Skeletal gene expression in the temporal region of the reptilian embryos: implications for the evolution of reptilian skull morphology

Differential expression of early osteoblast marker, Runx2, in the head of crocodile and turtle embryos

Osteogenic mesenchymal precursor cells that express Runx2 are first detected in the temporal region of crocodile embryos at stage 14 (Additional file 1) and an almost identical pattern of Runx2 expression was observed in a subsequent embryonic stage (stage 15; Figure 2A). These Runx2-positive cells were localized at the domain dorsal to the oral cavity where the ventral part of the braincase and future palatine and pterygoid bones develop, as well as in a limited domain dorsolateral to the orbit where the future dorsal projection of the postorbital bone forms (Figure 2B). We also detected a population of Runx2-positive cells at the domain ventrolateral to the orbit where future jugal bone and ventral projection of postorbital bone are formed. At this stage, the precursor of the jugal and ventral projection of the postorbital were dorsoventrally continued as a layer of cells but it was thin mediolaterally, especially at the middle part. In the posterior part of the head, we observe a population of Runx2-positive mesenchymal cells that later differentiate into the main body of the postorbital bone (Figures 2E and 2H). In these stages of crocodile embryogenesis, jaw muscle precursors that were derived from cranial mesoderm migrated to the first pharyngeal arch and expressed MyoD was clearly detected at the central domain of the jaw primordia (Figures 2C and 2F; Additional file 1). Expression of Sox9 was detected at cartilage precursors that later differentiate into quadrate and Meckel's cartilages at the domain ventral to jaw muscle precursor, as well as in chondrocytes that form the future braincase (Figures 2D and 2G). Expression of Scx was detected in tendon precursor cells that are distributed within the primordia of the jaw muscles and in the connective tissue within the eye muscles (Additional file 1). Expression of Six2 was somewhat broader than that of other markers, expressed in mesenchymal cells surrounding the eyes, cartilaginous precursors of the braincase, quadrate, and Meckel's, as well as in the mesenchyme at the interface between muscle precursors and the skeletal tissues to which the muscles attach and in a population of the mesenchyme that dorsally surrounds the brain (Additional file 1).

In crocodile embryos at stage 17 where none of the dermatocranial elements were positive for Alizarin red in previous studies (Rieppel, 1993b;Vickaryous & Hall,, Vickaryous & Hall, Vickaryous & Hall,2008) (Figure 3A), we could detect Runx2 expression in the cell populations that were localized to the area where the future dermatocranium differentiates (Figure 3B). Although the domain where Runx2-positive cells were populated was almost identical to that in previous stages, the boundary of each precursor of the dermatocranial elements became clearer. Although differentiation of the parietal bone is delayed compared to other dermatocranial elements as described previously (Rieppel 1993b;Vickaryous & Hall,, Vickaryous & Hall, Vickaryous & Hall,2008), a pair of precursors of the parietal were recognized as Runx2-positive cell aggregation at the domain dorsolateral to the cartilaginous braincase (Figures 3B and 3E). We could detect Runx2-negative cell populations between the nascent parietal precursor and the morphologically more developed postorbital precursor and also between precursors of the postorbital and quadratojugal located lateral to Sox9-positive quadrate cartilage (Figure 3E). In this stage, MyoD was expressed in differentiated jaw and eye muscles (Figures 3C and 3F) and Sox9 was expressed in differentiated chondrocranium and splanchnocranium components, including the braincase, quadrate, and Meckel's (Figures 3D and 3G). Expression of Scx was detected in tendinous tissues accompanying MyoD-positive muscles as in previous stages (Figure 3H). Expression of Six2 was detected in the mesenchyme localized around the jaw articulation between quadrate and Meckel's, as well as in adjacent mesenchyme of the braincase, postorbital bone, and jaw muscles (Figure 3I).

Next, we examined cranial morphogenesis of turtles that have an anapsid skull, using the same method to identify the distribution pattern of precursors of each tissue that constitutes the cranial musculoskeletal system. In turtle embryos at stage 14 (Additional file 2) that correspond to crocodile embryos at stage 14 in external morphology, and in turtle embryos at stage 15 (Figure 4A) that are comparable to crocodile embryos at stage 15, we observed almost identical patterns of expression for each gene. We detected MyoD expression specifically in the primordia of jaw adductor and eye muscles (Figures 4C and 4H; Additional file 2) and Sox9 expression in precursor cells of the braincase, quadrate, and Meckel's cartilages (Figures 4D and 4I; Additional file 2) as in stage-matched crocodile embryos. Expression of Scx was first detected in a layer of mesenchymal cells that was located at the periphery of the jaw adductor and eye muscle precursors in stage 15 turtle embryos (Figure 4E). Expression of Six2 was observed in the mesenchyme surrounding the eye and adjacent mesenchyme of the braincase and jaw cartilages, as well as in some mesenchymal cells within jaw muscle precursors (Figure 4F; see Additional file 2), as in stage-matched crocodile embryos. Interestingly, we observed expression of the early osteoblast marker, Runx2, in a broader domain at the temporal region of the head of turtle embryos, compared to that in stage-matched crocodile embryos. In stage14 turtle embryos, Runx2 expression was detected in a population of cells medial to the precursor of the jaw adductor muscles and the mesenchyme localized at the domain dorsolateral and ventrolateral to the orbit (Additional file 2). The domain of Runx2 expression became further expanded in the head of turtle embryos at stage 15. A thick layer of the mesenchymal cells that express Runx2 completely covered the brain and the precursor of jaw adductor muscle laterally (Figures 4B and 4G).

In turtle embryos at stage 17 (Figure 5A) that correspond to crocodile embryos at stage 17 in overall morphology, we observed MyoD expression in differentiating cranial muscles, including external adductor muscles (Figure 5C) and Sox9 expression in the cartilaginous tissues that constitute the braincase, quadrate, and Meckel's (Figure 5D). Scx was specifically expressed in tendinous tissues at the periphery of jaw adductor muscles, as well as in the precursor of the bodenaponeurosis (central tendon of external adductor) just appeared within the jaw adductor muscular tissue (Figure 5E). The expression domain of Six2 was broader in the temporal region of the head compared to that of Scx, diffusively expressed in the mesenchymal cells surrounding jaw adductor muscles, braincase, and jaw cartilages (Figure 5F). A thick layer of Runx2-positive mesenchymal cells that surrounds the braincase and jaw adductor muscle laterally was observed (Figure 5B). Runx2-expressing mesenchyme was also distributed around the quadrate cartilage and the ventral part of the braincase.

Expression of potential upstream osteogenic regulatory genes in the head of crocodile and turtle embryos

Through comparative analysis of expression patterns of tissue-specific marker genes, we noticed a difference in the spatial pattern of expression of the early osteoblast marker, Runx2 in the head of crocodile and turtle embryos. To reveal potential mechanisms that account for such differential distribution of osteogenic mesenchymal precursor cells between two reptilian lineages with or without temporal fenestrae, we next examined expression patterns of some candidate genes that are known to regulate cranial osteogenesis. In the present study, we focused on Bmp4, Msx1, and Msx2. Bmp4 is a signaling molecule and plays a key role in the Bmp signaling pathway. Because exogenous Bmp4 increases tissue volume in calvarial bone tissue culture, this protein is considered to be involved in calvarial bone growth (Kim et al., 1998, Rice et al., 2003). Both Msx1 and Msx2 are members of the muscle segment homeobox (msh) gene family of transcription factors and both loss-of- and gain-of-function analyses of these genes suggest their essential roles in vertebrate cranial osteogenesis (Satokata & Maas, 1994, Satokata et al., 2000).

In the present analysis, we found that Bmp4 and Msx1 showed almost identical expression patterns through cranial osteogenesis in crocodile and turtle embryos. In crocodile embryos we examined (through stage 14 to stage 17), Bmp4 was strongly expressed in the epithelium of cochlear canal, the mesenchyme surrounding the eye, the mesenchyme distributed in the medial part of jaw primordia, the precursors of the palatine bones, and a population of mesenchymal cells that covered the brain dorsally (Figure 6A; Figures 7A and 7D). In turtle embryos we examined (through stage 14 to stage 17), Bmp4 was expressed in a spatially limited domain: the epitthelium of cochlear canal, the mesenchyme dorsolateral and ventrolateral to the eye and a limited population of the mesenchyme in close proximity of the jaw articulation (Figure 6G; Figure 7G). We observed Msx1 expression in the epithelium of the cochlear canal, the mesenchyme that occupies the domain close to the jaw articulation and lateral to the quadrate and Meckel's cartilages, and a thin layer of mesenchymal cells that covers the brain dorsally in crocodile embryos examined (Figure 6B; Figures 7B and 7E). In turtle embryos, Msx1 was expressed in the epithelium of the cochlear canal, the mesenchyme distributed around the jaw articulation and lateral to quadrate and Meckel's cartilages, as well as in the mesenchyme that populates the domain dorsal to the eye (Figure 6H; Figure 7H).

In contrast to Bmp4 and Msx1, we detected differential expression patterns of Msx2 in the head of crocodile and turtle embryos. In crocodile embryos at stage 14 and 15, Msx2 was expressed in a thin layer of mesenchymal cells surrounding the dorsal aspect of the brain (Figures 6C and 6E). In the posterior part of the head, the ventral edge of this Msx2-expressing cell population is located dorsal to the eye. In these crocodilian embryos, Msx2 expression was also observed in a population of the mesenchyme that occupied the domain between the ventrolateral part of quadrate cartilage and surface epidermis (Figures 6C and 6E). These mesenchymal cells expressed Msx1 as well (Figure 6B) and appeared to differentiate into the quadratojugal bone later. In crocodile embryos at stage 17, specific expression of Msx2 was detected at a population of mesenchymal cells in close proximity of Runx2-expressing precursors of postorbital and quadratojugal bones, as well as in a thin layer of the mesenchyme surrounding the brain dorsally where future parietal bones were developed (Figures 7C and 7F). We observed that the space adjacent to Msx2-positive precursors of these dermatocranial elements was filled with Msx2-negative mesenchymal cells. Interestingly, we observed broader expression of Msx2 in turtle embryos, compared to that in stage-matched crocodile embryos. In turtle embryos examined, Msx2 was expressed in mesenchymal cells that populate lateral aspect of the head of embryos (Figures 6I and 6K; Figures 7I and 7J). The ventral edge of the Msx2-expressing mesenchymal layer was terminated ventral to the eye and these cells covered MyoD-expressing external adductor muscle laterally. Showing its dorsoventrally broadened expression pattern, the domain of Msx2 expression largely overlapped with that of Runx2 in turtle embryos (Figures 4B and 4G; Figure 5B).

Expression of Runx2 and Msx2 in the head of snake embryos

To verify a correlation between the expression patterns of Runx2 and Msx2 and reptilian skull morphology, we finally examined expression patterns of these genes, as well as of marker genes for muscular and cartilaginous tissues, in cranial morphogenesis of a snake species where their temporal fenestrae are not encircled by the temporal bars. In snake embryos at stage 26 (Figure 8A) that morphologically correspond to crocodile and turtle embryos at stage 14 or 15, MyoD was expressed in the primordia of the first arch muscles (Figures 8C and 8G). Sox9 was strongly expressed in the precursors of quadrate and Meckel's cartilages and the base of the braincase, as well as in a layer of mesenchyme surrounding the brain laterally (Figures 8D and 8H). In these snake embryos, early osteoblast marker, Runx2 was expressed in the mesenchyme that occupied the space medial to the Sox9-positive quadrate precursor and in a limited population of mesenchymal cells ventral to the orbit (Figure 8B). We also detected Runx2 expression in a layer of mesenchymal cells that surround the brain laterally. In the posterior temporal region, Runx2 was only faintly expressed in the adjacent mesenchyme of the Sox9-positive quadrate precursor (Figure 8F). Expression of Msx2 was detected in the mesenchyme medial to the precursor of the quadrate (Figure 8E). Its expression domain was spatially overlapped with the domain where Runx2-positive cells were distributed, but the former was narrower. We also detected Msx2 expression in a mesenchymal layer that surrounded the brain dorsally (Figures 8E and 8I). In snake embryos at stage 29 (Figure 8J), which morphologically correspond to crocodile and turtle embryos at stage 17, well-differentiated cranial muscular and cartilaginous tissues were specifically labeled by expression of MyoD (Figures 8L and 8P) and Sox9 (Figures 8M and 8Q), respectively. We observed spatially overlapped expression of Runx2 and Msx2 in these embryos: both genes were expressed in the precursors of palatine and pterygoid bones medial to Sox9-positive jaw cartilages and in mesenchymal cells accompanying jaw cartilages, as well as in a layer of loose mesenchyme that later forms a precursor of parietal bones that cover the dorsal part of the brain (Figures 8K, 8N, 8O, and 8R). However, only Runx2 was expressed in the mesenchyme that populated the domain ventral to the orbit, which possibly corresponds to precursors of the maxilla bones of the upper jaw. No Runx2 and/or Msx2 expressing osteogenic mesenchymal precursor cells populated the domain lateral to jaw adductor muscles. In more developed snake embryos (at stage 31; Figure 8S), both Runx2 and Msx2 were specifically expressed in the precursors of the dermatocranial elements, including the parietal that encases the brain dorsally (Figures 8T and 8W). As in previous stages, expression of the former was more expanded. The lateral aspect of the external adductor muscles was never covered by the skeletal tissues that express Runx2 and/or Msx2. The results on expression domains of each gene analyzed in crocodile, turtle, and snake embryos are summarized in Table 1.

Potential developmental basis of anapsid skull in turtles

Skull morphology, especially the osteological configuration of the temporal region, has historically been treated as the most important character in the classification of major lineages of reptiles. Based on their anapsid skull, turtles have been regarded as a sole descendent of stem reptiles (Williston, 1917, Gregory, 1946, Romer, 1968, Gaffney, 1980;Reisz & Laurin,, Reisz & Laurin, Reisz & Laurin,1991, Lee, 1993;Laurin & Reisz, 1995, Lee, 1996 1997, Reisz, 1997, Lee, 2001) despite the contrary argument that turtles were derived from an ancestor with a diapsid skull (Lakjer, 1926, Goodrich, 1930). Recent phylogenetic studies where the interrelationships of both extant and extinct reptiles were surveyed through comprehensive analysis of multiple osteological traits concluded that turtles were closely related to lepidosaurian diapsids (Rieppel & deBraga, 1996;deBraga & Rieppel , deBraga & Rieppel deBraga & Rieppel 1997Rieppel 2000, Hill, 2005, Li et al., 2008). Furthermore, results of molecular phylogenetic studies have strongly suggested diapsid affinity of turtles (Hedges & Poling, 1999;Kumazawa & Nishida,, Kumazawa & Nishida, Kumazawa & Nishida,1999, Iwabe et al., 2005, Hugall et al., 2007, Shedlock et al., 2007Shen et al. 2011, Tzika et al., 2011, Chiari et al., 2012, Crawford et al., 2012, Fong et al., 2012, Lyson et al., 2012, Wang et al., 2013). If turtles were derived from a diapsid ancestor, then the anapsid skull of turtles evolved independently from that of ancestral lineages of reptiles by secondary closure of the temporal fenestrae. However, although the phylogenetic position of turtles within amniotes still remains inconclusive (Lyson et al., 2010 2013, Kuratani et al., 2011), there has been no study in which the process of development of their anapsid skull is described with molecular markers for labeling precursor cells of the dermatocranium. In the present study, we examined early cranial morphogenesis of representative reptilian species through comparative analysis of gene expression patterns and found unique expression patterns of Runx2 and Msx2 in turtle embryos that are not observed in crocodile and snake embryos.

Runx2 is widely known as a transcription factor that plays a fundamental role in osteoblast differentiation in vertebrate embryos (Ducy et al., 1997, Komori et al., 1997, Mundlos et al., 1997, Nakashima et al., 2002, Ishii et al., 2003, Dobreva et al., 2006, Kerney et al., 2010) and its transcript has been used as a molecular marker for preosteoblasts or osteoblast progenitors (Ducy et al., 1997, Karsenty, 2001, Bobola et al., 2003, Ishii et al., 2003, Abzhanov et al., 2007, Han et al., 2007;Tokita & Schneider , Tokita & Schneider Tokita & Schneider 2009). In turtle embryos where mineralization of dermal bones of the skull has not yet occurred, mesenchymal cells that express Runx2 were broadly distributed in the lateral domain of the head, from the top of the head to the ventral margin of the jaw (Figure 9). This means that turtle embryos have wider distribution of the cells that have a potential to differentiate into bones in the temporal region of the head, compared to other lineages of reptiles. The reduction of the amount of Runx2 mRNA causes developmental defects in calvarial bones, called cleidocranial dysplasia, in mouse embryos (Lou et al., 2009). In contrast, early onset of Runx2 expression that eventually results in an increase of the amount of its mRNA in the cranial mesenchyme accelerates the timing of mineralization of cranial dermal bones in mouse embryos and brings about craniosynostosis characterized by overgrowth of bones (Maeno et al., 2011). Similar result has been obtained from experiments using avian embryos (Merrill et al., 2008). We speculate that heterotopy in the preosteoblast distribution observed in early stage turtle embryos may lead to the increase of the amount of Runx2 expression that results in the increase of the level of ossification in the temporal region of their skull and this would be a primary factor to build the anapsid skull where the temporal region is completely covered with bone.

In this study, we focused on several candidate molecules that potentially regulate Runx2 expression and examined their expression patterns in reptilian embryos. Bmp4 is known to be involved in osteogenesis of vertebrates where it regulates expression of other osteogenic regulatory genes, including Msx1, Msx2, and Runx2 (Marazzi et al., 1997, Kim et al., 1998, Hollnagel et al., 1999, Tribulo et al., 2003, Zhang et al., 2003, Brugger et al., 2004). Msx1 is a transcription factor known to regulate growth and patterning of calvarial bones in mouse embryos (Satokata & Maas, 1994, Han et al., 2007, Roybal et al., 2010). Although, as previously reported in mouse embryos (Rice et al., 2003, Han et al., 2007), both Bmp4 and Msx1 are expressed in limited populations of cranial mesenchyme in embryos of crocodiles and turtles, we could not detect any substantial differences in their expression domains between the two species. On the other hand, we observed spatially different expression patterns of Msx2 in the head of embryos of all reptilian species we examined. Expression of Msx2 was detected in cranial mesenchyme and dermal bone precursors as reported in mouse embryos (Jabs et al., 1993, Ishii et al., 2003, Rice et al., 2003, Han et al., 2007, Roybal et al., 2010). Furthermore, its expression spatially overlapped with that of Runx2 in reptilian embryos, as in mouse embryos (Ishii et al., 2003, Rice et al., 2003, Han et al., 2007). In turtle embryos, expression domain of Msx2 in the mesenchyme distributed in the temporal region of the head was broad in a dorsal-ventral direction, showing similar pattern with Runx2 in the mesenchyme. A mutation in the homeobox of Msx2 gene causes craniosynostosis in human and mouse (Jabs et al., 1993, Liu et al., 1999). Similarly, overexpression of Msx2 promotes osteogenesis (Cheng et al., 2003, Ichida et al., 2004) and causes overgrowth of dermal bones of the skull by increasing the number of proliferative osteoblasts (Dodig et al., 1999, Liu et al., 1999). In contrast, loss-of-function of Msx2 results in defects of skull ossification in mammals (Satokata et al., 2000, Wilkie et al., 2000, Ishii et al., 2003, Antonopoulou et al., 2004, Han et al., 2007). Furthermore, Msx2 is known to positively regulate downstream Runx2 expression (Ishii et al., 2003, Han et al., 2007, Watanabe et al., 2008). Considering the evidence provided by previous studies, regulatory changes in Msx2 expression in turtle embryos may influence expression patterns of downstream Runx2, which regulate osteoblast differentiation. Dorsoventrally broadened distribution of osteogenic mesenchymal precursor cells in the temporal region of the head owing to the regulatory alteration of these osteogenic genes may allow this reptilian lineage to reacquire the anapsid skull. Although the precise mechanism underlying regulatory change of Msx2 expression in the head of turtle embryos has not been identified, recent findings that early stage arrest of Msx2 expression in neural crest-derived odontoblasts may account for the absence of teeth in turtles (Tokita et al., 2012) supports the hypothesis that this transcription factor may play a pivotal role in the development of their unique cranial morphology.

The development of the dermatocranium occurs in multiple steps (Ishii et al., 2003). The first phase includes the genesis, migration, and initial specification of osteogenic mesenchymal precursor cells. The second phase consists of the differentiation of the mesenchyme into osteoblasts. And the last phase includes deposition of osteogenic extracellular matrix around the osteoblasts and mineralization of the matrix. The dermatocranium of vertebrates is formed from cranial mesenchyme derived from two distinct embryonic sources: neural crest and mesoderm (Jiang et al., 2002;Gross & Hanken , Gross & Hanken Gross & Hanken 2005;Noden & Trainor,, Noden & Trainor, Noden & Trainor,2005). Unfortunately, fate mapping studies of each dermatocranial element as performed in avian and mammalian embryos (Le Lièvre, 1978, Noden, 1978 1983, Couly et al., 1993;Köntges & Lumsden,, Köntges & Lumsden, Köntges & Lumsden,1996, Jiang et al., 2002) have not been done in non-avian reptiles. Interestingly, the pattern of migration and distribution of cranial neural crest cells from which some cranial dermal bones should form is almost identical in early stage embryos of crocodiles and turtles (Meier & Packard, 1984;Hou & Takeuchi,, Hou & Takeuchi, Hou & Takeuchi,1994;Kundrát,, Kundrát, Kundrát,2008). Such data may support that differentiation or maturation processes of osteogenic mesenchyme are more responsible for producing diversity of reptilian skull morphology. We speculate that the developmental program, which determines cranial mesenchymal populations where early-phase osteogenic transcription factors Msx2 and Runx2 are activated, may be important in the patterning of reptilian skull morphology.

There exists substantial diversity in the skull morphology within turtles and most living turtle species do not have fully anapsid skulls and instead possess varying degrees of dorsal and/or ventral emargination on their skull (Jones et al., 2012, Werneburg, 2012a). In the present study, we could not sample and analyze the embryos of turtle species with fully anapsid skull, such as marine turtles (Kuratani, 1999, Jones et al., 2012), alligator snapping turtle () (Sheil, Macrochelys temminckii 2005), and big-headed turtle (Platysternon megacephalum), owing to difficulty in the access to the materials. Instead, we analyzed the embryos of a soft-shelled turtle species with highly emarginated skull. In fact, soft-shelled turtles have only a narrow bar of bone across the temporal region lateral to the external adductor muscles due to large scale emargination from the dorsal and ventral margins of the cheek (Ogushi, 1911, Sheil, 2003). In normal development of soft-shelled turtles, the postorbital bone does not grow in a posterior direction significantly, keeping its relatively small size within the dermatocranium (Sheil, 2003;Sánchez-Villagra et al.,, Sánchez-Villagra et al., Sánchez-Villagra et al.,2009). Therefore, the small postorbital bone of soft-shelled turtles does not largely contribute to the formation of a bony roof at the temporal region of the skull.

It is interesting that we observed dorsoventrally broadened distribution of the mesenchymal cells that express Runx2 at the temporal region of the embryos of a soft-shelled turtle species with highly emarginated skull. Dermal bone development occurs through a multi-step molecular pathway regulated by different transcription factors (Zhang, 2010). As an initial step, Runx2 is required for the differentiation of mesenchymal cells into preosteoblasts. In subsequent stage where these preosteoblasts differentiate into mature osteoblasts, Osx, a downstream gene of Runx2, is necessary (Nakashima et al., 2002, Nishio et al., 2006). Furthermore, in the later stages where the osteoblasts produce osteogenic extracellular matrix and the mineralization of these extracellular matrix is occurred, many additional molecules such as bone sialoprotein, osteopontin, and osteocalcin are involved (Zhang, 2010). We speculate that in soft-shelled turtles only a limited population of cells within Runx2-positive preosteoblasts distributed in the temporal region of the head is allowed to differentiate into mature osteoblasts and eventually osteocytes through regulation of expression of down stream genes (e.g. Osx), to form a pair of relatively small postorbital bones. Although the regulatory mechanism of Osx expression in osteogenic mesenchyme is not fully understood, both Runx2-dependent and -independent pathways have been suggested (Lee et al., 2003;Celil & Campbell,, Celil & Campbell, Celil & Campbell,2005, Maehata et al., 2006, Xing et al., 2007, Zhang, 2010). Histological analysis reveals that late stage soft-shelled turtle embryos have a layer of (non-muscular) fibrous connective tissue lateral to the external adductor muscles (Additional file 3). Judging from its position, the connective tissue layer appears to be derived from Runx2-positive preosteoblasts and have a potential to ossify themselves as other connective tissues represented by tendons and ligaments (Okawa et al., 1998;Tokita et al., 2007). Interestingly, similar type of connective tissue layer is absent in the temporal region of crocodile and snake embryos (Additional file 3). Those histological observations support the above hypothesis that later processes of cranial osteogenesis may largely contribute to the construction of the main body of each dermatocranial element from the osteogenic mesenchymal progenitor pool.

The dorsoventrally broadened distribution of preosteoblasts observed in turtle embryos might be a developmental synapomorphy re-acquired by the common ancestor of turtles. In the course of chelonian evolution, each chelonian lineage may develop the temporal dermal bones (e.g. postorbital, parietal, jugal) with various sizes and shapes, through regulatory changes of the osteogenic down stream molecules. Future studies should investigate expression pattern of Runx2 and Msx2 in the head of embryos of turtle species with fully anapsid skull, as well as expression pattern of downstream genes that regulate differentiation of mature osteoblasts and osteocytes in turtle embryos, to verify a correlation between the gene expression pattern and their skull morphology.

Heterotopy in distribution of osteogenic mesenchymal precursor cells and diversification of reptilian skull morphology

The frame-like skulls possessed by diapsid reptiles evolved in response to functional forces (Rieppel, 1993a, Moazen et al., 2009, Herrel et al., 2007, Curtis et al., 2011) and several studies have suggested heterochrony as a driving force for producing this morphological diversity (Rieppel, 1993a, Whiteside, 1986, Irish, 1989). The ancestral lineage of diapsid reptiles possessed upper and lower temporal bars that encircle temporal fenestrae (Müller, 2003, Moazen et al., 2009). The lower temporal bar that encloses lower temporal fenestra ventrally was probably lost once in the common ancestor of lepidosaurs and archosaurs, possibly as the outcome of paedomorphosis: incomplete ossification of a quadrato-maxillary ligament between jugal and quadratojugal bones (Rieppel, 1993a;Müller,, Müller, Müller,2003). If this is true, the lower temporal bar that possibly results from peramorphosis (hypermorphosis): complete ossification of a quadrato-maxillary ligament was independently re-acquired in the lineages of tuatara and crocodiles, as well as in several extinct reptilian lineages (Rieppel, 1993a;Müller,, Müller, Müller,2003). Furthermore, disappearance of upper temporal bar, which is regarded as an extreme condition of reduction of the dermatocranium in reptiles, may have independently evolved in the skull of geckos (Gekkonidae), miniaturized fossorial lizards (e.g., Typhlosaurus, Dibamus), amphisbaenian, and snakes, as the outcome of paedomorphosis represented by the retardation of ossification (Rieppel, 1993a, Irish, 1989;Cundall & Irish,, Cundall & Irish, Cundall & Irish,2008). In the present study, we revealed a possible correlation between distribution pattern of Runx2 and/or Msx2-expressing osteogenic mesenchymal precursor cells and the skull morphology of each reptilian lineage (Figure 9). In early stage crocodile embryos, we observed focal distribution of osteogenic mesenchyme around the domain where future temporal bars are formed. In early stage snake embryos, osteogenic mesenchymal cells were primarily found adjacent to the primordium of the braincase and the spatial pattern presaged the absence of bony temporal bars in the temporal region of adult animal.

Sample collection and staging of embryos

Fertilized eggs of Chinese soft-shelled turtle, Pelodiscus sinensis, were purchased commercially from a local breeder in Japan. Fertilized eggs of Siamese crocodile, Crocodylus siamensis, were provided by a local breeder in Thailand. Fertilized eggs of corn snake, Pantherophis guttatus, were obtained by the first author after mating several pairs of the reproductively mature adults in the laboratory. Staging of P. sinensis embryos was performed after Tokita and Kuratani (2001). Because there is no embryonic staging system for C. siamensis at present, we used the system for Alligator mississippiensis embryos (Ferguson, 1985) where each stage was determined based on external morphology of the embryos, for staging of this species. Staging of P. guttatus (Zehr, embryos was performed on the basis of staging table of Thamnophis sirtalis 1962). Interspecific comparisons of gene expression pattern were performed in the embryos that are comparable to each other in terms of overall external morphology. Because snake embryos are limbless, we mainly employed external features of the head of the embryos as primary criteria for determining the stages for comparison. All animal experiments were approved by the University of Tsukuba Committee for Animal Care (No.10-034).

Molecular cloning

Total RNA was extracted from embryos using ISOGEN reagent (NIPPON GENE CO., LTD).

RT–PCR was performed to amplify fragments of P. sinensis Runx2, Six2 and C. siamensis Bmp4, Msx2, MyoD, Runx2, Scleraxis (Scx), Six2, Sox9 and P. guttatus Msx2, Runx2, Sox9 messenger RNA. Primer sequences used for isolation of the fragments of these genes are available upon request. Because Bmp4, Msx1, Msx2, MyoD, Scx, Sox9 of Pelodiscus and MyoD of Pantherophis were already sequenced and sequence data were deposited in the database by other researchers, we isolated the orthologous fragments by RT–PCR with primers constructed by referring to the reported sequence data. The fragments were isolated using the pGEM T-easy vector systems (Promega) or TOPO® TA cloning kit (Invitrogen) and sequenced using an ABI 3130 sequencer (Applied Biosystems). To identify the orthologous genes of the isolated fragments, comparable sequence data were surveyed using a BLAST search, and phylogenetic trees with neighbor joining method were constructed after sequence alignment using the CLUSTALX software. All new DNA sequence data were deposited in the DDBJ database (AB811933-AB811944).

Gene expression analysis

Embryos were fixed in 4% PFA, dehydrated using an methanol series, placed in xylene, embedded in paraffin, and sliced with a microtome. Serial sections were hybridized with digoxigenin-labeled RNA riboprobes as described in Neubüser et al. (1995) with slight modifications. To identify the expression domain of Msx1 in crocodile tissues, chicken Msx1 antisense riboprobe was hybridized. Generally, hetero-specific RNA probes easily hybridize among reptilian lineages (Harris et al., 2006, Tokita et al., 2012). In this study, we only analyzed reptilian embryos at the ontogenetic stages where early cranial osteogenesis occurs. To confirm the expression pattern of each gene in the cranial tissues, two to five individuals representing each embryonic stage were sampled for analysis. The consistency of the gene expression patterns among all individual embryos at the same stage was confirmed. Multiple sections representing several longitudinal (anterior-posterior) planes prepared from the same individual were hybridized with the probes and the sections prepared at corresponding longitudinal planes were compared between different individuals. Corresponding longitudinal planes between different reptilian species were determined based on overall histological configuration of the head of the embryos. For visualization of each cranial tissue and interspecific comparison of general histology of the head, Miligan's Trichrome staining was performed following standard protocols. To identify each anatomical structure in cranial musculoskeletal tissues of the embryos, we took the results of other's researches into account: (Schumacher, 1973, Rieppel, 1993b;Vickaryous & Hall,, Vickaryous & Hall, Vickaryous & Hall,2008;Bona & Desojo,, Bona & Desojo, Bona & Desojo,2011) for crocodile, (Schumacher, 1973, Rieppel, 1990, Rieppel, 1993c;Sánchez-Villagra et al.,, Sánchez-Villagra et al., Sánchez-Villagra et al.,2009, Werneburg, 2012a 2012b) for turtle, and (Kamal et al., 1970, Haas, 1973, Zaher, 1994;Buchtová et al.,, Buchtová et al., Buchtová et al.,2007) for snake.