Modified from: Kimmel et al., 1995. Developmental Dynamics 203:253-310. Copyright © 1995 Wiley-Liss, Inc. Reprinted only by permission of Wiley-Liss, a subsidiary of John Wiley & Sons, Inc.

A wonderful variety of morphogenetic movements now occur, the somites develop, the rudiments of the primary organs become visible, the tail bud becomes more prominent and the embryo elongates (Fig. 15). The AP and DV axes are unambiguous. The first cells differentiate morphologically, and the first body movements appear.

We might have called this period the "tail bud period" because the tail bud appears at the caudal end of the lengthening axis during the whole time. Indeed, because tail morphogenesis is so prominent during this period of development, simply noting the overall shape of the embryo is a surprisingly effective way to determine its stage. Furthermore, as the tail extends, the overall body length of the embryo very rapidly increases, reasonably linearly, such that length is a fairly useful staging index after about 15 h (Fig. 16).

Embryo length (EL), at any stage, refers to the embryo's longest linear dimension. This is important to understand, for reference to Fig. 1 and Fig. 15 will show that the tail does not extend directly opposite to the head. Rather, the extending axis (on the dorsal side) curls around the original vegetal pole, incorporating cells that were ventrally located in the gastrula, and eventually the tail bud grows towards the head. The direction reverses during the latter part of the segmentation period, for as the tail lengthens it rapidly straightens (Fig. 15L-N). Straightening of the head occurs later, for the most part during the next developmental period, the pharyngula period.

Somites appear sequentially in the trunk and tail, and provide the most useful staging index (Fig. 17). Anterior somites develop first and posterior ones last. The earliest few somites seem to form more rapidly than later ones, and we approximate the rate at 3 per hour for the first six, and 2 per hour thereafter at our standard temperature (Hanneman & Westerfield, 1989). Figure 18 shows the curve, but if one keep in mind that 18 somites are present at 18 h and two form per hour over most of the period, then the figure is unneccessary. To accurately determine somite number use Nomarski optics, and include in the total the last (posterior-most) somite that appears to be fully segmented (see Fig. 17). The earliest furrow to form is the one posterior to the first somite, not anterior to it as as one might think. The anterior margin of the first somite becomes visible not long afterwards, roughly coincidentally with the formation of somite two.

Shortly after each somite forms, its surface appears epithelial. The epithelium surrounds a more mesenchymal region (e.g, the third somite in Fig. 17A, the last two somites in Fig. 17D). The large majority of the interior cells in each somite will develop as a myotome (sometimes termed a myomere), or muscle segment (Fig. 17C; see also below, Fig. 34B). Myotomes retain the metameric arrangement of the somites, adjacent myotomes becoming definitively separated by a transverse myoseptum consisting of connective tissue.

In some vertebrates, including some teleosts, the earliest somites seem to be transient structures. However, a cell marked with lineage tracer dye in an early zebrafish somite (e.g. at the 3-somite stage) produces a muscle clone in the corresponding myotome later (Fig. 19). Thus there are no transient somites in the zebrafish; the first somite forms the first definitive myotome and so on.

The earliest cells to elongate into muscle fibers appear to derive from a part of the medial somitic epithelium, the "adaxial" region (Thisse et al., 1993) adjacent to the developing notochord, and in the middle, dorsoventrally, of each somite. The precocious subset of adaxial cells may include a special cell class, the muscle pioneers (Felsenfeld et al, 1991), recognizeable by enhanced expression of specific markers (Hatta et al., 1991a). The muscle pioneers are so named because of their early development, and the name does not imply they play any necessary morphogenetic role such as shown for muscle pioneers in insects (Ho et al., 1984), which have many different features from the muscle pioneers in zebrafish. As the muscle pioneers elongate, the myotome takes on a chevron shape. The V points anteriorly, with the muscle pioneers present at its apex (Fig. 17C). Later a horizontal myoseptum develops at the same position.

A second derivative of the somite is the sclerotome, giving rise to vertebral cartilage. In a young somite, the sclerotome cells develop ventromedially in the somitic epithelium. Late in the segmentation period the sclerotome cells delaminate, take on a mesenchymal appearance, and migrate dorsally, along a pathway between the myotome and notochord (personal communication, B. Morin-Kensicki).

We do not know if a dermatome also develops from somitic cells.

Pronephric kidneys develop bilaterally deep to the third somite pair. They are difficult to visualize in the live preparation until later in development when they enlarge and hollow out, but their ducts, the pronephric ducts, are easy to see (Fig. 20). Each pronephric duct primordium, at first without a lumen, grows posteriorly during the early part of the period, and ventrally around the posterior end of the yolk sac extension (see below), to reach the ventral midline site near the future anus where the bilateral pair of ducts will join together. The lumen appears, and eventually the ducts open to the outside.

The notochord differentiates, also in an AP sequence. Some of its cells vacuolate and swell to become the structural elements of this organ (Fig. 21, and compare with Fig. 34C below), and others later form a notochord sheath, an epithelial monolayer that surrounds the organ.

The endoderm becomes morphologically distinctive at about the onset of the segmentation period, as revealed in sections through the trunk (personal communication from R. Warga). At first, only a few rather disorganized-looking cells can be identified as being endodermal by their positions; they contact the YSL at or near the midline on the dorsal side of the embryo deep to the main part of the hypoblast. During this period more endodermal cells appear, and form a more compact and orderly-looking epithelial arrangement. Endoderm may develop on only the dorsal side of the embryo, beneath the axial and paraxial mesoderm.

The epiblast, now exclusively ectodermal, undergoes extensive morphogenesis during the segmentation period. As gastrulation ends, the primordium of the central nervous system, the neural plate, is already fairly well delineated, because of its prominent thickness. Indeed, expression of putative patterning genes, e.g. krox20 (Oxtoby and Jowett, 1993) and pax2 (Krauss et al., 1991) show that the neural plate is already regionalized as the segmentation period begins. The anterior region where the brain will form is particularly thick (Fig. 15B and C). Whereas the epiblast of the early and midgastrula appears several cells thick (Fig. 13), labeling studies (Papan and Campos-Ortega, 1994) reveal that by the late gastrula period the epiblast becomes a monolayer, a pseudostratified epithelium. The thickness of the neural plate is due to the fact that its cells take on a columnar shape, whereas the more lateral and ventral ectodermal cells, in the epidermal primordium, remain cuboidal (diagrammed in Fig. 22A). Formation of the neural tube then occurs by a process known as "secondary neurulation", as recently studied in the zebrafish by Papan and Campos-Ortega (1994). Seondary neurulation contrasts with "primary" neurulation, the textbook version of neurulation in vertebrates where a hollow tube forms from the neural plate by an uplifting and meeting together of neural folds. In teleosts the lumen of the neural tube, the neurocoele, forms 'secondarily', by a late process of cavitation. An intermediate and transient condensed primordium with no lumen, the neural keel, forms first (Fig. 15E). The neural keel rounds into a cylindrical, still solid neural rod (Fig. 15K, 21A), and only afterwards hollows into the neural tube (Fig. 15 and Fig. 21B).

More covertly, primary and secondary neurulation are similar to one another (Papan and Campos-Ortega, 1994). The neural plate transforms topologically into the neural tube in fish in the same way it does in amphibians, as revealed by fate mapping in salmonids (Ballard, 1973) and zebrafish (Kimmel et al., 1990b), where it has now been studied in detail (Papan and Campos-Ortega, 1994). The medial part of the neural plate (originally from dorsal epiblast in the gastrula) forms ventral structures in the neural tube, and the lateral part of the plate (from lateral and ventral gastrula gastrula epiblast) forms dorsal tube. The similar transformation occurring in primary and secondary neurulation is due to what seems to be fundamentally similar morphogenesis; the neural keel develops by a process of epithelial infolding at the midline, equivalent to the folding of the plate in primary neurulation of amphibians (Papan and Campos-Ortega, 1994). The cells remain epithelial throughout neurulation in both types. They converge towards the midline, and then, in the zebrafish, they tip obliquely on both sides of the midline in a manner that indicates infolding (Fig. 22B), paralleling the neural folds in primary neurulation. The original outer surfaces, the apical ones, of the epithelial cells from the left and right sides of the neural plate then meet to form the midline of the developing neural keel, and subsequently the midline of the neural rod (Fig. 22C). It is then that these apical surfaces, rather late during the segmentation period, pull away from one another to form the neurocoele. This process begins at the floor plate (Fig. 21), a distinctive row of cells occupying the neural primordium's ventral midline (Hatta et al., 1991b).

Because the times of neurulation and segmentation overlap so extensively the zebrafish does not have a distinct "neurula period" of development, such as occurs largely before segmentation in amphibian embryos. Another prominent difference between zebrafish and Xenopus, revealed by labeling studies, is that in the amphibian one side of the neural plate forms the corresponding lateral wall of the neural tube, whereas in the zebrafish cell divisions occurring in the midline during keel formation (typically division 16; Kimmel et al., 1994) often distribute sister cells originating from one side of the neural plate to both sides of the neural tube. We presently do not understand the significance of this redistribution.

Even before cavitation, the anterior part of the neural keel that will form the brain undergoes regional morphogenesis. At the beginning of the segmentation period, the brain rudiment, although larger than the spinal cord rudiment, appears uniform along its length (Fig. 23A). Then, during the first half of the segmentation period, about ten distinctive swellings, termed neuromeres, form along it (Fig. 23B). The first three are large, and become prominently sculptured. They correspond to the two forebrain subdivisions, the diencephalon and telencephalon, and the midbrain or mesencephalon. Additionally, the rudiments of the eyes, the optic primordia, develop very early from the lateral walls of the diencephalon (Fig. 15D and Fig. 24), eventually giving a very lovely arrowhead-like shape to the brain rudiment in dorsal view (Fig. 15K). During the last part of the segmentation period the ventral diencephalon expands as the primordium of the hypothalamus, and the primordium of the epiphysis appears as a small, well-delineated swelling in the midline of the diencephalic roof (Fig. 23C). The midbrain primordium subdivides horizontally to form the dorsal midbrain (optic) tectum and the ventral midbrain tegmentum (see below, Fig. 30).

The remaining seven neuromeres, termed rhombomeres (r1-r7), subdivide the hindbrain (Fig. 23B & C and Fig. 25). The primordium of the cerebellum forms near the end of the period as a very prominent dorsal domain in the region of the hindbrain-midbrain boundary.

Associated with the neural primordium are changes in its periphery. Neural crest, its early position indicated by expression of the homeobox gene dlx3 (Akimenko et al., 1994), delaminates and migrates from the dorsolateral wall of both the brain (Fig. 26) and spinal cord primordia, beginning at neural keel stage, and continuing well after the tube has appeared. Just adjacent to the keel, thickened ectodermal placodes appear bilaterally at specific locations, also predicted by dlx3 expression (Akimenko et al., 1994) within the head region, and which later form sensory tissues (Fig. 26 and Fig. 27). They include the olfactory placode overlying the telencephalon, the lens placode, associated with the optic primordium, and the otic (ear) placode overlying rhombomere 5 of the hindbrain (Fig. 25 and Fig. 27).

Following closely upon the early morphogenesis of the neural tube, the first neurons begin to differentiate, and in comparison with what will come later the pattern of cells and axonal pathways is not overwhelmingly complicated. So far as we know all of the early developing cells, termed "primary neurons" (Grunwald et al., 1988), develop large cell bodies, and extend long axons that project between different regions of the neural tube (e.g. between the hindbrain and spinal cord), or, in the cases of sensory neurons and motoneurons, between the neural tube and the periphery. Early sensory neurons, that will mediate tactile sensibility, are located bilaterally both in a discontinous dorsal column in the spinal cord (Rohon-Beard cells) and in a ganglion (trigeminal) in the head. Beginning at about 16 h the peripheral sensory axons extend to blanket the skin and the central axons from the same neurons pioneer a single long tract located dorsolaterally along the wall of the neural tube (Metcalfe et al., 1990). Motoneurons differentiate in the ventral spinal cord only slightly later (Eisen et al., 1986; Myers et al., 1986). A marvelous feature of their development is that as individual motor axons grow out to meet the developing myotomal muscle cells, the axons elicit contractions in the muscle, presumably by transmitter release from the growth cones (Grunwald et al., 1988). Hence at this time an observer can very easily appreciate just where these neurons are in their development. The primary motoneurons have been particularly well-studied. There are three, sometimes four (Eisen et al., 1990) of them in each spinal hemisegment, and their axons grow without error along precise pathways to innervate restricted domains within the overlying myotome (Myers et al., 1986). By carrying out meticulously timed transplantions of single postmitotic cells to new locations in the ventral spinal cord, Eisen (1991) has shown that position of the motoneuronal cell body plays a crucial role in specification of its identity, which becomes fixed 1-2 hours before axonogenesis.

The centrally-located interneurons form a relatively simple pattern at first, one that is clearly related to the neuromeric organization of the neural tube: Bilateral pairs of single interneurons or small clusters of them begin to differentiate at the centers of each of the brain neuromeres indicated in Fig. 23B (reviewed in Kimmel, 1993), and in each segment of the spinal cord (Bernhardt et al., 1990). Primary interneurons in both the spinal cord and hindbrain are individually dentifiable. The timing of interneuronal development is generally not known as precisely as for the sensory neurons and motoneurons, but the earliest central axons appear at nearly the same stage as the peripheral ones. These central axons form a simple scaffold, organized orthogonally (Wilson et al., 1990; Chitnis and Kuwada, 1990; Metcalfe et al., 1990): A primary bilateral pair of ventral longitudinal tracts extend through the length of the brain and spinal cord, with the two sides connected by commissural axons crossing the midline. Each set of interneurons, or in the hindbrain and spinal cord each individual interneuron, pioneers a part of this system or joins to it in a way that is highly stereotyped (e.g. Wilson and Easter, 1991), apparently matching the precision of motor axonal pathfinding in the periphery. Notably, when and where the cells develop and where their axons extend correlate closely with the spatial domains and temporal patterns of expression of regulatory genes, including genes of the pax, krox, eph, forkhead and wnt families (Macdonald et al., 1994; Krauss et al., 1991; Oxtoby and Jowett, 1993).

Function follows upon morphological development of the neuronal systems, and the early pattern of synaptic interconnections between the neurons seems also to be very precise. In a particularly clear example, an identified dendrite of a specific interneuron, the Mauthner neuron located in the hindbrain's fourth rhombomere, begins to extend late in the segmentation period (18 h), nearly simultaneously with the arrival to it of a set of axonal growth cones coming from trigeminal sensory neurons. As the dendrite begins its outgrowth it tightly envelops the small bundle of growing sensory axons, and at that location specifically, just at the base of a dendrite which eventally becomes quite long, synapses are made (Kimmel et al., 1990a). This connection appears to be a key one in a reflexive relay between primary sensory neurons in the head and primary motoneurons in the spinal cord, to which the axon of the Mauthner cell projects. The behavioral reflex, a motor response to a light touch to the head, becomes apparent a few hours later.

Ventrolaterally to the brain primordium, and largely posteriorly to the eye, a primordium appears that will form the series of pharyngeal or visceral arches. Including an inner epithelial lining of pharyngeal endoderm, the pharyngeal arches derive from all three germ layers, head neural crest contributing prominently to the arch mesenchyme (Schilling and Kimmel 1994). Both morphogenesis and differentiation of the pharyngeal walls occurs most dramatically later, during the hatching period.

The pharyngeal arches and rhombomeres are visible components of the head segments, as the somites are compontents of the trunk and tail segments. Early head mesoderm might also be segmentally organized, but we have no evidence for it. Head mesodermal condensations and vesicles, such as described classically as "preotic somites" from sections of embryos of sharks and other kinds of fish, have not been found in teleost embryos, as confirmed in a recent reinvestigation (Horder et al., 1993). On the other hand, head mesodermal "somitomeres" were characterized in an SEM study of the medaka, Oryzias (Martindale et al., 1987), a teleost fish distantly related to the zebrafish. Head mesodermal segments, if present in early zebrafish embryos, are not readily visible in live material.

Detailed description of Segmentation stages.