Retinoid signaling is required for the establishment of a ZPA and for the expression of Hoxb-8, a mediator of ZPA formation

Vertebrate limb development begins with the specification of limb position and polarity, and the initiation of a limb bud. Tissue transplantation experiments and fate mapping studies indicate that in the case of the chick limb, position and polarity along the anteroposterior axis are specified as early as Hamburger-Hamilton stages 8 to 11 (Chaube, 1959). An important event during the first phase of limb development is the appearance of polarizing activity in the lateral plate mesoderm (Hornbruch and Wolpert, 1991). Polarizing activity is defined as the ability of tissue to induce additional digits when grafted to the anterior wing bud margin (e.g. Saunders and Gasseling, 1968). By stage 14/15, the lateral plate mesoderm locally condenses and, subsequently, through a ventral involution of the somatopleure identifiable limb buds form around stage 16/17. At this time, the distalmost portion of the ectoderm that encloses the limb bud thickens and forms the apical ectodermal ridge (AER). The AER is required for further limb outgrowth (Saunders, 1948; Summerbell, 1974). By stage 18, polarizing activity becomes concentrated to the posterior mesenchyme of the limb bud, thereby giving rise to the zone of polarizing activity (ZPA). It is believed that the ZPA directs patterning along the anteroposterior limb axis (Tickle et al., 1975) and, in conjunction with signals produced by the AER (Summerbell et al., 1973), promotes a rapid distal elongation at the tip of the limb bud. Around stage 22, differentiated tissues such as cartilage or muscle precursors appear.

Much has been learned about molecules guiding limb bud growth and patterning. The list of important factors includes SHH, FGF-4, BMP-2 and Wnt-7a (Riddle et al., 1993; Francis et al., 1994; Laufer et al., 1994; Niswander et al., 1994; Yang and Niswander, 1995; Duprez et al., 1996; for reviews see Tickle and Eichele, 1994; Tabin, 1995; Tickle, 1995). By contrast, much less is known about the nature of the factors that control the early phase of limb development. It is proposed that Hox genes of paralog groups 5 to 8 specify forelimb position, possibly in a combinatorial fashion (Tickle and Eichele, 1994; Burke et al., 1995; Tickle, 1995). It is conceivable that certain Hox genes also mediate the development of the polarizing region; the chief evidence for this view is that ectopic expression of Hoxb-8 at the anterior margin of the forelimb bud induces an ectopic ZPA (Charité et al., 1994). The fibroblast growth factor FGF-8 is implicated in controlling the initial outgrowth of limb buds. The main evidence for this notion is that ectopic FGF-8 applied to the lateral plate evokes the formation of an additional limb and that fgf-8 is expressed in the intermediate mesoderm at the level of the future limb buds as well as in the presumptive and, later, in the definite AER (Crossley et al., 1996; Vogel et al., 1996). Retinoids act early in limb development because disruption of the retinoid signaling pathway by retinoid receptor antagonists prior to limb bud outgrowth results in truncated limbs (Helms et al., 1996). Furthermore, alltrans-retinoic acid (RA) locally applied to limb buds, rapidly induces genes implicated in early limb development (Hoxb-6/8, Lu et al., 1997) and, at a substantially slower rate, genes of the later phase (shh, bmp-2, fgf-4, Riddle et al., 1993; Francis et al., 1994; Helms et al., 1994; Niswander et al., 1994).

In this study, we have explored the role of retinoids during the early limb development. We provide evidence that Hoxb-8, a gene implicated in the establishment of the ZPA (Charité et al., 1994), is a primary response gene to the retinoid signal. We show that the expression domain of Hoxb-8 prefigures two important features of limb development. (1) Already by stage 8, the rostral boundary of its expression domain falls within the future wing region. (2) The initially broad expression domain of Hoxb-8 in lateral plate mesoderm develops into a domain co-extensive with the tissue that exhibits polarizing activity. Targeting retinoid receptor antagonists into this domain not only abolishes Hoxb-8 expression, but also prevents the establishment of a ZPA.

RT-PCR was used to generate a chicken Hoxb-8 fragment containing the homeobox of this gene (Scotting et al., 1990). A chick stage 14-17 embryonic λZAPII cDNA library was screened with this fragment and positive clones were isolated and sequenced. Two overlapping cDNAs defined the complete coding region of the Hoxb-8. Sequence comparison was performed using the software Wisconsin Package, version 8, 1994, Genetics Computer Group, Madison, Wisconsin. The GenBank accession number of the Hoxb-8 sequence is U81801.

White Leghorn chicken embryos were used as donors and hosts. To analyze for polarizing activity, posterior tissue of either normal or antagonist-treated wing buds was transplanted to the anterior margin of a stage 20 embryo wing bud (Fig. 1B). Host embryos were incubated to day 10 and processed as described (Wedden et al., 1990).

AG1-X2 ion-exchange beads of 200 μm diameter were soaked for 5 hours in a mixture of 2.4 mg/ml LG754 (RXR selective antagonist; see Lala et al., 1996) and 2.4 mg/ml LG629 (RAR selective antagonist, identical to Ro 41-5253, Apfel et al., 1992). Embryos with 20 to 24 somites (stage 13 to 14, Hamburger and Hamilton, 1951) were slightly stained with Neutral Red and two incisions were made in the lateral plate at somite levels 15 and 20; beads were placed into these slits. The soaking concentration used in the present study was 5 times higher than that used in an earlier study (0.5 mg/ml; Helms et al., 1996). With 2.4 mg/ml, complete hand plate truncations were observed in 67% of wings, 24 % of the wings had just a digit 3 and/or 2 and 9% of the wings were normal (n=12). Furthermore 58% of the wings had no ulna. All-trans-retinoic acid (RA, agonist, 100 μg/ml), LG754 or LG629 treatment of stage 20 wing buds was performed as previously described (Helms et al., 1994) using AG1-X2 beads of 250-300 μm diameter.

The RXR antagonist LG754 is a potent inhibitor of RXR homodimermediated transactivation, but can induce a reporter gene through RXR-RAR heterodimers (Lala et al., 1996). However, the gene expression data reported here show that LG754 and LG629 are not acting agonistically. We applied beads soaked in LG629 and/or LG754 to the anterior margin of stage 20 wing buds and monitored, by in situ hybridization, the expression of RARβ, a direct RA target gene (de Thé et al., 1990; Sucov et al., 1990). LG754 and LG629, applied alone or in combination, did not induce RARβ. Instead, the combined treatment with LG629 and LG754, or that with LG629 alone, markedly decreased the expression of RARβ in the vicinity of the bead (data available upon request).

Synthesis of digoxigenin-tagged riboprobes of shh (Riddle et al., 1993; Roelink et al., 1994), bmp2 (Francis et al., 1994), fgf-8 (Crossley et al., 1996), Hoxb-8 (nt −51 to nt 726), Hoxd-11, Hoxd-13 (Izpisúa-Belmonte et al., 1991) and RARβ (nt 1 to nt 612, Smith and Eichele, 1991), was carried out with a Stratagene RNA transcription kit following the procedure described in Albrecht et al. (1997). Whole-mount in situ hybridization was performed with digoxigeninlabelled riboprobes as described by Albrecht et al. (1997).

Whole limb buds from stage 20 chick embryos were dissociated to single cells by trypsin treatment. Cells were cultured at 37 ° C in DMEM containing 10% fetal calf serum (Gibco, BRL) until they had attached to the culture plate. Thereafter, medium containing either 1 μM RA, or 30-50 μM cycloheximide (Sigma), or the combination of RA and cyloheximide was added. Cells were harvested for RNA purification after 6 hours of treatment and total RNA was isolated using RNAzol (Tel-Test). Northern analysis was carried out with 20 μg total RNA per lane (Sambrook et al., 1989). Filters were probed with ³²P-labelled DNA probes corresponding to Hoxb-8, RARβ and Pax-3 (Goulding et al., 1994).

When retinoid receptor antagonists are applied to the presumptive wing region, the resulting wings lack either all three or the two posteriormost digits and one or both forearm elements (see Material and Methods and Helms et al., 1996). Thus, proximodistal and anteroposterior axes are both affected. Since signals from the ZPA are required for anteroposterior patterning as well as distal growth (reviewed by Tickle and Eichele, 1994; Tabin, 1995; Tickle, 1995), the dual effects of anti-retinoids could best be understood if retinoids are needed for the formation of the ZPA. To test this hypothesis, we assayed for polarizing activity in anti-retinoid-treated wing buds. Antagonists were applied from beads implanted into the lateral plate at the rostral and caudal limits of the wing region opposite somites 15 and 19/20 (Fig. 1A). When treated embryos had reached stage 18, tissue surrounding the posterior bead was excised and grafted to stage 20 host wing buds (Fig. 1B). Of a total of 17 wings that received antagonist-exposed grafts, ten exhibited a normal digit pattern, five had an additional cartilage rod (d) within the hand plate, one had an additional digit 2 and one exhibited a 4334 duplication (Fig. 1C). Thus, with the exception of one specimen, anti-retinoid-treated tissue had its polarizing activity greatly reduced. Control ZPA taken from untreated stage 18 embryos resulted in nine out of ten cases in full digit pattern duplications. This functional assay thus demonstrates that anti-retinoids block the establishment of a ZPA.

The absence of ZPA was further substantiated by an analysis of molecular markers that are normally expressed in posterior limb bud mesenchyme. We first examined how anti-retinoid treatment affected the expression of shh, a polarizing signal (Riddle et al., 1993; Chiang et al., 1996). We found that, in six out of seven embryos, shh mRNA was not detectable in the treated wing bud while expression was unaffected on the contralateral side (Fig. 1D, arrowhead). The other marker analyzed was bmp-2, which is normally expressed in the posterior mesenchyme of limb buds in and around the shh expression domain and throughout the AER (Francis et al., 1994). Since ZPA grafts (Francis et al., 1994), ectopic expression of shh (Laufer et al., 1994) or RA application (Francis et al., 1994; Helms et al., 1994) activate bmp-2, this gene is considered to be a reporter for polarizing signals (Francis et al., 1994; Tabin, 1995). We found that retinoid receptor antagonists applied at stage 13/14 prevented the expression of bmp-2 in the posterior mesenchyme (n=7; Fig. 1E), but had no effect on the expression of bmp-2 in the AER. The absence of a ZPA and the loss of shh and bmp-2 expression could be accounted for if anti-retinoids abolished posterior mesenchymal cells. This is not the case, since Hoxd-13, normally expressed in the posteriormost wing bud mesenchyme (Fig. 1F; see Izpisúa-Belmonte et al., 1991; Nohno et al., 1991) was still expressed in antagonist-treated buds (n=13, Fig. 1F). Anti-retinoidtreated embryos were also stained with Nile blue vital dye to inspect for cell death. There was no evidence for an abnormal pattern of cell death in treated embryos at any stage examined (stage 16-22). Taken together, our gene marker studies show that shh, a polarizing signal, as well as bmp-2 are not expressed in retinoid receptor antagonist-treated wing primordia. By contrast, expression of Hoxd-13 as well as Hoxd-11 (data not shown) was not significantly affected, arguing against a loss of posteriormost mesenchymal cells.

When wing buds first appear (stage 16/17), the morphology of anti-retinoid-treated and control buds do not differ. The absence of morphologic differences does not preclude that antagonists already by this early stage compromise outgrowth. To investigate this possibility, we examined the expression of fgf-8, which is considered to be the prime candidate for initiating limb bud outgrowth (see Introduction). We found that, at any stage examined (stage 16-22), fgf-8 expression was not visibly affected by retinoid receptor antagonists (n=10); however, in the same specimens, expression of shh was not detectable (Fig. 1D). Taken together, the absence of change in early limb bud morphology and in the expression of fgf-8 suggest that anti-retinoids do not directly interfere in any obvious way with limb bud outgrowth. Instead, anti-retinoids effectively block the establishment of the ZPA as revealed by grafting experiments and the loss of expression of shh and bmp-2. The absence of the ZPA then precludes the establishment of a feed back loop between AER and the mesenchyme (reviewed e.g. by Tabin, 1995) and distal growth is markedly reduced, eventually resulting in an absence of distal skeletal elements.

Our experiments show that retinoid signal transduction is required for the establishment of a ZPA and for the expression of shh and bmp-2. Although shh and bmp-2 expression are inducible in the limb bud by retinoids (Riddle et al., 1993; Francis et al., 1994; Helms et al., 1994), the activation of both these genes requires ≥20 hours of treatment (Helms et al., 1994). This suggests the existence of intermediate factors that link retinoid signaling with the activation of shh and bmp-2 genes. It has long been proposed that Hox genes provide positional information along the anteroposterior body axis (reviewed in McGinnis and Krumlauf, 1992) and Hox genes may determine the position of limbs (Tickle and Eichele, 1994; Tabin, 1995; Gérard et al., 1996). Initial studies have shown that Hoxb-8 is activated in the wing bud by RA within 6 hours (Lu et al., 1997). Moreover, it was shown that the ectopic expression of Hoxb-8 at the anterior margin of the mouse forelimb bud results in the formation of an ectopic polarizing region (Charité et al., 1994). This finding, and the retinoid responsiveness of Hoxb-8, make the gene product one of the candidates for transducing the retinoid signal. In what follows, we provide evidence for this idea.

A Hoxb-8 cDNA encompassing the entire protein coding region was isolated as described in Material and Methods. Sequence comparison with the mouse Hoxb-8 gene confirmed that the isolated chick cDNA indeed encoded Hoxb-8 (Fig. 2). The homeodomains of mouse Hoxb-8 and our chicken protein are identical, and the N-terminal domains exhibit 91% identity. By contrast, the identity between our sequence and the N-terminal domains of mouse paralogues Hoxc-8 and Hoxd-8 is only 55% and 42%. Therefore, we conclude that the isolated chicken cDNA encodes Hoxb-8.

The anterior wing bud margin of a stage 20 embryo is a convenient model system for testing the inducibility of limb patterning genes by polarizing agents. We used this feature to assess the time course of inducibility of Hoxb-8 by RA. We found that RA released from an implanted bead, rapidly and transiently induced the expression of Hoxb-8 in the mesenchyme surrounding the implant (Fig. 3A-C). The loss of Hoxb-8 expression is not due to degradation of RA; earlier studies have shown that, in the presence of a bead continuously releasing RA, this compound persists in the wing bud for at least 20 hours (Eichele et al., 1985). We note that Hoxb-8 transcripts are found throughout the mesenchyme, bordering the RA source and at a distance from the AER (Fig. 3A,B, the AER straddles the bead). This is in contrast with the RA-induced ectopic expression domain of shh, bmp-2 and Hoxd genes, which is limited to tissue distal to the bead and adjacent to the AER (Helms et al., 1994, 1996). This suggests that unlike shh, bmp-2 and Hoxd, Hoxb-8 expression is independent of signals from the AER. ZPA grafts present for 6 hours (Fig. 3E, arrowhead), or even 12, 18 or 24 hours (data not shown) do not induce Hoxb-8, suggesting that this gene is not responding to, but is upstream of, the polarizing signal. Intriguingly, the induction of Hoxb-8 by RA is site-specific since anteriorly but not posteriorly implanted beads induce this gene (Fig. 3B). This lack of Hoxb-8 induction is not due to a inability of posterior tissue to respond to RA, since RARβ is rapidly induced in posterior cells by exogenously applied RA (data not shown). More likely, posterior cells, as a result of having become a ZPA, are prevented from activating a gene potentially involved in the induction of a ZPA.

The induction of Hoxb-8 is as rapid as that of RARβ, a direct retinoid target gene (compare Fig. 3A and D; de Thé et al., 1990; Sucov et al., 1990). It is thus plausible that like RARβ, Hoxb-8 is a direct response gene. To investigate this possibility, we dissociated stage 20 wing buds into single cells and exposed these primary cultures to 1 μM RA for 6 hours in the presence of cycloheximide. Optimal concentrations of cycloheximide not affecting cell viability were empirically determined as 30 μM to 50 μM. Fig. 4 shows that, in the absence of RA, RARβ mRNA levels are virtually undetectable (lane 1).

Adding cycloheximide augmented mRNA levels presumably as result of increasing the stability of mRNA (Lund et al., 1991). RA by itself induced RARβ, but combination treatment with RA and cycloheximide strongly induced RARβ expression. Hoxb-8 induction exhibited a very similar behavior (Fig. 4). Untreated cells showed low levels of transcript, cycloheximide and RA-augmented expression. Combination treatments strongly induced Hoxb-8 expression. Taken together, these in vitro and in vivo studies demonstrate that the induction of Hoxb-8 by RA is rapid and does not require protein synthesis, a behavior also typical for direct target genes such as RARβ.

So far, we have shown that ectopic RA rapidly induced Hoxb-8 in the anterior mesenchyme of stage 20 wing buds, some of which will subsequently express shh andbecome a ZPA (Noji et al., 1991; Wanek et al., 1991; Helms et al., 1994). This raises the possibility that Hoxb-8 transduces the retinoid signal that is required for the formation of a ZPA during normal development. The pattern of expression of Hoxb-8 in the lateral plate supports this idea.

The expression profile of Hoxb-8 can be divided into five phases (Figs 5, 6).

Hoxb-8 mRNA appears by stages 5 and 6 in the posterior third of the primitive streak (Fig. 5A) in a region, part of which gives rise to the lateral plate (Psychoyos and Stern, 1996). After the formation of the first few somites (Fig. 5B), Hoxb-8 transcripts appear in the neuroectoderm in which a distinct expression boundary at the border between somites 6 and 7 is established (arrow in Fig. 5B). This boundary is maintained throughout development (Fig. 5 and data not shown). There is a high level of Hoxb-8 expression in the region surrounding Hensen’s node, a previously identified source of retinoic acid (Chen et al., 1992; Hogan et al., 1992). By early stage 9, very low levels of expression are also seen in the lateral plate (Fig. 5B).

This phase is characterized by a significant up-regulation of Hoxb-8 expression in the lateral plate (Figs 5C-F, 6A). Initially (Fig. 5C,D) expression in this tissue extends slightly anterior of the rostral limit of the wing field (the boundaries of this field are indicated by green dots; see Chaube, 1959). Subsequently, expression in the lateral plate slightly retracts and the anterior boundary of expression now coincides with the anterior border of the wing region (Fig. 5E,F). The expression of Hoxb-8 in the lateral plate is graded with the gradient diminishing in a posterior-to-anterior direction (e.g. Fig. 5E). During phase II, the anterior expression boundary of Hoxb-8 in the presomitic mesoderm shifts rostrally (Fig. 5E,F). Transverse sections (not shown) illustrate the presence of Hoxb-8 mRNA throughout the mesoderm of the somatopleure and splanchnopleure.

The anterior boundary of Hoxb-8 expression in the lateral plate mesoderm retracts towards the level of somite 17 (center of wing field, Figs 5G, 6B) and the expression in posterior mesoderm around the tail bud begins to decrease (data not shown).

A wing bud appears (Fig. 5H, between green lines) and, in the lateral plate, marked anterior and posterior Hoxb-8 expression boundaries are established at the level of somites 17 and 24, respectively (Fig. 6C). Hoxb-8 is also expressed in somitic mesoderm from somite 16 on, back to somite 25/26.

This phase is characterized by rapid disappearance of Hoxb-8 expression in the lateral plate and in wing bud mesoderm (Fig. 5I), a reduction of expression in the somites, and the definition of a posterior boundary of Hoxb-8 expression in the neural tube at the level of somite 24. Temporally, the down-regulation of Hoxb-8 expression in the nascent wing bud and the lateral plate posterior to it coincides with the initial activation of the shh gene (Riddle et al., 1993). Hoxb-8 mRNA is just disappearing at the time when shh transcripts are detectable by in situ hybridization. The dynamic expression of Hoxb-8 relative to anatomical markers of the wing field and to the maps of polarizing activity is summarized in Fig. 6.

The expression pattern of Hoxb-8 in the lateral plate is remarkable for two reasons. First, the anterior boundary of Hoxb-8 expression is located within the wing region (red box in Fig. 6) at a time when this region is being defined. Transplantation studies (Chaube, 1959) show that the anteroposterior axis of the wing region is irreversibly specified as early as stages 8 to 11, a period when Hoxb-8 expression undergoes retraction and an anterior boundary within the wing region is established (Figs 5C-F, 6A,B). These observations suggest a correlation between the establishment of an anterior expression boundary of Hoxb-8 and the determination of the anteroposterior axis within the wing field. Second, Hornbruch and Wolpert (1991) have mapped the appearance of polarizing activity in the lateral plate mesoderm of the chick. These studies were based on grafting blocks of lateral plate tissue to the anterior wing bud margin of a host embryo, followed by analysis of digit pattern duplications. While very weak polarizing activity was demonstrated in the lateral plate mesoderm as early as stage 81. Such activity begins to augment between stages 11 and 13 and eventually encompasses the entire wing field. Subsequently polarizing activity begins to recede to the posterior third of the wing bud (between phases III, IV, Fig. 6B, C). During phases III and IV, this dynamic distribution of polarizing activity closely reflects the spatiotemporal expression pattern of Hoxb-8 (Fig. 6B,C, compare yellow and purple areas in the lateral plate). We conclude that the anterior boundary of Hoxb-8 expression in the lateral plate prefigures the establishment of a wing field in much the same way as e.g. the anterior boundary of Hoxa-1 and b-1 expression anticipates the cranial boundary of rhombomere 4 in the hindbrain (Sundin and Eichele, 1990; Murphy and Hill, 1991). What is more, the expression domain of Hoxb-8 coincides, at least during phases III and IV, and within the accuracy of the available maps, with the domain of polarizing activity in the wing region.

We have shown that blocking retinoid signal transduction prevents the formation of a ZPA. Because polarizing activity and Hoxb-8 expression are associated (Figs 5 and 6 and Charité et al., 1994), we predict that retinoid receptor antagonists should not only abolish the ZPA but also the expression of Hoxb-8 in the wing region. Anti-retinoids were locally applied to the prospective wing region (Fig. 1A) and, after 8-10 hours, embryos were analyzed for Hoxb-8 mRNA. Treated embryos either lacked Hoxb-8 transcripts (n=10) or showed a marked reduction (n=8) of expression in the treated wing primordium but not on the contralateral side (arrowhead in Fig. 7). Beads soaked in DMSO, the solvent for antagonists, had no effect on Hoxb-8 expression (data not shown). Thus, blocking of the retinoid signaling pathway results either in a significant downregulation or a loss of endogenous Hoxb-8 expression paralleling the loss of polarizing activity of this tissue.

Retinoid are required for normal limb development since an interruption of the retinoid signaling pathway with retinoid receptor antagonists prevents the formation of a normal limb pattern (Helms et al., 1996). Similarly, blocking RA synthesis in the prelimb bud stage using the dehydrogenase inhibitor disulfiram, blocks limb formation (Stratford et al., 1996). Effects of retinoid receptor antagonists are particularly striking when antagonists are provided to the wing region prior to bud formation, suggesting that retinoids operate early during limb development. This prompted us to search for factors that transduce the retinoid signal at this developmental stage. We suggest that Hoxb-8, a gene product previously associated with polarizing activity (Charité et al., 1994), is such a molecule. We show that Hoxb-8 is rapidly induced by RA in the absence of protein synthesis, and that retinoid receptor antagonists abolish the expression of Hoxb-8 in the lateral plate. Retinoid receptor antagonist treatment of the presumptive wing bud did not abolish wing bud initiation, and the expression of fgf-8, a mediator of limb bud initiation (Crossley et al., 1996; Vogel et al., 1996), was not affected. However, such buds lacked a functional ZPA and did not express sonic hedgehog and bmp-2, two signaling molecules implicated in limb patterning (Riddle et al., 1993; Duprez et al., 1996).

This study provides evidence that Hoxb-8 is directly regulated by retinoids. Retinoid receptor antagonists block Hoxb-8 expression in the lateral plate mesoderm (Fig. 7). Additionally, exogenous RA rapidly induces Hoxb-8 in wing bud mesenchyme (Fig. 3), and cultured limb bud cells turn on this gene in response to RA treatment in the presence of a protein synthesis inhibitor (Fig. 4). Although our studies have not attempted to identify a retinoid responsive element in Hoxb-8, the transcriptional regulation of several Hox genes by RA is well documented (see Hofmann and Eichele, 1993 for a review). Retinoid response elements have been identified e.g. in Hoxa-1, Hoxb-1, Hoxd-4 (for review see Krumlauf, 1994) and Hoxd-10/11 (Gérard et al., 1996). Endogenous RA is produced at the right time and location in the embryo to activate Hoxb-8 in the lateral plate. We have previously shown that lateral plate tissue synthesizes RA at a high rate (Helms et al., 1996). Furthermore, transgenic reporters under control of a retinoid responsive element reveal the presence of RA in the lateral plate (Reynolds et al., 1991; Rossant et al., 1991; Balkan et al., 1992).

Ectopic expression of Hoxb-8 in the mouse forelimb induces an ectopic ZPA. Based on this, it is proposed that Hoxb-8 has a role in the establishment of the ZPA in the lateral plate (Charité et al., 1994). Our studies in the chick now show that polarizing activity is for the most part co-extensive with Hoxb-8 expression. During stages 11 and 12, weak polarizing activity appears in tissue extending lateral of somite 14/15 to prospective somite 23/24 (Hornbruch and Wolpert, 1991). Hoxb-8 is expressed in this region (Fig. 5D-F). After stage 13, polarizing activity in the lateral plate substantially increases (Hornbruch and Wolpert, 1991). Fig. 6B and C show that, during phases III and IV (stage 13-17), the rostral and caudal limits of polarizing activity and the expression of Hoxb-8 essentially coincide within the resolution of the polarizing activity mapping data. Charité et al. (1994) have suggested that Hoxb-8 expression is correlated with polarizing activity. Since there are no maps of polarizing activity for the mouse, this remained a conjecture that the present study affirms.

This association of polarizing activity with Hoxb-8 expression, on the one hand, and the dependence of Hoxb-8 transcription on retinoid signaling, on the other hand, suggests that Hoxb-8 is a link between RA signaling and the establishment of polarizing activity. It is possible that Hoxb-8 is not the only RA-regulated Hox gene that has this capacity. For example, Hoxb-6 and Hoxc-6 are also induced by RA (Oliver et al., 1990; Lu et al., 1997) and are expressed in the lateral plate in a pattern similar to that of Hoxb-8 (DeRobertis, 1994; H.-C. Lu and G. Eichele unpublished data). It is well known that Hox genes can function in a combinatorial manner (Horan et al., 1995; Rancourt et al., 1995; Davis and Capecchi, 1996). Therefore, Hoxb-8 may be only one of several Hox genes capable of specifying polarizing activity. This would imply that a loss of function in Hoxb-8 alone might not affect limb patterning. In fact, the deletion of Hoxb-6 gene has no effect on the limb pattern (Rancourt et al., 1995).

What is the relationship between shh and Hoxb-8 expression? It has been proposed that shh is expressed in those Hoxb-8-positive cells that are in proximity to the AER and thus receive a FGF signal (Charité et al., 1994; Tabin, 1995). While the general idea of AER proximity is correct, expression data suggest a more complex regulation. During stages 16 and 17, Hoxb-8 expression in the lateral plate is very strong and extends up to somite level 17/18 (Fig. 6C). shh transcripts are not found at stage 16 (Crossley et al., 1996) but, by stage 17, a small number of shh-positive cells are detected in the posteriormost wing bud mesenchyme, opposite the boundary somites 19/20. By stage 18, about 5 hours later, the population of shh-expressing cells has considerably expanded, but now Hoxb-8 transcripts have disappeared. Apparently, there seems to be only a brief temporal overlap between Hoxb-8 and shh expression and furthermore, only a subset of Hoxb-8-positive cells will eventually express shh. Thus there are mechanisms limiting shh expression to a subset of Hoxb-8-positive cells. The expression of shh may depend on the simultaneous action of several Hox genes (Hox code), or Hoxb-8 function could require locally expressed co-activators (Pöpperl et al., 1995). Another hypothetical mechanism to limit shh expression would be a localized release of a growth factor by the posteriormost AER cells. Recent work from our laboratory suggests the existence of a RA-induced ectodermal factor that is required for the formation of a ZPA (Helms et al., 1996).

The dynamic behavior of Hoxb-8 expression in the lateral plate of chick and mouse is very similar (this study; Deschamps and Wijgerde, 1993; Charité et al., 1994). In the mouse, forelimbs arise between somites 7 and 11; in the chick, the wings arise between somites 15 and 20 (see Burke et al., 1995 for a discussion of axial position of vertebrate limbs). In a 9-somite mouse embryo, the anterior boundary of Hoxb-8 expression in lateral plate is at somite level 9 (Charité et al., 1994). In a 13-somite mouse embryo, this boundary has retracted to somites 10/11. In the chick, there is an analogous caudal retraction (Fig. 6). First, the anterior expression boundary resides opposite somite 15/16 then it repositions to opposite of somite 17/18. The anterior shifting of the caudal expression boundary of Hoxb-8 is also very similar between chick and mouse. Thus, within the resolution of the expression data, the establishment of Hoxb-8 expression boundaries in the lateral plate is conserved between chick and mouse, despite the fact that the two species have their forelimbs at different axial levels.

Fate maps in the chick define limb-forming regions well before limb buds appear (Chaube, 1959). The present study shows that the anterior boundary of Hoxb-8 expression in the lateral plate resides within the prospective wing region from the time such a region can be delineated (Fig. 5, green dots; Fig. 6, red box). Additional evidence for a role of Hox genes in limb positioning comes from comparisons of Hox gene expression patterns in different species (see above and e.g. Burke et al., 1995). Furthermore, mutations in Hoxb-5, d-10 and d-11 genes can shift limb position (Rancourt et al., 1995; Gérard et al., 1996). Hoxd-10 and 11 are regulated by retinoids (Gérard et al., 1996) and Hoxb-5 is inducible by RA in the limb bud (G.E. unpublished data). Of note, local treatment of the wing-forming region with disulfiram, a RA synthesis blocker, can result in a posterior displacement of the wing (Stratford et al., 1996). We suggest that absence of RA in the wing region prevents the expression of Hox genes in the wing field, but since the blocking agent acts locally, the expression of these Hox genes are not affected more posteriorly, and thus the wing arises at a more posterior site. The application of retinoid receptor antagonists to very early chick embryos may further clarify this interpretation. Taken together, it is possible that retinoids are not only involved in the establishment of polarizing activity but are also mediate, through genes such as Hoxb-8, the specification of limb position.

We would like to thank Dr Randy L. Johnson for critical reading and review. Dr Richard Heyman at Ligand Pharmaceuticals, San Diego CA provided LG629 and LG754. This work was supported from a grant by the NIH to G. E. (HD-20209) and a fellowship by Swiss National Science Foundation to J.-P. R. (823A046697). H.-C. Lu is a recipient of a fellowship from the Markey Charitable Trust.