Regeneration potency of mouse limbs


*Author to whom all correspondence should be addressed.


Mammalians have a low potency for limb regeneration compared to that of amphibians. One explanation for the low potency is the deficiency of cells for regenerating amputated limbs in mammals. Amphibians can form a blastema with dedifferentiated cells, but mammals have few such cells. In this paper, we report limb formation, especially bone/cartilage formation in amputated limbs, because bone/cartilage formation is a basic step in limb pattern regeneration. After the amputation of limbs of a neonatal mouse, hypertrophy of the stump bone was observed at the amputation site, which was preceded by cell proliferation and cartilage formation. However, no new elements of bone/cartilage were formed. Thus, we grafted limb buds of mouse embryo into amputated limbs of neonatal mice. When the intact limb bud of a transgenic green fluorescent protein (GFP) mouse was grafted to the limb stump after amputation at the digit joint level, the grafted limb bud grew and differentiated into bone, cartilage and soft tissues, and it formed a segmented pattern that was constituted by bone and cartilage. The skeletal pattern was more complicated when limb buds at advanced stages were used. To examine if the grafted limb bud autonomously develops a limb or interacts with stump tissue to form a limb, the limb bud was dissociated into single cells and reaggregated before grafting. The reaggregated limb bud cells formed similar digit-like bone/cartilage structures. The reaggregated grafts also formed segmented cartilage. When the reaggregates of bone marrow mesenchymal cells were grafted into the stump, these cells formed cartilage, as do limb bud cells. Finally, to examine the potency of new bone formation in the stump tissue without exogenously supplied cells, we grafted gelatin gel containing BMP-7. BMP induced formation of several new bone elements, which was preceded by cartilage formation. The results suggest that the environmental tissues of the stump allow the formation of cartilage and bone at least partially, and that limb formation will be possible by supplying competent cells endogenously or exogenously in the future.


Mammalians have limited potency for regenerating limbs compared to the regeneration potency of amphibians. For example, urodela amphibians can regenerate limbs completely after amputation at any level, whereas humans and mice can regenerate only digit tips (Illingworth 1974; Borgens 1982). Urodela amphibians can regenerate not only appendages but also other body parts such as lens, tail, lower jaw, spinal cord and forebrain (Brockes & Kumar 2002).

Regeneration can be classified into: (1) epimorphic regeneration, which accompanies organ morphogenesis in the regenerating process; and (2) tissue regeneration, which only regenerates lost tissue regions. Liver regeneration in humans is classified as tissue regeneration, and limb regeneration in urodeles is called epimorphic regeneration. The regeneration of digit tips is a rare case of epimorphic regeneration in mammals, and it has therefore been considered as a model for organ regeneration in higher vertebrates.

The mechanisms of amphibian epimorphic regeneration in limbs have been studied in detail, but little is known about mammalian epimorphic regeneration in limbs because the regeneration potency is limited. Thus, comparison with the urodele limb regeneration process is useful for studying mammalian limb regeneration.

The regeneration process in amphibian limbs is divided into several steps. First, the wound formed at the amputated site is rapidly enclosed by epidermis. After this wound healing process, multiple tissue interactions, especially epidermis–mesenchyme interactions, induce a dedifferentiation step. The wound epidermis becomes thickened to form an apical epidermal cap (AEC), and dedifferentiated limb cells and/or stem cells accumulate beneath the AEC to form a blastema. AEC and blastemal cells secrete growth factors that are trophic for each other, as in limb bud development. Blastema cells proliferate and redifferentiate by these interactions to restore the limb pattern (Stocum 1998; Bryant et al. 2002). Therefore, the mechanism of amphibian limb regeneration is similar to the reformation of a limb bud at the tip of the stump. It is known that the blastema has nerve dependency because it has been shown that denervation treatment prevents the proliferation of blastema cells (Singer 1952; Endo et al. 2000). Thus, the process of amphibian limb regeneration can be divided into wound healing, dedifferentiation and redifferentiation steps.

In the case of mammalian limbs, only the wound healing process is observed after amputation. The stump is first covered by a scab and then the wound epidermis encloses the stump slowly compared to the speed of the process in amphibians. It is also known that the formation of the wound epidermis accompanies the formation of the dermis in the enclosure step of amputated mouse limbs (reviewed by Martin 1997). The dermis may prevent direct interaction of the epidermis with stump tissue. The dermis expresses fibroblast growth factor (FGF)7 (Werner et al. 1992) and FGF10 (Tagashira et al. 1997) in the wound healing process. However, the interaction that is observed in limb bud development or amphibian limb regeneration is not known in amputated mouse limbs. In contrast to amphibian limbs, digit tip regeneration in mouse occurred even after denervation (Mohammad & Neufeld 2000). This suggests that the dedifferentiation step is not necessary in mammalian digit tip regeneration.

It has been reported that the regenerative region of the mouse digit tip is consistent with the Msx1-positive area (Reginelli et al. 1995). Msx1 is expressed in the limb bud progress zone where cells are kept in an undifferentiated state, and Msx1 expression is maintained by progress zone maintenance factors such as FGF2, FGF4 and FGF8 (Kostakopoulou et al. 1996). Thus, Msx1 is thought to be related to the maintenance of an undifferentiated state. Thus, colocalization of Msx1 and the regenerative compartment may indicate that undifferentiated cells contribute to digit tip regeneration. Planarians are known for their high regenerative potency; however, their cells do not dedifferentiate after amputation, instead they regenerate their bodies by somatic stem cells (Baguñàet al. 1989; Agata et al. 2003). Thus, it seems that the dedifferentiation step is specific to amphibians and that the existence of undifferentiated cells is sufficient for redifferentiation. That is, if a sufficient cell source is supplied, the dedifferentiation step is not necessary for epimorphic regeneration.

In this study, we examined the responses of neonatal mouse limbs to amputation at several levels and compared the responses with those of amphibian limbs. Next, we supplied limb bud mesenchymal cells into the stump as a source for limb regeneration to examine the limb environment for limb regeneration. Finally, we observed bone formation in the stump by the addition of BMP to examine the possibility of bone formation with endogenous limb cells.

Materials and methods


DDY mice (SLC) were used in the experiments. Green fluorescent protein (GFP) mice were generated from CAG-EGFP 129SvJ mice (a gift from Dr Okano) and backcrossed to the DDY strain for at least seven generations. These GFP mice were used for transplantation experiments to distinguish host tissue and graft.

Limb amputation

Before the operations, neonatal mice were anesthetized by icing and adult mice were anesthetized by diethyl ether. Forelimbs were amputated at increasingly proximal non-regenerating levels. The mouse limb consists of three regions: the autopod, zeugopod and stylopod, from distal to proximal. We amputated at three levels (Fig. 1A): at the autopod (metacarpus) level, at the zeugopode (radius and ulna) level and at the stylopod (humeras) level.

Figure 1.

Hypertrophy of amputated bones in mouse forelimbs. (A) Amputation level at stylopod (blue line), zeugopod (red line) and autopod (green line) of neonatal mouse. Amputation was performed in right forelimbs and left forelimbs were used as control (B, H). Samples were collected at day 0 (C), day 3 (D), day 5 (E) and day 10 (F, G, I, J). Neonatal limbs amputated at the zeugopod level (C–F) showed strong hypertrophy. In contrast, hypertrophy was weak in adult limbs amputated at the same level (G). Hypertrophy was also observed in the limbs amputated at the stylopod (I) and autopod (J) levels of neonatal mice. Cartilage and bone were stained with alcian blue and alizarin red, respectively.

Alcian blue/alizarin red staining

To observe the skeletal pattern, samples were stained with alcian blue and alizarin red for cartilage and bone, respectively. Prior to staining, skin was removed and limbs were fixed in 100% ethanol for 4–6 h. Fixed limbs were stained with 60% ethanol/5% acetic acid/0.015% alcian blue/0.005% alizarin red at room temperature for 12–24 h and then placed in 1% KOH for several hours to wash the dyes and dissolve the connective tissue (Parr & McMahon 1995). After the skeletal pattern had become obvious, samples were transferred into 50% glycerol and stored in 100% glycerol.

Immunohistochemistry of collagen II

Limbs were fixed in 4% paraformaldehyde for 12 h and then embedded in paraffin. Sections were sliced into 6 µm thickness. The sections were dewaxed in xylene, rehydrated through graded ethanol, rinsed in phosphate-buffered saline (PBS) and then treated with 25 mg/mL hyaluronidase for 30 min at room temperature. After washing twice in PBS, sections were blocked with 1.5% skim milk (Meiji Nyugyo, Tokyo, Japan)/PBS for 90 min at room temperature. After blocking, sections were incubated with 1:200-diluted anticollagen II antibody (LSL, LB-1297) overnight at 4°C. On the next day, after washing twice in PBS, the sections were incubated with fluorescein isothiocyanate (FITC)-conjugated antirabbit IgG for 30 min at room temperature. After washing twice in PBS, sections were covered with mounting medium and mounted with cover grasses.

Assay of DNA synthesis

Animals were injected with bromodeoxyuridine (BrdU) at 50 µg per gram of bodyweight and killed 2 h later. Limbs were fixed in Bouin's fixative for 4 h at room temperature and embedded in paraffin. Tissue sections were cut into 6 µm thickness. BrdU incorporation into DNA was detected by anti-BrdU antibody (OB0030F; AbD Serotec, Oxford, UK). FITC-conjugated antirat IgG was used as a secondary antibody.

Alcian blue-hematoxylin-eosin staining

Sections were sliced into 6 µm thickness. After removal of paraffin, sections were washed in 70% ethanol with 1% HCl and then stained with 0.015% alcian blue/70% ethanol with 1% HCl for 30 min. After washing twice with 70% ethanol, sections were stained with hematoxylin and eosin (HE).

Whole limb bud transplantation

Limb buds for grafting were collected from E11-13 embryos and kept in Tyrode at 4°C. The forelimbs of host mice were amputated at the digit joint level and the metacarpus was removed to allow placing of the graft on the space of the stump. One whole limb bud was transplanted to one stump. After the transplantation, instant glue was applied to close the wound.

Reaggregated cell implantation

Collected limb buds were treated with dispase (2000 unit/mL) at room temperature for 10 min and the ectoderm was removed. After rinsing with Ca2+ Mg2+-free Tyrode, the mesenchyme was dissociated by pipetting. In the case of limb bud dissociation at a stage later than E11.5, collagenase treatment and a filtration step were added for sufficient dissociation. Dissociated cells were centrifuged and reaggregated in a fibrin clot, with 3 × 104 cells being reaggregated in one clot for grafting.

Isolation and grafting of bone marrow-derived mesenchymal cells

Bone marrow-derived mesenchymal cells were collected from femur bones of 4-week-old mice. After washing in PBS several times, the bones were treated with 0.25% trypsin-ethylenediaminetetraacetic acid (EDTA) solution for 30 min at 37°C to remove surrounding soft tissue. Both sides of the bone were cut and the bone marrow was washed by a syringe stream to flush out bone marrow vascular zone cells. The cleaned bones were dissected with scissors and treated with 0.4% collagenase/Dulbecco's modified Eagle's medium (DMEM) for 30 min at 37°C. After the addition of an equal volume of 10% fetal bovine serum (FBS)-containing medium to stop the reaction, tissue suspension was pipetted to detach the cells from bone pieces, passed through a 40 µm cell strainer, and centrifuged at 400 × g for 7 min at 4°C. The pellet was suspended by pipetting and cultured in 10% FBS/DMEM for 1 week. Cells that had adhered to the dish were harvested with trypsin and used for grafting after clotting in fibrin, with 3 × 104 cells being reaggregated in one clot.

Isolation and grafting of femur cartilage cells

Femur bones were collected from E16 embryos and surrounding connective tissues were mechanically removed. Collected femur bones were dissected into small pieces and then treated with 0.4% collagenase for 30 min at 37°C. After the addition of an equal volume of DMEM containing 10% FBS, cells were dissociated by pipetting. The suspension was passed through a 40 µm cell strainer and centrifuged at 400 × g for 7 min at 4°C. Pellets were suspended by pipetting and used for grafting after clotting in fibrin, with 3 × 104 cells being reaggregated in one clot.

Implantation of acidic gelatin containing BMP

Acidic gelatin containing BMP-7 was implanted into the spaces of amputated limbs from which cartilage/bones at the distal radius/ulna level and carpus level had been removed. Acidic gelatin was used to supply space for cartilage cells to form a pattern and for gradual desorption of BMP (Tabata et al. 1999). Acidic gelatin was a gift from Nitta Gelatin (Osaka, Japan).


Induction of bone hypertrophy after amputation

To examine whether or not a regenerative response occurs in the mouse, we amputated forelimbs of a neonatal mouse at the zeugopod level (Fig. 1A, red line). The amputation site contained no cartilage (Fig. 1C, compare with Fig. 1B). The stump at the amputation site was immediately enclosed by the scab (Fig. 1D). After wound enclosure, the edge of the amputated bone showed strong hypertrophy. Hypertrophy started at day 4–5 (Fig. 1E) and the degree of hypertrophy increased day by day (Fig. 1F). When amputated at the stylopod level (Fig. 1A, blue line) or the autopod level (Fig. 1A, green line), hypertrophy of the amputated bone was similarly observed (Fig. 1I,J). However, when the limb was amputated at the joint level to avoid bone amputation, no hypertrophy was observed (data not shown). These results suggest that hypertrophy is the response to bone amputation and is independent of the amputation level of bone. In contrast to the neonatal mouse, hypertrophy was significantly weak in the adult mouse (Fig. 1G). This age-dependency of hypertrophy suggests that cells responsible for hypertrophy are reduced in adult mice. Thus, it is thought that the cells responsible may be tissue-specific progenitor cells or stem cells that abundantly exist in neonatal mice.

It was thought that cell proliferation should have occurred before hypertrophy. Thus, we examined DNA synthesis by BrdU incorporation to determine which cells contribute to the repair of stump tissues and hypertrophy. Forelimbs were amputated at the zeugopod level and BrdU incorporation was observed 5 days after amputation (Fig. 2A). The results showed that active cell proliferation was only observed around the stump (Fig. 2B) and that cell proliferation activity in the proximal region was low, as in normal limb tissues (Fig. 2C). These results suggest that the amputation signal affected the region just around stump tissues. Therefore, only a limited number of progenitor cells and/or stem cells could respond to the signal and contribute to the repair process.

Figure 2.

Sections of hypertrophy region. (A–C) Cell proliferation was examined by BrdU incorporation 2 days after amputation. BrdU was detected by FITC-conjugated anti-BrdU antibody. Green dots in (B) and (C) show proliferating cells. White boxes B and C in Figure 2A indicate position of high magnification images. (B, C) Cell proliferation was detected around the stump and not in the proximal region. White arrowheads in Figure 2A indicate the boundary of the proliferating region and the dormant region. (D–F) Chondrogenesis was observed 5 days after amputation. Chondrocytes were stained immunohistochemically with anticollagen II antibody (green). White box in Figure 2D indicates the position of high magnification image (E). Cartilage was also stained by alcian blue, thus this section was stained by alcian blue/hematoxylin-eosin (F). Arrowheads in (E) and (F) show that cartilage was newly formed distal to the amputated edge. R, radius; U, ulna.

Because limb bones are formed by endochondral ossification, it is thought that hypertrophic bones are also formed through chondrogenesis. Indeed, we observed accumulation of chondrocytes covering the edge of amputated bone (Fig. 2D–F). Interestingly, this repair process is different from reports of fracture in adult mammals. In the repair process of fractured bone in mammals, chondrocytes accumulate beside the amputated bone plane because chondrocytes are thought to be supplied from the periosteum. However, in the neonatal mouse, we observed accumulation of chondrocytes covering the amputation plane (Fig. 2D–F). Thus, hypertrophy of amputated bone in the neonatal mouse is different from normal tissue repair in the adult mouse and is also different from epimorphic regeneration in amphibians. Hypertrophy in amputated neonatal mouse limbs might indicate the amount and quality of immature cells.

These findings indicated that the addition of pluripotent cells, which can differentiate into various limb tissues like blastema cells, may help to determine the limb regeneration ability of postnatal mice.

Transplantation of limb bud cells

One explanation for the low potency of limb regeneration in mammals is the deficiency of a cell source that contributes to regeneration. In amphibians, cells in the stump tissues dedifferentiate in order to prepare a sufficient cell source. Thus, we implanted substitute cells to confirm the limb regenerative potency in the tissue environment of neonatal mice. However, there are no cells known to be the equivalent of limb blastema cells in mammals. Desirable properties for the replacement of blastema cells are the ability of cells to respond to the signal from the tissue environment and to have multipotency for differentiation. Thus, for grafting we selected limb bud cells, which have the capacity to respond to morphogenic signals in development and differentiate into all limb tissues.

At first, we transplanted whole forelimb buds into the stumps of neonatal mice (Fig. 3). Forelimb buds were collected from E11-13 GFP mouse embryos and implanted into the pocket of each stump that had been made by the removal of the metacarpus in advance (Fig. 3A, red box). After surgery, the animals were killed at 10 days or 20 days and the skeletal pattern of GFP tissues was examined. Because a limb bud at a stage later than E11.5 had grown too large to implant the whole limb bud, we grafted the distal region of the autopod (Fig. 3B). Results showed that grafts could form digit-like structures that seemed to be continuous to amputated bones. Grafted limb buds formed a segmented pattern and some showed joint-like structures that had cartilage between the segments. Ossification of grafted tissue was completed at 20 days (Fig. 3F,J,N,R). By grafting limb buds at various developmental stages, a tendency for the skeletal pattern to become complicated when limb buds at advanced stages were used was observed (Fig. 3C–R). However, well-developed limb buds behaved independently from host tissue (Fig. 3P,R). It is thought that these differences correlate to the differentiation state of the graft. As illustrated in Figure 3B, E11.5-E13 limb buds have already started chondrogenesis. Thus, a younger graft may be more flexible and suitable for a regenerative cell source.

Figure 3.

Skeletal pattern formed by transplantation of whole limb bud. (A) Forelimb buds were collected from GFP mouse embryos. Forelimbs of host mice were amputated at the digit joint level, and the distal halves of metacarpus were removed to prepare the space for grafting (red box). One limb bud was transplanted to each animal. (B) Developmental stages of limb buds used for grafting. Distal region of autopod was grafted at advanced stages. According to the developmental stages of the limb buds, cartilage pattern is gradually formed from proximal to distal (indicated in blue). (C–R) Skeletal pattern. E11 (C–F), E11.5 (G–J), E12 (K–N), E13 (O–R) limb buds were grafted. Animals were killed at day 10 (C, D, G, H, K, L, O, P) or day 20 (E, F, I, J, M, N, Q, R), fluorescent images of the limbs were taken after the removal of skin (C, E, G, I, K, M, O, Q), and limbs were stained by alcian blue/alizarin red (D, F, H, J, L, N, P, R). GFP signals indicate graft-derived tissues.

However, it is also thought that the grafts develop autonomously, like limb buds in organ culture (Kochhar & Aydelotte 1974). Thus, we tested whether or not the grafts interact with host tissues. An E11.5 forelimb bud was grafted and sections were prepared on day 7 (Fig. 4A–D). The graft-derived fluorescent tissues showed a continuous pattern with host tissues, and it was difficult to distinguish host tissues and graft-origin tissue in an HE-stained image (Fig. 4B) and a bright field image (Fig. 4D, overlaid image). We proposed that there must be some interaction between grafts and amputated host tissues, and then grafts differentiated according to the host tissue pattern.

Figure 4.

Interaction between host stump tissues and grafts. E11.5 forelimb bud was implanted into amputated autopod as shown in Figure 3A and killed at day 7. (A) Fluorescent image of autopod. Graft was observed in green. (B–D) Tissue section of the autopod. (B) Hematoxylin and alcian blue staining. (C) Fluorescent image. (D) Overlaid image of bright field and fluorescence. Tissues originating from the graft showed continuous pattern with host tissues.

Each limb bud cell has positional information along the anterior-posterior, proximal-distal and dorsal-ventral axes. This provides spatial identity of cells and decides developmental fate. Our next approach to prevent autonomous development of the graft is dissociation of the limb bud. Limb bud mesenchyme was completely dissociated and reaggregated by fibrin clotting and then implanted into the stump (Fig. 5A). In the reaggregates, limb bud cells were surrounded by differently fated cells, making autonomous development impossible. To examine the effect of amputation, clots with limb bud cells were also implanted into control normal limbs. The results clearly showed that grafted cells formed a digit-like structure at the stump (Fig. 5C,F,I,L), whereas the same cells differentiated into large cartilage aggregates at the control side (Fig. 5D,G,J,M). This strongly suggests the presence of signals from damaged host tissue and the response of grafted mesenchymal cells to the signals. The grafting of limb bud cells at advanced stages resulted in segmented and branched cartilage patterns (arrowheads in Fig. 5F,I,L), as observed in whole limb bud transplantation (Fig. 3).

Figure 5.

Implantation of dissociated limb bud cells. (A) After removal of ectoderm, limb bud mesenchymal cells were completely dissociated by pipetting and then reaggregated in fibrin clot. The reaggregate was implanted as in the case of whole limb bud transplantation. (B–M) Grafts were implanted into both limbs; one was implanted into the amputated right-forelimb, the other was implanted into normal left-forelimb and used as a control. Animals were killed at day 10. We examined with fluorescent images (B, E, H, K) and alcian blue/alizarin red staining. Compared to control grafts (D, G, J, M), the grafts implanted into the stump showed digit-like structures (C, F, I, L). Arrowheads indicate segmental structure in the grafts. Reaggregates were prepared from E11 (B–D), E11.5 (E–G), E12 (H–J), E13 (K–M) limb bud cells.

Alternative source of limb bud mesenchymal cells

Transplantation experiments showed the possibility of mammalian limb regeneration by the grafting of limb bud cells. However, using limb bud cells as a cell source for human limbs is practically impossible because of ethical issues and difficulty in the collection of a sufficient amount of cells. Thus, we tested several adult somatic cells as alternative cell sources. As a result, we found that bone marrow-derived mesenchymal cells behaved like limb bud mesenchymal cells (Fig. 6). Implanted bone marrow-derived mesenchymal cells formed a cartilage pattern at the stump. After 10 days, implanted cells had formed a digit-like cartilage structure distal to the amputated bone (Fig. 6B). In contrast, a graft that was constituted from embryonic femur cartilage cells differentiated into a huge cartilage mass with no definite pattern (Fig. 6C). Thus, femur cartilage cells could not contribute to amputated bone regeneration. It becomes apparent that bone marrow-derived mesenchymal cells have the ability to respond to amputation signals. Using these cells for autografting may help to improve mammalian regeneration.

Figure 6.

Alternative cells as sources for regeneration. (A, B) Bone marrow-derived mesenchymal cells were implanted into the stumps. Fluorescent image (A) and skeletal pattern (B) showed that implanted cells differentiated into cartilage and formed digit-like structures at the end of amputated bone. (C) Skeletal pattern of control limbs constituted of structures at the edge of amputated bone (arrowhead). (C) Skeletal pattern of control limbs constituted from femur cartilage cells of E16 embryos. These cells formed a huge cartilage mass.

Bone formation in the stump with BMP-7

No new cartilage/bone formation was observed after only amputation of limbs, although grafted limb bud cells or bone-marrow cells formed a cartilage pattern after implantation. Thus, we examined if the limb cells of neonatal mouse retain their chondrogenic potency by inserting gelatin gel containing BMP-7.

A gelatin gel containing BMP-7 was inserted into the space made by removing cartilage/bone at the zeugopod level. After 5 days, cartilage formation was observed and the cartilage changed to bone after 9 days (Fig. 7). The skeletal pattern formed distal to the amputated bone varied, but there was a tendency for bone in the proximal region to be large and bone in the distal region to be small, suggesting the formation of basic patterning. Thus, the limb tissues of neonatal mouse retain the potency of cartilage/bone formation.

Figure 7.

Induction of cartilage/bone element formation by BMP-7. (A) Limbs were amputated at the palm level (red arrows), and bone/cartilage at distal radius/ulna level and carpus was removed (black arrows). No formation of cartilage/bone was observed on the hypertrophied radius and ulna within 9 days. (B) Acidic gelatin gel was inserted into the space in the distal forearm and wrist. No formation of cartilage/bone was observed on the hypertrophied radius and ulna within 9 days. (C) Acidic gelatin gel containing 500 µg/mL BMP-7 was inserted into the space in the distal forearm and wrist. Several cartilage/bone elements were formed on the amputated radius and ulna (arrows) within 9 days.


To determine the potency of mouse limb regeneration, we amputated the limbs at increasingly proximal levels from the digit tip. Regardless of amputation level, the limbs of neonatal mice showed strong hypertrophy (Fig. 2F,I,J), although those of adult mice did not (Fig. 2G). This age-dependent hypertrophy indicates that the number of cells that can contribute to the bone formation process decreases in adult mice. Thus, adult stem cells and/or progenitor cells (ASCs) might be responsible for the hypertrophy. ASCs, which are known to differentiate into chondrocytes or osteoblasts, exist in the periosteum (Neufeld 1985; Nakahara et al. 1991; Bruder et al. 1994), bone marrow (Prockop 1997; Pittenger et al. 1999; Jiang et al. 2002), adipose tissue (Zuk et al. 2001) and skeletal muscle (Tamaki et al. 2002). Interestingly, the hypertrophy process of amputated neonatal mouse limbs was similar to the early step of the amphibian epimorphic regeneration process. By responding to the amputation signal, cell proliferation was activated around the stump and then proliferating cells were accumulated in the distal tip (Fig. 2). However, after the accumulation of proliferated cells, stump tissues did not show dedifferentiation or morphogenesis, such as segmentation or elongation to normal limb length. As a result, terminal phenotypes of the mouse limbs were different from those of regenerated amphibian limbs. It is known that non-cultured ASCs can contribute to various tissue regenerations (Kawada et al. 2004; Tamaki et al. 2005). ASCs may support the regeneration of amputated tissues, but are not available for epimorphic regeneration. It is thought that ASCs can respond to amputation signals and start proliferation, but that ASCs and/or the stump environment lack something for epimorphic regeneration.

To answer this question, we transplanted limb bud mesenchymal cells because these embryonic cells have pluripotency for limb cell differentiation and are sensitive to morphogenic signals, as are amphibian blastema cells. As a result, the implanted whole limb bud formed digit-like structures (Fig. 3). The intact graft showed a continuous pattern with the host tissues (Fig. 4). These results suggest that grafted tissues interacted with host tissues and formed digit-like structures. Furthermore, we prepared dissociated limb bud mesenchymal cells and implanted reaggregated mesenchymal cells to exclude the possibility of autonomous development. Dissociated cells lose the developmental fate of limb structures so that the graft can not develop autonomously, but digit-like structures were formed (Fig. 5C,F,I,L). Conversely, control reaggregates grafted to non-amputated limbs could not form any structures and only differentiated into nodules (Fig. 5D,G,J,M). Thus, the difference in the morphogenetic potency indicates the existence of some signals from the stump to the graft. All grafts implanted into stumps showed distalward extension. The distalward extension may indicate that stump tissue has positional information along the proximo-distal axis that provides signals from surrounding tissue or amputated bones.

In amphibian limbs, the cells at the amputation site dedifferentiate and form blastema, and then these blastema cells restart morphogenesis. Conversely, damaged tissues are repaired by existing stem cells in mammals. We stimulated ASCs to induce morphogenesis by the addition of growth factors, such as fibroblast growth factor (FGF), as morphogenic signals. These growth factors were found to induce cell proliferation or osteogenesis as previously reported (Kostakopoulou et al. 1996), but epimorphogenesis was not observed (data not shown).

We also showed that bone marrow-derived mesenchymal cells formed digit-like cartilage on the stump (Fig. 6). These cells were collected from the bone marrow osteoblastic zone where stem cells exist abundantly (Dobson et al. 1999; Arai et al. 2004; Suzuki et al. 2006) and were enriched by their adhesiveness to a plastic dish. These cells are thought to include primal mesenchymal stem cells because mesenchymal stem cells can be established in the same conditions (Peister et al. 2004). In contrast to the femur cartilage cells (Fig. 6C), bone marrow-derived mesenchymal cells showed interaction with host tissues (Fig. 6A,B). As described above, the stem cells can differentiate in response to damaged tissues as a regenerative source (Tamaki et al. 2005). Therefore, these stem cells might be an alternative for blastema cells.

We conclude that stump tissues exhibit some signals for responding cells to form digit-like structures, and that dissociated limb bud cells can act like stem cells or dedifferentiated cells as a cell source for limb regeneration. Our experiments suggest the possibility of autonomous mammalian limb regeneration by supplying pluripotent stem cells.

Finally, we applied BMP-7 by implanting gelatin containing BMP. Bone and cartilage elements were formed on the hypertrophied stump bones. BMP is known to induce cartilage/bone formation in mesenchymal tissue and muscle (Wozney et al. 1988). Thus, the formation of cartilage/bone itself is probable. However, the formation of limb skeletal pattern-like elements suggests the regeneration of patterned cartilage/bone from remaining tissue. A combination of the addition of other signal molecules may induce cartilage/bone structures with improved limb patterns. This finding also suggests that endogenous ASCs partially have the potency for reaction to morphogenic signals.

These results suggest that the environmental tissues of the stump allow the formation of cartilage and bone at least partially, and limb formation will be possible by supplying competent cells endogenously or exogenously.