Pattern formation in the facial primordia

At early stages in embryonic development, the vertebrate face has a common plan. A series of small buds of tissue, the facial primordia, forms around the primitive mouth (Fig. 1). The upper jaw develops from five main buds of tissue: a central primordium, the frontonasal mass (sometimes present as the median nasal processes), the two lateral nasal processes on either side and flanking these, the two maxillae. The lower jaw develops from the paired mandibular primordia. The same plan of facial primordia is found in the embryos of both birds and mammals. However, most of the information about how the face develops is based on work carried out in chickens because the embryos are readily accessible for experimental manipulations.

The facial primordia are made up mainly of neural crest cells that have migrated from the cranial crest and settle to form the facial primordia (Noden, 1975). The neural crest cells give rise to the connective tissues of the face. The myogenic cells of the facial muscles constitute a separate cell lineage. The myogenic cells originate from the paraxial mesoderm and also migrate into the facial primordia (Noden, 1983a; see also Noden, this volume).

The development of the facial primordia involves the four fundamental processes that underlie all of embryonic development. These processes are growth, morphogenesis, cell differentiation and pattern formation. Here, we will consider pattern formation - the process that leads to the spatial ordering of cell differentiation. The relationship between cell differentiation and pattern formation is illustrated by the development of the upper and lower beaks of chicken embryos. The same cell types, cartilage, muscle, bone and other connective tissues differentiate in both beaks but are arranged in different patterns (Fig. 2). We wish to understand how the patterns of differentiated cells are generated. What controls cell differentiation in the facial primordia to give the typical skeleton and associated muscles in the lower beak but a different pattern of tissues in the upper beak?

Pattern formation can be considered from the viewpoint of positional information (Wolpert, 1969; 1981). According to this view, cells are informed of their position within the embryo and acquire a positional value. Positional values are ultimately interpreted in terms of spatially appropriate cytodifferentiation. Development proceeds in a step-wise fashion because positional cues can only operate over small distances (Crick, 1970). Positional values acquired at an early stage in development affect the interpretation of positional values given later. For example, in chick limb development, cells must acquire positional values at an early stage that dictate whether they belong to a wing or a leg. Interactions within the developing limb bud determine which part of the limb they will form. If these same principles are applied to chick face development, the cranial neural crest cells that give rise to the upper and lower beaks must first be informed to which facial primordia they belong and then of their position within that facial primordium.

How are cranial neural crest cells informed to which facial primordium they belong? In general, the mesoderm of the embryo appears to act as the template that carries positional information along the head-to-tail axis of the embryo. For example, it is the mesoderm that determines whether a chick limb bud is a leg bud or a wing bud. In addition, the information carried by the mesoderm appears to dictate the regional differentiation of the ectoderm and the endoderm. For example, the specific pattern of the feather tracts is determined by the mesoderm (Sengel, 1975). The appropriate spatial differentiation of migratory neural crest cells in the trunk could also be controlled by the mesoderm. The trunk neural crest cells differentiate according to their final position rather than their origin (see for example Le Douarin, 1980). However, in the head where neural crest cells settle to form the facial primordia there is virtually no mesoderm to convey positional information and the cells must be informed to which primordium they belong by some other means.

One possibility is that the neural crest cells in the head are informed to which facial primordium they belong, prior to emigration. The individual facial primordia are populated by neural crest cell populations that arise in different regions of the head neural folds. The origins and pathways of these cranial crest cells have been extensively investigated in chick embryos. Recently, the remarkable feat of mapping cranial crest migration in a mammalian embryo has been accomplished (Tan & Morriss-Kay, 1986). In chick embryos, the neural crest cells that settle to form the frontonasal mass first migrate from the prosencephalic region (reviewed Le Douarin, 1982) and are joined by cells migrating later mainly from the anterior mesencephalic region (Noden, 1978). The cells in the maxillae come from the posterior mesencephalic region (Noden, 1975), whereas the cells in the mandibular primordia come mainly from the region of the anterior rhombencephalon but there is also a contribution from cells that arise in the posterior mesencephalon (Noden, 1975). In the trunk, exchanges between different regions of the neural crest almost invariably lead to normal development (Le Douarin, 1980a; reviewed Le Douarin, 1982). This suggests that trunk neural crest cells are initially equivalent and that their subsequent behaviour depends on where they settle (see also Bee & Newgreen, this volume). Until recently, the same appeared to be true of cranial neural crest (see, for example Noden, 1978). However, it has now been shown, in a series of transplantations of crest from forebrain and midbrain regions to the hindbrain, that ectopic beak-like structures can develop. For example, an upper ‘beak’ can form in the neck and a set of mandibular structures can arise in addition to the normal set, in the arch below (Noden, 1983b). The formation of ectopic beak structures strongly suggests that, in the head, the neural crest cells are already programmed before emigration.

A second proposal recently put forward by Thorogood (1987) is that cranial epithelium acts as a template to inform cells of their position in the head. For example, in the development of the tissues surrounding the chick eye, transient synthesis of type II collagen at the mesenchymal-epithelial interface of the periocular mesenchyme and the pigmented retina, correlates with the time at which neural crest cells become committed to form the scleral cartilage (Thorogood, Bee & Von der Mark, 1986). Specification of cranial neural crest cells that form the facial primordia could also involve cues provided by epithelial interfaces encountered during migration. The neural crest cells of each facial primordium have followed different routes to reach their destinations. The neural crest cells that settle in the frontonasal mass move at first rostrally above the prosencephalon and then around the front of the developing brain whereas the cells that settle in the maxillae and mandibles migrate laterally from the neural crest (see Johnston, 1966; Noden, 1975). Transient synthesis of type II collagen can be detected at mesenchymal-epithelial interfaces on some of these pathways (Thorogood et al. 1986). This may be involved in preparing the neural crest cells to subsequently differentiate as cartilage in the facial primordia but it is not clear whether this could characterize cells as belonging to, say, a frontonasal mass as opposed to a mandibular primordium.

A problem for the idea that positional information is conveyed during the migration of neural crest cells is the development of ectopic beak-like structures that result in certain cranial crest transplantation experiments (see above) since the routes taken by the crest cells must clearly be abnormal. It therefore appears most likely that the cues that inform neural crest cells to which facial primordium they belong operate before the cells start migrating.

The facial primordia of chick embryos contain apparently homogeneous populations of undifferentiated cells (Fig. 3) The cells in frontonasal mass primordia appear virtually indistinguishable from those in the primordia. However, when the mesenchyme cell populations from each of the facial primordia are placed in high-density (micromass) cultures, there are distinct differences in the extent and pattern of chondrogenesis (Wedden, Lewin-Smith & Tickle, 1986). (Micromass cultures have been extensively used to study the chondrogenic potential of cell populations in developing limbs (Ahrens, Solursh & Reiter, 1977)). At early stages (stage-20 to -24 chicken embryos), the mesenchyme cell population of the frontonasal mass undergoes extensive chondrogenesis and forms an almost continuous sheet of cartilage; in mandibular cultures, cartilage differentiation is less extensive and discrete nodules form; and in cultures of maxillae cells, virtually no chondrogenesis occurs at all (Fig. 4). Therefore, it is clear that with respect to cartilage differentiation in culture, there are distinct differences in the mesenchyme cell populations of the frontonasal mass, mandibles and maxillae at an early stage in the development of the facial primordia.

It is possible that interactions within the facial primordia have already taken place in chick embryos by stage 20 and that these have differentially modified the cell populations. For example, it may be that the frontonasal mass and mandibles at first contain equivalent cell populations but interactions within the mandible have led to some cells no longer having the potential to form cartilage. Flowever, cultures that have been established with cells from the mandibles of embryos at stages 17/18 have a sparse nodular pattern of chondrogenesis and in cultures from even earlier mandibles (from stage 16/17 chick embryos), hardly any cartilage differentiates at all. Cultures of frontonasal mass cells from stage-17 chick embryos still form a sheet of cartilage (J. R. Ralphs, unpublished observations).

The pattern of chondrogenesis in micromass cultures of the individual facial primordia could be interpreted as reflecting differences in the neural crest cell populations. However, the facial primordia contain, in addition to neural crest cells, cells of mesodermal origin that will give rise to the myogenic cells of the muscles (Noden, 1983a). Therefore, the differences in the extent of cartilage differentiation in the cultures may reflect, not characteristic populations of neural crest cells, but different proportions of presumptive myogenic cells. The effects of myogenic cells on the extent and pattern of chondrogenesis in cultures of chick limb bud cells have recently been demonstrated by Cottrill, Archer, Hornbruch & Wolpert (1986). They irradiated the somites at the level where the wing buds will develop to eliminate the myogenic cell lineage. In cultures of cells from the subdistal regions of muscle-less limb buds, a sheet of cartilage forms rather than the nodular pattern that is normally obtained. However, it should be noted that in cultures from different regions of chick limb buds, there appears to be no correlation between the extent of chondrogenesis and the number of muscle cells that differentiate (Swalla & Solursh, 1986).

Recently, we have investigated the possibility that the proportion of myogenic and potentially myogenic cells can account for the different patterns in chondrogenesis in cultures of chick facial primordia (Ralphs, Dhoot & Tickle, 1988). In the intact face, no differentiated myogenic cells can be detected in any of the primordia in chick embryos at stages 20 and 24. However, when the cells are placed in micromass culture, myogenic cells differentiate and this reveals the presence of potentially myogenic cells in the primordia. The myogenic cells are recognized by antibodies to the heavy chains of skeletal muscle myosin (Dhoot, 1986) and desmin, the intermediate filament protein characteristic of muscle (Osborn, Geisler, Shaw, Sharp & Weber, 1981).

The number of muscle cells that differentiate in the cultures can be compared with the extent and pattern of chondrogenesis. With cells from stage-20 embryos, cultures of all the facial primordia contain about the same number of myogenic cells despite the distinct differences in cartilage differentiation. With cells from embryos at stage 24, cultures of the frontonasal mass and the mandible now both contain an increased number of myogenic cells compared with the cultures from the earlier embryos. However, the same number of myogenic cells differentiate in cultures of both primordia. In the frontonasal mass cultures, the myogenic cells are distributed more or less singly throughout the sheet of cartilage (Fig. 5A), whereas, in the mandibular cultures, the myogenic cells are clustered between the nodules (Fig. 5B). In maxillae cultures, very few myogenic or chondrogenic cells differentiate (Fig. 5C). Therefore we conclude that the distinct patterns of chondrogenesis in micromass cultures of cells from chick facial primordia are not a reflection of dilution of the neural crest cell populations with different proportions of potentially myogenic cells. Instead, there appear to be real differences in the neural crest cell populations in each facial primordium in terms of the ability of the cells to differentiate into cartilage when placed in high-density culture.

Once the facial primordia have been established, cells must be informed of their position within the facial primordium to which they belong. The laying down of the pattern of cellular differentiation along the proximodistal axes of the beaks (this axis runs from the base to the tip) appears to involve a set of epithelial-mesenchymal interactions. This has been shown by Wedden (1987) in a series of transplantation experiments. She grafted fragments of facial primordia with and without their associated epithelium to holes cut in wing buds and assayed their development. With just the mesenchyme from the frontonasal mass or mandibular primordia, outgrowth and pattern formation is inhibited. In control fragments with intact epithelium considerable outgrowth occurs and beak-like structures are formed.

Pattern formation along the proximodistal axis of the developing limb is also coupled with outgrowth. A pronounced thickening in the epithelium at the tip of the limb bud, the apical ectodermal ridge (Saunders, 1948; Summerbell, 1974) is required for outgrowth and patterning. When the apical ectodermal ridge is removed from a limb bud, further outgrowth is inhibited. The limb that develops is truncated and lacks distal structures. In the limb, the apical ectodermal ridge maintains a region of undifferentiated mesenchyme at the tip of the bud as it elongates. This region has been called the progress zone. The progress zone model suggests that pattern along the proximodistal axis of the limb may be specified by the length of time cells spend at the tip of the limb (Summerbell, Lewis & Wolpert, 1973). Cells that leave the progress zone early form proximal structures whereas cells that leave later form distal ones such as digits. It is not clear whether the epithelial-mesenchymal interactions in the facial processes also involve an apical ectodermal ridge and a progress zone mechanism.

A second set of epithelial-mesenchymal interactions that may be involved in the patterning of cellular differentiation during development of the facial primordia is suggested by experiments in culture. Using micromass cultures of chick facial primordia, we have shown that face epithelium locally inhibits cartilage differentiation in these cultures (Wedden et al. 1986). Epithelium from either mandibular or frontonasal mass primordia is inhibitory when tested on cultures of frontonasal mass mesenchyme. Non-ridge epithelium from the limb bud also inhibits cartilage differentiation in its immediate vicinity in micromass cultures of limb bud cells (Solursh, Singley & Reiter, 1981). These interactions between epithelium and mesenchyme demonstrated in micromass cultures could serve to confine cartilage differentiation to the core of a developing limb bud (Solursh, 1984) or facial primordium. However, recent experiments in which chick limb buds have been permanently denuded or dorsal epithelium show that this, surprisingly, has no effect on the cartilage pattern that develops (Martin & Lewis, 1986).

The inhibition of cartilage differentiation induced by epithelium in the cultures of postmigratory neural crest cells should be contrasted with the epithelial stimulation of chondrogenesis of premigratory cranial crest cells in culture (Bee & Thorogood, 1980). However, it should be noted that in these organ cultures of premigratory crest cells and epithelium, cells differentiate into cartilage within the explant and a rim of nonchondrogenic tissue develops immediately below the epithelium.

Finally, we should consider the mechanisms that lead to specification of the mediolateral axis of the face. For example, cartilage differentiation is confined to the centre of the frontonasal mass in chick embryos to give the prenasal cartilage and the midpoint of the upper beak is also defined by differentiation of the epithelium to form an egg tooth. The formation of the egg tooth by the epithelium requires a signal from the mesenchyme (Tonegawa, 1973). How are cells informed of their position with respect to this mediolateral axis?

To explore the problem of specification of the midline of the chick upper beak, Wedden has isolated fragments from the frontonasal mass and assayed their development when grafted to limb buds. In fragments taken both from the midline and extreme lateral regions of the frontonasal mass of embryos at early stages (18-21), a cartilage rod and an egg tooth can develop. Therefore, from one frontonasal mass it is possible to obtain at least three cartilage rods and egg teeth. After stage 21, only central regions of the frontonasal mass give rise to midline structures. This suggests that the signals that confine cartilage differentiation to the centre of the frontonasal mass begin operating at stage 21 (Wedden & Tickle, 1986a).

The basis of the regulatory behaviour of lateral fragments of the frontonasal mass at early stages that results in the formation of midline structures is not clear. The new midpoint could be formed by cells in lateral positions now taking on central characteristics following removal of the native midpoint. Alternatively, cell proliferation could generate cells to reform midpoint structures. The frontonasal mass can apparently regulate in the intact face. In a series of experiments to explore the effects of the growing eyes on interorbital structures in chick embryos, both optic vesicles were removed and in one case a fragment of amnion was also inserted into the anterior neuropore. This resulted in formation of a bifid upper beak (Silver, 1962).

Another interesting feature of the behaviour of grafts of the frontonasal mass primordium is the development of duplicated prenasal cartilages and egg teeth from a single fragment (Fig. 6). This occurs with fragments taken from a range of developmental stages when the primordium is divided into either two or three pieces (S. E. Wedden: unpublished observations). At present, the basis of these duplications is not known.

When all-trans-retinoic acid is applied to chick embryos at early stages in the development of the facial primordia, facial defects result (Tamarin, Crawley, Lee & Tickle, 1984). The upper beak is missing but the lower beak develops normally (Fig. 7). The frontonasal mass appears to be specifically affected. Pattern formation and outgrowth of this primordium is inhibited. Fusion with the maxillae fails to occur leading to clefting of the primary palate (Tamarin et al. 1984). The full defect is produced when chick embryos are treated between stages 18 and 21 (Wedden & Tickle, 1986b) and can also be caused by applying a synthetic analogue of retinoic acid, (E)-4-2-(5, 6, 7, 8-tetrahydro-5, 5, 8, 8-tetramethyl-2-napha enyl)-l-propenyl) benzoic acid (TTNPB: Loeliger, Bollag & Mayer, 1980), which is metabolically rather stable. In mammalian embryos, 13-cis-retinoic acid may have a similar effect: applied when the facial primordia are populated with neural crest cells, development of the upper jaw is affected and median cleft lip can result (Goulding & Pratt, 1986).

Our working hypothesis for pattern formation in the face is that cells first acquire a positional value that informs them to which facial primordium they belong and then are informed of their position within that primordium. This framework may help interpretation of the specific defect. One possibility is that the response to retinoids is determined by the positional values that characterize the cells of the frontonasal mass. A second possibility is that retinoids interfere with the positional cues that inform cells of their position within the frontonasal mass and operate specifically within this primordium.

We can investigate how retinoids affect cells from the facial primordia in culture and find out whether the origin of the cells affects the response. We have compared the effects of retinoids (all-trans-retinoic acid and TTNPB) when added to the medium of the micromass cultures of cells from either the frontonasal mass or mandible. The retinoids inhibit cartilage differentiation in the cultures. There is a dose response. Cartilage differentiation in cultures of frontonasal mass and mandible cells shows the same sensitivity to each retinoid. Either l×10⁻⁶M-retinoic acid or 1×10⁻⁸M-TTNPB abolishes chondrogenesis (Wedden, Lewin-Smith & Tickle, 1987). Therefore, in culture, the response of frontonasal mass and mandible cells to the addition of retinoids appears very similar. However, there is a subtle difference when the amount of matrix secreted and the area of cartilage is examined in the retinoid-treated cultures. In mandible cell cultures, the dose response to retinoids involves a progressive reduction in both parameters whereas in frontonasal mass cultures, the area of cartilage-producing cells is much less sensitive to increasing doses of retinoid than the amount of matrix secreted.

In the intact primordia in the embryo following a treatment with TTNPB that would lead to the specific defect, the concentration of the retinoid in the frontonasal mass is 5·5X10”⁸M and in the mandible 4·6X10⁻⁸M (Wedden et al. 1987). This concentration of TTNPB applied to the mandible cells in culture would be sufficient to markedly reduce chondrogenesis. Therefore, the puzzle is why the mandibles of treated embryos nevertheless develop normally.

Retinoids could interfere with positional cues that operate as the facial primordia develop. The truncated upper beaks that result from retinoid treatment are reminiscent of the effect of removing the epithelium from transplants of the frontonasal mass. Wedden (1987) has investigated whether the epithelium is the target of retinoid action by making combinations between the epithelium and mesenchyme of facial primordia from treated and untreated chick embryos, and assaying the development of the recombined tissues. In combinations of frontonasal mass tissues, outgrowth and pattern formation is inhibited when the mesenchyme is taken from treated embryos, whereas from combinations of treated epithelium and untreated mesenchyme, upper beak-like structures develop. She therefore concluded that it is the mesenchyme of the frontonasal mass that is affected by retinoids and not the epithelium. It is interesting that retinoid treatment of either the mesenchyme or epithelium of mandibular primordia reduces the extent of outgrowth. In combinations of mandibular tissues from treated and untreated embryos, lower beak structures develop but these are shorter than controls.

In chick limb development, retinoids may act as morphogens to spatially control the pattern of cell differentiation. When retinoic acid is locally applied to the anterior margin of a chick limb bud, a gradient of retinoic acid is established across the bud (Tickle, Lee & Eichele, 1985). This signal leads to the formation of a duplicate set of digits that develops in mirror-image symmetry with the normal set. The signal generated by local application of retinoic acid mimics the effect of grafting an additional polarizing region. The polarizing region is a signalling region consisting of a small group of cells found at the posterior margin of the bud (Saunders & Gasseling, 1968; Tickle, Summerbell & Wolpert, 1975) and controls the pattern of structures that develop across the anteroposterior axis of the limb. Recently, endogenous retinoic acid has been demonstrated in developing chick wing buds (Thaller & Eichele, 1987). The concentration of this endogenous retinoid is similar to that which experimentally brings about pattern changes. Furthermore, the posterior half of the wing bud where the polarizing region is located contains more retinoic acid than the anterior half.

These experiments with chick limbs may be relevant to the effects of retinoids in the face. The concentration of TTNPB in the frontonasal mass of treated embryos is 5·5×10⁻⁸M. A similar retinoid concentration, 10× 10⁻⁸ M-TTNPB, has been found to bring about pattern changes in developing chick limb buds (Eichele. Tickle & Alberts, 1985). Therefore, it is tempting to speculate that retinoids may act as signalling substances in pattern formation in the frontonasal mass. The facial defect might thus arise because the applied retinoid distorts the normal retinoid distribution that signals the pattern of structures that develop in the frontonasal mass.

S.E.W. carried out the work reported here during the course of an MRC studentship. The research of C.T. and J.R.R. is supported by the MRC. C.T. thanks Dr A. Tamarin for introducing her to the face. We thank Professor L. Wolpert, J. Richman and S. Croucher for reading this manuscript.