Cell rearrangement during gastrulation of Xenopus: direct observation of cultured explants

The dorsal axial mesoderm is of central importance to the development of the vertebrate body plan. This region has long been known as the inducer of the nervous system and the’ organizer’ of the axial mesodermal tissues (see Spemann, 1938; Gerhart, 1980). Moreover, this region serves as the principal engine of gastrulation, around which other events are organized (Keller, 1986). Its autonomous extension (lengthening) and convergence (narrowing), together with similar movements in the prospective neural plate, drive involution of the marginal zone, close the blastopore and elongate the body axis. The power of these movements in Xenopus and other amphibia has been known for some time (see Vogt, 1929; Schechtman, 1942; Keller, 1986). However, their role was only recently worked out in detail by Keller and Danilchik (1988), who used double, or sandwich expiants to explore their regional distribution and timing.

Convergent extension of the axial mesoderm appears to be driven by active cell intercalation, both in the radial direction (normal to the surface of the embryo), and in the mediolateral direction (Keller, 1980; Keller et al. 1985a,b). Cell rearrangement underlies many other morphogenetic processes, in groups ranging from insects to mammals, yet in no case is its cellular basis understood (Keller, 1987; Fristrom, 1988). Thus our continued investigation of the mechanism and function of cell intercalation promises not only to yield greater understanding of amphibian gastrulation, but to shed light on a developmental strategy of great generality.

We have developed a method to observe directly and record the behavior of deep cells of the gastrula for the first time, based on culturing expiants in a special saline, Danilchik’ s solution, which allows normal development of dorsal axial tissues (Keller et al 1985a,b). Using a combination of video and time-lapse 16 mm micrography to produce high-contrast, high-resolution recordings of cell behavior in these expiants, we have begun to analyze directly the cellular events underlying axial morphogenesis (Keller and Hardin, 1987). We have shown how radial and mediolateral cell intercalation function in the morphogenesis of the somitic mesoderm (Wilson et al. 1989) and of the notochord (Keller et al. 1989a) of the neurula stage. Here, using the same methods, we analyze the function of these cell behaviors in the morphogenesis of the dorsal axial mesoderm of Xenopus laevis during gastrulation. We now have a detailed picture of what actually occurs, at the cellular level, during the mysterious elongation and narrowing of the axial mesoderm. This understanding comes at an exciting time, since as molecular and genetic investigation of vertebrate development advances, it will become increasingly important to have a sophisticated understanding of how cells actually build embryos.

Xenopus laevis embryos were obtained by standard methods and dejelhed in 3· 5% cysteine hydrochloride, pH7·9. Embryos were reared in one-third strength modified NiuTwitty medium (phosphate buffer has been replaced by 5 DIM Hepes, pH7·4). Vitelline envelopes were removed manually with sharpened watchmaker’ s forceps (Dumont no. 5) shortly before the embryos were used. All staging was according to Nieuwkoop and Faber (1967). Stages of expiants were inferred from control embryos cultured in the same dish.

Expiants comprising about 90 degrees of the dorsal marginal zone were cut from stage 10 to 10·5 embryos in modified Danilchik’ s solution as shown in Fig. 1, and their inner surfaces cleaned of involuted mesodermal cells with a hair loop, as described previously (Fig. 3 in Keller and Damlchik, 1988). Each explant was then placed face-down in a 35 mm plastic Petri dish hd containing the same solution and covered with a covershp fragment, supported on both ends by stopcock grease The Ud was then pushed against an inverted 60 mm Petn dish and the entire preparation inverted again, so that the deep side of the explant could be viewed from above on the stage of a compound microscope. These expiants are called ‘open-faced’ to distinguish them from double or ‘sandwich’ expiants, in which the mesoderm is not exposed (Keller and Damlchik, 1988).

Danilchik’ s solution, first described in Keller et al (1985a,b), has been modified slightly It differs from the original formulation in that gluconate has been substituted for isethionate and the bicarbonate concentration has been lowered The new version contains 53DIM NaCl, 15·0mM NaHCOs, 4 5mM potassium gluconate, 1·0mMCaCh, 1·0mM MgSO₄, 5·0mM bicine buffer, 13·5 mM NaCO₃, and Hepes buffer to bring the pH to 8· 3. This solution is more stable than the original version, which tended to form a precipitate over time. Whole-mount antibody staining was carried out as described in Wilson (1990).

Expiants were observed on a Zeiss 16 upright microscope, using 10× and 20× objectives and low angle fiber optic lamps Two kinds of time-lapse recordings were made. A combination of high-resolution video and 16 mm film was used to maximize both resolution and contrast. This system is described m Wilson et al. (1989). Some video recordings were made as well, using a Dage MTI 81 video camera and a Panasonic TQ-2028F optical memory disc recorder (OMDR). In some instances, the Image II video image processor (Universal Imaging Corporation, Media, PA) was used to average 16 video frames before recording. For recording of whole embryos, a chamber allowing counter rotation of the embryo was used (see Keller, 1978) to permit recording of the behavior of the NIMZ and IMZ regions as they move toward the blastopore hp Areas of traced regions of the embryo were measured using the MacMorph morphometries program (written by leff Hardin), run on a Macintosh II computer with a graphics tablet.

Open-faced expiants begin to extend along the animal-vegetal axis early in gastrulation (stage 10·25–10·5) and continue to lengthen until late neurula stages (Fig. 2). The rate of elongation is usually greatest in late gastrula and early neurula stages, but averages about 100μmh⁻ . After about the middle of gastrulation, extension is accompanied by convergence, or narrowing toward the dorsal midline. Only the involuting marginal zone (IMZ), consisting of prospective mesoderm and superficial endoderm (Fig. 1), participates in these movements. The animal portion of the explant, called the non-involuting marginal zone (NIMZ), is prospective neural tissue, which extends autonomously in double or sandwich expiants (Keller and Danilchik, 1988), but not in open-faced culture. There is no clear morphological boundary between IMZ and NIMZ when the expiants are made, but the tissues become distinct during extension of the explant. The cells of the IMZ are larger and more yolky than those of the NIMZ, and a fissure comes to separate the two populations in most specimens (see Fig. 3A). It is not known whether cells in the region of the boundary are committed to different fates before the distinction becomes manifest.

Longitudinal fissures appear within the extending IMZ toward the end of gastrulation, dividing the central notochord from the somitic mesoderm on either side (Fig. 3A). This occurs at about the same time in expiants as in intact embryos (see Keller et al. 1989a). Both mesodermal tissues advance well along their developmental paths in the open-faced expiants: the somitic mesoderm expresses the somite-specific antigen 12–101 (Kintner and Brockes, 1984) (Fig. 3B) and segmental fissures appear, while the cells of the notochord become long and thin, align transverse to the long axis of the notochord, and form characteristic vacuoles (pointers, Fig. 3B). We have studied these later aspects of mesodermal behavior and differentiation in explants of a different type, made at the beginning of neurulation; these results are presented elsewhere (Wilson et al. 1989; Keller et al. 1989a)

The deformation of the IMZ during convergence and extension involves dramatic flows of tissue. These gross patterns of movement can be mapped by tracing cells in low-power recordings. As Fig. 4 shows, cells at the midhne move vegetally (in the prospective anterior direction) as the expiant extends, while cells that begin laterally move first toward the midline and then vegetally. In some expiants, lateral material may actually be displaced posteriorly before sweeping around to join the central extending axis. The amount of convergence toward the midline is greater in the prospective posterior part of the mesoderm than at the leading edge. As a result, an anterior-posterior line, drawn parallel to the midline on a fully extended explant, would lie at an angle at the start of gastrulation. This accords well with the stage 10 fate map (Keller, 1976) in which both the notochord-somite and the somite-lateral mesoderm boundaries slant away from the midhne, as suggested by the dotted lines in Fig. 1. Moreover, the notochords that form in incompletely extended expiants are wider posteriorly than anteriorly, offering further confirmation of the arrangement of primordia in the fate map.

During the first half of gastrulation, the dorsal mesoderm extends primarily by radial intercalation. Cells repack to form fewer layers of greater area, expanding the tissue and thinning it. We have traced cell movements and intercalation in detail in six timelapse recordings of extending expiants, and studied many others less exhaustively In these recordings, radial intercalation manifests itself as a dramatic traffic of cells into and out of the exposed surface of the explant (which was the innermost surface of the mesoderm in the embryo). These movements are already underway at the start of recording (stage 10.25 or earlier). After a brief period of heavy traffic in both directions, the appearance of new cells from deeper layers strongly predominates over disappearance, and the tissue begins to expand, driving the leading edge forward. Fig. 5 shows a patch of cells at several stages of radial intercalation; Fig. 6 traces the rate at which new cells and area are added to the surface layer of this patch. The growth is explosive at first: the area of the patch increases by 70% in the first hour of culture. It slows somewhat in the next hour to 37 %, while in the third hour the area of the patch actually falls slightly. This drop has two causes. Radial intercalation falls off to where the few new cells still joining the surface are almost balanced by departing cells. In addition, the area of cells already at the surface shnnks somewhat, presumably reflecting the onset of thickening or the columnanzation of individual cells that occurs later, during mediolateral intercalation (see below). Until this point (about the middle of gastrulation), cell area has remained quite constant at about 1000 μm². As this implies, the number of cells in a surface patch grows at about the same rate as the patch’ s area during the period of rapid radial intercalation (Fig. 6).

The detailed pattern of cell rearrangement varies among expiants, but our analyses confirm the following general conclusions. Radial intercalation begins soon after the onset of gastrulation. The rate of intercalation into the explant surface soon begins to exceed the rate of disappearance, thus adding new cells and area to the surface. Expansion is rapid at first, tapering off by the middle of gastrulation. Exchange of cells between layers then continues at a much slower rate, and contributes little to further extension. This is the overall sequence of events, but there are significant local differences in timing. In particular, radial intercalation begins and ends earlier in the vegetal (prospective anterior) region of the mesoderm (see The timing of intercalation along the anterior-posterior axis below.)

Before turning to mediolateral intercalation and the second half of gastrulation, we must consider one remaining issue. Other things being equal, radial intercalation should produce an isotropic expansion, widening the explant as well as lengthening it. In general this does not happen, although patches of the mesoderm sometimes widen during intense radial intercalation. This may be because some mediolateral intercalation always accompanies radial intercalation but is obscured by the explosive appearance of new cells

After the middle of gastrulation, convergent extension derives mostly from mediolateral intercalation, the rearrangement of cells within each layer to form a longer, narrower array (As we have mentioned, mediolateral intercalation may take place deep within the explant from the start, reaching the exposed surface by the middle of gastrulation.) Fig. 7 follows a patch of cells near the dorsal midhne and near the anterior edge from the middle to the end of gastrulation. During this period only one cell (33) joins the surface; two cells (29 and 21) go under. Thus exchange of cells with deeper layers is no longer important in reshaping the tissue. Instead, the cells of this region undergo about one round of mediolateral intercalation during the recording, lengthening the patch by about two thirds while narrowing it by almost half. Adjacent cells along the axis of extension in the first tracing become separated by initially lateral neighbors. Consider for example cells 18, 20, 2 and 8.

This analysis confirms that mediolateral intercalation is not highly ordered: some anterior-posterior neigh-, hors remain together while others are separated by two or three intercalating cells. This feature of rearrangement in the mesoderm was first revealed by studying the intercalation of patches-of fluorescently labelled cells grafted into unlabelled gastrulae (Keller and Tibbetts, 1989). Furthermore, cells intercalate between mediolateral neighbors as well as anterior-posterior ones. It is the statistical predominance of rearrangement along one axis that produces net extension and convergence, much as a sheet of cells expands when entering cells outnumber leaving cells during radial intercalation. The degree of statistical bias can be estimated by comparing the frequency with which anterior-posterior and mediolateral neighbors are separated. In the patch followed in Fig. 7, 20 pairs of cells are initially in primarily antenor-postenor contact; only 7 are still neighbors at the end. In contrast, 16 of 23 predominantly mediolateral neighbor pairs stay together.

During the period of mediolateral intercalation the notochord-somite boundary forms, and intercalating cells begin to elongate mediolaterally, normal to the axis of extension. This change in cell shape, visible in the anterior cells of Fig. 7, continues during neurula stages in both the notochord and the somitic mesoderm (Keller et al 1989a; Wilson et al. 1989). Fig. 8 illustrates how the three processes -mediolateral intercalation, cell shape change and boundary formation -take place together during the second half of gastrulation. Cells of both the prospective notochord (shaded) and somitic mesoderm intercalate mediolaterally, driving extension, especially of the notochord. For example, both cell 21 and cell 6 in the notochord come between cells 2 and 18. In the somitic mesoderm, cell 54 separates cells 60 and 61. Cells in both tissues change shape as they intercalate, elongating transversely and narrowing along the axis of extension. The axis thus extends despite changes in the shape of individual cells. Finally, a fissure forms between the notochord and the prospective somitic mesoderm. Before this boundary appears, at around stage 12 or 12·5 in most expiants, the two prospective cell types are not obviously distinct, although subtle differences can be discerned in SEM at slightly earlier stages. In fact, it is not known at what stage cells become committed to one tissue or the other, or whether the two populations are mixed in the early gastrula.

Our initial analysis of radial and mediolateral intercalation suggested that each behavior is initiated at the anterior (formerly vegetal) end of the tissue, from where it moves posteriorly towards the boundary with the NIMZ. To investigate this hypothesis further, we traced cell rearrangement in two patches along the dorsal midline during several hours of video recording. This study, presented in Fig. 9 and supported by examination of other specimens, confirms that the progression of cell behavior is more advanced in anterior tissue.

As the recording begins, at stage 10·25–10·5, radial intercalation is occurring throughout the dorsal midfine, which at this point is only about 12 cells long. (Some material would have already involuted by this stage, and would not have been included in the explant). During the second half of gastrulation, radial intercalation continues in the posterior part of the axis. In the anterior region, however, almost no new cells are added to the surface, and further extension is driven by mediolateral intercalation. In the next interval, comprising early neurula stages, mediolateral follows radial intercalation in the posterior mesoderm as well. Although the anterior patch is not followed in detail through this period, rearrangement within the surface continues, here accompanied by cell shape change. The axis is now about 30 cells long and the two patches, which at first were only a cell diameter or two apart, are now at a considerable distance from one another.

Thus radial intercalation gives way to mediolateral intercalation sooner in the anterior part of the axis, and the first process continues at the posterior end for some time after the second has begun at the leading edge. It cannot be established from the recording analyzed in Fig. 9 whether radial intercalation is also initiated first at the anterior end, since the explant was not made early enough in gastrulation. The analysis of other recordings, however, strongly suggests that this is the case. The small number of cells constituting the prospective anterior-posterior axis at the beginning of gastrulation and the statistical nature of cell intercalation ensure that spatial pattern at this stage can only be crude. The nature of the work involved, moreover, prevents the detailed analysis of large numbers of expiants. Nonetheless, we are now persuaded that active cell rearrangement begins at the future anterior of the axis and moves posterior during gastrulation, and thus that an anterior-posterior gradient of development timing is already established in the prospective notochord before involution. We have not attempted a similar gastrula-stage analysis of timing differences in intercalation within the prospective somitic mesoderm, whose geometry is far more drastically transformed during extension. However, in expiants that contain enough lateral material to form several somites, expression of the 12–101 muscle antigen at late neurula stages falls off toward the posterior boundary with the NIMZ, suggesting an anterior-posterior gradient in the timing of somite differentiation (see Fig. 3B).

It is important to understand how the patterns of cell behavior observed in expiants relate to movements in the marginal zone of intact gastrulae. In particular, when and where do radial and mediolateral intercalation occur with respect to involution around the lip of the blastopore? To answer this question, time-lapse recordings were made of early gastrulae, and the distortion of the dorsal IMZ and the neighboring NIMZ was mapped. It is of course the outer, epithelial surface of the IMZ that is followed in these recordings, rather than the deep, mesodermal cells. Furthermore, the IMZ can only be followed until it involutes and disappears from view. Nonetheless, the analysis of these recordings, presented in Fig. 10, shows that each region of the IMZ begins to expand before involution, extending rapidly without much convergence. These observations confirm that the rapid radial intercalation and consequent extension seen in explants occur in embryos as well, and show that it takes place in mesoderm prior to its involution. Each region begins to converge more dramatically as it prepares to involute, suggesting that mediolateral intercalation occurs primarily at the lip and after involution.

The inductive powers of the dorsal mesoderm, first revealed by the organizer grafts of Hilde Mangold and Hans Spemann (Spemann and Mangold, 1924; Spemann, 1938), are once again the object of intense study. Both the induction of the nervous system and the assimilation of surrounding mesoderm into the dorsal axis are being re-examined, often with the aid of molecular markers of differentiation (Kintner and Melton, 1987; Sharpe et al. 1987; Dale and Slack, 1987; Savage and Phillips, 1989). The crucial morphogenetic function of this tissue, which builds the embryonic axis as it patterns it, has received much less attention. We know that convergent extension of the organizer elongates the axis and closes the blastopore, but we know little about the cellular behaviors that bring about these movements, and even less about how these activities are patterned and coordinated Furthermore, we do not understand how the morphogenetic undertakings of the organizer assist or constrain its patterning functions, although they clearly must. Extension presumably aids in neural induction by bringing the mesoderm below the ectoderm (but see Kintner and Melton, 1987), while convergence may facilitate dorsalization by drawing lateral and ventral tissue under the influence of the organizer (Keller and Danilchik, 1988; Slack et al. 1988). At the same time, the rapid cell intercalation underlying convergent extension must limit the refinement of early pattern formation. Finally, the characteristic and coordinated morphogenetic behaviors of the dorsal mesoderm must themselves result from pattern-forming interactions. How is this ‘dynamic determination’ of the organizer as extending tissue related to its ‘material determination’ as notochord and somites, and to its acquisition of organizer function?

To answer these questions, we will need to know in far greater detail what the cells of the dorsal mesoderm actually do during axis formation, both at the level of individual cell behavior and at the level of pattern and coordination of behavior. The method of explant culture in Danilchik’ s solution makes this kind of analysis possible for the first time. We will now consider some of these issues in the light of the results presented here.

Our analysis of gastrula expiants, together with our study of neurula-stage expiants (Wilson et al. 1989; Keller et al. 1989a), establishes by direct observation that active cell rearrangement underlies convergent extension of the axial mesoderm. In each region of the prospective notochord and somitic mesoderm, extension begins with rapid radial intercalation, causing thinning and expansion. This is followed by mediolateral intercalation, which drives further extension, now coupled to convergence. Different regions of the mesoderm begin this sequence of axial behaviors at different times. Along the dorsal midline, radial intercalation begins early in gastrulation, giving way to mediolateral intercalation at mid-gastrula stages, with anterior cells beginning the sequence sooner than posterior cells. Material that will form the first few somites has also advanced to mediolateral intercalation by the end of gastrulation. But prospective somitic mesoderm from more lateral positions, recruited into the emerging axis during gastrulation or neurulation, initiates intercalation only in neurula stages (Wilson et al. 1989).

There are pitfalls in inferring mechanical relationships from the movements and deformation of embryonic cells and tissues. For example, one could argue that the deep cell behavior analyzed here is passive, and that forces produced elsewhere drive convergent extension. However, the evidence weighs heavily against this argument. The epithelial component of the marginal zone shows no capacity to extend on its own, and its behavior is consistent with its being stretched passively by the underlying mesoderm (Keller, 1978, 1981, 198b). It is thus unlikely that the force of extension comes from the epithelium. Within the deep cell population (the mesoderm), it may be the cells closest to the epithelium that are the most active in convergent extension, rather than those exposed to view in the expiants that we have studied here. However, recent work in our lab shows that these cells behave much as the exposed cells do, although mediolateral intercalation may be more ordered and consequently more powerful among these cells, and may begin earlier (Keller et al. 1989b, 1991). Thus we believe that active intercalation of the entire deep cell population drives convergent extension, and that it is this force-generating behavior that our recordings of open-faced expiants have enabled us to capture and analyze.

Although we now understand in broad outline the pattern of cell behavior that drives convergence and extension and builds the dorsal axis, two areas of ignorance remain. First, we know little about how mesodermal cells move between one another. Involved here are cell biological questions about specialized protrusive activity, cytoskeletal changes, and the roles of extracellular matrix and cell-cell adhesion. Second, we do not understand how cell intercalation is organized and coordinated. We know from our analysis of time-lapse recordings that cells seem to move as individuals rather than in groups, neither conforming to a rigid geometry of intercalation nor moving in obvious coordination. Yet both radial and mediolateral intercalation result in directed and sustained extension of the tissue as a whole. Such coordination may require communication among the intercalating cells, perhaps in the form of cell polarizing signals, or even an attractive signal from an organizing region at the dorsal midline. We are currently exploring the establishment and maintenance of directional extension.

The associated epithelium itself may influence some aspects of mesodermal cell behavior. Shih and Keller have demonstrated that the dorsal epithelium can induce dorsal structures and morphogenetic behavior in ventral and UV-irradiated mesoderm, suggesting that it may direct or sustain intercalation and axial differentiation in the dorsal mesoderm as well (unpublished data) However, the epithelium is not strictly necessary for all aspects of dorsal mesodermal development, as shown by the fact that early gastrula expiants deprived of the epithelium differentiate into notochord and somite and undergo cell intercalation, although their extension is greatly reduced (Wilson, 1990).

It is well established that convergent extension of the dorsal marginal zone is responsible for many aspects of gastrulation and neurulation, shaping the dorsal axis and closing the blastopore (Keller, 1986; Keller and Danilchik, 1988). The role of these forces in involution has not been as clear. The results presented here, however, which clarify the roles of radial and mediolateral intercalation, allow us to propose the following model (presented diagrammatically in Fig. 11). The initial expansion of the IMZ by radial intercalation, now known to take place before involution, pushes the vegetal end of the marginal zone toward the blastopore lip. Once near the lip, mediolateral intercalation and convergence begin in earnest, reducing the circumference of the IMZ just as it involutes. Since involution involves decrease in circumference, this convergence may help to drive material around the hp. The involuted IMZ continues to converge and extend on the inside, squeezing the circumblastoporal region and thus closing the blastopore. It is likely that the extension of the NIMZ and the migration of the leading edge of the mesoderm toward the animal pole assist in involution, pushing on the outside and pulling on the inside. However, experiments in which the blastocoel roof and even the NIMZ are removed have shown that neither of these processes is required for involution (Keller et al. 1985a b); our model explains how forces produced within the IMZ itself could account for this essential aspect of gastrulation.

The anterior-posterior polarity of the vertebrate embryo apparently originates in the dorsal mesoderm, passing to the nervous system during neural induction (reviewed in Hamburger, 1988). One of the earliest manifestations of polarity along this axis is the gradient of developmental timing in the axial mesoderm. Our analysis of neurula stage expiants demonstrated that this gradient is well-established by the end of gastrulation. In the somitic mesoderm, cells progress through a sequence of cell behaviors (Wilson et al. 1989), beginning with radial intercalation, followed by cell elongation and mediolateral intercalation, and culminating in segmentation and somite rotation (see Fig. 10 in Wilson et al. (1989). Cells at the anterior end of the axis are further along in this sequence than posterior cells, and each activity sweeps caudally through the tissue as development proceeds. A similar but less pronounced anterior-posterior timing pattern is manifest in the notochord as its cells progress through several stages of shape change and rearrangement (Keller et al. 1989a)

The work presented here shows that the axial timing gradient is already present in rough form at the start of gastrulation, at least along the dorsal midline. Cells at the most vegetal (prospective anterior) end of the marginal zone initiate both radial and mediolateral intercalation before more animal or posterior cells. Thus, the events of gastrulation, in particular involution around the lip of the blastopore, are apparently not responsible for initiating these timing differences. It is tempting to hypothesize a link between the local timing of gastrular behavior and the local timing of mesoder-malization during blastula stages, perhaps reflecting the gradual advance through the marginal zone of a signal from the vegetal endoderm. Cooke, however, noted that the stage at which ectopic mesoderm induced in the blastocoel roof by injected XTC-MIF undergoes a morphological transition associated with gastrulation is not influenced by the timing or strength of the inductive signal, arguing against the simplest notions of anteriorposterior patterning during mesoderm induction (Cooke et al. 1987).

The movements of convergent extension may have a profound effect on the dorsoventral structure of the mesoderm as well. Slack and his colleagues have demonstrated that much of the prospective somitic mesoderm, located in lateral and even ventral sectors of the marginal zone, is not yet committed to differentiate as muscle at the start of gastrulation (Dale and Slack, 1987). A dorsalizing influence from the organizer region apparently acts during gastrulation to bring this material into the dorsal axial structures. The movements of convergence almost certainly enhance this process, bringing more tissue under the influence of the dorsal midline, as recently suggested by Keller and Danilchik (1988) and by Slack and his co-authors (1988). We are currently testing this notion directly by studying tissue differentiation in expiants in which morphogenetic movements have been arrested. Convergent extension and dorsalization by the organizer are linked in a second way. We believe that dorsalization not only confers a dorsal tissue fate (muscle) on more ventral mesoderm, but also induces dorsal morphogenetic behavior (convergent extension) by active cell rearrangement. John Shih in our laboratory has recently demonstrated that grafts of mid-dorsal material can induce convergent extension in normally non-extending ventral mesoderm (unpublished data). Thus the organizer is involved in dynamic as well as material determination of the surrounding mesoderm. Differentiation as somite or notochord and convergent extension may be manifestations of a single dorsal determination, or they may represent independent, potentially separable inductive events. It is worth noting in this context that blastula animal caps induced by XTC-MIF or by purified growth factors such as FGF and activin acquire morphogenetic behavior reminiscent of the dorsal marginal zone as well as mesodermal tissue fates (Smith, 1987; Symes and Smith, 1987; Slack et al. 1987).

This work was supported by National Institutes of Health grants HD18979 and HD25594 and National Science Foundation grant DCB89052 to Ray Keller, NSF grant DMS 8618975 to Paul Wilson and George Oster; Paul Wilson was supported in part by National Institutes of Health Training Grant HD7375 We thank John Shih for his insights and advice, Jessica Bolker and Steve Minsuk for reading the manuscript, and Paul Tibbetts for technical assistance.