Formation of the notochord in living ascidian embryos

Several cell behaviours considered to be important in forming embryonic structures are displayed during formation of the notochord. Cell proliferation, growth, migration, rearrangement, shape changes, and alteration of the extracellular environment occur as the notochord is transformed from a loose plate or mass of cells into an elongated rod surrounded by a sheath of fibrous material. Notochord formation is thus an ideal event to examine the roles that these behaviours play in how this and other embryonic structures are formed and how the embryo is shaped.

There have been a number of descriptive studies on notochord morphogenesis in a variety of different chordates (Mookerjee, Deuchar & Waddington, 1952; Lesson & Lesson, 1958; Bancroft & Bellairs, 1970; Ruggeri, 1972; Malacinski & Youn, 1982). Notochord formation has not been examined directly in these organisms, however, because of the size and opacity of their embryos. Studies of internal cell behaviour in intact, living embryos are possible in a number of species of simple ascidians since they are relatively transparent. Prior studies on ascidian notochord formation have characterized the lineage of presumptive notochord cells by endogenous (Conklin, 1905; Reverberi, 1971) or applied cell markers (Ortolani, 1955; Nishida & Satoh, 1983), cellular interactions and the determinative factors responsible for specifying notochord differentiation (Reverberi, 1971; Whittaker, 1979), structure of the notochord in the larva (Welsch & Storch, 1969; Katz, 1983), changes of the notochord during metamorphosis and tail resorption (Cloney, 1969; 1982), and involvement in neural plate induction (Rose, 1939; Reverberi, Ortolani & Farinella-Ferruzza, 1960; Reverberi, 1971). Early descriptions of notochord formation (Conklin, 1905; Berrill, 1955) do not examine suggested morphogenetic mechanisms in a critical manner whereas ultrastructural studies of Cloney (1964) and Mancuso & Dolcemascolo (1977) do not examine directly any of its dynamic aspects. This study takes advantage of the relative transparency of embryos of the ascidian, Ciona intestinalis, the ability of Differential Interference Contrast (DIC) microscopy to optically section and visualize details in thick specimens, and the improved contrast resulting from current videotechnology (Allen, Allen & Travis, 1981), to follow presumptive cells from mid-gastrulation through tadpole formation and describe their dynamic behaviour as they form the notochord.

Gametes of Ciona intestinalis were obtained from mature adults, fertilized, dechorionated, and cultured according to procedures described by Crowther & Whittaker (1983). Embryos to be studied were placed in a perfusion chamber constructed out of microscope slides, coverglasses, No. 18 syringe needles, intramedic tubing and epoxy cement. The chamber was thin enough (1–1·2 mm) to use oil immersion and other high numerical aperature objectives while maintaining a temperature of about 19 °C and a continuous flow of fresh sea water past the embryo. Newly fertilized embryos placed in this chamber developed normally to the time of hatching (18–20 h). Development times reported are the actual age of the embryo since fertilization at the culture temperature. Development occurs at the rate reported by Whittaker (1979) for 18 °C. Because the tail begins to twitch around 13 to 15 h of development, embryos were anaesthetized by perfusing 0·1 mM-nicotine through the chamber beginning at 12 h postfertilization. This treatment had no adverse effects on development whereas other anaesthetics tried (MS-222, Chloro tone) stopped both tail elongation and further development of the notochord.

A Leitz microscope equipped with Nomarski DIC optics was used to make the recordings. A Dage Model 65 Newvicon camera connected to a time-lapse videorecorder (either a Sony model TVO-9000 (⁠ inch tape) or a Panasonic model NV-8050 (⁠ inch tape)) with time–date generator, and high-resolution videomonitors comprised the videosystem. Tapes were recorded at speeds ranging from 1/12 to 1/36 actual speed. Photographs were taken of the monitor screen with a 35 mm camera as described by Allen et al. (1981). Scan lines were eliminated in the photographs by using a Ronchi grating, 50 cycles/inch (Rolyn Optics, Covina, CA) either during the making of the negative or during printing. The mapping of cell rearrangement as well as the measurements of tail length, cell diameter, and cell length were done from tracings of the monitor screen of selected time points using transparent acetate sheets. Cell volume was calculated by assuming cells were perfect cylinders of constant diameter.

The emphasis of this study was on observing the dynamic behaviour of cells, events that can only be seen when they are speeded up during playback of the videorecordings. Events that occurred in a relatively brief moment of time could be captured and observed using various playback features of the recorder (different speeds, forward and reverse playback, stop field and one shot advance). Some of the changes are such that even the most patient of still photographers would have missed them. Other advantages of the technique are (1) developmental changes in the same cells could be followed continuously through time; (2) constant refocusing was unnecessary since videorecording is vibration free; (3) only low levels of illumination are required, reducing the amount of heat generated that can kill the embryo; (4) a record of time is provided by the time-date generator; and (5) results can be viewed immediately upon completion of taping. The major problem with the technique is that the videorecorder has a horizontal resolution of only 300+ lines. The resolution of the camera and monitor is 700 fines. Stopping on a particular ‘frame’ of the tape in order to photograph the screen resulted in further loss of definition as only one set of scan lines is displayed.

In vivo studies were complemented by histological studies to examine some of the changes difficult to see with the video. Styela clava, another species of ascidian, was used to take advantage of its distinctive differences in pigmentation and cytoplasm that mark cells of different presumptive fates (Conklin, 1905). Gonads from two different adults were minced together to fertilize the eggs and debris was filtered out using nylon mesh cloth. Various stages were fixed with 2% glutaraldehyde in sea water, embedded in JB-4 resin (Polysciences), and serially sectioned at 4 μm using glass knives on a AO rotary microtome. Measurements and reconstructions were done from drawings of complete sets of serial sections made using a projection microscope.

Ciona embryos at selected stages of development were fixed in 2·5 % glutaraldehyde in a 0·2 M-phosphate buffer (pH 7·2), containing 0·34M-NaCl for 60 mins, and postfixed in 2% osmium tetroxide in 1·25 % bicarbonate buffer (pH 7·2) for 15 mins. After a buffer wash, they were then dehydrated through an ethanol series, cleared with propylene oxide, and embedded in an Embed 812-Araldite mixture (Mollenhauer, 1964) at 60 °C in flat embedding moulds. Thin sections were cut with a diamond knife on a Sorvall MT-2 B ultramicrotome and stained at room temperature with alcoholic uranyl acetate (7·5 % in 50 % ethanol) for 3 to 5 mins and aqueous lead citrate (Reynolds, 1963) for 3 mins. Preparations were examined and photographed in a Zeiss IOC transmission electron microscope at 80 kV.

Early changes during notochord formation are as described by Conklin (1905). At the start of gastrulation presumptive notochord cells are arranged in an arc that lies anterior and lateral to the endoderm (Fig. 1A). As invagination begins, these cells elongate, extending radially from the anterior margin of the blastopore. As blastopore closure occurs, cell division converts the initial single arc of cells into two (Fig. 1B) and then three tiers of cells.

Examination of the time-lapse videorecordings shows closure of the blastopore to be due mostly to the posterior movement of the anterior margin of notochord and overlying neural plate cells between flanking bands of muscle and mesenchyme (Fig. 1B). Consequently the descendants of the central cells of the arc (Fig. 1A) that are still attached to the blastopore margin become the posterior-most cells of the rudiment formed by the notochord cell lineages established by Conklin (1905) and Ortolani (1955) (Figs 1B, 1C). However, the blastopore does not end up at the very posterior tip of the embryo but rather stops anterior to it (Fig. 1D). As the blastopore moves posteriorly and becomes smaller, notochord cells that line it become wedge shaped, but their constricted apical surfaces do not show any activity such as blebbing. Cell contours remain smooth and no surface extensions are seen that can be interpreted to be motile.

As the neural tube is formed, the width of the notochord rudiment decreases as it elongates. This change is accompanied by cell rearrangement (Fig. 2). Cells exchange neighbours, moving toward the midline of the rudiment and causing longitudinal separation and displacement. As cells elongate across the axis of the rudiment they become wedge or spindle shaped with tapered ends that extend between neighbouring cells. Once they reach the opposite side they become disc shaped, appearing rectangular in the recordings (Figs 1F, 2H). The resulting sequence of stacked cells appears to be random. Cells that start from the same side of the primordium may or may not end up as neighbours once interdigitation is complete (Fig. 2H), an observation that agrees with the results of Nishida & Satoh (1983). Vertical rearrangements also occur. Beginning as a single layer (Fig. 3A), the notochord becomes a bilayer (Fig. 3B) and then becomes circular to oval in cross section (Fig. 3C).

The interdigitating movement of cells appears to be due to the penetration between neighbouring cells of the tapered cell projection (Figs 1E, 2) as reported by Cloney (1964). During the early stages of interdigitation (Figs 2A–D) the image is not sharp enough to see the cellular activity associated with interdigitation. As the diameter of the notochord and tail decreases and the image becomes clearer, careful study of the videotapes show the cells to be pulsating as rearrangement takes place (Figs 2E–H). The entire surface of the notochord cell adjacent to the perinotochordal sheath bulges in and out in a rhythmic fashion. Cells and cell contents exhibit back- and-forth movements that run parallel to the direction of cell translocation. In the best of the recordings, the leading, tapered extensions appear to show minute dilations and attenuations while at other times they have an undulating or wave-like quality.

By the time notochord cells have completed their rearrangement, the embryo is 10–11 h old and is comma shaped (Fig. 4H). Vacuoles begin to appear in some of the cells around 13 h (Figs 1G, 5A, 5B). At first not all cells have these vacuoles (terdigitating movement of cells appears to be Figs 1G, 5A) and they appear and disappear from view. By 14–14·5 h, however, they are larger and stable, arranged in pairs to either side of a separating boundary (Fig. 1H). These boundaries remain distinct as the vacuoles increase in size (Figs 1H, 4A). The vacuoles are at first irregular in shape (Fig. 5A, 5B). Their behaviour is very dynamic as they are constantly bulging inward and outward in the recordings. As they become larger they become ellipsoid or spherical (Fig. 4A) and their activity declines in intensity.

At about 16 h after fertilization the boundary separating adjoining vacuole pairs becomes less distinct (Figs 4B,C) and disappears as they fuse to form one larger vacuole (Fig. 4D). In shape and position these fused vacuoles correspond to the intercellular vacuoles described by Cloney (1964). Vacuolar growth continues (Fig. 4E,F) as the cytoplasm of notochord cells become less granular. At the time that hatching would normally occur (18–19h after fertilization), cells shift toward the periphery of the notochord and vacuoles merge (Figs 4F,G). At this point the tail has almost reached its maximum length (Figs 4F, 6) and the notochord is a central core of clear material that is surrounded by attenuated notochord cells (Fig. 4G). Surrounding the notochord cells is a sheath of fibrillar extracellular material (Fig. 5C; Cloney, 1964; Katz, 1983).

After gastrulation is complete, tail elongation in Ciona proceeds at a constant rate of about 1·3 μm/minute (Fig. 6). At the magnifications that were used, the tip of the tail moves out of the field of view as it elongates. Most recordings were made near where the tail joins the body of the developing tadpole (Fig. 4H) in order to monitor cellular changes taking place continuously. Cell diameter decreases and cell length increases in this area as the tail elongates (Figs. 1E–H; 7A). These changes occur mostly during the early part of tail elongation and begin to decline in extent at about 14h (Fig. 7A). After adjoining vacuole pairs have fused (16h) there is no indication of any further longitudinal movement or cell lengthening in this region (Figs. 4D–G; 7A). Since the rest of the tail is still elongating, it appears that cell shape changes are completed first in anterior part of the notochord before they are in the posterior part.

There is a gradual rise in the estimated cell volume up until about 14 h (Fig. 7B). Comparing the data for this period (10–14 h) by means of analysis of variance resulted in F-values that were not significant at a 95 % confidence level. The interpretation of this result is that growth is at best a minor component and notochord elongation during this period is principally due to changes in cell shape. After 14 h the average cell volume rises as intracellular vacuoles grow in size (Fig. 7B). This change is not uniform, however, since the amount of variability in cell volume is greater for this period than the one preceding (Fig. 7B). Comparison of the length, diameter, and volume measurements shows that the increase in volume is principally the consequence of an increase in notochord diameter after 14 h (Fig. 7A). More importantly, the major component of apparent growth or swelling that occurs during notochord formation happens after cells have completed most of their lengthening.

Cell surfaces adjacent to the other surrounding tissues of the tail being to bleb extensively as notochord cells begin to elongate (Fig. 1H). Blebbing occurs only after interdigitation is complete and is therefore not a component of cell movement involved in interdigitation. Transmission electron micrographs show that during this period there are membrane-enclosed vesicles adjacent to these surfaces that contain fibrillar and amorphous material (Fig. 5C; Crowther, unpublished data; Cloney, 1964; Mancuso, 1973). Cloney (1964), and Mancuso & Dolcemascolo (1977) suggest that this period is one of very active secretion by notochord cells of perinotochordal sheath material. The blebbing observed here, therefore, is believed to be associated with this secretory activity.

Chalk marks located on the anterior lip of the blastopore do not end up at the posterior tip of the notochord but rather some distance from it (Ortolani, 1955; Reverberi, et al. 1960). Cell lineage studies that mark early blastomeres by injecting them with horseradish peroxidase (HRP) suggest that the most posterior notochord cells are not products of the established notochord cell lineage, but are derived instead from posterior blastomeres of the 8-cell stage (Nishida & Satoh, 1983). The time-lapse videostudies of morphogenetic movements corroborate that finding. They also help in understanding Nishida & Satoh’s (1983) report that HRP-stained muscle cells appear on the side contralateral to the one injected since the strict separation between muscle and mesenchyme bands of the left and right sides that results from the passage of notochord rudiment between them does not occur past the point where the blastopore stops its movement (Fig. ID), the point at which these particular muscle cells appear to be located.

Blastopore closure is associated with apical constriction and the formation of wedge-shaped cells in a fashion similar to that described for other examples (see Trinkaus, 1976, 1984). Posterior movement of the margin, however, does not appear to be due to cell motility that involves motile cell extensions (see Trinkaus, 1976, 1984). No surface activity that could be associated with active motility was observed. Scanning electron micrographs of ascidian gastrulae also give no indications of motile activity (Satoh, 1978).

Cloney (1964) and later Mancuso & Dolcemascolo (1977) report that the large vacuoles between cells are formed by direct accumulation of material between cells and that these intercellular structures later merge as a consequence of cells shifting toward the sheath as the central matrix core is formed. In contrast, Berrill (1947) reports that the vacuole formation is intracellular and vacuoles grow ‘until the ends of adjoining cells break down and the notochord becomes essentially a long continuous cylindrical vacuole, enclosed by fused peripheral cell walls’. With video time lapse the development of particular vacuoles in Ciona could be followed continuously from the blastula stage until completion of tadpole formation. The recordings clearly show that vacuoles develop initially as adjoining pairs that fuse to form a single, larger vacuole. The interpretation of this observation is that vacuoles initially develop as intracellular ones and that these then fuse to form intercellular ones. That early vacuoles are indeed intracellular could be shown by using electron microscopy to examine embryos at the particular times indicated by the videorecordings (Fig. 5). The thin and fragile boundary that separates vacuole pairs later in development appears to be easily disrupted, possibly because of osmotic stress produced during preparation for electron microscopy (Crowther, unpublished observations). Showing it in an intact state by electron microscopy during these stages has not been possible. Subsequent events, as observed by continuous videorecording, agree with Cloney’s and not with Berrill’s description. Intercellular vacuoles continue to increase in size and later merge as cells shift toward the periphery of the notochord.

Why do these observations differ from those of Cloney (1964)? The micrographs presented in Cloney’s (1964) paper show a very early stage in vacuole formation where adjoining cell membranes have just begun to separate and a much later stage, just prior to hatching, where the intercellular vacuoles are well formed. There is a 4 h gap between the two sets of micrographs in which there is only a description of the intermediate stages of vacuole formation that is based on observations with a dissecting microscope. It may be that Cloney missed examining closely the critical stages indicated by continuous time-lapse recording. Another possible explanation is that formation of the notochord is not the same for all species of ascidians. For example, in the tunicate Dendrodoa no vacuoles or central matrix core are formed (Welsch & Storch, 1969) whereas in a number of other species only small vacuoles are formed that nevermerge to form an inner matrix core (Berrill, 1947;Cloney, 1964). In most vertebrates miracellular vacuoles are formed but no matrix core is formed (Mookerjee, et al. 1952; Lesson & Lesson, 1958; Jurand, 1962, 1974; Waddington & Perry, 1962; Bancroft & Bellairs, 1970; Ruggeri, 1972). Other species need to be examined by time lapse in order to determine if this is the reason for the discrepancy.

Tail elongation depends on the formation of the notochord. Partial embryos, from which all presumptive notochord cells have been deleted, form rudimentary tails that do not elongate completely (Reverberi, et al. 1960). The small amount of elongation that does occur is probably due to rearrangements and shape changes of the muscle cells that are present (see Cavey & Cloney, 1974). Partial embryos that contain presumptive notochord cells and lack muscle cells form elongate tails, the length of which depends somewhat on the number of notochord cells present (Reverberi, et al. 1960). The quantitative data presented in this paper indicates that cell growth or swelling is not a major cause of elongation. Berrill’s (1955, 1975) proposal that elongation is due to the colloidal matrix imbibing water and generating a pressure that is restricted by the perinotochordal sheath to exert its force along the axis of the embryo is not correct. Most of the elongation in a region occurs when the vacuoles are still small and before the first indications of major growth or swelling are apparent. The fact that not all urodele species of ascidians form intercellular vacuoles or a matrix core in the tail (Berrill, 1947; Cloney, 1964) is another argument against swelling or growth of the notochord being important in shaping the tail.

After formation of the initial rudiment by cell division, notochord cells interdigitate and rearrange to form a rod shaped structure in which they are lined up end to end (Conklin, 1905; Cloney, 1964). Blastomere deletion and isolation experiments show that notochord cells have a tendency to line up end to end even when normal structural relationships with other presumptive tissues have been greatly disrupted (Von Ubisch, 1939; Rose, 1939; Reverberi, et al. 1960; Reverberi, 1971). These experiments suggest that the ability to rearrange is intrinsic to notochord cells and is not the result of external forces. This implies that the mechanism of neighbour exchanges is an active one. The videorecordings suggest two possible mechanisms by which interdigitation may be occurring. The pulsating movements could be indicative of forward surges of cytoplasm that ‘push’ the tapered extension between neighbouring cells. New cell adhesions are formed, followed by a period of reorganization of the cytoskeleton, and then another forward surge occurs. The other possibility is that the undulating movements of the tapered extension represent membrane ruffling that is modified by the geometry of the substrate (neighbouring cells) encountered (see England & Wakely, 1979). In this case interdigitation is postulated to be due to active cell motility, since ruffling is a behaviour associated with motility in other cells (see Trinkaus, 1976; 1984).

Cell rearrangement appears to be an important component of the shaping of embryonic structures in other organisms (Fristrom, 1976; Fristrom & Chihara, 1978; Keller, 1980; Poole & Steinberg, 1981; Kageyama, 1982; also see Trinkaus, 1976; 1984). While cell rearrangements in some cases appear to be passive and the consequence of other active morphogenetic changes (Kageyama, 1982), in other examples there is evidence that it requires the active translocation of cells past one another (Fristrom, 1976; Fristrom & Chihara, 1978). Only in the case of imaginal disc évagination has the question of how cells rearrange been well studied. Fristrom (1983) has documented changes in the junctional complexes between imaginal disc cells by freeze-fracture studies and has proposed that these changes are what are responsible for rearrangement. Thus it is very possible that the behaviours exhibited by notochord cells as they rearrange are not related at all to their movement. Further study is needed to clarify this point for the ascidian notochord.

Subsequent to the completion of cell interdigitation, further tail elongation correlates with a shape change of the notochord cells. Videorecordings of anterior half embryos show that notochord cells are ‘pushed’ out of the posterior end of the fragment and form chains (Miyamoto, unpublished observations, see also Reverberi, 1971). The spewing out of cells occurs during the period of cell shape changes and not during the period of cell interdigitation and suggests that they are actively producing an elongating force. Cell rearrangement and shape changes, therefore, appear to be principal behaviours responsible for the elongation of the ascidian tail. Ascidians are ideal organisms to study further these behaviours and the cellular mechanisms that cause them. In these embryos these dynamic cell changes can be correlated with ultrastructural changes and probed experimentally using manipulations that already have been used to study other aspects of ascidian embryogenesis.

This study was made possible in part by a Steps-Toward-independence Fellowship from the Marine Biological Laboratory, a University Research Council Grant from Seton Hall University, and National Science Foundation Grant No. RII-8210021 (Equipment for Two and Four Year Colleges and Universities) to D. M. Miyamoto. Additional support was provided by Grant No. HD-16547 from the National Institute of Child Health and Human Development to J. R. Whittaker. We would like to thank J. R. Whittaker, Tom Meedel, and Jane Loescher of the Boston University Marine Program for the assistance they provided. Our thanks also go to Sandy Bóchese and Pat Schall of the Seton Hall Educational Media Centre for their greatly appreciated assistance during the preparation of the figures.