Transdifferentiation in bilateria

Transdifferentiation occurs when a differentiated cell switches its identity to another differentiated cell type (Merrell and Stanger 2016; Tosh and Slack 2002). The most frequently reported cases of transdifferentiation result from genetic manipulations (Halley-Stott et al. 2013; Zhou and Melton 2008) as well as chemically or surgically induced organ regeneration, such as liver, heart, and lens repair described in mammals, teleosts, and amphibians, respectively (Day and Beck 2011; Godwin 2014; Makino et al. 1990; Suetsugu-Maki et al. 2012; Zhang et al. 2013). Remarkably, transdifferentiation can also occur under physiological conditions, as has been described in the developing Xenopus laevis pancreas and the Drosophila melanogaster heart (Mukhi and Brown 2011; Schaub et al. 2015). In addition, the nematode Caenorhabditis elegans hindgut Y cell transdifferentiates into a motor neuron, while hemocytes give rise to adult-born neurons in the crustaceans Procambarus clarkii and Pacifastacus leniusculus (Benton et al. 2014; Zuryn et al. 2014). These natural cases of transdifferentiation, occurring during the development of noninjured specimens, provide spectacular examples of cell-fate plasticity in both protostomes and deuterostomes. Nevertheless, such events have rarely been associated with a major evolutionary transition (Arenas-Mena 2010).

Transdifferentiation in the mouse skeleton

Mouse skeletogenesis occurs via two distinct developmental routes according to the anatomical location of the skeletal elements being considered. On the one hand, bones of the skull and part of the clavicle form in the absence of cartilage through intramembranous ossification, a process consisting in the direct differentiation of osteogenic mesenchymal condensations into osteoblasts, the bone-forming cells (Hartmann 2009; Komori et al. 1997). From a genetic perspective, osteoblasts (i) respond to a plethora of extrinsic signals, such as the Wnt, BMP, and Hh signaling pathways activated by ligands of the wingless/invected1, bone morphogenetic protein, and hedgehog families, respectively; (ii) integrate inputs from intrinsic cues such as the Runx2 and Sp7 (also known as Osterix) transcription factors; and (iii) express a variety of bone extracellular matrix components such as osteocalcin (also coined bone gla protein, bglap or bgp) and the type I fibrillar collagen 1a1 (Long 2011; Marcellini et al. 2012; Nakashima and de Crombrugghe 2003). On the other hand, the development of endochondral skeletal elements, such as the limbs, ribs, and vertebrae requires the differentiation of chondrocytes (the cartilage-forming cells) from mesenchymal condensations to form a cartilaginous scaffold for eventual replacement with bone (Hartmann 2009). Figure 1a–b summarizes the major events occurring during mouse long bone development. Resting and proliferative chondrocytes (located closer to the epiphyses, at each extremity of the long bone cartilaginous scaffold), express Sox9 and produce a nonmineralized immature type of hyaline cartilage matrix enriched in specific extracellular proteins such as the fibrillar collagen 2a1 (type II collagen) and aggrecan (Kozhemyakina et al. 2015). As skeletal development proceeds, the cartilage located at the level of the diaphysis (the central region of the long bone cartilaginous scaffold) undergoes a maturation process involving chondrocyte hypertrophy, the transcriptional activation of a variety of genes such as ihh (coding for a ligand of the Hh signaling pathway) and collagen 10a1 (a member of the network-forming collagen family), and extracellular matrix calcification (Gomez-Picos and Eames 2015; Kozhemyakina et al. 2015). Subsequently, hypertrophic chondrocytes are eliminated by apoptosis and the resorption of mineralized cartilage by osteoclasts allows the formation of a central cavity invaded by blood vessels and filled with bone marrow (Mackie et al. 2008).

Fig. 1
figure 1

Endochondral bone development in mouse and other extant osteichthyan representatives. a Legend of the symbols and color codes used to represent the distinct skeletal cell types and tissues. b Patterns of growth and differentiation during mouse long bone development, shown here from embryonic day 12 to 1 month of age. The boxed area indicates the location of similar regions shown at higher magnifications in c and d. c Simplified procedure of the lineage tracing experiment, merely shown here as a proof of principle. The temporally and spatially controlled activations of Cre-lox cassettes mediate the specific and irreversible genetic labeling of chondrocytes and osteoblasts with a GFP and mCherry reporter, respectively. Expected outcomes are shown here for two hypothetical species whose endochondral bones either develop with or without transdifferentiation of chondrocytes into osteoblasts. d Schematic drawings of the epiphyseal regions of developing cartilaginous skeletal elements of teleosts (adult D. rerio and juvenile C. carpio), amphibians (6–month-old froglet and 3-year-old adult X. tropicalis, and metamorphosed P. waltl), and newborn M. musculus. Note the presence of endochondral bone in all species. A cladogram is shown at the bottom. Skeletal elements are not drawn to scale and are adapted from Miura et al. 2008, Quilhac et al. 2014, Weigele and Franz-Odendaal 2016, Witten and Huysseune 2009 and Zhou et al. 2014. See text for details

Full size image

At this point, it is important to make the distinction between the perichondral and endochondral bone types that develop in close association with the cartilage tissue. The perichondral bone, also known as the cortical bone or perichondrial bone, is a compact type of bone deposited by osteoblasts whose differentiation from the perichondrium (a mesenchymal layer surrounding the cartilaginous scaffolds) is induced by Ihh emanating from neighboring chondrocytes (see Fig. 1a, b and Shimoyama et al. 2007; St-Jacques et al. 1999). Perichondral bone development occurs through perichondral ossification, a process contingent on adjacent cartilage but otherwise similar to skull intramembranous ossification. By contrast, the endochondral bone, whose development is also Ihh dependent (Long et al. 2004), consists of bone matrix trabeculae initially laid upon calcified cartilage remnants lying within the marrow cavity. In mouse, the chondro-osseous junction refers to the sharp transition delineating a zone of intense chondrocyte death and mature cartilage resorption, and a zone of trabecular bone deposited by endochondral osteoblasts (i.e., the subset of osteoblasts in charge of endochondral ossification). Deciphering if all endochondral osteoblasts derive from perichondral osteoblasts invading the marrow cavity or if some of them also derive from transdifferentiated hypertrophic chondrocytes has been a source of debate and controversy in the field for more than 150 years (Quilhac et al. 2014; Tsang et al. 2015). In 1925, Fell already compared the conflicting views regarding the fate of hypertrophic chondrocytes, favoring, at the time, complete cartilage disintegration (Fell 1925). Chondrocyte elimination was later shown to occur through apoptosis and to involve members of the Bcl-2 family (Hatori et al. 1995; Oshima et al. 2008). As an alternative to the aforementioned scenario, several authors, including Fell himself in 1932, proposed that chondrocytes can give rise to osteoblasts (Bianco et al. 1998; Crelin and Koch 1967; Fell 1932). Four recent studies reporting sophisticated lineage tracing experiments in Mus musculus have settled the matter, demonstrating that a subset of chondrocytes do survive the chondro-osteo junction and transdifferentiate into endochondral osteoblasts during early skeletal development and postnatal growth, as well as fracture healing (see Fig. 1c and Park et al. 2015; Tsang et al. 2015; Yang et al. 2014a; Yang et al. 2014b; Zhou et al. 2014). Being restricted to one mammalian representative, these findings pose the question of the origin of chondrocyte transdifferentiation into osteoblasts and its implications regarding the evolutionary emergence of endochondral ossification.

Endochondral bone origin

How does the timing of origin of the endochondral bone relate to the evolution of other skeletal tissues? There is ample evidence that an immature type of nonmineralized cartilage predated vertebrate emergence because it is also found around the amphioxus mouth region and in some protostomes (see Fig. 2 and Donoghue and Sansom 2002; Gomez-Picos and Eames 2015; Jandzik et al. 2015; Kaneto and Wada 2011; Tarazona et al. 2016; Yong and Yu 2016). The dermal bone (e.g., exoskeletal intramembranous bones and teeth) and perichondral bone evolved in the stem-gnathostomes, although their exact order of appearance is still a matter of debate (see Fig. 2 and Donoghue et al. 2006; Enault et al. 2015; Gomez-Picos and Eames 2015; Janvier 2015; Johanson et al. 2010; Johanson et al. 2012; Min and Janvier 1998; Nian-Zhong et al. 2005; Sire et al. 2009). A mineralized type of mature cartilage is found in extant gnathostomes and has also been reported in Euphanerops longaevus and Palaeospondylus gunni, two cyclostome fossils (Enault et al. 2015; Hirasawa et al. 2016; Janvier and Arsenault 2002; Johanson et al. 2010; Johanson et al. 2012). However, because the cartilaginous skeleton of extant cyclostomes is unmineralized, it has been proposed that cartilage calcification evolved independently in cyclostomes and gnathostomes (Janvier and Arsenault 2002). It therefore still remains to be determined if a mineralized type of mature cartilage evolved before or after dermal and perichondral bones (Fig. 2). Finally, according to the fossil record, the endochondral bone evolved relatively recently, in the lineage leading to osteichthyans and after it diverged from chondrichthyans (see Fig. 2 and Donoghue and Sansom 2002; Hirasawa and Kuratani 2015).

Fig. 2
figure 2

Polarizing the evolution of the major skeletal tissue subtypes. A schematic phylogenetic tree shows the evolutionary relationships of the nonvertebrate amphioxus outgroup (Amp.); cyclostomes, such as hagfish and lampreys (Cyc.); the stem-gnathostomes (Gna.), shaded in gray, a polyphyletic group of extinct jawed and jawless vertebrates (daggers); chondrichthyans such as sharks and rays (Cho.); and osteichthyans such as mammals, amphibians, and teleosts (Ost.). Unmineralized immature cartilage (i) was already present in the last common ancestor of all chordates. Dermal bone (d) and perichondral bone (p) evolved in the stem-gnathostomes (bracket). Mineralized mature cartilage (m) might have evolved in the stem-gnathostomes or at the base of the vertebrates (dotted arrows). Endochondral bone (e) evolved in the osteichthyan lineage, after it separated from chondrichthyans. See text for details and references

Full size image

In agreement with the above scenario, evidence of the endochondral bone has been reported in the cartilaginous skeletal elements of all osteichthyans groups analyzed so far. For instance, paleohistological examinations reveal that endochondral ossification is well conserved between cartilaginous skeletal elements of extinct and extant tetrapods (Danto et al. 2016; Hubner 2012; Scheyer 2007). The limbs of the amphibians Xenopus tropicalis (a frog) and Pleurodeles waltl (a urodele) also harbor an endochondral bone, albeit to a lesser extent than that of mammals (see Fig. 1d and Castanet et al. 2003; Miura et al. 2008; Quilhac et al. 2014). Likewise, teleosts such as Astatotilapia elegans, Cyprinus carpio, and Danio rerio exhibit a moderate amount of endochondral bones, most of their marrow cavity being occupied by adipocytes (see Fig. 1d and Benjamin et al. 1992; Bruneel and Witten 2015; Huysseune and Verraes 1986; Witten and Huysseune 2009).

In summary, both fossil and living specimens suggest that the endochondral bone represents a shared feature of sarcopterygians and actinopterygians and should therefore be considered an osteichthyan synapomorphy. Can we, then, pinpoint the cellular and genetic changes that led to the emergence of this evolutionary novelty?

The transdifferentiation hypothesis

Taken together, the aforementioned evidence raises the intriguing possibility that the transdifferentiation of hypertrophic chondrocytes into osteoblasts played a fundamental role in the evolution of endochondral ossification. According to this hypothesis, hypertrophic chondrocytes acquired the ability to transdifferentiate into osteoblasts in the osteichthyan lineage, after it separated from chondrichthyans, thereby contributing to the origin of bone trabeculae built upon calcified cartilage remnants of the bone cavity. If this turned out to be the case, one would expect to detect the presence of chondrocyte-derived endochondral osteoblasts in most, if not all, examined actinopterygian and sarcoptergigyan species. This model is supported by the presence, in D. rerio larvae, of endochondral osteoblasts induced by Ihh signaling and expressing high levels of osteocalcin and sp7 (Hammond and Schulte-Merker 2009). Nevertheless, because such cells conserve a chondrocytic phenotype, are not located within a bone cavity, and do not secrete trabecular bone, older D. rerio specimen should be examined to ensure that mature chondrocytes represent a potential source of bona fide endochondral osteoblasts. (Hammond and Schulte-Merker 2009; Weigele and Franz-Odendaal 2016).

It is also possible that, ancestrally, endochondral ossification exclusively relied on the invasion of periosteal osteoblasts. For instance, the transdifferentiation of chondrocytes into osteoblasts might have evolved in amniotes and mammals in response to higher mechanical forces experienced by the long bones of terrestrial vertebrates, thereby providing an additional osteoblast supply. A prediction of this scenario is that, for the vast majority of osteichthyans, endochondral osteoblasts will be found to derive exclusively from perichondral osteoblasts. If, in addition to M. musculus, only a minority of species offer evidence of chondrocytic transdifferentiation into osteoblasts, it should be concluded that this phenomenon appeared relatively late during evolution and perhaps even convergently according to the phylogenetic distribution of these species. This would imply that a hidden developmental heterogeneity exists between homologous skeletal elements whose endochondral ossification might rely on transdifferentiation in some species but not in others. In this respect, independent gains (or losses) of transdifferentiation of chondrocytes into osteoblasts would represent an additional case of developmental system drift already reported for diverse skeletal elements (Hirasawa and Kuratani 2015; Piekarski et al. 2014).

One way to discriminate between these models would be to reproduce, in a range of osteichthyans, lineage tracing experiments based on the irreversible genetic labeling of chondrocytes using adequate Cre-lox drivers (Fig. 1c). Because two of the most commonly studied vertebrate genetic models, D. rerio and X. tropicalis, only exhibit a moderate amount of late-developing endochondral bone (see Fig. 1d and Hammond and Schulte-Merker 2009; Miura et al. 2008; Weigele and Franz-Odendaal 2016), examining alternative amphibian (e.g., P. waltl) and teleost (e.g., C. carpio) species, as well as coelacanth, lungfish, bowfin, or spotted gar might provide a more straightforward strategy to address this problem.

A unifying view of skeletal cell plasticity

In the past decade, several studies have blurred the border between chondrocytic and osteoblastic molecular identities. Amongst the genes that were initially recognized as “chondrocyte-specific,” collagen 2a1, collagen 10a1, and sox9 have been found to be expressed in osteichthyan osteoblasts, albeit to varying degrees according to the species, anatomical location, and developmental stage being considered (Abzhanov et al. 2007; Albertson et al. 2010; Aldea et al. 2013; Bertin et al. 2015; Eames et al. 2012; Enault et al. 2015; Hilton et al. 2007). Likewise, “osteoblast-specific” genes are now known to be expressed (e.g., collagen 1a1 and osteocalcin) as well as functionally required (e.g., runx2, sp7) during osteichthyan chondrogenesis (Chen et al. 2014; Gavaia et al. 2006; Hammond and Schulte-Merker 2009; Kerney et al. 2007; Kirsch et al. 1997; Nishimura et al. 2012; Viegas et al. 2013). In addition to the intermingling of the expression patterns described above, several remarkable cases of hybrid skeletal cell types have been reported. Firstly, in contrast to the hyaline type of cartilaginous tissues, the fibrocartilage is deposited by fibrochondrocytes producing densely packed arrays of collagen 1a1 and mineralizing their matrix in successive layers (Dyment et al. 2015). Secondly, the chondroid bone, present in many teleosts as well as a few tetrapod species, consists of chondrocyte-like cells embedded within a mineralized bone matrix and coexpressing high levels of osteoblastic and chondrocytic genes (Benjamin et al. 1992; Huysseune and Sire 1990; Mizoguchi et al. 1997; Paul et al. 2016; Vickaryous and Hall 2008). Thirdly, as reported in a subset of tetrapod cartilaginous skeletal elements, some chondrocytes survive the chondro-osseous junction and remain enclosed within lacunar spaces called globuli ossei (Castanet et al. 2003; Quilhac et al. 2014). In P. waltl, such cells lose their chondrocytic phenotype and modify their surrounding matrix by secreting collagen fibrils whose diameters are intermediate between type I and type II collagens (see Fig. 1d and Quilhac et al. 2014). Interestingly, globuli ossei-like structures (intact chondrocyte stacks surrounded by mineralized cartilage matrix) form in mutant mouse long bones of reduced osteoclastic activity (Tonna et al. 2016). This observation, together with the fact that globuli ossei are frequently observed in amphibian and mammalian bones experiencing low rates of remodeling (Quilhac et al. 2014), raises the possibility that osteoclasts contributed to the evolution of the endochondral bone, not only by actively carving the inner cavity but also by sparing areas of mature cartilage matrix sheltering surviving chondrocytes. This might have allowed globuli ossei-confined chondrocytes to experience osteogenic signals for a prolonged period of time and become poised to adopt an osteoblast-like phenotype, or even to fully transdifferentiate into osteoblasts. One good candidate for such a signal is Ihh because it is required for the differentiation of endochondral osteoblasts in mouse (Long et al. 2004) and because forced Hedgehog signaling dramatically increases the expression of osteocalcin and sp7 in teleost hypertrophic chondrocytes (Hammond and Schulte-Merker 2009).

Some authors have proposed that the emergence of osteoblasts involved the exaptation of a gene regulatory network (GRN) which originally evolved in chondrocytes (Fisher and Franz-Odendaal 2012; Gomez-Picos and Eames 2015). Considering that the osteoblastic and chondrocytic GRNs share a common origin might help design sound experimental strategies to explore the genetic basis of the skeletal cell diversity found both within and between species. For instance, a systematic comparison of the distinct GRNs deployed in osteoblasts (both of perichondral and chondrocytic origins) and mature chondrocytes as well as cells of intermediate phenotypes will help identify the transcription factors and signaling pathways acting as crucial switches at distinct depths within the regulatory hierarchy. We postulate that changes affecting downstream nodes of these GRNs will modify discrete aspects of cell physiology and give rise to the aforementioned “hybrid cell types,” while the genes located on top of the genetic cascade will be found to be involved in the transdifferentiation phenomenon recently reported during mouse endochondral ossification.

Conclusion

Transdifferentiation is a pervasive feature of bilateria development, as it occurs between cells belonging to all three germ layers of protostome and deuterostome representatives. A long-standing debate regarding the possible transdifferentiation of chondrocytes into endochondral osteoblasts was recently put to an end by independent lineage tracing studies in mouse long bones. While most major vertebrate skeletal tissues are ancient, the endochondral bone represents a relatively recent osteichthyan innovation whose emergence and/or diversification might be intimately linked to skeletal cell transdifferentiation. Particularly intriguing is the possibility that a common evolutionary origin between the osteoblastic and chondrocytic gene regulatory networks underlies skeletal cell plasticity, ranging from mixed expression patterns to seemingly hybrid cell types and complete transdifferentiation. While solving these issues undoubtedly represents a major challenge, one thing is certain in the skeletal evo-devo field: many unexpected and exciting findings lie ahead as we delve deeper into the cellular, molecular, and genetic underpinning of how the exquisitely complex vertebrate skeletal system evolved.