Peripheral nerve development was studied in the cat shark, Scyliorhinus torazame, using whole-mount and sectioned embryos. Nerve fibers were immunohistochemically stained using a monoclonal antibody against acetylated tubulin and, in early embryos, cephalic crest cells were observed by scanning electron microscopy. The initial distribution patterns of crest cells were identical to the typical vertebrate embryonic pattern, in that three crest cell populations were associated with even-numbered rhombomeres, prefiguring the pattern of the branchiomeric nerve roots. In older pharyngula, however, the trigeminal and postotic branchiomeric nerve roots were found to have shifted caudally along the neuraxis: the trigeminal nerve root finally arose from r3, while the glossopharyngeal nerve root arose from the presumptive region r7 of the hindbrain. The shift apparently takes place between the root fibers and the dorsolateral fasciculus. From observing the topographical relationships between the peripheral nerves and other epithelial structures (for example, the otocyst and the mesodermal head cavities - the anlagen of extrinsic ocular muscles), it was assumed that the shift was the result of an epigenetic effect caused by allometric growth of the otocyst, the mandibular cavity, and the spiracle epithelia anchoring the mandibular branch of the trigeminal nerve. It was concluded that the deviated morphological pattern of elasmobranch cranial nerves is a secondary phenomenon caused by the well-developed head cavities. In those animals whose head cavities are degenerated, the original pattern of the cranial nerve-rhombomere assignment, which is intact in lamprey embryos, is retained.
Among extant vertebrates, elasmobranchs are generally regarded as an outgroup to all other gnathostomes. Comprised of sharks and rays, the Elasmobranchs are characterized by unossified skeletons, odontoid scales in the skin, and a low grade of morphological differentiation of the pharyngeal arches (reviewed by Mallat, 1996). The eggs of oviparous species contain a large yolk and, by the pharyngula stage, the embryos attain a comparatively large size for vertebrates. For this reason, as well as their phylogenetic position, elasmo-branch embryos have been extensively studied by comparative embryologists. These studies established the basic concept of the segmental organization of the vertebrate head (Balfour, 1878; van Wijhe, 1882; Dohrn, 1888; Goodrich, 1918; de Beer, 1922, 1937; reviewed by Goodrich, 1930; Bjerring, 1971, 1977; Jarvik, 1980; and Jefferies, 1986).
A unique aspect of the elasmobranch embryo is the appearance of conspicuous head cavities, with transient epithelial mesoderm appearing in the head region (Balfour, 1878; van Wijhe, 1882; Scammon, 1911; Bjerring, 1971, 1977; reviewed by Jefferies, 1986; E. Gilland, personal communication). Similar cavities have been recognized in a number of vertebrate embryos, including those of amniotes, and they are believed to represent the primordia of extrinsic eye muscles (Brachet, 1935). Generally, the mesoderm in the head region develops three pairs of epithelialized pouches, the premandibular, mandibular, and hyoid cavities, or somites (Marshall, 1881; Lamb, 1902; Platt, 1891; Neal, 1918a,b; reviewed by Goodrich, 1930; Neal and Rand, 1942; de Beer, 1937; Starck, 1963; Jarvik, 1980; and Jefferies, 1986). The most rostral premandibular cavity is connected to its counterpart by a median bridge near the rostral tip of the notochord. In a few species, there is also an additional head cavity found more rostral than the premandibular cavity, known as Platt's vesicle (Platt, 1891). These three or four cavities were once thought to represent head mesomeres (mesodermal segments). Thus, the vertebrate head was assumed to originally consist of three or four preotic segments (reviewed by Goodrich, 1930; de Beer, 1937; Neal and Rand, 1942; Jarvik, 1980; and Jefferies, 1986). The presence of these segmental mesodermal components is a curious problem for modern biology, since the theory of head segmentation has gained support through molecular studies (reviewed by Kuratani et al., 1999). Thus, the development of the shark should be reevaluated with techniques that were unavailable to the classical embryologists.
Segmental patterning of the embryonic head is not restricted to mesodermal elements. In the embryos of several gnathostomes, as well as in lampreys, the first three bran-chiomeric nerves primarily arise from even-numbered rhombomeres as a result of the selective association of crest cells with those rhombomeres (Kuratani et al., 1997; Horigome et al., 1999 and references therein). Regardless of such features, which may be ancestral, late pharyngula elasmobranch embryos exhibit a peculiar feature in their cranial nerve patterning. The trigeminal and facial nerve roots arise very close to each other on the hindbrain, such that the trigeminal nerve root appears to be placed too posteriorly (typically observed in 13-mm Squalus acanthias; Neal, 1914; see below). In the present paper, we chose to study the cat shark (Scyliorhinus torazame), whose eggs are easiest to collect in Japan. Histo-logical and immunohistochemical methods were used to observe the staged development (Table 1) of the morphology of embryonic peripheral nerves to elucidate the apparently inconsistent cranial nerve configuration specific to this animal group. We also aimed to re-evaluate the morphology of elasmobranch embryos in the context of vertebrate head segmentation.
Staging list of S. torazame embryos.
MATERIALS AND METHODS
Embryos of S. torazame (a gift from Marine World Umino-Nakamichi, Fukuoka, Japan) were brought into the laboratory and kept in sea water at 15°C. Embryonic stages were determined according to body lengths, which were consistent with the post-oviposition age (Table 1). Embryos were prepared as previously described (Kuratani et al., 1997) with minor modifications. After fixation with 4% paraformaldehyde in 0.1% phosphate-buffered saline (PFA/PBS; containing 3% NaCl to adjust osmolarity) at 4°C for 1 day, embryos were washed in 0.9% NaCl in distilled water, dehydrated in a graded series of methanol solutions (50%, 80%, 100%) and stored at −20°C. The samples to be stained were placed on ice in 2 ml of dimethyl sulfoxide (DMSO)/methanol (1:1) until they sank. Then 0.5 ml of 10% Triton X-100/distilled water was added, and the embryos were incubated for an additional 30 min at room temperature. After washing in Tris-HCl-buffered saline (TST: 20 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.1% Triton X-100), the samples were sequentially blocked using aqueous 1% periodic acid and 5% non-fat dried milk in TST (TSTM).
A monoclonal antibody (mAb) raised against acetylated tubulin (monoclonal anti-acetylated tubulin, No. T-6793, Sigma Chemical Co., St. Louis, MO) was used to stain the cranial nerves. Embryos were incubated in the primary antibody (diluted 1/1,000 in spin-clarified TSTM containing 0.1% sodium azide) for 2 to 4 days at room temperature while being gently agitated on a shaking platform. Horse-radish peroxidase (HRP)-conjugated goat anti-mouse IgG (ZYMED Lab. Inc., San Francisco, CA) diluted 1/200 in TSTM served as the secondary antibody. After a final washing in TST, the embryos were pre-incubated with the peroxidase substrate, 3,3′-diaminobenzidine (DAB, 100 μg/ml), in TST for 1 h and then allowed to react for 20 to 40 min at 0°C in TST with the same concentration of DAB also containing 0.01% (v/v) hydrogen peroxide (35% aqueous solution). The reaction was stopped, and the embryos were simultaneously clarified, by addition of 30% glycerol in 0.5% KOH. The stained embryos were transferred to 60% glycerol/water for observation. Line drawings were prepared based on photographs. Norris and Hughes (1920), Tanaka (1976, 1979), Tanaka and Nakao (1979), Northcutt (1993, 1997) and Northcutt and Brändle (1995) were consulted for anatomical terms.
Immunostaining of sectioned specimens
Specimens, fixed in Bouin's fixative (containing 3% NaCl to adjust osmolarity), were embedded in paraffin and sectioned at 5 μm. The sections were then deparaffinized, treated with 1% periodic acid for 5 min at room temperature, and washed in TST. Anti-acetylated tubulin (diluted 1/1,000) and anti-HNK-1 mAbs (Leu-7, Becton Dickinson, San Jose, CA; diluted 1/50) were used to label the developing nervous system on the paraffin-sectioned specimens. Primary antibodies were diluted in TSTM and applied to the sections for 1.5 h at room temperature. After washing with TST, the respective secondary antibodies, HRP-anti mouse IgG and IgM (ZYMED Lab. Inc.), were diluted 1/200 in TSTM and applied to the specimens for 40 min. After the peroxidase reaction, the sections were counter-stained with either cresyl violet or hematoxylin.
Scanning Electron Microscopy
Embryos, fixed in 2.5% glutaraldehyde/0.01 M phosphate buffer (PB: pH 7.6), were post-fixed with 4% osmium tetroxide. The specimens were then rinsed in PB, where they were also skinned with a sharpened tungsten needle. They were then dehydrated in a graded series of ethanol solutions and freeze-dried in t-butyl alcohol. Finally, the embryos were placed on an aluminum stub, sputter-coated with gold-palladium alloy, and viewed by scanning electron microscopy (SEM; JSM-5800, JEOL).
anterodorsal lateral line placode
branchial crest cells
rostral subpopulation of branchial crest cells differentiating into the glossopharyngeal nerve anlage
branchial arches 1 to 5
dorsoventral diencephalic tract
preotic lateral line ganglion
hyoid crest cells
subpopulation of hyoid crest cells forming the acusticofacial nerve complex
trigeminal crest cells
medial longitudinal fasciculus
otic vesicle or otic pit
pectoral fin bud
posterior lateral line placode
pharyngeal pouches or slits 1 to 4
rhombomeres 1 to 6
dorsal branch of the first spinal nerve
hyomandibular branch of the facial nerve
mandibular branch of the maxillomandibular nerve
maxillary branch of the maxillomandibular nerve
superficial ophthalmic branch of the preotic lateral line nerve
posttrematic branch of the glossopharyngeal nerve
posttrematic branches of X1 and of X2
supraoptic branch of the trigeminal nerve
ventral rami of spinal nerves
spinal nerve 1
spinal dorsal root
tract of anterior commissure
trigeminal crest cells
trigeminal crest cell subpopulation representing trigeminal nerve anlage
tract of the habenular commissure
tract of the postoptic commissure
truncus arteriosus of the heart
mesencephalic trigeminal tract
trigeminal nerves or root
ophthalmic profundus nerve
acusticofacial nerve root
Stage I (19-day or 3.5-mm embryos)
The stage I S. torazame embryo appears similar to the stage F Scyllium embryo (Goodrich, 1918), corresponding to the early pharyngula. It exhibits two rostral pharyngeal pouches (Fig. 1A, B). The neural tube of the embryo contains a large optic vesicle and at this stage has formed a slight cephalic flexure at the level of the future midbrain. In the hindbrain, a pair of clear neuromeric boundaries can be seen. These represent the rostral and caudal boundaries of rhombomere 3 (r3).
Although at this stage, whole-mount immunostaining with the anti-acetylated tubulin mAb reveals no neurites, the anti-HNK-1 mAb specifically binds to an epithelial mesodermal pouch, located between the optic vesicle and the first pharyngeal pouch. This mesoderm will later become the mandibular head cavity (Fig. 1B). Consistent with earlier reports on elasmobranch embryos (Scammon, 1911; reviewed in Jefferies, 1986), histological sections reveal that this cavity arises as a typical enterocoel grown from the endodermal foregut. This antibody does not label neural crest cells.
Stage II (22-day or 3.5-mm embryos)
By stage II, overt cephalic crest cells can be recognized by SEM. Three putative crest cell populations are found in the embryo (Fig. 1C, D). These cells are superficially located in the head region and are easily distinguished from the more deeply embedded mesodermal cells. This implies a dorsolateral developmental pathway, characteristic of cephalic crest cells from embryos of other vertebrates. The overall morphology and crest cell distribution show a similarity to the stage G Scyllium embryo described by Goodrich (1918). In S. torazame at this stage, rhombomeric boundaries can be seen at the levels of r1/2, r2/3, r3/4, r4/5 and r5/6, but the mid-hindbrain boundary is not detectable (Fig. 1D). Each crest population is associated with a particular rhombomere. The most rostral and largest cell population, the trigeminal crest cells, covers the anterior part of the neural tube rostral to the r2/3 boundary. The second cell population, the hyoid crest cells, adheres to r4. The most caudal cell population, the branchial crest cells, has just begun emigration from the neural tube at the level of r5 or more caudally. No crest cells are seen on the surface of r3 (Fig. 1C, D). The trigeminal crest cells have already migrated into the mandibular arch and premandibular regions. It is not clear whether the absence of crest cells from the lateral aspect of the midbrain is an artefact of preparation, or whether it reflects the normal distribution pattern of crest cells in this species at this stage. However, in the 3.5-mm embryo of Squalus acanthias, a crest cell-free space in the same region divides trigeminal crest cells into two sub-populations: trigeminal crest cells posteriorly and thalamic crest cells anteriorly (Scammon, 1911). In either case, crest cells rapidly fill this space during subsequent stages (see below).
No acetylated tubulin immunoreactivity is observed in any neurites at this stage. Nonetheless, observation of the whole-mount embryo enables us to distinguish three compact streams of crest cells representing the initial primordia of three branchiomeric nerves (the trigeminal, acusticofacial and glossopharyngeal nerves; Fig. 1D). As seen in several other vertebrate embryos, each cranial nerve primordium is associated with an even-numbered rhombomere, with the exception of the vagal nerve primordium, which cannot be detected at this stage.
At the dorsal aspect of the mandibular cavity, the cell stream representing the trigeminal nerve anlage bifurcates into two substreams, representing the profundal and maxillomandibular divisions, respectively (Fig. 1D). Of the two divisions, the latter penetrates into a narrow space between the caudal surface of the mandibular cavity and the rostral aspect of the first pharyngeal pouch. Similar topographical relationships have been described in the stage K Galeus canis embryo by van Wijhe (1882) and in the 4.5-mm Squalus acanthias embryo by de Beer (1922).
Stages III and IV (8.0-to 9.5-mm embryos)
Cephalic crest cell populations exhibit the typical vertebrate pattern at stage III. This is similar to the 5-mm Torpedo embryo described by Froriep (1905) and the 8-mm Squalus acanthias embryo described by de Beer (1922). Trigeminal crest cells cover the entire area innervated by the trigeminal nerve, including the mandibular arch, the premandibular (infraoptic) regions, the cavum epiptericum, and the supraoptic region. Hyoid crest cells extend from r4 to the hyoid arch, and the postotic pharyngeal arches are entirely covered by branchial crest cells (Fig. 2A). Proximally, each crest cell population maintains a relationship with its respective rhombomere similar to that seen during the previous stage, except that the cells appear to have been excluded from r5, thereby leaving two crest cell-free spaces lateral to r3 and r5. In addition, when compared with stage II, there is a caudal shift in the relative position of the branchial crest cells associated with the growth of the otocyst (see below).
The epiphysis and telencephalon first begin to emerge in the forebrain at stage IV (Fig. 2B); this roughly corresponds to the 10-mm stage of Scyllium (Goodrich, 1918). At this point, whole-mount immunostaining of neurites becomes possible, and a significant change is observed in the position of the branchiomeric nerve primordia: the trigeminal nerve root is now found on the rostral portion of r3 (Fig. 2B). Distally, trigeminal nerve branches show a growth pattern analogous to that of the trigeminal crest cell streams at stage II (Fig. 1D). The profundal nerve passes dorsal to the aspect of the mandibular cavity, and the maxillomandibular branch passes ventrally between the caudal aspect of the mandibular cavity and the rostral wall of the spiracle (Fig. 2B). The mandibular cavity is located close to the epidermis, in the same plane as the trigeminal nerve ganglion. The other head cavities, the premandibular and hyoid, are observed to be located more medially. In S. torazame, no additional head cavities develop rostral to the premandibular cavity (Fig. 2B).
As compared with stage III, the positions of the postotic branchiomeric nerve roots (IX, X) have shifted caudally on the rhombencephalon in stage IV (Fig. 2B). Because the glossopharyngeal nerve root is now located caudal to r6, this shift does not appear to result simply from the local growth of the hindbrain at the level of the otocyst.
The stage IV embryo also exhibits extensive neurite growth along the body. Numerous neurites are observed growing in an irregular pattern in the embryonic skin (Fig. 2B). These neurites belong to the Rohon-Beard cell class and have no morphological relationship to, or connection with, spinal nerves peripherally (see below). Moreover, in each branchiomeric nerve, the post-trematic branch seems to be the first to extend neurites (Fig. 2B).
Stage V (16-to 17-mm embryos)
By stage V, external gills have begun to develop on the caudal edges of hyoid and more caudal pharyngeal arches. The basic anatomical configuration of vertebrate peripheral nerves is emerging. Thus, the trigeminal nerve has grown neurites into the maxillary and mandibular regions, and the superficial ophthalmic branch grows from the preotic lateral line ganglion (Fig. 3A). Histological examination reveals that the contributions to the formation of the sensory ganglia made by the epibranchial placodes are similar to those in Torpedo embryos of the same body length (Froriep, 1891). A few epithelial placodal cells exhibit HNK-1-immunoreactivity, but acetylated tubulin immunoreactivity is associated exclusively with the cells within the ganglion (not shown). The trigeminal nerve root is found on r3, and the acusticofacial nerve root is on r4.
Graphic reconstruction of the head cavities from histological sections reveals typical topographical relationships with surrounding structures. The premandibular cavity is connected to its counterpart medially, between the hypothalamus and the hypophysis at the point where the rostral tip of the notochord attaches. Of the three cavities, the mandibular is located most laterally, between the ophthalmicus profundus and maxillomandibular branches of the trigeminal nerve; and the hyoid cavity is located medial to both the trigeminal and acusticofacial nerve roots (Fig. 3B, C). Moreover, the premandibular and hyoid cavities have connections with the oculomotor and abducens nerves, respectively (not shown). The trochlear nerve has not yet developed at this stage (Fig. 3A).
The overall anatomical configuration of the stage V S. torazame embryo is very similar to the 6.5-mm Squalus acanthias embryo described by Wedin (1955), the 12-mm Acanthias embryo described by Lamb (1902), and the stage K Scyllium canicula embryo described by Goodrich (1930). Thus, this pattern is apparently conserved among elasmo-branch pharyngula.
Stage VI (53-day or 20-to 22-mm embryo)
The trochlear and olfactory nerves, as well as the posterior lateral line placode are visible by stage VI (Fig. 4A). The preotic lateral line ganglion has grown into a bipartite configuration: the buccal branch now grows from the ventral portion of the ganglion (buccal ganglion) into the maxillary region. In the brain, the appearance of nerve tract has made regionalization apparent. In addition to the mesencephalic trigeminal tract and the medial longitudinal and dorsolateral fasciculi, which have already been seen at stage IV (Fig. 2B), the tracts of the anterior, postoptic and habenular commissures, the dorsoventral diencephalic tract, the supraoptic tracts and the posterior commissure are all now present (Fig. 5). At this stage, the nerve tract morphology is comparable to that seen in late pharyngula of several other vertebrates including lampreys (Kuratani et al., 1998), teleosts (Chitnis et al., 1992), amphibia (Hartenstein, 1993; Burril and Easter, 1994), birds (Windle and Austin, 1936; Figdor and Stern, 1993), and mammals (Windle, 1970; Easter et al., 1993; reviewed by Kupffer, 1906).
As in the previous stage, the trigeminal and acusticofacial nerve roots are found on r3 and r4, respectively (Fig. 5A). These nerve roots are now located on the dorsolateral fasciculus, and direct contact between the crest cells and neurectoderm have long been lost (Fig. 6D, E; also see Fig. 2B). Epithelial head cavities retain the same topographical relationships as in the previous stage (Fig. 4B; compare with Fig. 3). Presumably, from the contribution made by the dorsolateral placodes to the preotic lateral line, the acusticofacial nerve complex is located far lateral to the trigeminal nerve. The topographical association between the head cavity and the cranial nerve is much closer between the trigeminal nerve and the mandibular cavity than between the acusticofacial nerve and the hyoid cavity, which are medio-laterally separated from each other (Fig. 4A, B).
Rohon-Beard neurites are still present in stage VI embryos (Fig. 6A, B, C). In the whole-mount stained embryos, these neurites are found to issue close to the midline of the neural tube (Fig. 6C). In most cases, HNK-1-positive, Rohon-Beard cell bodies are found within the neural tube (Fig. 6B), but in a few cases, cell bodies can also be found extra medullarly (Fig. 6C). Interestingly, the Rohon-Beard neurites do not supply the skin of the pectoral fin buds, which first become apparent at this stage (Fig. 6A). These neurites also invade the wall of the umbilical cord, a region never innervated by spinal nerve branches (Fig. 6A; also see Fig. 3A).
As a consequence of the macrolecithal nature of the development of embryos with umbilical cords, the shark pharyngula looks more like the pharyngula of an amniote embryo than that of a teleost or a lamprey embryo at a similar stage (Keibel, 1906). For example, the topographical relationship between the profundal nerve, the oculomotor nerve and the optic vesicle in the 9-mm S. torazame embryo is similar to that seen in the late pharyngula of amniotes (Keibel, 1906; Halley, 1955; Kuratani and Tanaka, 1990; Kuratani, 1997; Easter et al., 1993). One might expect, therefore, that the morphotype of the elasmobranch embryo, formulated through vertebrate comparative embryology, is more directly representative than other embryos with morphotypes formulated on the basis of the anatomical features of the adult peripheral nerves (Tanaka, 1976, 1979; Tanaka and Nakao, 1979). However, a major difference is evident in the preotic region of the head, namely, the positions of the trigeminal and acusticofacial nerve roots, which appear to be associated with embryonic features unique to elasmobranchs. Importantly, the early pharyngula of S. torazame exhibits three cephalic crest cell populations extending from even-numbered rhombomeres to the superficial portion of the embryo. Thus, elasmobranch embryos exhibit the same head-specific, dorsolateral, crest cell migration pathways and the same conserved relationships with rhombomeres as are exemplified by amniote embryos (reviewed by Le Douarin et al., 1984; and by Bronner-Fraser, 1995).
The branchiomeric nerves of vertebrates (e.g., chick: Tello, 1923; Lumsden and Keynes, 1989; Kuratani, 1991; Lumsden et al., 1991; Niederländer and Lumsden, 1996; Lampetra: Kuratani et al., 1998; Horigome et al., 1999; reviewed by Kuhlenbeck, 1935; and by Gilland and Baker, 1993) characteristically develop on even-numbered rhombomeres. This is a consequence of the selective association of crest cells with those rhombomeres. Trigeminal crest cells adhere to r2, hyoid crest cells to r4, and branchial crest cells to r6 and further posterior. In amniote embryos, cephalic crest cells are known to be generated at all levels of the neuraxis, but are then secondarily excluded from the r3 and r5 levels (Serbedzija et al., 1992; Sechrist et al., 1993; Graham et al., 1993). Thus, the typical tri-population pattern of cephalic crest cells appears to be established in various vertebrate species including lampreys (Horigome et al., 1999), elasmobranchs (Froriep, 1891, 1902; Goodrich, 1918; de Beer, 1922), teleosts (Sadaghiani and Vielkind, 1990), amphibians (Froriep, 1917; Stone, 1926; Starck, 1963; Jacobson and Meier, 1984; Zackson and Steinberg, 1986; Hall and Hörstadius, 1988; Mayor et al., 1995; Northcutt and Brändle, 1995; Olsson and Hanken, 1996), reptiles (Meier and Packard, 1984), birds (Anderson and Meier, 1981; Kuratani and Kirby, 1991; Sechrist et al., 1993) and mammals (Halley, 1955; Müller and OíRahilly, 1980; Tan and Morriss-Kay, 1985; Maden et al., 1992; Osumi-Yamashita et al., 1994). Similar exclusion of crest cells from r5 is also apparent in S. torazame (Figs. 1C, 2A). The transient division of trigeminal crest cells into thalamic and trigeminal sub-populations (Scammon, 1911; and see Fig. 1C) raises the possibility that originally crest cells formed four cell populations rather than three. Trigeminal crest cell populations in lamprey embryos do not show such division, however (Horigome et al., 1999). Presumably, therefore, this feature is transient and may be specifically associated with elasmobranch embryos.
Segmental rhombomere-crest cell relationships serve as a critical developmental pre-pattern in branchiomeric nerve patterning. For example, in the chick embryo, heterotopic transplantation results in a re-patterning of the nerve roots that stands in contrast to spinal nerve metamerism, which primarily depends on somitic metamerism (Detwiler, 1934; Keynes and Stern, 1984; Rickmann et al., 1985; Tosney, 1987; Bronner-Fraser and Stern, 1991; Kuratani and Eichele, 1993). Moreover, adhesion of the trigeminal ganglionic primordium to the hindbrain is required for normal neuronal patterning and differentiation of the trigeminal nerve in the chick (Moody and Heaton, 1983a, b, c).
Epigenetic translocation of the cranial nerve roots
The present study confirms that, during development, the shark trigeminal nerve root shifts along the hindbrain neuraxis from r2 to r3. Similar aberrant morphologies have been described in several other elasmobranch embryos, in which the trigeminal nerve root was either found on r3 or was found too close to the root of the acusticofacial nerve, for example the 13-mm Squalus acanthias embryo (Neal, 1914) and, perhaps, the 12-mm Torpedo ocellata embryo (Froriep, 1905). However, as seen in various other vertebrates, in early pharyngula embryos of elasmobranch species, including S. torazame, the proximal portion of the trigeminal crest cell population (the trigeminal nerve primordium) attaches at a level corresponding to r2 or more anteriorly (Torpedo ocellata: Froriep, 1905; Scyllium: Goodrich, 1918; Squalus acanthias: Scammon, 1911; de Beer, 1922). This means that the shift in the position of the nerve roots along the brain stem is a secondary phenomenon that deviates from the vertebrate morphotype. This modification of the rhombomere-cranial nerve root relationship is illustrated in Fig. 7. The transposition of the trigeminal nerve root takes place between the 3.5-mm and the 9.0-mm stages in S. torazame. Histological observations reveal that the translocation takes place along the dorsolateral fasciculus, implying that the adhesion between the crest cell population and the hindbrain neurectoderm did not serve as a pre-pattern for the later anatomical configuration.
One possible explanation for the peculiar behavior of the trigeminal nerve root in elasmobranch embryos is that it is epigenetic in nature and may involve the unique feature of the elasmobranch pharyngula, namely, the head cavities. Among vertebrate groups, elasmobranch embryos develop enormously expanded head cavities during the pharyngula stages (Figs. 2, 3). Although epithelial head cavities are also observed in many gnathostome embryos, only in elasmobranchs do a complete set of cavities arise during development (see below). Important topographical relationships between the PNS and the head cavities are seen between the trigeminal nerve and the mandibular cavity. By the stage at which the trigeminal nerve anlage-the characteristic distribution pattern of the trigeminal crest cell population-is first visible, the primitive nerve root is found on r2 of the hindbrain and, distally, the cell streams are bifurcated dorsal to the expanding mandibular cavity (Fig. 1D). Of the two cell streams, the caudal stream corresponds to the site of mandibular branch development and is confined to a narrow space pressed between the caudal wall of the mandibular cavity and rostral wall of the first pharyngeal pouch. It is easy to imagine that the trigeminal nerve complex is continuously pulled caudally with respect to this fixed point as the proportions of the components of the head change, especially those changes resulting from caudal growth of the hypothalamus (Fig. 8). Also involved in this phenomenon would be the absence of a focal dorsolateral placode associated with the trigeminal nerve at these stages. In the facial nerve region, the preotic, antero-dorsal and anteroventral dorsolateral placodes all contribute to branches associated with the acusticofacial nerve complex.
It is important to note that the morphological topography of the head cavities differ: the mandibular cavity is located superficially, and the other two develop medially (Goodrich, 1930). With dorso-laterally migrating cephalic crest cells paving the way, the branchiomeric nerve anlagen develop in the lateral part of the embryonic body. A close relationship, therefore, is established only between the trigeminal nerve and the mandibular cavity, which also explains the relatively stable position of the acusticofacial nerve root lying far lateral to the hyoid cavity.
A similar embryonic translocation of the nerve roots is associated with the glossopharyngeal and vagus nerves: both nerves appear to shift caudally during development, perhaps as a result of being pushed in that direction by the enlarging otocyst (Figs. 7, 8). As already noted, this phenomenon is not due to the local growth of the hindbrain neurepithelium.
In conclusion, the early elasmobranch pharyngula with its odd/even pattern of rhombomeres and cranial nerve roots initially exhibits the characteristic vertebrate phylotype, but this pattern is secondarily lost in late pharyngula states. Presumably, the caudal translocation of the trigeminal nerve root may be of a secondary nature brought about by the presence of head cavities: a unique feature of elasmobranchs.
Evolutionary history of head cavities: a hypothetical scenario
According to the classical view of comparative embryology, selachian head cavities are equivalent to somites or myotomes in the trunk (reviewed by Goodrich, 1930; and by de Beer, 1937). It is indeed true that myogenesis from an epithelial state is reminiscent of the development of the myotome in the trunk. This has served as the basis for theories aimed at explaining the segmental organization of the vertebrate head (Bertmar, 1959; Jollie, 1977; Bjerring, 1977; reviewed by Goodrich, 1930; Jarvik, 1980; Jefferies, 1986; and by Gilland and Baker, 1993). Among known vertebrate groups, however, complete sets of clear and large epithelial cavities are restricted to elasmobranchs: pharyngula state lampreys do not possess mesodermal epithelial cavities (Damas, 1944; Horigome et al., 1999). Consequently, we are left with the question of whether these structures are primitive or acquired characteristics among vertebrates.
The occasional presence of vestigial head cavities in bony fishes and amniotes (Fraser, 1915; Wedin, 1949, 1952, 1953; Jacob et al., 1984; personal communication with Edwin Gilland; reviewed by Brachet, 1935) might infer that possession of these cavities is a shared, derived characteristic of gnathostomes that is secondarily diminished in the majority of members (Fig. 9). At the same time, the typical even/odd pattern of rhombomeres has been clearly established by the divergence of lampreys and gnathostomes (Horigome et al., 1999; Kuratani et al., 1999). The peculiar branchial nerve root morphology of elasmobranchs would thus represent a unique and derived character state caused by mandibular cavities remaining as epithelial structures of appreciable size for a long period during development.
One possible scenario that appears consistent with the above concept is that elasmobranchs maintain their head cavities because they are one of the earliest direct developers among vertebrates (with no larval stage), and thus need to differentiate extrinsic eye muscles extremely rapidly. Although no data are available on the development of the direct ancestors of elasmobranchs, extrinsic eye muscles do not appear at pharyngula stages in outgroups of gnathostomes. Hagfishes, a sister group of vertebrates, do not possess these eye muscles, nor do they seem to develop head cavities (Dean, 1899). Lampreys, another sister group of gnathostomes, on the other hand, possess a larval state that does not require a visual system or the rapid development of eye muscles. It has been speculated that lamprey extrinsic ocular muscles might originate in rostral myotomes (Neal, 1914, 1918; Damas, 1944; Bjerring, 1977; also see Hardisty and Rovainen, 1971). However, these myotomes are, in fact, postotic elements that have nothing to do with the preotic components of the head (Kuratani et al., 1997, 1999). The origin of the extrinsic ocular muscles in the lamprey is still unknown (Hardisty and Rovainen, 1971), although they do appear to arise before metamorphosis (unpublished personal observation in Lampetra japonica; see also Koltzoff, 1901).
As observed in sea urchins, which develop directly (without larval stages), acceleration of adult organogenesis involves rearrangement of cell lineages and embryonic primordia (reviewed by Wray and Raff, 1991; and by Wray, 1995). Elasmobranch head cavities may thus be embryonic structures primarily adaptive for accelerated eye muscle development and not mere vestigial metameric units of paraxial mesoderm. Similar concepts were presented by Adelmann (1926) and Wedin (1955), who assumed that the head cavities were an adaptation to accelerated head development in the shark. If the elasmobranch head cavities are adaptive anlagen for accelerated head development, and the apparent mesodermal segmentation of the head is merely an epigenetic secondary phenomenon (Starck, 1963; Kuratani et al., 1999), then cranial nerve morphology may have initially diverged from the pre-existing developmental plan of vertebrates, only to merge again later in those groups in which the cavities are diminished.
We are grateful to Hisoka Hiruta and Koji Takada in Marine World Umino-Nakamichi, Fukuoka, for providing embryonic materials. We also thank Edwin Gilland for communicating his unpublished observations and valuable discussions. Sincere gratitude is extended to Shigenori Tanaka for his valuable comments on the manuscript. This work was supported by grants-in-aid from the Ministry of Education, Science and Culture of Japan.