The dre, uki, and lep mutant class has shown that multiple regulators secure the activity of the Hh signaling pathway. This was illustrated by the fact that the concurrent loss of function of the negative regulators Sufu, Hip and Ptc2 did not result in the typical Hh overexpression phenotype, mainly concerning the chevron shape of the somites 14 . Presumably, the slight increase in Hh signal is enough to increase Ptc1 expression via a sensitive auto-regulatory loop, thereby maintaining the pathway at a certain level of activity, which does not exceed the threshold for Hh activation that will lead to patterning defects.

To test this hypothesis we have generated a target-selected knockout of the Ptc1 gene. An ENU-induced mutation library of approximately 12,000 F1 fish was screened for mutations in this gene using the TILLING method 36 . This resulted in the identification of a splice donor mutation, changing the 5' consensus sequence of the intron after exon 10, from GT, to AT (Fig. 1A, referred to as ptc1 ^hu1602 ). The removal of this splice site results in the insertion of the 81 bp intron into the transcript (Fig. 1B). RT-PCR (Fig. 1C) and sequencing experiments on a fragment containing exons 10 to 13 confirmed that this complete intron is inserted in frame into the ORF, resulting in an expansion of the second extracellular loop of the Ptc1 protein by 27 amino acids (Fig. 1D).

Zebrafish homozygous for ptc1 ^hu1602 display a subtle somite phenotype at 32 hours post fertilization (hpf), where the angle of the chevron shaped form of the somites becomes more obtuse (Fig. 2A,B). The average angle of the somite is 84° in a wild type situation (n = 6, 4 somites measured), compared to 99° for the ptc1 ^hu1602 mutant (n = 6, 4 somites measured). However, the typical flattened somite phenotype, as observed after Shh overexpression in zebrafish, was not observed, probably due to redundancy with Ptc2. Additionally, at 72 hpf a partially penetrant eye phenotype is observed, where the retinal pigmented epithelium (RPE) extends into the diencephalon (Fig. 2C,D). This phenotype has already been described for the blowout (blw ^tc294z ) mutant 37 , suggesting that blowout may also be a mutation in ptc1. Complementation analysis indicated that ptc1 ^hu1602 and blw ^tc294z are allelic (data not shown). Subsequent sequence analysis of all exons of ptc1 identified a premature stop codon (W1039X) in the blw ^tc294z allele, located after the eighth transmembrane domain (Fig. 1D). Comparing ptc1 ^hu1602 with blw ^tc294z revealed that ptc1 ^hu1602 has a higher penetrance of the eye phenotype (21.4% in total, 3.4% single eye, 18% both eyes affected, n = 261) than blw ^tc294z (5% in total, 4.4% one eye, 0.6% both eyes, n = 521). Additionally the RPE in the ptc1 ^hu1602 extends more severely into the diencephalon of the embryo compared to blw ^tc294z (Fig. 2E), suggesting that the ptc1 ^hu1602 allele is stronger. As transcription of ptc1 is known to be upregulated by increased Hh signaling, we decided to test whether our mutations resulted in an increase in ptc1 transcripts. In situ hybridization (ISH) experiments showed that both alleles resulted in an increase in ptc1 transcripts. In addition, the blw ^tc294z mutant (Fig. 2F–H) shows milder upregulation than the ptc1 ^hu1602 mutant (Fig. 2I–K) at 18 hpf, again indicating that the latter allele is stronger. An alternative explanation for this observation might be that the splice donor mutation in ptc1 ^hu1602 increases the stability of the ptc1 transcript, subsequently resulting in a higher expression level after ISH. Therefore we analyzed the expression levels of gli1, another downstream transcriptional target of Hh signaling 6 to confirm the overactivation of the pathway in the ptc1 ^hu1602 mutant. The expression levels of gli1 at 19 hpf distinguishes wild types from heterozygotes and mutants (Fig 2L–N). This confirms that the pathway is indeed activated, and that the increase in ptc1 expression is not a consequence of altered stability of the transcript, due to the induced mutation, but a representation of the activity of Hh signaling in this mutant.

The ptc1 ^hu1602 allele results in an in frame insertion of 27 amino acids into the protein sequence, and therefore we investigated whether this is a full null or a partial loss of function allele. To test this hypothesis we injected a MO against ptc1 25 into a ptc1 ^hu1602 mutant background, which did not enhance the phenotype. Additionally, we checked whether the ptc1 ^hu1602 allele has a dominant negative effect, by injecting wild type embryos with ptc1 ^hu1602 mRNA. No phenotypes comparable with the ptc1 ^hu1602 mutant were observed, excluding a dominant negative effect. Together the data suggest that ptc1 ^hu1602 is at least a strong loss of function allele. Since the ptc1 ^hu1602 allele is stronger than blw ^tc294z , we used this allele for subsequent analysis and therefore we will refer to ptc1 ^hu1602 when ptc1 is mentioned.

By the identification of a zebrafish ptc1 mutant we are now able to investigate whether the inhibition of the pathway is dependent on the presence of both Ptc homologues. Based on the current model, the absence of both Ptc proteins will constitutively activate Smo, normally inhibited by Ptc, which in turn activates the entire downstream pathway. Double mutants for lep(ptc2) ^tj222 and ptc1 ^hu1602 were generated (hereafter referred to as ptc1;ptc2 mutants) and these were expected to show the consequences of increased Hh signaling. From previous studies it has been shown that overactivation of the Hh pathway, after injection of Shh or dnPKA 16 25 , results in a flattened somite phenotype. Phenotypically, ptc1;ptc2 mutants clearly exhibit a flattened somite phenotype at 18 hpf (Fig. 3A,B) confirming an activation of the pathway, by losing both Ptc homologues. At 24 hpf, the primitive eye field is present but the lens is missing (Fig. 3C,D), where at 48 hpf a complete absence of the eyes and the nose could be observed (Fig. 3E,F). To confirm an activation of the pathway we analyzed the expression level of nkx2.2a 15 and gli1 6 . The expression of nkx2.2a is only induced when Hh activity is high. A clear increase in nkx2.2a (Fig. 3G–J) ptc1 and gli1 (See Additional file 1) expression could be observed between the different genotypes, confirming that the activity of the Hh signaling pathway is severely upregulated in ptc1;ptc2 mutants. Administration of 10 mM of cyclopamine to ptc1;ptc2 partially rescued the phenotype and resulted in a decrease in ptc1 expression levels (data not shown). This confirms that the phenotype seen in ptc1;ptc2 mutants is a result of increased Hh pathway activation.

The eye phenotype of a single ptc1 ^hu1602 mutant, a protrusion of the pigmented epithelium towards the medial region of the brain, could be a result of a disturbed balance between optic cup versus optic stalk differentiation. This process is reported to be under control of Hh signaling 17 38 . Possibly, the absence of the eyes in the ptc1;ptc2 double mutants could be a further disturbance of cup versus stalk differentiation. To test this hypothesis, we analyzed the expression pattern of pax2 and pax6, identifying the presumptive optic stalk and cup respectively. Indeed, ptc1;ptc2 mutants expand pax2 expression and lose pax6 expression in the presumptive optic cup region (See Additional file 1), confirming a differentiation defect during eye development. Finally, ptc1;ptc2 mutants show an expansion of the lateral floor plate, indicated by foxa ISH, whereas shh expression in the medial floor plate was unaffected (See Additional file 1). Again, this is in line with previous reports 39 , and confirms the overactivation of the Hh pathway in the ptc1;ptc2 mutants. These observations justify the assumption that the Hh pathway can be constitutively activated in a genetic manner, by losing activity of both Ptc homologues.

In zebrafish, somites give rise to several muscle types. The earliest event concerns the formation of the adaxial cells, which will later go on to form the slow muscles, whereas the rest of the somite will mainly produce fast muscle. Adaxial cells that are located at the apex of the somite will become muscle pioneers, cells strongly positive for Engrailed (eng1) 40 . Hh signaling from the midline induces both adaxial cells and muscle pioneers. As predicted by Hh injection experiments we find that both myod (Fig. 4A,B) and eng1 (Fig. 4C,D) 16 25 41 are strongly induced in ptc1;ptc2 mutants, indicating that these mutants mainly form slow muscle type fibers. Consistent with this somites appear to be elongated in the D/V direction and more cells appear to stack in the medial somite (See additional file 2). To gain further proof for a conversion of the myotome towards the slow muscle fate, we analyzed the expression of prdm1, known to be a key regulator for slow muscle cell differentiation 26 . In a wild type situation, prdm1 is expressed in the most posterior somites and spinal cord neurons (Fig. 4E,F). The expression of prdm1 is regulated by Hh signaling 27 , which is confirmed by the fact that the expression level increases in the ptc1 mutant, and to an even higher level in the ptc1;ptc2 mutants. This strongly suggests that the differentiation between fast and slow muscle cell differentiation in the myotome is shifted towards slow muscle type fiber formation, possibly by upregulating prdm1 expression. Upregulation of myoD correlates with upregulation of skeletal muscle alpha actin (acta1) gene expression as a marker for muscle differentiation, and downregulation of Pax3/7 expression (See Additional files 1 and 2) a marker for the undifferentiated dermomyotomal cells 42 .

Additionally, we also investigated whether the anteroposterior patterning of the somites was affected. Although myod is expressed in the posterior part of recently formed somites, the expansion of the adaxial domain of myod expression in ptc1;ptc2 mutants means that this marker cannot be used as A/P marker. Therefore, we analyzed uncx4.1, a marker whose expression is normally restricted to the posterior part of the somite, and surprisingly, we found that it was lost (Fig. 4G,H). In addition, expression of fgf8, which demarcates the anterior somite, was also lost (Fig. 4I,J), The loss of markers for both the anterior and posterior halves suggests that the somites have lost their polarity. Polarity defects are also corroborated by myf5 expression, which has a more complex expression pattern during somite maturation (Fig. 4K–N).

Since A/P polarity of the somites is likely to be determined during- and in fact may be necessary for, proper somite formation (epithelialization) we analyzed the expression of deltaC (dlc) (Fig. 4O,P), her1 (data not shown), and cellular morphology (See Additional file 2). In wild types, the expression of these genes shows oscillation in the presomitic mesoderm and is required for proper segmentation. These markers showed a normal expression pattern in presomitic mesoderm. Furthermore, epithelialization does occur, albeit less regularly. Furthermore dlc expression showed that although it is normal in presomitic mesoderm and during somite formation, it is not maintained in the posterior part of older, more anterior somites (Fig. 4O,P).

In an attempt to identify a link between increased Hh signaling and the loss of polarity markers, we analyzed a possible role for prdm1 in this process. Prdm1 can act as a transcriptional repressor, whose expression is controlled by Hh signaling 27 . Morpholino knockdown of prdm1 did not rescue the expression of fgf8 or uncx4.1 (data not shown), indicating that additional negative regulators are involved in the loss of polarity markers in the ptc1;ptc2 mutant somites. All these data suggest that the somites of ptc1;ptc2 mutants become apolar after somite boundaries have been established. To our knowledge, this is the first instance in which the process of somite formation has been genetically uncoupled from A/P patterning of the formed somite.

The Hh signaling pathway is known to regulate the formation and patterning of the vertebrate limb. However, Hh signaling is supposed to perform a rather late function in maintenance and outgrowth of the fin bud. Unexpectedly, the ptc1;ptc2 fin buds are completely lost. To investigate at which level Hh is acting on the induction of the fins, we investigated the expression pattern of several genes involved in fin development, like fgf24 and prdm1. Both genes were absent in the presumptive fin field (Fig. 5A–D). Additionally, the expression level of the transcription factor tbx5, one of the earliest markers for fin bud initiation was hardly detectable at 20 hpf, and a slight scattered expression was observed at 40 hpf (Fig. 5E–H). Initiation of tbx5 expression, labeling presumptive heart and pectoral fin primordia appeared normal, however (See Additional file 1). Hand2 (or dHAND) is another important factor in limb bud formation in zebrafish 43 . The expression of tbx5 and hand2 is mutually dependent since hand2 expression is reduced in tbx5 mutants and tbx5 expression is reduced in hand2 mutants 43 44 . We therefore tested whether we could see defects in hand2 expression. ISH at 20 hpf, shows the expected expression pattern of hand2 in posterior lateral mesoderm with a subtle difference in the future fin bud region where hand2 is absent (Fig. 5I–L). Published analysis of early stages of fin bud formation suggests that tbx5 is only mildly reduced in a hand2 mutant and fin buds are still formed 43 . The strong reduction of tbx5 in ptc1;ptc2 mutants suggests that although hand2 is reduced it is probably not the only causal factor. Since both hand2 and tbx5 are reduced we sought to identify defects in signals that might act upstream of these two genes. One factor that has been implicated in tbx5 expression is wnt2b 32 , however no difference could be observed in wnt2b expression at 18 hpf (See Additional file 1). Finally, fin bud induction is known to be under direct control of RA signaling, as shown by the neckless (nls) mutant encoding aldh1a2 45 . Furthermore, mouse Aldh1a2 -/- embryos neither express tbx5 nor hand2 46 . Manipulation of RA signaling in the nls mutant, has shown that it is required before 10 somite stage (ss) 45 for development of the pectoral fin bud, as indicated by dlx2 expression. Additionally, RA signaling has also been shown to be necessary during early segmentation, when it is produced in the anterior somites, to induce the fin field 29 , where it rescues tbx5 expression in the neckless mutant. In the ptc1;ptc2 mutant aldh1a2 expression, required for the production of RA, was not significantly altered at the 60% epiboly stage or at 18 somites in ptc1;ptc2 double mutants (data not shown). In an attempt to rescue the fin phenotype, we treated double mutant embryos with RA. Concentrations ranging from 10^-7 to 10^-6 M did not rescue the formation of the pectoral fins (n = 60 per concentration, including 3 to 4 ptc1;ptc2 mutants). However, the expected morphological defects in axis formation were obtained, showing that RA can signal in the ptc1;ptc2 mutants. We conclude that the Hh signaling pathway represses induction of the fin bud, and acts independent of RA signaling.

Since the expression of wnt2b is not affected in the ptc1;ptc2 mutants and exogenous RA treatment is unable to rescue the fin phenotype, we tried to determine in which time window Hh is preventing fin field induction. Therefore, we treated 90 embryos, obtained from a ptc1 ^+/-;ptc2 ^+/- incross, with 10 μM cyclopamine between 50% epiboly and different developmental stages. This concentration of cyclopamine does not severely affect wild type siblings. By using this rather low concentration we do not affect the later role for Hh in outgrowing of the fin bud, but specifically focus on its supposed early role in fin bud induction. As a readout for a possible rescue, we examined tbx5 expression at 40 hpf, showing a clearly distinct finbud in the wild types (Fig. 6A–D,E–H) and an absence of tbx5 in the ptc1;ptc2 mutant (Fig. 6I,M). Transient inhibition of Hh by cyclopamine between 50% and 100% epiboly (Fig. 6B,F,J,N) did not re-establish a localized tbx5 expression in the presumptive fin field. However, low tbx5 expression could be detected, in contrast to the untreated ptc1;ptc2 mutant (Fig. 6I,M). Localized expression of tbx5 was detected when Hh was inhibited between 50% epiboly and 5 ss (Fig. 6K,O). This can be enhanced when ptc1;ptc2 mutants are treated with cyclopamine between 50% and the 10 ss (Fig. 6L,P). Thus inhibiting Hh signaling during early segmentation stages clearly rescued tbx5 expression in the fin field. However, it is not clear how quickly the cyclopamine is washed out. Furthermore no obvious outgrowth of the pectoral fin bud was observed in these embryos at 40 hpf, which indicates that high levels of Hh signaling can still inhibit pectoral fin outgrowth after the 10 ss in the ptc1;ptc2 mutant. To further investigate the time window in which Hh is negatively regulating fin induction, we treated embryos with 10 μM cyclopamine from 12 ss, 18 ss, 24 ss and 24 hpf towards 40 hpf (Fig. 6Q–T). Restricted expression of tbx5 can only be rescued when Hh activity is inhibited starting from the 12 (data not shown) or 18 somite stage (Fig. 6R). Adding cyclopamine at 24 ss or 24 hpf results in a scattered tbx5 expression (Fig. 6S,T) which is nearly similar to the untreated ptc1;ptc2 mutants (Fig. 6Q). These experiments suggest that Hh signaling is inhibiting recruitment of tbx5 positive cells during the segmentation period (from 100% epiboly until 24 somite stage). During this time, the closest source of potential Hh signaling is at the midline. The normal fin primordium is located approximately 130-150μm, this is somewhat further than the estimated active range of midline hedgehog signaling (in chick neural tube approximately 100μm 47 ) and would suggest such signals would normally not occur in the fin region.

Our previous research has shown that the Hh signaling pathway is strictly controlled in zebrafish to prevent overactivation, since the concurrent loss of Sufu, Hip and Ptc2 does not result in the severe morphological defects 14 , that have been described for Hh overactivation. We previously hypothesized that a combined loss of both Patched proteins should result in a constitutive activation of the Hh pathway. Contrary to Ptc2, there was no zebrafish Ptc1 mutant available to confirm this idea. Here we report the identification of a zebrafish Ptc1 mutant via a reverse genetic approach. The identified splice donor mutation, results in the insertion of intron 10 consisting of 81 base pairs into the transcript, enlarging the second extracellular loop of the protein by 27 amino acids and the inserted amino acid sequence has no homology to known proteins or domains. Previous reports have described that the first and fourth extracellular loops are required to bind Hh 47 . There are two likely explanations for the observed effect. Firstly, the modified protein is not inserted correctly in the membrane, and can therefore not exert its normal function. An alternative possibility is that the extended loop mimics the effect that Hh binding normally has, or that it somehow affects the sterol-sensing domain. This scenario might result in a correctly localized protein, which would however, lose its repressive capacity on Smoothened. Although the consequence of this insertion on a protein level is hard to predict, the Hh pathway is clearly constitutively activated, since ptc1, nkx2.2a and gli1 expression are severely increased in ptc1 ^hu1602 mutants.

Zebrafish mutants homozygous for ptc1 ^hu1602 , show a somite defect, as expected from Hh overexpression experiments, where the angle of the somite becomes more obtuse. However, the strength of the phenotype is surprisingly mild, since mouse data suggest that Ptc1 is the main inhibitor of Hh signaling. In mice, ptc1 knockouts are embryonic lethal 48 and ptc2 mutants are viable without obvious defects 49 , indicating that Ptc1 is the major regulator in mammals. This could indicate a functional shift between zebrafish and the mouse homologues of Ptc. This shift may also be reflected by the fact that the comparisons between the mouse and fish genes does not result in a clear one-to-one relationship between the Ptc1 and Ptc2 homologs 50 .

The ptc1 mutant enabled us to confirm our hypothesis that in zebrafish Ptc1 is the key negative regulator of the Hh pathway, conferring a stronger phenotype than the triple mutant made with Sufu, Hip and Ptc2. The ptc1;ptc2 double mutants showed the expected flat somite phenotype, which is the classical effect of Hh overexpression. Additionally, mutants homozygous for both the ptc1 ^hu1602 and uki(hip) ^hu418B mutations also resulted in flattened somites confirming the idea that ptc1 expression is induced above a certain activity, thereby preventing the pathway from further activation (data not shown). This is a likely scenario since subtle increases in ptc1 expression could be observed in ptc2 and hip mutants 14 , confirming the idea that Ptc1 has a pivotal role in controlling the activity of the pathway. The constitutive activation of the pathway in ptc1;ptc2 double mutants results in similar phenotypes as described after injection of shh or dnPKA, concerning the optic stalk versus cup differentiation in the developing eye, patterning of the floor plate and differentiation of fast and slow muscle cell types, thereby validating these mutants as a model system to study the consequences of aberrantly activated Hh signaling.

Somite A/P information has previously been shown to be necessary for proper segmentation 13 51 . In the zebrafish mutants fused somites (fss) 13 , anterior information is lost and the complete somite is posteriorized 51 , resulting in the lack of somite boundary formation. When anterior and posterior polarity genes are expressed throughout the complete somite, segmentation is disturbed. This has been shown for the beamter (bea), deadly seven (des) and after eight (aei) mutants 13 , which do only form the first four, seven and eight somites respectively, as a consequence of mutated components of the Notch signaling pathway 52 53 . The A/P axis is thought to be established in the presomitic mesoderm, and is thought to be required for morphological boundary formation 51 .

The ptc1;ptc2 mutants show that A/P patterning in the presomitic mesoderm/nascent somite can be separated from maintenance of A/P patterning in existing somites. We analyzed expression of her1 and dlc in the presomitic mesoderm and found no defects, indicating that the oscillations in their expression occur normally. This is confirmed by a normal morphological progression of somitogenesis in double mutants. From the known role of Hh signaling in paraxial mesoderm, and myod labeling experiments, we expect that the paraxial mesoderm is completely induced to become adaxial. We conclude that adaxial cells are capable of executing the oscillations required for segmentation and are capable of forming somite boundaries. However, analysis of dlc has shown that the somites are not able to maintain the A/P pattern in the somites. Surprisingly, such somites do not show a default A or P identity, rather they appear to be apolar (loss of fgf8/anterior; uncx4.1/posterior).

Since the Hh signaling pathway leads to activation of transcription via the Gli genes, repression of the former genes has to occur via intermediate repressors. prdm1 was tested as an good candidate, but morpholino-knockdown of this gene could not rescue fgf8 or uncx4.1 expression, indicating that additional repressors are induced. Since we have tested only a limited set of markers we cannot exclude that the somites have some currently undetectable A/P polarity.

Most markers that show A/P differences are not expressed in the adaxial cells or are expressed in all adaxial cells without displaying A/P differences. Thus it might be assumed that polarity is lost because adaxial cells do not participate in this patterning, and Hh signaling from the midline induces adaxial fates. Loss of hh signaling has not been shown to affect somite A/P patterning, thus such a interference model appears attractive. Even though the effect on myf5 expression suggests that A/P is directly by the hh signal, our data currently cannot exclude an indirect model where premature muscle differentiation prevents expression of such markers.

Somite A/P pattern has been reported to guide patterning of the motoneurons of the neural tube 54 . In mutants ptc1;ptc2 neural tube patterning itself is affected by ectopic Hh signaling, thus in order to study this tissue specific knockouts are required. Nevertheless, we find defects that are at least consistent with defects in A/P patterning (See Additional file 2).

One of the most surprising phenotypes of the ptc1;ptc2 double mutant is the absence of the pectoral fins. Thus far, only a positive role for Shh signaling in limb development has been described, and this function is supposed to be later in development after the establishment of the limb bud 55 56 57 58 . Indeed, zebrafish mutants that abolish Hh signaling, like syu or smu, do have fin buds, but these fail to grow out 35 59 60 . Although Shh is required late in limb formation, Gli3, a downstream component of the Hh pathway, may act earlier. This has been shown convincingly by several reports 61 62 63 . The results indicate that the repressor form of Gli3 (Gli3R) prevents expression of hand2 in the anterior limb bud, thereby prepatterning this structure. According to this model activation of Hh signaling should lead to processing of Gli3R to Gli3Act and hand2 should become expressed in the anterior limb bud. This leads to expansion of posterior genes and ectopic shh expression. Morphologically it should lead to polydactylous limbs rather than absence of limbs, similar to what has been observed in mouse Gli3 deficient limbs. This is in fact the opposite of what we find in ptc1;ptc2 double mutants. We find that hand2 is suppressed and that shh expression is lost in the prospective fin bud. One possible explanation for this discrepancy is that there are evolutionary differences between fish and more advanced vertebrates. Therefore, it will be interesting to create a mouse ptc1;ptc2 double mutant to test whether the findings from zebrafish hold up in mammals. Recently, a mouse ptc2 mutant has been generated 49 , but since the Ptc1 mutant 64 is embryonic lethal, more sophisticated inducible mutants should be generated to test whether our finding on fin induction are conserved in mammals.

Retinoic acid may have an early role in establishing the fin bud territory, since RA has been found to affect tbx5 and hand2 expression. During segmentation, RA formed in the first few somites is necessary to induce fin induction with a critical time window between gastrulation and 12 hpf 29 . We have not found any evidence that the effect of Hh acts through a RA signal. The expression of aldh1a2 is not altered by Hh activation and exogenous RA does not rescue pectoral fin formation. Furthermore, wnt2b acts downstream of RA to establish the pectoral fin territory, and its expression is also unaffected in the ptc1;ptc2 mutant. Our results suggest that Hh signaling may have a very early negative effect on the establishment of the fin field in the zebrafish.

At the 10 somite stage when tbx5 expression is just established 65 66 and normal in double mutants, we find ptc2 expression both in the somites and lateral plate, whereas ptc1 is only detectable in the somites, however, in ptc1;ptc2 double mutants ptc1 becomes expressed more laterally as well (See Additional file 1). Thus it is possible that Patched proteins function around this stage in the fin primordium to maintain tbx5 expression. This does not exactly fit with our data that cyclopamine rescues already at the 50%- 5 somite stage. However, rapid development, in combination with perdurance of the cyclopamine after wash-out could account for this. It might be possible that Hh signals from the midline could put a limit to the fin bud territory at the medial side. However such a model would imply that loss of Hh signaling should lead to an increase in the tbx5 expression domain towards the midline, which has not been reported 65 . To accommodate for this, such a model would require a second unknown signal that inhibits expansion in parallel to Hh.

An alternative explanation would be that a subset of cells in the somites is transformed into adaxial cells/slow muscle, or intermediate mesoderm could be affected. As a result of this transformation a cell type may be lost that is capable of inducing the pectoral fin primordia, or gain a signal that blocks formation of these primordia, creating a neomorphic phenotype. Indeed, the lbx1 positive population of premigratory limb muscle cells is lost in ptc1;ptc2 mutants (See Additional file 1), but there is currently no evidence that these particular cells are required to form a limb bud in mice or fish 67 68 Again this model would require postulation of a second unknown signal. In the case of a positive signal, it would be required parallel to the "RA/Wnt2b" signal. Elegant experiments from Gibert et al. 29 , suggest that there is no second positive signal from the somites, since application of RA in embryos in which the paraxial mesoderm is genetically ablated can induce early fin markers.

We currently favor the first model where Hh signaling can inhibit tbx5 directly in the fin primordia but that this is normally redundant and may help to make pectoral fin development more robust. Transplantation experiments may help to resolve where ptc1 and ptc2 exert their inhibitory function through the somites or directly in fin primordium, but such experiments will be challenging due to the low frequency of ptc1;ptc2 double mutants, and the inability to identify such animals at the stage these experiments are normally performed.

Zebrafish were maintained and staged according to standard protocols 69 . Zebrafish lines used: ptc1 ^hu1602 , lep ^tj222 , blw ^tc294z .

An ENU induced mutation library was screened for Ptc1 mutants, analyzing an amplicons covering exon 9–12, using the TILLING protocol 36 . Primer sequences can be obtained upon request.

MOs for Ptc1 and LynGFP mRNA (a kind gift of R. Kim) were injected at the 1-cell stage according to 25 in a range from 0.5 to 3 ng for the MO and approximately 1 nl of 25 ng/μl for LynGFP. Prdm1 MOs were injected according to Baxendale et al, 2004 26 .

ISH experiments were performed as described in Thisse et al, 1993 70 . The following probes where generated according to the indicated articles: ptc1 15 , fgf8 71 , myod 72 , tbx5 31 , hand2 42 , uncx4.1 73 , eng1 41 , dlc 52 ), dlx2 74 , gli1 6 , foxa1 75 , ptc2 50 , shh 76 , islet1 77 , acta1 (a gift of S. Baxendale), pax2, pax6 78 , lbx1 79 . The prdm1 and nkx2.2a probes were generated from PCR product, using T3-tailed primers and subsequent transcription with T3-RNA polymerase. Whole mount immunohistochemistry was performed according to standard protocols 80 we used a pan-islet monoclonal antibody 39.4D5 (Developmental Studies Hybridoma bank) and monoclonal DP312 81 to label pax3/7 cells 42 Pax3/7 was detected using Cy3 conjugated donkey anti-mouse secondary antibody (Jackson Immuno Research), mounted in Vectashield+DAPI (Vector Labs) and viewed under a Zeiss LSM510 confocal microscope.

Progeny of ptc1 ^hu1602 /+ ; lep(ptc2)+/- was treated with all-trans RA as described before 82 . Concentrations were ranging from 10^-6 to 10^-7 M, diluted from a 10^-2 M stock in DMSO. Embryos were treated and analyzed from 4 hpf till 48 hpf. Genotypes were determined subsequently on the complete clutch to confirm the presence of ptc1;ptc2 double mutants.

Cyclopamine was dissolved in 96% EtOH to a concentration of 10 mM. Zebrafish embryos were treated from 5.5 hpf with varying concentrations of cyclopamine and controls with an equal volume of 96% EtOH.