Sample collection

E. burgeri embryos were collected as described previously (Ota et al., 2007), staged according to the method of Dean (1899), and fixed with Serra's fixative or 4% paraformaldehyde.

Collection of lamprey samples

Ammocoete and adult lampreys, Lethenteron reissneri, a brook lamprey species closely related to L. japonicum, were collected from the rivers in Nagano Prefecture and from Lake Biwa in Saga Prefecture by the Kurosawa Inc. (Saku, Sinsyu, Japan) and the Biwako Satellite-area Research Group, Japan. They were fixed in Serra's fixative overnight at 4°C and then gradually dehydrated in a graded series of methanol and phosphate-buffered saline (PBS). Finally, they were stored in 100% methanol at -20°C.

Histology

To avoid distortion of histologic sections, and especially to preserve the cartilage, we used Kawamoto's Film Method and a Paraffin Section Preparation Kit (Section Lab Co. Ltd.; see  http://sectionlab.jp/English.htm). The adhesive side of the paraffin transfer film (fine type) was attached firmly to the cut surface of the paraffin block, and each section was cut slowly at 6 to 8 mm thick with a microtome. One side of the cut “film-sections” was firmly placed onto a glass slide with the sectioned side down, and the slide was bathed in water at 45°C for 10 s. The films were then pressed onto, and attached to, the glass slide with a rubber roller and dried on a hot plate at 50°C for 24 h to transfer the embryonic tissues onto the slides. The adhesive compound and paraffin were removed by treatment with xylene for 1 to 3 days and then used in histological, immunohistochemical, and in situ hybridization procedures. Images were recorded with a DP70 digital camera (Olympus Inc., Tokyo, Japan) attached to a light microscope and reconstructed with a computer graphics program Avizo (Visualization Sciences Group).

In situ hybridization

In situ hybridization was performed either by using a manual standard protocol or a Ventana automated instrument (Roche, Japan). In the standard protocol, serial sections were fixed for 10 min in 4% paraformaldehyde in PBS at room temperature, washed twice in PBS, treated with proteinase K in 0.01 M Tris buffer for 10 min, and then fixed again for 10 min in 4% paraformaldehyde at room temperature. After rinsing twice in PBS, the sections were incubated with 0.25% acetic anhydride and 0.1 M triethanolamine (pH 8), washed in PBS, air dried, and hybridized with riboprobes at 51°C for 16 to 20 h. The sections were then washed in 5 × saline sodium citrate (SSC) buffer at 55°C, treated with 50% formamide in 2 × SSC at 60°C for 20 min, then washed once in 2 × SSC and twice in 0.2 × SSC, at 60°C each, for 20 min each. After being blocked with 1.5% blocking reagent (Roche) in 0.1 M Tris buffer with 0.15 M NaCl (pH 7.6), the sections were incubated with alkalinephosphatase-conjugated anti-digoxigenin (DIG) antibody (Roche). After final washes of the sections with Tris buffer, positive cells were stained purple with nitroblue tetrazolium salt (NBT) and 5-bromo-4chloro-3-indolyl phosphate toludinium salt (BCIP). With the Ventana instrument, signals were detected and counterstaining performed by using a BlueMap NBT/BCIP substrate kit (Roche) and a nuclear fast red-equivalent reagent, ISH RED, (Roche), as described previously (Ota et al., 2007).

Immunohistochemistry and histochemistry

Histologic observations were made on hematoxylin- and eosinstained sections (thickness, 6 to 8 µm), some of which were stained further with 0.1% Alcian blue to show the cartilage in older embryos of hagfish and adult lamprey and the mucocartilage in ammocoetes. To detect axon bundles, anti-acetylated tubulin was applied to sections after in situ hybridization or Alcian blue staining. Anti-mouse lgG1 was used as the secondary antibody. All histologic images were recorded with a DP70 digital camera (Olympus) attached to a light microscope.

Stages 40 to 50

In vertebrate embryos, neural crest-derived ectomesenchyme occupies primarily the rostroventral portion of the head, including the pharyngeal arches, whereas mesodermal mesenchyme occupies the dorsal and caudal perineural portions of the head (reviewed by Noden, 1988). To obtain insights into the distribution pattern of neural-crestderived ectomesenchyme in the prechondogenic stages of E. burgeri embryos, we first observed the expression patterns of four Dlx genes (EbDlx2/3/5B, EbDlx2/3/5C, EbDlx1/4/6A, EbDlx1/4/6C[not shown]; Fujimoto et al., 2013) in stage 45 E. burgeri embryos, corresponding to the mid-pharyngula stage of this animal. Dlx genes, which are homeobox-containing regulatory genes homologous to Distal-less in Drosophila, are ectomesenchymal markers in the region including, and caudal to, the mandibular arch (see Kuratani et al., 2012). The transcripts of all the Dlx genes showed similar distribution patterns: Dlx-positive ectomesenchyme was found in the mandibular and posterior pharyngeal arches, as well as in the caudal and lateral regions of the posthypophyseal process (PHP), which is a cyclostome-specific craniofacial primordium (Oisi et al., 2013). The mesenchyme in the anterior nasal process (ANP), which is mainly of premandibular neural crest-origin in the lamprey (Kuratani, 2012), did not express any Dlx genes (Fig. 2A–G). Notably, the periotic mesenchyme was also negative for Dlx gene expression (Fig. 2A–C).

As noted previously (Shigetani et al., 2002), the premandibular ectomesenchyme (the ectomesenchyme distributed rostral to the mandibular arch) is specifically Dlx-negative in jawed vertebrate embryos. In contrast, in the lamprey embryo, part of the premandibular ectomesenchyme destined to form the upper lip of the ammocoetes larva expresses Dlx genes as specifiers of the oral apparatus (Shigetani et al., 2002). However, this phenomenon can be ascribed to the secondary migration of mandibular arch crest cells into the premandibular (including PHP) domain covering the premandibular ectomesenchyme laterally, as implied by cell lineage studies (Kuratani, 2012; Kuratani et al., 2012). Thus, the presence of Dlx-positive and -negative mesenchyme in the hagfish PHP is consistent with observations of the lamprey.

To identify mesodermal mesenchyme in the hagfish embryos, we also observed the expression patterns of EbTbx1/10A (the hagfish homolog of Tbx1/10, the marker gene for prochondrogenic head mesoderm; Oisi et al., 2013) and of EbSoxE (a Sox10 homolog) in E. burgeri (Fig. 2H–M; Ota et al., 2007). At stage 40 there was no sign of chondrification, as detected by SoxE expression in the embryonic head (Fig. 2H). Instead, a clear accumulation of mesenchyme expressing EbTbx1/10A was seen to surround the otocyst, prefiguring the future otic capsule (Fig. 2I). This mesenchyme appeared to correspond to the periotic mesenchyme found in gnathostome embryos (Ladher et al., 2005; Monks and Morrow, 2012). At stage 50, the above-noted EbTbx1/10A-positive mesenchyme could be followed rostrally, as a longitudinal strand of cells, toward the ventral aspect of the eye (Figs. 2J, K, 3), possibly prefiguring trabeculae of this animal (Fig. 3C, D).

Fig. 2.

Embryonic expression of genes for craniofacial structure in Eptatretus burgeri. Histologic sections of E. burgeri at stages 45 (A–E), 40 (H, I), 50 (J, K), and 51 (L, M), hybridized with Dlx1/4/6A (A), Dlx2/3/5B (B), Dlx2/3/5C (C–E), Tbx1/10A (l–K, M), or SoxE (H, L) riboprobes. Also dorsal view of reconstruction (F) and schematic diagram of Dlx genes expression in the mid-sagittal region (G) of a stage 45 embryo (dark gray). At stage 40 (I), Tbx1/10A, the mesodermal neurocranial marker gene, is expressed in the pharyngeal arch muscle anlage as well as in the mesenchyme surrounding the otocyst, representing the otic capsule anlage. By stage 50 (J, K), the mandibular arch muscle anlage has differentiated into tentacular and lingual muscle primordia, and the periotic prochondrogenic mesenchyme has grown rostrally to form the common anlage for trabeculae and the dorsal longitudinal bar. (L) SoxE expression at stage 51 depicts a rostrally growing longitudinal prochondrogenic anlage, continuous with the otic capsule. Abbreviations: anp, anterior nasal process; e, eye; mm, mandibular mesoderm; ot, otocyst; p1, 2, pharyngeal pouches 1, 2; php, posthypophyseal process; pom, periotic mesenchyme. See the list in the text for other abbreviations. Bars = 100 µm.

Fig. 3.

Mesenchymal cranial primordium of a stage 50 Eptatretus burgeri embryo. 3D reconstructions were based on sectioned specimens hybridized with Tbx1/10A (A–D) riboprobes. (A) Dorsal view. (B) Ventral view. (C) Left lateral view. (D) Oblique posterior view. The ectodermal oronasohypophyseal cavity is colored light blue, the pharyngeal endoderm is colored yellow, and the Tbx1/10A positive-mesoderm colored pink, mphp, muscle primordia in the php; vm, velum muscle; nt, notocord; trab, trabecula. See the list in the text for other abbreviations.

Comparison of this pattern of expression with that of the Dlx genes at stage 45 suggested that this mesenchyme represented either the head mesodermal derivative or the premandibular ectomesenchyme (Fig. 2A–G).

Stage 51

By stage 51, when EbSoxE was being upregulated in the otic capsule, EbTbx1/10A had begun to be downregulated in the above-noted periotic cartilage (Fig. 2L, M). Stage 51 was the youngest stage at which reconstruction of the prechondrogenic primordial cranium could be performed on the basis of the expression of both EbSoxE and EbTbx1/10A (Fig. 4). This stage of E. burgeri appeared to resemble “Stadium I” (i.e., stage I) of the “Bdellostoma” (Eptatretus stouti) embryo described by Neumayer (1938) and was younger than the Myxine embryo described by Holmgren (1946). Comparison of Neumayer's Stadium I and our stage 51, however, was not simple solely on the basis of cranial morphologic examination, as our observation was based on EbSoxE-positive mesenchyme, part of which was not recognized as a cartilage precursor by these previous authors.

By this stage, the anterior nasal process of the E. burgeri embryo had developed a transverse cartilage primordium rostral to the nasohypophyseal plate (nhp), the common placode that differentiates into the nasal epithelium and adenohypophysis (Kupffer, 1899, 1900). From its position relative to the forebrain, this cartilage appeared to correspond to a common precursor of the future “anterior and posterior vertical nasal bars” (avnp and pvnb) of Holmgren (1946) and the “cartilago nasalis posterior et anterior” of Neumayer (1938) (Fig. 4A, F).

The mesenchyme ventral to the otocyst formed an overt primordium of the otic capsule (otc), which had a large dorsal fenestra and in which the otocyst was embedded (Fig. 4A, B). This cartilaginous capsule was medially continuous with the cranial floor on both sides of the notochord (Fig. 4A, B). This cranial base may have contained the skeletal elements generally called the parachordals; Neumayer (1938) thought that the medial wall of this capsule represented the parachordals. Holmgren (1946), on the other hand, described an independent primordium for the parachordals as a separate element medial to the otic capsule. However, we were unable to detect any independent anlage—even by the expression of EbSoxE in the prochondrogenic mesenchyme—that would have implied the separate origin of parachordals distinct from the otic capsule.

From the anterolateral aspect of the otic capsule, a thick longitudinal prochondrogenic mesenchymal bar grew rostrally toward the area ventral to the eye primordium and toward the ventrolateral aspect of the forebrain (Fig. 4A–C); this bar corresponded to the common anlage for the trabecule and the dorsal longitudinal bar described in the hagfish (Holmgren, 1946; see below).

In the ventral part of the mandibular arch, which differentiates into the tongue apparatus of this animal (Yalden, 1985), an extensive sheet-like prochondrogenic primordium was observed on the oral floor, representing the early development of the lingual plate (lp; Fig. 4D, F). The position of this prochondrogenic mesenchyme corresponded to the site occupied by the ventral plate of mucocartilage (ventrolateral plate) in the ammocoete larva of the lamprey ( Supplementary Figure S1 online (10.2108.zsj.30.944.s1.pdf); Gaskell, 1908; Holmgren and Stensiö, 1936). Curiously, this plate in the hagfish is turned upward to form a hook at this stage (Fig. 4E, F). Also, there was an arch-like connection—the extrahyal (exhy) of Holmgren (1946)—linking the lateral aspect of the lingual plate (Fig. 4D–F) and the above-noted longitudinal bar at the level of the hyoid arch and representing a hyoid arch element (Fig. 4E).

Fig. 4.

Prochondrogenic cranial primordium of a stage 51 Eptatretus burgeri embryo. 3D-reconstructions were based on SoxE and Tbx1/10A expression. (A, F) Reconstructions of the cranial primordium. (B–E) Reconstructions with epithelial structures (ectodermal oronasohypophyseal cavity, light blue; pharyngeal endodermal lining, yellow; Tbx1/10A-positive head muscle primordium, pink). (A, B) Dorsal views. (D) Ventral views. (E, F) Left lateral views. lp, lingual plate; otc, otic capsule; t1 and t3, common anlage for tentacular cartilages I and III; trab + dlb, common anlage for trabeculae and the dorsal longitudinal bar; vb, velar bar. See the list in the text for other abbreviations.

Rostrally, the lingual plate continued into a pair of prochondrogenic nodules below and posterior to the nasal cartilage primordium noted above. Situated in the rostral part of the early PHP, these nodules appeared to differentiate into tentacular cartilages, most probably corresponding to those for tentacles 1 and 3 (Fig. 4F; see below).

As a mandibular arch element, a ventral process grew medially from the common anlage of the trabecule and the dorsal longitudinal bar (Fig. 4A, C), close to the junction of the ectodermal oronasohypophyseal cavity and the pharyngeal endoderm (Fig. 4E). This mesenchymal process represented the extrapalatoquadrate of Holmgren (1946) and was associated with the primordium of the velar bar (vb in Fig. 4C, E, F).

Stage 53

From this stage onward, reconstruction of the hagfish embryos was based on cartilaginous tissues, which are easily detected in histologic sections (Figs. 5 and 6). As described previously (Oisi et al., 2013), the PHP-derived oronasohypophyseal septum had grown completely, separating the nasohypophyseal duct and oral cavity dorsoventrally by this stage (Figs. 5A, D and 6A, C). Some authors (e.g., Stensiö, 1927) also used the term “palato-subnasal lamina”.

The overall morphology of the chondrocranium of stage 53 E. burgeri embryo (Figs. 5 and 6) had a conspicuous resemblance to that of the Myxine embryo described by Holmgren (1946). By this stage, the nasal epithelium had differentiated to acquire the characteristic zigzag, comb-like morphology (ne; Fig. 6B–D). On the anterior and the posterior aspect of this cavity, two transverse cartilaginous bars had appeared (compare Figs. 5B, D to 6B–D). The posterior bar corresponded to the “posterior vertical nasal bar (pvnb)” and the anterior one to the “anterior vertical nasal bar” (avnb) of Holmgren (1946), or to the “Vordere Nasalknorpel” (anterior nasal cartilage; Neumayer, 1938). As noted above, these cartilages appeared to have been derived from the single common anlage observed in the ANP of the previous stage (Fig. 4A, F). Unlike the posterior bars, which remained separated bilaterally, the anterior bar by this stage had formed a complete arch rostral to the nasal cavity (Figs. 5B, 6B, D).

Fig. 5.

Chondrocranium of a stage 53 Eptatretus burgeri embryo. Ventral (A, C), dorsal (B), and left lateral (D) views of a 3D-reconstructed model. Lingual cartilage is removed in (C). See the list in the text for abbreviations.

The anterior part of the oronasohypophyseal septum—the derivative of PHP—contained three pairs of rod-like cartilages, corresponding to the supporting skeletons for tentacles 1 to 3 (H to t3 in Figs. 5A, D and 6A, C, D; the tentacular blastema of Holmgren, 1946). The cartilage bars in the second tentacle were fused posteriorly to form a median longitudinal rod, or subnasal cartilage (snc), which is also situated within the oronasohypophyseal septum, below the nasohypophyseal duct (compare Figs. 5C, D to 6A, C, D). The cartilaginous rods for tentacles 1 and 3, on the other hand, were posteriorly united with each other and further connected posteriorly, by means of a string of labial cartilage (lb), with the lingual cartilage complex located clearly on the oral floor or in the ventral part of the mandibular arch (Fig. 5A, D).

By this stage, trabeculae (“trab” in the definition of Holmgren, 1946) were seen as conspicuous longitudinal cartilages separated from the more ventrally located dorsal longitudinal bar (dlb; see below) and had become a pair of rods, posterior parts of which developed in a similar plane as that of the notochord. At the level of the adenohypophysis the rod sank below the nasohypophyseal duct and grew a medially oriented commissure that would later unite with its counterpart to form the “hypophyseal commissure” (hypcom; Holmgren, 1946; Figs. 5B–D and 6B–D). This commissure was found slightly rostral to the adenohypophysis (ah) (Figs. 5B, C and 6B, C). Rostral to the above commissure, the trabeculae grew more rostrally to unite with the posterior vertical nasal bar described above (Figs. 5B–D and 6B–D).

In addition to the trabecula, there was another longitudinal cartilaginous rod lateral and slightly ventral to the trabecula and the otic capsule (Figs. 5B–D). This rod corresponded to what was called the “dorsal longitudinal bar” by Holmgren (1946), although this is not the dorsalmost cartilage in the hagfish chondrocranium. By this stage, the dorsal longitudinal bar connected to the trabecula and otic capsule by means of cartilaginous communications at three places: with the rostral part of the trabecula (con1), with the posterior part of the trabecula (con2), and with the posteroventral part of the otic capsule (con3) (Figs. 5B–D and 6B–C). The major part of the dorsal longitudinal bar was located lateral and ventral to the nasohypophyseal duct; rostrally it continued into a cartilage called the “cornual cartilage” (Holmgren, 1946) and into a palatine bar, indicating that this cartilage, like the above-noted tentacular cartilages, develop in the derivative of the PHP of the earlier embryo (Oisi et al., 2013). From the Dlx-expression patterns of earlier embryos (Fig. 2A–G), as well as from the topographic position of the dorsal longitudinal bar, this cartilage appeared to have been derived either from the dorsal part of the original mandibular arch ectomesenchyme that had migrated rostrally into the lateral part of the PHP (for its possible similarity to the gnathostome palatoquadrate, see the Discussion) or from the more medially located mesenchyme. As observed in the patterns of EbTbx1/10A and EbSoxE expression, however, it became clear that the above-noted trabecula and the dorsal longitudinal bar differentiated from a common longitudinal prochondrogenic anlage observed at stage 51 (compare Fig. 4C to Fig. 6C′). Namely, the trabecula and the dorsal longitudinal bar were dorsolaterally separated from each other by fenestration out of this mesenchyme. There were two sites of fenestration; therefore, the above-noted three communications do not represent secondarily established commissure cartilages but instead the remnant of the original common anlage that has not been absorbed (Fig. 6C′).

Fig. 6.

The same chondrocranium as shown in Fig. 5, reconstructed with endodermal (yellow) and ectodermal (light blue) linings associated with the cranium. The Tbx1/10A-positive head muscle primordium is coloured pink.Ventral (A), dorsal (B), lateral (C, C′), and oblique anterior (D) views of a 3D-reconstructed model. Note in the lateral view (C) that most of trabeculae (trab) lies below the nasohypophyseal duct, or within the oronasohypophyseal septum (space between nasohypophyseal duct and oral cavity). The lingual plate lies below the oral cavity, or within the ventral part of the mandibular arch. See the list in the text for abbreviations.

Beneath the oral cavity, in the ventral mandibular arch domain that will differentiate into the lingual apparatus, the anlage of the lingual cartilages could still be identified as an undivided single sheet of cartilage (Figs. 5A, D and 6A, C). This cartilage appeared to contain two types of cartilaginous primordia. The rostral paired cartilage corresponded to the primordium of the “latero-rostral part of the basal plate (lvp)” (Holmgren, 1946). Holmgren labeled this structure at this embryonic stage “basil”. Medial and dorsal to these cartilages, the primodium of the dental plate (dp) was developing (dental cartilage of Marinelli and Strenger, 1954). Holmgren referred to the latter as the “medio-rostral part of the basal plate (mvp)” (“bas1m” of Holmgren, 1946). Caudal to the lvp, there was another pair of longitudinal plates labeled “bas3” by Holmgren (1946). According to Holmgren, there was another element at the junction of lvp and bas3, called the “medial part of the basal plate” (“bas2” of Holmgren, 1946); this was not discerned in our embryo at this stage.

Rostrally, the rostrolateral part of the lvp developed a rod-like communicating cartilage that grew laterally and rostrally to invade the PHP domain and establish a connection with tentacular cartilages 1 and 3 (Figs. 5D and 6C). This communicating cartilage, therefore, represents a commissure between two different craniofacial modules in the hagfish embryo.

Another group of cartilages to be noted at this stage is the pharyngeal archassociated cartilages of the hagfish. First, there are two pairs of thin cartilaginous arches connecting the dorsal longitudinal bar and the lingual cartilage primordium. They correspond to “extrabranchiale I (exbr1)” and “extrabranchiale II (exbr2)” of Holmgren (1946), belonging to the third and fourth pharyngeal arches, respectively (Figs. 5A, C, D and 6A, C). Anterior to these cartilages, another vertical cartilage had developed, connecting the dorsal longitudinal bar at the level of the posterior commissure dorsally with another newly formed ventral cartilage that had developed ventral to the rostralmost part of the pharynx (Figs. 5A, C, D and 6A, C). The former is called the extrahyal (exhy) and the latter the ventral branchial bar (vbrv) by Holmgren (1946). The topographic relationship between the main branch of the facial nerve, often called the hyomandibular, has been investigated to determine the second arch origin of the extrahyal (Neumayer, 1938); this was also ascertained by our study here (not shown).

Stage 60

The E. burgeri embryo at stage 60 particularly resembles “Stadium II” of E. stouti by Neumayer (1938); however, the rostrum of the latter chondrocranium appears to be somewhat compressed as a result of secondary distortion. Our reconstructions (Fig. 7) appear to reflect the normal proportion of the hagfish embryonic head at this stage. Although the morphology of the chondrocranium of E. burgeri at stage 60 was highly complex, it could readily be derived from that of the previous stage.

The most conspicuous change was the appearance of the nasal duct cartilages (ng), which developed rostral to the anterior vertical nasal band as a series of inverted U-shaped cartilaginous bands along the nasohypophyseal duct (Fig. 7A, E, F, H). Apparently, these cartilages support the hag-fish-specific elongated nasohypophyseal duct, and they are functionally and morphologically reminiscent of the tracheal rings of amniotes. In the lamprey, the nasohypophyseal duct of which does not elongate during development, no equivalent cartilages are present. The above-noted chondrocranium in E. stouti; does not possess nasal duct cartilages, although the specimen examined may have been at a slightly younger stage. Alternatively, as another possibility, the late development of the nasal duct cartilage may represent a species-specific heterochronic variation, because this cartilage in E. stouti is again only poorly developed in “Stadium III” (Neumayer, 1938), the overall chondrocranial morphology of which resembled that of the prehatching stage of E. atami.

Rostrally, the supporting cartilage appeared as an independent cartilaginous rod for the fourth tentacle (Fig. 7B, C, F, G). Nasal capsule had appeared as several longitudinal cartilaginous rods along the sulci of the nasal epithelium (Fig. 7A, E, H). In the velum, the velar bar had now grown along the longitudinal axis with the change in orientation of the velum characteristic to the hagfish (Fig. 7F).

Fig. 7.

Chondrocranium of a stage 60 Eptatretus burgeri embryo. (B, C, F, G) Reconstruction of the chondrocranium. (A, D, E, H) Reconstruction with epithelial structures (ectodermal oronasohypophyseal cavity, light blue; pharyngeal endodermal lining, yellow). (A, B) Dorsal views. (C, D) Ventral views. (E, F) Left lateral views. (G, H) Oblique anterior views. See the list in the text for abbreviations.

Prehatching stage

For the embryonic stage close to hatching (about six months after stage 60) we used E. atami, a species closely related to E. burgeri (Figs. 8 and 9). This specimen was one of a number of specimens donated by the Kasai Marine Aquarium, Tokyo, that were still in their shells; its siblings hatched in the laboratory. As described below, the morphology of the embryo examined was entirely consistent with the adult morphology of E. burgeri, and we considered that it represented the equivalent developmental stage in that species.

The snout of E. atami at this stage had lifted upward to form a longitudinally elongated head characteristic of the adult hagfish described by Müller (1834, 1839) and Parker (1883a, 1883b) (Figs. 8 and 9). In every aspect, the configuration of the chondrocranium at this stage was nearly identical to that of the adult E. burgeri (Fig. 1). Owing to the secondary shift in head morphology, the position of the dorsal longitudinal bar had also been lifted to reach a level approximately identical to the horizontal level of the trabecula, giving the impression that these longitudinal cartilages together form the lateral wall of the neurocranium of this animal, similar to the orbital cartilage in the adult lamprey or the taenia marginalis, spheno-orbital commissure, and orbitotemporal commissure of jawed vertebrates (De Beer, 1937).

Fig. 8.

Chondrocranium of a prehatching-stage Eptatretus atami embryo. (B) Reconstruction of the chondrocranium. (A, C) Reconstruction of the chondrocranium with epithelial structures (ectodermal oronasohypophyseal cavity, light blue; pharyngeal endodermal lining, yellow). (A. B) lateral views. (C) Oblique anterior view. See the list in the text for abbreviations.

Marinelli and Strenger (1954) have also called this cartilage the taenia dorsolateralis. However, the above-mentioned cartilages do not appear to correspond to the neurocranial lateral wall; instead, they more likely to represent an ectomesenchymal derivative (see below).

Another cartilage that first appeared at this stage was the subnasal cartilage found as a single median cartilage beneath the nasohypophyseal duct (and therefore within the oronasohypophyseal septum) and continuing rostrally to bifurcate bilaterally into the supporting cartilage for tentacle 2 (Figs. 8 and 9). As suggested by our own previous study (Oisi et al., 2013), this cartilage appears to correspond to the dorsomedial supporting tissue including rostral dorsal plate in the ammocoete larva of the lamprey ( Supplementary Figure S1 online (10.2108.zsj.30.944.s1.pdf)).

Between the anterior and posterior vertical nasal bars, along the longitudinal depressions of the nasal cavity, at this stage there were several longitudinal cartilage rods, connecting the two vertical bars to form the nasal basket (Figs. 8 and 9). Curiously, some of the longitudinal cartilages are reported to develop in “Stadium II” (stage II) of E. stouti (Neumayer, 1938), implying that there are heterochronic variations in chondrogenesis among members of the hagfish group (Hypetotreti, or Myxiniformes).

The problem of hagfish and lamprey trabeculae

Trabeculae of the hagfish arise as dorsal parts of the common, prochondrogenic mesenchymal anlage for the trabeculae and the dorsal longitudinal bar, which grows from the rostral aspect of the otic capsule. In the developmental context the nature of the so-called trabeculae in the lamprey chondrocranium is important. In terms of morphology (a pair of rod-like cartilages surrounding the adenohypophysis), this cartilage has often been compared with trabeculae in jawed vertebrates (Damas, 1944; reviewed by De Beer, 1937). As is now well established experimentally, these cartilage are neural crest-derived prechordal cranial elements (Couly et al., 1993; Wada et al., 2011). The lamprey trabeculae, however, were observed by Koltzoff (1901) and Filatoff (quoted by Sewertzoff, 1916) to differentiate from the head mesodermal element (reviewed by De Beer, 1937). Morphologically, as well, the cyclostome trabeculae have been compared with the parachordals of crown gnathostomes (Sewertzoff, 1916; Neumayer, 1938). Johnels (1968) also supported the mesodermal origin of these elements from a morphologic and embryologic viewpoint (the earliest primordium appears dorsal to the first aortic arch, lateral to the notochord). The position adjacent to the notochord is more suited to mesodermal neurocranial elements that require notochordally derived signals to chondrify (Couly et al., 1993). A cell-labeling study using a vital dye has revealed that the greater part of the lamprey trabeculae is derived from the mandibular mesoderm (Kuratani et al., 2004; but see also Newth, 1956; Langille and Hall, 1988). Thus the trabeculae in the lamprey appears to be a misnomer and are more likely to represent mesodermally derived parachordals that extended anteriorly beyond the level of the hypothalamus (Kuratani et al., 2004).

From its position in the embryonic head, the trabeculae of the hagfish, described here, do not appear entirely equivalent to those of the lamprey. First, as described above, in the hagfish components corresponding to the lamprey trabeculae are composites of the trabecula and the dorsal longitudinal bar (Fig. 4C). Previously, only the dorsal component of this complex, lying rostral to the rostral tip of the notochord, was defined as hagfish trabeculae. The portion of the cranial base associated with the notochord was called the hagfish “parachordals” (Neumayer, 1938; Holmgren, 1948). Unlike the lamprey trabeculae, the posterior part of which is found adjacent to the notochord, most of the hagfish trabeculae is located rostral to the notochord in later development (Fig. 4). Importantly, however, the putative paraxial mesodermal component of this cartilage is found lateral to the notochord in the prechondrogenic stages (Fig. 3A, C). Therefore, rostral extension of the parachordals is likely to take place in the hagfish. At least the posterior portion of the hagfish trabeculae that grow from the rostral part of the otic capsule and lies slightly dorsal to the nasohypophyseal duct would correspond to the parachordals in jawed vertebrates, as a large part of the lamprey trabeculae do. The rostral halves of the hagfish trabeculae, on the other hand, may be more similar, if not entirely homologous, to trabeculae of jawed vertebrates (see below). Its position within the PHP appears to be consistent with this assumption.

Second, the hagfish trabeculae are, as a whole, widely separated from each other and do not appear to have originated from paraxial mesoderm in the head, although the possibility remains that the hagfish trabeculae correspond to the rostral, widely expanded portion of the trabeculae in the lamprey.

Third, especially in terms of the position of the hypophyseal commissure and the longitudinal commissure connected to the posterior nasal cartilage, the rostral portion of the hagfish trabeculae more closely resemble the typical prechordal cranium in jawed vertebrates. Moreover, as inferred from the embryology of the lamprey, the homolog of the jawed vertebrate trabeculae in the hagfish should be sought in cartilages derived from the hagfish PHP or from the oronasohypophyseal septum (Kuratani et al., 2004; Oisi et al., 2013; see below). Thus, although the neural crest origin of this cartilage needs to be demonstrated, it is possible that rostral parts of the hagfish trabeculae+dorsal longitudinal bars are similar to the crown gnathostome trabeculae. As noted above, examining the Dlx gene expression pattern will not answer this question, and further labeling studies are needed to determine the nature of the hagfish trabeculae.

The problem of the cyclostome pharynx and the mandibular arch

Crown gnathostomes are characterized by differentiation of the mandibular arch into upper and lower jaws. This dorsoventral division and morphological specification are shared by all jawed vertebrates; the cartilaginous primordium of the dorsal moiety is generally called the palatoquadrate, and the ventral moiety is called Meckel's cartilage (reviewed by Goodrich, 1930 and by De Beer, 1937).

Recent molecular genetic studies have shown that the identity of the mandibular arch as the rostralmost element of branchiomerism (the metameric arrangement of pharyngeal arches) is specified by the absence of Hox gene transcription in this arch (Hox code-default state; Rijli et al., 1993), which is shared by the lamprey and gnathostomes (Takio et al., 2004). Thus, the morphological identity of the mandibular arch appears to be universal among vertebrates. For the more caudally located pharyngeal arch skeletons in hagfish, the previous homologizations between hagfishes and lampreys performed by several authors are mostly correct in terms of cranial nerve innervation patterns (Lindstrom, 1949; Homma, 1978; Koyama et al., 1987; Nishizawa et al., 1988; Ronan, 1988; Song and Boord, 1993) and the distribution of pharyngeal arch muscles described by Marinelli and Strenger (1954, 1956); both of which were confirmed by our anatomic studies of adult hagfish and lamprey (Oisi et al., 2013). Division of the mandibular arch into upper and lower elements, on the other hand, would require careful treatment.

It was also molecular genetic studies that elucidated the basic mechanism specifying dorsoventral polarity of the mandibular arch in gnathostomes. Dlx genes are expressed in a dorsoventrally nested pattern in the pharyngeal arch ectomesenchyme (Beverdam et al., 2002; Depew et al., 2002; Minoux and Rijli, 2010; Gillis et al., 2013; reviewed by Takechi et al., 2013). Dlx5 and Dlx6 in the mouse are specifically expressed in the ventral half of the mandibular arch, and their simultaneous disruption leads to the transformation of lower jaw morphology into that of the upper jaw (maxillary process derivatives) (Depew et al., 2002). Gain of function of their upstream gene, Ednra, in the upper jaw domain, on the other hand, transforms the upper jaw morphology into that of the lower (Sato et al., 2008). Thus the dorsoventrally nested expression pattern of Dlx genes (i.e., the Dlx code) parallels morphological specification of the pharyngeal arch skeleton.

In the lamprey, the presence of the Dlx code is enigmatic. There are at least six Dlx genes (DlxA to -F) in the lamprey, five of which are specifically expressed in an ubiquitous fashion in the pharyngeal ectomesenchyme, including in the mandibular arch (Kuraku et al., 2010; also see Neidert et al., 2001). To date, there have been no reports from which we can infer dorsoventrally nested expression, although a dorsoventrally symmetrical nested pattern of expression around the gill pores has been suggested (Cerny et al., 2010). Moreover, LjBapxA, a homolog of Bapx1, the specifier of the jaw joint in gnathostomes (Miller et al., 2003), is not expressed in the mandibular arch of the lamprey (Cerny et al., 2010; Kuraku et al., 2010), suggesting that gnathostome-type basic topographic specification is absent in the lamprey. Although the expression pattern of dHand cognate, a ventral pole specifier, suggests the presence of dorsoventral polarity in the lamprey, the apparently unpolarized expression of Dlx genes in this taxon is consistent with the dorsoventrally symmetrical morphology of its posterior pharyngeal arch skeletons.

In the hagfish, too, dorsoventral polarity is not apparent from our preliminary analyses (Fig. 2A–G), although the morphologic pattern of the pharyngeal arch skeleton is less clear than that in the lamprey. This may be due partly to the secondary posterior shift of the posterior portion of the pharynx, which takes place in the late phase of organogenesis in the hagfish (Stockard, 1906; Oisi et al., 2013). The only conspicuous differentiation along the dorsoventral axis in the hagfish visceral skeleton is that of the lingual apparatus, which is derived from the ventral portion of the mandibular arch. As noted above, homology of this structure to that of the lamprey is well established at the morphologic level (Yalden, 1985). This skeletomuscular complex, however, does not appear to depend on any localized expression of Dlx genes; instead, it may develop through a different mechanism. Therefore, although this structural complex has often been homologized with Meckel's cartilage in gnathostomes, the developmental mechanisms in the two taxa would not be identical at the molecular level.

No homolog of the lower jaw elements therefore seems to exist in the cyclostomes. Upper jaw homology requires different consideration, because in the jawed vertebrates the upper jaw is specified by Dlx1 and Dlx2, which are expressed ubiquitously in the pharyngeal arch ectomesenchyme. Thus, the upper jaw in gnathostomes is likely to represent the default state of the Dlx code. Even if dorsoventrally nested expression is lacking in the cyclostomes, the default specification mechanism may have been acquired before the lower jaw specification program. This question, however, will require consideration of another candidate palatoquadrate homolog (“pq” in the hagfish; Holmgren, 1946).

Homology of the velum in lamprey and hagfish has been called into doubt by several authors (Goodrich, 1909; Strahan, 1958; Janvier, 1981, 1996). However, we have shown that this structure arises in a very similar pattern in both animals: the velum in both animals arises in the mid-part of the mandibular arch, between the oral ectoderm and rostral endodermal wall of the first pharyngeal pouch (von Kupffer, 1895; Oisi et al., 2013).

Comparison of chondrocrania between the hagfish and the lamprey: developmental architecture of the cyclostome cranium

The chondrocranium of the hagfish consists largely of the following elements: the nasal capsule cartilages, neurocranial base, otic capsule, lingual cartilages, other pharyngeal arch cartilages, and premandibular cartilages, including tentacle-supporting cartilages. One of the earliest skeletal elements to chondrify is the otic capsule (Fig. 2H–M), as is the case in the lamprey embryo (reviewed by De Beer, 1937). Quite interestingly, in the presumed Late Devonian stem lamprey (or stem cyclostome) Euphanerops, the otic capsule has also been suggested to be the first skeletal element to calcify in the developmental series (Janvier and Arsenault, 2007). This element in the hagfish, along with the trabecula and the dorsal longitudinal bar, is most likely to represent, possibly entirely, the mesodermally derived neurocranial elements.

In the more ventral part of the chondrocranium, we can establish homologies on the basis of the pancyclostome embryonic pattern (Oisi et al., 2013). We can compare components derived from the anterior nasal process, namely the dorsal wall posterior to the nostril in the lamprey head and the supranasal region in the hagfish. This domain in the hagfish contains nasal duct cartilages and many cartilaginous elements constituting the nasal capsule. In the lamprey chondrocranium, part of the nasal capsule will correspond to these components.

Components derived from the PHP will also contain cartilages derived from both the mandibular arch and the premandibular crest cells. To this category belong all of the tentacular cartilages (although there remains a problem as to the nature of the T4 cartilage that develops between the oronasohypophyseal septum and the lower mandibular arch; this T4 cartilage is similar to the cornual cartilage) and the subnasal cartilage of the hagfish chondrocranium, as well as the palatine bar, hypophyseal commissure, and possibly the rostral part of the dorsal longitudinal bar and the trabecula.

In the lamprey chondrocranium, the mucocartilage in the upper lip and possibly the rostral part of the trabeculae appear to arise from the equivalent anlage; in the hagfish, all of the above PHP-derived cartilage elements combined will be homologous with the upper lip (rostral dorsal plate and lateral wall of the upper lip) and trabecula, as well as a part of the nasal capsule. More precisely, the distribution of the trigeminal nerve branches suggests that the upper lip in the ammocoete will further be subdivided into a dorsal median portion (rostral dorsal plate) innervated predominantly by the ophthalmic nerve and a lateral wall innervated by the rostral branch of V2, 3 (Oisi et al., 2013). On the basis of the similarity of the innervation pattern in the hagfish to that in the lamprey, the lateral wall in the lamprey may correspond to the T1, T3 and T4 cartilages in the hagfish, whereas the T2 cartilage may be more similar to the dorsal roof. At any rate, the questionable cornual cartilage in the hagfish is located in such a way as to divide the cutaneous innervation area into a domain innervated by V1 and one innervated by the dorsal V2, 3 branch (Marinelli and Strenger, 1956; Oisi et al., 2013). On the basis of the above discussion, we summarized the homologous relationships between the lamprey and hagfish crania (Fig. 10, Table 1).

Fig. 10.

Homology of chondrocranial elements in cyclostomes. Hagfish and lamprey chondrocrania were compared on the basis of our results. Hagfish chondrocranium was redrawn from the work of Holmgren and Stensiö (1936), and those of the lamprey from the work of Marinelli and Strenger (1954) and Fontaine (1958).

Table 1.

Comparison of chondrocranial elements in cyclostomes.

There are a number of cyclostome-specific traits, namely differentiation of the lingual apparatus in the ventral mandibular arch region; differentiation of the velum in the mid-portion of the mandibular arch; absence of an intertrabecula homolog; absence of occipital vertebrae; close association between the otic capsule and parachordals; presence of lateral (external) pharyngeal arch skeletons; and presence of skeletal elements in the PHP derivatives. Some of these features will be counted as cyclostome synapomor- phies, but others will potentially represent plesiomorphies of vertebrates established by the common ancestor of cyclostomes and gnathostomes but secondarily lost in gnathostome lineages (see Oisi et al., 2013). To describe the evolutionary sequence of craniogenesis, it will be crucial to analyze the crania of gnathostome stems, the embryology of which remains unknown even in the best-studied cephalaspids (Osteostraci) and galeaspids (Galeaspida). Further collaborations spanning paleontology and evolutionary developmental biology, along with comparative embryology, will be needed to further our understanding of cranial evolution.