Locomotion and cell-substratum contacts of Xenopus epidermal cells in vitro and in situ

The spreading of epithelial sheets of cells is an important feature of animal morphogenesis, as well as wound repair and malignant invasion, yet the mechanism of epithelial spreading is poorly understood. A major obstacle to studying these movements has been that most tissues are opaque, hence their cells cannot be observed directly in situ. To avoid this problem, I have studied the migration of epithelial cells during wound closure in the transparent tail fin of Xenopus laevis tadpoles, where the locomotion and contact behaviour of individual cells can be seen clearly (Radice, 1980; for studies of other cell types in the tadpole fin see Billings-Gagliardi, 1977; Lehman, 1953; Speidel, 1964; and especially Clark, 1913, 1916; Clark, Clark & Rex, 1936). As favourable as this tissue is, however, certain experiments or observations, such as those using interference-reflexion microscopy, can only be made on cells spreading in culture. Although the molecular basis for motility is undoubtedly the same in both cases, there may be subtle yet important modifications of behaviour imposed by the culture environment. Since Xenopus epidermal cells will also migrate from explants when cultured on glass and plastic (Bereiter-Hahn, 1967; Radice, 1978), they, like the fibroblasts from the chick cornea (Bard & Hay, 1975) and Limulus amoebocytes (Armstrong, 1979), provide the opportunity to test directly the influence of culture conditions on the motile behaviour of tissue cells. In the present study, I compared the movements of these cells in situ and in vitro. Of particular interest were the contacts that spreading cells formed with their substrata under both conditions. Parts of this work have been presented in preliminary form (Radice, 1978).

Tadpoles were obtained from induced matings, raised, and staged by the method of Nieuwkoop & Faber (1967). Cell motility was studied in vivo during wound closure employing procedures published elsewhere (Radice, 1980).

The procedure for culturing tadpole epidermal cells was modified from Bereiter-Hahn (1967), and Chen et al. (1978). Tadpoles from stages 47–50 were starved for 1 day to reduce the number of intestinal contaminants, then transferred to a dilute solution of potassium permanganate (about 0 · 005 M) for 1 to 2 min to sterilize the epidermis. All further steps were aseptic. The posterior one-half to one-third of the tail was severed with a scalpel, transferred with a wide mouth pipette through sterile Steinberg’s solution, and placed in culture medium. Then, 2 or 3 fragments were transferred separately in small drops of medium and spaced evenly on a 25-mm² coverglass that lay in a 60-mm Petri dish. Two or three small pieces of coverglass were placed at the comers as spacers, then a second 25-mm² coverglass was placed on top of the explants to hold them in place. At this point, the volume of medium was adjusted so that it just spread to the edges of the coverglass ‘sandwich’; surface tension thus kept the 2 coverglasses immobile. Finally, the cover was placed on the Petri dish and the culture was transferred to an incubator at room temperature and 100% humidity for 24–30 h.

The culture medium consisted of a 1:1 mixture of Steinberg’s solution and calf serum (Gibco no. 220-6440), pH 7·8, with 100 units/ml antibiotic-antimycotic (Gibco no. 600–5240). As noted by Bereiter-Hahn (1967), this high concentration of serum seemed to promote the spreading of epidermal cells on the coverglass, since at lower serum concentrations (i.e., 10–20 %) fewer cultures were successfully established.

For light-microscopic observation, fresh medium was added to the side of the culture until the upper coverglass floated free. Then the coverglass with its attached epidermis was sealed with silicone grease over a 20-mm diameter hole in glass slide. When inverted, this formed a shallow well that was filled with medium containing 50 % serum, or, in some cases, 30 % serum as noted in the text. For micromanipulation, the well was left open and the cells were observed with water-immersion objectives. For all other observations, the filled well was sealed with another coverglass and the chamber was then inverted so that the cells could be observed directly through the coverglass with dry or oil-immersion objectives. Cultures appeared to remain healthy in these sealed chambers for several hours, more than sufficient time for these studies.

Cells were observed with phase-contrast and interference-contrast optics. For interference-reflexion microscopy a Zeiss Universal microscope was modified according to the design of Curtis (1964) and Izzard & Lochner (1976). An illuminating numerical aperture of 107 was used for all photography. A complete analysis of the technique can be found in Izzard & Lochner (1976), Gingell & Todd (1979) and Bereiter-Hahn, Fox & Thomell (1979).

A Bolex H16 movie camera coupled with a Sage intervalometer was used for time-lapse cinemicrography. Most sequences were filmed at 2 frames/s with Kodak Plus-X reversal film. Kodak SO-115 technical pan film was used for recording interference-reflexion images. Prints from 16-mm reversal films were made via an internegative of 3 5-mm Kodak So-115 film made with a Honeywell Repronar slide duplicator.

For scanning electron microscopy of cultured cells, the medium was carefully removed from the culture chamber and the cells were rinsed with Steinberg’s solution. They were then fixed with 3 % glutaraldehyde (Polysciences, EM grade) in Millonig’s phosphate buffer with 2% sucrose at pH 7·3 and room temperature. Cells were fixed for 20 min to 4 h, after which they were rinsed with buffer and postfixed with 1 % osmium tetroxide in the same buffer for 5-15 min. Then, they were rinsed with buffer, dehydrated through ethanol, and dried by passage through the critical point of CO,. After coating with approximately 20 nm of gold-palladium, they were examined with an ETEC Autoscan SEM.

The procedure for transmission electron microscopy of sectioned tadpole epidermal cells in situ has been reported elsewhere (Radice, 1979). For TEM of cells maintained in vitro, cells were fixed with glutaraldehyde and postfixed as for SEM. Then, the cells were treated with i % tannic acid in buffer for 30 min, dehydrated through ethanol, and embedded by laying the coverglass that supported the cells face down on a thin layer of Epon. After the plastic polymerized, the coverglass was separated by immersing it in liquid nitrogen. The wafer of Epon that contained the cells was then re-embedded in a second layer of Epon and sectioned. Thin sections were stained with uranyl acetate and lead citrate and examined with a Philips 300 microscope.

Cells were manipulated with tungsten microneedles mounted on a Leitz micromanipulator.

The general characteristics of the epidermal sheet that migrates from fragments of Xenopus tadpole tails have been described by Bereiter-Hahn (1967) and will be summarized here only briefly. Within 12–24 h after culturing, the epidermis attaches to the substratum and begins to spread outward, usually beginning at the distal tail tip, but often proceeding from the cut, proximal edge as well. Of the several cell types in the tail fragment, only the cohesive epidermal epithelium and a few isolated cells (probably fibroblasts) migrate from the explant during the first 24-36 h of culture. The epithelial cells are easily identified by their contiguity and by the desmosomes within their stable lateral contacts (Figs. 1–3). In the early stages of migration, they spread as a bilayered sheet with basal cells on the substratum and outer cells facing the culture medium, so that the normal polarity of the epithelium is maintained. Later, in some cultures, the outer layer of cells detaches from the basal layer and floats freely in the medium, leaving an intact layer of basal cells attached to the substratum (Fig. 1). Both monolayered and bilayered regions were studied in these experiments. No mitosis is seen during the first 24 h of culture; in consequence, as the epithelium spreads, the individual cells also become more spread, the outer ones averaging 394 ±115 μm²in vitro, compared with 245 ± 50 μm-in situ at the same stage (mean ± standard deviation). The areas of the basal cells are usually difficult to determine by light microscopy because their margins are indistinct when they are covered by the outer cells, as is usually the case in situ and in vitro. However, in thin sections they appear to have about the same area as the outer cells. It is interesting, therefore, that in cultures where the outer layer has detached, the basal cells are much more spread, averaging 629±169 μm². They may spread more when outer cells detach.

To determine in what ways the spreading of epithelial sheets in vitro resembles their behaviour in the animal, I compared the locomotion of epidermal cells under 3 conditions: during wound closure in situ, spreading at the outer edge of a sheet in vitro, and at the margin of microwounds made by removing 1 to 4 cells from the central portion of a sheet spreading in vitro. In all cases, the basal cells spread by extending lamellipodia over the substratum (Figs. 1, 2, 4, 5), those observed in vitro being somewhat broader than those in situ. Microspikes and filopodia are never seen, and blebs are rare, both in vitro and in situ.

To see whether the locomotory activity of the lamellipodia was also the same, I used the method of Abercrombie, Heaysman & Pegrum (1970) to analyse the movements of the leading edge of several cells under each of the 3 conditions. Table 1 shows that the rate of advance (net displacement) of epidermal cells during wound closure in situ is 11·5± 1·8μm/min. This is due entirely to a steady rate of protrusion, since the cells do not stop spreading or withdraw until they meet other cells at the centre of the wound. In contrast, cells at the outer edge of an epidermal sheet in vitro advance much more slowly, at 1·5 ± 0·55μm/min. This reduced rate has 2 components; cells protrude more slowly (5·9± 2·1 μm/min) and are stationary or withdrawing much of the time. In this respect they move much like chick gut epithelial cells in culture (Table 1).

It is not clear why cells at the edge of the epidermal sheet in vitro move slowly and hesitantly compared with their rapid and direct movement in situ. A partial explanation may be that since cells at the outer edge of the sheet are more spread, they are under more tension than are cells spreading into wounds in situ, and, being ‘at the end of their rope’, they resist further spreading. The behaviour of cells spreading into microwounds in vitro is consistent with this hypothesis. When such wounds are made, the margins of the wound immediately retract and round off, indicating that the epithelium is indeed under tension (Figs. 4, 5), as apparently are spreading epithelial sheets in general (e.g., Holmes, 1913; Trinkaus, 1951; New, 1959; Beloussov, Dorfman & Cherdantzev, 1975). In contrast, wound borders do not retract after wounding in situ, indicating that the normal, intact epidermis is not under significant tension prior to wounding and cell migration (Radice, 1979). If tension retards the advance of cells in vitro, then cells spreading into microwounds should migrate faster at the beginning of closure, that is, after the wound margins have retracted and tension has decreased. Such is the case. Cells spreading into microwounds in vitro advance at 6·5 ± 4·9 μm/min during the first 2·5 min of spreading, several times faster than cells at the outer edge of an unwounded sheet. This increased rate is not due to an increased rate of protrusion, but rather to an increase in the amount of time each cell spends protruding compared with standing still or retracting (Table 1). After this initial burst of spreading, the rate of movement slows, consistent with the proposed increase in tension in the sheet, although the increasing contacts between cells as the wound closes may also contribute to this reduction in the rate of movement. These rates of movement are for cells spreading in medium containing 50% serum; cells fluctuate more and translocate slower in medium containing 30% serum (see Figs. 4, 5), suggesting that serum may affect spreading, perhaps by modulating the adhesion of cells to the substratum.

One should note that the structure I have described as a lamellipodium is larger in Xenopus epidermal cells than the lamellipodia of chick and mouse fibroblasts in vitro (Abercrombie et al. 1970) and of chick epithelial cells in vitro (DiPasquale, 1975a). In those cells, lamellipodia are 1–2·5μm long and 0·11-0·16μm thick, whereas in Xenopus epidermal cells they may be 10–15 long and 0·15-0·40 μm thick. Like other lamellipodia, however, they protrude from the cell margin, are motile, are of a uniform thinness, and exclude most cytoplasmic organelles, containing instead a meshwork of 4–6 nm diameter microfilaments. Unlike lamellipodia of other cells, however, Xenopus lamellipodia extend parallel to the substratum, apparently in contact with it (see below), and rarely lift from the substratum or fold back to form a ruffle (cf. Ingram, 1969; Abercrombie, Heaysman & Pegrum, 1971; Lochner & Izzard, 1973; Harris, 1973b).

In the intact epidermis in situ, the entire basal plasma membrane of the basal cells lies less than 20 nm from the lamellated bodies, and in many places appears to contact them (Radice, 1980). It is not clear, however, whether the cells are adhering to the lamellated bodies or to the fine fibrils that traverse the adepidermal space and connect with the basal lamina (see Kelly, 1966; Overton, 1979). In any case, at the developmental stages studied, there are no hemidesmosomes between the basal cells and the basement membrane, although they do appear later. While migrating during wound closure, the basal epidermal cells remain in close apposition to the substratum over much of their lower surface (Fig. 6). The distal margin of the lamellipodium, in particular, is almost always less than 10–15 nm from the lamellated bodies (50–60 nm from the basal lamina, see Radice, 1980). Interestingly, a wider separation from the lamellated bodies (about 30–80 nm) typically is seen about 0·1–1·5μm back from the leading edge. The separation narrows again to less than 20 nm under the remainder of the cell body to the point where the cell’s trailing edge has detached from the substratum. The transition in this region is acute, as if the cell’s trailing edge is ‘peeling’ from the substratum (Fig. 7). This is consistent with time-lapse films that show detachment to be smoothly continuous (Radice, 1980) rather than sporadic and abrupt as is characteristic of fibroblast locomotion in culture (Harris, 1973a; Chen, 1977).

Localized, electron-dense plaques joining cells with substratum, such as those seen in fibroblasts moving in vitro (Abercrombie et al. 1971; Brun, Ericsson, Ponten & Westermark, 1971; Goldman, Schloss & Starger, 1976) have not been observed among these epidermal cells migrating in situ, even at regions of cell-substratum separation of less than 10–15 nm-Nor do these cells form hemidesmosomes with the wound surface during migration, as apparently do mammalian epidermal cells and gingival cells during wound closure (Krawczyk, 1971; Pang, Daniels & Buck, 1978; Sciubba, 1977).

Epidermal cells migrating in vitro show a similar pattern of cell-substratum contact when observed with TEM although the separation is usually greater: most of the lamellipodium and much of the cell body is 10–40 nm from the glass substratum (Figs. 8, 9). In no case were the cells’ plasma membranes actually in contact with the protein film that coats the substratum. In contrast to the case in situ, however, these cells do form small electron-dense plaques similar to those mentioned above (not illustrated), but they are not numerous or as large as those described in other cell types.

Although the high resolution of TEM makes it a valuable tool for studying cell contacts, thin sections reveal only a srrtall sample of a fixed cell surface. To examine the pattern of cell-substratum contacts in living cells migrating on glass, interference-reflexion microscopy (IRM) was used. With these optics, a consistent pattern of interference is observed (Fig. 10). When viewed with monochromatic light, most of the leading lamella and some of the cell body produce a broad, grey image consistently darker than the background intensity. This grey region represents a separation of about 30–40 nm between cell and substratum (Izzard & Lochner, 1976) and is known as a close contact. Darker areas within the region of close contact may be somewhat closer than 30 nm, while the lighter grey portions may be more than 40 nm. Within the close contacts of some but not all of the cells are dark spots or streaks, about 1·5 μm long and 0·2–2 μm wide, that represent separations of 10–15 mm and are known as focal contacts. Much of the rest of the cell is brighter than background intensity, which indicates a separation in these areas of 100 nm or more.

Fig. 10 shows how the pattern of contact changes as a cell at the edge of the sheet moves across a glass substratum. Note that the protruding region of the lamellipodium always maintains close contact with the substratum right up to its outer edge as it protrudes. This agrees well with the results of electron microscopy (Fig. 8), micromanipulation (see below, Figs. 13–15), and time-lapse cinematography of these cells in vitro and in situ, all of which show that the margin of the lamellipodium extends close to the substratum. In contrast, the leading margin of the lamellipodia of fibroblasts in vitro usually extends at a slight angle from the substratum, and then either moves down onto the substratum or folds back onto the upper surface as a ruffle (Ingram, 1969; Harris, 1969; Izzard & Lochner, 1976). The constant adhesion of the leading edge to the substratum may prevent Xenopus epidermal cells from ruffling. Consistent with this idea, the leading margin of the lamellipodium often does fold back like a ruffle whenever it is dislodged from the substratum either by another lamellipodium or by a microneedle (Figs. 5, 13–15).

Within the close contact of a protruding lamellipodium there typically is a band of lighter intensity, 0·5–1·5 μm wide, that appears 1–3 μm back from the edge, maintains that distance as the lamellipodium advances, and disappears when the lamellipodium retracts (Fig. 10). Although the significance of this zone of wider separation is unclear, it is interesting that a similar region of slightly wider separation is also often seen a short distance back from the edge in sections of lamellipodia of cells spreading in situ (Fig. 6).

Although the pattern of close contacts changes rapidly as the cell moves, focal contacts do not move relative to the substratum (Fig. io) as has been noted before in studies of fibroblasts (Lochner & Izzard, 1973; Abercrombie & Dunn, 1975; Rees, Lloyd & Thom, 1977). As is the case with other cells, focal contacts of Xenopus epidermal cells appear within the larger region of close contact, become slightly longer or shorter, wider or narrower as the cells move, persist for several minutes, then fade (Fig. 10). It is important to note that not all cells have focal contacts. In fact, most of the cells do not, and those with focal contacts generally move more slowly than those without.

Interference-reflexion microscopy is particularly useful for examining the contacts that submarginal cells form with the substratum, since the image obtained is not obscured by overlying cells. Fig. 11 shows that submarginal cells of the epidermal epithelium also form extensive close contacts with the substratum in vitro, especially beneath their lamellipodia and often beneath their nuclei as well. And the close contact of these cells also extends to the edge of the lamellipodium as it protrudes and withdraws, just as it does beneath the marginal lamellipodia. Separations of 100 nm or more also are seen, not beneath the lamellipodia, but in the perinuclear region. Focal contacts, though present, do not appear as frequently beneath the submarginal cells as they do beneath the marginal cells. That these submarginal cells not only make contact with but also adhere to the substratum is shown by the bits of cytoplasm left behind when a submarginal cell is removed with a microneedle (^Fig-5)-

To analyse further the pattern of adhesive contacts with the substratum, several cells spreading in vitro and in situ were manipulated with tungsten microneedles. With this technique, one assumes that migrating cells are under tension, stretched between their sites of adhesion to the substratum; thus dislodging part of a cell from its adhesive site causes that part of the cell to retract sharply. The cell then takes a new shape determined by the location of its remaining adhesions (Goodrich, 1924; Chambers & Fell, 1931; Harris, 1973a; DiPasquale, 1975a).

When the entire lamellipodium of a cell is dislodged, the cell retracts into the sheet, indicating that the cell is under tension and that it adheres strongly to the substratum by its lamellipodium, and less strongly beneath the cell body. This retraction is obvious in vitro but less dramatic in situ, probably because, as noted above, the cells migrating in situ are initially under less tension. To see whether the entire lamellipodium adhered, only a small part of the leading edges of several cells was dislodged. In each case, only the part of the cell touched by the needle retracted, the rest of the lamellipodium retained its form, indicating that most of the lamellipodium, both in vitro and in situ, adheres to its substraum (Figs. 13 – 15). Thus, it appears that Xenopus epidermal cells indeed adhere to the substratum at regions that TEM and IRM show are generally 30 – 40 nm or less from the substratum.

The most obvious parallel in the behaviour of Xenopus epidermal cells in situ and in vitro is that in both cases cells spread by extending broad lamellipodia and are devoid of other protrusions usually associated with moving cells, such as microspikes or filopodia. The cells also lack microvilli, although they do possess surface folds. The general absence of cylindrical protrusions may reflect the organization of the underlying cytoskeleton.

Even though the shapes of migrating cells are similar, their behaviour differs under the 2 conditions, since lamellipodia in vitro protrude more slowly and tend to fluctuate so that their net rate of translocation is less than that in situ. Why cells protrude more slowly in vitro is not clear, but may be due to the higher tonicity of the tissue culture medium. Consistent with this, when the epidermis is wounded in situ while immersed in full strength Steinberg’s solution, or in 10% Niu-Twitty solution containing 0-2 M sorbitol, the rate of protrusion of the marginal cells is reduced to near the maximum rate of protrusion in culture (data not shown). Raising the sorbitol concentration to 0-5 M reversibly inhibits protrusion altogether. Whether this response is due to a reduced hydrostatic pressure within the cell (Harris, 1973b; DiPasquale, 1975 b), or a disruption of essential enzymic processes, or some other mechanism is not known.

Although the rate of protrusion is reduced by these treatments, the relative amounts of time spent protruding, withdrawing, or remaining stationary are not, so this fluctuation of the leading edge must depend on other factors. One such factor appears to be the degree of tension within the sheet, since epidermal cells fluctuate less in vitro when tension is reduced by retraction of the sheet. A similar phenomenon has recently been discovered independently in fibroblasts moving across a plane substratum (retraction-induced spreading; Chen, 1979). This relationship between degree of spreading and rate of translocation may help to explain the extraordinarily high rate of movement during wound closure in situ. Since the epidermis is apparently not under significant tension prior to wounding, and the wounds are relatively small (less than 150/tm in diameter), the cells may not migrate far enough to generate tension sufficient to retard their further movement. If this is true, one would predict that in larger wounds in the Xenopus epidermis, fluctuation would be more frequent as wound closure nears completion and tension builds. Preliminary results suggest that this may be the case. But, as these larger wounds close, the basement membrane buckles, and the flatness and clarity of the tissue are lost. Thus, although this folding is an indication that tension does increase with spreading in situ, it makes it impossible to follow the delicate movements of the marginal cells.

Whether tension modulates the spreading of other epithelial cells is not clear. The evidence is conflicting. Lash (1955), for example, found that in wounds about 1 mm² in the epidermis of larval urodeles the rate of movement was relatively constant throughout closure. Thus, if tension was increasing, it apparently did not slow the advance of these cells. Takeuchi (1979) found that in the case of chick corneal epithelial cells spreading on Millipore filters in culture, applying tension to the sheet of cells actually increased the rate of spreading. On the other hand, Downie (1976) found that although the normal increase in tension that appears as the chick blastoderm spreads is not enough to limit its rate of advance, artificially increasing the tension does limit the sheet’s advance. A slightly different example of a delay in spreading as a result of tension is found in the teleost Fundulus (Betchaku & Trinkaus, 1978). During epiboly of this embryo, the epithelial enveloping layer (EVL) is normally firmly attached to the independently spreading yolk syncytial layer (YSL) and spreads with it over the surface of the yolk. However, when the EVL is experimentally detached, the YSL speeds up, and may spread over a full developmental stage faster than do control embryos. This example is instructive because the YSL appears restrained by tension within the EVL, a separate but attached layer of cells. A similar situation may occur during the closure of wounds in a stratified squamous epithelium such as the mammalian epidermis, where the basal layers of cells may be detained by their adhesion to the outer, keratinized layers, which do not spread (McMinn & Pritchard, 1972). The slower rate of migration of mammalian epidermis compared with amphibian epidermis is consistent with this idea.

Micromanipulation demonstrates that spreading Xenopus epidermal cells do not detach from the substratum (cf. Lash, 1955, 1956), but adhere to the wound surface in situ and to the glass substratum in vitro by their lamellar margins, as do various other cells migrating over plane substrata in culture (Goodrich, 1924; Chambers & Fell, 1931; Harris, 1973a) and in vivo (Downie & Pegrum, 1971). In transmission electron micrographs (TEM) of thin sections, these regions of the Xenopus cells generally are separated from the lamellated bodies in situ by 0-20 nm, and from the glass substratum in vitro by about 10—40 nm. The separation seen in vitro is probably not an artefact, since it agrees well with the separation calculated from interference-reflexion microscopy (I RM) of living cells, and studies of other cells in vitro with TEM (see Curtis, 1973, for review) that show the closest approach of cell to an extracellular substratum of about 10 nm. That the plasma membrane of the cells in situ contacts the lamellated bodies in many places is thus unusual. Little is known about the composition of the lamellated bodies, other than they probably contain lipid (Kelly, 1966), and the nature of the interaction between the cell and these structures is not known. In fact, it is not clear whether the cells actually adhere to them, or instead to the fine fibrils that cross the adepidermal space and connect with the basal lamina (Kelly, 1966; Overton, 1979).

Contrary to expectation, no electron-dense plaques were found with TEM in regions of cell-substratum contact in situ. Such plaques are commonly observed in a variety of cell types in culture (Abercrombie et al. 1971; Brunk et al. 1971; Goldman et al. 1976) and have been seen in cells of the chick blastoderm spreading on the vitelline membrane (Downie & Pegrum, 1971). Of course, it could be that such plaques are present in Xenopus epidermal cells in situ, but are small and sparse and so were missed in the sample of sections examined. It is also possible that these cells do not form plaques during migration in situ. Many of the epidermal cells in vitro do not form focal contacts (equivalent to plaques) when viewed with IRM, particularly if they are moving rapidly. Indeed, it now appears that rapidly migrating cells in general, such as rabbit polymorphonuclear leukocytes (Armstrong & Lackie, 1975), amphibian leukocytes, and freshly seeded chick heart fibroblasts (Shure, Young, Kolega & Chen, 1979; Couchman & Rees, 1979) form fewer or no focal contacts with the substratum than do slowly moving or stationary cells. If this is so, then such cells should also have fewer microfilament bundles (stress fibres), since it is known that most such bundles terminate at focal contacts (Heath & Dunn, 1978). Couchman & Rees (1979) have recently confirmed this correlation for chick heart fibroblasts in culture, and I, also, have noted that there are few microfilament bundles in Xenopus epidermal cells migrating during wound closure in situ (Radice, 1978). Together, these results support the suggestion of Couchman & Rees (1979) that bundles of microfilaments are less important for cell motility per se than is widely believed. Instead, they may be more important in stabilizing cell adhesions and cell shape.

The absence of detectable focal contacts in many migrating cells indicates that cells adhere at their regions of close contact, as suggested by Abercrombie & Dunn (1975). Close contacts no doubt contain weaker adhesions than focal contacts: not only is the separation greater, but time-lapse films also show them to change their distribution continuously as the cell moves, and to be dislodged easily when underlapped by neighbouring cells. It appears, therefore, that the adhesion in these regions is strong and stable enough to hold the cell on the substratum and give it traction at the leading edge, but not strong enough to retard the advance of the trailing edge, as is the case with focal contacts (Chen, 1979).

This investigation clearly demonstrates that submarginal basal cells of Xenopus epidermis adhere to the substratum and spread actively in vitro and in situ (see also Bereiter-Hahn, 1967). During wound closure in situ this behaviour contributes to the advance of the sheet, since all the lamellipodia of the submarginal cells extend in the direction the sheet is spreading. In vitro, however, each submarginal cell may extend a lamellipodium across the substratum in any direction, not only in the direction of the marginal cell ahead of it. The reason for this lack of coordination of cells in vitro is not known, but the result is that the expansion of the sheet in culture is due primarily to the movement of the marginal cells. This is not because the submarginal cells are passive, however, as they apparently are in embryonic chick epithelial cells (DiPasquale, 1975 a), but because the marginal cells have larger, more consistently oriented lamellipodia, and apparently because of this overcome the less-oriented movements of the smaller submarginal lamellipodia (see also Vaughan & Trinkaus, 1966).

This study has shown that Xenopus epidermal cells move in much the same manner during the closure of microwounds in situ and in vitro. The differences in rates of movement may in part be due to the increased tension within the sheet in vitro. They may also reflect the presence of focal contacts or adhesive plaques seen in vitro but not in situ, although whether this means that the presence of focal contacts reduces the rate of translocation or that slower-moving cells tend to form more focal contacts is not known. Because the Xenopus epidermis can be observed both in situ and in vitro, and because the cells can be stimulated to move in a known direction simply by wounding, they appear to offer unique advantages for further studies of tissue cell motility.

It is a pleasure to acknowledge the advice and encouragement of Professor J. P. Trinkaus, in whose laboratory this work was carried out. I also thank Dr Wen-Tien Chen and Mr Michael Shure for valuable discussions, and Dr Peter Armstrong for reading the manuscript and suggesting improvements. This work represents part of a dissertation presented to the Graduate School of Yale University in partial fulfilment of the requirements for the Doctor of Philosophy degree. It was supported by grants from the NIH (USPHS-HD-07137 and CA-22451) and NSF (BNS 70-00610) to J. P. Trinkaus. The author was supported by a Yale Teaching Fellowship and a Graduate Fellowship from the Danforth Foundation.