Comparisons of multiple vertebrate shh loci indicate a high degree of sequence similarity between zebrafish, fugu, chick, mouse, and human (Figure 1). A global alignment using shuffle Lagan algorithm and visualization by VISTA plot clearly identifies all three exons of shh orthologs and paralogs throughout vertebrate evolution (Figure 1). The CRMs identified previously are conserved among shh(a) genes (orange peaks), and the degree of their conservation is in accordance with the evolutionary distance between the species compared. In contrast, the zebrafish shhb gene exhibits no obvious conservation with the shha ar-A, ar-B, ar-C, and ar-D CRMs. Apart from Shuffle Lagan, Valis 36 has also failed to detect conserved putative CRMs of shhb (data not shown). Taken together, these findings indicate that although orthologous regulatory elements may exist between shhb and shha, they are much less conserved at the DNA sequence level than are shha elements, as detected by the applied alignment programs.

To characterize individual regulatory elements better, we focused on a single enhancer element ar-C, which is conserved between fish and mouse (SFPE2) and which has been analyzed in considerable detail in both species 266366. To this end, first we addressed whether the ar-C enhancer or its mouse ortholog SFPE2 is detectable across shh(a) loci in various vertebrate species from different lineages that diverged before and after the gene duplication event leading to the evolution of shh paralogs in zebrafish. Because the zebrafish shha ar-C enhancer is located in the second intron of shha and exhibits high sequence similarity to human and mouse counterparts, candidate ar-C containing intronic fragments of several vertebrate species were amplified by polymerase chain reaction (PCR) with degenerate oligonucleotide primers. We cloned and sequenced the relevant genomic DNA fragments from several fish species that experienced the genome duplication, such as the cyprinid tench (Tinca tinca), fugu, and medaka 45. In addition to actinopterygian fishes, several species of sarcopterygians such as chick, mouse, and the early sarcopterygian lineage Latimeria menadoensis were used in the analysis. All sarcopterygians diverged from the common ancestor with actinopterygians before the fish-specific genome duplication in the ray-finned fish lineage. A sequence comparison of intron 2 sequences from the available vertebrate model systems revealed a high degree of sequence similarity in all species specifically in the region that spans the ar-C enhancer in zebrafish and the SFPE2 enhancers of mouse (Figure 2a). This analysis also indicated that the orthologous Latimeria genomic region also contains a highly conserved stretch of sequence in the ar-C region, which is consistent with the hypothesis that ar-C is an ancestral enhancer of shh genes.

To test whether the sequence similarity observed between ar-C enhancers of different lineages of vertebrates is also indicative of conserved tissue-specific enhancer function, we carried out transgenic analysis of enhancer activity in microinjected zebrafish embryos. We utilized a minimal promoter construct (containing an 0.8 kilobase [kb] upstream sequence from the transcriptional start site with activity similar to the -563shha promoter described by Chang and coworkers 67, linked to green fluorescent protein (GFP) reporter. Transient mosaic expression of GFP was measured as read-out of reporter construct activity by counting fluorescence-positive cells in the notochord and floor plate, where the ar-C enhancer is active, in the trunk of 1-day-old embryo (Table 1). This approach was a reliable substitute for the generation of stable transgenic lines, as reflected by the identical results obtained with transient analysis and stable transgenic lines made for a subset of the constructs used in this study (Additional data file 1).

As described previously, the zebrafish ar-C enhancer is primarily active in the notochord and only weakly in the floor plate (Figure 2c). Intron 2 sequences of tench, chick, and Latimeria shh genes gave strong enhancer activity in the notochord (Figure 2d-f). However, the mouse intron 2 (with the SFPE2 enhancer) was found to be inactive in zebrafish (data not shown), suggesting that SFPE2 had functionally diverged during mammalian/mouse evolution either at the cis-regulatory or the trans-regulatory level. All together, these data indicate a high degree of functional conservation between ar-C sequences among vertebrates.

The evolutionary functional divergence of paralogous ar-C enhancers was tested through the isolation of the shhb intron 2 from zebrafish. Because a genome duplication event has taken place early in actinopterygian evolution, it was predicted that the ostariophysian and cyprinid zebrafish as well as all acanthopterygian fish model species whose genomes are known (medaka, stickleback, green spotted pufferfish, and fugu) may contain a shhb homolog. Analysis of the available genome sequences of these four species of teleost fish indicated that none of them carries a discernible shhb homolog, suggesting that these lineages (which evolved some 290 million years after cyprinids 68) may have secondarily lost this shh paralog. Synteny is observed between the medaka genomic region surrounding shh on chromosome 20 and a region on chromosome 17; however, chromosome 17 lacks shhb (Additional data file 2). This finding further supports the hypothesis that a shhb gene was originally present after duplication but has been lost secondarily during evolution.

However, we were able to detect and isolate shhb and its intron 2 from another cyprinid species, tench, by PCR using degenerate oligonucleotides that were designed in conserved exon sequences. Importantly, the isolation of more than one shhb intron 2 sequences from cyprinids allowed for phylogenetic footprinting of shhb genes and a search for a putative ar-C homolog. We have compared the shha and shhb intron 2 sequences between zebrafish and tench (Figure 3a). The shha orthologs between zebrafish and tench exhibit a high degree of sequence similarity, which is strongest in the region in which ar-C resides. In contrast, comparison of intron 2 from shhb and shha paralogs of either species revealed no conspicuous conservation. The apparent lack of sequence similarity, however, does not necessarily rule out the possibility that a highly diverged ar-C homolog enhancer may still reside in shhb intron 2. A sequence comparison between zebrafish and tench shhb intron 2 reveals a striking sequence similarity in the 3' region close to exon 3, where a positionally conserved ar-C would be predicted to be located. This suggests that intron 2 of shhb genes of cyprinids may contain a functional enhancer, which has diverged significantly from the shha ar-C. Furthermore, the apparent sequence divergence suggests that the function of the shhb enhancer may also have diverged.

To test whether the conserved sequence in the intron 2 of shhb genes is indeed a putative enhancer element, we tested several shhb fragments representing approximately 10 kb of the locus in transgenic reporter assays. The shhb proximal promoter and 2.7 kb of upstream sequences can activate GFP expression in the notochord (Figure 3b) but only very weakly in the floor plate, similarly to previously reported data 69. Because shhb is only expressed in the floor plate and never in the notochord, this GFP expression of the reporter is an ectopic activity and reflects the lack of a notochord repressing functional element, probably located elsewhere in the unexplored sequences around the shhb locus. The weak expression in the floor plate suggests that other CRMs are required for floor plate activation. In shha a floor plate enhancer resides in intron 1 26. To check whether a similar enhancer exists in shhb, intron 1 of shhb was attached to the promoter construct. It was found that it did not enhance the promoter's activity, indicating no obvious enhancer function in this transgenic context (Figure 3c). Interestingly, the addition of shhb intron 2 does result in enhancement of expression in the floor plate (Figure 3d). This finding indicates that intron 2 of shhb contains a floor plate enhancer.

The 2.7 kb upstream and proximal promoter sequence of shhb may have influenced the autonomous function of an enhancer in intron 2. To address the activator functions of the identified shha and shhb enhancers without influence of potential upstream regulatory elements, a series of injection experiments was carried out in which the enhancer activities were analyzed with a minimal promoter containing only 0.8 kb of the shha promoter (Figure 3e-j). Moreover, activity of intron 2 sequences from shha and shhb genes from both zebrafish and tench were systematically compared. Shha intron 2 fragments of both species consistently resulted in comparable notochord activity (Figure 3f and Additional data file 1 [parts B and C]), wheres the shhb intron 2 fragment from both species exhibited distinct enhancement of expression in the floor plate and reduction in GFP activity in the notochord (Figure 3g,h). The presence of a highly conserved region within the intron 2 of zebrafish and tench shhb genes strongly suggests that the floor plate enhancer activity is the property of this conserved sequence. To test this prediction a set of deletion analysis experiments was carried out. Zebrafish shhb intron 2 was cleaved into a 1,026 base pair (bp) fragment of nonconserved and a 380 bp conserved sequence. As shown in Figure 3i,j, the floor plate specific enhancer effect is retained by the conserved fragment but not by the non-conserved sequence, verifying the prediction of the location of the floor plate enhancer. Taken together, a diverged, floor plate active ar-C enhancer has been discovered in the shhb intron 2, which is consistent with the floor plate specific expression of shhb in zebrafish.

Transcription factor binding sites may be more conserved than the surrounding sequences 70. We have hypothesized that sequence similarity between fish and human ar-C sequences may indicate conserved motifs, which may reflect conserved transcription factor binding sites 66. We postulated that putative transcription factor binding sites and changes in them may be detectable by identification of motifs using local alignment of ar-C from large numbers of pre-duplicated and post-duplicated shh orthologs and paralogs. To this end, a CHAOS/DIALIGN 71 alignment was used to compare the functionally active ar-C enhancer of zebrafish (as described by Muller and coworkers 26) and equivalent sequences from all major vertebrate classes. The alignments were arranged according to phylogeny (Figure 4).

A pattern of conserved motifs is detected in the form of homology blocks extending to 20 to 30 bp. These conserved motifs exhibit distinct distribution characteristics, which reflect phylogenic as well as paralogy and orthology relationships between shh genes. C1 and C3 are homology blocks, which are present in all shh sequences, including shhb paralogs, in all species analyzed. In contrast, C2 and C4 are homology blocks that are present only in shh(a) genes but absent in shhb genes. Because C2 and C4 are present in pre-duplicated enhancers of sarcopterygians, the lack of C2 and C4 in shhb enhancers is probably due to a secondary loss of these elements after the fish-specific gene duplication. The two sets of putative binding sites (C1/C3 and C2/C4, respectively) may thus be targets for transcription factors that regulate the differential enhancer activities of shh(a) (predominantly notochord expression) and shhb (predominant floor plate expression). In conclusion, we identified a set of putative targets of mutations that may contribute to the divergence of ar-C enhancer functions after gene duplication.

To test the functional significance of the two sets of homology blocks, we conducted a systematic mutation analysis of the C1 to C4 conserved homology blocks in both shha and shhb genes. Furthermore, we carried out exchange of homology blocks between shha and shhb ar-C enhancers to test whether evolutionary changes after gene duplication can be modeled in a transgenic zebrafish system.

As shown in Figure 5b-f, mutations inserted into homology blocks (C1 to C4) result in dramatic changes in shha ar-C enhancer activity. Replacement of C1 with random sequence results in total loss of ar-C enhancer function, indicating that this binding site is critical for shha ar-C activity (Figure 5b). By contrast, loss of C3 results in no observable effect, suggesting that this conserved block is either not required for enhancer function or only necessary for functions that are not detectable in our transgenic system (Figure 5d). Importantly, removal of C2 or C4 (the blocks that are only present in shha genes) results in strong expression of GFP in the floor plate (Figure 5c,e). In the case of C4 removal, a reduced reporter expression in the notochord has also been observed (Figure 5e). The obtained expression pattern strongly resembles the activity of the wild type shhb ar-C enhancer (compare panels e and g of Figure 5). Thus, removal of shha-specific motifs from the shha ar-C mimics shhb ar-C enhancers. Moreover, this result is consistent with a model in which the C2 and C4 elements are targets for repressors of floor plate expression in the shha ar-C enhancer.

The multiple alignment of ar-C homolog sequences revealed a noticeable modification in the C4 element of acanthopterygian fishes, which do not have a shh paralog (fpr example, medaka and fugu; see Figure 4 and Additional data file 3 for alternative alignment results). The divergence in the C4 motif of acanthopterygians may reflect a functional change in the ar-C enhancer in these species, potentially leading to the relaxation of the floor plate repression observed in ar-C of shha genes. To test whether the modification of the C4 motif of acathopterygians may reflect the loss or modification of C4 repressor function, we have replaced the C4 of zebrafish shha with that of medaka shh. The resulting hybrid construct activated strong expression in the floor plate (Figure 5f), suggesting that the medaka C4 motif is unable to rescue the repressing activity of zebrafish shha C4 in zebrafish embryos.

We next asked whether shhb ar-C is active in the floor plate because it contains the general midline activator site C1 and lacks the floor plate repressor elements C2 and C4 that are present in the shha ar-C enhancer. To this end, we first tested whether the C1 and C3 of shhb are required for the function of the shhb enhancer. Similar to the results obtained with shha, C1 was found to be critical for the activity of shhb ar-C (compare panels b and h of Figure 5), whereas loss of C3 had no effect, thus mimicking the findings in shha (Figire 5i). We then introduced C2 or C4 into the shhb enhancer in order to test the functional significance of the lack of C2 and C4 motifs in shhb. When a shh-derived C2 was introduced into shhb ar-C, no effect was observed (Figure 5j), but introduction of the C4 putative floor plate repressor motif from shha did result in a dramatic shift in shhb enhancer activity (Figure 5k). The effect was a repression of floor plate expression while notochord activity was retained, thus resembling the wild-type or C2 mutant shha ar-C enhancer (Figure 5a,c). In a control experiment, random DNA sequence was introduced at similar positions into the shhb ar-C enhancer. However, this manipulation had no effect on the activity of shhb ar-C (data not shown), indicating that the changes observed with the C4 insertion are due to the specific sequence of C4. These results together strongly suggest that the function of C4 is to repress floor plate activation by the shha ar-C enhancer. Together, these findings are consistent with a model in which loss of the C4 motif in the evolution of the shhb ar-C has contributed to its floor plate specific activity.

Using phylogenetic footprinting of intron 2 of shhb genes we have identified a conserved ar-C homolog enhancer in two species of cyprinids. The results of our transgenic analysis indicate that the ar-C sequences in intron 2, together with the promoter activity of shhb 69, contribute to this gene's activity in the floor plate. Although shh(a) enhancers retained significant sequence similarity with their orthologs, the whole of the shhb gene and its ar-C enhancer is grossly changed from that of shha paralogs. This paralog-specific change happened despite the fact that shhb had equal time and chance to diverge as did shha after duplication from an ancestral sonic hedgehog gene. This result is in accordance with observations indicating selective pressure on the CRMs of paralogs in invertebrates 79 as well as in vertebrates 19208081. Our results, together with the reports cited above, provide experimental support to the notion that differential divergence of noncoding conserved elements of paralogs may be a general phenomenon in vertebrates 35.

Use of a local sequence alignment approach of representative species of major vertebrate lineages allowed us to predict functionally relevant motifs within the ar-C enhancers. Our findings are most consistent with a model in which these motifs are individual or multimeric transcription factor binding sites. Mutation and transgenic analysis verified the functional relevance of these motifs in driving expression in the midline, and therefore the most parsimonious explanation for the conservation of these sequence elements is that they represent functional binding sites for developmental regulatory transcription factors.

The ar-C enhancer is composed of motifs with different regulatory capacities (Figure 6a). Motifs exist that are crucial for the overall activity of the enhancer (C1), whereas other repressor motifs refine enhancer activity (C2 and C4). This indicates that the overall activity output of an enhancer in midline tissues is subject to both activator and repressor functions acting in concert. These results are in accordance with the previously proposed grammar of developmentally regulated gene expression 11828384858687. Importantly, the order and combination of motifs of ar-C are conserved. This is a very different result from that proposed for the stripe 2 enhancers of drosophilids, in which the functional conservation of CRMs was a result of stabilizing selection of reshuffled transcription factor binding site composition 177. The evolutionary pressure to keep the order and composition of binding sites within enhancers may be limited to transcription factor and developmental regulatory genes 8889. The high conservation level, however, may be a consequence of selective pressure acting on a secondary function of enhancer sequences 90.

Previously, individual binding sites were identified through comparative approaches in vertebrates (for instance, see 669192). These examples, together with our systematic analysis of conserved motifs in the ar-C enhancers, demonstrate that functionally relevant motifs detected by sequence alignment may aid in identifying as yet unknown and uncharacterized functional transcription factor binding sites.

The use of large numbers of species spanning long evolutionary distances allowed us to generate a phylogenetic reconstruction of enhancer divergence before and after gene duplication (Figure 6b). By generating artificial enhancers with mutations that mimic the predicted lineage-specific changes in motif composition of shhb and shha enhancers, we were able to reconstruct the probable evolutionary events leading to divergence of the ar-C enhancer function. For example, insertion of the floor plate repressor C4 element into shhb resulted in enhancer activity reminiscent of shha ar-C, in which the C4 site had been identified. These findings indicate that the very changes that resulted in the divergence of the enhancer function have been identified.

An open question remains, however; why should the ar-C enhancer of shha be repressed in the floor plate while the shha gene is well known to be active in this tissue? The level of the Hedgehog morphogen signal emanating from the embryonic midline is critical for correct patterning of the ventral neural tube 93. Animals with only one gene encoding the Sonic hedgehog protein (sarcopterygians and fishes without shhb) achieve this by controlled activation of shh in the notochord and floor plate as a result of a combination of several synergistic enhancers 6263. In zebrafish and other ostaryophisian species (for instance, tench and Mexican cavefish) a second copy of shh paralog (shhb) also contributes to Shh production in the floor plate. At least in zebrafish, controlled levels of the floor plate expressed shhb are required, together with the notochord and floor plate derived shha, for normal patterning of branchiomotor neurons and the somites 565758. The combined activity of two shh genes emerging from the floor plate and notochord may thus result in one of the paralog floor plate enhancers being subjected to selection pressure. For example, to counter the overproduction of Hedgehog levels, the reduction in transcription can occur by blocking the activity of one of the synergistically active enhancers (in this case ar-C). It is important to note, however, that the shh(a) ar-C enhancers are not exclusively expressed in the notochord, and retained a weaker but still noticeable capacity to activate expression in the floor plate. Thus, the output of Shh levels in zebrafish appears to be a subject of quantitative regulation of paralog enhancer activities. Alternatively, it is feasible that there are time points when the two paralog genes are not overlapping in expression and the complementing specificities of shhb and shha ar-C enhancers reflect the non-overlapping production of Hedgehog proteins in the two midline tissues 53.

Recent reports have provided experimental verification of subfunctionalization of Hox gene enhancers 1718. Our report adds to those findings by contributing evidence for the diversity of subfunctionalization mechanisms that may act on paralog enhancers during evolution. Here, we propose that the presence or absence of the C4 site functions as a binary switch to modulate ar-C enhancer activity specific to one of two midline tissues after gene duplication. By selective removal of repressor and activator binding sites, subfunctionalization of the ar-C enhancer to floor plate or notochord can thus occur (Figure 6b). This model is reminiscent of those proposed for subfunctionalization of CRMs 15.

The subfunctionalization model would argue for the existence of a preduplication (sarcopterygian) ar-C enhancer that is equally active in both floor plate and notochord. Interestingly, the mouse ar-C homolog SFPE2 enhancer is mainly active in the floor plate of the mouse 63 and can activate notochord expression in a multimerized form 66 (Figure 6b). However, in fish all shh ar-C enhancers from sarcopterygian lineages exhibit notochord-specific enhancer activity. The differences between zebrafish and mouse may be explained both by subfunctionalization mechanisms as well as by trans-acting factor changes. In support of trans changes the mouse SFPE2 enhancer exhibited no activity in the fish. In the converse experiment, the mainly notochord-specific zebrafish shha ar-C exhibited both floor pate and notochord activity in mouse 26. Thus, the subfunctionalization of duplicated ar-C shh enhancers is a composite result of selective loss of several motifs, including negative regulatory elements in one enhancer (shhb) paralleled by modifications either on the cis or on the trans level to restrict activity of the less diverged sister paralog enhancer (shha). The prediction from this model is that fish species without shhb gene (acantopterygii) may have floor plate active ar-C enhancer. Interestingly, the floor plate repressor elements (C2/C4) of shha ar-C of acanthopterygians (for example, medaka and fugu) are present but diverged from all other shh(a) homologs (Figure 4), and they may thus represent the evolutionary changes that lead to retention of shh ar-C floor plate activity in these fish lineages. Our experiments with the medaka shh C4 element replacing that of zebrafish shha provide further support to the model outlined above. The hybrid zebrafish shha ar-C construct with the modified medaka C4 motif cannot rescue the loss of the zebrafish shha C4 element and does not function as a repressor site in zebrafish. These findings are in line with a predicted compensatory relaxation of repressor function of shh ar-C in medaka.

The combination of both negative and positive regulatory sites within a single enhancer indicates the integration of activating and repressing signals to modulate the resulting transcriptional activity. This could be achieved through multiple trans-acting factors that interact with a series of binding sites within the ar-C enhancer. Determining which transcription factors bind to the C1 to C4 blocks remains a challenge for future research. Predictions can be made based on known transcription factor recognition sequences. For instance, C1 contains a foxA2 binding sequence, which is consistent with the previously suggested role of this factor in regulating shh gene expression in the midline of mouse 6694, frog 95, and fish 67. Interestingly, C4 carries a sequence identical to the homeobox binding site that has been described to be present in the mouse SFPE2 enhancer 66. This binding site is required for floor plate activity in the mouse. The identity of the mouse binding factor and whether the same transcription factor acts (probably by repressing floor plate activity) in the ar-C enhancer in zebrafish are unknown. The relevance of specific transcription factors from large protein families in binding to the ar-C binding sites remains a challenging question. It is important to note, however, that the functionally relevant sequences in SFPE2 that are responsible for floor plate activity in the mouse (HR-c) 66 only partially overlap with ar-C sequences that are functionally relevant in zebrafish, and this divergence may explain, at least in part, the different results obtained with mouse and fish enhancers.

The tench shha and shhb intron 2 fragments were isolated by using degenerative oligonucleotides, designed based on conserved amino acid blocks in the second and third exons of shh(a) and shhb genes from several vertebrate species. The PCR products were directly cloned into pCRII-TOPO vector (Invitrogen, Carlsbad, CA, USA), and the clone containing the right insert was identified by sequencing.

The Latimeria intron 2 was isolated by screening of genomic bacterial artificial chromosome (BAC) library from Latimeria menadoensis 96 (Lang and coworkers unpublished data), kindly provided by Chris Amemiya. The positive BAC clone, containing the shh locus, was shotgun sequenced and relevant genomic regions were secondarily amplified by gene specific primers. The correct PCR product was identified by sequencing. The mouse and chick intron 2 were directly amplified from genomic DNA with specific oligonucleotides.

The 0.8shha:gfp plasmid was constructed by cutting out the SalI/HindIII fragment from 2.4shha:gfp plasmid 62 (described as 2.2shha:gfp in 2667) and subsequent blunting and relegating. The 0.8shha:gfp:z-shha-I2, 0.8shha:gfp:z-shha-arC, and 0.8shha:gfp z-shhb-I2 were created by subcloning the respective NotI/KpnI fragments from 2.4shha:gfp:C 62, 2.4shha:gfp:ΔC, and 2.4shha:gfp:shhb C (Müller and coworkers, unpublished data) into 0.8shha:gfp plasmid. The plasmids 0.8shha:gfp:t-shha-I2 and 0.8shha:gfp:shhb-I2 were made by reamplifying the respective intron 2 fragments from pCRII-TOPO:t-shha-I2 and pCRII-TOPO:t-shhb-I2, and subcloning them in 0.8shha:gfp using NotI/KpnI restriction sites. The 0.8shha:gfp:l-shh-I2 was constructed by reamplifying the intron 2 part from the correct PCR fragment isolated from the BAC clone and cloning it into 0.8shha:gfp (NotI/KpnI). The 0.8shha:gfp:c-shh-I2 and 0.8shha:gfp:m-shh-I2 palsmids were made by direct cloning of the respective intron 2 fragments, amplified from genomic DNA, into 0.8shha:gfp (NotI/KpnI). The 0.8shha:gfp:z-shhb-non-cons and 0.8shha:gfp:z-shhb-arC were made by cloning the PCR-amplified nonconserved 5' part of z-shhb I2 (1032 bp) and the 380 bp 3' part containing the conserved region (ar-C) into 0.8shha:gfp (NotI/KpnI). All plasmids (0.8shha:gfp:z-shha-arCΔC1, 0.8shha:gfp:z-shha-arCΔC2, 0.8shha:gfp:z-shha-arCΔC3; 0.8shha:gfp:z-shha-arCΔC4, 0.8shha:gfp:z-shha-arC+C4m, 0.8shha:gfp:z-shhb-arCΔC1, and 0.8shha:gfp:z-shhb-arCΔC3) containing z-shha-ar-C or z-shhb ar-C carrying mutations in one of the conserved motifs (C1 to C4) were created by replacing the respective wild-type sequence of each conserved block with random sequence using a PCR-based approach. The same method was used to introduce the C2 and C4 from z-shha ar-C or random sequence into z-shhb ar-C (0.8shha:gfp:z-shhb-arC+C2, 0.8shha:gfp:z-shhb-arC+C4, 0.8shha:gfp:z-shhb-arC+C2rnd, and 0.8shha:gfp:z-shhb-arC+C4rnd). The PCR products were cloned into 0.8shha:gfp (NotI/KpnI) and verified by sequencing.

The zebrafish shha ar-C 26 and shhb ar-C sequences can be found in GenBank under the following accession numbers: AL929206 (gi|34221785|, emb|AL929206.6|, region: 111,511 to 111,717 bp) for the shha ar-C and BX510360 (gi|46518135|, emb|BX510360.8|, region: 88,241 to 88,620 bp) for the shhb ar-C. The GenBank accession numbers for tench shha and shhb and Latimeria shh intron 2 sequences are as follows: EF593170, EF593171, and EF593172. For more detailed information about the sequences, which have been mutated and introduced in shha and shhb ar-Cs, see Table 2. The plasmid 2.7shhb:gfp was constructed by replacing the 2.4shha promoter fragment (SalI/XhoI) from 2.4shha:gfp with the PCR-amplified 2.7 kb shhb promoter fragment (upstream from the translation start site). The plasmid 2.7shhb:gfp:z-shhb-I1 and 2.7shhb:gfp:z-shhb-I2 were made by subcloning the shhb I1 and I2 from 2.4shha:gfp:shhb-I1 and 2.4shha:gfp:shhb-I2 (Müller and coworkers, unpublished data) into 2.7shhb:gfp (NotI/KpnI). For sequence information on the oligonucleotides that were used, see Table 3. More detailed information about the plasmid constructions is available upon request.

All microinjection experiments were performed with injection solution containing circular plasmid at a concentration of 10 to 15 ng/μl, supplemented with 0.1% phenol red. The solution was injected trough the chorion into the cytoplasm of zygotes. The GFP expression was analyzed on 24-hour-old embryos using Leica MZ FLIII fluorescent stereomicroscope (Leica Microsystems GmbH, Wetzlar, Germany). The level of expression was quantified by counting the number of GFP-positive cells in notochord and floor plate, as well as the number of ectopic GFP-positive cells in tissues where shh(a) and shhb are normally not expressed.

Pair-wise sequence aliments were performed using one of the global alignment algorithms, namely AVID 97 in the case of intronic sequences (Figures 2a and 3a) and Shuffle-Lagan 98 in case of the whole hh loci (Figure 1), and visualized using Vista 99100.

The multiple alignment of the intronic sequences was made using two algorithms, namely CHAOS/DIALIGN 71 or MUSCLE 101102, and visualized using BioEdit (sequence alignment editor written by Tom Hall, Ibis Therapeutics, Carlsbad, CA, USA).