Chicken homeobox gene Msx-1: structure, expression in limb buds and effect of retinoic acid

Homeotic genes are the groups of clustered homeobox genes that control distinct steps of the pattern formation along the anteroposterior axis during Drosophila development by determining the identity of each segment (reviewed by Gehring, 1987a; Scott et al. 1989). Homeobox genes carrying homeodomains are also isolated from vertebrate genomes and, especially in clustered homeobox genes, are structurally similar to Drosophila homeotic genes (reviewed by Graham et al. 1989). In addition, ectopic expression of this type of gene in transgenic animals produced a gain-of-function homeotic mutation phenotype (Kessel et al. 1990), and experimental removal of the protein product from the embryo induced loss-of-function type homeotic mutation (Wright et al. 1989). This evidence strongly suggests that clustered homeobox genes are the vertebrate homologues of Drosophila homeotic genes.

In contrast, individual homeobox genes like muscle segment homeobox (msh; Gehring, 1987b), H2.0 (Barad et al. 1988) and msh-2 (Bodmer et al. 1990) are expressed in a. particular group of mesodermal cells indicating their function is lineage specification in Drosophila. Homeobox genes like cut (Blochlinger et al. 1988) and rough (Saint et al. 1988; Tomlinson et al. 1988) define unique sensory neurogenic cell lineages. Unlike homeotic genes these homeobox genes are not included in cluster structures. Some of these gene homologues were also isolated from the vertebrate genome (Cdx:Duprey et al. 1988; Hox-7/7.1-. Robert et al. 1989; Hill et al. 1989) and, from their tissue-specific expression pattern, they are expected to have different functions from clustered homeobox genes.

One of the most interesting problems in the morphogenesis of vertebrates is pattern formation of the limb bud, and information has been obtained at both the descriptive and experimental levels particularly with chicken limb bud. At the beginning of limb development, specialized mesenchymal cells are released from the somatic layer of the lateral plate mesoderm, migrate laterally and accumulate under the epidermal tissue. Identity of the potential limb-forming area and axis are determined earlier than visible accumulation of limb mesenchyme (Hamburger, 1938; Saunders and Reuss, 1974). The limb mesenchymal cells then induce the ectodermal cells to elongate and form an apical ectodermal ridge (AER) at the distal tip of the limb bud. Transplantation experiments show that the AER and limb bud are not induced by mesoderm other than limb mesenchyme (Reuss and Saunders, 1965; Dhouailly and Kieny, 1972). If the AER is removed at any time, limb development ceases. However, grafting of an extra AER onto another limb bud induces supernumerary structures toward the distal end of the limb. Thus the AER is responsible for outgrowth along the proximodistal axis and development of the limb (Saunders and Gasseling, 1968; Tickle et al. 1975; Summerbell, 1979) by interacting with mesenchymal cells underneath. The positional information along the anteroposterior axis originates from the zone of polarizing activity (ZPA) localized at the posterior margin of the limb bud. Duplicated digits are induced as mirror images of normal digits when ZPA is grafted to the anterior limb bud region (Saunders and Gasseling, 1968; Tickle et al. 1975). In addition to this positional signaling, generation of limb-specific mesenchymal cells from lateral plate mesoderm initiates limb formation and lineage restriction through programmed cell death of limb mesenchymal cells in a specific area (Hinchliffe, 1981) that controls the supply of precursor mesenchymal cells. This may also affect cartilage pattern formation in the limb.

In vertebrate limb development, several homeobox genes have been shown to have a role in pattern formation. Recent progress demonstrated that the spatial expression pattern of homeobox genes in the upstream region of murine Hox-4 cluster (Dollé et al. 1989a), chicken Chox-4 (Nohno et al. 1991; Izpisúa-Belmonte et al. 1991) and Xenopus XlHboxl (Oliver et al. 1989) is strongly correlated with the positional information along the anteroposterior axis. In the latter two cases, expression of the gene at the anterior margin of the limb bud is affected by local RA treatment (Oliver et al. 1990) or ZPA transplantation (Nohno et al. 1991; Izpisúa-Belmonte et al. 1991) indicating the existence of a control mechanism for the spatially restricted expression of the homeobox genes by positional signaling systems. In addition, murine Hox-7/7.1 is expressed in distal marginal mesenchyme, a so-called progress zone (PZ) (Hill et al. 1989; Robert et al. 1989) where mesenchymal cell proliferation takes place; after leaving the progress zone, proximodistal values are specified in mesenchymal cells (Summerbell and Lewis 1975).

Here, we report detail on the cloning, structure and developmental expression pattern of chicken Msx-1 gene (previously designated as ch-mshl ; Kuroiwa and Yokoyama, 1989; Yokouchi and Kuroiwa, 1990), one of the chicken homeobox gene belong to the msh gene family. The results suggest that Msx-1 gene expression is closely linked to somatopleure mesenchyme cell lineage and presumptive non-cartilage-forming lineage of limb mesenchymal cells. The different expression pattern in the AER indicates a relationship to ectoderm-mesoderm interactions. We also report that local application of retinoic acid repressed the Msx-1 expression prior to ectopic cell growth followed by limb duplication

A day 10 embryonic cDNA library was purchased from Clonetech (Polo Alto). A day 4 embryonic library was constructed as described (Sasaki et al. 1990). To isolate a chicken Msx-1 cDNA clone, a 600 base pair of EcoRV to Xhol fragment containing the homeobox sequence from Drosophila msh cDNA (Barnard et al. unpublished data) was used as probe for low-stringency screening of 1.2 × 10⁶ plaques of the day 10 embryonic cDNA library in λgt11 vector. The isolated cDNA fragment was used for screening 1.0 ×10⁶ plaques of λgt10 library. For a cosmid library, the DNA purified from chicken liver nuclei was partially digested by Sau3A1, size fractionated with NaCl density gradient centrifugation to obtain an average 40 kb DNA fragment, treated with calf intestine alkaline phosphatase and integrated into pcosEMBL2 as suggested by Poustka et al. (1984). The msx25 cDNA fragment was used for screening 1 ×10⁶ colonies of the cosmid library. Genomic or cDNA fragments were subcloned into BlueScript (Stratagene) and were sequenced by dideoxysequencing method (Sanger et al. 1977).

RNA was extracted by the guanidium isothiocyanate-CsCl method according to Chirgwin et al. (1979) and poly(A)⁺RNA was selected by two cycles of oligo (dT)-cellulose column chromatography. Samples containing 10 μg of poly(A)⁺RNA were separated on 1% agaroseformaldehyde gels and transferred to Gene Screen plus (DuPont). Nucleic acid was crosslinked under 309 nm UV light and the filters were then hybridized under highly stringent conditions. Hybridization and washing conditions were as described earlier (Kuroiwa et al. 1984).

Plasmid containing the Msx-1 cDNA was linearized and sense strand RNA was synthesized with T3 RNA polymerase as described in the manufacturer’s manual of Blue Script (Stratagene). The in vitro made RNA was purified by DNAase I treatment followed by phenol extraction and ethanol precipitation. Approximately 0.5 μg RNA was translated in a rabbit reticulocyte lysate according to the manufacturer’s specification (Amersham) using [³⁵S]methionine (Amersham) to label the protein.

Sense and antisense RNA probes were transcribed in vitro from linearized plasmid containing the msx6 cDNA fragment using digoxigenin-labeled UTP and T3 or T7 RNA polymerase as described in the manufacture’s manual (DIG RNA Labeling Kit SP6/T7; Boehringer). The probes were treated with alkaline to reduce them to an average size of 150 nucleotides (Cox et al. 1984). Fixation of embryos, preparation of frozen sections and section pretreatment were done according Hogan et al. (1986) with the slight modifications below. Chicken embryos were dissected and fixed in phosphate-buffered saline (PBS) containing 4 % paraformaldehyde for 2 –6 h at 4 °C. Then the fixative was replaced with 30% sucrose containing PBS at 4°C for 2h. Whole embryos or dissected limbs were embedded in OCT compound (Miles) and frozen in a dry ice-ethanol bath, then stored at –80°C. 5 gm sections were collected on a gelatin-chrome alum-subbed slide, immediately dried under an air dryer for 1h, then further dried in an oven at 45°C for 2h. Sections were stored at –80°C until used for in situ hybridization. Sections were rehydrated in PBS containing 0.3% Triton X-100 for 5 min, treated with 0.2 N HC1 for 20min, washed with PBS for 5 min, treated with PBS containing 1 μgml^-1 of proteinase K (Boehringer) for 5 min at 37 °C, postfixed with 4% paraformaldehyde in PBS, quenched twice with PBS containing 0.1 M glycine for 15 min and equilibrated in 50% formamide containing 2 ×SSC for 1h. The sections were hybridized overnight at 45 °C in a humid box using a hybridization solution (150 μl/slide) that contained 20 mM Tris –HCl pH 8.0, 2.5 MM EDTA, 50% formamide, 0.3 M NaCl, 0.02% BSA, 0.02% Ficoll, 0.02% polyvinylpyrrolidone, 1mgml^-1E. coli tRNA, 10 % dextran sulfate and probe at a concentration of 0.7 μgml^-1. After hybridization the sections were washed for 1h in 50% formamide and 2xSSC at 45°C followed by RNAase digestion in 0.5 M NaCl, 10 mM Tris –HCl pH 8.0 and 20 μgml^-1 of RNAaseA. Subsequent washes were done in 50% formamide and 2 ×SSC at 45°C for 1h, then 50% formamide and 1 ×SSC at 45°C for 1h. Immunological detection was done as described by Tauz and Pfeifle (1989).

AG1-X2 beads (200 μm in diameter) were soaked in retinoic acid (Sigma) dissolved in dimethyl sulfoxide (DMSO) at 0.01 mgml^-1, 0.1mgml^-1 or 1mgml^-1, rinsed in tissue culture medium and then implanted beneath the apical ridge of wing buds as described by Tickle et al. (1985). As a control, beads were soaked in DMSO alone.

We obtained one Msx-1 cDNA clone (msx25) from the day 10 embryonic cDNA library and another clone from the day 4 chicken embryo library (msx6). The two independent cDNA clones slightly differ in size at both ends, but have essentially the same putative proteincoding region. The complete nucleotide sequence (1083 bp) of the cDNA clone (msx6) is shown in Fig. 1A. The longest open reading frame consists of 777 bp (nucleotide position 64 –840). The oligo(dA) stretch is found at the same position from the 3 ’ end of both cDNA, but no polyadenylation signal (AATAAA) precedes it. The oligo(dA) stretch was also seen in genomic DNA at the same position (results not shown), indicating that this stretch in cDNA does not originate from the poly(A) tail of mRNA. Since Northern blot analysis (below) suggests that the message size is 2600 nt, our cDNA clones are missing about 1500 nt of noncoding region.

Two ATG codons (nucleotide positions 64 and 184) were found in the frame with the homeobox. Since the surrounding nucleotide sequence of the first ATG codon better matches the consensus sequence established for eukaryotic translation start (CCA/ GCCATGG; Kozak, 1987) than that of the second ATG, the first ATG is likely to be the initiation codon. If protein synthesis starts from the first ATG, the putative Msx-1 protein consists of 259 amino acids, and carries the homeodomain relative to its carboxyl terminus.

Four overlapping cosmid clones containing Msx-1 cDNA sequence were isolated from the cosmid DNA library of chicken genome. The structure of the genomic region encoding Msx-1 cDNA is shown in Fig. IB. By comparing size and restriction sites of the cDNA fragment and homologous genomic fragment, an intron of approximately 3.3kb is expected. The fragments containing splicing junctions were isolated from the genomic clone and sequenced. The sequence at the 5 ’splice site (AA/GTGAGT) and the 3 ’ splice site (CCCTTCCTTGCAG/GA) are in good agreement with the splice consensus (Shapiro and Senapathy, 1987). Thus the protein-coding region of Msx-1 gene consists of two exons separated by a single intron.

The overlapping cosmid clones, which span approximately 20 kb up- and downstream of the Msx-1 coding region, were used to search for homeobox gene(s) other than the Msx-1 gene. Using Drosophila Antennapedia, Sex combs reduced, engrailed and msh homeobox sequence for low stringency Southern blot analysis, we did not find any specific cross-hybridizing signals other than Msx-1 (results not shown).

The deduced protein of Msx-1 consists of 259 amino acids and has a predicted relative molecular mass (M_r) of 28.4 ×10³. In order to analyze the protein product encoded by the Msx-1 cDNA clone on SDS gel, the cDNA sequence was transcribed in vitro and translated in a reticulocyte lysate system in the presence of [³⁵S]methionine. As shown in Fig. 2B, one major protein band (34 ×10³M_r) and two additional bands (31 and 32 ×10³M_r) appeared to depend on Msx-1 RNA. As in the case of the other homeodomain protein (Sasaki et al. 1990), Msx-1 proteins migrate slower on gel than expected from their molecular mass. Production of the two smaller proteins was not inhibited by addition of protease inhibitors mixture (leupeptin, chymostatin, pepstatin, antipine and bestatin) in the translation reaction. The difference in size would be due to use of the second ATG for translation initiation in vitro at lower efficiency than the first ATG.

The amino acid sequence of Msx-1 homeodomain showed highest homology to Drosophila msh, mouse Hox-7.1 and Hox-8. The amino acid sequence homology of entire Msx-1 protein to murine Hox-7.1 (Monaghan et al. 1991) and Hox-8 (Monaghan et al. 1991) was 61.3% and 83.8% (217/259), respectively. Taking account of the synonymical relationship of amino acids, the homology to Hox-8 amino acid sequence reached 91% (236/259) (Fig. 1C). In addition, we also observed a cross-hybridizing band on Southern blot hybridization using a fragment containing Msx-1 homeobox indicating the presence of a related gene also in the chicken genome. From these, Msx-1 is most likely to be a chicken homologue of murine Hox-8. As shown in Fig. 1D, the amino acid sequence homology of Quox-7 (Takahashi and LeDouarin, 1990) to chicken Msx-1 is 98.5 % (255/259) and the length of the putative protein is identical. We therefore conclude that the Msx-1 is the chicken homologue of Quox-7. The nucleotide sequence of both 3 ’ and 5 ’ noncoding regions also shows high homology indicating functional significance.

The expression of Msx-1 during chick development was analyzed by northern blot analysis using poly(A)⁺RNA from chick embryos and adult tissues. Fig. 2A shows the hybridization under highly stringency conditions of the Msx-1 cDNA probe to RNA from day 2 –11 embryos. A single transcript of 2.6 kb is detected. The Msx-1 transcript is most abundant in day 2 –7 embryos. After day 8, expression decreases gradually. These results demonstrate that Msx-1 is expressed on day 2 of embryogenesis at the latest. We also examined poly(A)⁺RNA derived from a variety of adult chick organs including brain, retina, spinal cord, lung, liver, kidney and testis for Msx-1 expression and found no detectable expression in these organs (data not shown).

The spatial expression pattern of Msx-1 was determined by in situ hybridization using the digoxigenin-labcled antisense RNA probe (Tautz and Pfeifle, 1989) as described in Materials and methods. Essentially the same results were obtained by using the ³⁵S-labeled antisense RNA probe transcribed from a template DNA consisting of the region from 5 ’ terminal of the cDNA clone to the EcoRI site in the homeobox where fewer sequence homologies to Hox-7/7.1 are found. Fig. 3 shows the result of hybridization to cross-sections of 2-day-old (stage 12; Hamburger and Hamilton, 1951) chick embryos. A dorsal view of stage 12 embryo is illustrated in Fig. 3E to show the levels of the transverse sections. Somatopleure expression was observed at the level where somite formation takes place (Fig. 3B) and in unsegmented mesoderm of more posterior and more undifferentiated regions (Fig. 3C,D), but not in somites, intermediate mesoderm or splanchnic mesoderm. In addition, in the nervous system, the expression was observed in the neural crest region of the spinal cord (Fig. 3A,B), neural fold of the neural plate (Fig. 3C) and its presumptive anlagen in the primitive streak (Fig. 3D), but not in the residual part of the neural tube or neural plate.

The first sign of limb bud formation is visible from the outside in stage 16 embryo as a thickening of the ventral ectoderm which will form the apical ectodermal ridge (AER), and from inside as cells from the lateral mesoderm migrating into the limb bud region. At stage 17, both fore- and hindlimb buds lift off blastoderm by enfolding the lateral body-folds. Fig. 4 shows typical examples of the expression in serial transverse section of fore-(A – C) and hind-(D –F) limb buds in stage 17. As shown in Fig. 4A, just anterior to the forelimb buds, expression was observed in the lateral mesoderm and overlying ectoderm layer. The expression was seen in the anterior to middle part of forelimb mesenchyme and lateral mesoderm as well as in their overlying ectoderm. Fig. 4B shows a graded expression along the proximodistal axis. In the posterior region of the forelimb bud, mesenchymal expression and in the overlying ectoderm disappeared (Fig. 4C); however, strong expression in the thickening the AER remained as indicated by arrows. These results and the pattern in the horizontal section (not shown) indicate that expression was graded along the anteroposterior axis of the limb bud. The same expression pattern was observed in hindlimb buds (Fig. 4D –F). Expression was also observed in the roof plate of the neural tube as indicated by triangles in Fig. 4, and in facial mesenchyme cells derived from neural crest (not shown).

The AER is distinguishable from other ectoderm at stage 20 because of its thickened or columnar morphology. Strong expression was observed in the AER and continued to stage 25 (Fig. 5E). At stage 22, the limb extended more distally and, as shown in the transverse section in Fig. 5A, the expression pattern itself was similar to that of stage 20, but mesenchymal expression was more characteristic. In the forelimb bud, the mesenchymal expression was clearly seen in the anterior proximal and the posterior distal regions, whereas in the hindlimb bud, the expression was only observed in the anterior proximal region. A continuous expression in lateral mesoderm to anterior limb mesenchyme was observed at the same level (Fig. 5A,B). The expression pattern seemed to be local rather than graded as seen in stage 17. As shown in Fig. 5C,D, at the anterior part of limb bud, Msx-1 was expressed in mesenchymal cells underneath the AER, whereas at the posterior part, expression in mesenchymal cells underneath the AER was not observed (Fig. 5E).

At stage 24, limb bud growth was further toward the distal posterior. As shown in Fig. 6A, Msx-1 expression was observed in mesenchymal cells of both anterior and posterior margins of wing limb bud. However, in hindlimb bud, expression was observed only in the anterior margin. The anterior expression region seemed to move distally during limb growth, resulting in the disappearance of expression at the anterior proximal region. At the boundary of expressing and non-expressing regions, fade-out-type expression was observed; however, the graded expression area did not extend across the whole limb bud. At this stage, differences in expression pattern along dorsoventral axis began to be visible. The expression was strong in the ventral and relatively weak in the dorsal side (result not shown). The Msx-1 expression in the AER as well as the expression in ectoderm accompanied by under-neath mesenchymal expression were commonly observed in both wing and leg buds.

After stage 26, expression was gradually restricted in the distal margin (results not shown). At stage 26, expression was observed in the region distal to the zeugopodium and, at stage 28, expression was seen distal to the autopodium. Expression in the AER gradually decreased. As shown Fig. 6C,D, at stage 28, Msx-1 expression is observed in the distal periphery of mesenchymal cells where it had never been seen earlier. Since migration of anterior or posterior marginal mesenchymal cells to the distal peripheral region has not previously been reported, signals observed in this region must be due to the newly induced expression.

To analyze the effect of retinoic acid on the Msx-1 expression in developing limb bud, beads soaked in DMSO solution containing retinoic acid were implanted in the anterior margin of the forelimb buds of stage 20 embryos. Using 0.1mg ml^-1 RA, 24h after implantation, the expression of Msx-1 in anterior mesenchymal cells and ectodermal layer had disappeared (Fig. 7C), whereas expression in posterior mesenchyme showed no change in pattern but was weakened. In our experiments, this concentration of RA results in mirror-image duplication of posterior digit reproducibly. The same type of change in expression was observed using Imgml^-1 retinoic acid treatment (results not shown). When control beads presoaked in DMSO (Fig. 7A) were implanted, no change in mesenchymal cells or ectodermal Msx-l expression was detected. At lower concentration of retinoic acid (0.01 mg ml^-1), which can induce duplication of second digit, no visible change occurred (Fig. 7B). At any RA concentration, the β-actin mRNA accumulation pattern, monitored by in situ hybridization, was not altered (results not shown). Fig. 7D – I shows the time course of change in the expression pattern after 0.1mg ml^-1 retinoic acid treatment. The first sign of decrease of the Msx-1 mRNA was found in the nearest AER at 2h after implantation (results not shown). 6h after treatment, expression in the entire AER was remarkably reduced (compare Fig. 7D and E), but no sign of morphological change was observed. A decrease of transcripts in the AER of the distal tip was also observed where furthest from the implanted position (Fig. 8). At this stage, expression in the mesenchymal cells distal to the bead disappeared. 12h after treatment, the hybridization signal in the remaining cells was also weakened significantly (Fig. 7F,G). The posterior marginal expression disappeared 12 h after treatment but reappeared 24 h after treatment while in the anterior region expression was still declining. At the same time, the morphology of the cells in the AER just overlying the bead became flatter (results not shown, Tickle et al. 1989). The expression in both ectodermal and mesenchymal cells reappeared 48 h after treatment (see Fig. 7H and I). At this stage, retinoic-acid-induced ectopic outgrowth of the limb bud was visible and Msx-1 expression was also observed in the peripheral mesenchyme of this region. However, the expression was not observed in the distal part of the extra outgrowth region

During segmentation, the mesoderm differentiates into four different types of structurally and spatially distinct mesodermal layers along the proximodistal axis. The paraxial mesoderm, which gives rise to somites, is the most proximal. The intermediate mesoderm produces genital –urinary mesenchyme, forming a thin layer connecting the paraxial mesoderm and the lateral mesoderm. The lateral mesoderm consists of two layers. The upper layer is somatic mesoderm and gives rise to flank mesoderm and limb mesenchyme. The somatic mesodermal layer is closely connected with overlying ectoderm both at physical and physiological levels. The lower layer is splanchnic mesoderm and gives rise to visceral organs in combination with the underlying endodermal layer. This subdivision process takes place in an anterior-to-posterior direction at the region around Hensen’s node.

In stage 12 embryo, there is uniform Msx-1 expression in the unsegmented mesoderm at the posterior part of the embryo. At the more anterior region, where somite formation and morphological subdivision of the mesodermal layer is occurring, Msx-1 expression is only found in the somatic mesodermal layer. This shows that the Msx-1 expression is restricted in somatic mesoderm during the process of mesoderm subdivision and indicates that Msx-1 is closely related to the determination of somatic mesoderm identity. The expression in the overlying ectoderm is nearly equal in strength to the somatic mesoderm. This suggests a mesodermectoderm interaction for mutual maintenance.

Recently rnsh-2, a Drosophila msh-related gene, was isolated and shown to be expressed transiently in mesoderm and then restricted in visceral mesoderm (Bodmer et al. 1990). Genetical analysis of this locus indicates that the msh-2 is crucial for determination of visceral mesoderm lineage. This evidence and our finding of restricted Msx-1 expression in somatopleure lineage indicate that the vertebrate homeobox gene is closely related to the subdivision of mesoderm lineage in mesodermal cells.

In chick embryos, the thickening and mesenchyme accumulation are continuous throughout the whole length of the body in the form of a horizontal ridge, the Wolffian ridge. The most anterior and most posterior parts of the ridge are thicker than the intermediate part, and it is only these anterior and posterior parts that develop progressively, giving rise to the forelimbs and hindlimbs (Balinsky, 1981). As described above, Msx-1 expression becomes restricted in the Wolffian ridge after mesodermal subdivision takes place. At the beginning of limb bud formation, a condensation of mesenchymal cells derived from the somatopleure is observed where the limb bud will emerge. Then, thickening of the overlying ectoderm becomes visible. At stage 17, where the limb bud is clearly distinguishable by its slightly expanded morphology from the flank, mesenchymal cells located in the central part of the bud do not express Msx-1, whereas mesenchymal cells in the Wolffian ridge still express it. This pattern is more obvious in later stages, where mesenchymal cells in the flank express Msx-1, but limb mesenchymal cells at the same dorsoventral position do not. Fig. 9 shows a schematic representation of the change in expression of Msx-1 during limb bud formation. The spatial expression pattern changes at the limb level, although almost no change is found at the flank level. A plausible mechanism for this change in expression is that a specific portion of somatopleure mesenchymal cells, which receives a signal to differentiate to a specific lineage of limb bud mesenchyme cells, ceases to express Msx-1 in response to the signal.

After limb bud formation is distinct, the area of the anterior expression region is dramatically reduced and restricted to the anterior margin during the development from stage 17 to stage 24. The regional expansion of limb bud is unequal during this period (Bowen et al. 1989). The central part of the limb expands more rapidly than the anterior and posterior margins, also the posterior region expands faster than the anterior region. Since Msx-1 is expressed in the anterior proximal part at stage 17, apparent gradual restriction in the expression seems to arise from this regional bias of limb growth. Alternatively, the protein products of the Hox-4 cluster gene, which are expressed in the posterior half of limb bud (Dollé et al. 1989a and Yokouchi et al. submitted), and the anterior proximal expression of XlHboxl gene would suppress the Msx-1 expression. On the contrary, Hox-7/7.1 gene is expressed in the posterior part of limb mesenchymal cells (Hill et al. 1989) indicating opposite sensitivity to the positional signaling or other homeobox gene products.

Expression in the AER starts just after limb bud formation and continues until the regression of the AER. A strong expression in the AER is contrary to the rule that ectodermal expression is always accompanied by underlying mesenchymal expression. It is possible that the role of the AER expression is different from those of other parts. If the AER is removed from a developing limb bud, cartilage pattern formation following the removal is prevented. Therefore the AER is thought to secrete a kind of growth factor that promotes mesenchymal growth in a distal direction. Recently the bone morphogenetic protein 2 (BMP2), a member of the TGF-βfamily, was demonstrated to be expressed in the AER of mouse limb bud (Lyons et al. 1990). The strong and characteristic expression of Msx-1 in the AER might be related to the production of such a growth factor(s) for pattern formation of mesenchymal cells in cartilage-forming lineage. The Msx-1 expression in the AER contrasts with Hox-7/7.1 expression in the progress zone. The exclusive expression of related genes in physically and functionally closed layers indicates that these genes mediate ectoderm-mesenchyme interactions in a different way to that in the somatopleure. Recent observations on Hox-7.1 and Hox-8 expression during murine eye development also strongly suggest involvement of these genes in interaction of the two cell layers (Monaghan et al. 1991).

The first sign of limb pattern formation is recognized in the cartilage architecture. At the beginning of the cartilage formation, limb mesenchymal cells form an aggregate, which subsequently grows distally by recruitment of cells at the growing distal end. The limb cartilage pattern develops by a sequential segmentation and branching of the precartilaginous condensation during outgrowth of the cell aggregate under the control of positional signaling (Oster et al. 1988). In their model, the size of the cell aggregate and adhesivity of precursor cells are important parameters for timing the bifurcation processes. The positional signaling would determine the cell adhesivity which affects the size of the cell aggregate and cell growth. At the same time, change of mesenchymal cell supply to the aggregate would be a key factor in controlling the rate of cell aggregate formation and the timing of the bifurcation processes. During limb development, programmed cell death occurs in several regions designated as necrotic zones (reviewed by Hinchliffe, 1981). The necrotic zones are located at the proximal marginal region in the early stage and this necrosis is thought to remove or trim away excess mesenchymal cells, which is not needed in the potentially cartilage-forming cell lineage. In fact, in animals carrying talpicr²⁺³ mutation, neither anterior nor posterior necrotic zones appear, and this results in polydactylous limbs with up to eight digits. In addition, local retinoic acid treatment prevents programmed cell death (Lee, 1986) and results in extra digit formation (discussed below). In both case, a larger amount of mesenchyme results in the limb than in normal development. At stage 23 –24 forelimb bud, the necrotic zone of both anterior and posterior proximal margins overlap with the Msx-1 expression domains. The posterior necrotic zone is not distinct in the hindlimb bud (Lee, 1986) and the Msx-1 expression is not seen in this region. The Msx-1 expression precedes cell death because marginal Msx-1 expression is observed as early as stage 16 and cell death becomes visible from stage 23. This indicates that Msx-1 expression is closely related to the identity of the mesenchymal cells that do not differentiate into limb cartilage and also suggests that Msx-1 contributes to limb pattern formation through controlling the mesenchymal cell supply to the limb field.

The Msx-1 expression domains were slightly larger than the necrotic zones, indicating that not all cells in the Msx-1-expressing region are programmed to die. It might be that the Msx-1 product acts to prevent marginal mesenchymal cells entering the potential cartilage-forming lineage. A specific expression of Hox-7/7.1 in interdigital necrotic zone is reported (Robert et al. 1989; Hill et al. 1989); however, Msx-1 expression is not observed in that region indicating functional difference between two genes.

Local application of RA at the anterior margin of limb bud results in a mirror-image duplication of digits (Tickle et al. 1982). We found that the Msx-1 mRNA level in the anterior margin begins to decrease 6h after RA treatment and, 12 h after treatment, it was significantly diminished. Disappearance of the functional Msx-1 protein product may be more delayed than that of mRNA. For the induction of partial mirror image duplication of digits, limb cells have to be exposed to RA for 10 h and at least for 18 h for complete duplication (Eichele et al. 1985). Since this period is enough for the Msx-1 gene product to disappear and for the ectopic induction of the clustered homeobox gene observed after Msx-1 mRNA disappearance (Nohno et al. 1991), it is possible that repression of Msx-1 transcription followed by the disappearance of its product is one of the initial events for mirror-image duplication induced by RA. Our model is that RA releases the cells in anterior necrotic zone from the necrotic lineage restriction generated by Msx-1 protein, and allows mesenchyme cells to have a chance to reprogram toward potentially cartilage-forming lineage. The finding that this type of RA treatment abolished cell death at the anterior margin of the wing bud (Lee, 1986) supports this model. The expression in the distal AER is weakened by RA treatment but never disappears. This correlates with the finding that the manipulation generates ectopic induction of duplicated digits but does not affect the intrinsic development of the digits (Tickle et al. 1989).

One possible mechanism for Msx-1 mRNA disappearance after local RA treatment is an effect of transcriptional repression of the Msx-1 gene by a high concentration of RA. The other possibility is a facilitation of mRNA degradation. Since accumulation of cytoplasmic βactin mRNA is not affected, we favor the first possibility. The RA binds to its specific receptor (RAR) for the regulation of transcription and the mode of the regulation by the ligand –receptor complex differs from gene to gene. The transcription of RAR βgene itself is activated by retinoic acid (de Thé et al. 1990) whereas EGF receptor transcription is repressed (Hudson et al. 1990). In addition, accumulation of chicken homologue of Xenopus XlHbox1 homeo-domain protein is increased by treatment with RA in limb bud (Oliver et al. 1990). Induction of RAR βin limb mesenchymal cells by local application of RA already reaches a maximum 4h after the treatment (Noji et al. 1991) and the induction precedes the decrease of Msx-1 transcripts. Since accumulation of RAR β transcripts in Msx-1 expression domain of normal limb is very low (Noji et al. 1991), it is reasonable that overproduced RAR β or RAR β/RA complex repress the Msx-1 transcription.

Since the Msx-1 expression domain in the limb does not directly correspond to the distribution of any RAR mRNA (Dollé et al. 1989b), the regulative interaction of RAR on Msx-1 in normal development is puzzling. The presence of a variety of RAR genes is already reported (for example Mangelsdorf et al. 1990) and other types of retinoid are reported to show interesting distributions in the limb bud (Thaller and Eichele, 1990). It is still possible that these retinoid and unidentified RARs control Msx-1 expression during normal limb development. The neural crest and its derived cells contain RAR (Osumi-Yamashita el al. 1990; Dollé et al. 1990) where Msx-1 expression is also observed, and RA treatment is reported to affect the differentiation of these cells (for example Sulik et al. 1988). We do not know the effect of RA on the Msx-1 expression in these cells, however, it is possible that RA and RAR only function in Msx-1 regulation in these cells during normal development.

We cannot exclude the possibility that RA induces degradation of specific type mRNA. A molecular approach is ongoing to identify the RA receptor complex responsible element in the cis regulatory region of Msx-1.

The authors thank Dr H. Ide for critical reading of this manuscript and Dr M. Obinata for helpful discussion. We are grateful to Dr W. J. Gehring for supplying the Drosophila msh cDNA clone and Dr R. E. Hill for providing us with information on murine Hox-8 before publication. This research was supported by the Naito Foundation and Grants-in-Aid from the Ministry of Education, Science and Culture of Japan.