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

Reptiles have achieved highly diverse morphological and physiological traits that allow them to exploit various ecological niches and resources. Morphology of the temporal region of the reptilian skull is highly diverse and historically it has been treated as an important character for classifying reptiles and has helped us understand the ecology and physiology of each species. However, the developmental mechanism that generates diversity of reptilian skull morphology is poorly understood. We reveal a potential developmental basis that generates morphological diversity in the temporal region of the reptilian skull by performing a comparative analysis of gene expression in the embryos of reptile species with different skull morphology. By investigating genes known to regulate early osteoblast development, we find dorsoventrally broadened unique expression of the early osteoblast marker, Runx2, in the temporal region of the head of turtle embryos that do not form temporal fenestrae. We also observe that Msx2 is also uniquely expressed in the mesenchymal cells distributed at the temporal region of the head of turtle embryos. Furthermore, through comparison of gene expression pattern in the embryos of turtle, crocodile, and snake species, we find a possible correlation between the spatial patterns of Runx2 and Msx2 expression in cranial mesenchymal cells and skull morphology of each reptilian lineage. Regulatory modifications of Runx2 and Msx2 expression in osteogenic mesenchymal precursor cells are likely involved in generating morphological diversity in the temporal region of the reptilian skull. Electronic supplementary material The online version of this article (doi:10.1186/2193-1801-2-336) contains supplementary material, which is available to authorized users.


Introduction
Amniotes (Amniota) consist of two large groups of tetrapod vertebrates, Synapsida and Reptilia, that diverged from one another over 300 million years ago (Ma) (Carroll, 1988, Modesto & Anderson, 2004, Benton, 2005. The synapsids are represented today by mammals while reptiles by extant turtles, tuatara, lizards, snakes, crocodiles, birds, and their extinct relatives, including dinosaurs and pterosaurs. Over time, reptiles have evolved highly diverse morphological and physiological traits that allow them to exploit various ecological niches and resources on the land, in water, and in the air. Morphology of the skull of reptiles, especially the temporal region is highly diverse (Figure 1). This morphological diversity observed in the temporal region is broadly categorized into three architectural patterns. The anapsid condition of the skull is observed in basal amniotes such as Scutosaurus and Captorhinus, and in turtles. In these animals, the temporal region of the skull is completely roofed by bones, without temporal openings (fenestrae). The synapsid condition of the skull is recognized in ancestral lineages of extant mammals and is characterized by the presence of a temporal fenestra at lower position of either side of the skull. The diapsid condition of the skull is seen in "non-turtle" extant reptiles and in their extinct relatives. In these animals, two temporal fenestrae exist on either side of the skull. In reptiles with fully diapsid skulls, the upper temporal fenestra is dorsally bordered by the parietal bone, anteriorly by the postorbital bone, and posteriorly by the squamosal bone. The lower temporal fenestra is dorsally surrounded by both the postorbital and squamosal bones, and ventrally by both the jugal and quadratojugal bones. These temporal fenestrae are thought to have evolved to allow space for accommodating the enlarged jaw-closing muscles that enable powerful biting, or to minimize stresses exerted by the contraction of jaw muscles on the skull, or to reduce the weight of the skull itself (Frazzetta, 1968, Carroll, 1982, Rieppel, 1993a, Benton, 2005. During the course of diapsid evolution skull morphology has been rearranged repeatedly, resulting in a variety of modified patterns (Rieppel, 1993a, Rieppel & Gronowski, 1981, Müller, 2003, Moazen et al., 2009). Among extant reptiles, fully diapsid skull is only seen in tuatara and crocodiles. Because the lower temporal bar that encloses the lower temporal fenestra ventrally is regarded to be lost once in the common ancestor of lepidosaurs (lizards, snakes, tuatara) and archosaurs (crocodiles and birds) (Rieppel, 1993a, Müller, 2003, these reptilian lineages possibly acquired the lower temporal bar secondarily (Müller, 2003, Moazen et al., 2009. Both snakes and birds have lost the upper temporal bar so that their temporal region is free from any bony frames (Pough et al., 2005).
In reptiles, phylogenetic position of turtles is highly controversial. Traditionally, turtles have been regarded as the only surviving clade of stem reptiles based on the pattern of their skull morphology: an anapsid skull whose temporal region is completely roofed with bones (Williston, 1917, Gregory, 1946, Romer, 1968, Gaffney, 1980, Reisz & Laurin, 1991, Lee, 1993, Laurin & Reisz, 1995, Lee, 1996, Reisz, 1997, Lee, 2001. However, recent comprehensive analysis of morphological traits (Rieppel & deBraga, 1996, deBraga & Rieppel, 1997, Rieppel, 2000, Hill, 2005, Li et al., 2008, but see Lyson et al., 2010Lyson et al., , 2013 for opposed conclusion) and molecular phylogenetic studies (Hedges & Poling, 1999, Kumazawa & Nishida, 1999, Iwabe et al., 2005, Hugall et al., 2007, Shedlock et al., 2007, Shen 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 suggest that there is a close relationship of turtles to diapsid reptiles, implying that the temporal fenestrae were secondarily closed in turtles. In this study, we employ the hypothesis that turtles are descendent of diapsid reptiles. Although skull morphology has been regarded as an important character in classification of reptiles and in understanding the ecological and physiological aspects of each reptilian species, the developmental mechanism underlying diversification of reptilian skull morphology is poorly understood (Rieppel, 1993a, Evans, 2008. As a consequence, a general genetic and developmental model of reptile skull diversity does not yet exist. In this paper, we test the hypothesis that changes of skeletal gene expression patterns cause diversification of reptilian skull morphology through comparative analyses of gene expression in the Figure 1 Simplified phylogeny of the Reptilia highlighting diversity of their skull morphology. Paleontological evidence suggests that all reptiles, including extant lizards, snakes, tuatara, crocodiles, birds, and turtles, were derived from ancestor whose temporal region was completely roofed by bone. Earliest diapsid reptiles such as Petrolacosaurus (Araeoscelidia) acquired two temporal openings (fenestrae) on either side of the skull (red vertical bar). Recent molecular phylogenies indicate that turtles (Testudines) were derived from diapsid ancestor, which would require secondary closure of temporal fenestrae (green vertical bar). The bone surrounding anteroventral border of the upper temporal fenestra and anterodorsal border of the lower temporal fenestra in diapsids: the postorbital was colored in pink. The bone surrounding posteroventral border of the upper temporal fenestra and posterodorsal border of the lower temporal fenestra in diapsids: the squamosal was colored in blue. The bone surrounding the anteroventral margin of the lower temporal fenestra in diapsids: the jugal was colored in yellow. Lizards do not have the lower temporal bar. Both upper and lower temporal bars are absent in snakes and birds. Other extinct diapsid lineages such as Ichthyosauria and Sauropterygia were not included in phylogeny for simplicity. embryos of representative reptilian species and reveal a potential developmental basis underlying reptilian skull evolution. First, we describe the pattern of early phases of cranial morphogenesis in a crocodile species with both upper and lower temporal bars surrounding temporal fenestrae, using molecular markers specific for musculoskeletal tissue precursors. Then, we compare these data with cranial morphogenesis in a turtle species. We found a broader expression of the early osteogenic genes, Runx2 and Msx2 in the mesenchymal cells at the temporal region of turtle embryos, compared to that in crocodile embryos. Finally, to obtain a broader picture of reptilian skull morphogenesis, we examined expression patterns of Runx2 and Msx2 in cranial morphogenesis of a snake species without temporal bars on the skull and compared with the patterns in crocodile and turtle embryos. Our findings suggest that there is a possible correlation between the expression patterns of Runx2 and Msx2 and the architectural pattern seen in the temporal region of the reptilian skull.

Results
In previous studies in which cranial osteogenesis of reptilian embryos was described, whole-mount clearing and staining with Alizarin red was used to detect mineralization of intramembranous bones that comprise the dermatocranium (Kamal et al., 1970, Haluska & Alberch, 1983, Rieppel, 1993b, Rieppel, 1993c, Kuratani, 1999, Rieppel & Zaher, 2001, Sheil, 2003, Sheil, 2005, Vickaryous & Hall, 2008, Werneburg et al., 2009). However, this method is unable to identify the distribution of the precursor cells of bones: osteoblasts, as reported by others (Kerney et al., 2010). To overcome this, we conducted section in situ hybridization analysis, which labels tissues located deep inside of the embryonic body and is effective for detecting tissue-specific domains of expression. We used a probe to Runx2, which is a molecular marker for osteogenic mesenchymal precursor cells (Ducy et al., 1997, Bobola et al., 2003, Abzhanov et al., 2007, Han et al., 2007, Kerney et al., 2010 and described its expression pattern in the temporal region of reptilian embryos where mineralization of bones has not been initiated. Furthermore, to describe distribution pattern of "non-osteoblast" cell lineages relative to that of osteoblasts in the cranial tissue of the embryos, we also examined expression of other tissue-specific markers: MyoD for skeletal muscle precursors (Hacker & Guthrie, 1998, Noden et al., 1999, Sox9 for cartilage precursors (Wright et al., 1995, Bell et al., 1997, Scleraxis (Scx) and Six2 for precursor of connective tissues, including ligaments and tendons (Oliver et al., 1995, Schweitzer, 2001, Dreyer et al., 2004, Edom-Vovard & Duprez, 2004, Schweitzer et al., 2010. 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, 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, 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 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. 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  Figure 4A. (A and G) In both crocodile and turtle embryos, expression of Bmp4 is detected at the mesenchyme distributed in medial part of the jaw primordia (arrows). (B and H) Expression of Msx1 is detected at the mesenchyme that occupies the domain close to jaw articulation and lateral to quadrate and Meckel's cartilages (arrows). (C and E) In crocodile embryos, Msx2 is expressed in a thin layer of mesenchymal cells surrounding dorsal aspect of the brain and in a population of the mesenchyme that occupies the domain between ventrolateral part of quadrate cartilage and surface epidermis (arrows). (I and K) In turtle embryos, Msx2 is expressed in mesenchymal cells that populate lateral aspect of the head (arrows). In contrast to the condition in crocodile embryos, the ventral edge of Msx2-expressing mesenchymal layer is terminated ventral to the eye in turtle embryos and these cells cover MyoD-expressing jaw adductor muscle precursor (J and L) laterally. The red outlined domains in (A-C, and E) indicate the location of the anlagen of the pseudotemporal muscle deduced from adjacent sections where muscular tissues are labeled by MyoD probe. The red outlined domains in (G-I, and K) indicate the location of the anlagen of jaw adductor muscle. Scale bars are 0.5 mm.
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, 1991, Lee, 1993Laurin & Reisz, 1995, Lee, 1996, 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 1997, Rieppel 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, 1999, Iwabe et al., 2005, Hugall et al., 2007, Shedlock et al., 2007, Shen 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, 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 2009). In turtle embryos where mineralization of dermal bones of the skull has not yet occurred, mesenchymal cells that express Runx2 were broadly  St.14/15: A population of cells medial to the precursor of the jaw adductor muscles; the mesenchyme localized at the domain dorsolateral and ventrolateral to the orbit; a thick layer of the mesenchymal cells that completely covers the brain and the precursor of jaw adductor muscle laterally.

St.26:
The mesenchyme occupying the space medial to the quadrate cartilage precursor; the mesenchyme ventral to the orbit; a layer of mesenchymal cells surrounding the brain laterally.
St.17: A thick layer of mesenchymal cells surrounding the braincase and jaw adductor muscle laterally; the mesenchyme associated with the quadrate cartilage and the ventral part of the braincase.
St.29: The precursors of palatine and pterygoid bones; the mesenchyme accompanying jaw cartilages; a layer of loose mesenchyme that later forms a precursor of parietal bones; the precursors of the maxilla bones.
St.31: The precursors of the dermatocranial elements, including the parietal.
MyoD St.14/15: Precursor cells of each jaw muscle in the first pharyngeal arch; eye muscle precursors.
St.14/15: The primordia of jaw and eye muscles.
St.26: The primordia of jaw and eye muscles.

St.26: The precursors of quadrate and
Meckel's cartilages and the braincase; a layer of mesenchyme surrounding the brain laterally.
St.14/15: A layer of mesenchymal cells located at the periphery of the jaw adductor and eye muscle precursors.
St.17: Tendinous tissues at the periphery of jaw adductor muscles; the precursor of the bodenaponeurosis (central tendon of external adductor) within the jaw adductor muscular tissue.

Six2
St.14/15: The mesenchyme surrounding the eyes and cartilaginous precursors of the braincase, quadrate, and Meckel's; the mesenchyme between jaw muscle precursors and the skeletal tissues to which the muscles attach; the mesenchyme that dorsally surrounds the brain.
St.14/15: The mesenchyme surrounding the eye; the mesenchyme associated with the braincase and jaw cartilages; the mesenchyme within the jaw muscle precursors.
-St.17: The mesenchyme localized around the jaw articulation between quadrate and Meckel's; the mesenchyme associated with the braincase, postorbital bone, and jaw muscles.
Bmp4 St.14-17: The epithelium of cochlearcanal; the mesenchyme surrounding the eye; the mesenchyme distributed in the medial part of jaw primordia; the precursors of the palatine bones; a population of mesenchymal cells covering the brain dorsally.
St.14-17: The epithelium if cochlear canal; The mesenchyme dorsolateral and ventrolateral to the eye; a limited population of the mesenchyme in close proximity of the jaw articulation. -

Msx1
St.14-17: The epithelium of the cochlear canal; the mesenchyme adjacent to the jaw articulation; the mesenchyme lateral to the quadrate and Meckel's cartilages; a thin layer of mesenchymal cells covering the brain dorsally.
St.14-17: The epithelium of the cochlear canal; the mesenchyme adjacent to the jaw articulation; the mesenchyme lateral to quadrate and Meckel's cartilages; the mesenchyme that populates the domain dorsal to the eye.
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 , 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, 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, 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. 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. 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 Table 1 Expression domains of the genes in the head of crocodile, turtle, and snake embryos (Continued) Msx2 St.14/15: A thin layer of mesenchymal cells surrounding the dorsal aspect of the brain; the mesenchyme located dorsal to the eye; the mesenchyme occupying the domain between the ventrolateral part of quadrate cartilage and surface epidermis.
St.14-17: Mesenchymal cells populating lateral aspect of the head (lateral to external adductor muscle).

St.26:
The mesenchyme medial to the quadrate precursor; a mesenchymal layer surrounding the brain dorsally.
St.17: A population of mesenchymal cells in close proximity of postorbital and quadratojugal bone precursors; a thin layer of the mesenchyme surrounding the brain dorsally where future parietal bones form.
St.29: The precursors of palatine and pterygoid bones; the mesenchyme accompanying jaw cartilages; a layer of loose mesenchyme that later forms a precursor of parietal bones.
St.31: The precursors of the dermatocranial elements, including the parietal.
embryos has not been identified, recent findings that early stage arrest of Msx2 expression in neural crestderived 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 Figure 9 Potential developmental basis that generates morphological diversity in the temporal region of the reptilian skull. All extant reptilian lineages are considered to be derived from ancestor with diapsid skull. In crocodiles that have both upper and lower temporal bars like the stem Diapsida (e.g., Petrolacosaurus), osteogenic mesenchymal precursor cells which express Runx2 and/or Msx2 are distributed at the domain where future temporal bars are formed in the head of early stage embryo (top of the middle column). Through ontogeny (black arrow), these osteoblast precursors may differentiate into the dermatocranial elements including upper and lower temporal bars (bottom of the middle column). Between these bony bars, both upper and lower temporal fenestrae are clearly recognized. In turtles (left white arrow), distribution of osteogenic mesenchymal precursor cells is broadened in a dorsal-ventral direction, filling the whole lateral portion of the head of early embryos (top of the left column). Through ontogeny, these osteoblast precursors may differentiate into the dermatocranial elements roofing the temporal region of the head (bottom of the left column). In snakes that have modified diapsid skull where temporal bars are absent, osteogenic mesenchymal precursor cells do not fill lateral domain of the head of embryo. Rather, these cells are mainly distributed in the vicinity of the brain (top of the right column). Through ontogeny, these osteoblast precursors may differentiate into the dermatocranial elements accompanying the braincase, without forming bony temporal bars (bottom of the right column). A condensed mesenchymal layer that differentiates into the braincase in later stages is highlighted by dotted line in the head of embryos. 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 2005, 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, 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, 1994, 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 earlyphase 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 (Macrochelys temminckii) (Sheil, 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. 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 softshelled 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, 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 quadratomaxillary ligament between jugal and quadratojugal bones (Rieppel, 1993a, 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 reacquired in the lineages of tuatara and crocodiles, as well as in several extinct reptilian lineages (Rieppel, 1993a, 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, 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.

Conclusions
Regulatory modifications of Runx2 and Msx2 expression in osteogenic mesenchymal precursor cells are likely involved in generating morphological diversity in the temporal region of the reptilian skull, including secondary closure of the temporal fenestrae in turtles. Our findings demonstrate that not only heterochrony in ossification of the dermatocranium that has been traditionally regarded as the major factor producing diversity of reptilian cranial morphology but also heterotopy in distribution of the osteogenic precursor cells may play a fundamental role in this process and it should be further investigated in future studies of reptilian cranial development and evolution.

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 embryos was performed on the basis of staging table of Thamnophis sirtalis (Zehr, 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 Msx1antisense 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, 2008, Bona & Desojo, 2011 for crocodile, (Schumacher, 1973, Rieppel, 1990, Rieppel, 1993c, Werneburg, 2012a, 2012b for turtle, and (Kamal et al., 1970, Haas, 1973, Zaher, 1994 for snake.

Competing interests
The authors declare that they have no competing interests.

Authors' contributions
MT conceived and designed the study. MT, WC, and JS contributed to collection and preparation of biological materials. MT performed experiments and analyses and wrote the paper. All authors read, discussed and approved the final manuscript.