Spatial regulation of a zebrafish patched homologue reflects the roles of sonic hedgehog and protein kinase A in neural tube and somite patterning

In Drosophila, cell patterning is controlled by the segment polarity genes, a molecularly heterogeneous group that includes the components of two signal transduction pathways (Klingensmith and Nusse, 1994; Ingham, 1995). One of these pathways is responsible for transducing the activity of the secreted protein Hedgehog, itself encoded by a member of the segment polarity class (Lee et al., 1992; Mohler and Vani, 1992; Tabata et al., 1992; Tabata and Kornberg, 1994). Several genes closely related to hedgehog (hh) have now been described in various vertebrate species, the best characterised of these being sonic hedgehog (shh) (Echelard et al., 1993; Krauss et al., 1993; Riddle et al., 1993; Roelink et al., 1994). The spatiotemporal deployment of shh is highly conserved from fish to mouse: expression is initiated during gastrulation in the embryonic shield or node and persists in the midline mesoderm, the notochord precursor, as the main body axis extends. Subsequently, expression of shh is activated in the floorplate cells of the ventral neural tube that overlie the notochord, where it continues to be expressed throughout somitogenesis (reviewed by Fietz et al., 1994).

Experimental manipulations have identified both the notochord and floorplate as sources of inducing activities that control the patterning of the neural tube and somites (van Straaten et al., 1989; Placzek et al., 1990, 1991; Yamada et al., 1991; Pourquié et al., 1993; Goulding et al., 1994) and several lines of evidence suggest that shh is a major component of such activities. First, in the zebrafish, loss of midline signalling in various mutants is closely correlated with loss of shh expression (Krauss et al., 1993; Ekker et al., 1995; Macdonald et al., 1995). Second, ectopic expression of shh in zebrafish (Krauss et al., 1993; Barth and Wilson, 1995; Macdonald et al., 1995), as well as in mouse and frog (Echelard et al., 1993; Ruiz i Altaba et al., 1995), leads to the inappropriate expression of floorplate and/or ventral brain markers. Such ectopic expression has also been shown to result in the inappropriate activation of the sclerotomal and myotomal markers pax1 and myoD, respectively, in the developing somites of the chick (Johnson et al., 1994) as well as of myoD in paraxial mesoderm of the fish (Weinberg et al., 1996). Finally, and most definitively, recombinant Shh protein is itself capable of inducing floorplate and motor neuron differentiation in neural plate explants (Marti et al., 1995; Roelink et al., 1995) and sclerotomal or myotomal differentiation in explants of presomitic mesoderm (Fan et al., 1995; Munsterberg et al., 1995).

At later stages of development, transcription of shh is activated in the posterior mesenchyme of the developing limb buds, a region corresponding to the signalling centre known as the zone of polarising activity (ZPA). Mis-expression of shh in the anterior of the limb bud results in digit duplications similar to those induced by heterotopic grafts of ZPA material (Riddle et al., 1993; Chang et al., 1994), suggesting that shh mediates the signalling activity of this region of the vertebrate limb.

Although the importance of shh during vertebrate embryogenesis is thus well established, little is known about the way in which the activity of the Shh protein is transduced. In Drosophila, by contrast, genetic analysis has identified a number of putative components of the hh-signalling pathway, such as the cAMP-dependent protein kinase (PKA) (Jiang and Struhl, 1995; Lepage et al., 1995; Li et al., 1995; Pan and Rubin, 1995; Strutt et al., 1995) as well as the products of several segment polarity genes, including patched (ptc) (Hidalgo and Ingham, 1990; Ingham et al., 1991; Ingham and Hidalgo, 1993; Capdevila et al., 1994), a novel multipass membrane spanning protein (Hooper and Scott, 1989; Nakano et al., 1989). Inactivation of either Ptc or PKA in the developing limbs of Drosophila has remarkably similar consequences; in both cases, various hh target genes are inappropriately activated, suggesting that Ptc and PKA normally act to suppress the Hh response pathway. Whilst PKA most likely acts by phosphorylating other components of this pathway, the molecular mechanism of Ptc activity has remained enigmatic. One suggestion, based upon its predicted topology and membrane localisation, is that it functions as a receptor for Hh, binding of the latter antagonising Ptc activity and thus relieving the repression of different target genes (Ingham et al., 1991). Although consistent with the spatial deployment of the two proteins and the interactions between them as deduced by genetic analysis, this model has so far received no direct biochemical support.

Whatever the identity of the Hh receptor, it seems likely to have been highly conserved through evolution since vertebrate hh homologues are capable of activating the response pathway when expressed in transgenic Drosophila (Krauss et al., 1993; Chang et al., 1994; Ingham and Fietz, 1995). Moreover, the discovery that the response of tissue explants to Shh activity can be attenuated by drugs that activate PKA (Fan et al., 1995; Hynes et al., 1995) suggests that at least some of the intracellular components of the pathway may have also been conserved. In this study, we have explored the extent of this conservation further by cloning a zebrafish homologue of ptc and analysing its expression during embryogenesis. In Drosophila, ptc is itself one of the principal targets of hh activity, its transcription being upregulated in hh-responding cells. Using midline mutants and overexpression strategies, we have investigated the relationship between shh activity and transcription of the zebrafish ptc1 gene. Our results support a role for shh and ptc1 in the specification of both neural and mesodermal cell fates in the zebrafish; in addition, we show that both aspects of shh signalling are mediated by PKA activity.

In situ hybridisation was performed as described by Oxtoby and Jowett (1993) with the following modifications: hybridisation were performed at 70°C and post-hybridisation washes were as suggested by Henrique et al. (1995). Stained embryos were dehydrated through an ethanol: butanol series, embedded in Fibrowax and sectioned (610 μm). Double stainings were performed essentially as described (Jowett and Lettice, 1994).

Probes used for in situ hybridisations were synthesized using the following templates: shh, (Krauss et al., 1993); pax-[b] (Krauss et al., 1991); nk2.2, (Barth and Wilson, 1995); MyoD (Weinberg et al., 1996). Unincorporated DIG-UTP was removed by centrifugation through a Nuc50 column (Kodak). Specimens were analysed using a Zeiss Axioplan microscope and photographed with Kodak Ektachrome 64T film. Images were scanned on a Sprintscan 35 slidescanner and processed using Adobe Photoshop software.

Wild-type Danio rerio were bred from a founder population obtained from the Goldfish Bowl, Oxford. The flhⁿ¹ strain was obtained from T. Jowett (Newcastle University, UK). The cyc and ntl strains were obtained from C. Kimmel (University of Oregon, USA). Fish were maintained in a constant recirculating system at 28°C on a 14 hours light/10 hours dark cycle.

Two pairs of primers (rev2, rev4 and genie1, genie2, see below) were used separately to perform PCR starting from random-primed cDNA samples synthesised using either bud stage or somitogenesis stage zebrafish RNA.

Rev2: ggacgaattcTSYTCNRGCCARTGCAT

Rev4: ggacgaattcYTNGANTGYTTYTGGGA

Genie1: ggacgaattccGAYGGNATNATNAAYC

Genie2: ggacgaattcRTAYTGYTCCCARAANA

(large lettering corresponds to sequences conserved between mouse and different insect ptc sequences, small lettering corresponds to an EcoRI site and arbitrary flanking sequence introduced to facilitate subsequent cloning of PCR products). Amplification was performed using 35 cycles of 94°C 30 seconds; 50°C 30 seconds; 72°C 90 seconds. Fragments of the expected size were subcloned into Bluescript KS and sequenced. One fragment was obtained with the Genie1 and Genie2 primers (called Genie) and two with the Rev2 and Rev4 primers (called Rev1 and Rev3). Two primers identical to Rev1 and Rev3 fragments, respectively, a primer complementary to the Genie fragment and the Genie2 primer were used in different pairwise combinations to perform PCR starting from the same original cDNA samples. Two fragments of 1.8 and 2 kb were amplified using the Rev1and Genie-derived primers and the Rev3-derived and Genie2 primers respectively, which after sequencing proved to be homologous to both mouse and Drosophila ptc.

2×10⁶ plaques of a λgt11 library made from 33-hour-old zebrafish embryos (K. Zinn) were screened at low stringency according to standard procedures using the 1.8 and 2 kb fragments as probes. We did not obtain any positive clones with the 2 kb probe using either the 33 hour library or libraries from other stages. 20 positive plaques from the screen using the 1.8 kbp fragment were purified and analysed by PCR using oligonucleotides against the 1.8 kb probe and against the lambda vector. The longest clone, λ105, was sequenced on both strands using either Pharmacia or Applied Biosysytems automated sequencers with primers from the cloning vector and internal primers. Sequence analysis was performed using Geneworks (Intelligenetics) and GCG software packages.

The nucleotide sequence of the zebrafish ptc1 gene described here is deposited in the EMBL database under the accession number: X98883.

The open reading frames of a dominant negative mutant cDNA of the regulatory subunit of Drosophila melanogaster cAMP-dependent protein kinase (Li et al., 1995) and of a constitutively active subunit of the mouse PKA catalytic subunit (Orellana and McKnight, 1992) were subcloned into the CS2 expression plasmids where cDNA transcription is driven by the ubiquitous CMV enhancer/promoter region (Turner and Weintraub, 1994) to create pCS2dnPKA and pCS2cPKA, respectively. The open reading frame of shh was subcloned into the CDM8 expression vector (Seed, 1987).

Injections were performed on 2to 4-cell-stage embryos using backfilled capillaries (Clarks Microelectrical Instruments, Reading) and a pressure-pulsed Narishige microinjector. RNA, synthesized in vitro from linearized plasmids p64Tshh (Krauss et al., 1993), pCS2dn PKA and pCS2cPKA, respectively, was injected at empirically determined concentrations which were of approximately 100 μg/ml. Plasmid DNA, purified using Qiagen columns, was injected at a concentration of approximately 100 μg/ml.

Vertebrate homologues of ptc were first identified in mouse using a polymerase chain reaction (PCR) ptc homologues from distantly related insect species (Goodrich et al., 1996). Two pairs of degenerate oligonucleotide primers based upon sequences conserved between the mouse and insect genes were used to amplify related sequences from zebrafish cDNA (see Materials and Methods). Two fragments of approximately 320 bp and one fragment of around 330 bp, whose sequences proved to be related to both mouse and insect ptc sequences, were amplified. Oligonucleotides based on sequences within these three PCR fragments were then used to amplify two fragments of approximately 1.8 kb and 2.0 kb and each of these fragments was used as a probe to screen a 33-hour-old embryonic zebrafish cDNA library. One full-length cDNA, designated ptc1, was isolated and is described in this paper. Similar clones corresponding to the other ptc homologue, ptc2, have yet to be isolated and this gene will not be considered further here.

The deduced amino acid sequence of ptc1 shows 64% identity to both the mouse and chicken ptc genes and 39% identity to Drosophila ptc. Comparison of the different Ptc sequences reveals that eight cysteine residues are conserved in all four species. The Drosophila protein has seven potential Nglycosylation sites of which three are conserved in mouse and chick but only one is present in zebrafish. Hydropathy analysis predicts that each protein contains 12 potential transmembrane domains (see Fig. 1A). There are two regions in the amino acid sequences that are particularly divergent; in the first of these, the vertebrate sequences have insertions relative to the fly sequence, between the sixth and seventh putative transmembrane domains. This region is also relatively divergent between zebrafish and mouse although it is well conserved between chick and mouse (Marigo et al., 1996). The second significant divergence occurs towards the C-terminus, which in zebrafish is considerably shorter than in either Drosophila or mouse; the region absent in zebrafish shows little conservation between the latter two species (see Fig. 1A). Database searches have also revealed homology to a ptc-related sequence identified by the C. elegans genome project (Wilson et al., 1994); the zebrafish ptc1 gene is as similar to this gene as it is to Drosophila ptc (Fig. 1B).

The expression pattern of ptc1 during zebrafish embryogenesis was analysed by in situ hybridisation and compared to that of shh. Transcription of ptc1 is first apparent at around 70% epiboly in the presumptive mesodermal cell layer in two stripes of cells flanking the axial mesoderm, which at this stage already shows robust expression of shh. As the convergence extension movements of gastrulation proceed, low-level ptc1 expression is apparent throughout the presomitic mesoderm while the two stripes of high-level expression, corresponding to the so-called adaxial cells (Thisse et al., 1993), extend along the entire axis; in addition, high levels of ptc1 transcript are detectable in the neurectoderm overlying the axial mesoderm and in the ventral part of the future brain (Fig. 2A). With the onset of somitogenesis, high-level expression persists in the tissues surrounding the notochord in the trunk as well as in the ventral neuroectoderm of the brain (Fig. 2C). Distinct patches of high-level mesodermal expression become apparent adjacent to the notochord as the somites form (Fig. 2M), a pattern similar to that of α-tropomyosin at the same stage (Thisse et al., 1993). During later stages of somitogenesis when shh expression is lost from the notochord, expression of ptc1 persists throughout the ventral neural tube, except in the floorplate, which at this stage still expresses shh at high levels. Notably, expression in the mesoderm also persists lateral and ventral to the notochord (Fig. 2G,M-P). In the brain, ptc1 expression extends dorsally in the diencephalon, paralleling the dorsal extension of shh expression in the same region and shows a modulation in the hindbrain, which becomes more defined at later stages (Fig. 2G; see also Fig. 5C).

The relationship between the ptc1 and shh expression domains in the developing brain was analysed directly using double-labelling techniques to visualise both transcripts simultaneously. At all stages, high-level expression of ptc1 occurs in a domain delimiting the expression of shh. By 24 hours, expression of shh no longer occupies the most ventral part of the rostral forebrain and ptc1 expression can now be detected ventrally to shh-expressing cells (Fig. 3A). High-level expression of ptc1 is also detected around the lumen of the neural tube.

By 36 hours, additional sites of ptc1 transcription appear in the first branchial arch (not shown) and the posterior mesenchyme of the fin buds (Fig. 2Q), in both cases close to domains of shh expression (Fig. 2R; see also Krauss et al., 1993). By 48 hours, both genes exhibit a complex expression pattern in the brain with a persistence of high-level ptc1 expression adjacent to shh-expressing cells. High-level expression of ptc1, is, however, also observed some distance away from cells expressing shh in a number of places, including most notably a domain of intense expression in the hindbrain (Figs 2S,T, 3B). Expression of both genes is now obvious in the foregut (Fig. 2S-V) as well as in more posterior domains.

The relationship between the patterns of expression of shh and ptc1 described above is highly reminiscent of that between their Drosophila counterparts (Taylor et al., 1993; Tabata and Kornberg, 1994) suggesting that, as in Drosophila, transcription of ptc1 may be induced in response to hh signalling. Since no mutation of shh is currently available in the zebrafish, we took advantage of a number of mutants that eliminate shh expression at different levels along the anteroposterior axis to investigate this possibility.

Mutation of the cyclops (cyc) gene disrupts the specification of the prechordal plate mesoderm (Thisse et al., 1994) and concomitantly the induction of the overlying neurectoderm, resulting in defective midline signalling in the developing brain which gives rise to the cyclopic phenotype (Hatta et al., 1994; Ekker et al., 1995; Macdonald et al., 1995). In addition, cyc mutants lack a floorplate, apparently due to a defect in the response of cells to the inductive signal from the underlying notochord (Hatta et al., 1991). In line with these phenotypic effects, expression of shh is completely absent from both the prechordal plate mesoderm and the overlying neurectoderm that gives rise to the ventral floor of the brain (Krauss et al., 1993) while, posterior to the midbrain, expression is normal in the axial mesoderm but absent from the ventral neural tube.

Expression of ptc1 in cyc embryos mirrors these changes in shh expression precisely: thus by the time the body axis is fully extended, no ptc1 transcripts are detectable anterior to the domain of expression of pax[b] (Fig. 4A), which marks the future midbrain (Krauss et al., 1991), and the forebrain and midbrain remain devoid of ptc1 expression throughout the rest of embryogenesis (Fig. 4B). The only exception is a short dorsoventral stripe in the abnormal neural fold of the diencephalon that appears at around 24 hours of development surrounding a stripe of shh-expressing cells that appears at the same stage (Fig. 4B-D). Posterior to the midbrain/hindbrain boundary, ptc1 is expressed normally in the somites and ventral neural tube (Fig. 4B), except that it persists along the ventral midline of the latter, presumably reflecting the failure of the floorplate to differentiate.

In contrast to cyc, mutation of the floating head (flh) gene has no effect on prechordal plate mesoderm but instead disrupts notochord specification; this leads to a premature loss of shh expression from the axial mesoderm posterior to the midbrain-hindbrain boundary and the resultant disruption of floor plate induction is reflected in patchy expression of shh along the ventral midline of the neural tube (Talbot et al., 1995; see Fig. 5D). In addition, muscle pioneer cells that derive from the adaxial cells adjacent to the notochord (Felsenfeld et al., 1991; Thisse et al., 1993) fail to differentiate (Halpern et al., 1995; Talbot et al., 1995). A similar effect on muscle pioneer differentiation is caused by mutation of the notail (ntl) gene, which also lacks a notochord. Unlike flh, however, ntl mutants do not lack the floorplate (Halpern et al., 1993). These differences in phenotype seem to reflect differences in the effects of each mutation on shh expression, which persists longer in the axial midline of ntl embryos (Krauss et al., 1993) than in that of flh embryos (Talbot et al., 1995).

In flh embryos at 24 hours, expression of ptc1, like that of shh, is completely normal in the brain and anterior neural tube (Fig. 5C,D). More caudally, however, expression is restricted to small clusters of cells distributed sporadically along the ventral neural tube in a manner reflecting the intermittent expression of shh (Fig. 5; compare C to D and E to F). In addition to the effects on neural expression, the broad stripes of ptc1 expression typical of wild-type somites are absent from flh homozygotes, with only scattered mesodermal cells expressing the gene in close association with expressing cells in the neural tube (Fig. 5E). Thus expression of ptc1 in the mesoderm, as in the neural tube, seems to depend upon proximity to a source of shh activity. To investigate this relationship further, we analysed the expression of ptc1 at earlier stages of development in flh embryos. At the bud stage, shh is still expressed along the axial midline in flh embryos, in cells of uncertain origin (Fig. 5B); at the same stage, ptc1 is expressed in a single stripe of cells along the midline instead of in the two stripes typical of wild type (Fig. 5A). A similar change in the early pattern of ptc1 expression is seen in ntl homozygotes (data not shown). However, at 24 hours, in ntl mutants high-level expression persists in distinct stripes in each somite adjacent to the floorplate that expresses shh (Fig. 5G,H).

The analysis of ptc1 expression in midline mutations suggests that expression of shh is necessary for the induction of high-level ptc1 transcription. To test whether shh is sufficient to induce such transcription, we next examined the effects of ectopic shh activity on ptc1 expression. Previous studies have shown that injection of synthetic shh mRNA into the 2to 4-cell-stage embryo results in the ectopic activation of genes such as axial and nk2.2 whose expression domains are normally restricted to the ventral region of the neural tube and brain (Krauss et al., 1993; Barth and Wilson, 1995). In the brain, the distribution of ptc1 transcripts around the shh expression domain closely mirrors the expression domain of nk2.2 (Fig. 6A,B). To determine whether ptc1 expression can similarly be ectopically induced by shh in the nervous system, embryos from the same cohort were injected with shh mRNA, fixed after 27 hours of development and hybridised with probes for ptc1 and nk2.2 As expected, nk2.2 is expressed ectopically in dorsal and lateral regions of the brain as well as in the eye rudiment of such embryos (Fig. 6D). Similarly, high-level expression of ptc1 is also detected ectopically in the diencephalon and midbrain of shh-injected embryos as well as in the eye rudiments (Fig. 6C). Notably, ectopic expression of ptc1 was never detected in the telencephalon. Similar effects on ptc1 expression were observed in small groups of cells following injection of a shh expression plasmid (data not shown) supporting the interpretation that the induction of transcription is a direct consequence of shh activity.

To examine the effects of ectopic shh activity in the mesoderm, embryos from the same injection cohorts were fixed at the onset of somitogenesis and hybridised with probes for myoD and ptc1. Like ptc1, expression of myoD is restricted to the adaxial cells in the presomitic mesoderm (Fig. 6F) and recent studies have suggested that its expression is regulated by shh (Weinberg et al., 1996). In a significant proportion of injected embryos, we found that the expression domains of both ptc1 (64%; n=112) and myoD (71%; n=90), are expanded so that they occupy most of the lateral mesoderm on one or both sides of the midline (Fig. 6G,H). In addition, ptc1 is ectopically expressed throughout much of the neurectoderm at this stage following shh injection (Fig. 6G).

In Drosophila, removal of PKA activity from cells mimics their response to hh signalling, activating the transcription of ptc and other effectors of hh activity, suggesting that PKA normally acts to repress expression of hh target genes (Jiang and Struhl, 1995; Lepage et al., 1995; Li et al., 1995; Pan and Rubin, 1995; Strutt et al., 1995). Since the relationship between hh signalling and ptc expression appears to be conserved from flies to fish, we wondered whether regulation of ptc1 and other shh targets might also depend upon PKA activity. PKA exists as an inactive heterodimer comprising a catalytic and a cAMPbinding regulatory subunit; binding of cAMP to the latter dissociates the heterodimer allowing the catalytic subunit to phosphorylate its target substrates. To modulate PKA activity, we made use of two previously described dominant mutations of PKA subunits. In the first of these, the cAMP-binding sites in the regulatory subunit are mutated such that it cannot be dissociated from the catalytic subunit (Li et al., 1995). Expression of this mutated subunit, which we refer to as dnPKA, thus results in the inactivation of the catalytic subunit and hence in the reduction or loss of PKA activity. In the second case, the catalytic subunit is mutated such that it cannot bind the regulatory subunit, thus rendering it constitutively active (Orellana and McKnight, 1992); we refer to this mutant form as cPKA. Amongst embryos injected at the 2to 4-cell stage with synthetic mRNA encoding the dnPKA mutant and fixed at early somitogenesis stages, more than half (55% n=120) exhibit high-level expression of ptc1 transcription throughout the lateral mesoderm on one or both sides of the midline (Fig. 7C). In addition, a similar proportion of embryos from the same injection cohort exhibit ectopic expression of myoD either unilaterally or bilaterally (Fig. 7D). At later stages (27 hours), injected embryos express both ptc1 (Fig. 7G) and nk2.2 (Fig. 7H) at ectopic locations in the diencephalon and mesencephalon. All of these effects are similar to, though somewhat less robust than, those induced by ectopic shh expression (compare Figs 6 and 7). Similarly, embryos injected with dnPKA mRNA also consistently show ectopic activation of pax[b] in the eye, again similar to, though less extensive than, that induced by ectopic shh expression (Fig. 8A-C).

By contrast, injection of embryos with mRNA encoding the cPKA mutant has the opposite effect on the expression of hh target genes. At 27 hours, expression of ptc1 and nk2.2 is almost totally eliminated from the brain and ventral neural tube of injected embryos, with only a stripe of expression of both genes persisting in the diencephalon (Fig. 9D,E). This stripe corresponds to the dorsal extension of the normal diencephalic expression domain, which also persists in cyc mutant embryos (compare with Fig. 4). Like the latter, most cPKA-injected embryos (60%, n=90) also exhibit varying degrees of cyclopia, involving fusion of the retina alone or both the retina and lens (Fig. 9F), a further indication that midline signalling is attenuated or eliminated by the unregulated activity of PKA.

To explore the relationship between shh signalling and PKA further, we co-injected mRNAs encoding shh and cPKA and analysed the eyes of resultant embryos at 27 hours. Injection of shh RNA alone leads to a high proportion of embryos with rudimentary eyes (77% n=183) as previously described (Krauss et al., 1993; Ekker et al., 1995; MacDonald et al., 1995). Amongst embryos co-injected with both shh and cPKA RNAs, only 22% (n=70) exhibited such reduction of the eyes. Moreover, the frequency of cyclopia was much reduced (11%) relative to that induced by injecting cPKA alone (60%), suggesting that the activities of shh and PKA are mutually antagonistic.

Signalling by Hh family proteins is used repeatedly and in various contexts during animal development. In Drosophila, signalling by hh in the embryonic segments as well as in the imaginal discs, retina and ovary is intimately associated with the activity of the segment polarity gene ptc (Ingham and Hidalgo, 1993; Ma et al., 1993; Capdevila et al., 1994; Heberlein et al., 1995; Tabata et al., 1995; Forbes et al., 1996a,b). In each instance, the two genes are expressed in complementary domains (Taylor et al., 1993; Tabata and Kornberg, 1994; Capdevila et al., 1994; Forbes et al., 1996b), the activity of ptc suppressing the transcription of hh target genes wherever it is expressed. Thus when ptc activity is eliminated, cells behave as though they have received the hh signal (Ingham and Hidalgo, 1993; Capdevila et al., 1994; Lepage et al., 1995; Tabata et al., 1995), even in the total absence of hh function (Ingham and Hidalgo, 1993). These properties have led to the notion that hh acts by antagonising the activity of ptc (Ingham et al., 1991; Ingham and Hidalgo, 1993; Basler and Struhl, 1994; Tabata and Kornberg, 1994) and, because ptc encodes a transmembrane protein, it has been suggested that it may act as a receptor for the secreted Hh protein (Ingham et al., 1991). Since the signalling activity of Hh family proteins has been highly conserved during evolution (Krauss et al., 1993; Chang et al., 1994; Ingham and Fietz, 1995), we anticipated that proteins involved in their reception at the cell surface would similarly be well conserved. The isolation of ptc homologues from zebrafish reported here, as well as from mouse (Goodrich et al., 1996) and chicken (Marigo et al., 1996), is in line with this expectation. Whether or not there is a direct interaction between these proteins and the respective Hh family proteins, however, remains a moot point. Although the membrane association of Drosophila Ptc (Taylor et al., 1993) is consistent with its hypothesised role as a receptor, the predicted topology of both the invertebrate and vertebrate Ptc proteins is more typical of that of ion channels or transporter proteins (Hooper and Scott, 1989; Nakano et al., 1989; Goodrich et al., 1996).

A defining feature of Drosophila ptc is its transcriptional up regulation in response to hh signalling. The highest levels of ptc expression are typically found in cells immediately adjacent to those expressing hh (Hooper and Scott, 1989; Nakano et al., 1989; Taylor et al., 1993; Tabata and Kornberg, 1994; Capdevila et al., 1994; Forbes et al., 1996b) and this pattern of expression depends upon hh activity (Hidalgo and Ingham, 1990; Capdevila et al., 1994) whilst ectopic expression of hh induces inappropriate levels of ptc transcription (Ingham, 1993; Basler and Struhl, 1994; Capdevila and Guerrero, 1994; Diaz-Benjumea and Cohen, 1994; Tabata and Kornberg, 1994; Ingham and Fietz, 1995; Forbes et al., 1996b). High-level expression of ptc is thus an indication of a response to hh signalling that is independent of cell or tissue type. Although the functional significance of this regulatory relationship is currently unclear, one possibility is that Ptc activity acts as a sink for the Hh protein; in this case, its upregulation in response to Hh would effectively limit the range of the Hh signal. Alternatively, the increase in ptc expression could act as a feedback mechanism that attenuates the response of a cell to the hh signal.

Our analysis of ptc1 expression in the zebrafish embryo does not allow us to distinguish between these or other possibilities; however, it does provide compelling evidence that this regulatory interaction has itself been conserved during evolution and is therefore likely to be fundamental to the mechanism of hh signalling. Thus as in flies, the highest levels of ptc1 transcription are invariably associated with shh-expressing cells; for instance, in the pectoral fin buds and the foregut high-level ptc1 expression is found immediately adjacent to cells expressing shh, while in the CNS, the levels of ptc1 transcript are at their highest in the ventral region of the neural tube adjacent to the shh-expressing cells of the axial mesoderm and floorplate. A similar relationship exists between the expression domains of ptc homologues and shh both in the chick (Marigo et al., 1996) and mouse (Goodrich et al., 1996). However, in each of these organisms, additional sites of elevated ptc expression are observed in tissues associated with the expression of two other hh family members, Indian (Ihh) and Desert (Dhh) hedgehog (Goodrich et al., 1996; Marigo et al., 1996; Bitgood et al., 1996). In zebrafish, the expression of tiggy-winkle hedgehog (twhh), the only other hh family gene characterised to date in this organism, although initiated slightly earlier, is entirely included within the shh expression domain (Ekker et al., 1995). Thus, in contrast to amniotes, there is no expression of ptc1 that can be specifically associated with the activity of a hh family gene other than shh. On the contrary, except in the fin buds, where twhh is not expressed, it is possible that at least some aspects of the transcriptional regulation of ptc1 may in part be mediated by twhh. In midline mutants that lack shh expression at specific positions along the body axis, we observe changes in the pattern of ptc1 transcription consistent with a regulatory interaction between the two genes. In cyc mutants, high-level ptc1 expression is completely absent from the brain, consistent with the lack of shh expression in the ventral floor of the brain typical of this mutant (Krauss et al., 1993; Ekker et al., 1995; MacDonald et al., 1995). Since cyc embryos also lack twhh expression along the entire midline (Ekker et al., 1995), we cannot rule out the possibility that this gene may also be necessary (and indeed sufficient) for ptc1 transcription in the brain. However, the essentially normal expression of ptc1 along the ventral neural tube of cyc mutants, suggests that in this part of the body at least, shh activity is sufficient for the regulation of ptc1 transcription. In flh mutants by contrast, shh expression disappears from the axial mesoderm at a relatively early stage, many of the cells along the ventral neural tube failing to differentiate into floorplate presumably as a consequence (Talbot et al., 1995). In line with this widespread loss of shh expression, transcription of ptc1 is also severely reduced, being maintained only around the few scattered islands of cells that still express shh. Taken with our finding that transcription of ptc1 can be induced ectopically in the neural tube by mis-expression of shh, these data strongly suggest that shh is both sufficient and necessary for transcriptional activation of ptc1 transcription in the neural tube.

Whereas low level transcription of ptc1 is detectable throughout the presomitic mesoderm, its expression is significantly elevated in the adaxial cells that flank the midline mesoderm. Several lines of evidence suggest that this mesodermal expression of ptc1 depends upon shh rather than twhh activity. First, at this stage, expression of twhh is restricted to the neuroectoderm (Ekker et al., 1995) while shh is expressed at high levels throughout the axial mesoderm (Krauss et al., 1993). Second, in cyc embryos, which lack twhh expression along the entire midline (Ekker et al., 1995) but have normal levels of shh expression in the axial mesoderm, expression of ptc1 in the somitic mesoderm is normal. Conversely, the absence of shh expression along the axial midline caused by the flh mutation, is accompanied by a significant loss of ptc1 expression in the mesoderm. When considered with our finding that mis-expression of shh induces high-level transcription of ptc1 throughout the paraxial mesoderm, these data strongly suggest that the mesodermal expression of ptc1 is regulated by shh.

Interactions between the notochord and paraxial mesoderm have previously been well documented in higher vertebrates (Dietrich et al., 1993; Koseki et al., 1993; Pourquie et al., 1993; Goulding et al., 1994) and indeed the activity of Shh has been directly implicated in these interactions (Fan and TessierLavigne, 1994; Johnson et al., 1994; Fan et al., 1995). However, whereas in zebrafish, the induction of ptc1 expression in the adaxial cells appears to involve a very shortrange effect of shh, in the chick, Shh protein acts over relatively long distances (Fan and Tessier-Lavigne, 1994; Fan et al., 1995). Moreover, adaxial cells give rise to a myotomal lineage (Felsenfeld et al., 1991; Thisse et al., 1993) and indeed express the muscle-specific transcription factor myoD, which can itself be induced by shh activity (this report; Weinberg et al., 1996). By contrast, in the mouse and chick, Shh induces expression of the sclerotomal marker Pax1 (Fan and TessierLavigne, 1994; Johnson et al., 1994; Fan et al., 1995); in line with this, notochord grafts have been shown to induce differentiation of sclerotome at the expense of myotome, actually repressing MyoD expression (Goulding et al., 1994). Thus, it appears that the outcome of shh signalling on somite patterning differs between lower and higher vertebrates. The different organisation of the somites in amniotes and teleosts may thus be accounted for, at least in part, by a modification of the inductive signals that control their specification. Interestingly recent studies have revealed that, in chick, MyoD expression can be induced by Shh in presomitic mesoderm when presented in combination with signals from the dorsal neural tube (Munsterberg et al., 1995), an activity that may reflect its original role in the evolution of the vertebrate body plan.

Intriguingly, as in Drosophila, we have found that the effects of ectopic shh activity on ptc1 expression can be mimicked by reducing or eliminating the activity of PKA. Although PKA has previously been implicated in signalling by Shh protein in the brain and somites of rats and chicks, respectively (Fan et al., 1995; Hynes et al., 1995), our finding that transcription of a ptc homologue in both the neuroectoderm and paraxial mesoderm is repressed by PKA provides the most direct indication to date of the extent to which the elements of the hh signalling pathway have been conserved. Similar effects are seen on the transcription of other targets of shh activity, both in the mesoderm, where myoD transcription is induced, and in the neural tube, where nk2.2 is inappropriately expressed. In addition, we consistently observe the ectopic activation of pax[b] in the developing eye, an effect that presages eye abnormalities similar to, though less extreme than, those induced by ectopic shh expression (unpublished observations). Analogous effects of a dnPKA on zebrafish eye development have also recently been described by Hammerschmidt et al. (1996).

The implication of PKA in the regulation of ptc1 is further supported by our finding that constitutive activity of PKA suppresses ptc1 transcription throughout the brain and neural tube. This effect is accompanied by a similar suppression of nk2.2 expression and by cyclopia, a condition clearly associated with a loss of midline signalling (Krauss et al., 1993; Hatta et al., 1994; Ekker et al., 1995; MacDonald et al., 1995). Together, all of these effects strongly support the notion that the shh signal acts by antagonising the repressive activity of PKA on various target genes, just as in Drosophila, hh counteracts the repressive effect of PKA on wg, dpp and ptc transcription. In Drosophila, it has been argued that PKA acts in parallel to the hh signalling cascade (Jiang and Struhl, 1995; Kalderon, 1995). In this regard, it is interesting that co-expression of shh and the cPKA mutant results in a mutual suppression of the two contrasting eye phenotypes induced by either activity alone. This result implies a balance between two opposing signals rather than a strictly linear relationship between the two, supporting the conclusions from the Drosophila studies. Further analyses will be required to elucidate the precise relationship between these activities.

We are especially grateful to T. Simmonds for running the fish facility, G. Clark and A. Davies for help with the automated sequencing, I. Goldsmith for oligonucleotide synthesis and our colleagues on the 4th floor at the ICRF for interest and advice. We thank T. Jowett and C. Kimmel for providing mutant fish strains, A. Barth, D. Henrique, T. Jowett, D. Kalderon, S. McKnight, D. Simmons and E. Weinberg for various clones and K. Zinn for providing the cDNA library. L. V. G. is an HHMI predoctoral fellow, R. L. J. is a fellow of the Damon Runyon-Walter Winchell Foundation Cancer Research Fund (DRG 1218) and M. P. S. is supported by the HHMI. The experimental part of this study was funded by the Imperial Cancer Research Fund and by grants from the Human Frontiers Science Programme and the European Community Human Capital and Mobility Programme to P. W. I.