Patterns of cell motility in the organizer and dorsal mesoderm of Xenopus laevis

In this paper, we examine the pattern of morphogenetic cell behavior in the dorsal, axial mesoderm of Xenopus. Mesoderm plays a major role in the organization and patterning of the embryo. Spemann and Mangold (1924) demonstrated the role of the dorsal mesoderm (the Spemann organizer) in neural induction (also see Spemann, 1938; Mangold, 1933; Holtfreter, 1933a-d; Keller et al. 1992a,b; see Nieuwkoop, 1973 for a review). Additional studies have clarified the role of the dorsal mesoderm in assimilative induction, the recruitment of lateral and ventral mesoderm into dorsal axial structures (see page 163, Spemann, 1938; Cooke, 1972, 1975, 1985, 1989; Dale and Slack, 1987; Stewart, 1991; see Cooke, 1991 and Gerhart et al., 1991). The morphogenetic functions of notochordal and somitic mesoderm, which form the structural framework of the vertebrate body plan, have been studied extensively (Schechtman, 1942; Holtfreter, 1939, 1943a, b, 1944; Nakatsuji, 1975, 1984; Keller, 1981, 1984; Keller et al., 1985a,b; Nakatsuji and Johnson, 1983; Shi et al., 1987, 1989; Winklbauer, 1990; Winklbauer and Nagel, 1991).

There is evidence that proper gastrulation movements depend on precise patterning of mesoderm morphogenesis (Keller, 1986; Keller et al., 1991a,b). The early-involuting, leading edge of the mesodermal mantle shows directed migration across the roof of the blastocoel (see Nakatsuji, 1975; Winklbauer, 1990; Winklbauer and Nagel, 1991), where it differentiates into head (prechordal), heart, lateral plate and blood islands. In contrast, the late-involuting, axial and paraxial mesoderm shows the powerful convergence and extension movements that alone can drive involution and blastopore closure (Keller et al., 1985a,b; Keller and Jansa, 1992). Convergence and extension are driven by radial and mediolateral intercalations of deep mesodermal cells (Keller and Tibbetts, 1989; Wilson et al., 1989; Keller et al., 1989a; Wilson and Keller, 1991; Shih and Keller, 1992a). These two types of cell intercalations are highly patterned in the neurula (Wilson et al., 1989; Keller et al., 1989a), suggesting that patterning of cell behaviors also might be a fundamental part of the preceding gastrulation movements. But observations of mesodermal cells in open- faced explants reveal only a weak anterior-posterior pro- gression of cell behavior during gastrulation (Wilson and Keller, 1991).

However, it is not the deepest mesodermal cells, which were viewed in the open-faced explants, but the ones imme- diately beneath the overlying endodermal epithelium that are strongly patterned. In a companion paper, we show that the endodermal epithelium can organize convergence and extension of the underlying deep cells, suggesting that those cells closest to the organizing epithelium might be the first to respond with specific cell behaviors driving convergence and extension (Shih and Keller, 1992b). The mesodermal cells immediately beneath the endodermal epithelium of the DMZ show a highly organized morphology and arrange- ment (Keller et al., 1989b). When these cells are exposed in ‘shaved’, open-faced explants, they show a highly orga- nized sequence of events, called mediolateral intercalation behavior (MIB), that drive mediolateral intercalation (Shih and Keller, 1992a). MIB consists of bipolar, medially and laterally directed protrusive activity, which exerts traction on adjacent mesodermal cells’ surfaces, and thus produces mediolateral cell elongation, alignment and intercalation (Shih and Keller, 1992a).

In this paper, we use the shaved, open-faced explant preparation to follow the progression of mediolateral inter- calation behavior and subsequent tissue differentiation over large fields of deep mesodermal cells through gastrulation and neurulation. These observations reveal that this behav- ior is highly patterned with respect to the lateral-medial and anterior-posterior axes, and with respect to the notochord and somitic territories. These patterns have important impli- cations for how the mesodermal tissues are organized and how convergence and extension function to bring about involution and extend the embryonic axis.

The methods used in this paper follow closely those described in the companion paper (Shih and Keller, 1992a). Eggs were obtained, dejellied and labelled with 10-15 nl of FDX 10,000 (Sigma Chemical Company) made up at 25 mg/ml in 0.2 N KCl (Gimlich and Braun, 1985) as described previously (Shih and Keller, 1992a). Shaved open-faced explants of the dorsal marginal zone (DMZ) were made as described previously (Shih and Keller, 1992a). For low-light filming, FDX-labelled deep mesodermal cells were added to the inner (deep) surface of unlabelled ‘host’ explants. Labelled cells were taken from sibling FDX-labelled embryos by peeling back the suprablastoporal endodermal epithe- lium and teasing out a small number of cells from the presump- tive chordamesoderm lying beneath the epithelium. These labelled cells were added to the same region of unlabelled, shaved explants of the same stage, and the host explants were cultured in Shih- modified Danilchik’s solution (SMD: see Keller, 1991).

Cell behaviors are described below with respect to the prospec- tive lateral-medial and anterior-posterior axes of the notochordal and somitic mesoderm, as they map back to large, shaved explants of the involuting marginal zone (IMZ) (Fig. 1). Note that the prospective anterior-posterior axis of the notochord runs from veg- etal-to-animal at the midline but in the prospective lateral region it curves into the lateral sector of the IMZ. This curvature is extreme in the case of the prospective somite files, which have their prospective anterior-posterior axes curved laterally and ven- trally in the IMZ. Note also that the prospective lateral edges of the somitic mesoderm lie next to where the vegetal endoderm was attached. The medial edges of the first six or seven somites bound the notochord, whereas the remaining ones do not, at least at the surface immediately beneath the endoderm; the notochord shears posteriorly along the medial edges of these posterior somites in the neurula stage (see the text and Fig. 5, Keller et al., 1991a, for details).

Whole-mount staining of somitic mesoderm, microscopy under epi-illumination, fluorescence microscopy, video recording and image processing were done as described previously (Shih and Keller, 1992a). The embryos were fixed, dissected and prepared for scanning electron microscopy by methods described previously (Keller et al., 1989a). We digitized cell shapes by tracing images from video monitors to mylar sheets and digitizing them on a Summasketch tablet (Summagraphics). Parameters of cell mor- phology were quantified from the digitized data as described pre- viously (Shih and Keller, 1992a).

We will describe the patterns of expression of MIB in the mesoderm by using cell elongation and alignment as indi- cators of MIB. Cell elongation and alignment are reliable and useful indicators of MIB because they always accom- pany MIB and presence of elongated, aligned cells was always associated with the bipolar, mediolaterally directed protrusive activity that defines MIB (Shih and Keller, 1992a). We distinguish between MIB and mediolateral intercalaton itself because MIB always produces elongation and alignment but not necessarily mediolateral intercala- tion. Patterns of MIB will be examined in explants that do extend, in which case MIB produces elongation, alignment and intercalation, and in those that do not extend, in which case MIB produces elongation and alignment, but little or no effective intercalation (see Shih and Keller, 1992a).

Most or all of the dorsal mesodermal cells divide in the first part of gastrulation (Shih and Keller, 1992a). Then MIB begins during the early part of stage 10.5, two hours after the onset of gastrulation, and results in formation of the vegetal alignment zone (VgAZ). The cells at the vege- tal end of the explants begin to elongate, increasing their length/width ratio (LWR) and aligning parallel to one another in an arc lying 3 to 7 cells from the vegetal end of the explant (Fig. 2; also see Fig. 2, Shih and Keller, 1992a). The lateral ends of the bilateral arcs lie about 15 degrees on each side of the dorsal midline. The elongated cells within the arcs are aligned perpendicular to the prospective anterior-posterior axis of the mesoderm (Fig. 2). The cell elongation and alignment that define the VgAZ develop progressively, beginning laterally and proceeding medially (Fig. 2, stage 10.5-11). Through stage 10 the mean LWR of cells in the prospective VgAZ is about 1.48 (Shih and Keller, 1992a). In early stage 10.5, the average LWR of cells in lateral sectors of the VgAZ rises above that of medial sectors. This trend continues into stage 10.75 (Fig. 3) when the LWR of the medial sector 5 differs signifi- cantly from that in lateral sectors 1 (P<0.002) and 9 (P<0.005). After another hour (stage 11+), the mean LWR of medial cells approaches that of lateral cells (Fig. 3) and medial and lateral sectors may or may not show significant differences. For example, at stage 11, positions 6 and 11 differ significantly (P<0.033) while positions 1 through 9 have a greater than 50% probability of being the same data set (average P>0.5). The lateral-to-medial progression is due to local changes in cell shape rather than movement and sorting of elongated cells. Local mediolateral interca- lation occurs, but no long-range movements and sorting of cells by shape occurs. Slight left-right asymmetries in both directions are typical of this and other parameters of meso- dermal development.

Ignoring the medial-lateral differences, the average LWRs of all VgAZ cells is 1.8 at stage 10.5, 2.0 at stage 10.75 and 2.3 at stage 11+. These differ significantly from one another in either paired or unpaired Student’s analysis (P<0.01 for stage 10.5 versus 10.75, and P<0.003 for stage 10.75 versus 11+). By stage 12, the mean LWR of the VgAZ has reached 3.2 (Shih and Keller, 1992a). By stage 11+, the notochordal-somitic mesodermal boundaries form within the VgAZ, as described below, thus separating the lateral somitic from the medial notochordal region.

The VgAZ is defined as an arc of cells with noticeably larger LWRs at stage 11 or 11+. Development of the VgAZ is described by measuring the LWRs in mediolateral sec- tors of the same cell population at earlier stages (Figs 2, 3). The LWR of the definitive VgAZ at stage 11+ (above), is significantly different (P<0.001) from the mean LWR (1.3±0.04) of the population immediate animal to the VgAZ. If the boundaries of the VgAZ had been chosen too broadly, cells with low LWRs would have been included and the difference between cells inside and outside the VgAZ would have been masked. If the boundaries had been chosen too conservatively, elongated VgAZ cells would have been included in the external region, again masking the difference between it and the VgAZ.

The VgAZ expands animally in the prospective posterior direction (see Fig. 1), growing from 2-6 cells in animal- vegetal extent to 4-12 cells (depending on the amount of extension) when the notochordal-somitic mesoderm boundary forms at stage 11+ to 12. In explants that do not con- verge and extend, this lengthening is due solely to expression of mediolateral intercalation behavior in more posterior cells. In explants that do converge and extend, the VgAZ also lengthens by mediolateral intercalation and thus extension of its original population.

The notochordal-somitic boundaries first appear as distinct, optically dark regions (under epi-illumination) between no more than two or three prospective notochord and somite cells lying within the VgAZ during stage 11 (Fig. 4, 0 and 2 hours). These initial boundaries are very short and stable, and form at the same sites within the VgAZ in all explants. In explants that converge and extend very little or not at all, the boundaries lengthen primarily by arcing animally and laterally (Fig. 4, 2.2-6.2 hours), along the anterior-pos- terior axis of the notochord, tracing out the notochord in a shape predicted by the fate map (Keller, 1976). When con- vergence and extension occur, mediolateral cell intercalation reduces the width of the notochord. In the embryo, this is extreme, reducing the notochord to less than two cells in width and greatly increasing its length (see Keller et al., 1989a; Wilson et al., 1989). The notochordal-somitic meso- dermal boundaries also extend a short distance anteriorly, through the rest of the VgAZ. With the exceptions noted below, cells do not cross these boundaries.

Two distinct cell populations are involved in formation of the notochordal-somitic boundaries. The initial, anterior boundaries form in the VgAZ among cells that are already expressing MIB and intercalating (see Shih and Keller, 1992a). In this region, tongues of notochordal and somitic cells often extend into the opposite territory and are with- drawn as development proceeds (note the tongue of noto- chord, white pointer, Fig. 4, 3-6.5 hours). In contrast, the boundaries in the posterior region form among cells not yet expressing MIB, and these tongues do not occur. In either case prospective notochordal and somitic cells are not mixed extensively when traced back to stage 10.5, prior to boundary formation (Fig. 5). Although the notochord pop- ulation itself is fairly contiguous, the notochord cells that form the initial boundary at stage 11.5-12 do not form a coherent population when traced back to the pre-boundary stage 10.5 (Fig. 6). Thus within the prospective notochord tissue, there does not appear to be a contiguous, special- ized population of cells that will form the initial boundary. Ultimately, all notochord cells become boundary cells by first intercalating between one another and finally interca- lating between those constituting the initial boundary (Keller et al., 1989a).

Individual cells cross the notochord-somite boundary in both directions, although this occurs infrequently. Only two cells crossed from the notochord into the somitic mesoderm in five recordings, each spanning several hours with hun- dreds of cells in view. For example (Fig. 7), the boundary (small white pointer, 0 min) formed three cells lateral to a cell (large white pointer, 0 min), placing it within the noto- chord, but the cell subsequently migrated laterally (20 min), crossed the boundary, and took up permanent residence in the somitic mesoderm (40 min). We observed only one cell moving from the somitic territory into the notochord, and its behavior was similar to that of cells moving in the other direction.

After formation of the VgAZ and the notochordal-somitic mesodermal boundary, mediolateral intercalation behavior is expressed in a vegetal-to-animal (prospective anterior-to- posterior) and a lateral-to-medial progression in the noto- chord (see Fig. 1, for axial directions). This progression is reflected in the elongation and alignment of cells posteri- orly along the notochordal-somitic mesodermal boundary and inward (medially) from the boundary in explants that converge and extend (Figs 8, 9). Mean LWRs were deter- mined for mediolateral and anterior-posterior sectors by morphometrics (Fig. 10A,B). The mean LWR of the lateral zone (no. 1) increases first, followed by the mean LWRs of the lateral/medial (no. 2) and the medial (no. 3) sectors (Fig. 10A). The three mediolateral zones do not differ sig- nificantly from one another at stage 12 (comparing: lat with lat/med P<0.407, lat/med with med P<0.807, lat with med P<0.238). At stage 13, mean LWRs are higher in lateral sectors than those in medial sectors. The boundary cell pop- ulation has a significantly higher mean LWR than the two medial cell populations (P<0.003), while the lateral-medial and medial populations do not differ significantly (P<0.511). By stage 14 (for simplicity, data for this stage is not plotted in Fig. 10A) and continuing to stage 16, the cell LWRs of these three cell populations no longer differ significantly (P<0.26 to P<0.756). However, the cell LWR values of the three time points, in Students t-value analy- sis both as paired and unpaired groups, differ significantly from one another (stage 13/stage 14, P<0.019; stage 14/stage 16, P<0.008; and stage 13/stage 16, P<0.000; stage 12 in comparison with stages 13, 14, and 16 P<0.000).

In the vegetal-animal (prospective anterior-posterior) direction, the mean LWR is initially low everywhere except in the vegetal end of the notochord, where cells have already participated in VgAZ formation (Fig. 10B, stage 12). Thereafter, cell elongation proceeds animally or posteriorly in the notochord (stage 12-16, Fig. 10B). The mean LWR rises first in mid-body sectors (stage 13-14, Fig. 10B) and then in posterior sectors (stage 16, Fig. 10B). We com- pared mean LWRs of anterior-posterior zones in pairs through the four stages and found that the transition from unelongated to elongated cells moves animally (posteriorly), as revealed by significant differences between mean LWRs. At stage 12, the LWR values makes a sharp transition between zones 2 and 4 (comparing: 1 with 2,P<0.117; 4 with 6, P<0.117; 1 with 4, P<0.000). Subsequently, the transition moves animally, between zones 6 and 8 at stage 13 and between 8 and 10 by stage 14. At stage 16, LWR values had risen throughout, and zones 2 through 10 no longer differed significantly from one another (minimum P<0.27) although position 1, the most vegetal, still differ significantly from all others. These analyses show that the cells in the notochord elongate and align progressively from the vegetal end animalward (from anterior-to- posterior) and at any animal-vegetal level, from lateral-to- medial.

Boundary cells in the notochord elongate perpendicular to the notochordal-somitic boundary (Figs 4, 8 and 9). Medial cells are elongated and aligned parallel to arcs across the width of the notochord, thus forming a curved pattern in which all cells are both parallel to the mediolateral axis and perpendicular to the anterior-posterior axis of the notochord (Figs 8, and 9; refer to Fig. 1 for axes). This is demonstrated by the wider notochords in explants show- ing little convergence and extension, as discussed below (Figs 13, 14).

Low light fluorescence microscopy of FDX-labelled cells reveals that the cell elongation and alignment observed here, and used as an indicator of mediolateral intercalation behavior, results from the medially and laterally directed bipolar protrusive activity observed previously (Shih et al., 1992a). Cells abutting the notochordal-somitic boundary show a modified mediolateral intercalation behavior. The ends of these cells lying at the notochordal-somitic bound- ary or those entering this boundary become quiescent, as reported earlier (Keller et al., 1989a). Time-lapse recording of cells during boundary formation reveals that protrusive activity at the notochordal-somitic boundary differs, depending on whether the boundary forms before or after the cells express mediolateral intercalation behavior. The boundary-facing ends of the cells bleb rapidly when the boundary forms within the VgAZ, or when internal noto- chordal or somitic cells intercalate into a boundary that has already formed (triangle, Fig. 11, 0 and 45 min), whereas their opposite ends (the medial ends of the notochord cells and lateral ends of the somitic cells) make large filiform and lamelliform protrusions (arrows, Fig. 11, 0-135 min). Blebbing at the boundary is soon followed by quiescence (note stable boundary ends of labelled cells indicated by white bars, Fig. 11, 90-225 min), which is associated with formation of stable filiform protrusions within the plane of the boundary (see Keller et al., 1989a). The second type of protrusive activity occurs when the boundaries form among rounded cells not yet expressing mediolateral intercalation behavior, in which case, the cells do not bleb during or after boundary formation. Instead, the boundary-facing ends of these cells remain quiescent and stabilize the boundary sur- face while their opposite ends show the polarized fili- form/lamelliform protrusive activity associated with inter- calation. In both cases, protrusive activity of the type used in cell translocation is suppressed at the boundary, and thus most but not all cells stay at the boundary. In a 2 hour period beginning in stage 13 and ending in stage 14, the total cell number making up a segment of boundary increased from 24 cells to 37 cells, an increase of 54%, for an intercalation rate of 27% per hour. A low rate (2% per hour) of cells leave the boundary.

The vacuolation of notochord cells in shaved explants also progresses in a vegetal-to-animal and a lateral-to-medial order (Fig. 12). This is particularly evident in explants that have not been allowed to converge and extend, as described below. Thus the histological differentiation of notochord cells follows the preceding morphogenetic progression.

Normally the progression of MIB and vacuolation in the notochord is not easily recognized because convergence compresses the mediolateral axis. Convergence and exten- sion can be reduced or stopped with increased coverslip pressure, presumably by increasing friction. Cells in these explants show all the protrusive activity and cell shape changes seen in normal, converging and extending explants, except that mediolateral intercalation is minimal and the elongation of the cells is exaggerated (Shih and Keller, 1992a). Tissues differentiate in place in these explants, without benefit of convergence and extension movements.

Under these conditions, the notochordal-somitic bound- aries form approximately in the positions predicted by the fate map (Keller, 1976), and thus the boundaries of the two tissue territories appear to be normal, or nearly so. How- ever, the posterior, medial tissue in the notochord territory does not differentiate, a phenomenon that also occurs in the somitic mesoderm (see below). The notochord remains very wide and the lateral and anterior cells elongate and align (white triangles, Fig. 13A), whereas those in the corre- sponding medial, posterior region do not (large black pointer, Fig. 13A). Cell elongation and alignment do occur in the original, narrow VgAZ (black arrows, Fig. 13A). Subsequently, in the tailbud stages, cells vacuolate in the lateral regions, next to the notochord boundary (white point- ers, Fig. 13B), and across the original VgAZ (large point- ers, Fig. 13B) but not in the posterior, medial region of the notochord (large pointers, Fig. 13B).

The progression of mediolateral intercalation behavior and tissue differentiation in the somitic mesoderm parallels their progression in the notochord, proceeding from anterior to posterior and from lateral to medial (Fig. 14A). The ante- rior-posterior axis of the somitic mesoderm forms an arc,which originates just lateral to the anterior notochord and passes laterally, and ventrally (see Fig. 1 and Keller et al., 1989a). The lateral edge of the somitic mesoderm lies next to the vegetal boundary of the explant and medial edge is next to the notochord (see Fig. 1 and Keller et al., 1991a).

Cell elongation, indicating expression of MIB, occurs first in the prospective anterior somitic mesoderm (no. 2, Fig. 14A) and then the prospective anterior notochordal mesoderm (no. 1, Fig. 14A) during formation of the VgAZ. Mediolateral intercalation behavior progresses posteriorly along the lateral margin of the somitic mesoderm (no. 2, no. 3, Fig. 14A), and from there, spreads medially toward the notochord. This medial progression is illustrated by the greater expression of mediolateral intercalation behavior laterally (no. 3) than in the corresponding medial region (no. 4, Fig. 14A). The cells of the next arc, representing a slightly more posterior level, show no elongation or align- ment (no. 5, Fig. 14A). Cell outlines and mean LWRs show that MIB is initiated at the anterior, lateral margin and pro- gresses posteriorly along the lateral margin (Fig. 15). At given anterior-posterior level, cell elongation and alignment progress in arcs, beginning laterally and proceeding medi- ally, toward the notochord (Fig. 15).

Segmentation of somitic mesoderm becomes visible in shaved explants at stage 16. The anterior intersomitic fissures form rapidly and follow the preceding pattern of MIB. They progress from the prospective anterior to posterior (no. 1-no. 5, Fig. 14B) and from the lateral margin of the somite (the vegetal edge of the explant) towards the prospective medial edge (the notochord-somite boundary; white arrows, Fig. 14B). Differentiation of somitic meso- derm, as seen by staining with the 12-101 antibody, shows the same anterior-posterior and lateral-medial progression (Fig. 16). As in the notochord, when little or no conver- gence and extension occurs, the mediolateral axis of the somitic mesoderm remains relatively long. The markers of somitic differentiation, cell elongation and alignment, inter- somitic furrows, and 12-101 staining of the somitic meso- derm, fail to proceed medially far enough to reach the noto- chord except in the narrow anterior region.

In previous work, we described the bipolar, mediolaterally elongated cells, bearing large medial and lateral protrusions, in the postinvolution dorsal mesoderm (Keller, 1984) and just outside the dorsal lip of the blastopore (Keller et al., 1989b). We speculated that this morphology was involved in mediolateral intercalation, a speculation subsequently supported by indirect evidence that mediolateral intercala- tion of mesodermal cells is primarily a postinvolution behavior (Keller and Tibbetts, 1989; Wilson and Keller, 1991). To reaffirm where the characteristic elongate,aligned cells reside during gastrulation, we measured the LWRs in SEMs of selected regions. Animal cap ectoder- mal cells and preinvolution mesodermal cells are approxi- mately isodiametric (Fig. 17, column 1) and remain so throughout gastrulation (data not shown; see Keller, 1980). Cells just beginning to involute over the blastopore lip at stage 10.5 have a slightly greater length-width ratio (Fig. 17, column 2). In subsequent stages, the deep cells imme- diately beneath the endodermal epithelium and the deeper mesodermal cells at the blastoporal lip and in the postin- volution mesoderm become progressively more elongate, with LWRs averaging from 2.5 to just over 3.0 by stage 11.5 (Fig. 17, columns 3-7).

MIB drives convergence and extension but, in turn, is mod- ified by convergence and extension. Thus we will summa- rize the expression of MIB in explants that converge and extend and those that do not. The pattern of expression of MIB in explants that are not allowed to converge and extend is shown diagramatically (Fig. 18A). (1) After the cells divide once between stage 10 and 10.5, MIB begins later- ally and progresses medially to form the arc-shaped VgAZ, which includes the prospective anterior notochordal and somitic tissue at the vegetal end of the explant (Fig. 18A, stage 10.5-10.75). In these explants, MIB results in cell elongation and alignment but no intercalation of the type that produces convergence and extension. (2) The noto- chordal-somite boundaries are initiated in the lateral parts of the VgAZ at the midgastrula stage (Fig. 18, stage 11). (3) The notochordal-somitic boundaries arc animally and laterally in the prospective posterior direction during gas- trulation, outlining the notochord as it appears in the fate map (Fig. 18A, stage 11+) (cf. Keller, 1976). (4) Noto- chordal cell elongation and alignment proceed posteriorly from the VgAZ, along the notochord-somite boundaries (open, curved arrows, Fig. 18A, stage 11+). (5) At the same time, somitic cell elongation and alignment proceed later- ally around the marginal zone, in the prospective posterior direction, from the VgAZ, along the lateral boundary of the somitic mesoderm (open, straight arrows, Fig. 18A, stage 11+). (6) In both somitic and notochordal mesoderm, cell elongation and alignment also proceed medially from their respective lateral boundaries, toward the midline in the case of the notochord, and toward the notochord in the case of the somitic mesoderm (solid arrows, Fig. 18A, stage 11+).(7) Vacuolation of notochord cells proceeds medially from the boundaries of the notochord and progresses from ante- rior to posterior (not illustrated). (8) Segmentation proceeds from anterior to posterior within the somitic mesoderm, the intersomitic furrows beginning at the lateral margins and proceeding medially (not illustrated). When convergence and extension are not allowed to occur, tissue at the pos- terior medial regions of both notochordal and somitic ter- ritories (asterisks, Fig. 18A, stage 11+) do not undergo cell elongation and alignment, and they do not differentiate.

Fig. 18. Schematic diagrams show the progression of mediolateral intercalation behavior across the shaved explants of the marginal zone of Xenopus, which were not (A) and were (B) allowed to converge and extend. The expression of MIB is indicated by elongated, aligned, red outlines of cells. The notochordal-somitic boundary (dashed arrows), the notochordal (n), and the somitic (s) territories are shown. Within the explants, the open arrows indicate the progress of MIB in the anterior-posterior direction and the solid arrows indicate progress of MIB in the lateral-medial direction (see the text). Outside the explant, the solid arrows indicate convergence and extension movements. A vegetal view of the midgastrula (C) shows MIB expression progressing posteriorly just outside the blastoporal lip (open arrows), where convergence or constrictive forces tend to roll the lip downward and over the lip (solid arrows) (see text for a description). The lines along which convergence or constrictive forces are generated by the progression of MIB are represented by red arcs (D). The explant is depicted as if the vegetal endoderm (green, D) was not removed; each arc originates laterally and is normally anchored laterally near the vegetal endoderm. The role of the posterior progression of constrictive forces in generating involution is illustrated schematically by the shortening of the red arcs (E). Note that the separated arcs are used only for illustration; normally the progression is spatially continuous rather than in the form of discrete, separated arcs. Also, the mesoderm is normally covered by the suprablastoporal endodermal epithelium. Only anteriorly do the notochordal and somitic components of the arcs stay in register; posteriorly the notochord shears posteriorly along the medial boundary of the somitic mesoderm (see Keller et al., 1991b, Fig. 5, for an explantation of this geometry), but the tissues maintain mechanical continuity at their boundaries.

When less mechanical resistance is imposed, MIB results in mediolateral intercalation, and convergence and exten- sion occur (Fig. 18B). In this case, MIB and tissue differ- entiation occur to the midline of the notochord and to the medial region of the somitic mesoderm (Fig. 18B, stage 10.75-11+). Our hypothesis is that MIB is organised medi- ally from the lateral boundaries of both the somitic and notochordal mesoderm, these being the sources of the orga- nizing signals. Thus in converging explants the signals organizing MIB (and tissue differentiation) are made more effective as MIB brings medial prospective notochordal and somitic cells within their range, whereas in nonconverging explants these cells never fall within range of the signals (cf. Fig. 18A,B). The lateral-to-medial progression of seg- mental furrows is also seen in the embryo and in open- faced explants (see Wilson et al., 1989; Wilson and Keller, 1991). The arc-shaped pattern of posterior progression encompasses the entire notochord region and the first six or seven somites relatively fast and early, making the ante- rior-posterior timing difference less obvious, but the pos- terior progression continues for a longer time in the pos- terior somitic mesoderm, as this mesoderm moves around the blastopore and comes to lie along both sides of the noto- chord during neurulation (see Wilson et al., 1989; Keller et al., 1989a, 1991a).

The progress of mesodermal cell behavior and tissue differ- entiation from lateral to medial and from anterior to pos- terior in both somitic and notochordal tissue may reflect the order in which the cells were instructed to adopt these phe- notypes, and thus may reflect the geometry of the signals that instruct them. This turns our attention to the lateral boundaries of the somitic and notochordal mesoderm as possible sources of organizing signals. Both tissues develop medially from the lateral boundaries, and medial, posterior cells of both notochordal and somitic territories do not dif- ferentiate in the absence of convergence. This suggests that morphogenetic and histological differentiation requires con- vergence of the signal-generating boundaries towards the medial cells in order for these cells to respond. If progress of the cell behaviors and differentiation reflects the order of commitment, then the posterior, medial cells within both notochordal and somitic territories should be the last com- mitted to a behavior or tissue phenotype. Likewise, if orga- nizing signals are proceeding inward from the boundaries, competent cells grafted into notochordal or somitic fields should express MIB and tissue differentiation in a pattern reflecting this progresion. We are testing this by relocating cells within shaved explants and monitoring their behavior and differentiation. Animal cap cells grafted into central notochord region at the early gastrula adopt MIB and form notochord from anterior to posterior (C. Domingo, personal communication; see Keller et al., 1992c), as if they were exposed to a notochord-organizing signal in this same order.

It is not clear how the signals from vegetal endoderm (the sub-blastoporal endoderm) that function in the induction and patterning of mesoderm (Nieuwkoop, 1969a,b; Sudarwati and Nieuwkoop, 1971) generate the highly organized patterns revealed above, nor is it clear how these pat- terns are related to the three-signal model of mesoderm organization (Dale and Slack, 1987) or to the current think- ing about how growth factors possibly involved in mesoderm induction generate fate maps (see Green and Smith, 1992; Cooke et al. 1987; Slack et al., 1987; Smith et al., 1988; Kimelman et al., 1988; Cooke, 1991). The progres- sion of cell behaviors described here may reflect the final sequence of events in determining cell behavior and pheno- type rather than earlier steps in organizer differentiation (see Gerhart et al., 1991; Cooke, 1991). The idea that com- mitment of cells to specific behaviors and tissue identities may still be going on progressively in the dorsal mesoderm during gastrulation is consistent with the organizer activity of this region when grafted to other sites (Spemann, 1938; Gimlich and Cooke, 1983). Moreover, it is perhaps no sur- prise that the organizer patterns cell behaviors in a highly ordered progression, since such a progression may be essen- tial for proper involution movements during gastrulation (see below).

Boundary formation seems to be superimposed on the mesodermal cells independent of their motility. The ante- rior notochordal-somitic boundaries form among cells that are in the midst of MIB, whereas the posterior boundaries form among cells not yet showing this behavior. These two cell populations respond differently to the boundaries. Ante- rior cells already undergoing mediolateral intercalation bleb wildly at their site of contact with the boundary, both when the boundary forms among them and when they contact the boundary after it forms; only later are protrusions stabilized on the boundary, making the cells monopolar (see Keller et al., 1989a; Weliky et al., 1992). In contrast, when the boundary forms among the posterior cells not yet express- ing MIB, they do not bleb. When they do begin MIB, they immediately stabilize their protrusions on the boundary. Moreover, the boundaries appear to be directional sources of signals, resulting in different behaviors and states of cell commitment on their notochordal and somitic sides. Cells on the notochordal side express mediolateral intercalation behavior and are perhaps committed to notochordal fates, while cells at the same level but on the somitic side will not express this behavior for some time and are perhaps pluripotent and uncommitted. Finally, the notochordal- somitic boundaries form in about the right positions, whether or not convergence and extension occur, and thus, unlike MIB, they appear to be relatively independent of changes in tissue dimensions.

The distribution of elongated cells in embryos, which reflect expression of MIB, show that MIB is initiated in the postin- volution position and progresses posteriorly, producing arcs of constriction that press on the inner surface of the IMZ, and act to roll the IMZ over the blastoporal lip (Fig. 18C). As noted here (Fig. 17) and some time ago (Keller, 1984; Keller et al., 1989b), cells of the involuted dorsal meso- derm and those cells just outside the blastoporal lip, which are about to undergo involution, are elongated mediolater- ally and aligned parallel to one another, to the mediolateral axis, and to the margin of the blastopore from stage 10.5 onward. The biomechanical consequence of progressive posteriad expression of MIB can be visualized as the pro- gressive shortening of a series of arcs that span the the IMZ and end laterally where the somitic mesoderm is anchored next to the vegetal endoderm; Fig. 18D shows an extend- ing explant as if the vegetal endoderm were still present. In the explant, where these arcs are no longer anchored, their shortening results in pulling the lateral margins toward the midline, as shown in Fig. 18B. In the embryo, where these arcs are anchored near the vegetal endoderm, their progressive shortening by MIB in a progressive anterior-to- posterior order rolls the IMZ over the lip (involution) and at the same time extends the IMZ in the postinvolution posi- tion (Fig. 18E). Thus convergence, extension and involu- tion all result from the mechanics of shortening of these arcs in the spherical geometry of the embryo (Fig. 18E). This explains why convergence and extension of the invo- luted IMZ (see Keller et al., 1985a,b; Keller and Jansa, 1992). The continuity of these arcs is essential; if they are interrupted in the dorsolateral regions, involution fails, the notochord extends outward, and the somitic mesoderm extends and differentiates along both sides of an open blastopore, as shown in Fig. 18C (see Schechtman, 1942; Keller, 1981, 1984). Note that mediolateral intercalation in Xenopus involves shortening of mechanically continuous arcs by local cell traction, which pulls all dorsal tissue closer to the lateral anchorage points near the vegetal endo- derm rather than directed cell migration toward the dorsal midline. Thus ‘convergence’ is a property with an axial (mediolateral) orientation, but there is no evidence at this time that it has a direction. Convergence is ‘dorsal’ only in that it occurs in dorsal tissue; it is not directed dorsally, based on the evidence uncovered thus far.

Since mediolateral intercalation is underway when the ante- rior notochordal-somitic boundaries form, cells are moving across the path of the boundary as it forms. Nevertheless, the prospective notochord cell population is contiguous, for the most part, at the beginning of gastrulation. Thus differ- entiation of cell affinities (Holtfreter, 1939), or some other mechanism that blocks cell mixing, would have to follow very quickly on the heels of commitment of cells to one tissue or another. Otherwise, the incipient or ongoing medi- olateral intercalation would move cells out of range of their homotyic neighbors. Alternatively, some mixing might occur but misplaced cells would be called back across the boundary to their own territory. The rarity of cells cross- ing the boundary suggests that this is not a major factor. There is no contiguous subset of specialized boundary cells; prospective initial boundary cells are distributed through- out the notochord in the early gastrula in no definable, con- sistent pattern. Any cell entering the boundary region changes its behavior to correspond to that of other bound- ary cells, and eventually all notochord cells become bound- ary cells (Keller et al., 1989a). This suggests that bound- ary formation is the response of cells to a local condition, generated by a subset of cells that happen to be in the boundary-forming region at a specific time. This local con- dition then modifies the behavior of internal notochord cells as they intercalate into the boundary, transforming the bipo- lar mediolateral intercalation behavior (Shih and Keller, 1992a) into a monopolar one (Keller et al., 1989a).

This work was supported by NSF grant DCB89052 and NIH grant HD25594. We thank J. P. Trinkaus for his inspiration, his encouragement, and his pioneering work on the study of motility in embryos. We also thank Paul Wilson for his pointed criticism and support. Finally, we thank Jessica Bolker, Dr John Gerhart, Mary Constance Lane, Nan Ho, Susan Purcell, and Amy Sater for their insightful comments and suggestions.