Expression of the mouse goosecoid gene during mid-embryogenesis may mark mesenchymal cell lineages in the developing head, limbs and body wall

Work on Drosophila and vertebrate embryos has shown that morphogenesis and the development of pattern is often presaged by the expression, within undifferentiated tissues, of spatially restricted patterns of homeobox gene (homeogene) expression (reviewed, for example, by Gaunt, 1991). The body is thus divided into zones, within each of which a homeogene product, or unique combination of homeogene products, may act as ‘selector’ of a particular route of morphogenetic development (Garcia Bellido, 1975). Such a zone has been variously described as a ‘field’ (De Rober-tis et al., 1991; Ingham and Martinez Arias, 1992) or ‘com-partment’ (Lawrence, 1990; Gaunt, 1991), the chief dis-tinction being that the term ‘compartment’ also implies a strictly self-contained unit of cell lineage.

goosecoid is a homeobox-containing gene present in a wide variety of vertebrate species (Blum et al., 1992). It is expressed in the dorsal lip of the blastopore in Xenopus (Blumberg et al., 1991; Cho et al., 1991) and in the devel-oping primitive streak region of the mouse (Blum et al., 1992). At these sites, goosecoid is expressed in anteriorly migrating cells that spearhead gastrulation movements. Microinjection of goosecoid mRNA into Xenopus embryos leads to formation of a new body axis, suggesting that this gene is part of the biochemical pathway leading to ‘orga-nizer’ activity. The goosecoid-expressing cells of the early mouse embryo are fated to form the head process, which later gives rise to anterior notochord and endoderm, and head mesoderm (Beddington, 1983; Lawson et al., 1991). In both mouse and Xenopus, goosecoid expression ceases to be detectable in later stages of gastrulation (Blumberg et al., 1991; Blum et al., 1992). We have found, however, that goosecoid expression recommences during organogenesis stages of mouse development, from 10.5 days, and it is these patterns of expression, seen in restricted regions of the facial processes and arches, limbs and body wall that are the subject of this paper.

Fig. 1 shows the arrangement of the branchial arches and facial processes as seen in a mouse embryo of about 10.5 days. The mesenchymal component of these is derived almost entirely from cephalic neural crest cells (Noden, 1988; Hunt et al., 1991d). Expression of the Antennapedia-like homeobox (Hox) genes probably accounts for at least some of the positional programming known to be intrinsic within neural crest-derived mesenchyme (Noden, 1983, 1988; Kirby, 1989; Richman and Tickle, 1989). Thus, ante-rior boundaries of Hox gene expression correspond with junctions between the branchial arches, and arches 2, 3 and 4 each display a unique combination of Hox gene products (Hunt et al., 1991a-d). Hox genes are not, however, expressed rostral to branchial arch 2 (Hunt et al., 1991a-d), and other genes must therefore be responsible for pattern-ing in the mandibular, maxillary and frontonasal processes. We now show how the goosecoid gene is expressed in discrete spatial domains within the facial processes and arches. These goosecoid-expressing domains, formed pre-sumably within neural crest-derived mesenchyme, give rise to parts of the tongue, mandible, nasal chambers, palate and ears. Our findings also indicate that the goosecoid gene may play a role in spatial patterning within proximal regions of the limb and adjacent body wall.

Methods used for ageing of embryos, tissue fixation, sectioning, preparation of ³⁵S-labelled riboprobes, and in situ hybridization were all as previously described (Gaunt, 1987).

The mouse goosecoid gene (Blum et al., 1992) contains three exons. Two different goosecoid probes (‘probes 2 and 3’), pre-pared from genomic DNA, were used in the experiments of this study. Probe 2 (as described in Blum et al., 1992) was prepared from a 909 bp DNA fragment that comprises 180 bp of exon 2 together with exon 3 and the intervening 347 bp intron. Probe 3 was prepared from a 349 bp Sau3A-HincII DNA fragment (nucleotides 1808-2157; Blum et al., 1992) that contains most of exon 3 plus 3′ trailer sequences upstream of the polyadenylation signal. In all in situ hybridization experiments, probes 2 and 3 gave identical results for the expression pattern of goosecoid. Probe 2 gave slightly higher background over all tissues than probe 3, but it also gave a stronger signal that was better suited to low-power photography of autoradiograms. The autoradigrams shown in this paper were obtained using probe 2. Localized expression of goosecoid within the tongue, nasal chambers and limbs was also noted using a non-overlapping exon 1 probe (‘probe 1’, as described by Blum et al., 1992), Probe 1 was not, however, used extensively since it gave excessive background labelling over all tissues.

Dlx-1 and Hox-1.5 probes were as used by Dollé et al. (1992) and Gaunt (1987) respectively.

At 10.5 days, strong expression of goosecoid was seen in the mandibular process of the first branchial arch, especially more posterior parts, and the anterior one-third of the second arch (Fig. 2A,B). Within the second arch, the pos-terior boundary of expression was sharp (arrowed in Fig. 2B). Expression of goosecoid was therefore abundant within the region of the first branchial cleft (hyomandibu-lar cleft; Fig. 2B), the precursor of the eustachian tube and auditory meatus (Fig. 1). Expression was detected only within medial parts of the first arch (Fig. 3B,C) and, apart from the rostral-most limits of this arch (Fig. 3B), was con-tinuous from the left to right sides (Figs 2B, 3C). Expression was not evident at this stage within the maxil-lary process of the first arch (Figs 3B, 4B). Fig. 4 allows comparison of the goosecoid expression domain with that of Dlx-1, another transcription factor known to be spatially restricted within the first arch (Dollé et al., 1992). goosec -oid transcripts (Fig. 4B) were generally seen to lie medial to those of Dlx-1 (Fig. 4C). In rostral parts of the first arch (seen on the right hand side of Fig. 4), goosecoid and Dlx-1 transcripts occupied apparently exclusive and comple-mentary domains. However, in more posterior parts (left hand side of Fig. 4) there was apparent overlap between the goosecoid and Dlx-1 domains. Expression of goosec -oid within the second arch was detected both medially and laterally (Fig. 3D), with medially located expression extending posteriorly beyond the limits of more lateral labelling (Fig. 3E). It was not clear whether or not this medial labelling (located within the floor of the pharynx) extended posteriorly beyond the limits of the second arch. The third and fourth arches were only clearly distin-guishable in their more lateral parts (Fig. 3E,F) and these regions showed no expression of goosecoid. At 10.5 days, strong expression was also seen in the lateral and medial nasal processes surrounding the base of the nasal pits (Fig. 3A).

By 12.5 days, strong labelling persisted within the hyomandibular cleft (cf. Figs 5C, 2B). Strong labelling also persisted in the derivatives of the mandibular process (see Fig. 1), namely the lower jaw (Figs 5C, 6B) and parts of the tongue (Figs 5A,B, 6B). The ventral-most parts of the tongue were not labelled (Fig. 6B). Within the tongue, there was a sharp boundary between labelled anterior and non-labelled posterior parts (Fig. 6B). This boundary, being at the level of the thyroid duct, lay at the anatomical bound-ary between the ‘body’ and ‘root’ of the tongue (Fig. 10). A continuous tract of labelled tissues was seen to extend posteriorly from the base of the thyroid duct to below, and around, the aortic arch (Fig. 6B). These tissues have been commonly observed to be labelled with Hox gene probes (e.g. Hox-1.4, -2.6, -4.2, -1.5; Gaunt et al., 1988, 1989) and were tentatively identified as the thyroid (anterior) and thymus (posterior) glands. Fig. 6 allows comparison of the goosecoid (Fig. 6B) and Hox-1.5 (Fig. 6C) expression pat-terns within the tongue. As shown earlier (Gaunt et al., 1988), Hox-1.5 is expressed in the root of the tongue, and in the sulcus terminalis (tissue lying just rostral to the thy-roid duct; Fig. 10), but is not expressed in more rostral parts of the body of the tongue. By 12.5 days, the earlier expression around the olfactory pit (Fig. 3A) had developed into an extensive zone of labelled mesenchyme surround-ing each nasal passage (Fig. 5A). This zone of expression extended continuously into the ingrowing palatal shelves (Fig. 5A).

By 14.5 days, strong labelling persisted around the nasal passages, in the palate and around the lower jaw (Fig. 7A). As seen in a cross-section through the mandible, labelling became markedly reduced in central parts already trans-formed into cartilage (Fig. 7A). Strong expression of goosecoid was seen at 14.5 days in mesenchymal tissues around the developing eustachian tube and base of the audi-tory meatus (Fig. 7C). Both of these regions are derived from tissues around the hyomandibular cleft (see Fig. 1), already noted to be expressing goosecoid at 10.5 and 12.5 days (Figs 2B, 5C). Strong labelling was seen in the devel-oping malleus (Fig. 7C) but not in other middle ear ossi-cles (incus and stapes; Fig. 7C).

At all stages, the central nervous system was unlabelled. However, one exception to this was found at 14.5 days in a small region of tissue within the floor of the diencephalon (Fig. 7A).

Neural crest-derived cells, which account for most of the mesenchymal component of the face (Noden, 1988), have completed their migration into the branchial arches by 9 days (Hunt et al., 1991d). At 9.5 days, however, no signif-icant expression of goosecoid above background was detected (Fig. 8). We therefore conclude that expression of goosecoid in the developing tissues of the head commences after migration of neural crest from the brain, at some time between 9.5 and 10.5 days.

goosecoid expression within the branchial arch tissue did not, apparently, extend into the overlying branchial epithe-lium (Fig. 2C). Similarly, at 12.5 days, the epithelium over-lying the tongue and lining the anterior edge of the thyroid duct showed little or no labelling above background (Fig. 6B). At 10.5 days (Fig. 3A), 12.5 days (Fig. 5A) and 14.5 days (Fig. 7A) labelling around the developing nasal cavi-ties was confined to mesenchymal tissues, and did not include the lining epithelium.

goosecoid expression was detected in a discrete ventral region of the otic vesicle (precursor of the inner ear) at 10.5 days (arrowed in Fig. 3D,E) and 12.5 days (arrowed in Fig. 5B). By 14.5 days, the labelling persisted within discrete regions of the cochlear duct (Fig. 7A-C). This labelling was restricted to part of the wall adjacent to, but not including, the sensory organ of Corti (Fig. 7B). At all stages, expression appeared to be confined solely to the epithelium. Thus, although some spread of silver grains was apparent in the immediately underlying mesenchyme, the range of this spread (about half a cell width) was no greater than that seen in the overlying lumen (not shown).

No expression of goosecoid was detected in 9.5-day limb buds (Fig. 8B). However, strong expression was seen by 10.5 days (Figs 2A, 3A, 3D-F). In both fore- and hindlimb buds the expression was restricted to proximal parts, and serial sections through the forelimb bud (Fig. 3D-F) showed that the bulk of labelling was proximal, ventral and ante-rior. This domain of labelling therefore excludes that part of the limb (distal, dorsal and posterior) that includes the zone of polarizing activity, around which Hox-4 gene tran-scripts radiate in a series of partially overlapping domains (Dollé et al., 1989). Expression within the proximal limb buds was seen to extend into the adjacent body wall (Fig. 3A,D-F). This expression within the body wall did not extend continuously between the hind- and forelimbs (Fig. 3B,C). Anterior to the forelimb, expression was seen within lateral parts of the thoracic body wall overlying the heart (Fig. 3F, left hand side of Fig. 3D,E) but not within more dorsal (not shown) or ventral (right hand side of Fig. 3D) parts of the thoracic wall.

By 12.5 days, the limbs contain mesenchymal conden-sations destined to give rise to skeletal elements (Figs 5D, 9). goosecoid expression at this time was restricted within patches that lay mainly between, rather than within, these condensations (Figs 5D, 9). The proximal restriction of goosecoid expression seen at 10.5 days (Fig. 3D-F) was maintained, so that 12.5-day expression was seen within the region of the developing hip, tibia and fibula (Fig. 5D), shoulder, radius and ulna (not shown), ankle and wrist (Fig. 9), but not around the digits (Fig. 9). At 12.5 days, expression of goosecoid was also maintained within the lateral body wall. This was confined to expression around the lateral and ventral, but not dorsal, parts of the developing ribs. As already described above for the limbs at 12.5 days (Fig. 5D), and the mandible at 14.5 days (Fig. 7A), expression around the ribs was excluded from the condensing skeletal elements themselves (Fig. 5D).

By 14.5 days, goosecoid expression within the forelimb was seen to be reduced to a few flecks, seen mainly within the shoulder and wrist regions (Fig. 7A).

From our findings, we note four striking features of goosec -oid expression during the period of organogenesis in the mouse. These features are also shown by the Hox genes, a group of homeobox-containing genes known to be required for spatial programming in the developing embryo (e.g. Gaunt, 1991). First, goosecoid expression is spatially restricted within tissues that otherwise appear morphologi-cally homogeneous in cellular composition (for example, the branchial arches and limbs). Second, expression of goosecoid commences in undifferentiated tissues and then persists as these undergo morphogenesis. Third, earlier (10.5-day) and later (12.5-to 14.5-day) patterns of expression are, at least to some extent, clearly linked by cell lineage (for example, goosecoid is expressed in first branchial arch tissue, and also its derivatives such as the mandible and tongue). Finally, goosecoid is expressed according to the position of a cell rather than its tissue-specific pathway of differentiation (for example, cells in the mandible and tongue pursue different pathways of differ-entiation). These findings, taken together with the fact that goosecoid encodes a homeodomain protein (Blumberg et al., 1991), presumably a transcription factor (Treisman et al., 1992), lead us to suggest that goosecoid product may be required for spatial programming within discrete embry-onic fields (De Robertis et al., 1991; Ingham and Martinez Arias, 1992), and possibly lineage compartments (Lawrence, 1990; Gaunt, 1991), during organogenesis stages of mouse development.

Already by 10.5 days, the branchial arches and limbs contain both mesenchymal (Hunt et al., 1991d) and myo-genic (Ott et al., 1991; Sassoon et al., 1989) cell popula-tions. In facial processes, mesenchymal and myogenic cell types are derived from cephalic neural crest (except, in branchial arches, for a minor component from lateral meso-derm; Noden, 1988) and head mesoderm respectively (e.g. Noden, 1988). In limbs, these two cell populations are derived from lateral plate mesoderm and trunk myotomes respectively (e.g. Tabin, 1991). We believe that the pat-terns of goosecoid expression now described are largely, if not entirely, attributable to expression within mes-enchyme. Thus, many of the areas of goosecoid expression form mesenchymal rather than muscular derivatives: for example, the mandible, walls of the nasal chambers and eustachian tubes, palate and ventral ribs. However, owing to a lack of available cell markers, and to the limited res-olution of the in situ hybridization technique, we cannot rule out the additional possibility of myogenic cell expression.

The principal sites of goosecoid expression seen in the head are within anterior arch 2, arch 1, and the frontonasal mass. The neural crest-derived mesenchymal cells within all of these structures first arise by an orderly ventral streaming of cells from the dorsal surface of the brain (data from cell lineage studies in the chick; e.g. Lumsden et al., 1991). By analogy with the chick (Le Douarin, 1982; Lumsden et al., 1991; Lumsden, personal communication), we conclude that goosecoid expression in the head spans a region derived from the anterior part of the fourth rhombomere of the hind-brain (which contributes cells to the anterior part of the second branchial arch) to the anterior midbrain (which con-tributes cells to the frontonasal mass).

We found that the goosecoid gene is expressed in dis-crete spatial domains within the facial arches and processes. This suggests that goosecoid has a role in patterning within facial arches and processes. For Hox genes in contrast, expression patterns within mesenchyme have so far suggested a role in differential patterning between arches. Thus, anterior boundaries of Hox gene expression coincide with junctions between adjacent branchial arches (Hunt et al., 1991a-d; Frohman et al., 1990).

For Hox genes, the expression patterns within branchial arches are first established in neural crest cells prior to their migration from the central nervous system (Hunt et al. 1991a-d). This is not, apparently, the case for goosecoid, since goosecoid expression is not detected until after neural crest migration is completed (Hunt et al., 1991d). Most of the goosecoid expression domain is rostral to the limits of Hox gene expression (the anterior boundary of arch 2; Hunt et al., 1991a-d), yet it is known that such rostral tissue (e.g. arch 1 mesenchyme) is, prior to migration, already deter-mined, at least to some extent (Noden, 1983, 1988), with respect to anteroposterior position. Some yet-unknown pat-terning mechanism must therefore precede goosecoid in this event. As one possibility, this primary mechanism, distin-guishing neural crest rostral to rhombomere 2, might simply be absence of any Hox gene product. Following this pri-mary anteroposterior patterning of cephalic neural crest, we assume that the discrete domains of goosecoid expression within head mesenchyme must be established by secondary events that take place in the environment of the facial processes and arches. These might, for example, include interaction between mesenchyme and overlying epithelia (reviewed by Hall, 1987), or diffusion of signalling polypeptides within the facial mesenchyme (e.g. Wnt-5; Gavin et al., 1990).

The possibility that the goosecoid-expressing cells seen from 10.5 days are, alternatively, simply descendants of the goosecoid-expressing cells seen earlier in the developing primitive streak (Blum et al., 1992) and fated to form the head process does not seem likely. Thus, no signs of goosecoid expression were detected in the embryo between 7.5 and 10.5 days, and goosecoid-expressing cells within the head process are not known to contribute to the domains of goosecoid expression seen later within the limbs and body wall.

In addition to goosecoid and Hox genes, other genes encoding transcription factors are also expressed in neural crest-derived head mesenchyme (Dlx-1, Dollé et al., 1992; Hox-7, Hill et al., 1989; Robert et al., 1989; Mackenzie et al., 1991; AP-2, Mitchell et al., 1991; M-twist, Wolf et al., 1991; Pax-3, Goulding et al., 1991). These also display spatial restriction in expression, suggestive of a role in spatial programming, although all of the genes are appar-ently expressed in all of the branchial arches. Most such genes (Hox-7, AP-2, M-twist, Pax-3), like the Hox genes, commence expression prior to neural crest migration. How-ever, the homeobox-containing gene Dlx-1, like goosecoid, commences expression in facial mesenchyme at about 10.5 days. Moreover, we have now shown that the spatial domains of Dlx-1 and goosecoid are apparently comple-mentary at sites within the first branchial arch.

Expression of goosecoid in the developing tongue was noted anterior, but not posterior, to the level of the thyroid duct (which opens to the pharynx through the foramen cecum). It is of interest to consider the position of this expression boundary in relation to the patterns of cell lin-eage that contribute to the tongue.

Fig. 10 shows how, according to the commonly held view, the tongue develops from branchial arches 1, 3 and 4. From this, we suggest that patterns of goosecoid expression in the developing tongue may simply follow the known patterns of cell lineage. Thus, goosecoid expression is seen in the body of the tongue (that is, anterior to the foramen cecum) at 12.5 days because this is derived from goosecoid-expressing parts of the first arch as seen at 10.5 days. goosecoid is not expressed in the root of the tongue (posterior to the foramen cecum) at 12.5 days because this is derived from goosecoid non-expressing regions in the third and fourth arches.

Furthermore, we suggest that the pattern of Hox-1.5 expression in the developing tongue similarly follows the pattern of cell lineage. Thus, Hox-1.5 is expressed in the root of the tongue, and the sulcus terminalis (just anterior to the foramen cecum), because these parts contain cells derived from Hox-1.5-expressing tissues located in the third arch (Hunt et al., 1991a). Hox-1.5 is not expressed anterior to the sulcus terminalis because this region is derived from Hox-1.5 non-expressing first arch tissue (Hunt et al., 1991a). One finding not easily accommodated within this interpretation is the expression pattern noted for Hox-2.6. Although Hox-2.6 is expressed in the fourth, and not third, branchial arches (Hunt et al., 1991a-d) its expression pat-tern in the tongue has been seen to be similar to, although much weaker than, the pattern for Hox-1.5 (Gaunt et al., 1989). Dlx-1 (Dollé et al.,1992), a gene expressed in lat-eral parts of the first arch at 10.5 days, is not expressed in the tongue or in medial parts of the mandible at 13.5 days. This observation, taken together with our results for goosec -oid, suggest that it is medial first arch tissue that gives rise to the tongue and to medial parts of the mandible.

Commencing at the thyroid duct, we observed a tract of goosecoid-expressing tissue extending posteriorly towards the base of the heart. We tentatively identified the labelled structures within this tract as the mesenchymal components of the thyroid, thymus and aortic arch. Expression within thyroid mesenchyme is readily understood, since this struc-ture originates at the ventral junction of the first and second branchial arches (at the level of the foramen cecum; Fig. 10), a region seen to be expressing goosecoid in the 10.5-day embryo. Expression within the thymus and aortic arch (derivatives of the third and fourth branchial clefts, and the fourth branchial arch respectively; Hamilton and Mossman, 1972; Sadler, 1985) is not so readily explained since we have not, so far, located goosecoid expression within any part of the third and fourth arches of 10.5-day embryos.

As for the tongue and mandible, there was evidence of a lineage relationship between cells that express goosecoid at earlier and later stages of development of the ear. Thus, both the mesenchyme surrounding the hyomandibular cleft at 10.5-12.5 days and also its derivatives, seen at 14.5 days as tissue around the eustachian tube and part of the exter-nal auditory meatus (Hamilton and Mossman, 1972; Sadler, 1985), were found to express goosecoid. The ossicles of the middle ear showed expression of goosecoid within the malleus, but not incus or stapes. This may be explained by our finding at earlier times (10.5 days) of more widespread labelling within the mandibular process (origin of the malleus; Sadler, 1985) than in the maxillary process or hyoid arch (origins of the incus and stapes respectively; Sadler, 1985). At 14.5 days, goosecoid expression within the cochlear duct was restricted to a discrete part of the epithelial lining adjacent to, but not including, the devel-oping sensory organ of Corti. These labelled cells may be derived from the discrete patch of labelled cells seen at ear-lier times (10.5-12.5 days) in the ventral part of the otic vesicle. In keeping with this proposal, it is known to be the ventral part of the otic vesicle that gives rise to the cochlear duct (Sadler, 1985). It is of interest to note that int-2, another gene likely to be involved in spatial patterning within the inner ear, is expressed in the cochlea only within the sensory regions (Wilkinson et al., 1989).

From 10.5 days, goosecoid was found to be expressed strongly in the proximal limb buds and adjacent body wall. So far, all transcription factor genes found to be expressed in facial mesenchyme (including Hox-7, AP-2, M-twist Pax-3 and Dlx-1) have also been seen to be expressed in the limbs. Of these genes, however, only goosecoid has been found to be expressed proximally, rather than distally. At present, the significance of this remains unclear. However, if the limb is patterned by similar mechanisms to those of the main body axis (Dollé et al., 1989; De Robertis et al., 1991) then it is of interest to note that goosecoid expression within these structures occurs in parts that have been deemed to be analogous: that is, the anterior part of the body and the proximal part of the limb.

We thank M. Price and D. Duboule for the Dlx-1 probe, and A. Lumsden for discussion on the origin of the cephalic neural crest. This work was funded in part by NIH grant HD 21502-07. M.B. was supported by a postdoctoral fellowship of the Deutscher Akademischer Austauschdienst (DAAD, NATO program).