Sonic hedgehog is not required for polarising activity in the Doublefoot mutant mouse limb bud

Within the vertebrate limb bud, three interacting tissues are responsible for the organisation of skeletal pattern. These are the apical ectodermal ridge (AER), a thickened rim of distal ectoderm, (2) the progress zone (PZ), the mesenchyme underlying the AER, and (3) the zone of polarising activity (ZPA), a specialised area of posterior distal mesenchyme. By maintaining the PZ cells in a proliferative state, signals from the AER maintain proximodistal outgrowth of the limb bud, while cells leaving the PZ differentiate to form the limb skeletal pattern in a proximodistal sequence. Signals from the ZPA and AER maintain each other, and the ZPA governs the anteroposterior pattern of the limb, as shown most clearly by the three distinctly different digits of the chick wing. In most mammalian limbs, including mouse and human, the anteroposterior digital pattern is characterised by an anterior digit with two phalanges (numbered I) and four similar triphalangeal digits (numbered II-V).

The mouse mutant Doublefoot (Dbf), which shows polydactyly with 6-8 digits per limb, is a particularly informative resource for investigating the molecular mechanisms governing limb pattern. The phenotype is inherited in a single, dominant Mendelian manner and is fully penetrant; the mutation has been mapped to chromosome 1 (Lyon et al., 1996). There is no digit I, all digits invariably being triphalangeal. Some limbs have one or two digits that are not in contact with the carpus/tarsus (Fig. 1B); digital bifurcation is also sometimes present (Fig. 1C). The mutant also shows tibial defects and craniofacial abnormalities, including a broadened, bulbous skull, and facial clefts in homozygotes. Dbf limbs are morphologically distinguishable by E10.5, being enlarged along the anteroposterior axis (Fig. 2).

The protein Sonic hedgehog (Shh) has been implicated as the active signalling molecule of the ZPA (Riddle et al., 1993). The ZPA is defined as a region of posterior limb mesenchyme with the ability to induce mirror-image duplications when transplanted to the anterior margin of a chick wing bud (Saunders and Gasseling, 1968). Ectopic Shh gene expression mimics the effect of ZPA grafts and can induce the ectopic expression of downstream genes such as bone morphogenetic protein 2 (Bmp2), 5’HoxD genes and fibroblast growth factor 4 (Fgf4) (Laufer et al., 1994). These downstream targets only possess a subset of the properties of Shh itself (Francis et al., 1994; Duprez et al., 1996). This information, together with the localization of Shh transcripts to the ZPA, has been taken to suggest that Shh is the active ZPA signal (Riddle et al., 1993). This interpretation is supported by the observation that preaxial polydactyly is associated with ectopic preaxial Shh expression in Strong’s Luxoid and other mutant mouse strains (Chan et al., 1995; Masuya et al., 1995, 1997).

Current understanding of the molecular nature of Shh- mediated patterning of the limb is incomplete. It is now known that the active amino peptide of Shh (Shh-N), even when cleaved from the C-terminal domain, is anchored to the cell surface by cholesterol (Porter et al., 1995, 1996). This lack of mobility, which is also indicated by co-localisation of Shh peptide and mRNA (Martí et al., 1995), spatially limits the range of direct influence of Shh at the molecular level in contrast to its functional influence, which spreads across the full width of the limb bud. Shh is the ligand for the cell surface receptor patched (Ptc) (Marigo et al., 1996a; Stone et al., 1996), which inhibits activation of another predicted cell surface receptor-like molecule, smoothened (Smo) (van den Heuvel and Ingham, 1996). Ptc also serves to reinforce the limited spatial distribution of hedgehog by sequestration of the ligand (Chen and Struhl, 1996). Binding of Shh to Ptc lifts the repression of Smo, leading to activation of the transcription factor Gli, and the transcription of direct target genes of Shh signalling. These include Ptc and Gli themselves (reviewed by Ruiz i Altaba, 1997).

Analysis of Dbf mutant limbs demonstrates that the as yet unidentified product of the Dbf locus ectopically activates the Shh signal transduction pathway in the absence of Shh gene expression. The resultant cascade of molecular genetic events is sufficient to confer polarising activity on anterior and distal limb bud mesenchyme cells. Our results identify a new element of the Shh pathway whose molecular function is integrated with those of Ptc and Smo, but which (at least in its mutated form) does not require Shh in order to activate signal transduction.

Embryos were obtained from matings between a Dbf/+ male or female and a C3H strain mouse. The day of the vaginal plug was designated E0. E10.5-11.5 Dbf/+ embryos were distinguished phenotypically from their wild-type littermates by their broader limb buds at E11.0/11.5 and by the characteristic diencephalic bulge on E10.5 (C. H., M. F. L. and G. M. M.-K., unpublished data).

Mice were fixed overnight in 95% alcohol and eviscerated; the skin was then removed under a dissecting microscope with watchmakers forceps and the mice were stained as previously described (Watson, 1977).

An 810 bp fragment from the full-length clone of the murine Patched gene (M. Scott) was amplified using the following PCR primers: 5’-TGACTGTTCCCGAGA- GTGC-3’ and 5’-CGAATAGACA- GAGGTAGGGTGG-3’ with a thermal cycler profile of (1) 94°C, 2 minutes; (2) 94°C, 30 seconds; (3) 60°C, 45 seconds; (4) 72°C, 45 seconds; (5) 72°C, 5 minutes; (6) an additional 34 cycles of steps 2-4. The resulting blunt-end-polished fragment was inserted into the SrfI site of pCR-script (Stratagene, La Jolla, CA), flanked by T3 and T7 promoters for the generation of antisense and sense riboprobes.

E10.5-11.5 embryos were dissected into Tyrode’s saline and their forelimbs were removed and processed for in situ hybridization (Wilkinson, 1992) with digoxigenin-labelled riboprobes for the genes Shh, Fgf4 (A. Mc Mahon), Hoxd13 (D. Duboule), Bmp2 (M. Jones), Gli (C. Hui) and Ptc (see above).

Mesenchymal cubes were cut from hindlimbs of freshly dissected E10.5-11.0 embryos (see above) following removal of the ectoderm; they were used as donor tissue for grafting under the anterior ectoderm of host Hamburger and Hamilton (1951) stage 20 chick wing buds (Izpisua-Belmonte et al., 1991b).

Shh and Fgf4 have been implicated as the main mediators of pattern along the anteroposterior and proximodistal axes of the developing limb (Laufer et al., 1994). We therefore investigated the expression domains of these two genes in Dbf embryonic limb buds. In order to make the most efficient use of the limited mutant tissue available, in situ probes for both genes were hybridized on the same specimen. Shh was detected only in its normal posterior mesenchymal domain (Riddle et al., 1993), in both wild-type (C3H) and Dbf/+ limbs (Fig. 2A,B) (n=20). The double detection method Shh signal was comparatively weak; therefore, in order to rule out the possibility that the anterior mesenchyme possessed low levels of Shh transcripts, or that ectopic expression was delayed as in other polydactylous mouse mutants (Masuya et al., 1995), mutant limbs were hybridized with a probe for Shh alone at a range of stages from E10.5 hindlimbs to E11.5 forelimbs showing early digital condensation (n=20). Again, Shh expression was detected only in the posterior mesenchyme of Dbf/+ limb buds (Fig. 2C and data not shown). Fgf4 was expressed uniformly throughout the apical ectodermal ridge (AER) of Dbf/+ limbs (n=20), in contrast to its normal posterior domain (Niswander and Martin, 1992) as detected in wild-type AERs (Fig. 2A,B). Expression of Fgf4 has been shown to be induced and polarised by Shh (Laufer et al., 1994). The presence in the mutant limbs of an Fgf4 expression domain with both Shh-dependent and Shh-independent components indicates that the Dbf mutant gene product is sufficient to induce and maintain Fgf4 expression in the overlying ectoderm. The observation that Fgf4 is induced throughout the AER in the mutant limbs, whereas Shh shows its normal pattern of expression, places the Dbf gene downstream of Shh.

For further information on the position of the Dbf mutation in the polarising region signalling pathway, we investigated the expression of Bmp2 and Hoxd13, two downstream targets of Shh. In Dbf/+ limbs, Bmp2 showed ectopic expression extending throughout the anterodistal mesenchyme, in addition to its normal domain in the posterior and posterodistal mesenchyme (Hogan, 1996) (Fig. 2D,E). This pattern is reminiscent of that of the polydactylous chick mutant, talpid, in which Bmp2 is expressed throughout the distal mesenchyme (Francis-West et al., 1995). Hoxd13, which is responsive to the polarising region (Izpisua-Belmonte et al., 1991a), was expressed uniformly throughout the distal mesenchyme of mutant limbs at E11, rather than being restricted to the posterodistal mesenchyme as in wild-type limbs (Fig. 2F,G). This information places the Dbf mutation genetically upstream of Bmp2 and Hoxd13.

Recent studies have shown ectopic Shh expression in several well- characterized mouse mutants, which phenotypically resemble the Dbf mutant, suggesting that the polydactylous phenotype is due to ectopic polarising activity in the anterior limb bud mesenchyme. These mutants show ectopic Fgf4 in the overlying anterior AER and, in some cases, ectopic (anterior) HoxD gene expression, all concomitant with ectopic Shh expression (Chan et al., 1995; Masuya et al., 1995, 1997). The Dbf mutant differs significantly from these previously described mutants in having supernumerary digital elements suggestive of ectopic polarising activity in spite of the absence of Shh expression.

In order to establish unequivocally whether the anterior mesenchyme of the Dbf mutant limbs possesses polarising activity, we grafted portions of the anterior mesenchyme from E10.5 and E11 Dbf/+ and wild-type limbs into the anterior margin of stage 20 chick wing buds (Izpisua-Belmonte et al., 1991b); posterior and mid-distal mesenchyme was also grafted (Fig. 3A). The results are summarised in Table 1 and illustrated in Fig. 3B-F. By definition, the presence of duplications arising as a result of the grafts indicates polarising activity (Honig and Summerbell, 1985; Hinchliffe and Samson, 1985). As expected (Tickle et al., 1976), surviving limb grafts from wild-type anterior limb bud mesenchyme (Fig 3B) had no effect, whereas grafts of wild- type posterior limb bud mesenchyme induced mirror-image duplication (Fig. 3C, 4/5). Grafts from all three sites in Dbf/+ limbs induced duplications (Fig. 3D-F; 5/7 posterior, 14/17 anterior, 3/5 mid-distal), although the polarising activity of mid-distal mesenchyme was present in early (E10.5, 3/3) but not later stage (E11, 0/2) embryos.

A significant observation concerning the pattern of the induced digits is that, although the size and number of the induced digits was variable, grafts from all three mesenchymal sites from Dbf/+ embryos induced a greater number of supernumerary elements than grafts of posterior wild-type mesenchyme. Some successful mutant grafts, in addition to inducing extra elements in the region of the graft, had effects on the ‘host’ digits (i.e. those formed posterior to the graft site), e.g. thickening and bifurcation (Fig. 3E).

To further elucidate the functional position of the Dbf gene and to rule out the possibility that the duplications and polarising activity were the result of a mutation in a parallel signalling pathway, we investigated the expression of elements of the Shh signal transduction pathway. The genes selected were (1) Ptc, which has been implicated as receiving the Shh signal by direct binding of Shh (Stone et al., 1996; Marigo et al., 1996a,b), and the transcription factor Gli, a transactivator of Shh targets in cells responding to Shh (Alexandre et al., 1996; Hepker et al., 1997; Lee et al., 1997; von Ohlen et al., 1997), both of which are direct transcriptional targets of Shh, and thus immediate indicators of Shh pathway activation.

Ptc is normally expressed in the posterior limb mesenchyme, in a domain overlying but more extensive than that of Shh (Goodrich et al., 1996; Marigo et al., 1996b; Fig. 4A). In the Dbf/+ limbs, Ptc was expressed throughout the distal mesenchyme in a continuous posterior-to-anterior domain (Fig 4B). Gli also showed an exclusively posterior expression domain in wild-type embryos (Fig. 4C), but was expressed throughout Dbf/+ limb bud mesenchyme (Fig. 4D). In both Dbf/+ and wild-type limbs a specific downregulation of these two genes was observed in a small area of posterodistal mesenchyme corresponding to the Shh expression domain, as reported by others (Fig. 4 arrows; Marigo et al., 1996b,c; Goodrich et al., 1996). There was no equivalent region of lower transcript abundance in the ectopic anterior domain of either Ptc or Gli (n=40). This supports the Shh gene expression evidence indicating that there is no Shh present outside the posterodistal mesenchyme of Dbf limbs.

Analysis of the timing of Ptc expression revealed that, in younger Dbf/+ limbs, before the limb has significantly widened along the AP axis, Ptc is expressed in a broad anteroposterior band that then extends further anteriorly (compare E10.5 hindlimbs, Fig. 5A,B with E10.5 forelimbs, Fig 4A,B). In E11.0 forelimbs (not shown), the transcript level was lower mid-distally, indicating early downregulation in this site. In even older limb buds (E11.5 forelimbs, Fig. 5D), when Ptc expression has been down-regulated to barely detectable levels in wild-type limb buds, mutant limb buds showed downregulation in all regions except for the anterior domain, where relatively high transcript levels remained.

The results of these experiments indicate that polarising activity is located in a continuous band around the distal margin of the Dbf mutant limbs.

Analysis of polarising activity and gene expression in the polydactylous mutant Doublefoot has shown that ectopic activity is present throughout the distal (progress zone) mesenchyme of the Dbf limb bud; it is not confined to the normal posterior ZPA region, nor is the ectopic activity localised to a discrete anterior region. This finding is reflected in the gene expression patterns. The results suggest that the Dbf mutation affects a gene encoding a key component of the Shh signalling pathway. The mutation represents a gain of function, showing both ectopic and enhanced activity of a normal function.

Detailed genetic mapping of the chromosomal assignment of the Dbf locus (Lyon et al., 1996) excludes the possibility that Dbf represents a mutation in a previously identified component of the vertebrate Shh signalling pathway (see Ruiz i Altaba, 1997 for a recent summary). Similarly, there is no known component of the Drosophila hedgehog pathway that fulfils a role equivalent to that of the Dbf gene (see Perrimon, 1995). Our analysis of the Dbf mouse indicates that the mutation is in a gene encoding a novel element of the Shh signalling pathway. The mutant gene product is able to fulfil the roles of Shh (polarising activity and induction of target genes) in the limb, in the absence of Shh. In our grafting experiments, wild-type ZPA grafts induced mirror-image duplications comparable to those induced by mouse-chick ZPA grafts (Tickle et al., 1976); in contrast, mesenchyme from all regions of the limb, including posterior mesenchyme, induced irregular patterns of extra digits that are very similar to those induced by the ectopic retroviral expression of Shh in the anterior of chick wing buds (Fig. 9 of Riddle et al., 1993). The failure of both Shh of retroviral origin and mesenchyme from Dbf limb buds to induce the mirror-image duplication pattern so consistently obtained with ZPA grafts, suggests that other factors, in addition to Dbf and Shh, are required to elicit the complete developmental sequence of events controlled by the endogenous ZPA.

We suggest that the activity of the normal Dbf gene product is under the control of Shh, confining activity within the limb to the region of the ZPA. In contrast, the mutant Dbf protein does not require Shh for its activity, which is therefore not restricted to the region of Shh ‘activation’. The fact that the ectopic polarising activity is widespread suggests that the Dbf gene product itself is widespread and that its normal function is closely related to the propagation and/or inhibition of signalling through the Shh signalling pathway.

The Dbf phenotype strongly resembles that of transgenic mice overexpressing Shh, which have polydactylous limbs with supernumerary interdigital elements as well as bifurcations of the distal phalanges (Oro et al., 1997). The ability of a gain- of-function mutation at the Dbf locus to phenocopy a transgenic mouse ubiquitously expressing Shh supports the interpretation that the Shh and Dbf genes participate in a common pathway and that Dbf gene expression is widespread in the mutant limb bud mesenchyme. This is significant in that, if the normal Dbf peptide is also widespread in the developing limb, it cannot be under the transcriptional control of the Shh signalling pathway, unlike Ptc and Gli. This would account for the Dbf limbs having a single enlarged ZPA throughout the distal mesenchyme, in contrast to other polydactylous mouse mutants, which possess a second discrete anterior polarising region (Chan et al., 1995; Masuya et al., 1995, 1997). The widespread polarising activity correlates with the phenotype of the Dbf limbs as a fan-shaped array of identical triphalangeal digits (Fig. 1; C. H., M. F. L. and G. M. M.-K. unpublished data).

Grafts of posterior limb bud mesenchyme from both chick (Saunders and Gasseling, 1968) and mouse (Tickle et al., 1976) limb buds show a characteristic complete or partial mirror- image pattern. This pattern was only shown by the wild-type posterior grafts in our experiments; it was not induced by any of the Dbf mutant mesenchyme grafts. Some grafts from Dbf/+ limb buds, including those from the posterior region, showed both extra digits (mainly the more posterior digits, i.e. frequently III and more rarely IV) and an effect on the host digital pattern. These results indicate that the Dbf/+ mesenchyme, including that of the region expressing Shh, has a greater ability to repattern the chick limb bud mesenchyme than that of wild-type embryos and that the effect is independent of Shh. Nevertheless, the ectopic upregulation of Ptc and Gli indicates that Dbf is signalling through the pathway that is normally Shh-dependent.

In terms of the spatial extent of polarising activity, the Dbf mutant shows some similarity to the polydactylous chick mutant talpid, although the ectopic polarising activity of talpid wing buds becomes progressively weaker in a posterior-to- anterior gradient (Francis-West et al., 1995), whereas Dbf limbs do not show this graded pattern of polarising activity. The number of Dbf mutant mesenchyme- induced skeletal elements is also greater than that induced by the ectopic polarising region of Strong’s Luxoid mice, which in similar grafting experiments showed weak activity (extra digit II or other skeletal abnormality) in only 4 out of 39 grafts (Chan et al., 1995). Taking together the evidence from the limb phenotype, the observation of polarising activity in both anterior and mid-distal grafts, and the continuous anteroposterior expression domains of downstream Shh target genes in the limb bud, our results indicate that, in the Dbf mutant, polarising activity is located in a continuous band around the distal limb bud margin.

Current understanding of the Shh signalling pathway is summarised in Fig. 6A. Outside the ZPA, Ptc is present at basal levels; the low level of Ptc here is presumably sufficient to repress the pathway in the absence of Shh. Within the ZPA, the presence of Shh lifts the repressive influence of Ptc on Smo, enabling the signal to be transduced, leading to Gli-mediated transcription of target genes, including Ptc and Gli, and ultimately more downstream targets such as Fgf4, Bmp2 and the 5’ HoxD genes. This model makes Shh, through its binding to Ptc, solely responsible for polarising activity and hence pattern formation in the developing limb. This model does not fit the new information derived from analysis of the Dbf mutant.

We have considered the possibility that the Dbf mutation could represent the inactivation of a negative regulator. Negative regulators are a feature of the Shh/Hh pathway, including Patched (Ingham et al., 1991), cAMP-dependent protein kinase A (PKA) (Jiang and Struhl, 1995; Hammerschmidt et al., 1996;) and Costal2 (Robbins et al., 1997). Our grafting experiments demonstrate the ability of the Dbf gene product to act in a non-cell-autonomous fashion, i.e. to signal to surrounding cells and to divert their developmental fate. This property strongly suggests that the Dbf gene encodes a secreted signalling molecule that acts as a positive regulator of Shh-related signalling, although the possibility that it acts as a negative regulator cannot categorically be excluded.

We therefore propose the scheme summarised in Fig. 6B, in which the Dbf gene product interacts with Smo. In this model, in the ZPA Shh would activate Dbf, possibly via a Shh/Ptc/Smo/Dbf ligand-receptor complex, the activated Dbf protein then activating Smo in and around the ZPA; this would result in signal transduction and Shh-dependent gene expression in an extensive posterior domain reflected by the domain of Ptc gene expression. According to this interpretation, the normal expression domain of Ptc indicates the sphere of influence of activated Dbf protein. A possible route for the Dbf-Smo-activated pathway is through PKA. PKA is epistatic to smo, which itself has similarities to G- protein-coupled receptors, which signal through PKA (van den Heuvel and Ingham, 1996). This model requires Shh for the activation and spatial control of the pathway, but not for signal transduction in all cells, thereby taking into account the non- diffusibility of Shh. The Dbf mutant phenotype would then result from generalised activation of Dbf protein, leading to widespread Dbf-Smo-generated signal transduction. In this molecular context, the upper limit of the number of digits formed would be dependent on the amount of mesenchyme that can be generated during the developmental period in which the genes are active.

We thank Kate Storey for facilities for grafting experiments, Sue Morse for animal care, A. McMahon, D. Duboule, M. Jones, C.-C. Hui and M. Scott for materials, Simon Ward and Marcel van den Heuvel for critical reading and helpful comments on the manuscript. C. H. is in receipt of a studentship from the Anatomical Society of Great Britain and Ireland; J. M. B. is funded by the Wellcome Trust; M. F. L. is supported by the MRC.