, followed by amputation through the strip. Such limbs failed to regenerate (Fig. 4b). However, when short longitudinal skin strips from three or four limb quadrants were rotated 90° and grafted to the circumference, regeneration took place normally. Furthermore, blastema-like accumulations induced by deviating nerves to limb skin wounds grew only if pieces of skin from opposite circumferential sites met to cover the wound (Endo et al., 2004).

Collectively, these results suggest that regeneration depends on the sensing, by dedifferentiated cells, of discontinuity in all dimensions of the positional identity map. Although experiments altering the spatial relationships of limb tissues are most conveniently described in terms of AP, DV, and PD axes, the results suggest that positional identity is composed of PD, circumferential (angular) and radial components, and that these components are inter-dependent. Without sufficient discontinuity in positional identity, blastema cells are satisfied that they are in contact with normal neighbors and will not respond to mitogens. While most investigators consider that the interacting positional identities are those of mesodermal blastema cells, Campbell and Crews (2008) have suggested that AEC cells have asymmetric positional identities that must interact as well.

Positional identity is expressed on the cell surface.

Positional identity of blastema cells, as might be expected, is expressed on the blastema cell surface. A Steinberg-type (Steinberg, 1978) in vitro assay in which proximal and distal blastemas were juxtaposed at their bases showed that the proximal blastema always tended to engulf the distal one, suggesting a distal (stronger) to proximal (weaker) gradient of cell adhesivity (Nardi and Stocum, 1983). The activity of this gradient was demonstrated in vivo by an “affinophoresis” assay in which blastemas from wrist, elbow, and mid-upper arm levels were grafted to the blastema-stump junction of hindlimbs regenerating from the mid-femur (Fig. 5). The wrist and elbow blastemas always moved to their corresponding level on the regenerating host blastema (ankle and knee, respectively) while the mid-upper arm blastema remained at its original level (Crawford and Stocum, 1988a; Egar, 1993). This sorting behavior was confirmed in experiments grafting marked cells from an early wrist blastema into the base of an early humeral blastema, where they sorted out to participate in hand formation (Echeverri and Tanaka, 2005). Genetic marking experiments showed that the adhesive differentials exist at the single cell level (Kragl et al., 2009). Position-dependent adhesion of cells in the developing and regenerating limb bud of the early Xenopus tadpole has also been demonstrated by the sorting out of distal and proximal cells from an initial homogeneous mixture (Ohgo et al., 2010).

thumbnail image

Figure 5. Right: “Affinophoresis” assay in vivo for blastema cell PD adhesive gradient (after Crawford and Stocum, 1988a). Medium bud stage blastemas derived from the wrist (W), elbow (E), or mid-upper arm (UA) were autografted to a wound bed made at the junction of the stump and a medium bud blastema derived by amputation (Amp) at the mid-femur (F) level of the hindlimb. Left: As the host hindlimb blastema grows and redifferentiates, the wrist and elbow blastemas sort distally to their corresponding tarsal and knee levels, respectively, while the mid-upper arm blastema remains at the amputation level.

Download figure to PowerPoint

Retinoic acid modifies positional identity in all three axes.

A major discovery, initiated by the work of Niazi (Niazi and Saxena, 1978; Sharma and Niazi, 1979; Niazi et al., 1979, 1985; Niazi, 1996) and extended by Maden (1982a, 1997; Keeble and Maden, 1984) and others (Thoms and Stocum, 1984) was that retinoic acid (RA) proximalizes the positional identity of blastema cells. Other experiments showed that RA also posteriorizes and ventralizes positional identity of blastema cells (Stocum and Thoms, 1984; Kim and Stocum, 1986a, b; Ludolph et al., 1990; Monkmeyer et al., 1992). Retinoic acid exerts its effects by binding to retinoic acid receptors (RARs) that activate or inhibit retinoic acid response elements in the promoters of target genes (Mendelsohn et al., 1992). The RAR that mediates proximalization is δ2 (Pecorino et al., 1996).

Retinoic acid changes the transcriptional program of blastema cells in a way that coordinately alters adhesiveness and positional identity. The distal movement of wrist blastemas proximalized by retinoic acid and grafted to the blastema-stump junction of regenerating hindlimbs is abolished (Fig. 6), as is intercalary regeneration after grafting a RA-treated wrist blastema to a more proximal stump level, thus demonstrating a link between RA target genes and the cell surface (Crawford and Stocum, 1988b). Prod-1, which was discovered in a screen of cDNA libraries from normal and RA-treated newt wrist level blastemas, forms a gradient that is opposite the gradient of adhesiveness from the distal to proximal aspect of the limb (Morais da Silva et al., 2002). Antibodies to Prod1, or its removal from the blastema cell surface by phosphatidylinositol-specific phospholipase C (PIPLC) inhibit the recognition of adhesive differentials between distal and proximal blastemas in the Nardi and Stocum (1983) in vitro engulfment assay (Morais da Silva et al., 2002). Furthermore, over-expression of Prod1 causes distal blastema cells to sort to a more proximal (less adhesive) position when grafted into proximal blastemas (Echeverri and Tanaka, 2005). These results suggest that Prod-1 plays a role in recognizing gaps in positional identity between non-neighboring cells that could stimulate cell dedifferentiation and mitosis in response to AGPand thus be a crucial link between blastema growth and patterning. Other surface molecules that may be involved in position-dependent adhesion of blastema cells are ephrins and cadherins. Antibodies to the EphA4 receptor and to N-cadherin, or cleaving of ephrin A ligands from the cell surface with phospholipase C, abolish the sorting of proximal and distal chick limb bud cells from one another (Wada et al., 1998; Yajima et al., 1999; Wada, 2011, pages 969–978, this issue). It would be instructive to investigate the expression and function of these molecules in the regenerating urodele limb.

thumbnail image

Figure 6. Coordinate proximalization of adhesive quality and positional identity of blastema cells (after Crawford and Stocum, 1988b). Right: RA-treated wrist and elbow blastemas (W, E) were grafted to the blastema-stump junction of a hindlimb regenerating from the mid-femur (F). Left: As the hindlimb blastema grows and redifferentiates, the proximalized wrist and elbow RA-treated blastemas develop from the upper arm level (UA) and remain at the site of grafting instead of sorting distally.

Download figure to PowerPoint

Genes associated with pattern specification.

Molecular data on blastema patterning is incomplete compared to that of the chick and mouse limb bud, but what is known generally conforms to what has been found in limb buds (Gardiner and Bryant, 1996). Forelimbs and hindlimbs express distinct T-box genes (Simon et al., 1997). Hoxa and d, and Meis genes are involved in PD patterning (Gardiner et al., 1995; Torok et al., 1998; Mercader et al., 2005). Hoxd 10 is expressed at a 2–3-fold higher level in proximal versus distal blastemas (Simon and Tabin, 1993; Simon et al., 1995, 1997). Hoxa 9 is expressed throughout both proximal and distal blastemas, but Hoxa 13 expression is 30% higher in distal versus proximal blastemas (Gardiner et al., 1995). Meis 1 and 2 are expressed preferentially in the stylopodial region of a blastema derived from the upper arm (Mercader et al., 2005). Hoxd 10 and Meis 1 and 2 expression is upregulated in distal blastemas proximalized by RA (Simon and Tabin, 1993; Mercader et al., 2005), while expression of Hoxa 13 is reduced (Gardiner et al., 1995). Furthermore, distal blastema cells are proximalized by overexpression of Meis 2 and relocate more proximally (Mercader et al., 2005). These observations implicate Hoxa 9, Hoxd 10, and Meis 2 in specifying the stylopodium, and Hoxa 9/13 in specifying the autopodium.

An important element in specifying anteroposterior (AP) pattern in chick limb buds is the zone of polarizing activity (ZPA) on the posterior edge of the bud (MacCabe et al., 1973). The ZPA induced twinning of wing tips after limb bud tip rotation (Fallon and Crosby, 1975). Cameron and Fallon (1977) used this technique to demonstrate that the posterior tissue of Xenopus tadpole limb buds has polarizing properties. The effector molecule of the ZPA is sonic hedgehog (Shh) (Riddle et al., 1993). Shh expression was demonstrated in the posterior blastema tissue of newt and Xenopus early tadpole limbs, as well as in posterior stump tissue following reversal of the blastema AP axis by contralateral grafting (Imokawa and Yoshizato, 1997, 1998; Endo et al., 1997; Torok et al., 1999). Furthermore, the transfection of Shh into anterior blastema tissue by vaccinia virus to posteriorize anterior cells and set up an AP confrontation resulted in the production of a supernumerary limb (Roy et al., 2000). Another protein prominently expressed along the AP axis of the developing and regenerating Xenopus tadpole limb is XlSALL4, a member of the spalt family (Neff et al., 2005; Neff et al., 2011, pages 979–989, this issue). In the regenerating limb buds of Xenopus tadpoles, the Lmx-1 gene is expressed in the dorsal mesenchyme and is involved in specifying DV axial polarity (Matsuda et al., 2001).

Models of pattern formation.

The 1970s and 80s saw the development of several models to explain the mechanisms of pattern formation in urodele limbs (for review, see Tank and Holder, 1982). The most prominent were the polar coordinate model (French et al., 1976; Bryant et al., 1981) and the boundary model (Meinhardt and Gierer, 1980; Meinhardt, 1982, 1983a, b). These models used existing data to explain (1) how the PD axis is restored after amputation and (2) the position and handedness of supernumerary limbs that arise either after axial reversal of the blastema on the limb stump and after rotation of skin cuffs or cross-transplantation of muscles followed by amputation through the manipulated region (Carlson, 1974, 1975a, b; Iten and Bryant, 1975).

Polar coordinate (PC) model

The PC model assigned angular and radial values to cells to describe their position in the cross-section of the limb. The original model (French et al., 1976) proposed that the radial positional values specified the PD axis, with the cells at the center of the cross-section representing the distal tip of the limb. However, the radial arrangement of PD positional values made the regenerate a two-dimensional cylinder, instead of a three-dimensional solid. Supernumerary limbs evoked by reversal of the AP and DV axes were postulated to arise by the intercalation of radial (PD) positional identities within the complete circumference created by juxtaposition of anterior and posterior or dorsal and ventral half circumferences. The model correctly predicted the origin of the supernumeraries at the axial poles and their stump handedness, and linked circumferential and PD positional information in the “complete circle rule.” This rule stated that a complete set of circumferential values must be present for distal transformation to occur.

The complete circle rule was tested immediately by experiments on half and double half limbs and found to be untenable, since half anterior or posterior zeugopodia regenerated a half anterior or posterior autopodium, and double half anterior or posterior zeugopodia regenerated double half anterior or posterior zeugopodia and autopodia (Stocum, 1978; Maden, 1979a, Wigmore, 1986). However, a double half anterior stylopodium regenerated only a symmetrical spike of cartilage, whereas a double half posterior stylopodium regenerated symmetrical double posterior zeugopodia and autopodia with up to six digits, suggesting an inequality of positional information in these halves of the stylopodium (Bryant, 1976; Stocum, 1978; Maden, 1979a; Holder et al., 1980; Krasner and Bryant, 1980; Wigmore and Holder, 1985). Subsequently, double half limb experiments showed that double dorsal stylopodia regenerate PD-complete symmetrical limbs, whereas double ventral stylopodia fail to regenerate (Ludolph et al., 1990).

The second version of the PC model (Bryant et al., 1981) modeled the regenerate as a three-dimensional solid and introduced a new idea for how distal transformation occurred (Fig. 7). Dedifferentiated cells from different points around the limb circumference were postulated to migrate toward the center of the wound surface and undergo circumferential and radial intercalation to form a new set of cross-sectional positional identities. The new set of cells generated in this way was assumed to adopt the next PD positional identity in the sequence, a process that would be reiterated until the PD axis was restored. The production of supernumerary limbs of stump handedness after AP or DV reversal was again explained by radial intercalation, but the radial intercalation now produced a complete limb cross-section that was distalized by successive rounds of circumferential/radial intercalation. No evidence was provided, however, that successive rounds of radial/circumferential interactions are sufficient, as well as necessary, to generate sequential PD identities, nor was any physical mechanism proposed by which newly reiterated sets of radial/circumferential cells adopted progressively more distal positional identities.

thumbnail image

Figure 7. Right: Basic features of the polar coordinate model (after Bryant et al., 1981). Cells from the circumferential arcs of all four quadrants of the amputated limb migrate centripetally (red arrows) and interact to generate circumferential and radial intercalation (yellow arrows), forming a complete cross-section of blastema cells that assume the next distal positional identity. Successive repetitions of this process regenerate the PD axis (dark blue arrow). Left: Upper: Dorsal view of a right limb stump (light blue) to which a left blastema (green) has been grafted, reversing the AP axis. Lower: View from the lateral side of the above construct showing how this graft confronts two half circumferences (orange brackets, DPV of the stump and DAV of the graft). The graft (primary limb) develops with its handedness of origin (blue arrow). A complete right-handed cross-section of blastema cells is generated by radial intercalation within the half circumferences. Repetitive circumferential and radial intercalation/distalization, as in the diagram at the right, regenerates the PD axis. A, anterior; P, posterior; D, dorsal; V, ventral.

Download figure to PowerPoint

APDV reversal produced two supernumeraries, one of stump handedness, the other of opposite handedness (Bryant and Iten, 1976). The supernumeraries arose at posterodorsal and anteroventral positions rather than at AP or DV axial poles. This was explained in the PC model by non-uniform spacing of circumferential positional identities, such that supernumerary complete circumferences could be generated by intercalation only at posterodorsal and anteroventral positions. At these loci, intercalation could theoretically take place over a long or short route. Since the short route seemed to explain the results, a “shortest intercalation rule” was postulated to explain position and handedness of the supernumeraries. The hypomorphic regeneration of double anterior stylopodia was explained as the result of a premature halt in distalization caused by the progressive convergence of positionally identical circumferential cells at the tip of the blastema, due to fewer positional values in that half. This convergence does not occur in double posterior stylopodia and they regenerate PD-complete double posterior limbs. The equality in number of positional values in double half anterior and posterior zeugopodia prevents convergence and they regenerate PD-complete symmetrical limbs.

Meanwhile, several other investigations had reported that up to three supernumeraries arise after APDV rotation over a wide range of angles (Wallace, 1978; Maden, 1978b; Wallace and Watson, 1979), though the maximum frequency of formation was at 180°, falling off on either side (Stock et al., 1980; Turner, 1981). Furthermore, the loci of origin of these limbs were unpredictable. Maden (1980), Maden and Mustafa (1982), and Papageorgiou and Holder (1983) examined the musculoskeletal structure of the supernumeraries at the carpal/metacarpal levels and found that they fell into four classes: anatomically normal (left or right), mirror-imaged (double dorsal or ventral), part normal/part mirror-imaged, and part normal/part inverted (mixed handed) limbs. The latter two categories are contrary to what is expected if discontinuities are eliminated strictly by intercalation, because discontinuities in the DV axis (but never in the AP axis) are clearly tolerated in these cases. Thoms and Fallon (1980) also obtained mixed handed limbs after rotation of stage 38+–40 urodele limb bud tips using the triploid marker to identify host and graft cell contributions. The second version of the PC model was thus incompatible with these results.

Boundary model

The boundary model (Meinhardt, 1983a, b) proposed that regeneration of an amputated limb requires the intersection of anterior/posterior and dorsal/ventral boundaries that define four compartments, with the A/P boundary being flanked by a ZPA (Fig. 8). In essence, this is the same as saying there must be interactions between cells from different quadrants of the limb for distal transformation to take place. The model proposes that these interactions result in the production of a morphogen that spreads out from the intersection point. However, whether such a morphogen exists or whether the ZPA actually has a role in the initial interactions that result in regeneration needs further investigation, because no intersect morphogen has been identified, and shh expression does not occur until the conical (medium bud) stage of regeneration (Imokawa and Yoshizato, 1997, 1998; Endo et al., 1997). Nevertheless, an analysis by Maden (1983) showed that the boundary model not only correctly predicted the handedness and position of supernumeraries regenerated after blastema AP and DV reversal, but most of the anatomical classes of supernumeraries created by APDV reversal, the hypomorphic regeneration of double anterior limbs, and the symmetrical, PD-complete regeneration of double posterior limbs.

thumbnail image

Figure 8. The boundary model (after Meinhardt, 1983b). The intersection of AP and DV boundaries (yellow lines) to form four compartments is necessary for distal transformation. A strip of polarizing tissue flanks the AP boundary. When the circumferences of the limb stump and axially reversed blastemas are “unrolled,” the cellular contributions of each compartment in the host and graft predict most of the known anatomical configurations of supernumerary limbs.

Download figure to PowerPoint

The boundary model proposed a physical mechanism for the generation of PD positional identity in which the AEC produces a morphogen (Meinhardt, 1983c). The higher the concentration of morphogen, the more distal is the specification of identity. The concentration of the morphogen is lowest at the earliest stages of blastema formation, so proximal identities will be specified first. The blastema cells produce an apical epidermal maintenance factor (AEMF; see Zwilling and Hansborough, 1956) that controls production of the PD morphogen by the AEC. More distally determined cells produce more AEMF, resulting in a higher concentration of PD morphogen. Progressively more distal PD positional identities are thus “bootstrapped” into existence by a stepwise increase in AEMF and PD morphogen. This idea has yet to be tested experimentally. The model accounts for intercalary regeneration after grafting a distal blastema to a more proximal limb level, as well as the failure of a proximal blastema grafted to a distal limb level to elicit intercalary regeneration (Stocum, 1975; Iten and Bryant, 1975; Stocum and Melton, 1977; Pescitelli and Stocum, 1980). In addition, it explains why a normal blastema induces distal regeneration from a symmetrical anterior limb stump (Stocum, 1981).

Holder and Weekes (1984) have pointed out that the prediction of symmetrical anatomies such as double dorsal and double ventral supernumeraries is actually problematic for the boundary model because these symmetrical regions appear to have no D/V boundary, whereas the model predicts that no outgrowth can occur unless such a boundary is present to intersect (interact) with an A/P boundary. Thus, the boundary model too has shortcomings, although it has better accounted for the data overall.

Mosaic model

The boundary model seems to reflect another idea that may better provide a unified explanation for the anatomy of supernumeraries after AP, DV, or APDV axial reversal. This idea supposes that the blastema is a mosaic of differing cellular contributions from each quadrant of the limb stump, a concept that first began to emerge from axial reversal experiments in which either host or graft tissue was irradiated (Maden, 1979b; Holder et al., 1979). Supernumerary formation occurred in these experiments, in which the extra limbs were always composed of unirradiated tissue and were incomplete, having mostly 2–3 posterior digits. This result suggested that posterior unirradiated tissue contributed all of the cells to the supernumeraries and that the supernumeraries could not have arisen by radial intercalation, otherwise they should have been complete. Tank (1981) noted the production of mixed handedness supernumeraries after APDV rotation and suggested that these limbs might be explained by assuming that the limb regenerates as a mosaic. Maden (1982b) and Maden and Mustafa (1982) demonstrated that all the classes of supernumeraries formed after APDV rotation are explainable by assuming that they are anatomical mosaics of host and graft tissues that reflect the polarities of these tissues with respect to one another at whatever positions the supernumeraries arise. Likewise, the results of double half anterior versus posterior limb regeneration are explainable by differential contributions of the limb quadrants to the regenerate, with posterior quadrants providing more cells than anterior quadrants. Thus the anatomical structure of supernumerary limbs and the regenerates of half and double half limbs would be the product of (1) which quadrants are contributing to the blastema, (2) the number of cells contributed by each of these quadrants during normal regeneration, and (3) the polarity of the contributions with respect to one another.

Evidence consistent with this interpretation has been provided by experiments with chimeric triploid and diploid limbs. Each half of axolotl zeugopodia that were surgically constructed from anterior and posterior halves of different ploidy contributed about half the cells of the regenerate after amputation (Muneoka et al., 1985). Supernumerary limbs regenerated after grafting AP reversed triploid blastemas to diploid stumps and vice versa were composed of approximately equal numbers of cells from graft and host (Muneoka and Bryant, 1984). Maden and Mustafa (1984) inverted triploid blastemas onto ipsilateral diploid limb stumps and vice versa, and analyzed the cellular contribution of stump and graft cells to the four classes of supernumeraries that resulted. They found different percentages of contribution from graft and stump for each class of supernumerary ranging from all stump to all blastema. Only one class did not fit this progression, double dorsal and ventral supernumeraries, which were much more variable.

However, mosaicism cannot explain all the experimental results, because amputation of surgically constructed mixed-handed zeugopodia (half normal, half inverted in the AP axis) resulted in regeneration of the same classes of anatomical pattern as after APDV rotation (Holder and Weekes, 1984; Muneoka et al., 1986b), not just mixed-handed limbs as might be expected. The use of the triploid marker in one of the halves of these constructs showed evidence of a directionally biased intercalation to smooth out discontinuities and also showed cell mixing (Muneoka et al., 1986), something that was noted by Thoms and Fallon (1980) as well in their urodele limb bud tip rotation experiments. These results suggest a model in which both mosaicism and intercalation are at work. Sorting of blastema cells (Kragl et al., 2009) might also contribute to the final anatomical and histological configuration of different classes of regenerates. An interesting question in this regard is why wrist and elbow blastemas grafted to the stump/blastema junction of a regenerating hindlimb (Crawford and Stocum, 1988a) do not evoke intercalary regeneration of a supernumerary limb, but instead sort to their normal position on the host regenerate. This result suggests that sorting might be preferred to intercalation when cells are given a choice of mechanism to re-establish normal neighbors.

Without further data to distinguish different possible mechanisms of blastema construction, model building to explain blastema patterning could go no further and was essentially abandoned in the late 1980s in favor of collecting molecular data. With the compilation of this data, and the advent of newer marking techniques, the time may be ripe for a renewed attempt at examining blastema cell behavior to explain the mechanisms by which tissue limb patterns are regenerated.

The blastema is self-organizing.

A major research issue from the 1920s on has been whether the blastema is a nullipotent mass of cells whose development depended on gradients of signals diffusing from parent differentiated tissues or whether it was a self-organizing structure that contained all the information required for its development, or a combination of both. To answer this question, cone stage stylopodial blastemas were grafted to ectopic locations (for review, see Stocum, 1984). Such grafts were either reported to regress, remain undeveloped, or develop into autopodial structures only. Older blastemas developed according to origin and with a complete PD array of limb structures.

Faber (1960) fate-mapped cone stage axolotl blastemas derived from the stylopodium by placing carbon marks at different PD levels of the blastema and then observing where the marks ended up in the regenerated limb. He found that cells representing the prospective stylopodium, zeugopodium, and autopodium were all present in serial order. Similar mapping experiments using GFP or DsRed plasmids confirmed that distal, intermediate, and proximal cells of cone stage stylopodial blastemas corresponded to the prospective autopodium, zeugopodium, and stylopodium, respectively (Echeverri and Tanaka, 2005).

To examine the self-differentiation potential of axolotl cone stage stylopodial blastemas, Faber (1960) placed a carbon mark in the prospective zeugoppodial region and grafted the blastemas to the back. The grafts formed only digits. In the majority of cases, the mark was found in the back tissues, suggesting resorption of prospective stylopodial and zeugopodial cells, leaving only autopodial cells. In a minority of cases, however, the mark was found in the digits. This led Faber to conclude that the conical blastema had autopodial differentiation tendencies, thus accounting for the fact that the grafts formed only hand parts. Stylopodial and zeugopodial qualities would be induced in their prospective regions by inductive signals from the stump. It is likely, however, that in the few cases where the mark wound up in the digits, it had been mis-placed in the prospective digital mesenchyme. In fact, when resorption was countered by grafting multiple blastemas, structures proximal to the autopodium developed at much higher frequency (DeBoth, 1965, 1970; Michael and Faber, 1961, 1971).

These ectopic grafting experiments were carried out on large, slowly regenerating animals. When the rapidly developing early cone-stage stylopodial blastemas of young A. maculatum larvae were grafted onto wound beds made in the skin of the dorsal fin, they were able to self-organize all of the limb segments distal to the plane of amputation in nearly 70% of the cases (Stocum, 1968). Self-organization has been confirmed in many other experiments involving the exchange of undifferentiated forelimb and hind limb blastemas of axolotls (Holder and Tank, 1979; Stocum, 1980, 1982). The grafted blastemas developed according to origin with a full array of PD structures. To test whether freshly dedifferentiated blastema cells could self-organize, proximal halves of upper arm blastemas undergoing early redifferentiation were grafted to the tarsus of the hindlimb (Stocum and Melton, 1977). The cells underwent complete dedifferentiation in the presence of distal hindlimb tissues, but self-organized into all of the arm structures distal to the amputation plane. These results indicate that the information for self-organization is set up very early during the phase of blastema formation.

Whole-mount studies on the expression of Hoxa 9 and 13 in regenerating axolotl limbs have again raised the possibility that autopodial cells are the first to be specified after a stylopodial amputation (Gardiner et al., 1995). Uniform expression of both genes was reported during formation of the accumulation blastema. By the cone stage, expression of Hoxa 13 was restricted to a distal domain corresponding to the prospective autopodium, while the prospective zeugopodial and stylopodial regions expressed Hoxa 9 only. The Meis 1 and 2 genes are also expressed at a high level in the prospective stylopodium of cone-stage blastemas derived from the upper arm (Mercader et al., 2005). Assuming that the Hoxa 9/13 combination is pivotal in specifying an autopodial domain and Hoxa 9 and Meis in specifying a stylopodial domain, how can we reconcile the uniform expression of both Hoxa 9 and 13 in the accumulation blastema with complete PD self-organization (we do not know if Meis genes are expressed earlier than cone stage), and how are these domains separated?

Figure 9 illustrates two alternative mechanisms of domain separation. The first postulates that each dedifferentiated cell of the accumulation blastema is expressing Hoxa 9/13 and that expression of Hoxa 13 requires factors secreted by the AEC. As the blastema grows, its more proximal cells are left out of range of these factors and these cells cease expression of Hoxa 13, creating a boundary between prospective autopodium and zeugopodium/stylopodium. It is likely that a third boundary, separating prospective stylopodium and zeugopodium, is also established and that transcription factors such as Hoxa11 may specify this boundary (see Yakushiji et al., 2009; Tamura et al., 2010; Ohgo et al., 2010). Resorption of proximal blastema cells or failure of the accumulation blastema to grow would prevent separation of autopodial and more proximal domains and lead to formation of hand parts only. Thus, the information for complete self-organization is present in the early blastema, but blastema growth is required to express it to the fullest extent. Alternatively, there may be two subsets of cells at the accumulation blastema stage uniformly mixed with one another, one expressing Hoxa 9, the other expressing Hoxa 9/13, and these sort out from each other via differential adhesion to form the autopodial and zeugopodia/stylopodial domains. Examining the expression profile of Hoxa 9, 13, and Meis genes at the level of individual blastema cells should give clues as to which alternative is correct.

thumbnail image

Figure 9. Models that link blastema self-organization and expression of Hoxa9/13, which is assumed to be a code for the prospective autopodium (after Stocum, 2006). A:Hoxa9/13 are expressed uniformly in all the cells of the accumulation blastema. Hoxa9 expression is autonomous, but that of Hoxa13 requires factors (arrow) from the AEC. As the blastema grows, its more proximal cells fall outside the range of the AEC factors and express only Hoxa9, the code for the prospective stylopodium. This separates stylopodial and autopodial domains. The code for the prospective zeugopodium is intercalated at the boundary between autopodium and styopodium. B:Hoxa9 and Hoxa9/13 are expressed by separate subsets of cells at accumulation blastema. Hoxa9 and Hoxa9/13 cells have a differential homotypic adhesive affinity and Hoxa9/13 cells have a stronger heterotypic affinity for the AEC. Prospective autopodial and stylopodial domains separated by cell sorting.

Download figure to PowerPoint

In either model, the proximal boundary of what is to be regenerated is set by the inability of blastema cells to adopt positional identities proximal to the Hoxa 9/Meis code representing the level of amputation. The distal boundary is represented by cells in contact with the AEC that are expressing Hoxa 13. Positional identities are intercalated between these two boundaries and the boundaries separating stylopodium and autopodium from the zeugopodium, and between the stylopodial/zeugopodial and zeugopodial/autopodial boundaries. Positional identities might be specified by a PD morphogen as suggested by Meinhardt (1983c), or by an averaging mechanism of short-range cell interactions as suggested by Maden (1977).

What Inhibits Limb Regeneration in Adult Frogs, Birds, and Mammals?

The limb buds of early anuran tadpoles regenerate perfectly, but lose the capacity for regeneration at progressively more distal levels as the limb parts differentiate in a proximal to distal sequence. The late tadpoles and froglets of some anurans form no blastema at all, whereas others such as Xenopus form a fibroblastic blastema that differentiates into a symmetrical cartilage spike without muscle (Dent, 1962; Korneluk and Liversage, 1984; Wolfe et al., 2000). Fibroblastema formation is associated with subnormal histolysis and dedifferentiation and a thinner AEC (Wolfe et al., 2000; Suzuki et al., 2005, 2006). Satellite cells are present in the myofibers at the amputation plane, but do not become part of the fibroblastema, a situation that can be remedied by transplanting cells that secrete hepatocyte growth factor (HGF) into the blastema (Satoh et al., 2005), or by ectopic shh expression or delivery of Shh and/or Fgf10 protein into fibroblastemas (Nye et al., 2011).

The regeneration of a symmetrical cartilage spike in amputated Xenopus limbs is correlated with deficiencies in the expression patterns of several patterning genes. The lack of asymmetry in the AP axis is associated with the failure of shh activation in the fibroblastema and thus failure of downstream targets of shh (Ptc-1, Ptc-2, Gli 1) to be expressed (Endo et al., 2000; Satoh et al., 2006; Yakushiji et al., 2007). Failure of shh activation is not due to lack of expression of upstream regulators of shh (Gli 3, dHAND), or to epigenetic changes in the shh promoter, but rather to an epigenetic modification in the long-range limb-specific shh enhancer sequence, MFCS1 (Yakushiji et al., 2007). This sequence is hypermethylated in the froglet limb and blastema, in contrast to medium levels of methylation in stage-53 tadpole hind limb buds and blastemas, and low to medium levels of methylation in newt and axolotl limbs and blastemas (Yakushijii et al., 2007). The pattern of MFCS1 methylation in regeneration-competent blastemas is region-specific, being high in anterior tissue, but much lower in shh-expressing posterior tissue.

There also appears to be a lack of PD positional information in the froglet blastema (Ohgo et al., 2010). Early blastemas of stage-53 hindlimb buds amputated through the knee or ankle/foot boundary exhibit nested expression of Hoxa11 and 13. As the blastema grows, the expression domains of these two genes separate in the knee blastema so that Hoxa 13 is expressed in the autopodial region, and Hoxa 11 is expressed in the zeugopodial region, whereas in the ankle blastema, Hoxa11 expression ceases, leaving only Hoxa 13 expression. This separated gene expression pattern is associated with the emergence of differential adhesive properties that enable distal and proximal blastema cells to sort out from one another in a mixture, reflecting the establishment of their differential PD positional identity. A similar nested expression of Hoxa 11 and 13 is initially observed in froglet fibroblastemas, but the expression domains fail to separate with blastema growth, and distal cells do not sort out from proximal cells. Thus, both the PD and AP components of the regenerative positional information system appear to be disrupted.

While we can correlate the lack of true blastema formation in Xenopus with certain deficiencies compared to urodeles, we do not yet know the fundamental physiological reasons as to why juvenile and adult urodeles and early anuran tadpoles are able to form a regeneration-competent blastema, whereas late anuran tadpoles and adults can form only a regeneration-deficient or incompetent blastema. Two main ideas have emerged. The first is that the regenerative program is suppressed in adult frogs, birds, and mammals as the immune system matures during development and the limb buds shift to a faulty regeneration response that is associated with a more intense inflammatory response to amputation (for review, see Harty et al., 2003; Mescher and Neff, 2005, 2006; Godwin and Brockes, 2006). Evidence supporting this idea is that urodeles do have a more primitive immune system than Xenopus (Cohen, 1971; Dupasquier and Flajnik, 1999) and that the immune system of Xenopus changes profoundly during tadpole development, coincident with loss of limb regenerative capacity. This was shown by the fact that skin taken from a regeneration-competent early tadpole and cold preserved was rejected when autografted to the donor after its metamorphosis (Izutsu and Yoshizato, 1993).

Further support for the idea of an inverse relationship between immune competence and limb regeneration comes from studies of mammalian fetal wounds. Mouse fetal limb buds have a limited capacity for regeneration (Wanek et al., 1989; Reginelli et al., 1995; Han et al., 2003), and mouse fetal skin regenerates until late in gestation, when it shifts to the adult scarring response to wounding (Martin, 1997; Ferguson and O'Kane, 2004). Skin regeneration in the mouse fetus is correlated with a minimal inflammatory response, reflected in lower numbers of platelets and macrophages, a lower ratio of TGF-β1, 2/TGF-β3, and type I/III collagens, lower levels of platelet-derived growth factor (PDGF) and its receptor, and higher levels of hyaluronic acid (HA) and its receptor (for review, see Stocum, 2006). Antibodies to TGF-β1, 2, or addition of exogenous TGF-β3 administered early in the course of adult mouse skin repair evoke a more regenerative response (Shah et al., 1994, 1995), while hyaluronidase and PDGF administered to fetal skin shifts the wound response toward scarring (Haynes et al., 1994; Mast et al., 1995). Skin wounds in antibiotic-maintained PU.1 null mice, which lack macrophages and neutrophils, are repaired by regeneration (Martin et al., 2003).

However, other evidence suggests that suppression of the regenerative capacity is not due to the maturation of the immune system, but rather how the changing developmental state of cells alters their response to injury. Evidence for this possibility was the demonstration by transplantation experiments that the ontogenetic decline in regenerative ability of Xenopus limb buds is the result of intrinsic changes in limb bud cells (Sessions and Bryant, 1988). Furthermore, fetal mouse skin fibroblasts maintain their regenerative response when grafted subcutaneously into adult athymic mice, even though these host mice heal by scarring (Lorenz et al., 1992; Lin et al., 1994) and the skin of early mouse limb buds cultured in vitro undergoes the transition from regeneration to scarring in response to wounding in the compete absence of circulating immune cells (Chopra et al., 1997).

The second idea is that urodeles have evolved (or retained) limb regeneration-specific genes not found in frogs, birds, and mammals that allow their limb cells to undergo dedifferentiation and accumulate as a blastema, or that frogs, birds, and mammals have evolved regeneration-suppressing genes not found in urodeles. Evidence for both has been presented. Firstly, a bioinformatic analysis failed to find homologues of Prod1 in other vertebrate taxa, suggesting that this crucial regeneration gene is special to urodeles (Garza-Garcia et al., 2009). Secondly, newt and mammalian myotube nuclei appear to differ in their ability to re-enter the cell cycle in response to serum stimulation. During muscle development, the cell cycle checkpoint protein, pRB, is hyperphosphorylated and inactive in primary myoblasts, allowing them to proliferate, but as myoblasts fuse and differentiate into myotubes, pRb becomes hypoposphorylated and actively suppresses cell cycle activity in myonuclei. Serum stimulation of newt myotubes inactivates pRb, promoting their re-entry into the cell cycle (Tanaka et al., 1997, 1999). However, myonuclei of normal mammalian myotubes do not re-enter the cell cycle in response to serum stimulation, suggesting that pRb inactivation is not sufficient to reverse cell cycle suppression. This was confirmed by experiments in which pRb was deleted (Huh et al., 2004) or knocked down by si RNA (Pajcini et al., 2010) in myotubes derived from primary myoblasts. The nuclei of these myotubes failed to re-enter the cell cycle in response to serum stimulation.

Based on these results, Pajcini et al. (2010) hypothesized that birds and mammals have evolved an additional cell cycle checkpoint gene not present in urodeles whose protein product synergizes with pRb to suppress cell cycle re-entry upon muscle cell differentiation. To get primary mammalian myotube nuclei to enter the cell cycle, this protein or its gene would have to be inactivated along with pRb. They focused on alternative reading frame (ARF), a tumor suppressor protein encoded by the ink4a locus, which is expressed only in taxa above the urodeles. Simultaneous knockdown of the Rb and ARF genes by siRNAs reversed cell cycle inhibition in differentiated myocytes. The myocytes dedifferentiated, re-expressed Pax-7, and formed myoblast colonies that retained the ability to undergo myogenic differentiation and form new myofibers after transplantation in vivo (Pajcini et al., 2010).

Myotubes derived from some abnormal myoblast cell lines in which pRb has been inactivated or lost have been reported to re-enter the cell cycle in response to serum stimulation, implying that ARF may not be an impediment to dedifferentiation in these cells. Serum stimulates C2C12 myotubes in which pRb has been knocked down to re-enter the cell cycle. However, C2C12 myotubes are derived from immortalized myoblasts and have a deletion in the ink4a locus that expresses ARF (Pajcini et al., 2010). Reversal of differentiation by serum stimulation was also reported for myotubes differentiated from CC42 muscle cell lines derived from embryonic lethal Rb−/− mice (Schneider et al., 1994). The differentiation of wild-type skeletal muscle requires pRb (Huh et al., 2004), but p107 can substitute for pRb in pRb−/− muscle differentiation (Schneider et al., 1994). This alternate differentiation pathway apparently does not involve placing a lock on cell cycle re-entry like it does in the differentiation of normal muscle. Thus, it is not surprising that pRb elimination alone allows both C2C12 and CC42 cells to re-enter the cell cycle upon serum stimulation.

In our view, experiments trying to understand how urodele and mammalian muscle cells (or other cells) differ in their intrinsic ability to enter the cell cycle should focus on normal (i.e., wild-type) cells or cell lines, which Pajcini et al. (2010) have done. It would now be of great interest to see whether transfection of mammalian ARF into urodele myotubes would render them refractory to entering the cell cycle upon serum stimulation. McGann et al.(2001) have reported that newt blastema extract promotes cell cycle re-entry and cellularization of C2C12 myotubes, and it would be of interest to determine whether this extract would do the same to mammalian myotubes derived from primary muscle cells. Still another interesting question is whether ARF is expressed in anurans such as Xenopus, which can regenerate amputated limb buds of early tadpoles, but then lose this capacity as the limb buds differentiate.

LOOKING DISTALLY: PERSPECTIVE FOR THE 21ST CENTURY

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. LOOKING PROXIMALLY: LIMB REGENERATION STUDIES FROM 1911–2011
  5. LOOKING DISTALLY: PERSPECTIVE FOR THE 21ST CENTURY
  6. NEW APPROACHES AND CHALLENGES
  7. TRANSLATION TO CLINICAL THERAPIES
  8. Acknowledgements
  9. REFERENCES

Virtually all of the major problems of limb regeneration investigated over the last 100 years leave unresolved questions that beg to be re-visited using more advanced technologies, new concepts and new approaches.

Origin of Blastema Cells

Several questions remain about the origin of blastema cells. First, what percentage of the total blastema mesenchyme in urodeles is contributed by cellular dedifferentiation versus adult stem cells? How much of the muscle of the regenerate is formed by satellite cell progeny and how much by mononucleate cells derived by myofiber dedifferentiation and cellularization? Do the latter cells have the same characteristics as satellite cells, such as expression of Pax-7? Current marking techniques do not distinguish between satellite cells and myogenic cells derived by dedifferentiation. Might it be that all the muscle of the regenerate is derived from satellite cells? Do mesenchymal stem cells that reside in the periosteum, endosteum, or bone marrow local to the amputation site contribute to the blastema, and if so, what is their percentage of contribution? Might there be a population of stem cells among the dermal fibroblasts that normally maintains connective tissues? Does the percentage contribution of stem cells versus dedifferentiated cells to the blastema change as larvae grow, metamorphose, and become adults?

Some of these questions can be answered by use of the transgenic GFP marking/grafting techniques reported by Sobkow et al. (2006), and/or the development of transgenics carrying genes for proteins that fluoresce with different colors. Others will require the use of much more precise transgenic techniques to mark individual cell populations, such as MSCs, in situ. In mammalian regenerating systems, inducible Cre/reporter gene constructs are used to label specific differentiated cell populations and then track their contributions to regenerated tissue (Branda and Dynecki, 2004). Can such systems be developed for amphibians, or can alternative methods be designed? The recent demonstration of a simpler and more efficient method of creating transgenic newts (Casco-Robles et al., 2010) suggests that this is possible.

Mechanisms of Histolysis and Dedifferentiation

Although we know most of the proteases involved in ECM remodeling, we know the cellular origin of only a few of them. Nearly all studies of MMPs have looked at transcripts rather than the actual proteins, leading to the question of whether or not there is differential translation of any of these transcripts. Furthermore, comparative information on the patterns and levels of MMPs and other proteolytic enzymes in regeneration-competent versus regeneration-deficient amphibians would be useful in determining why fewer cells are liberated to undergo dedifferentiation in the latter. Finally, we do not have good information on how histolysis is brought to a halt. While TIMPs are likely to be involved, regulation of MMP production at the transcriptional and/or translational level is possible as well.

A major question is what are the mechanisms that produce the shift in transcription that leads to the dedifferentiated state? What is the role that breaking contacts between ECM molecules and integrin receptors plays as the ECM is degraded by proteases? Dissolution of such contacts might lead to changes in cell shape and reorganization of the actin cytoskeleton (Juliano and Haskill, 1993) that activate signal transduction pathways to down-regulate phenotype-specific transcription programs and up-regulate programs characteristic of a less specialized state that allows blastema cell migration and response to proliferation and patterning signals.

Does the differential regulation of cell stress and apoptotic pathways play a role in the dedifferentiation of limb cells? Proteomic analysis suggests that reduced metabolic activity, up-regulation of the unfolded protein response (UPR), and differential regulation of apoptotic pathways may largely prevent apoptosis of limb cells (Rao et al., 2009), which is known to be minimal during blastema formation (Mescher et al., 2000; Atkinson et al., 2006). Studies on cultured chondrocytes, β-cells, and Muller glia cells of the retina suggest that cells dedifferentiate as part of a mechanism to combat apoptotic cell stress (for discussion, see Rao et al., 2009).

However, apoptosis within the first 24 hr after amputation, before dedifferentiation begins, may actually be necessary for regeneration. Two events that occur within 24 hr after amputation of stage-48 Xenopus tails are caspase-dependent apoptosis in the neurogenic tissues at the amputation site (Tseng et al., 2007), and the efflux of protons across the wound epidermis, driven by an ATPase pump (Adams et al., 2007). Caspase-dependent apoptosis also takes place in the amputated tails of the South American knifefish, Apternotus (Sirbulescu and Zupanc, 2009). These events are obligatory for Xenopus tail regeneration, because drug-induced inhibition of either H+ efflux or apoptosis prevents blastema formation; whether this is true for the knifefish is unknown. Whether these two requirements are linked, and how they might be involved in the regulation of gene activity accompanying the initiation of regeneration, is not known, but will make a fascinating line of investigation. Since inhibiting Na+ influx in amputated newt limbs prevents blastema formation, it would not be surprising if H+ efflux and early apoptosis proved to be obligatory for limb regeneration as well.

Another major question is the mechanism of intracellular remodeling during dedifferentiation. The details of internal structural remodeling in dedifferentiating cells are poorly understood. Dismantling of phenotypic structure and function is most visible in myofibers, but the molecular details of the process are largely uninvestigated for all limb cell types. A synthetic trisubstituted purine called myoseverin has been isolated from a combinatorial chemical library and found to cellularize C2C12 myotubes in vitro (Rosania et al., 2000; Chen et al., 2004). Myoseverin disrupts microtubules and upregulates genes for growth factors, immunomodulatory molecules, ECM remodeling proteases, and stress-response genes, consistent with the activation of pathways involved in wound healing and regeneration (Rosania et al., 2000), but does not activate the whole program of myogenic dedifferentiation (Duckmanton et al., 2005). Another synthetic small molecule isolated by a chemical library screen, a disubstituted purine dubbed reversine, can induce the dedifferentiation of some cell types. Reversine treatment of C2C12 myoblasts resulted in mononucleate cells that behaved like mesenchymal stem cells (MSCs); i.e., they were able to differentate in vitro into osteoblasts and adipocytes, as well as muscle cells (Anastasia et al., 2006). Myoseverin and reversine may be useful in analyzing the events of structural remodeling. A major question is whether they have natural counterparts that can be identified.

What mechanism insures that limb cells undergo only limited reprogramming to limb bud-like cells and no further? Do dedifferentiated cells retain some of the molecular features of their parent cells that may bias them toward redifferentiating into their tissue of origin? Interestingly, iPSCs have been found to carry parental epigenetic marks during early passaging even though they are reprogrammed to a pluripotent state (Kim et al., 2010; Polo et al., 2010). What then confers plasticity on dermal fibroblasts, enabling them to differentiate into chondrocytes during limb regeneration? Changes in epigenetic methylation and acetylation via chromatin-modifying enzymes will be crucial for understanding the mechanism of dedifferentiation in regenerating amphibian limbs. The characterization of blastema cell surface antigen and micro-RNAs will also be important. Might it be possible to directly reprogram dermal fibroblasts in situ to limb cell types other than cartilage, for example muscle, as has been reported for the reprogramming of mammalian neonatal cardiac fibroblasts into functional cardiomyocytes (Ieda et al., 2010)? This would be valuable as a therapy to reconstitute individual limb tissues damaged by injury.

Blastema Cell Proliferation

Axon-induced AGP expression has been observed at two stages of limb regeneration, dedifferentiation (in Schwann cells) and formation of the accumulation blastema (in epidermal or subepidermal gland cells). This and other data suggest that growth of the blastema is regulated by the continuous neural induction of AGP expression by the AEC. More evidence is needed, however, to sustain this conclusion. The expression pattern of AGP should be examined throughout the period of blastema growth and redifferentiation in the presence and absence of nerves.

Denervation at any stage should shut off AGP expression in the AEC, and stop blastema growth. The expression of AGP in regenerating aneurogenic and reinnervated aneurogenic limbs, as well as regenerating limb buds at various stages of normal development, should be investigated to help clarify the functional relationship between the AEC and nerves. Can AGP substitute for the function of the AEC if the AEC is removed? The Fgfs expressed by the AEC also play a role in blastema cell proliferation. Does nerve-induced AGP stimulate expression of Fgfs by the AEC, or are they independently regulated? Denervation of cone and later stage blastemas, or inhibition of AGP, should inhibit Fgf expression if AGP stimulates its expression. What roles do other molecules expressed by nerves such as transferrin, substance P, and neuregulin play in blastema growth? Are they responsible for the inductive effect of the nerve on the AEC, or do they synergize with AGP and other factors expressed by the AEC in promoting proliferation?

Models of Pattern Formation

Lewis Held (1992) wrote: “The question of how patterns originate is the Gordian Knot of developmental biology.” This remains true today for both limb bud development and for development of the regeneration blastema. It is important to realize that, given our relative lack of understanding of regeneration at the cell and molecular level, our models of regeneration are still superficial. Common to all limb regeneration models is the notion that cellular interactions driven by positional discontinuities in all three axes of the limb are essential for blastema formation, growth, and patterning. It is likely that the mosaic spatial relationships of cellular contributions, as well as intercalation and cell sorting to eliminate discontinuities, are all important for these processes. Newer molecular data and cell-marking technologies may provide a better foundation to answer some of our questions. For example, the origin of supernumerary limbs after AP, DV, and APDV axial reversal of the blastema at the level of origin and after distal to proximal and proximal to distal grafting, with and without axial reversal, should be revisited with GFP cell-marking technology.

How the PD axis is regenerated remains a major mystery. We lack the kind of molecular details for pattern formation in embryonic amphibian limb buds and regeneration blastemas that are available for chick and mouse limb buds. This is partly because fluorescent in situ hybridization in regenerating limbs has proven difficult compared to chick and mouse limb buds. Obtaining this data should be an objective. What genes in addition to Hoxa, Hoxd, Meis, and Prod1are involved in PD axis formation? Study of the expression patterns of the distal marker Hoxa 13 and the proximal markers Hoxa 9 and Meis in individual cells during blastema formation and growth will provide valuable information to test the idea that the first cells to dedifferentiate during blastema formation are autopodial and whether blastema growth or cell sorting is the key to manifestation of the ability to self-organize a full set of structures in the PD axis. Given the interdependence of AP, DV, and PD positional identity, does RARδ2 mediate the posteriorization and ventralization effect of RA, or are other RARs involved in these effects?

What is the physical mechanism by which positional identities are restored in the PD axis? Is a morphogen involved, and if so what is it? If local cell interactions are the mechanism, what is the molecular nature of these interactions? Finally, what is the role of the AEC in PD blastema pattern formation? Does the AEC have a distal boundary function and does it carry circumferential/radial positional identities that must interact during wound healing to initiate regeneration?

Why Frog, Bird, and Mammalian Limbs Are Regeneration-Deficient

Regeneration deficiency/incompetence refers only to the fact that limbs regenerate hypomorphically or not at all. There are many potential factors underlying this regeneration deficiency/incompetence that range from subnormal histolysis to intrinsic changes in the ability of cells to produce or respond to signals crucial to formation of a regeneration blastema, to inability of blastema cells to express or respond to patterning signals. These changes might involve the immune system, the loss of critical regeneration genes, or the gain of regeneration-inhibiting genes. If the immune system inhibits regeneration, one might expect to find differences in the ratios of growth factors, cytokines, and ECM components in amputated regeneration-competent versus deficient amphibian limbs, comparable to what is observed in regeneration-competent fetal versus regeneration-deficient mammalian skin. For example, fetal skin has lower ratios of TGF-β1 and 2/ TGF-β3 and type I/III collagens, and a lower level of PDGF (for reviews, see Stocum, 2006). Would antibodies to TGF-β3 retard or inhibit blastema formation in regeneration-competent limbs, and would augmenting TGF-β3 while simultaneously inhibiting TGF-β1, 2, enhance blastema formation in regeneration-deficient limbs?

If early anuran tadpoles can regenerate their limbs, but lose this ability as differentiation progresses from proximal to distal, how does this square with the idea that regeneration competence in urodeles is due to the gain or retention of regeneration-promoting genes, or the lack of genes that impose regeneration incompetence? Are regeneration-promoting genes suppressed and regeneration-inhibiting genes activated as the regenerative capacity of the Xenopus tadpole limb declines? Or is blastema formation on an amputated urodele or anuran limb bud not the same as blastema formation on a fully differentiated urodele limb? Can newt blastema extract (McGann and Odleberg, 2001) reverse the differentiation of mammalian myotubes derived from primary muscle cells, and if so, does this extract contain inhibitors of pRb and ARF? Will transfection of mammalian ARF into fully differentiated newt myofibers in vitro or in vivo render them unable to dedifferentiate upon serum stimulation or amputation? A further question is whether newt blastema extract or simultaneous inhibition of pRb and ARF would cellularize and reverse the differentiation of freshly isolated wild-type adult mammalian myofibers. It is these fully differentiated myofibers, and the other fully differentiated cells of the adult limb, that constitute the “hard targets” for forced dedifferentiation, not the “soft targets” represented by C2C12 myotubes or even primary myotubes differentiated in vitro. For example, preliminary results (Milner and Cameron, unpublished data) indicate that while myoseverin can cellularize myotubes derived from both C2C12 cells and primary mouse myoblasts, it fails to cellularize cultured adult mouse myofibers.

NEW APPROACHES AND CHALLENGES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. LOOKING PROXIMALLY: LIMB REGENERATION STUDIES FROM 1911–2011
  5. LOOKING DISTALLY: PERSPECTIVE FOR THE 21ST CENTURY
  6. NEW APPROACHES AND CHALLENGES
  7. TRANSLATION TO CLINICAL THERAPIES
  8. Acknowledgements
  9. REFERENCES

Granting agencies such as the NIH and NSF reject proposals that are not hypothesis-driven, and journals increasingly reject papers that editorial boards and referees feel do not test hypotheses about “mechanism” (O'Malley et al., 2009). Inductive reasoning from observations or experimentally derived data (“discovery science”) has almost come to be treated as unscientific. In reality, we need discovery science to generate hypotheses for testing. Glass and Hall (2009) have argued that when we do not know how a process works, the appropriate approach is not to formulate a hypothesis, but to ask a question, use the question as a basis to accumulate data, and build a model from the data that can be tested for its predictive ability. This is how models of blastema pattern formation have been constructed and tested. It is also the rationale behind the new global bioinformatic and systems biology approaches that integrate the molecular elements of large genomic and proteomic databases into functional networks and pathways from which hypotheses about mechanisms emerge (Nabel, 2009). For example, global genomic and proteomic data on blastema formation in regenerating anuran and urodele limbs has been generated by a number of laboratories (King et al., 2003; Habermann et al., 2004; Putta et al., 2004; Grow et al., 2006; Pearl et al., 2008; Smith et al., 2009; Monaghan et al., 2009; Rao et al., 2009) and in one case transcription factor pathways have been subjected to a systems biology analysis (Jhamb et al., 2011).

Proteomic analysis of blastema formation in amputated axolotl limbs has already begun to reveal novel features. For example, the low mitotic index observed during blastema formation is associated with the strong upregulation of the ecotropic viral integration factor 5 (Evi5) throughout blastema formation in regenerating axolotl limbs (Rao et al., 2009). Evi5 is a centrosomal protein that accumulates in the nucleus during early G1 phase in mammalian cells and, along with another protein Pin1 (Bernis et al., 2007), prevents the cells from prematurely entering mitosis by stabilizing Emi1, a protein that inhibits cyclin A degradation (Eldridge et al., 2006). At G2, Emi1 and Evi5 are phosphorylated by Polo-like kinase 1 (PLK1) and targeted for ubiquitin-driven degradation, allowing cells to enter mitosis. It is thus possible that high levels of Evi5 during blastema formation may restrain dedifferentiated cells from entering mitosis until they are present in enough numbers to form an accumulation blastema (Rao et al., 2009). Evi5 also renders the vesicle trafficking protein Rab 11 inactive, which would help prevent cells from entering mitosis by inhibiting the vesicular recycling of receptors that would otherwise transduce mitotic signals (Westlake et al., 2007; Dabbeekeh et al., 2007).

Ear tissue of the MRL/lpj mouse regenerates following punch wounds (Heber-Katz, 2004). Interestingly, the fibroblasts of uninjured ear tissue of this unusual mouse exhibit a high proportion of 4N (G2-arrested) fibroblasts, which is associated with a G1 checkpoint deficiency in p21 (Bedelbaeva et al., 2010). B6 mice do not regenerate ear tissue, but do so after knockout of p21. Lack of p21 appears to allow fibroblasts to traverse S and arrest in G2 and thus be ready for mitosis upon injury to the tissue. P21 prevents cyclins from phosphorylating pRB, and its loss theoretically would inactivate pRb, leaving ARF as the main impediment to cell cycle re-entry. It would be interesting to see whether p21 knockout alone would allow wild-type mammalian muscle cell nuclei to re-enter the cell cycle in response to serum factors. Or, are fibroblasts relatively undifferentiated cells compared to muscle so that re-entry into the cell cycle has fewer requirements? There are more questions than answers with regard to the relationship between cell cycling and dedifferentiation. It would be informative to do a comparative molecular analysis of cell cycle checkpoint proteins in the muscle, skeletal, and fibroblast cells of the urodele limb versus those of their mammalan counterparts to reveal what differences might exist that are associated with the ability to dedifferentiate in response to injury.

Combining new molecular technology with comparative analysis of regeneration in the limbs of a variety of regeneration-competent organisms and in regeneration-competent versus deficient/incompetent limbs will be a powerful approach by which to discover genetic and proteomic circuits common to regeneration-competent limbs, and to discover molecular differences that distinguish regeneration competence from deficiency or incompetence. Such discovery will allow the identification of potential intervention points to confer regenerative power upon regeneration-deficient tissues. However, amphibians present technical challenges that currently limit their use for understanding the molecular biology of regeneration. Molecular tools available for mammalian studies such as species-specific antibodies and gene chips are not available for urodele amphibians. While some mammalian antibodies recognize amphibian proteins, most do not. There is a great need for the development of commercial antibody panels to characterize the surface phenotype of blastema cells in regeneration-competent versus deficient limbs, and to investigate the expression of other proteins. Likewise, there is a need for genomic sequencing of urodele and anuran species, and the development of inexpensive gene chips to conduct comparative analysis of transcript expression.

TRANSLATION TO CLINICAL THERAPIES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. LOOKING PROXIMALLY: LIMB REGENERATION STUDIES FROM 1911–2011
  5. LOOKING DISTALLY: PERSPECTIVE FOR THE 21ST CENTURY
  6. NEW APPROACHES AND CHALLENGES
  7. TRANSLATION TO CLINICAL THERAPIES
  8. Acknowledgements
  9. REFERENCES

A question often asked by people intrigued by research on appendage regeneration in amphibians and fish is how long will it be before we are able to regenerate a human limb? The fact is, we don't know. It could be 10 years, 50 years, or a century from now. We will most likely figure out how to regenerate a finger or toe first, then scale up to regeneration of a full arm or leg. It is possible that we regenerationists will be beaten to limb replacement by the bioengineers. For example, the inventor of the Segway, Dean Kamen, is developing a sophisticated artificial arm with the potential for neural-machine interface. Nevertheless, nothing beats the real thing!

Some hold the opinion that we will never be able to regenerate large, complicated structures like limbs. Such absolutes have been pronounced many times before and have been proven wrong. The beauty of basic science is that it can generate unexpected advances over a short time frame. Four years ago, it was thought that the reprogramming of somatic cells to pluripotency would be so complicated that it was a decade or more away, yet the solution proved to be relatively simple (Takahashi et al., 2007; Yu et al., 2007). We may hope for a similar breakthrough in understanding the mechanisms of regeneration.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. LOOKING PROXIMALLY: LIMB REGENERATION STUDIES FROM 1911–2011
  5. LOOKING DISTALLY: PERSPECTIVE FOR THE 21ST CENTURY
  6. NEW APPROACHES AND CHALLENGES
  7. TRANSLATION TO CLINICAL THERAPIES
  8. Acknowledgements
  9. REFERENCES

We thank the W.M. Keck Foundation and the U.S. Army Office of Research for research support. We thank our colleagues Xiaoping Chen, Derek Milner, and Fengyu Song for a critical reading of the manuscript. We also thank Associate editor John Fallon and the reviewers for their most helpful comments and critique.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. LOOKING PROXIMALLY: LIMB REGENERATION STUDIES FROM 1911–2011
  5. LOOKING DISTALLY: PERSPECTIVE FOR THE 21ST CENTURY
  6. NEW APPROACHES AND CHALLENGES
  7. TRANSLATION TO CLINICAL THERAPIES
  8. Acknowledgements
  9. REFERENCES