Growth and differentiation of a long bone in limb development, repair and regeneration



Repair from traumatic bone fracture is a complex process that includes mechanisms of bone development and bone homeostasis. Thus, elucidation of the cellular/molecular basis of bone formation in skeletal development would provide valuable information on fracture repair and would lead to successful skeletal regeneration after limb amputation, which never occurs in mammals. Elucidation of the basis of epimorphic limb regeneration in amphibians would also provide insights into skeletal regeneration in mammals, since the epimorphic regeneration enables an amputated limb to re-develop the three-dimensional structure of bones. In the processes of bone development, repair and regeneration, growth of the bone is achieved through several events including not only cell proliferation but also aggregation of mesenchymal cells, enlargement of cells, deposition and accumulation of extracellular matrix, and bone remodeling.


Bone, a mineralized and vascularized tissue characteristic of vertebrates, is the main component of the skeleton, which supports the vertebrate body. In the Preface of his book “BONES and CARTILAGE”, Brian Hall (2005) noted, “But the skeleton is more than an articulated set of bones: its three-dimensional conformation establishes the basis of our physical appearance; its formation and rate of differentiation determine our shape and size at birth; its postnatal growth orders us among our contemporaries and sets our final stature”. Thus, bone formation, a continuous process during embryonic and postnatal stages, is a central issue for morphogenesis and size control of the vertebrate body. In this review article, we present an overview of the process of bone formation during its development and regeneration, and focus on bone formation of the limb skeleton, a long bone in particular.

Lateral appendages (limbs) in tetrapod (amniotes plus amphibians), consist of bones of many shapes; long, short, round, flat and indefinable-shape bones. For instance, the longest bone in the human body is the femur, a skeletal element of the stylopod in the hind limb. Bones in the limb develop during embryogenesis and re-develop during wound repair and regeneration after fracture or traumatic limb amputation. These two processes, development and regeneration of limb skeleton, share many cellular and molecular mechanisms and also have independent mechanisms.

The embryonic primordium of the tetrapod limb is called the limb bud, which is composed of mesodermal cells and an overlying ectoderm. Skeletal elements in the limb contain mesenchymal cells derived from the lateral plate while myogenic cells in muscle tissue are from the myotome of the somites. The limb bud emerges and protrudes at the lateral trunk, and outgrows laterally. The primordium of the regenerating limb is called a blastema, which is composed of undifferentiated mesenchyme derived from several types of cells, including bone, cartilage, dermis, connective tissue and muscle, and an overlying epithelium derived from the skin epidermis. This regenerative structure is considered to be equivalent to the limb bud since they share many cellular and molecular properties. In spite of their similarity, the origin of blastemal mesenchyme is not embryonic but derived from adult tissues, indicating that the initial process of blastema formation is not identical to that of the limb bud formation but includes dedifferentiation of adult mature tissues and/or stimulation of quiescent stem cells. In both the limb bud and the blastema, undifferentiated mesenchymal cells proliferate to develop or restore the limb skeleton. The outgrowth of the limb bud/blastema is known to be regulated by several key signaling molecules such as FGF, Wnt and Shh, and the early events and molecular cascades for their outgrowth are described elsewhere in other reviews (Stoick-Cooper et al. 2007; Nacu & Tanaka 2011). In the next section, we focus on relatively late events, cartilage and bone formation, in the developing limb bud, where we do not consider skeletal pattern and morphology of each skeletal element in the limb but focus on the developmental process of formation of a bone.

Growth and differentiation of cartilage and bone in a long bone

Bones in vertebrates are formed via two different processes; intramembranous and endochondral ossification (Fig. 1). Intramembranous ossification produces osseous tissues directly from mesenchymal progenitors. In the case of long bones in mammals, this process generates the bone collar and sequentially increases its diameter (Erlebacher et al. 1995) . In contrast, endochondral ossification progresses by replacing a cartilaginous mold with bone tissues, and is essential for epiphyseal morphogenesis and longitudinal growth of long bones. During long bone development, a skeletal element is formed initially as a condensation of precartilage cells in a histologically homogenous population of mesenchymal cells. Then cells in the condensation undergo differentiation and maturation into chondrocytes and perichondrial cells. Chondrocytes go through further growth and differentiation via the growth plate (ordered layers of several differentiation states of chondrocytes). Subsequently intramembranous ossification of the perichondrium and endochondral ossification of cartilage occur, followed by development of the marrow cavity. In postnatal life, longitudinal growth of the metaphyses continues until closure of the growth plate, and concomitantly circumferential growth of diameter goes on.

Figure 1.

An illustration of the development of a long bone. (A) Mesenchymal cells aggregate to form a precartilage condensation, which is a primordium of a long bone. (B) From the center of the primordium, mesenchymal cells differentiate into chondrocytes. Chondrocytes start to produce cartilaginous extracellular matrix (ECM). Condensation proceeds particularly around the presumptive epiphysis region. At the periphery of the cartilaginous anlage, mesenchymal cells begin to differentiate into perichondrial cells. (C) Chondrocytes undergo further maturation and become hypertrophic chondrocytes from the center of the cartilaginous anlage. The whole skeletal primordium including the articular surface becomes surrounded by the perichondrium, which eventually becomes bi-layered at the diaphyseal region. (D) Some perichondrial cells in the inner layer of perichondrium differentiate into osteoblasts and form the bone collar via intramembranous ossification. The zone of hypertrophic chondrocytes becomes vascularized, and perichondrial cells migrate into the anlage and differentiate into osteoblasts. Hypertrophic chondrocytes undergo apoptotic cell death and are replaced by osteoblasts, which subsequently form trabecular bones (endochondral ossification), resulting in formation of the marrow cavity. A small population of hypertrophic chondrocytes in a particular region are shown to differentiate into osteoblast-like cells; however, those cells are not a major source of osteoblasts (indicated by a dashed arrow). Proliferating, prehypertrophic and hypertrophic chondrocytes are arranged in columns. This chondrogenic zone is called the growth plate. Endochondral ossification spreads outward. The perichondrium transforms to the periosteum along the bone collar. (E) Eventually, endochondral ossification starts also in both epiphyses. Around this period, the perichondrium disappears from both articular ends, and the articular cartilage begins to differentiate.

Growth of skeletal elements occurs by the combination of two modes; appositional and interstitial growth. The former is growth by cell proliferation and addition of cells in the primordium, and the latter is growth by enlargement of the volume of substance (chondrocyte enlargement [e.g. hypertrophy] and extracellular matrix [ECM] deposition) (see also Williams et al. 2008). In the following section, we review these growth modes in each developmental phase.


Condensation is the aggregation of cells (Hall & Miyake 1995), and aggregation of mesenchymal cells into precartilage condensations is one of the earliest events in skeletogenesis (Fell 1925), prefiguring the skeletal elements. Consequently, cell density of the aggregate increases greatly (Thorogood & Hinchliffe 1975; Wezeman 1998), which is required for chondrogenesis. An experiment using cells from chick limb buds demonstrated that cell density is correlated with the extent of chondrogenesis in vitro (Ahrens et al. 1977; San Antonio & Tuan 1986), suggesting that the condensation is an important phase for chondrogenesis.

The condensation is not accompanied by and not due to a local increase of cell proliferation in the aggregate. In other words, the proliferating rate of cells in presumptive skeletal elements is comparable to that of surrounding cells before condensation (see Ede 1983). For the successful condensation growth, aggregation of cells toward a center and failure of cells to disperse from a center appear to be critical (Hall & Miyake 2000), which are mediated by cell–cell interaction by cell adhesion molecules such as N-cadherin and N-CAM (Chuong et al. 1993; Widelitz et al. 1993; Oberlender & Tuan 1994; Haas & Tuan 1999) and cell–ECM interaction by hyaluronan, tenascin and fibronectin (see De Lise et al. 2000).

For example, hyaluronan is considered to facilitate movement of mesenchymal cells and to be necessary for initial condensation. Meanwhile hyaluronan is suggested to be associated with inhibition of mesenchymal differentiation, and the removal of hyaluronan seems to enable cell–cell contact and interaction that is considered to be important for subsequent chondrocyte differentiation (Toole et al. 1972; Kosher et al. 1981; Knudson & Toole 1985; Li et al. 2007; see also Hall & Miyake 1992). Fibronectin, a player for cell–matrix interaction, promotes attachment and spreading of cells over the substratum, which seems to be necessary for condensation but inhibits chondrogenesis (Swalla & Solursh 1984; Zanetti et al. 1990; Gehris et al. 1997). TGF-β (transforming growth factor-β) promotes the expression of some of these molecules (N-cadherin, N-CAM, tenascin, and fibronectin) (Chimal-Monroy & Diaz de Leon 1999). Contrary to adhesion molecules discussed above, Syndecan-3 functions in regulating the switch from the mesenchymal condensation phase to a more differentiated state, which is supposedly mediated by repressing fibronectin function, and thus enables cells in the mesenchymal condensation to differentiate into chondrocytes, setting the boundary of the skeletal primordium (Koyama et al. 1996; Seghatoleslami & Kosher 1996). N-cadherin, N-CAM, hyaluronan and fibronectin are downregulated as expression of genes associated with chondrocyte differentiation begins. At the same time that cells of the skeletal primordium begin to differentiate, cells at the periphery of the forming skeletal element, particularly at the presumptive epiphyseal region, keep aggregating and expressing high levels of maker molecules for the less differentiated state (such as N-cadherin). This contributes to continuous aggregation of mesenchymal cells resulting in growth of the skeletal primordium (Oberlender & Tuan 1994).

Chondrocyte differentiation

Subsequent to mesenchymal condensation, cells located at the center of skeletal primordium differentiate into chondrocytes, which accompanies the morphological changes from fibroblastic to round (Fell 1925; Streeter 1949; Rooney et al. 1984). Cells at the periphery of the skeletal primordium form the perichondrium, which contains osteochondroprogenitor.

Proliferation of chondrocytes at this phase appears to modulate the appositional growth of the skeletal primordium, and to be regulated by insulin-like growth factor (IGF) signaling. Overexpression of IGFBP2 (IGF binding protein 2) in the chick limb bud results in shortening of the long bones without affecting the skeletal morphology, and is accompanied by a reduction of chondrocyte proliferation in the cartilaginous anlage (Fisher et al. 2005). Since IGF-I and IGFBP2 have opposing effects on chondrogenesis (McQueeney & Dealy 2001), a balance of endogenous IGF/IGFBP2 might be required for the appropriate pattern of growth of the cartilaginous anlage.

Changes of ECM composition contribute to the mode of interstitial growth. Condensed mesenchymal cells cease expressing some ECMs, after which differentiating chondrocytes gradually start to secrete other kinds of ECM molecules, including cartilage-specific isoform of collagen type II, collagen type IX and XI, matrix Gla protein, aggrecan and link protein (Mallein-Gerin et al. 1988; Nah et al. 1988; Stirpe & Goetinck 1989; Kulyk et al. 1991; Tsonis & Walker 1991; Swiderski & Solursh 1992; Luo et al. 1995; Yoshioka et al. 1995; Imamura et al. 2000). Collagen type I expression becomes restricted to perichondrium (Caplan & Pechak 1987). The analyses of mutant mice demonstrated that Sox9 is an essential transcription factor for chondrogenesis and one of earliest known marker molecules for this process (Bi et al. 1999; Akiyama et al. 2002). Its expression begins in cells undergoing condensation, continues during chondrogenesis, and ceases as cells begin to differentiate into prehypertrophic chondrocyte (Wright et al. 1995; Ng et al. 1997; Zhao et al. 1997). Data from in vitro micromass culture indicate that Sox9 is dispensable for initiation of condensation but is necessary for maintenance of it, leading to the further differentiation (Barna & Niswander 2007). Sox9 also inhibits the transition into hypertrophic chondrocyte, and its expression is excluded from hypertrophic chondrocyte (Bi et al. 2001; Akiyama et al. 2002, 2004). Two additional Sox family members, L-Sox5 and Sox6, are co-expressed with Sox9 during chondrogenesis. These three Sox proteins are required for the expression of some of the above-described ECM proteins prior to matrix deposition (Lefebvre & de Crombrugghe 1998; Lefebvre et al. 1998; Smits et al. 2001; Han & Lefebvre 2008), and for further chondrocyte differentiation and proliferation (Smits et al. 2001, 2004). Sox expression is regulated by Bmp signaling during chondrogenesis. Conditional knockout mice of Bmp receptors lack the expression of these three Sox proteins and decrease proliferation and cell survival in condensation, resulting in a severe and generalized chondrodysplasia (Yoon et al. 2005).

During differentiation into early chondrocytes, blood vessels are excluded from the condensation area, and this avascular state is required for proper chondrogenesis (Yin & Pacifici 2001). As a cartilaginous anlage transitions to the subsequent steps for bone formation, it becomes secondarily vascularized by invasion of blood vessels at the primary and secondary ossification centers (see below).

Growth plate development and ossification

At the next stage in bone formation, progressive maturation occurs from the center of a skeletal primordium and progresses towards the articular ends (Fell 1925; Streeter 1949; Rooney et al. 1984). In the center of the cartilaginous anlage, proliferating chondrocytes stop dividing and start enlarging to become hypertrophic chondrocytes (see Hall 2005; Gilbert 2010). Hypertrophic chondrocytes synthesize collagen type X, calcify the ECM around them and subsequently become vascularized, after which they start undergoing apoptotic cell death, and simultaneously osteoblasts appear at the site. Osteoblasts replace dying hypertrophic chondrocytes and produce bone matrix such as collagen type I. Subsequently, osteoblasts are embedded in the deposited bone matrix and become osteocytes of trabecular bones (Erlebacher et al. 1995). The same processes also occur in the epiphyses (the secondary ossification center), such that a zone of chondrocytes persists only in the metaphyses between the primary and secondary ossification centers. This chondrogenic zone is the only region that is responsible for the longitudinal growth of a long bone and called the growth plate. It remains until adulthood, at which time it becomes ossified and the long bone stops growing in length.

The growth plate is organized spatially from the epiphysis towards the diaphysis in the following order; a zone of periarticular proliferating chondrocytes, a zone of columnar proliferating chondrocytes, a zone of prehypertrophic chondrocytes and a zone of hypertrophic chondrocytes. Longitudinal growth of long bones mainly depends on enlargement of cell size in the zone of hypertrophic chondrocytes (Thorogood 1983; Noonan et al. 1998). Interestingly, short bones such as the carpals do not undergo this process (Wolpert 1982). Premature chondrocytes undergo successive differentiation in this order through the growth plate, resulting in longitudinal growth of skeletal primordium. Cells in the transitional zone between periarticular and columnar chondrocyte proliferate at the highest rate, and the rate of proliferation decreases towards the center of the diaphysis, suggesting that cells originating in this zone mainly give rise to the zone of columnar proliferating chondrocytes and furthermore the zone of hypertrophic chondrocytes (Smits et al. 2004).

Differentiation and proliferation of cells in the growth plate are regulated primarily by Ihh/PTHrP signaling. PTHrP (parathyroid hormone-related protein) is a paracrine factor expressed near the articular end and plays a role in maintaining cells in a premature state by suppressing hypertrophy through activation of Sox9 (Karaplis et al. 1994; Lee et al. 1995; Weir et al. 1996; Lanske et al.1999; Huang et al. 2001; Guo et al. 2002). Expression of PTHrP at the articular end is induced by a secretory factor, Ihh (Indian hedgehog) which is expressed in prehypertrophic and early hypertrophic chondrocytes. Ihh promotes chondrocyte proliferation in the growth plate (St-Jacques et al. 1999; Karp et al. 2000; Long et al. 2001; Hilton et al. 2007). Ihh signaling also regulates the transition from periarticular into columnar proliferating chondrocytes (Kobayashi et al. 2005; Hilton et al. 2007). Development of the growth plate is also regulated by other signaling pathways, including Wnt, Bmp, FGF, Notch and Runx (see Kronenberg 2003; Long & Ornitz 2013), some of which are associated with the Ihh/PTHrP regulatory axis, although others are independent of it.

Concurrently with chondrocyte maturation, cells in the inner layer of the perichondrium surrounding the hypertrophic chondrocyte zone, differentiate into osteoblasts and form bone collar. In addition, some but not major populations of hypertrophic chondrocytes locating between cartilage and osteogenic tissues are shown to have differentiation potential for osteoblast-like cells and do not undergo apoptosis (Maes et al. 2010; see also Bianco et al. 1998). In both cases, osteoblasts are derived from Sox9-positive mesenchymal progenitors originated from the lateral plate mesoderm (Akiyama et al. 2005; Day et al. 2005). This process is accompanied by vascularization of the zone of hypertrophic chondrocyte that is regulated by VEGF (vascular endothelial growth factor) (Gerber et al. 1999; Carlevaro et al. 2000). In association with vascular invasion, perichondrial cells migrate into the anlage and differentiate into osteoblasts, which develop trabecular bones, resulting in establishment of the primary ossification center (Maes et al. 2010). In this manner, perichondrial cells contribute to both bone collar and trabecular bones.

The master transcription factor Runx2 and its downstream target Osterix/Sp7 are required for osteoblast differentiation (Ducy et al. 1997; Komori et al. 1997; Otto et al. 1997; Nakashima et al. 2002). Runx2 expression is regulated by Ihh and Wnt/β-catenin signaling in osteogenesis (St-Jacques et al. 1999; Long et al. 2004; Mak et al. 2006; for review Kronenberg 2003). Runx2 also functions in the regulation of chondrogenesis in the growth plate. Thus, Runx2 is necessary for bone development by regulating osteoblast differentiation, chondrocyte proliferation and hypertrophy. Another transcription factor ATF4 (activating transcription factor 4) regulates expression of osteocalcin (Yang et al. 2004), which is one of the bone matrix proteins. During osteoblast differentiation, ATF4 can interact with Runx2 through Satb2 to synergize their activities (Dobreva et al. 2006). ATF4 promotes the import of amino acids required for protein synthesis such as collagen type I by osteoblasts (Yang et al. 2004). ATF4 also upregulates expression of RANKL (receptor activator of nuclear factor [NF]-κB ligand) (Elefteriou et al. 2005), which is essential for differentiation of osteoclasts. A portion of the osteoblasts differentiate into osteocytes that are embedded within the bone matrix; whereas, most osteoblasts differentiate into inactive bone-lining cells or undergo apoptosis (Bonewald 2011).

Bone homeostasis and fracture repair in mammals


After development, bone is continuously replaced throughout life. Osteoclasts, osteoblasts and osteocytes interact and control resorption and replacement of old bone. Bone remodeling is a balance between formation by osteoblasts and resorption by osteoclasts. Remodeling controls the bone shape and size during appositional growth, and repairs micro-damaged bones and fractured bones. It is accomplished by the basic multicellular units (BMU), which contain osteoblasts, osteoclasts, blood vessels, nerves and connective tissue. In the BMU, osteoclasts and osteoblasts make cell–cell contact and communicate via cell membrane proteins, such as EphB4 and ephrinB2 (Zhao et al. 2006; Matsuo & Otaki 2012), and paracrine factors, such as RANKL (Kobayashi et al. 2009). They also interact with bone matrix. An imbalance between bone resorption and formation results in bone remodeling diseases such as osteoporosis and osteopetrosis (Teitelbaum 2000; Boyle et al. 2003; Harada & Rodan 2003).

Fracture repair

Bone fracture occurs by traumatic accidents and diseases such as osteoporosis and cancer. Although the repair process from bone fracture is partially a recapitulation of embryonic bone development (Vortkamp et al. 1998, and see above) as well as an enhanced mode of bone homeostasis, the inflammatory responses and local cell proliferation are unique to the repair process.

In long bone fracture, the periosteum (the outer layers of cortical bone), endosteum (the inner layers of cortical bone) and blood vessels are also damaged. Several lines of evidence suggest that the periosteum and the bone marrow are the main local sources of multipotent skeletal stem cells/progenitors for bone repair (Utvåg et al. 1996; Ozaki et al. 2000; Colnot 2009; Murao et al. 2013). Myogenic cells can also be a source of cells for open fracture healing, but not closed fracture healing (Liu et al. 2011). The periosteum is composed of an outer fibrous layer containing cells that synthesize collagen fibers and an inner cambium layer consisting of osteoblasts, fibroblasts, and mesenchymal progenitor cells with the capacity to differentiate into osteoblasts and chondrocytes (Squier et al. 1990). The multipotent mesenchymal progenitor cells in the periosteum become committed to be Sox9-expressing osteochondroprogenitor cells shortly after a fracture occurs, and differentiate into chondrocytes, osteoblasts, and periosteal cells of a newly formed bone (Murao et al. 2013).

Bone fracture repair consists of a series of phases: inflammation, soft callus formation, hard callus formation and bone remodeling. During these phases, mesenchymal progenitor cells are activated to undertake migration, proliferation and differentiation to grow the bridging structure between broken bones as follows.

Inflammation phase

The mesenchymal progenitor cells are activated to proliferate and become recruited to the fracture site by inflammatory responses. Soon after bone fracture occurs, blood cells flow into the fracture site, and the hematoma (fibrin clot) is formed. TGF-β 1 and PDGF (platelet derived growth factor) are released into the fracture hematoma by platelets and recruit macrophages and other inflammatory cells (see Stocum 2012). Pro-inflammatory cytokines, such as TNF-α (tumor necrosis factor-α) and IL-1 (interleukin-1), peak in expression within the first 24 h after fracture. Both TNF-α and IL-1 are expressed primarily in macrophages and other inflammatory cells during the early inflammation phase. They enhance ECM synthesis, stimulate angiogenesis and recruit mesenchymal progenitor cells into the hematoma (Kon et al. 2001). Proliferating mesenchymal cells are observable in the inflammatory reaction region during the early inflammation phase (Fig. 2A). Subsequently, proliferating cells are localized in the fibrous layer and the cambium layer of the periosteum (Iwaki et al. 1997) (Fig. 2A,B). The cambium layer thickens to produce a mass of cartilage that eventually develops into an external callus (Fig. 2B,C) (see Cameron et al. 2013). Since the distribution of the proliferating cells is wide and homogeneous, it is likely that their proliferation is mediated by broadly diffused factors, such as platelet-derived TGF-β (Iwaki et al. 1997).

Figure 2.

An illustration of fracture healing processes and distribution of proliferating cells. (A) The hematoma is formed at the fracture site, into which inflammatory cells including macrophages are attracted. Mesenchymal progenitor cells are recruited into the hematoma by inflammatory cell-derived cytokines. Mesenchymal progenitor cells in the hematoma and cambium layer of the periosteum are proliferative during inflammation phase. (B) In the early phase of soft callus formation, the hematoma is invaded by blood vessels and osteochondroprogenitor cells, and become the granulation tissue. Those osteochondroprogenitor cells proliferate and differentiate into chondrocytes. The periosteum near the fracture edge thickens and forms a mass of cartilage outside of the granulation tissue. Some periosteal osteochondroprogenitor cells directly differentiate into osteoblasts and undergo intramembranous ossification. (C) Chondrocytes near the fracture edge and in fibrocartilage proliferate during the soft callus formation phase. Osteoblasts in the region of intramembranous ossification also proliferate. The callus attains its maximum size. (D) During hard callus formation phase, trabecular bones are formed via endochondral ossification in the callus, resulting in formation of woven bone. Subsequently, the woven bone is replaced by the lamellar bone and remodeled so that the normal bone shape and size are restored.

Soft callus formation phase

Before soft callus formation, the hematoma becomes invaded by newly formed blood vessels, immune cells and fibroblasts, resulting in formation of granulation tissue (see Cameron et al. 2013). Presumably these fibroblasts include the osteochondroprogenitor cells. The granulation tissue is replaced with fibrovascular tissue, and the progenitor cells differentiate into chondrocytes. This fibrocartilage tissue is called the soft callus, which functions to stabilize the broken bone ends. In this phase, proliferating cells are found in the population of osteochondroprogenitor cells and chondrocytes at the fracture edges (Fig. 2C). These cells proliferate until they become mature hypertrophic chondrocytes (Iwaki et al. 1997). Since the subsequent endochondral ossification takes place throughout the whole soft callus as a scaffold, the size of the bone healing tissue attains its maximum size by this time. As described above, Ihh signaling plays a central role in chondrocyte proliferation and differentiation in bone development. In fracture healing, chondrocytes in the soft callus express Ihh and its receptor, Patched-1, is expressed in the less differentiated chondrocytes around the callus, suggesting that Ihh signaling also functions in regulating endochondral ossification during fracture repair (Vortkamp et al. 1998). In addition, periosteal bone callus formation is significantly reduced in Smoothened-deleted mice, which indicates a significant role of Hh signaling in fracture repair (Wang et al. 2010). As for chondrocyte differentiation, Bmp-2 is essential, since callus formation does not occur in the absence of Bmp-2 (Tsuji et al. 2006). However, Bmp-2 knockout mice have normal skeletons, indicating that there are differences between skeletal development and fracture healing.

Hard callus formation phase

After the soft callus forms, cartilaginous and fibrovascular tissues are replaced by primary bones or woven bones through endochondral ossification (see 'Growth and differentiation of cartilage and bone in a long bone'). Woven bones consist of randomly oriented collagen fibers and are relatively weak in structure, compared to lamellar bones (see the next sub-subsection) (Shapiro 2008). In transition to hard callus formation, ingrowing vascular endothelial cells and osteoblasts on the newly formed trabecular bones are proliferative (Iwaki et al. 1997) (Fig. 2D).

In addition to endochondral ossification, intramembranous ossification is also involved in hard callus formation. Soon after injury, osteochondroprogenitor cells in the periosteum near the fracture edge differentiate directly into osteoblasts and start to produce bone matrix. Intramembranous ossification occurs continuously, and a periosteal hard callus is formed around the fracture edge. As a result, broken bones are bridged by the woven bone callus derived from both endochondral and intramembranous ossification.

Bone remodeling phase

Woven bones in the hard callus are replaced gradually by secondary bones or lamellar bones, which consist of orderly organized collagen fibers and are mechanically strong, by remodeling with osteoblasts and osteoclasts as described above for the process of bone homeostasis (Shapiro 2008). The balance of proliferation and apoptosis is shifted during this phase such that apoptosis of osteocytes is more active resulting in the elimination of cells and triggering the bone remodeling (Li et al. 2002). Eventually the normal bone shape and size are restored, although the remodeling phase continues for several years (Mountziaris & Mikos 2008; Cameron et al. 2013).

Fracture healing in amphibians

The mechanisms of bone fracture healing are highly conserved among mammal and amphibian species (see Cameron et al. 2013). Since amphibians are capable of regenerating many tissues and organs, they are animals of interest as an alternative model to study fracture healing. In spite of having an impressive regeneration ability, it appears that they use the same molecular mechanisms as mammals in fracture healing, and as is the case in mammals, there is a certain size of a gap between broken bones that is too large to undergo bone repair, called a critical size defect (CSD) (Hutchison et al. 2007; Satoh et al. 2010a; Feng et al. 2011).

Although bone fractures heal through similar processes, the timing of the transition between the four phases is different among species. In adult mice, a soft callus is formed by a week after fracture, and ossification of the callus is seen in about 2 weeks (Kon et al. 2001). In contrast, in Xenopus frogs a soft callus is formed quickly but is slow to ossify. As indicated in Figure 3, the whole callus is still cartilaginous 30 days after fracture, and there are no signs of either hard callus formation or intramembranous ossification from the periosteum. In the Mexican axolotl (a salamander), soft callus formation is even slower than Xenopus and takes 6–8 weeks (Hutchison et al. 2007). Interestingly, those animals show diverse ability of limb regeneration as described below.

Figure 3.

Fracture callus morphology in Xenopus froglets and mouse neonates. (A–C) Sections of Xenopus radio-ulna bone fracture. Five days post fracture (dpf) (A), 10 dpf (B) and 30 dpf (C). Xenopus forms a soft callus broadly from the fracture site, but it is slow to ossify. Formation of intramembranous bone around the fracture site is not obvious. (D) A section of mouse ulna bone fracture at 5 dpf. Simultaneously with soft callus formation, formation of intramembranous bone occurs. In all sections, elbow joint is on the left side. Black, red and blue dashed lines indicate bone shafts, outline of the soft callus and intramembranous bone areas, respectively. Skeletal tissues are indicated; R/U, radio-ulna; R, radius; U, ulna. Scale bars = 250 μm. All sections were stained with hematoxylin, eosin and alcian blue.

Epimorphic limb regeneration

In addition to fracture healing, bone is also regenerated in response to severe trauma to the limb in animals such as urodele amphibians, which can regenerate an entire amputated limb. After limb amputation, undifferentiated and proliferative cells residing within the limb stump accumulate on the amputation surface to form a regeneration blastema (see Carlson 2007). A blastema ultimately re-develops into a completely patterned and functional limb. This process appears to be a recapitulation of limb development, and blastema outgrowth and pattern formation are associated with gene expression and tissue interactions that are comparable to those in the developing limb bud (Bryant et al. 2002; Gardiner et al. 2002).

Compared to urodeles, regenerative ability is more limited in anuran amphibians. There are two general phenotypes of limb regeneration in adult anurans; non-regenerative and partially regenerative (see Stocum 1995). In some species, such as Ranidae, the skin heals over the amputation plane and a blastema does not form. Partial regeneration occurs in adult Xenopus frogs that form a blastema in response to limb amputation, but it eventually forms only a single cartilage rod, called a spike. Thus, Xenopus frogs have the ability to form a blastema even though the blastema is pattern formation-deficient (Suzuki et al. 2006).

During amphibian limb regeneration, cartilaginous structures form in the stump as well as in the blastema. In urodeles, small masses of cartilage are observed at the cut end of the cortical bones, before the missing skeletal elements are regenerated from the blastema. These extra structures become less evident as regeneration proceeds, and are not evident by the time that limb regeneration is complete. Ectopic cartilage formation around the amputated bone is also observed during Xenopus spike formation. Compared to urodeles, these structures are more prominent and extend almost to the remaining proximal epiphysis. With time, this cartilaginous callus becomes ossified, whereas, the regenerated spike remains cartilaginous except for a thin layer of cortical bone on the surface of the proximal base. Presumably this newly formed cartilage in the stump is the equivalent structure to the soft callus of fracture healing because of the similarity in appearance and position relative to the cut edge of the bone.

Skeletal regeneration in a blastema

Critical signals that are required for blastema formation are released from nerves. When nerves are severed at the shoulder level so as to denervate a whole forelimb during the early phase of blastema formation, the forelimb fails to regenerate (see Wallace 1981; Stocum 2011; Kumar & Brockes 2012). In such denervated limbs, the amputation surface becomes covered by regenerated skin with a differentiated dermis. However, in Xenopus, a cartilaginous callus around the stump bone is still formed in denervated limbs, indicating that this is a nerve-independent response. There are at least two possible pathways by which nerves induce and maintain blastema formation. One is that nerves produce mitotic factors for blastema cells (Brockes 1984; Mescher 1996). Several factors have been reported to meet the criteria for this function (Brockes & Kintner 1986; Mullen et al. 1996; Mescher et al. 1997; Wang et al. 2000; Kumar et al. 2007; Makanae et al. 2013). The other is that nerves induce the wound epithelium to function as the regeneration-permissive epithelium, the apical epithelial cap (AEC) (Satoh et al. 2008; Stocum 2011). The AEC is an analogous tissue to the AER (apical epidermal ridge) of developing limb buds, and secretes mitotic factors for blastema cells, such as FGFs (Christen & Slack 1997; Endo et al. 2000; Han et al. 2001; reviewed by Stoick-Cooper et al. 2007; Satoh et al. 2009). It also is reasonable to consider that proliferation of blastema cells during the initial phase of regeneration is maintained by the combination of these two pathways.

The fact that signals from nerves are sufficient to stimulate proliferation of the blastema precursor cells is demonstrated by the accessory limb model (ALM) (Endo et al. 2004). In this model, a piece of full-thickness skin is removed from the lateral side of the upper arm, and brachial nerves are surgically deviated to the wounded site. Signaling from the nerve and wound epithelium induce formation of a blastema that eventually regresses without making a limb. However, if a piece of skin from the opposite side of the limb is grafted to the nerve-deviated wound, then a well-patterned limb develops ectopically from the wound.

These ectopic blastemas are formed by the recruitment and proliferation of cells from the adjacent dermis and connective tissues in response to the nerve signals (Endo et al. 2004), as is the case when a blastema is formed on an amputated limb stump (Muneoka et al. 1986; Kragl et al. 2009; see Monaghan & Maden 2013 for review). Dermal and connective tissue fibroblasts are also the initial population of cells that migrates underneath the wound epithelium prior to cell proliferation (Gardiner et al. 1986). Dermis-derived blastema cells eventually differentiate into several types of tissues, such as cartilage, dermis and remaining connective tissues in normal limb regeneration (Muneoka et al. 1986) and the ALM (Hirata et al. 2010). In addition to blastema cells derived from connective tissue fibroblasts, there is cellular heterogeneity in the blastema that arise from contribution from other tissues in the stump such as muscles, nerves and vasculature (Kragl et al. 2009; Monaghan & Maden 2013). Since skeletal pattern can be regenerated independently of the existence of muscles (Dunis & Namenwirth 1977; Lheureux 1983), presumably it is the blastema precursor cells derived from fibroblasts that are stimulated to proliferate by the signals from the nerves and the AEC. They redifferentiate into the connective tissues including the dermis and cartilages of the regenerate. Precursor cells for other tissues, such as muscles and blood vessels, migrate into the blastema at the later stages of regeneration as occurs during limb development (Bryant et al. 2002; Nacu et al. 2013). Selective cell proliferation from different tissues might be achieved by tissue-specific expression of receptor genes for nerve factors. For example, FGF receptor 2 is expressed by the cells of the cartilage lineage in newts (Poulin et al. 1993; Poulin & Chiu 1995). Therefore, blastema precursor cells for skeletal cartilages might be activated preferentially.

Skeletal regeneration in the stump

In addition to nerve-dependent growth of cartilage precursor cells in the blastema, the ALM has provided insights as to how regenerated skeletal elements integrate into the host tissues. In the ALM, accessory limbs generally have a normal pattern at more distal segments of the skeletal tissues, but do not form the proximal structures that correspond to the level at which the accessory limbs were made (Maden & Holder 1984; Endo et al. 2004; Satoh et al. 2010b). Proximal structures are formed and connected to the host bone only when the ALM surgery involves making a deep wound in which the host bone is intentionally injured (Satoh et al. 2010b). As described above, injury to bones induces a fracture healing response, resulting in formation of a soft callus at the injured site. It seems that formation of a soft callus is necessary to connect the blastema-derived skeletal tissues and the host skeletal tissues. Similarly, it is likely that cartilaginous callus formation at the bone edge in amputation-induced limb regeneration function to connect the blastema-derived skeletal cartilages and the stump-derived skeleton.

In contrast to amphibians, mammalians are considered to be non-regenerative animals in terms of limb regeneration. Their regeneration ability is largely restricted to the digit tip, and amputation of the limb in rodents at more proximal levels results in failure of blastema formation. Nevertheless, there are some changes in the stump in response to limb amputation. A cartilaginous callus that subsequently ossifies forms around the cut end of the stump bones, which is noticeable in neonate (Masaki & Ide 2007). This callus is similar in appearance to what is observed in denervated Xenopus limbs, except that those in mouse neonates expand only laterally but not proximally. We have tested whether or not cartilaginous calluses also are formed in denervated mouse limbs and found that denervation does not affect cartilaginous callus formation (S. Miura, T. Endo, unpubl. obs).

Although nerve-independent callus growth in mammals may be related to the inflammatory responses, chronic inflammation in amputated amphibian limbs generally has negative effects on regeneration (King et al. 2012; Mescher et al. 2013). Inflammation and its resolution are controlled by the balance of different types of macrophages (reviewed by Mahdavian Delavary et al. 2011). Since depletion of macrophages in axolotls during the initial phase after limb amputation results in failure of limb regeneration, correct regulation of inflammation might be necessary for regeneration (Godwin et al. 2013). It may be that these two modes of skeletal regeneration (callus formation and epimorphic regeneration) commonly use some but not all of the steps of the inflammation response. There are similarities between the initial period after amputation and the inflammatory phase of mammalian wound healing (Mescher 1996), and thus the early events of inflammation, such as production of non-myeloid-derived reactive oxygen species (ROS), might be necessary for both modes (Niethammer et al. 2009; Love et al. 2013). After that, their cascades might diverge and respond differently to the later stages of the inflammatory reaction.

If callus growth at the cut end of the stump bones is a regenerative response that occurs even in mammals, then the difference between mice and Xenopus in the ability to grow a blastema might be attributed to differences in nerve-dependent responses. As discussed above, possible targets of nerve signals are the wound epithelium and blastema progenitor cells. Mice primarily have some problems that prevent the physical interaction between the wound epithelium and nerves from the stump just after amputation. For example, a blood clot is formed on the stump, and re-epithelialization of the wound surface occurs late, compared to rapid wound closure in amphibians. Both facts could physically prevent the interaction between nerves and the wound epithelium. Since appropriate interaction between the wound epithelium and nerves is necessary to induce the AEC, it will be important to reveal if the induction of the AEC is possible in mammals when those two tissues are placed side-by-side.

With regards to blastema precursor cells, there is evidence that transcription factor, Prrx1 (formerly referred to as Prx1), plays a central role. In Xenopus it is shown that prrx1 is expressed in fibroblasts during the early phase of skin wound healing and spike formation, and that its expression in blastema cells is maintained nerve-dependently at the later stage of spike formation (Suzuki et al. 2005). Prrx1 is also known to downregulate Runx2 and Osterix in osteoblast differentiation of a mammalian cell line (Lu et al. 2011). Those things imply that Prrx1 might be related to proliferative and undifferentiated state of blastema cells. Interestingly the mouse Prrx1 (Mprrx1) limb-specific enhancer is activated during scar-free wound healing and spike formation in transgenic Xenopus (Suzuki et al. 2007); however, this enhancer activity is not detected in fibroblasts associated with wound healing in mice (Yokoyama et al. 2011; see Kawasumi et al. 2013 for review). If the activation of the Mprrx1 limb-specific enhancer is a necessary prerequisite for blastema formation, then it will be necessary to discover therapies to activate it in mice.


The vertebrate mineralized bone first evolved as an “exoskeleton” beneath the skin, and an “endoskeleton” evolved later during vertebrate evolution (see Hall 2005; Shimada et al. 2013). Thus, the skeleton is a characteristic trait of vertebrates that ensures active locomotion, supports heavy weight and huge size of the body, and protects from intruders. How the skeleton is established is critical for regulating and even limiting the size and morphology of the vertebrate body. Although many of the molecular and cellular cascades/events that control bone formation are known, we are still far away from fully understanding how the size and shape of both the bone and body are controlled. The challenge is to discover how these molecular/cellular events are coordinated in time and space.

Although the regeneration ability of skeletal tissues is limited in mammals, they are important experimental models given the availability of tools and resources such as genetics, molecular and cellular biology, stem cell biology, medical science and tissue engineering. We have reviewed the molecular and cellular mechanism/process of bone repair and skeletal development. In spite of the information from basic science, attempts to induce endogenous skeletal regeneration in mammals have been of limited success. Moving forward, it should be of value to understand how the regenerative animals, amphibians in particular, exert their ability of skeletal regeneration and what difference lies between mammals and amphibians.


We thank Natsume Sagawa for her great help in drawing illustrations in figures and David Gardiner for his critical reading of the manuscript. This study was supported by Grant-in-Aid for Scientific Research on Innovative Areas to TE (22124006) and Funding Program for Next Generation World-Leading Researchers from the Cabinet Office, Government of Japan to KT (LS007). HY was supported by MEXT and JSPS KAKENHI (22124005), JSPS KAKENHI Grant (25870058), the Kurata Memorial Hitachi Science and Technology Foundation, and the Asahi Glass Foundation.