Transfilter studies on the mechanism of epithelio-mesenchymal interaction leading to chondrogenic differentiation of neural crest cells

In all types of vertebrate embryo studied to date it has been found that cranial neural crest (NC) cells migrate extensively within the developing head, give rise to much of the mesenchyme and later differentiate into various connective tissues. As a consequence there is a major contribution by NC-derived cells to the cranial and facial skeleton (Le Lievre, 1978; Morriss & Thorogood, 1978; Noden, 1978). It has been demonstrated in the avian embryo that NC cells are not ‘committed’² to a skeletogenic fate before migration and that instead they require tissue interactions with epithelia encountered during migration or at the presumptive site after migration if chondrogenic or osteogenic differentiation is to ensue (Bee & Thorogood, 1980; Thorogood, 1981). The interaction between mesencephalic NC cells and retinal pigmented epithelium (RPE) to form the scleral cartilage of the avian eye is one such epithelio-mesenchymal interaction and has been extensively studied at the phenomenological level (Weiss & Amprino, 1940; Amprino, 1951; Reinbold, 1968; Newsome, 1972; Stewart & McCallion, 1975). NC cells migrate to the optic cup by 52 h of incubation (stage 14 - Hamburger & Hamilton, 1951) and condense to form the periocular mesenchyme (Noden, 1975, 1978). This mesenchyme is committed to chondrogenic differentiation by to 4 days (Newsome, 1972) but the actual onset of chondrogenesis is not until day 7 (Romanoff, 1960) indicating that the RPE influence is not required to maintain the chondrogenic potential of the ectomesenchyme.

Newsome analysed the mechanism of interaction between RPE and NC and concluded that the mechanism was matrix-mediated (Newsome, 1976). This conclusion was based on the fact that NC or periocular mesenchyme cells formed cartilage when grown on filters bearing a distilled water lysate of cultured RPE cells which, it was claimed, was composed principally of extracellular matrix (ECM). However, in addition to matrix this lysate retained cytoskeletal components and membrane fragments (see fig. 4, Newsome, 1976) and hence these experiments only demonstrate that living RPE cells need not be present for the interaction to occur. However, there are at least three mechanisms (Saxen, Ekblom & Thesleff, 1980) through which epithelio-mesenchyme interactions could be mediated:

1. Transmission of signal substances by long range diffusion.

2. Action of extracellular matrix materials

   • If the morphogenetically active substance can move away from the matrix and the basement membrane of the epithelium then this would be an example of free diffusion of matrix components in the extracellular compartment.

   • If the active molecules are a structural part of the matrix and are unable to detach from it direct contact is required between the extracellular matrix and the responding cells.

3. Interaction mediated by direct cell to cell contact, i.e. direct plasmalemmal contact of epithelium cells with mesenchyme cells. The basement membrane of the epithelium must be discontinuous for direct cell contact of the two cell populations to occur.

To investigate more thoroughly the mechanism of interaction between RPE and NC we adopted the transfilter culture technique which has been used to analyse tissue interactions in amphibian primary induction (Toivonen & Wartio-vaara, 1976; Toivonen, Tarin & Saxen, 1976); avian limb bud chondrogenesis (Gumpel-Pinot, 1980,1981); lens induction (Karkinen-Jääskelainen, 1978); cornea development (Hay & Meier, 1976); early mammalian tooth morphogenesis (Thesleff, Lehtonen, Wartiovaara & Saxen, 1977) and kidney tubule formation (Wartiovaara, Nordling, Lehtonen & Saxen, 1974; Saxen et al. 1976). In essence the technique involves the cultivation of the interacting tissues on opposite sides of barrier membranes of known and constant pore size. Tissues can be combined across such membranes whose pore size is known to permit or prevent the passage of cell processes. Polycarbonate Nuclepore membranes are used in which the pores are straight-through channels resembling bullet tracks and thus if the filter is correctly orientated during sectioning it is possible to examine the whole length of individual pores for the presence or absence of matrix components and cell processes traversing the filter.

In order to study the mechanism of tissue interaction in the present system premigratory NC cells were grown transfilter to RPE explants of different developmental ages and on different pore size filters. Scanning electron microscopy observations of the undersurface of Nuclepore membranes on which NC or RPE explants had been grown for 24 or 48 h established which pore sizes allowed penetration of cell processes. Pores of 0·8 μm were found to allow the passage of cell processes through the filter whereas 0·2 μm pores did not allow such penetration. It was assumed that both pore sizes allow soluble ECM components to diffuse through the filter. If chondrogenesis occurred in transfilter cultures grown on Nuclepore filters that do not allow penetration of cell processes, this would indicate that the mechanism is mediated by free diffusion of extracellular matrix components or other signal substances. However, if differentiation occurred only in cultures grown on filters with pore sizes that allow penetration by cell processes, this would demonstrate that either direct plas-malemmal contact of RPE and NC cells or direct contact between RPE non-diffusable matrix components and NC cells is necessary for the interaction to occur.

All embryos used in this study were from eggs of the domestic fowl (Gallus domesticas), White Leghorn × Rhode Island Red, obtained locally (Ross Poultry, U.K.) or from eggs of the Japanese quail (Coturnix coturnix japónica). Eggs were incubated in a humid atmosphere in a forced draught incubator and at a temperature of 38 ± 0·5 °C; all embryos were staged according to the developmental table of Hamburger & Hamilton (1951).

Initial dissections were carried out in sterile Dulbecco ‘A’ phosphate-buffered saline (PBS) (Oxoid Ltd., England). To obtain neural crest tissue, stage-9+ embryos were removed from the vitelline membrane, placed in ‘alpha’ Eagle’s minimal essential medium (αMEM) supplemented with 10 % foetal calf serum (FCS), and the mesencephalic neural folds dissected out according to Bee & Thorogood (1980) (and see fig. 3 in Thorogood, 1981). Stage-18 periocular mesenchyme was dissected from the medial aspect of the optic cup of stage-18 embryos. Stage-17 presumptive RPE and stage-24 RPE and associated neural retina and periocular mesenchyme were dissected from the medial aspect of the optic cup (stage 17) or eye (stage 24) and incubated in sterile 0·25 % trypsin/ 1·25 % pancreatin in calcium- and magnesium-free Tyrodes solution for 30 min at 4 °C. Further enzyme activity was then blocked by adding an equal volume of FCS and the RPE was then dissociated from the adjacent tissues by mechanical separation.

Tissues were grown on a Nuclepore filter (marketed as ‘Unipore’ membrane filters by Bio-Rad Laboratories, U.K.) supported by a stainless steel grid platform with a central hole to facilitate observation. This assembly was contained within a 30 mm plastic culture dish containing 2·5 ml αMEM+ 10% FCS + 100units/ml penicillin and 100μg/ml streptomycin. To maintain humidity the culture dishes were placed inside a glass bacteriological dish, the bottom of which was covered with filter paper in which holes had been cut to accommodate two of the smaller dishes. The filter paper of the larger dish was then saturated with sterile distilled water containing 1 % Fungizone (Gibco, U.K.).

To establish the standard transfilter cultures pieces of stage-24 RPE or stage-17 RPE were placed individually on the Nuclepore filter and cultured for 48 h. A drop of agar (1 % in αMEM) was then put on the RPE with a micropipette, and the filter inverted. A fragment of premigratory mesencephalic NC or stage-18 periocular mesenchyme was placed transfilter to the RPE and the combination culture cultured for 12 days at 37 °C in an atmosphere of 5 % CO₂ in air, with a complete medium change every 3·4 days. After 11 days of culture a drop of 1 % agar in aMEM was put on the NC component to prevent the tissue being dislodged during processing. The culture was then fixed at 12 days and embedded in paraffin wax for histology. In transfilter cultures chick stage-9 NC or stage-18 periocular mesenchyme was grown in combination with quail RPE so that if necessary in later experiments the characteristic nucleolar marker of the quail cells (Le Douarin, 1973) could be employed as a marker for the cells of the quail tissue component. The apparently normal development of experimental chick chimaeras testifies to the absence of any species barrier to normal development in heterospecific combinations (Noden, 1975; Le Lievre & Le Douarin, 1975).

Transfilter culture and control explants were fixed in 80 % ethanol, stained with 1 % eosin in 80 % ethanol and dehydrated in an ethanol series, cleared in toluene, embedded in 56°C m.p. paraffin wax and serially sectioned at 5-6μm. Sections were stained with alcian blue and durazol red; Hansen’s haematoxylin was used as a nuclear stain. Any cartilage or bone was identified by the blue- and red-stained properties of their matrices and their characteristic morphology.

Sections were photographed with a Zeiss Olympus photomicroscope using Ilford Pan F.

After 24 or 48 h of culture explants of neural crest, stage-18 periocular mesenchyme or stage-17 RPE grown alone on 0·8 μm or 0·2μm Nuclepore filters were washed gently and briefly in sterile PBS and fixed for 45 min in 2·5 % glutaraldehyde in 0·1 M-sodium cacodylate buffer pH 7·3 at room temperature. In order to preserve extracellular matrix (ECM) components 0·5% w/v of cetyl pyridinium chloride (CPC) was added to the fixative when fixing RPE. The explants were then washed in three changes of 0·2 M-buffer pH 7·3, post fixed for 1 h in 1 % osmic acid in 0·1 M-cacodylate buffer pH 7·3, washed in distilled water and dehydrated through a graded series of ethanol (Gumpel-Pinot, 1980). The specimens were critical-point dried using liquid CO₂ and mounted on stubs so that either the tissue or the underside of the filter opposite to the tissue was visible, coated with gold and palladium in an SEM-PREP sputter coater and observed in a JEOL JSM-P15 Scanning Electron Microscope.

Transfilter combinations that had been cultured for 5 days were washed gently in PBS, then fixed for 2h in 2·5% glutaraldehyde in 0·1 M-cacodylate buffer+ 0·5 % CPC pH 7·3 at room temperature.

Specimens were then washed and dehydrated as described above, cleared in toluene and embedded in paraffin wax (Kaprio, 1977). The cultures were then sectioned at 10μm. Alternate sequences of 10–20 sections were then mounted on slides for standard histological analysis or attached to 11 mm small round glass cover slips (for SEM analysis) with a 1 % gelatin solution. Sections for SEM analysis were then dewaxed in three changes of toluene, dehydrated in a 1:1 toluene: ethanol mixture followed by three changes of absolute ethanol, critical point dried and coated with gold and palladium as described above.

Single explants of premigratory mesencephalic NC, stage-18 and stage-24 periocular mesenchyme and stage-17 and stage-24 RPE were each grown alone in organ culture for 12 days (see Table 1). Cultures of isolated NC and stage-18 periocular mesenchyme consisted predominantly of small patches of unpigmented mesenchyme cells; patches of melanocytes were sometimes present in NC cultures. In a small number of NC cultures some mesenchyme cells had an epithelioid morphology and were organized into vesicular structures; the lumen of such vesicles usually contained material which stained weakly with alcian blue. Stage-24 periocular mesenchyme control cultures grew into large masses of tissue containing a range of differentiated cell types, namely cartilage with an incidence of 96 %, bone – 65 %, melanocytes – 100 %, and muscle – 52 % (see Fig. 1). Stage-17 RPE cultures consisted of flattened masses of unpigmented, undifferentiated cells whereas stage-24 RPE cultures consisted of large masses of pigmented and unpigmented cells. The RPE of both developmental ages sometimes contained neural retina (NR). This had arisen during the culture period as care had been taken to explant RPE alone and it is well established that RPE is able to transdifferentiate into NR in culture (Tsunematsu & Coulombre, 1981). ECM was sometimes present on the underside of 0·8 μm filters but never when the tissue was grown on 0·2μm filters (see later).

These experiments demonstrate that premigratory NC, stage-18 periocular mesenchyme, stage-17 and stage-24 RPE are incapable of forming cartilage when grown alone in organ culture. However cultures of stage-24 periocular mesenchyme, which has been shown to be committed to chondrogenesis (Newsome, 1972), regularly formed cartilage thus demonstrating that the culture conditions used in this study are permissive for chondrogenic differentiation.

a) In order to establish whether there are differences in the behaviour of neural crest cells grown on 0·8 μm and 0·2 μm filters, single NC explants were cultured on 0·8 μ m and 0-2μm Nuclepore filters for 24 or 48 h and the lower or upper surfaces of these filters were then examined using SEM.

Observations on the upper side of 0·8 μm and 0·2 μm filters on which NC had been cultured revealed a layer of loosely associated cells which were often sperical and had many filopodia extending from them (see Fig. 2). Examination of the upper side of 0·8μm filters revealed filopodia from crest cells extending down through the pores of the filter (see Fig. 3). Sometimes many cell processes were observed extending from a single cell. Identical filopodia were extended by cells grown on 0·2μm filters but were never observed extending into the smaller pores, instead the ends of the filopodia rounded up (see Fig. 4).

Predictably, cell processes were never observed projecting out of the pores on the underside of 0·2 μm filters (see Fig. 5) but were regularly observed on the underside of 0·8 μm filters on which NC had been cultured for 24 h or 48 h (Fig. 6). The density of cell processes observed appeared to be the same at these two time intervals. Typically they comprised a filopodium which could be seen within the pore and which then bulged to form a bulbous end on the undersurface of the filter (see Fig. 7).

b) In order to establish whether stage-17 RPE deposited different amounts of ECM through the pores of 0·8 or 0·2 μm Nuclepore filters single stage-17 RPE explants were cultured on these filters for 24 h or 48 h. The lower surfaces of these filters were then examined noting the relative abundance and organization of any extracellular matrix. The experimental technique was the same as that used for other control cultures, with the addition of CPC to the glutaraldehyde fixative in order to preserve ECM components.

Neither cell processes nor extracellular matrix were observed on the lower surface of 0·2μm filters on which RPE had been cultured for 24 or 48 h. However, since ECM was observed between the RPE and the upper surface of the filter in such cultures (not shown) it is possible that soluble ECM components do pass through the pores of the filter but in the absence of cell processes may diffuse into the culture medium rather than consolidate into a matrix. Small quantities of ECM associated with a few cell processes were present on the lower surface of 0·8 μm filters on which RPE had been cultured for 24 h. Larger quantities of matrix including sheet-like, granular and fibrous components were observed at 48 h and all types of matrix components were usually associated with cell processes (Fig. 8). The ECM was localized in the region of the tissue and was often well organized (see Fig. 9). The unexpected presence of cell processes in such cultures is probably the result of the temporary removal of the basement membrane during tissue dissociation and prior to its regeneration in culture.

If the tissue is fixed without CPC included in the fixative, far less ECM is retained on the underside of the filter and consequently cell processes are far easier to observe. With RPE cultured on 0·2 μm filters there was little difference due to the relative paucity of matrix in such cultures but with cultures on 0-8 μm filters the removal of much of the abundant matrix revealed the presence of many cell processes. These varied in morphology from individual processes to groups of closely associated large bulbous structures often covered with small vesicles (see Fig. 10). Any matrix that was retained was associated with these groups of cell processes. The association of vesicles and ECM with cell processes may reflect the involvement of such processes in the synthesis and organization of matrix.

Mesencephalic NC and stage-17 or stage-24 RPE were cultured in transfilter combination on 0·8 μm or 0·2μm Nuclepore filters for 12 days to establish whether the interaction between NC and RPE is mediated by diffusable factors or cell contact.

Regardless of the developmental age of the RPE at the outset of the experiment there was virtually no cartilage formation in transfilter cultures grown on 0·2 μm Nuclepore filters (see Fig. 11 and Table 2). In marked contrast chondrogenesis did occur in transfilter cultures on 0·8 μm filters (see Figs 12 and 13 and Table 2) but the incidence of differentiation in the NC component was related to the age of RPE used. When stage-17 RPE was used as the epithelial component a dramatically improved growth of NC and higher incidence of cartilage formation compared to that in stage-24 RPE/NC transfilter cultures was observed (see Figs 12 and 13). These results suggest firstly that stage-17 RPE is a more potent stimulator of chondrogenesis than stage-24 RPE and secondly that the mechanism of interaction between RPE and NC is not diffusion-mediated and is instead mediated either by contact of NC cells with a non-diffusable matrix component of the RPE or by direct plasmalemmal contact of NC cells with RPE cells (see Discussion). The RPE component in transfilter cultures had the same histological appearance as RPE controls. Both stage-17 RPE and stage-24 RPE always grew well and sometimes contained NR. Cartilage was found in cultures containing both RPE and NR and also in cultures in which the neural tissue was absent. Since cartilage was seen to develop in the absence of NR it is clear that the development of cartilage is related to the presence of RPE.

Cartilage always formed against the filter in transfilter cultures (see Figs 12 and 13) which suggests that NC formed cartilage in response to RPE on the other side of the filter. In contrast the cartilage that formed in stage-24 periocular mesenchyme did not always form adjacent to the filter (see Fig. 1). Examination of serial sections revealed that in both types of culture cartilage always formed in plates or nodules - never in rods. The growth of the crest component was variable. Occasionally in both stage-17 RPE/NC and stage-24 RPE/NC cultures the NC grew very well but remained undifferentiated (Fig. 14), whereas in other cultures the NC grew far less but nevertheless differentiated into cartilage. Therefore, ectomesenchymal mass appears to be unrelated to whether or not cartilage formation occurs although cell packing density may be.

This experiment was designed to investigate the contact relationships between NC and stage-17 RPE cells when grown together in transfilter culture on 0 · 8 μm filters. Standard transfilter cultures were fixed at 5 days for SEM, embedded in wax and sectioned at 10 μm; dewaxed sections were then prepared for SEM to allow examination of the interface between the two tissues within the filter.

Most of the pores of the filter in the region of the tissues contained cellular and/or extracellular material (see Figs 15, 16,17). Thin cell processes, similar in appearance to the filopodia observed in control culture, were frequently observed in the pores (compare Figs 3, 17). Cell processes were occasionally observed traversing the whole width of the filter. However the incidence with which they were observed was reduced by the fact that the probability of a single section having a cut pore in it that traverses the whole width of the filter and contains a cell process is low. It is probable that most cell processes extended by the cells into the pores traverse the filter. Cell processes from NC cells were frequently seen projecting into pores and occasionally cell processes were associated with RPE cells but it was not possible to tell whether or not they were being extended by these cells. Sometimes NC cell processes were observed extending through a pore in the filter and spreading out on the surface of an RPE cell. It was not possible using this technique to establish whether direct plasmalemmal contact occurred through discontinuities in the basement membrane or alternatively if NC cells made contact with RPE-associated matrix. However these observations do provide evidence that a close association between NC and RPE cells is established during their interaction in a transfilter culture and is achieved by the passage of cell processes across the filter. Evidence from this study suggests that NC cell processes pass through the filter to contact the RPE but we cannot rule out the possibility that RPE cell processes may also participate in the interaction.

This experiment was designed to establish the minimum period of time necessary for the interaction to take place across a 0· 8 μ m filter. Neural crest was grown transfilter to stage-17 RPE for different lengths of time (7 days, 4 days, 2 days, 24 h). After these time intervals the RPE was removed from the transfilter cultures and the neural crest grown alone until a total culture period of 14 days had elapsed. When the duration of co-culture was 7 days the % incidence of cartilage formation was comparable to that found in standard transfilter cultures (see previous section). However, as the length of time of co-culture was decreased the % incidence of cartilage formation decreased rapidly (see Table 3). These results suggest that the interaction between RPE and NC is prolonged and is not completed in a short time.

When cartilage developed in these cultures it always formed against the filter and the amount that formed was generally less than in standard transfilter cultures. When NC was exposed to RPE for 7 days the cartilage that formed was fairly well organized but as the duration of co-culture decreased the cartilage which did form was less well organized. This observation suggests that continuous presence of the epithelium may be required for well-organized cartilage to form.

As described earlier when RPE was grown alone on a 0-8 μm Nuclepore filter extracellular matrix was present on the underside of the filter - it was visible with light microscopy after haematoxylin, alcian blue and durazol red staining and SEM observations on RPE controls grown for 24 or 48 h on 0· 8 μ m filters revealed that this ECM consisted of sheet-like, granular and fibrous components (see Fig. 9). Furthermore, ECM was never observed on the underside of the filter with both light microscopy and SEM when RPE was grown on 0· 2 μ m filters. The experiment was designed to investigate whether this matrix, components of which had diffused through from the epithelium on the other side of the filter, was capable of ‘inducing’ NC cells to differentiate into cartilage. Individual explants of stage-17 RPE were grown on 0· 8 μ m Nuclepore filters for approximately 2 weeks and the position of the RPE was marked on the filters with insoluble ink. The RPE was then removed and the filter inverted. Individual explants of NC were placed on the filters using the reference points to locate the former position of the RPE; i.e. the NC was put on top of the matrix that the RPE had deposited through the filter (see Table 4).

The results of this experiment demonstrated that the ECM deposited through a 0· 8 μ m filter by stage-17 explants cultured alone for 12– 14 days is not capable of ‘inducing’ NC cells to form cartilage. The majority of the cultures had a flattened pigmented outgrowth which often had discrete loci of darkly pigmented cells present in it. These loci have never been observed in NC control cultures and the incidence of melanogenesis was considerably higher than in NC control cultures. Possibly a component of this matrix induces melanocyte differentiation as extracellular materials have been shown to influence trunk neural crest cell differentiation in vitro (Loring, Glimelius & Weston, 1982). The lack of chondrogenesis provides further evidence that the interaction between stage-17 RPE and NC is not mediated by a diffusable matrix component. However, it should be noted that the matrix to which the NC is exposed has accumulated over a 12-to 14-day culture period and may not necessarily be identical in composition to that which the NC cells would encounter at the appropriate developmental time in vivo or indeed at the outset of the standard transfilter cultures described earlier.

Previous work has established that NC cells are relatively non-specific with respect to the type of epithelium that will promote chondrogenesis (Bee & Thorogood, 1980). A similar lack of specificity has also been demonstrated in the formation of mandibular membrane bone from ectomesenchyme (Hall, 1981; Tyler & McCobb, 1980). This lack of specificity in the nature of the epithelium in these interactions suggests that they may be ‘permissive’ (Wessells, 1977) and this interpretation is supported by the fact that such epithelia do not elicit chondrogenesis from tissues other than cranial NC or ectomesenchyme derived from it. For example, RPE does not ‘induce’ cartilage formation in chorioallantoic mesenchyme (Stewart & McCallion, 1975); if the interaction was ‘instructive’ RPE would direct the extraembryonic mesenchyme to form cartilage. The results of the present experiments clearly demonstrate that a permissive interaction leading to the chondrogenic differentiation of NC cells will occur only if the Nuclepore filters have a pore size large enough to permit the passage of cell processes. Furthermore, SEM observations have established that cell processes do actually traverse the Nuclepore filter when NC and RPE are grown in transfilter combination. We conclude that the mechanism is not mediated by diffusable factors but rather is mediated either by direct contact between NC cells and non-diffusable matrix closely associated with RPE or by direct plasmalemmal contact between RPE and NC cells through discontinuities in the basement membrane. In the light of this conclusion the claim (Newsome, 1976) that the interaction between the two tissues is simply ‘matrix-mediated’ is not tenable. Either that interpretation will need to be further qualified as we have proposed here (i.e. non-diffusable matrix) or the alternative possibility considered, that the contamination of the isolated matrix by cellular debris (see fig. 4 Newsome, 1976) led to a false positive result.

We have also demonstrated that younger (stage-17) RPE is more effective at eliciting chondrogenesis from premigratory NC cells than older (stage-24) RPE, in agreement with previous observations (Newsome, 1976). These observations are compatible with the sequence of events leading to scleral cartilage formation in vivo. NC cells arrive at the periocular region at stage 14 and are committed for chondrogenesis by stage 19 (Newsome, 1972). Hence, we would expect RPE to be most ‘inductively’ active between stage 14– 19. The apparent loss of inductive ability in the older RPE could be the expression of two possible control mechanisms. Firstly as the RPE becomes older and differentiates it loses its inductive ability: for instance it may produce a qualitatively different matrix that is less efficient at stimulating chondrogenesis. Alternatively as the epithelium grows older it may acquire inhibitory powers (Solursh, Singley & Reiter, 1981). The two proposed mechanisms are not mutually exclusive and could operate simultaneously.

The possibility of inhibition of cartilage formation may explain the apparent ‘induction at a distance’ which characterizes all epithelio-mesenchyme interactions leading to skeletogenesis. The differentiated tissue, be it cartilage (Bee & Thorogood, 1980; Gumpel-Pinot, 1981) or bone (Hall, 1981) arises at a distance from the epithelium within the mesenchyme, e.g. in vivo scleral cartilage is separated from the RPE by mesenchyme which becomes the fibrous choroid coat of the eye. Either the interface between epithelium and mesenchyme becomes occupied by another (migratory?) cell population after the interaction has been completed or the acquisition of short-range chondrogenesis-inhibiting ability by the epithelium might prevent adjacent mesenchyme cells from completing their differentiation into cartilage; a fibroblastic phenotype would result at the interface.

RPE matrix components which are freely diffusable and can pass through a 0 · 8 μ m filter and consolidate to form a matrix on the opposite side (see Figs 8 and 9) cannot elicit a chondrogenic response from NC cells (see Table 4). If the interaction is matrix-mediated rather than cell contact-mediated then clearly matrix components which are not mobile in this way are involved. We have referred to these putative factors as ‘non-diffusable matrix components’ and it is not obvious which molecules might be involved. RPE is unusual in that in vitro it produces a matrix of fibrillar, banded collagen in addition to its basement membrane (Newsome & Kenyon, 1973). As we can see no reason why procollagen would not pass freely through a Nuclepore filter and self-assemble in a matrix (see Fig. 9) and as such a matrix has been shown to be inactive, we feel that the most likely candidate is the basement membrane which characteristically remains associated with the basal surface of an epithelium both in vivo and (after regeneration) in vitro (Newsome & Kenyon, 1973). Indeed the importance of mesenchymal cell contacts with the epithelial basement membrane has been established in early tooth morphogenesis (Thesleff, Lehtonen & Saxen, 1978) and in corneal epithelial differentiation (Hay & Meier, 1976). It has been proposed, but not yet established, that a number of other epithelio-mesenchymal interactions are mediated via mesenchymal cell contact with epithelial basement membrane; e.g. cartilage formation in the avian limb bud (Gumpel-Pinot, 1981) and mandibular membrane bone formation (Hall & Van Exan, 1982). Precisely which component of the basement membrane is involved in any such interaction is not known and it is gradually being acknowledged that isolated purified matrix components cannot necessarily substitute for intact ECM. Newsome (1976) was unable to induce NC cell chondrogenesis with type I collagen and although sclerotomal cells will respond to chondroitin-4-sulphate, chondroitin-6-sulphate, type II collagen or proteoglycan by increased synthesis of matrix (Kosher, Lash & Minor, 1973; Kosher & Church, 1975; Lash & Vasan, 1978) it has not yet been possible to stimulate chondrogenesis from sclerotomal cells not already displaying limited synthesis of matrix components. Thus it appears that a normally organized and naturally synthesized matrix is a prerequisite if an interaction is to occur; in fact a matrix may only display morphogenetic potential in the presence of the living cells which have produced it and not when isolated on a filter. However it should be noted that the present results do not enable us to distinguish between the mechanism being mediated by the basement membrane or by actual plasmalemmal contact between the two cell populations.

The results presented also indicate that the interaction between RPE and NC cells is a prolonged one (see Table 3) and these in vitro observations are not incompatible with the timing of events leading to scleral cartilage formation in vivo. In the avian embryo NC cells start migrating from the tips of the neural folds at approximately stage 10 (36 h incubation), arrive at the periocular region at stage 14 (50 – 53 h) (Noden, 1975) and the periocular mesenchyme cells are committed to forming cartilage by stage 19 days) (Newsome, 1972). As the NC cells used in these experiments are premigratory NC cells it could be predicted that the interaction in vitro would be completed in approximately 2–3 days. However it was observed that the transfilter presence of RPE was needed for 7 days before the incidence of chondrogenesis was comparable to that found with the standard transfilter cultures. This longer period is probably artefactual and may result from either trauma during tissue isolation or a general retardation of growth and differentiation by the culture conditions. It should be noted that the RPE is isolated by an enzyme treatment which removes the basement membrane and as a consequence this experiment may, to some extent, measure the ability of the epithelium to regenerate its basement membrane. In addition, we do not know the effects of the migratory environment that the NC cells would normally experience ‘en route’ to the periocular region. Nevertheless, we predict that the interaction between RPE and NC in vivo is a fairly prolonged interaction of the order of days and not hours as in some systems such as the neuralizing effect of the dorsal mesoderm on the presumptive neurectoderm of the urodele gastrula which is completed after 10 h (Toivonen & Wartiovaara, 1976).

The cartilage formed in these experiments generally displayed a tissue-specific morphogenesis. In control cultures of stage-24 periocular mesenchyme, a tissue that is already committed to chondrogenesis having undergone an interaction in vivo, cartilage formed in plates or nodules away from the filter (see Fig. 1). In transfilter cultures of NC and RPE, where the interaction takes place in vitro, the cartilage likewise formed in plates or nodules, never in rods, and always formed adjacent to the filter (see Figs 12, 13). Morphologically, the scleral cartilage in the chick consists of a thin, curved plate of cartilage and our results and previous observations suggest that the basic form of cartilage is intrinsically determined (Weiss & Moscona, 1958; Newsome, 1972). Similarly cultured limb mesenchyme forms cartilage in rods (Weiss & Moscona, 1958; Gumpel-Pinot, 1981) resembling the long bone primordia which would have formed in vivo. We suggest that epithelio-mesenchyme interactions underlying skeletogenesis not only determine the differentiative fate of the responding cells but endow the tissue with morphogenetic autonomy. In fact the results of the ‘timing’ experiment indicate that the presence of an epithelium is required for well-organized cartilage to form because as the duration of co-culture of RPE and NC decreased, the cartilage which did form was increasingly less organized. The presence of an epithelium may be necessary for a critical time in order to organize the cells of the blastema in some characteristic way which is later expressed as a tissue-specific morphogenesis (Thorogood, 1982).

Observations on epithelio-mesenchymal interactions indicate that the means of transmitting the ‘inductive’ stimulus and the nature of that stimulus vary considerably. Nevertheless, in the only other transfilter analysis of an interaction causing chondrogenic differentiation (limb-bud chondrogenesis in chick embryos (Gumpel-Pinot, 1980, 1981)) it was established that close contact between epithelium and mesenchyme was necessary for cartilage formation to occur. In spite of apparent differences at the tissue and cell level of organization between somite chondrogenesis, limb chondrogenesis and cranial NC chondrogenesis it is possible that all are effected by the same molecular mechanism. However, given the very different developmental histories of the mesenchymes involved and our rudimentary understanding of most tissue interactions at the molecular level, it would be unwise at this stage to assume that the causal mechanisms in the different chondrogenic systems are necessarily identical or even similar.

This work has been supported by the Medical Research Council (U.K.). Chris Hawkins’ help enabled us to do the scanning electron microscopy, Norman Sylvester gave advice on the Appendix and David Johnston read the manuscript. Thank you.

A number of the physical parameters of the Nuclepore filters were measured during the course of the work and compared, where possible, with specifications provided by Bio-Rad Laboratories (U.K.) All measurements and counts were made on enlarged scanning electron micrographs of nuclepore filters, taken at calibrated magnifications.

In both filter types the diameters of approximately 80–180 pores were measured (see Figs 18 & 19) and the average diameter calculated. The average pore diameter in nominally 0·8 μm filters was found to be 0·698 μm (n = 79) and in nominally 0·2μm filters, 0·142μm (n = 189). The nominal sizes have been used throughout the text of the paper.

for 0·2μm filters,

pore density = 2·06 × 10⁸ pores/cm²

(manufacturer’s figure = 3 ×10⁸ pores/cm²)

This ratio represents the potential transfilter interface between two cultured tissues.

• total micrograph area = A x B cm²

• 

• non-pore area = (a) − (b)

For 0·8μm filters, pore : non-pore = 0·118.

For 0·2μm filters, pore : non-pore = 0·033.

The fact that the ratio of a 0·8 μm filter is approximately four times greater than that of a 0·2μm filter means that there is less potential interface between the interacting tissues in transfilter cultures grown on 0·2 μm filters than in similar cultures grown on 0·8μm filters. To achieve the same area of interface between the tissues would require a 0·2μm filter with approximately four times the pore density given in (ii). Such filters are not available. Although area of potential interface may be of secondary or even negligible importance, until that has been demonstrated the difference in ratio should not be overlooked when interpreting data from nuclepore transfilter cultures.