The cellular basis of epiboly: An SEM study of deep-cell rearrangement during gastrulation in Xenopus laevis

Embryologists have always been fascinated with the process of amphibian gastrulation and have directed much effort toward discovering its secrets. Roux (1894, 1896) ascribed the global rearrangements of gastrulation to the behavior of individual cells which he had isolated and observed in a simple culture medium. From tissue recombination experiments, Holtfreter (1939) concluded that the differentiation of a hierarchy of cell type specific affinities governs tissue behavior and thus the movements of gastrulation. On the basis of extensive observation of the structure of the gastrula and of the behavior of individual cells and groups of cells in vitro. Holtfreter (1943 a, b, 1944) set forth some ideas on cellular behavior and the mechanics of gastrulation. Schechtman (1942) and others (see Spemann, 1938) did surgical rearrangement and explantation experiments which showed both a regional autonomy and a cooperation on the part of various sectors of the gastrula. However, these studies did not produce a comprehensive description of cellular behavior in vivo or a mechanism relating specific patterns of cell behavior to the generation of forces appropriate to produce the observed distortion of the gastrula. The opacity of the amphibian embryo has prevented the direct observation in vivo of the cellular activities that actually occur during gastrulation, and therefore they remain unknown.

To establish the cellular mechanisms of gastrulation, it is necessary to answer several questions. What are the paths of cell movement? What cellular activities bring about these movements? How does cellular behavior generate the necessary mechanical forces? How is this cellular behavior controlled?

A series of studies on gastrulation in the African clawed frog, Xenopus laevis, has been directed at answering these questions. The paths of cell movements were mapped by vital dye marking of both the superficial and the deep layers of the embryo. Movements in both layers are similar in their major features to those found in other amphibians (compare Vogt, 1929, Pasteels, 1942). The animal region undergoes a nearly uniform expansion (epiboly). The dorsal marginal zone (DMZ) spreads rapidly toward the blastopore (extension), and concurrently it narrows (convergence). The lateral marginal zone (LMZ) and the ventral marginal zone (VMZ) also spread toward the blastopore but to a lesser degree. Both epiboly and extension consist of an increase in area; in the latter case the increase is anisotropic and directed toward the blastopore (see Keller, 1978). However, Xenopus differs fundamentally in that mesoderm is not found on the surface of the embryo but lies entirely in the deep marginal zone and is covered by a monolayer of suprablastoporal endodermal cells (see Nieuwkoop & Florshütz, 1950; Keller, 1975, 1976). The superficial layer remains intact during gastrulation, which is in contrast to the results of work on other amphibians (Vogt, 1929; but see Løvtrup, 1975). A quantitative analysis of the apical area, number, and shape of superficial cells by time-lapse cinemicrography of living Xenopus gastrulae (Keller, 1978) showed that these cells spread, divide, and undergo rearrangements and a temporary change in shape which accounts for the change in area and for the distortion of the superficial layer during epiboly, extension and convergence.

The activities of the deep cells could not be seen in the intact embryo, but a scanning electron microscopic analysis of fractured gastrulae suggests what their behavior might be. Regions of the gastrula undergoing epiboly and extension become thinner and show a decrease in the number of layers of deep cells (see Keller & Schoenwolf, 1977; Keller, unpublished data). There is no evidence of cytolysis or cell death. Indeed, this is unlikely since the total Xenopus cellular volume increases slightly during gastrulation (Tuft, 1965). Time-lapse cinemicrography shows that deep cells do not disappear from the deep layer by moving outward into the superficial layer. They cannot migrate more deeply into the interior because they are bounded internally by the interface, which is a clearly defined discontinuity with the underlying involuted cells. Vital dye mapping shows that a marked sector of the deep region expands as a contiguous mass (Keller, 1976). These facts suggest that deep cells undergo a local migration inward or outward, between one another, along radii on the embryo, and thus transform several layers of deep cells into one layer of greater area. Such a process of radial interdigitation would account for the spreading of the deep region during extension of the marginal zone and epiboly of the blastocoel roof (Fig. 1 ; see the Discussion and Fig. 13, Keller, 1978).

In the present study, measurements of indices which quantitatively describe the positions, the shapes, and the arrangements of the deep cells show that radial interdigitation occurs and that it and the subsequent flattening of a single layer of deep cells provide the increased area necessary for epiboly and extension of the deep layer. Furthermore, these data suggest a sequence of changes in deepcell shape and contact as a cellular mechanism of radial interdigitation. This paper and an earlier one (Keller, 1978) provide a quantitative description of cellular behavior of both the superficial and the deep cells during epiboly. From these data, models relating local cellular behavior, a global system of mechanical stresses, and the spreading movements of the deep and superficial layers were generated.

Xenopus embryos were obtained by standard mating procedures (see Keller, 1975) and dejellied as described previously (Keller & Schoen wolf, 1977) or by cysteine hydrochloride treatment (1·75 g in 50 ml distilled water with 0·3 g Tris buffer, adjusted to pH 7-8 with NaOH). They were then washed in 10% Steinberg or Niu-Twitty solution, staged according to Nieuwkoop & Faber (1967), and fixed in 2·5 % glutaraldehyde in 0-1 M sodium cacodylate (pH 7·4−7·6) at room temperature. They were then fractured with a blunt knife as described previously (Keller & Schoenwolf, 1977), dehydrated in ethanol, dried (liquid CO₂ critical-point-drying), mounted on stubs with silver paste, coated with gold-palladium (60:40) in a Polaron sputter-coater or a Denton vacuum evaporator, and photographed on Polaroid Type 55 film with an Etec Auto Scan electron microscope. Magnification of electron micrographs, vertically and horizontally, was determined from micrographs of a grid, the dimensions of which were measured with a light microscope fitted with a calibrated filar micrometer.

The regions to be discussed in this paper are found in the outer shell of the embryo (lightly shaded regions, Fig. 2). As this outer shell expands and moves vegetally, its lower margin is involuted. It is continuous with the involuted material (darkly shaded regions, Fig. 2) at the blastoporal lip where involution occurs, and is separated from the involuted material in all other regions by a clearly defined discontinuity, the interface (see Fig. 2). The outer shell consists of a monolayered epithelium of superficial cells and a deep region of one or more layers of non-epithelial deep cells (Fig. 3). There are two types of deep cells. Inner deep cells are those deep cells that bound the blastocoel or that form the interface with the involuted mesoderm (see Figs. 2 and 3). Inter deep cells are all those deep cells that do not bound the blastocoel or the interface (Fig. 3).

Indices that measure several relevant characteristics of superficial cells, inter deep cells and inner deep cells were defined (Fig. 3) and measured in scanning electron micrographs of known magnification (370-420). First, the maximum extent of the superficial cells, the inter deep region, and the inner deep cells in the direction perpendicular to the surface (along radii) of the embryo were measured and the positions of these points with respect to the surface of the gastrula were recorded, averaged, and the mean plotted graphically as the layer boundary (Fig. 3). Only cells for which the inner and outer boundaries could be clearly determined were included.. The superficial layer and the inner deep layer are, by definition, monolayers; in contrast, the inner deep region may consist of one or several layers or tiers of cells. In the latter case only the inner-most or outer-most boundaries were recorded. Second, individual superficial cells were chosen and the shortest route was traced to the interior, jumping from cell-to-cell (Fig. 3). The mean number of cells passed through in all micrographs of a given region and stage was recorded as the layer index. Third, the number of inner deep cells was divided by the total number of deep cells to give the per cent inner deep cells. Fourth, the number of inner deep cells extending through the entire deep region and contacting the inner surfaces of the superficial cells was divided by the total number of inner deep cells to give the per cent inner-to-superficial contact. Fifth, the mean overlap of inter deep cells and inner deep cells was calculated as a percentage of the thickness of the entire deep region and designated the per cent interdigitation (Fig. 3). Sixth, the maximum extent of cells along the animal-vegetal (meridian) lines of the embryo was divided into their maximum extent along radii (normal to the surface) of the embryo to give a height-length (h/l) ratio. Seventh, the length (as defined above) of the inner apices of the inner deep cells was divided by their maximum extent along latitude lines of the embryo to give a length-width (l/w) ratio. Note that these ratios reflect both cell shape and orientation. Eighth, the inner apices of the inner deep cells were traced and their areas determined with a planimeter, corrected for magnification, averaged, and the mean apical area plotted graphically. The mean, the standard deviation, the standard error of the mean, and the frequency distribution were determined for nearly all parameters for every stage and region considered. Means (and in some cases errors) are presented graphically, and other data will be noted where they are important. Lastly, the mean volume of inner deep cells was estimated from their apical area and height at several stages of development. All measurements were of fixed and critical-point-dried material which shows some shrinkage (Keller & Schoenwolf, 1977).

The changes in cell morphology and arrangement occurring in the DMZ of the late blastula and throughout gastrulation are shown in a series of representative electron micrographs (Fig. 4) and by quantitative data (Figs. 5 and 6).

At stage 8 the DMZ is about 125 μm thick and consists of a superficial cell layer and an inner deep cell layer, each about 50 μm thick, and a slightly thicker region of one and a half layers of inter deep cells (Fig. 6). By stage 9, the DMZ as a whole has thinned slightly due to thinning of all three of its component regions. The superficial cells, the inner deep cells, and probably the inter deep cells as well, decrease in their linear dimensions following cell division. These cells can be viewed as nearly cuboidal or as cylinders (roughly) with the height equal to diameter, since their h/l ratios are about 1·0, and both the inner deep cells and the superficial cells have a l/w ratio of about 1·0 (see page 212 and Keller, 1978). Thus the one cycle of inner deep cell division taking place from stage 8 to 9 (see Fig. 18) should reduce the linear dimensions of these cells by 21 %, assuming, correctly (Tuft, 1965), that cell growth does not occur. The observed decrease in height is 20 %. Concurrently, the thickness of the superficial layer is reduced by 35%. Time-lapse cinemicrography (see Keller, 1978) a population average of just under two anticlinal cell divisions occur in this period, which would reduce the linear dimensions of these cells by about 36 %. As a result of cell division there are more deep cells of smaller size and thus the layer index has risen to 4·1, even though the DMZ has become thinner.

The events in the period from stage 9 to 10 appear to be in preparation for the onset of gastrulation (Fig. 4). The DMZ thickens by nearly 20 /tm, all of it due to an increased number of layers (layer index is 5·5). The mean cell size decreases again; approximations of the inner deep cell volume indicate that these cells have divided twice in this period (compare stages 8 and 10 in Fig. 4; see Fig. 18). The mean h/l ratio of inner and inter deep cells has risen to 1·3 and 1·4 (Fig. 6), and both ratios are significantly greater than that of the superficial layer (non-parametric Wilcoxon Rank Sum test; P < 0·01).

In summary, stages 8 and 9 appear to comprise a preparatory phase in which the number of cells and layers of cells increases as the mean cell volume decreases by division, and in which the DMZ reaches its maximum thickness.

The period from stage 10 to 10+ marks the beginning of gastrulation. It is characterized by a rapid thinning of the DMZ and a rapid decrease in the number of cell layers in the inter deep region (Fig. 5). Thus it has been designated the multiple-layer interdigitation phase. The layer index decreases from 5·5 to just above 3·0 (Fig. 6), which means that the inter deep region decreased from 3·5 cell layers to just one. The mean h/l ratios of inner and inter deep cells are significantly greater than 1·0 at stage 10 and continues to increase. For the first time, a few (about 8 %) of the inner deep cells pass through the entire deep region and contact the superficial cells, and the per cent interdigitation of inner and inter deep cells shows a parallel increase (Fig. 6). During these events the DMZ expands and extends rapidly toward the blastopore (see Keller, 1978).

Since stage 9, important changes have occurred in the protrusions of the cell bodies of deep cells. It will be useful to make a distinction between macroprotrusions and microprotrusions of the cell body. The definition of these terms assumes that all cell shapes are modifications of a spherical cell body, which is, in fact, usually polymorphic but approaches a spheroidal or cuboidal shape on the average, much like those of the DMZ at stage 8 (Fig. 4). A large extension that comprises a major part of the cell body will be called a macroprotrusion. A small extension of either the cell body or of a macroprotrusion is called a microprotrusion. These are nearly always attached distally to other cells (see Fig. 10d).

At stage 8, the deep cells are polymorphic but generally cuboidal and show only microprotrusions of the cell body, which serve as attachments to adjacent cells (see stage 8, Fig. 4). Many deep cells of stage 9 and nearly all of them at stage 10+ bear macroprotrusions which are attached distally to adjacent cells by microprotrusions, usually filiform in shape (arrows, stage 10+ and ⁠, Dig. 4). By stage ⁠, these macroprotrusions tend to be aligned along radii of the embryo and thus contribute to the increased h/l ratio (Fig. 6). Macroprotrusions may be thought of as transitional structures in a process of changing cell shape. For example, when a macroprotrusion reaches the size of the cell body, the cell would no longer be considered a spheroidal or cuboidal cell with a macroprotrusion, but an elongate or columnar cell. At stage 10 + a few of the deep cells, particularly the inner deep cells, are already columnar, and many are by stage ⁠.

The DMZ thins slowly by 15 % in the next period (stage 10 + to ⁠). This is not due to thinning of the individual cell layers but is accomplished by a slow and partial interdigitation of inter and inner deep cells which produces an 18 % per cent decrease in the thickness of the deep region (Fig. 5). The inter deep region, thins by a reduction in the mean number of cell layers from 1·2 to 0·8 (the ;layer index decreases from 3·2 to 2,). In spite of the overall thinning of the deep region, both the inner and inter deep cells actually increase in height relative to their diameter (increased h/l ratio’, Fig. 8). The per cent interdigitation rises to nearly 30 % by stage ⁠. Concurrently, there is an increase in the per cent inner-to-superficial cell contact (Fig. 6) such that 30% of the inner deep cells extend through the deep layer by stage ⁠. By stage ⁠, most deep cells, particularly the inner deep cells, are elongate, aligned along radii of the embryo, and bear numerous terminal and lateral microprotrusions (see stage ⁠, Fig. 4).

To summarize, from stage 10+ to stage there is a slow decrease in the thickness of the DMZ due to partial interdigitation of a layer of inter deep cells and a layer of inner deep cells. Concurrently, both cell types become progressively more elongate and columnar.

From stage to 11, the DMZ thins by about 20 μm (29 %). About 6 μm of this is due to thinning of the superficial layer, and the rest is due to the rapid completion of the interdigitation of inner and inter deep cells (Fig. 6). As a result, the layer index decreases from 2·8 to 2·2, 88 % of the deep cells are inner deep cells, and the inner-to-superficial cell contact reaches 90% by stage 11. Concurrently, the h/l ratio increases to a maximum of 2·2 (Fig. 6), and the deep region then consists of one layer of tall, columnar cells (stage 11, Fig. 4). This and the previous period can be considered the two-layer interdigitation phase.

From stage 11 to ⁠, these columnar inner deep cells undergo flattening and spreading (compare stages 11 and in Fig. 4), and as a result the DMZ decreases from 50 to 36 μm in thickness. The flattening is indicated by a decrease in the h/l ratio of the inner deep cells from 2·2 to 1·6 (Fig. 6) and by a corresponding increase in apical area (see Fig. 8). By stage the layer index has reached 2·0 and the inter deep cells no longer exist, since nearly all have become inner deep cells by interdigitation. From stage to ⁠, the DMZ thins by 4 μm, losing an average of about 2 μm in each layer (Fig. 5), probably because of continued flattening and spreading of cells, but this time in the superficial layer as well as the deep layer. Thus at the end of gastrulation, the DMZ consists of a superficial layer about 14 μm thick and a deep layer about 21 μm thick (Fig. 5). The period from stage 11 to ⁠, characterized by shortening and flattening of the columnar array of deep cells, is called the deep-cell spreading phase.

Extension and convergence in the superficial layer occurs in part by a temporary rise in l/w ratio of superficial cells (Keller, 1978). Perhaps deep cells undergo a similar l/w change. Such a change should be reflected in the l/w ratios of the inner apices of the inner deep cells of the DMZ, which are exposed by digging out all involuted material from gastrulae fractured midsagittally (Fig. 7a, b). They are closely packed, relatively flat, and have filiform protrusions extending across their neighbors (Fig. 7 c). On the basis of the data thus far, the hypothesis that the mean l/w of these apices is equal to one could not be rejected (P = 0·01). The same was true when the length and width axes were rotated 45° to their proper orientation. Thus these polymorphic apices are iso-diametric when considered as a population. This means that the increase in h/l ratio during interdigitation is not due to decrease in length as the cells become relatively wider; instead, their height increases at the equal expense of length and width. Similarly, when the h/l ratio decreases during shortening and spreading (stage 11 onward), the height is decreasing to the equal benefit of both length and width.

These facts suggest that inner deep cells behave as cylinders in which the height increases at the expense of apical area during the interdigitation phases, and decreases to the benefit of apical area during the deep-cell spreading phase. Indeed, if the expected apical areas of inner deep cells are calculated on the basis of mean change in cell volume (Figs. 8, 18) and corrected for the observed changes in h/l ratio (Fig. 6), they lie close to the values actually observed (Fig. 8).

An increased mean h/l ratio of inter and inner deep cells is associated with interdigitation and a decreased h/l ratio is associated with their subsequent spreading. But does the entire population or only part of it show these changes in h/l ratiol The frequency distribution of h/l ratios of inner deep cells (Fig. 9) is not obviously bimodal at any stage except perhaps stage 11, and thus it appears that no large fraction of the population fails to show change in h/l ratio. The increasing broadness of the distribution (Fig. 9) might be explained in several ways. Each cell may undergo a single cycle of h/l increase and decrease during gastrulation but with individual cells some degree out of phase with one another. Also, some cells may show a large excursion of the h/l ratio and others much less. Alternatively, individual cells may be passing rapidly through many such cycles, again, out of phase with one another.

Cells undergoing division contribute to the lower h/l clases. Inner deep cells undergo anticlinal divisions by rounding up (h/l of about TO), dividing to form dumb-bells (h/l of less than TO), and finally forming pairs of rounded daughter cells (see Fig. 22, Keller & Schoenwolf, 1977). A large number of these rounded cells and dumb-bells are seen at stage 11. The relatively high frequency of cells with h/l ratios of less than TO at stage 11 (Fig. 9) may represent this population of dividing cells but the change in cell size (see Fig. 4) suggests that the majority of cells have divided earlier, somewhere between Stage and 11.

During interdigitation deep cells usually bear macroprotrusions which are aligned along radii of the embryo (Fig. 4) and thus contribute to the increased h/l ratio. By stage ⁠, 51 % of deep cells have h/l ratios above 2·0 (Fig. 9) and these cells either have large, oriented macroprotrusions or they are elongated and cylindrical (Fig. 4). Many inter deep cells of the DMZ have a rounded cell body with a long, gently tapered macroprotrusion extending between adjacent cells in the direction of interdigitation (see No. 1, Fig. 10c, d). Others have shorter, more blunt macroprotrusions of the cell body, also insinuated between cells to the inside (see No. 2 in Fig. 10a, b and e). Macroprotrusions of inter deep cells lying adjacent to the superficial layer often extend inward to the interface and form a small apex there (see No. 3 in Fig. 10a, e). Microprotrusions of several types are found on macroprotrusions, as well as on the cell bodies, and these connect the cells to one another. There are terminal filiform (Fig. 10a, e), lateral filiform (Fig. 10a, c, d and e), terminal lamelliform (Fig. 10b), and lateral lamelliform (Fig. 10b) types of microprotrusions. Some of these microprotrusions may be retraction fibres. Superficial and deep cells are connected to one another by microprotrusions extending from both cell types (Fig. 10a, e). In rare cases, all in the blastocoel roof, a macroprotrusion of a superficial cell extends through the deep layer and forms an apex bounding the blastocoel (Fig. 10 f). The inner apices of the inner deep cells are connected by marginal filiform protrusions which extend, often a considerable distance, across the apices of adjacent cells (Fig. 7 c). These are found without noticeable variation throughout the DMZ, VMZ, ANE and BR regions of the gastrula.

The formation of protrusions oriented along radii of the embryo may reflect an active locomotion of cells between one another in this direction. Cells may actively migrate between one another by using a cyclical locomotory process involving the extension of a macroprotrusion, which attaches distally by microprotrusions, to adjacent cells, and shortening of the macroprotrusion, thus pulling the cell body toward it. There is no direct evidence for this, but preliminary evidence on the behaviour of cells in an in vitro model system suggests that active interdigitation does, in fact, occur. The entire outer shell of the gastrula (shaded region in Fig. 2) is removed and pinned with glass needles, inside up, to a bed of Permoplast in a culture dish of modified Steinbergs solution. Clumps of cells are then added to this surface and their behavior recorded with time-lapse cinemicrography, followed by fixation and scanning electron microscopy. Deep cells of aggregates from regions thought to be undergoing active interdigitation do, in fact, migrate between the native inner deep cells on which they are placed and form a smooth pavement of polygonal apices indistinguishable from those of the native cells (see Fig. 7 c). In contrast, cells which have already undergone involution and which do not normally interdigitate further, remain as a clump sitting on the inner apices of the inner deep cells. Some of the marginal cells of the clump will migrate a short distance but they show no tendency to move between the cells serving a substratum.

The range and continuity of the h/l frequency distributions, and the appearance of a continuous morphological series from a rounded cell body, to a rounded cell body bearing an increasingly larger, often oriented, macroprotrusion, and finally an elongate or columnar cell, suggest that, in addition to first serving as locomotory structures, the formation of oriented macroprotrusions also serves as a step in deep-cell elongation and thus the formation of the columnar array of deep cells at stage 11 (Fig. 4).

Through stage ⁠, a large region on the dorsal side of the embryo appears to be nearly homogeneous, but from stage 11 onward, a region high in the dorsal sector, the prospective anterior neural ectoderm (ANE), remains thick while the region closer to the blastopore (the DMZ) thins (Fig. 11). The layer index and the total thickness of each region were compared statistically at stage 10, ⁠, 11, and ⁠, to determine at what stage the hypothesis that the two regions were identical with respect to these parameters would be rejected and thus indicating separate treatment of the two regions. These parameters were chosen because they showed a nearly normal frequency distribution. Both parameters were significantly different between the two regions at but not before stage 11, according to the Student’s t test (P < 0-01) and thus at stage 11 and later the two regions were treated separately (see Figs. 5 and 12).

At stage 11, the ANE remains 15 μm thicker than the DMZ; the layer index remains higher (2·7 versus 2·2); the percent interdigitation and percent inner-to-superficial cell contact remain less than 30% compared to 90% and 100% respectively, for the DMZ (compare Figs. 6 and 13). The h/l ratios of inner and inter deep cells rise to 3·6 and 2·5 respectively, by stage and decreases only slightly by stage (Fig. 13). At the end of gastrulation the ANE is 64 μm in thickness, or about double that of the DMZ. Thus in the ANE, deep-cell elongation occurs but, unlike the situation in the DMZ (Fig. 5), interdigitation does not follow (Fig. 12).

The ventral marginal zone (VMZ) could not be distinguished from the DMZ until stage 10 (Fig. 14) and therefore the data are combined and the plots of all parameters are identical for the DMZ and VMZ to stage 10. At the onset of gastrulation, the VMZ shows similarities in appearance to the DMZ (compare Fig. 4 with 13 and Figs. 5 and 6 with 15) but thinning is slower in the VMZ. This is to be expected since the extension and involution of the ventral sector is delayed relative to the DMZ (see Keller, 1976). Also, the VMZ shows an earlier but less pronounced rise in h/l ratio of the inner deep cells, and no rise at all among the inter deep cells (Fig. 15). Indeed, only the low VMZ (the region from which quantitative data was taken, Fig. 15) seems to behave like the DMZ in showing an increased h/l ratio of inner deep cells. At or after stage ⁠, the higher part of the VMZ (nearer the animal pole) behaves like the blastocoel roof (see Figs. 16 and 17); quantitative measurements from micrographs taken in the high VMZ are similar to those from the BR. For example, the mean total thicknesses for the low VMZ, the high VMZ, and the BR at stage are 76, 48 and 43 μm respectively. The correspondingly layer indices are 2·9, 2·2, and 2·1.

The changes in cell size, shape and arrangement in the course of epiboly of the blastocoel roof (BR) are shown in electronmicrographs (Fig. 16) and by quantitative data (Fig. 17). There are two periods of thinning of the blastocoel roof. The first, of 53%, occurs from stage 8 to stage 10+ and is distributed between all three layers. These changes are probably due to three rounds of cell division from stage 8 to 10 + (Fig. 18), and these alone would reduce the linear dimensions of the cells by 45 %. Since the cells remained cuboidal or approximately cylindrical with height and diameter equal (Fig. 17), change in cell shape probably has little effect on layer thickness in this period.

Concurrent with the decrease in their linear dimensions, complete interdigitation of inner and inter deep cells occurs by stage 10+. In the BR, the tendency of cell divisions in the pregastrular period to increase the number of cell layers is more than countered by thinning, and the number of cell layers actually decreases slightly (Fig. 17). In contrast, the DMZ shows no permanent thinning prior to the onset of gastrulation, and therefore cell division increases the number of layers. This is consistent with the spreading behavior of the two regions. Spreading, and presumably thinning, occurs throughout the blastula and gastrula stages in the blastocoel roof, but occurs only after the onset of gastrulation in the DMZ (see fig. 3, Keller, 1978). Indeed, it is likely that spreading of the deep region vegetally during the blastula stages actually supplies cells to the deep marginal zone (see Discussion, page 232).

The h/l ratio of inner and inter deep cells of the BR does not increase to the extent found in the DMZ (compare Figs. 6 and 17), and the decrease in the h/l ratio after stage brings these values to well below 1·0. This may be related to major differences between the DMZ and the BR in the mechanics and behavior of cell spreading (see Discussion, page 231). In fact, there is a rapid decrease in h/l ratio of inner deep cells of the BR between stage 11 and (Fig. 17), which may be related to events in the ANE. Above (toward the animal pole) and perhaps to the sides of the thickened ANE at stage ⁠, there is a region of the BR which shows very thin, flattened inner deep cells (Fig. 16). The margins of the inner apices of these cells seem to be overlapping one another in the direction of the thickened ANE. This may indicate a shifting and spreading of cells toward the ANE as the cells there become columnar and take up less area.

The layer index and the per cent inner deep cells were chosen as related but independently evaluated criteria of thickness in terms of cell layers, and they serve as checks on one another. The latter should equal one, divided by the layer index minus one. Using this relationship, the mean values of the layer index for a given region and stage was used to calculate the expected value of the per cent inner deep cells and this value was compared to the observed per cent inner deep cells. The correlation coefficient between the expected values and the observed mean values is 0·98 for the BR and DMZ and 0·96 for the VMZ, and the slopes are 1·04, 0·99 and 0·91 respectively when expected values (ordinate) are plotted against observed values (abscissa) (data not shown). It is also obvious from the graphed data that there is a rather sensitive inverse relationship between the layer index and the per cent inner deep cells (see Figs. 6, 12, 15 and 17). The per cent interdigitation and the per cent inner-to-superficial cell contact show coordinate behavior from stage 10 +onward (Figs. 6, 12, and 15) in the DMZ and at all stages in the BR (Fig. 17). At these stages two layers of deep cells interdigitate to form one layer. More inner deep cells contact superficial cells, and increasing numbers of inter deep cells, already in contact with superficial cells, become inner deep cells. Thus as the layer index approaches 2-0, the per cent interdigitation, the per cent inner deep cells, and the per cent inner-to-superficial cell contact approach 100 as expected (Figs. 6, 15 and 17). The coordinate behavior of these parameters suggest that they accurately reflect the positions and the layering of deep cells.

Since the apical area and the height of inner deep cells are known it is possible to calculate, approximately, the volume of inner deep cells. Since it is known that the total cell volume increases very little through gastrulation (see Tuft, 1965), any reduction in mean cell volume is due to division. Thus, the number and times of cell division cycles for all major sectors of the embryo can be estimated from change in mean cell volume. The minimum mean volume is reached at stage (12·8 h) for all regions. Working backwards to the points at which this minimum volume is doubled, it is estimated that the inner deep cells divide once during gastrulation, twice between 8 and 9·5 h, and once before that, for a total of four rounds of division in the period covered in this study. Inner deep cells tend to divide anticlinally during gastrulation, forming dumb-bells or pairs of rounded cells lying next to one another (see figure 22, Keller & Schoen wolf, 1977). Superficial cells also divide anticlinally during this period (Keller, 1978). It is not yet clear whether all divisions during gastrulation are of this orientation and what role this might have in spreading.

On the basis of the quantitative data in Figs. 5 and 6 cellular behavior can be related to the increase in area during extension of the marginal zone (19a). This model accounts for increase in area alone. The mechanism of directing this areal increase toward the blaso-pore will not be considered and for simplicity I have shown only the events from the three-layered stage (stage 10 +) onward. First the inner and inter deep cells interdigitate by extending protrusions between one another, along radii (and perpendicular to the surface) of the embryo, and concurrently they become columnar in shape. As interdigitation proceeds the mean number of cell layers decreases. As the inter and inner deep cells wedge between one another, the deep region becomes thinner and occupies greater area. When interdigitation is complete, there is only one layer of columnar deep cells. These cells then flatten and occupy more area, which results in additional expansion of the deep region. Concurrent with the expansion of the deep region the superficial layer expands by flattening, spreading, and division of the superficial cells (see Keller, 1978). The same process of interdigitation occurs during stage 10, but involves the interdigitation of multiple layers of inter deep cells with one another and with inner deep cells. A similar sequence of events occurs in the low VMZ and presumably in the lateral marginal zone as well, but at progressively later stages and perhaps with less change in the h/l ratio. In some respects this model may apply to the blastocoel roof, but in this case there is no marked increase in h/l ratio of deep cells and the cellular activities involved may be different.

It is appealing to think of the increase in h/l ratio as being related to the extension of protrusions of inner and inter deep cells between one another and thus to the process of interdigitation, but h/l ratio is not always associated with interdigitation and reduction in the number of layers of cells. The anterior neural ectoderm shows such an increase without interdigitation, and all other sectors of the gastrula show interdigitation and decrease in the number of cell layers, but only the DMZ shows a large increase in the h/l ratio of deep cells. In the DMZ interdigitation might be an active process requiring an increased h/l ratio, but in all other sectors it might be a passive process (see page 231) which does not require such an increase. Alternatively, the larger increase in the DMZ may be associated with the extension and convergence movements which are unique to this region. But extension and convergence would be expected to involve changes in the length and width dimensions, as is the case in the superficial layer (Keller, 1978), rather than changes in height. In fact, deep cells appear to be isodiametric and thus extension and convergence probably occurs solely by lateral interdigitation similar to that seen in the superficial layer (Keller, 1978), rather than by change in cell shape.

In order to make a mechanical model which relates this cellular behavior to the production of mechanical forces sufficient to bring about the increase in area during epiboly and extension, it is important to know whether the cell behavior in a given region is an active, intrinsic property of the cell which generates force, or if it is a passive response to forces generated elsewhere in the embryo. For example, the extension of the DMZ might be autonomous and due to local cell behavior, or it might be due to passive stretching in response to tension generated by constriction of the circumblastoporal region or by involution. The grafting and isolation experiments of Spemann (1931; see page 101, Spemann, 1938) and Holtfreter’s observations on exogastrulation (1933) suggest that spreading of the animal half and extension of the marginal zone are autonomous processes arising from forces generated by local cell behavior. But it is not clear whether the observed marginal zone extension is postgastrular notochord extension or the earlier gastrular extension (see Holtfreter, 1944). Also, there has been no quantitative comparison of DMZ extension in situ with involution proceeding and in situ with involution blocked. Schechtman (1942) observed elongation in the absence of involution but this may have been notochord elongation. Expansion of the ectoderm in isolated ventral halves (Spemann, 1931) or in exo-gastrulae (Holtfreter, 1933) was presumed from the folded appearances of the specimens and was not documented quantitatively. Time-lapse films of exo-gastrulae show that the ectoderm becomes corrugated by rapid constrictions of the apices of superficial cells and by the appearance of holes in the epithelium, perhaps as the result of powerful contractions of the cell apices (Keller, unpublished data). Thus, it is shrinkage, not expansion, that generates the folded morphology. Holtfreter’s observation that dorsal marginal zone explanted onto glass or a field of endoderm will elongate and that this process is dependent on the original cellular arrangement remaining intact (Holtfreter, 1944), is the strongest evidence for active extension. Until supporting evidence is available, though, it is prudent to consider both active and passive models of spreading.

Assuming that expansion is active, it is unlikely that this is due to active expansion of the superficial layer. When the superficial epithelium is cut, the wound margins gape open, indicating that it is under tension. The fact that the apices of superficial cells of the extending DMZ show an increase in l/w ratio followed by a decrease as the cells rearrange suggests that this region behaves as a viscoelastic system which is stretched under tension (see Keller, 1978). These cells have machinery for producing tension. Superficial cells contain circumapical bands of microfilaments which bind heavy meromyosin and are undoubtedly actin (Perry, 1975). Apparently, the cortical microfilament system can be made to contract by several conditions which raise the intracellular free calcium levels (Gingell, 1970). If EGTA is injected into the egg, the cortex relaxes, bulges outward, and finally bursts (Baker & Warner, 1972). These observations suggest that the cortical region of the egg and, at later stages, the apices of the superficial cells contain a calcium-regulated contractile system which is maintained under tension. If the superficial layer of the Xenopus gastrula is under tension and yet epiboly or extension is an active or intrinsic property, the force for spreading must arise in the deep region, probably from the active interdigitation and subsequent shortening of deep cells (Fig. 19a). In this model, the outer shell of the gastrula would consist of a bilayer system - a superficial layer under tension and an attached deep region with a tendency to expand forcefully (Fig. 19b). Spreading of the deep region would be resisted by tension in the superficial layer and outward curling of the bilayer would result. However, if the bilayer is constrained by attachment of its margins to the involuted material at the rim of the blastopore, and if the resistance to compression in the deep region is greater than the tension in the superficial layer, spreading of the entire bilayer would result from active expansion of the deep region (Fig. 19 c). Such a bilayer system is mechanically stable. This can be demonstrated by cutting a hole in a rubber raquet ball and turning it inside-out such that the superficial component is under tension and the deep component is under compression (Fig. 19 d).

There is considerable evidence in support of the bilayer model described above. It is a common observation that when pieces of the superficial-deep bilayer of the gastrula are excised, their immediate behavior is to curl to the outside, opposite the curvature of the embryo, suggesting that the deep layer is under compression and that the superficial layer is under tension. Second, cuts made in the superficial layer gape open, again suggesting tension. Third, the evidence presented in this paper shows that the proposed method of spreading of the deep region does occur. Fourth, preliminary evidence (see last section of the Results) shows that excised deep cells of the marginal zone prior to involution will actively interdigitate between cells of the inner surface of the blastocoel roof or the marginal zone. Furthermore, there is apparently a behavioral change upon involution, for cells which have involuted do not interdigitate between the inner deep cells of the uninvoluted marginal zone but instead use their apices as a substratum for migration (see Nakatsuji, 1974, 1975a, b, and Schoenwolf, 1977). This behavioural change from an interdigitating to a migrating cell type is correlated with the morphological differences between preinvolution and involuted cells, and is the basis for the formation of the interface between the two cell types (see Keller & Schoenwolf, 1977).

In contrast, a passive model of expansion would hold that the superficial epithelium is stretched by tension generated at the margin of the blastopore and that the superficial cells passively spread (Fig. 19e). As they do so the deep cells would actively rearrange themselves to occupy the increased area available to them but would not forcibly interdigitate or generate a spreading force. In this model, the entire shell of uninvoluted material would be pulled down over the involuted material by a dynamic locus of circumblastoporal constriction or perhaps by involution itself.

The total cell volume remains nearly constant during gastrulation (see Tuft, 1965), and the extracellular spaces in the regions undergoing epiboly extension do not appear to change, since cells in these regions seem to be closely packed at all stages. Therefore an inverse relationship should exist between the thickness and the amount of areal expansion in a given region. Overall, this expected inverse relationship is found in the superficial layer from the early gastrula stage onward, but the amount of thinning in the deep region is much greater than that in the superficial layer, implying that spreading in the deep region should far exceed that of the superficial layer. As calculated from its decrease in thickness, the area of the deep region of the DMZ should increase by a factor of 2·56 from the early gastrula (stage 10 + ) to the late gastrula (stage ⁠), whereas actual expansion of the superficial layer of this region is only by a factor of 1·95 (Keller, 1978). The same figures for the deep region of the blastocoel roof from stage 10+ to are 1·79 for expected and 1·2 for actual expansion. The thinning, and therefore spreading, of the deep region is most pronounced in the period from stage 10 to 10+, in which the deep region must expand by a factor of 1·85 in a short time. Where does this newly generated area go? Undoubtedly it is fed into the involuting mesodermal cell stream, which forms early in Xenopus (see Nieuwkoop & Florshiitz, 1950; Keller, 1976), and it probably also contributes to the rapid extension of the dorsal marginal zone vegetally during this period (see Keller, 1978).

Since the deep region expands more rapidly than the superficial layer it must move vegetally faster. Indeed, the upper border of the prospective mesoderm in the deep region lies above the corresponding upper boundary of the suprablastoporal endoderm in the early gastrula but coincides with it in the neurula (Keller, 1976). The great expansion of the deep marginal zone in the early gastrula stage suggests that its upper boundary lies much higher in the blastula at stage 8 or 9, than indicated on maps of the early gastrula (Keller, 1976).

Since the behavior of superficial cells can be followed by time-lapse cinemicrography and at least some aspects of deep cell behaviour can be studied by the methods described here, an experimental analysis of gastrulation movements at the cellular level is in order. First, it is important to determine whether cell behavior in both the deep and the superficial layers is regionally determined and autonomous, or if it is passive and dictated by systemic mechanical forces. This question can be answered unambiguously by grafting the superficial layer, the deep layer, or both from one region of the gastrula to another and determining with time-lapse cinemicrography and SEM whether cells of the grafted patch undergo spreading and rearrangement as they would have in their original location or adopt the pattern characteristic of the region into which they were grafted. Secondly, experiments must be done to map the mechanical stresses in the embryo and relate this to cellular behavior. Beloussov, Dorfman & Cherdantzev (1975) mapped mechanical stresses by recording the behavior of freshly isolated fragments of the gastrula. Additional work of this type must be done in order to understand the forces that actually bring about distortion of the embryo. Thirdly, the social behavior of cells with different or changing morphogenetic properties should be studied in vitro. Fourthly, these cells should be examined at the ultrastructural and molecular levels for differentiation of cytoskeletal, junctional, and cell surface components which might relate to these differences in morphogenetic behavior. It is important that these last two studies be directed toward cell populations with specific functions, such as preinvolution deep cells and deep cells after they have involuted, rather than traditionally important geographic regions such as the ‘dorsal lip’ which contains both these and probably more types of cells.

1 wish to thank Professors R. Briggs and J. P. Trinkaus for their support and suggestions, Dr W.-T. Chen, Dr G. Radice, and M. Shure for helpful discussions, Dr John Gerhart for reading the manuscript, and Dorothy Barone and Patricia White for technical assistance. This work was supported by NIH grant No. USPHS-HD-07137 to J. P. Trinkaus and No. R01 GM 05850-21 to R. W. Briggs, and by American Cancer Society Fellowship No. PF 1174 to the author.