Retinoic acid application to chick wing buds leads to a dose-dependent reorganization of the apical ectodermal ridge that is mediated by the mesenchyme

The analysis of the early events that follow retinoic acid application to the chick wing bud is important in understanding the cellular mechanisms involved in limb patterning and how these are correlated with morphogenesis. In chick limb development, the pattern of cellular differentiation is controlled by cell signalling and the shape of the bud is mediated by epithelial mesenchymal interactions (reviewed Wolpert, 1981). A striking finding is that retinoic acid, a chemically defined substance, can mimic one of the signalling systems that control pattern (Tickle et al. 1982; Summerbell, 1983). Furthermore, retinoid-induced pattern changes are linked to the morphogenesis of the limb (Lee & Tickle, 1985).

The effects of local retinoid application to early wing buds are both dose- and position-dependent (Tickle et al. 1985). When appropriate doses are applied at the anterior margin of a chick wing bud, a sequence of additional digits develops in mirror-image symmetry with the normal set. This change in pattern mimics the effects of grafting cells from the polarizing region (Saunders & Gasseling, 1968; Tickle et al. 1985). The polarizing region is a group of limb mesenchyme cells, which is located at the posterior margin of the normal bud and has been postulated to generate a signal that determines the pattern of connective tissue differentiation across the anteroposterior axis (this axis runs in the hand from thumb to little finger). Similar conditions are required for the induction of pattern changes by retinoic acid and polarizing region grafts (Tickle et al. 1985; Eichele et al. 1985). This has led to the suggestion that retinoic acid may be the signal produced by the polarizing region (Tickle et al. 1985). Recently, it has been shown that retinoic acid is present in developing chick wing buds and this raises the possibility that retinoic acid may act as an endogeneous signal in chick limb development (Thaller & Eichele, 1987).

The exposure to retinoic acid that experimentally brings about pattern changes in the wing has been established by removing beads soaked in retinoic acid at various time intervals after implantation (Eichele et al. 1985). During the first 12 h or so of exposure, any changes are reversible but during the next 6h, additional wing digits are sequentially specified. Therefore, to seek for the basis of retinoid-induced changes in wing pattern, the early events that follow retinoic acid application must be analysed.

One striking consequence of retinoid application that can occur well before cells differentiate into the cartilage of additional digits is that the shape of the bud changes (see for example, Wilde et al. 1987). The dose- and position-dependent effects of retinoic acid application allow us to produce a range of limb patterns and we can monitor the related changes in bud shape (Lee & Tickle, 1985). For example, under conditions when additional digits will develop (low doses applied to the anterior margin), the bud broadens but as higher doses are applied in the same position, the number of digits in the symmetrical wing progressively decreases and the buds are narrower.

The modifications in bud shape appear to be mediated by changes in the length of the apical ectodermal ridge, the thickened epithelium at the tip of the bud that is required for outgrowth (Saunders, 1948; Summerbell, 1974). In broader buds, the apical ridge lengthens whereas, in narrower buds, the ridge shortens. We can account for these changes in ridge length by assuming that where the concentration of retinoic acid exceeds some threshold value the ridge disappears but else-where lower concentrations of the retinoid lead to ridge maintenance. Therefore, the net effect on the ridge will depend on the dose of retinoid applied and also on the position of application. For example, when high doses of retinoic acid are applied at the bud apex, the ridge will be affected on both sides of the source and is entirely obliterated leading to truncations (Lee & Tickle, 1985; Tickle & Crawley, 1988).

We wish to understand how retinoic acid application affects the apical ectodermal ridge. Here we explore the cellular changes that lead to the disappearance of the apical ridge using slow-release carriers presoaked in retinoic acid implanted at the apex of wing buds. We show that application of the retinoid leads to dose-dependent changes in the height of the apical ectodermal ridge. Under conditions that obliterate the ridge, the epithelium loses its typical morphology of tightly packed pseudostratified cells and is converted into a loosely packed arrangement of cuboidal cells. The development of recombinations of treated and untreated limb bud tissues shows that retinoids appear to act via the mesenchyme. The change in ridge morphology that is mediated by the mesenchyme appears to involve a reorganization of the epithelial cells.

Two types of carriers were used to apply the retinoid: AG1-X2 beads, 200 μm in diameter (Eichele et al. 1984; Tickle et al. 1985), and small pieces of anion exchange membrane (BDH cat. no.: 55164). The sheet of anion exchange membrane is 100 μm thick and was cut to give flat square pieces between 200 and 400 μm per side.

The carriers (either AG1-X2 beads or pieces of membrane) were loaded with all-trans-retinoic acid (Sigma; batch numbers: 12F-0598; 63F-O476; 104F-0135) by a 20 min soak in a solution of the retinoid dissolved in dimethyl sulphoxide (DMSO). The carriers were then rinsed in minimal essential medium (MEM)-I-10% fetal calf serum (Gibco-Biocult) as previously described (Tickle et al. 1985). As controls, the carriers were placed in DMSO and then rinsed following the same protocol. The carriers were then implanted beneath the apical ridge at the apex of wing buds of stage 20 chick embryos (Hamilton & Hamburger stages). A slit was made at the base of the apical ridge and the ridge eased away from the mesenchyme to form a loop. The carrier was placed between the apical ridge and mesenchyme and held in place by the loop of epithelium.

To monitor the effects of retinoid treatment on wing pattern, the embryos were left to develop for a further 6 days. For comparison, carriers were also implanted beneath the apical ridge at the anterior margin of the bud in some experiments and the wing pattern examined. The wings were removed, fixed in 5 % trichloroacetic acid (TCA) and stained with alcian green. The pattern of skeletal elements was observed in whole-mount specimens cleared in methyl salicy-late.

At various times following implantation of carriers beneath the apical ridge at the bud apex, the buds were fixed in 1/2 strength Karnovskys fixative (Karnovsky, 1965) at 4 °C over-night. When AG1-X2 beads were used, these were sometimes removed from the buds with a sharp needle prior to fixation. The tissue was then rinsed in 0T M-cacodylate buffer. Following dehydration in a series of graded alcohols, the tissue was cleared in propylene oxide and embedded in Araldite. Sequential 1 or 2 gm sections were taken in a plane at right angles to the anteroposterior axis, starting at the anterior margin of the bud (Fig. 1), and stained with toluidine blue. Measurements of the height of the apical ectodermal ridge and distance between the carrier and the apical ridge were made using a calibrated eye-piece graticule.

For electron microscopy, wing buds to which anion exchange membrane had been implanted, were used. The tissue was postfixed in 1 % osmium tetroxide in 0·1 M-cacodylate buffer, embedded in Araldite and sectioned as before. At appropriate levels, the block was trimmed down and ultrathin sections were taken. The sections were stained with lead citrate (Reynolds, 1963) and viewed in a Philips 300 electron microscope.

Recombinations were made between tissues of treated and untreated wing buds. To separate the epithelium and the mesenchyme, buds were placed in 2% trypsin (Difco 1:250) for 1-1·5 h at 4°C. The treated tissues were obtained from the right wing buds in which beads had been implanted and the control tissues were either from contralateral left wing buds or from leg buds. Epithelial jackets from treated buds were recombined with the mesenchyme from the tip of untreated wing buds. In reciprocal combinations, the bead was removed from the mesenchyme of treated wing buds and this was then recombined with control epithelial jackets taken from leg buds. The tissues were recombined by an incubation of 0·5 h at room temperature followed by a further hour at 38°C. The tissue recombinations were then grafted to a bed cut in the dorsal surface of stage 23 wing buds to continue development. A few grafts were fixed 1 and 2 day(s) later, in 1/2 strength Karnovsky’s and processed for histology (as before). The majority of the grafts were fixed 6 days later. The host wing buds bearing the grafts were fixed in TCA and stained with alcian green to show the pattern of cartilage elements (as above).

With beads soaked in increasing concentrations of retinoic acid, the height of the epithelium progressiely decreases. Fig. 2 shows the height of the epithelium that tips the bud 21-24 h after bead implantation plotted with respect to the concentration of retinoic acid in which the bead was presoaked. With control beads, the average height of the epithelium in thesection that passes through the bead midpoint is 40 μm which agrees quite well with the data of Todt & Fallon (1984) for the height of the ridge in normal buds. There is a prominent thickened apical ectodermal ridge (Fig. 3A,B). With beads soaked in l mg ml^-1 retinoic acid, the thickened ridge epithelium is no longer discernible. Instead the epithelium is smoothly continuous with that on the dorsal and ventral surfaces of the bud and of more-or-less uniform height (Fig. 3C,D). Fig. 4 shows the height of the epithelium at the tip of the bud plotted with respect to the position of the implanted bead. With beads soaked in 0·01 mg ml^-1 retinoic acid, the reduction in ridge height appears to extend for about 400-500 μm on either side of the bead, whereas with higher concentrations, the epithelium across the entire anteroposterior axis is reduced in height.

The height of the apical ridge is related to the extent of outgrowth and the patterns that result. Table 1 lists the resultant wing patterns obtained. With beads soaked in 0·01 mg ml^-1 retinoic acid, duplicated patterns of digits result, whereas implantation of beads presoaked in l mg ml^-1 retinoic acid leads to limb truncations and no digits develop (see also Tickle & Crawley, 1988).

The time course of morphological changes in the apical ectodermal ridge at both light microscopical and ultrastructural levels was investigated following retinoid application. Since the height of the epithelium is most reduced by high doses of retinoic acid that lead to limb truncations (see above), we chose to study the time course of epithelial changes in such buds. However, AG1-X2 beads could not be used as carriers because, if left in place, the beads changed in volume during histological processing and often distorted the host tissue (Fig. 3). In addition, they could not be removed at short times after implantation without damaging the epithelium at the tip of the bud. Pieces of anion exchange membrane were found to be more suitable carriers because they withstood histological processing.

Before using anion-exchange membrane carriers to examine histological changes in the apical ridge, we had to establish a retinoid treatment regime that reproducibly leads to inhibition of bud outgrowth and subsequent limb truncation. Table 2 shows the wing patterns that resulted following implantation of pieces of anion-exchange membrane soaked in 10 mg ml^-1 retinoic acid to the apex of stage 20 wing buds. The effect depends on the size of the piece of membrane. With larger pieces (300-400 μm per side), truncations usually result. The most common truncation is the failure of development of all structures distal to the middle of the humerus. With smaller pieces of membrane soaked in the same retinoid concentration implanted at the apex, digits always develop. However, when the retinoid is applied to the anterior margin of the wing bud, duplicated digit patterns develop irrespective of the size of the piece of membrane.

To explore the time course of the effects on the apical ectodermal ridge and their relationship to bud out-growth, buds were fixed at 8, 16 or 24 h after implantation of the larger pieces of membrane presoaked in retinoic acid and sectioned. As controls, pieces of anion exchange membrane presoaked in DMSO were implanted. The number of wing buds sectioned is shown in Table 3.

The position of the carrier shows that outgrowth is inhibited in retinoid-treated buds. At 8h, the carrier still protrudes from the surface of the limb but by 16 h the piece of membrane has almost completely healed into the developing limb bud. The distance of the membrane from the mesenchymal-epithelial border was measured in thick Araldite sections of both retinoid-treated and control buds and the data are shown in Fig. 5A. Up to the 16 h time point, the position of the carrier in the host limbs appears to be the same. However, the distance between the implanted membrane and the mesenchymal-epithelial border hardly increases in the retinoid-treated buds between 16 and 24 h. In contrast, in control buds considerable outgrowth has occurred over the same time period and the distance of the membrane from the tip of the limb is much greater at 24 h compared with 16 h.

By 16 h, the height of the epithelium at the tip of retinoid-treated buds is reduced. Fig. 5B shows quantitative data on the height of the epithelium at the tip of the wing buds after implantation of the pieces of membrane. At 8h, in both control and retinoid-treated buds, there is a well-defined thickening in the epithelium at the tip of the bud (see later Fig. 6). However, at 16 and 24 h after implantation of the membrane, the epithelium at the tip of the control buds is much taller than that at the tip of the retinoid-treated buds. A small reduction of ridge height occurs in buds containing DMSO-treated carriers at 16 h, but the ridge appears to have returned to its original height by 24 h.

Pronounced cell death occurs shortly after implantation of the piece of membrane. At 8 h, signs of cell death are observed both in the periderm and the inner pseudo-stratified layer of the ridge in both control and retinoid-treated buds (Fig. 6A,B). Many of the epithelial cells contain electron-dense inclusions which are the result of ingestion of dying and dead cells. By 16-24 h, most of the cells look healthy and do not contain inclusions (see later Figs 7 & 8).

To find out if there is any difference in the extent of cell death in the epithelium of retinoid-treated and control buds, we counted the number of cells with inclusions in low-power electron-micrograph montages and the data are shown in Table 4. We detected no consistent difference between retinoid-treated and control limb buds. However, the number of cells with inclusions is much higher in the periderm than in the underlying pseudostratified layer in all cases.

By 16-24 h, there is a striking difference in the arrangement of the cells in the epithelium of retinoid-treated and control buds. In control buds, the ultrastructural features of the epithelium closely resemble those previously described for the ridges of normal buds; the cells in the thickened epithelium are tightly packed (Fig. 7A,B) and linked by extensive gap junctions (Fallon & Kelley, 1977; Fig. 7B,C). Desmosomes are present and are particularly associated with the borders of periderm cells but can also be found between cells deeper within the epithelium (Fig. 7D). In contrast, when the membrane has been presoaked in retinoic acid, the buds are tipped with a cuboidal epithelium in which the cells are loosely packed (Fig. 8A,D). The inner layer of epithelium below the periderm is 1-2 cells thick. Desmosomes are found both between periderm cells and the cells of the inner layer (Fig. 8C,B). However, no gap junctions are seen. In this respect, the epithelium at the bud tip resembles the nonridge limb ectoderm, in which gap junctions are small and scarce (Fallon & Kelley, 1977).

Striking changes in the cell packing and cell-cell contacts have occurred by 16 h in the epithelium of retinoid-treated buds. However, at 8h, in both reti-noid-treated and control buds, the epithelium at the tip of the bud is pseudostratified and consists of elongated tightly packed cells sitting on a well-defined basement membrane. To find out if we can detect any changes in cell-cell contacts at this early time, we examined the ridge in both retinoid-treated and control buds in more detail. We scanned the borders between epithelial cells and counted the number of gap junctions and desmosomes. Desmosomes are readily recognized (Fig. 9), but only extensive gap junctions, more than about 500 nm in length can be scored (Fig. 9). Furthermore, in some cases, the identity of gap junctions could only be confirmed by examining sequential ultrathin sections. Table 5 shows the quantitative data. The frequency of both types of junctions is reduced in the retinoid-treated ridges compared with control ridges. However, the frequency of gap junctions even in control ridges is lower than that reported by Kelley & Fallon (1983) in the ridges of normal buds using freeze-fracture techniques to identify gap junctions. The data on desmosome number show that this is reduced in both the periderm and underlying pseudostratified layer following retinoid treatment although our analysis shows that desmosomes occur much more frequently between periderm cells than between the cells in the rest of the epithelium.

The changes in epithelial organization that follow retinoic acid treatment could result from either a direct action of the retinoid on the epithelium or an indirect effect mediated by the mesenchyme. To investigate these possibilities, tissue recombinations were carried out. AG1-X2 beads were used to apply retinoic acid rather than anion exchange membrane because the beads can be cleanly removed from the bud tissue. The dose-time relationships necessary for irreversible changes in the outgrowth of treated buds were determined by implanting AG1-X2 beads soaked in either 0·lmgm1¹ or I mg ml^-1 retinoic acid to the apex of right wing buds and the length of exposure was varied by removing beads at a series of times after implantation. The wing pattern at 6 days after treatment can be used as an assay for the extent of outgrowth. The data in Table 6 show that the effect on wing pattern is completely irreversible after 21 h if the bead is presoaked in I mg ml^-1 retinoic acid. Irrespective of whether the bead is removed at 21 h or left in place, most buds develop into severely truncated limbs that consist of a shoulder girdle and a small fragment of cartilage. At this time point, the histological examination of sections of buds (which we described earlier) showed that the apical ridge has completely disappeared. Using beads soaked in the same retinoid concentration, a short treatment period (5h) has no effect on bud outgrowth, and an intermediate length of treatment (16 h) leads to the continued formation of some structures.

The data in Table 6 and Fig. 2 show that, 21 h after implantation of beads soaked in 1 mg ml^-1 retinoic acid, the effect on outgrowth of treated buds is irreversible and the apical ridge has disappeared. Therefore, tissues from treated buds at this time point can be used in combination with tissues from normal buds to explore whether the morphology of the ridge is dictated by the mesenchyme.

Table 7 shows the results of recombining tissues from retinoid-treated and control limb buds. When ectodermal jackets taken from buds 21 h after implantation of beads soaked in 1 mg ml^-1 retinoic acid are recombined with distal mesenchyme from control wing buds, digits develop in 10/10 cases (Fig. 10A). The grafts appear to develop autonomously and there is no interaction with the host limb. Furthermore, sections of the grafts at 1 day after recombination show that an apical ridge is now present at the tip of the ‘bud’ (2/2 cases - Fig. 10B) and can also be seen 2 days after recombination of treated epithelium and control mesenchyme (1 case). In contrast, recombinations of the mesenchyme from treated buds with control ectodermal jackets from leg buds do not give rise to digits (Table 7). Instead, small fragments, spikes or rods of cartilage develop. In 2/7 cases, the rods resemble the morphology of the humer-

As controls, tissues from buds treated for 5h with retinoic acid were also combined with tissues from untreated buds (Table 7). In this case, nearly all the recombinations of treated and untreated tissues continue outgrowth and distal structures develop (6/7 cases). Recombinations of treated mesenchyme and untreated leg epithelium give rise to a complete proximodistal set of wing structures (Fig. 10D).

The recombination experiments show that retinoic acid acts directly on the mesenchyme and 21 h later irreversible changes have taken place. The organization of the epithelium is dictated by the mesenchyme.

We have identified the mesenchyme of the limb bud as the tissue which is the primary target of retinoid action. Changes in the mesenchyme then lead to reorganization of the apical ridge. Therefore, we have begun to dissect the way in which patterning is linked to morphogenesis in the developing limb. The changes in ridge height are dose-dependent and we can demonstrate that the morphology of the apical ridge is related to bud outgrowth. The reorganization of the ridge may involve changes in cell-cell communication.

Recombinations of normal epithelium and treated mes-enchyme taken at a time when outgrowth of the bud is irreversibly inhibited do not develop distal structures. We can exclude the possibility that this failure of development is due to the retinoic acid contained in the treated mesenchyme. When mesenchyme is taken from the buds after a short exposure to the retinoid and recombined with normal epithelium, distal structures develop. Since the concentration of applied retinoic acid in the bud tissue is higher at early times after bead implantation than later on (Eichele et al. 1985), this experiment acts as a control to assess the effects of carry-over of retinoid in treated tissues. Therefore, we can conclude that the disappearance of the ridge following retinoid treatment must be mediated by irreversible changes in the mesenchyme. To understand the mechanisms of retinoid action, we now need to focus on the cellular and biochemical changes that take place in the mesenchymal cells of the limb bud.

An interaction between mesenchyme and epithelium that controls the morphology of the ridge in normal limb development is a crucial part of the ‘Saunders-Zwilling hypothesis’ (Zwilling, 1961; Saunders & Gasseling, 1968) and has been well documented. For example, when an ectodermal jacket is rotated through 180° and replaced on its mesenchymal core, the limb grows out and a normal sequence of digits develop (Zwilling, 1956). The morphology of the epithelium at the tip of the bud changes to conform with the mesenchyme (MacCabe & Parker, 1979). Some progress has been made in isolating a factor, which has been called apical ectodermal ridge maintenance factor, thought to be responsible for ridge maintenance (MacCabe & Richardson, 1982). The distribution of this protein would have to be graded along the anteroposterior axis to give the regional differences in ridge morphology. More recently, it has been found in culture experiments that insulin maintains the apical ectodermal ridge (Boutin & Fallon, 1984).

The interactions between mesenchyme and epithelium are not confined to the stages at which limb buds grow out. Interactions may also occur in prebud tissues that determine the ability of the mesenchyme to maintain the ridge as shown by studies on the wingless mutant (Carrington & Fallon, 1984).

Experiments in which the polarizing region is cut out of developing limb buds have given conflicting results which appear to depend on the amount of tissue removed (see for example, Fallon & Crosby, 1975). When extensive regions of the posterior of the bud are cut out, the apical ridge disappears and outgrowth is abolished (Hinchliffe & Gumpel-Pinot, 1981). This suggests that the signal from the polarizing region is necessary for ridge maintenance but does not distinguish between a direct or indirect action. Furthermore, in the normal bud, the epithelium that immediately overlies tissue with polarizing activity does not have a ridge-like morphology.

The results of the experiments reported here show that the applied retinoic acid does not directly modulate epithelial morphology. Therefore, if retinoic acid is the positional signal produced by the polarizing region, this must generate a second factor that maintains the ridge. However, the chain of signals could be more complex. For example, the high points of both the experimentally established and the endogeneously occurring gradients of retinoic acid (Tickle et al. 1985; Thaller & Eichele, 1987) may in turn trigger a signal as yet unidentified that specifies position across the anteroposterior axis of the limb. Finally, there is the intriguing possibility that the retinoid signal could operate in an autocatalytic fashion and that a primary response to the signal involves changes in retinoid metabolism in the mesenchyme.

In amphibian limb regeneration, in which retinoic acid also brings about changes in pattern, the mesenchyme rather than the epithelium appears to be affected (Maden, 1984). In addition, there is also a striking parallel between the effects of the retinoid treatment of wing buds described here, and those on a facial primordium, the frontonasal mass, that gives rise to the upper beak. Outgrowth and patterning in the frontonasal mass are inhibited when retinoic acid is applied to chick embryos (Tamarin et al. 1984). Recombination experiments show that it is the mesenchyme of the frontonasal mass that is affected by retinoid treatment (Wedden, 1987).

Our data show that the height of the apical ridge changes in a dose-dependent manner when retinoids are applied to chick wing buds. Therefore, our simple hypothesis (Lee & Tickle, 1985) that a threshold retinoid concentration determines an all-or-nothing response of the ridge - either maintenance or disappearance - appears to be incorrect. Using beads implanted at the apex of the bud, the reduction in ridge height over a range of retinoid concentrations suffices to truncate the limb. Only with beads soaked in the lowest doses of the retinoid, does the ridge appear to maintain sufficient height over a significant part of the limb bud margin to lead to sustained outgrowth.

When carriers soaked in appropriate concentrations of retinoic acid are implanted at the apex of wing buds, the apical ridge disappears across the entire bud and outgrowth is inhibited. The quantitative data obtained by monitoring the height of the epithelium at the tip of the bud and the position of the carrier after retinoid application allow a direct comparison between ridge morphology and bud outgrowth. When carriers are presoaked in retinoic acid, the height of the epithelium is dramatically reduced at 16 h and is still flat at 24 h. Between these two time points the carrier remains at the same distance from the epithelium. In contrast, in control buds, the ridge is slightly reduced in height at 16 h but is back to normal at 24 h. Accompanying the reestablishment of a taller ridge, the distance of the carrier from the epithelial-mesenchymal border increases dramatically between 16 and 24 h. These results clearly show that the morphology of the ridge is directly related to the outgrowth of the bud.

Inhibition of outgrowth and the failure to lay down distal structures occurs when the apical ectodermal ridge is surgically removed (Saunders, 1948; Summerbell, 1974). Therefore, the disappearance of the ridge following retinoid treatment could be due to cell killing followed by healing of the ventral and dorsal ectoderm over the limb bud tip. It has been suggested, in mouse embryos, that the limb defects produced by retinoid treatment are due to cell death in the apical ectodermal ridge (Sulik & Dehart, 1988). However, we cannot detect any difference in the extent of cell death between the ridges of treated and control buds. The cell death that does occur appears to be the result of the cutting of the ectoderm which is necessary to insert the carrier combined with the normally occurring cell death that has been observed in intact limb buds (Todt & Fallon, 1984).

When the epithelium from buds in which irreversible changes in outgrowth have been induced is recombined with mesenchyme from the tip of normal limb buds, a thickened epithelium is re-established and distal structures develop. Therefore, the retinoid effects on the epithelium of buds can be modified, even though outgrowth of the intact treated buds is irreversibly halted. The reappearance of the pseudostratified epithelium and resumption of outgrowth is accompanied by the specification of distal structures in the recombinations. This strongly supports the idea that the disappearance of the ridge is due to epithelial reorganization and not cell killing.

During disappearance of the ridge, the epithelial cells become more loosely packed and change in shape. In some cases, similar changes in epithelial morphology have been observed in the epidermis when skin is treated with retinoids. Retinoic acid treatment of adult mouse skin leads to an increase in intercellular spaces between the epidermal cells and a decrease in the frequency of desmosomes (Barnett & Szaba, 1973). However, when embryonic chick skin is treated, the number of desmosomes appears to be unaffected but there is an increase in the number of gap junctions (Elias & Friend, 1976).

In other systems, the organization of the epithelium has been shown to depend on extracellular components. For example, when Sertoli cells are cultured on plastic in serum-free defined medium, they form a monolayer of squamous cells, whereas, on thin reconstituted basement membranes, they form a columnar epithelium (Hadley et al. 1985). We did not observe any obvious morphological changes in the basement membrane of retinoid-treated ridges but we cannot exclude the possibility that alterations in basement membrane composition play a role in the modification of epithelial morphology.

Our analysis of the ultrastructure of the ridge after a short exposure to retinoic acid suggests another mechanism that may trigger epithelial reorganization. At 8h after treatment, the ridge cells are still tightly packed and elongated but there appears to be a reduction in the number of gap junctions between the cells. When early mouse embryos are treated with a gap junction antibody that prevents cell-cell communication, the cells de-compact (Lee et al. 1987). This suggests that cell-cell communication could be involved in the maintenance of a tightly packed epithelium. In this context, our observations on the ridge in retinoid-treated embryos are suggestive. However, we do not know whether the gap junctions have disassembled or whether they have broken up into smaller units that would not have been detected in our analysis. Furthermore, we need to understand how the functional coupling of the epithelium changes. Finally, we do not know whether the reduction in the number of gap junctions is a primary or secondary event and how this related to the accompanying decrease in desmosomal contacts. Nevertheless, the apparent reduction in the number of gap junctions in the ridge following retinoid treatment suggests that this may be an important step in the generation of a more loosely packed epithelium.

This work is supported by a grant from the MRC. J.F. was supported by an MRC award for an intercalated BSc. We are grateful to Professors L. Wolpert and A. Warner, Drs S. Wilde and F. Allen for their suggestions and helpful advice and to Dr M. Noble for his critical reading of the manuscript.