Introduction

How the extremely complex regulatory landscapes of certain developmental genes are originated and assembled in evolution is unclear. Although the presence of genomic regulatory blocks (GRBs) – in which key developmental factors are linked to bystander genes that contain regulatory information for the former – has been extensively described1,2,3, the origin and evolution of such syntenic blocks and their potential implications for organismal evolution is still poorly understood. One of the best characterized examples of a functional GRB involves Sonic Hedgehog (Shh)4,5, a major morphogen in animal development6,7. Shh has been implicated in a wide variety of ontogenetic processes, such as the dorso-ventral (D–V)8 and antero-posterior (A–P)9 patterning of the developing central nervous system (CNS), the development of limbs10, inner ear11, digestive system12, etc. Accordingly, Shh shows a remarkably complex expression pattern during development, comprising four major domains at early stages: CNS, notochord, epithelial sheet from the oral cavity to the hindgut and limbs13.

This complexity of developmental functions and expression patterns is paralleled at the genomic level. In mouse, Shh enhancers are scattered across a vast regulatory landscape spanning more than 850 Kbp, including its two introns, a gene desert of 729 Kbp in the upstream intergenic region and two upstream neighboring transcriptional units, the bystander genes Lmbr1 and Rnf32. This region constitutes a GRB around Shh conserved in most vertebrate species14 and comprises all Shh enhancers identified to date. A subset of these enhancers drives Shh expression to CNS domains conserved across jawed vertebrates (Figure 1). In the developing spinal cord, Shh is expressed all along the floor plate and this expression is crucial for proper D–V patterning of the neural tube and the differentiation of specific cell populations8. In mouse, this expression is directed by two enhancers (Shh Floor Plate Enhancers, SPFE1 and SFPE2) that are located proximally upstream of the Shh coding region and in the second intron, respectively15. Expression in the brain is more complex16,17 and is controlled by at least four different enhancers. In particular, within the diencephalon, Shh expression shows a characteristic dorsal expansion from the basal plate: the core of the Zona Limitans Intrathalamica (ZLI). The ZLI is an important secondary organizer that regulates specific diencephalic fates through the action of Shh9. ZLI expression, together with those in the midbrain and caudal diencephalon are driven by the Shh Brain Enhancer 1 (SBE1), also located within the second intron, as SFPE215. The other three brain enhancers, SBE2-4, control more rostral expression domains and are located far upstream from the Shh coding sequence18.

Figure 1
figure 1

Genomic location of tissue-specific Shh enhancers in mouse and zebrafish.

(A) Distribution of tissue-specific enhancers across the large upstream region and introns of Shh in mouse chromosome 5. Each enhancer is represented as a color block and its associated expression is shown in the same color in the schematic embryos above. SFPE1 (green) and SFPE2 (yellow) drive expression throughout the floor plate of the spinal cord; SBE1 (lile), to the midbrain and caudal diencephalon, including the ZLI; SBE2-4 (red and dark and light blue), to more anterior domains in the developing brain; MRCS1 (purple), MFCS4 (light brown) and MACS1 (dark brown) to epithelial linings; and MFCS1/ZRS (light orange) to limb buds. Two enhancers lay within the intronic sequence of bystander genes Lmbr1 (MFCS1/ZRS) and Rnf32 (MACS1) and two within the second intron of Shh (SBE1 and SFPE2). (B) Distribution of known enhancers in zebrafish shha gene. ar-A (light green) drives expression to the notochord and some brain structures; ar-B (dark orange), throughout the spinal cord floor plate; ar-C (dark green), to forebrain and notochord and weakly in the floor plate; and ar-D (yellow), to the anterior floor plate. Adapted from different sources4,18,22.

Full size image

Expression to other developing tissues is also driven by specific enhancers, recognizable on the basis of sequence conservation across different vertebrate groups (i.e. as highly conserved non-coding regions, HCNRs). Limb expression is controlled by an enhancer located within the bystander gene Lmbr1, ∼800 Kbp upstream of Shh in mouse (MFCS1, also called ZPA Regulatory Sequence (ZRS), Figure 1A)5,19,20,21. Similarly, expression to postpharyngeal linings is driven by an enhancer conserved from mammals to amphibians (MACS1), which is located within an intron of the Rnf32 gene (Figure 1A)4; two other enhancers (MFCS4 and MRCS1) are also located near Rnf32 and promote Shh transcriptional activation in more anterior linings. Finally, regarding notochord expression, an enhancer (SNE) has been identified upstream of mouse Shh, seemingly overlapping with SFPE1, and, although they have not been characterized, at least two notochord enhancers lie within the gene desert upstream of Shh, according to BAC screenings in mouse18.

Shh regulation has also been extensively studied in zebrafish. Perhaps surprisingly, the scenario is quite different, although most of the proximal enhancers can be traced by sequence similarity. Three HCNRs were identified within shha (two in intron 1, ar-A and ar-B and one in intron 2, ar-C, Figure 1B), plus a forth HCNR upstream, near the transcription start site (ar-D) (Figure 1B). Enhancers ar-D and ar-C correspond to SFPE1 and SFPE2, respectively. Their function, however, differs from the mouse counterparts, which drive expression throughout the floor plate: ar-D drives expression only to the anterior floor plate and ar-C promotes expression in forebrain and notochord and only weakly in the floor plate22. On the other hand, ar-B drives expression throughout the spinal cord floor plate22 and it has been lost in mammals23 and ar-A drives expression to notochord and some brain structures, similar to ar-C22. Phylogenetic footprinting using coelacanth – a slow-evolving, sister species of the tetrapods – show that these four HCNRs are ancestral; nonetheless, the enhancer function of the coelacanth sequences is more similar to the tetrapod counterparts23. The enhancer(s) responsible for other expression domains have not been characterized yet in zebrafish, although HCNRs orthologous to some of the mouse elements are present in teleost species14,18,20,24.

Despite the fact that Shh seems to have taken most ancestral Hedgehog functions6, tetrapods have two other paralogs, Indian hedgehog (Ihh) and Dessert hedgehog (Dhh), originated in the two rounds of whole genome duplication (WGD) occurred at the base of vertebrates25. The coding sequences of these paralogs are more divergent and their developmental expression domains and functions are much more restricted than those of Shh, especially in the case of Dhh6. Accordingly, the regulation of both Dhh and Ihh have received little attention and only one enhancer, responsible for the Ihh-specific expression during endochondral bone formation, has been identified so far26. This element is located within the longest intron of the upstream neighboring gene, Nhej1, suggesting that this gene is part of the Ihh GRB. In invertebrates, Hedgehog genes also show complex expression patterns and play crucial roles during development in all studied species6,27,28,29,30,31,32. In the basal chordate amphioxus, the best living proxy to the vertebrate-invertebrate ancestor bodyplan, Hh is expressed in four major developing regions at early developmental stages: CNS, notochord, tail bud and pharyngeal endoderm (including forming gill slits)32,33, some of which readily correspond to vertebrate Shh expression domains. In the developing CNS, amphioxus Hh is also restricted to the ventral side of the forming neural tube up to a rostral limit; however, in stark contrast to all vertebrates, no expression is found in the most anterior part of the amphioxus CNS, including no dorsal ZLI-like expansion32,33,34,35. This suggests important changes in the regulation of Shh/Hh during chordate diversification; however, the evolution of Hh regulatory landscape is still poorly understood.

Here, we have analyzed the amphioxus Hh genomic locus to get insights into the origin and evolution of the vertebrate Shh GRB. We have sequenced ∼55 Kbp of Hh loci in the European amphioxus, Branchiostoma lanceolatum, and performed phylogenetic footprinting analyses with two sister species (the Floridian and Chinese amphioxus) and several vertebrates. We found widespread conservation of non-coding sequences within the amphioxus Hh locus between the three cephalochordates, but we could not identify reliable orthologous sequences to any of the vertebrate HCNRs. In order to test cryptic regulatory conservation, we also generated transgenic zebrafish lines carrying amphioxus sequences spanning the whole Hh locus. One region drove reporter expression consistent with the endogenous Hedgehog genes in zebrafish and amphioxus (in developing pharyngeal endoderm and gill slits). This sequence and cis-regulatory function has no evolutionary correspondence to any described vertebrate enhancer, further supporting a general lack of regulatory conservation between vertebrates and amphioxus. Finally, we investigated the microsynteny associated with the Hedgehog locus in vertebrates and invertebrates. We found strong evidence for a vertebrate-specific genomic rearrangement affecting Shh/Dhh between the two rounds of WGD that configured a novel microsyntenic environment that included the enhancer-containing bystander genes Lmbr1 and Rnf32 as parts of a new GRB.

Results

Cloning and sequencing of B. lanceolatum Hh loci and comparison to other amphioxus species

We sequenced ∼55 Kbp of genomic sequence of the Hh locus from the European amphioxus, B. lanceolatum, following two main strategies (see Methods). A genomic fragment of 41,161 bp, containing the three coding exons of Hh, was sequenced from different phages identified through screening of a genomic library. In addition, 13,640 bp further upstream from the Hh loci were cloned using primers designed on lowly polymorphic B. floridae regions, based on haplotype comparisons. In total, the sequenced region comprised 29,069 bp upstream of the start codon, the three coding exons, the first and second introns (11,458 and 10,312 bp, respectively) and 2,714 bp downstream of the termination codon, including the full 3′ untranslated region (Figure 2A).

Figure 2
figure 2

Comparison of Hh locus among three amphioxus species.

(A) Schematic representation of the B. lanceolatum Hh genomic region and genomic fragments cloned and sequenced in this study. Coding exons are shown in dark blue and UTRs in light blue. Phage genomic fragments are named after λ and A–C indicates the exon(s) contained in the sequence. (B) VISTA plot comparing B. lanceolatum (reference, on top) with B. floridae (BfHh) and B. belcheri (BbHh) Hh loci using highly stringent conditions (LAGAN alignment, window size = 300 bp, minimum width = 300 bp, identity = 80%). Dark/light blue indicates coding/UTR exonic sequence and pink shows non-coding regions conserved above threshold.

Full size image

We next compared this sequence with the orthologous genomic regions from the Floridian and Chinese amphioxus species, Branchiostoma floridae and Branchiostoma belcheri, using LAGAN alignments visualized as VISTA plots (Figure 2B). Despite the three amphioxus species diverged at least 100 million years (my) ago36,37,38, we found widespread conservation of non-coding sequences, even using highly stringent conditions (calculation window = 300 bp, minimum width = 300 bp, sequence identity = 80%). The conservation of non-coding sequences was particularly striking within the two Hh introns, with regions having sequence similarity of 90% over >1,500 bp among the three amphioxus species (Figure 2B).

Comparison of Hedgehog loci from amphioxus and vertebrates

We next compared the amphioxus Hh locus with the vertebrate paralogs, Shh, Dhh and Ihh. One of the main differences between these loci is the massive upstream gene-free region in most Shh genes, compared to amphioxus Hh and the other two vertebrate paralogs. For instance, in mouse, the region upstream of Shh up to Rnf32 comprises 729 Kbp, whereas Ihh and Dhh have upstream intergenic regions of 16 and 5.4 Kbp, respectively and B. belcheri Hh has 27 Kbp. These differences suggest higher complexity in the regulatory landscapes for the Shh genes. On the other hand, the two conserved introns are more than twice the length in amphioxus Hh than in Shh genes, despite the fact that several enhancers have been described within these vertebrate introns15,22. Thus, it could be possible that much of the cis-regulatory information in amphioxus is also contained within these introns.

We attempted to identify deeply conserved HCNRs across chordates using VISTA plots of different alignment software for amphioxus Hh and several vertebrate Shh loci (Figure 3, see Methods). Several vertebrate- or tetrapod-specific HCNRs were identified in the Shh upstream region (Figure 3A), including all previously characterized enhancer elements4,23. However, none of these elements seems to be significantly conserved in amphioxus, even using highly relaxed conservation parameters (see Methods). For example, using LAGAN alignments, VISTA analysis detected a possible trace of conservation only for SBE4. (Figure 3A, light blue); nonetheless, this short sequence had low complexity and was not conserved between the three amphioxus species (see also Figure S1), despite their widespread general sequence identity, suggesting it is likely a false positive.

Figure 3
figure 3

Comparison of amphioxus Hh and vertebrate Shh loci.

(A) VISTA plot for the alignment of the Shh genomic region from six vertebrate species and amphioxus Hh using mouse Shh as reference with default conditions, except for amphioxus (window size of 50 bp, minimum width of 50 bp and sequence identity threshold of 60%). The aligned sequences comprise the genomic region between Lmbr1 and Shh, both included. (B) Detailed alignment within the Shh loci, as indicated in (A). Each reported tissue-specific enhancer is highlighted using the same color code as in Figure 1. Species abbreviations: HsShh, human, GgShh, chicken, AcShh, green anole lizard, XtShh, Xenopus tropicalis, OlShh, medaka and BfHh, Floridian amphioxus.

Full size image

Similarly, no conservation of non-coding regions was observed within the two Shh introns and for the sequence corresponding to the proximal floor plate enhancer SFPE1 (Figure 3B). Given that the two SFPE enhancers drive Shh expression to specific CNS subdomains that are likely homologous to those of amphioxus developing neural tube32, we performed specific local alignment of these sequences to the corresponding amphioxus regions. No trace of sequence conservation was found for any of the enhancers, including full sequence motif arrangements for previously described functional transcription factor binding sites39. Consistently, only a partial arrangement for SFPE2, with a putative FoxA2 binding site, has been previously reported40. Finally, comparison of amphioxus Hh to the other two vertebrate paralogs, Dhh and Ihh, gave similar negative results (Figure S2). Within jawed vertebrates, however, a few HCNRs were detected. In the case of Ihh, they were restricted to tetrapods (most of them to amniotes) and mostly located within the introns of its upstream neighboring gene, Nhej1, consistent with previous results26. Dhh presented a more extreme situation and no conservation could be identified outside mammals, in line with this paralog having a much simpler transcriptional regulation than Shh. Using amphioxus, medaka or Xenopus as reference genomes for the above analyses yielded similar results (data not shown).

Enhancer activity of amphixous Hh sequences in transgenic assays in zebrafish

The lack of non-coding sequence conservation over long phylogenetic distances is not particularly surprising, since it is known to be rare3,41,42. However, despite lack of sequence similarity, positive enhancer activity from amphioxus sequences has been successfully detected in zebrafish transgenic assays3, presumably reflecting conservation of ancestral chordate regulatory states. To investigate if this was the case for Hedgehog, we assayed >30 Kbp from the amphioxus Hh locus for enhancer activity using zebrafish transgenesis. Since the widespread conservation of non-coding regions among the amphioxus species precluded the identification of discrete candidate HCNRs, we generated zebrafish lines carrying overlapping fragments spanning both introns and ∼11 Kbp upstream from the transcription start site (Figure 4A). Only two fragments (D and F) drove consistent mosaic reporter expression at 24 hpf or 48 hpf embryos, but only F, within the second intron, was consistent with the endogenous shh expression (D drove expression to the hatching gland, Figure S3). To better determine the enhancer activity within the region F, we generated stable transgenic lines for this fragment. Three out of four different stable lines showed GFP expression in the developing pharynx and gill slits, confirming the results from the F0 assays. In situ hybridization of GFP transcripts confirmed reporter expression to developing pharyngeal endoderm and branchial arches, but not in notochord or CNS (Figure 4D and transversal section in Figure 4E). This expression is part of the endogenous expression pattern of zebrafish shh genes (arrow head in Figure 4C43,) and presumably homologous to the expression of amphioxus Hh in developing pharyngeal endoderm and gills slits33. In addition, we also generated stable transgenic zebrafish lines for fragments B and G, spanning the equivalent regions to those where the floor plate enhancers lay in vertebrates (Figure 1). None of these regions activated GFP expression in the transgenic embryos at these stages and only founders with control RFP expression were identified for each construct (data not shown).

Figure 4
figure 4

Transgenic reporter analysis of amphioxus Hh sequences in zebrafish.

(A) Schematic representation of the location and length of the seven fragments (A–G) spanning >30 Kbp of the amphioxus Hh locus tested by transgenesis in zebrafish. In red, ‘F’, the only fragment that drove GFP reporter expression consistent with the endogenous genes in zebrafish and amphioxus (D–E). (B) Conservation of the B. lanceolatum sequence compared to B. floridae, as in Figure 1. (C) In situ hybridization of shha in a zebrafish 30 hpf embryo. Arrowhead shows expression in pharyngeal endoderm and forming gill slits. (D) In situ hybridization of GFP in a stable transgenic embryo carrying fragment F. Expression is only observed in pharyngeal endoderm and gill slits (note that the seemingly dorsal expression domain correspond to the expression of the opposite side, as shown in (E)). (E) Section through dashed line in (D) showing expression of GFP is restricted to pharyngeal endoderm and gill slits and not present in notochord or CNS.

Full size image

Synteny analysis of vertebrate and invertebrate Hedgehog loci identifies a genomic rearrangement that has remodeled Shh regulatory landscape

To reconstruct the evolutionary history of Shh GRB, we studied the local synteny surrounding members of the Hedgehog gene family across metazoans (Figure 5). Within jawed vertebrates, we found a clear correspondence for general genetic neighborhood for the three vertebrate Hedgehog genes, with the region upstream of each Hedgehog gene containing at least one gene from three gene families (Des/Prph, DnaJB and Tub) (Figure 5A). This pattern suggests that this cluster of genes is an ancestral local linkage group, established before the WGDs that gave rise to the three Hedgehog paralogs in vertebrates. Interestingly, the genes immediately adjacent to the Hh paralogs showed a more complex pattern. As mentioned above, Shh is neighbored by the upstream bystander genes Lmbr1 and Rnf32, which contain important regulatory elements for Shh; however Dhh contains only one of these genes (the paralog Lmbr1l) and the region upstream of Ihh contains neither gene, but instead the phylogenetically unrelated Nhej1 gene.

Figure 5
figure 5

Syntenic organization of Hedgehog genes and vertebrate-specific genomic rearrangement.

(A) Genomic organization of the three Hedgehog paralogs in mouse (Shh, Dhh and Ihh) and in different selected invertebrates (red arrows, indicating the orientation of transcription). Lmbr1/Nhej1/Rnf32 orthologs are represented by black/yellow/white arrows, respectively. Vertical bars represent intervening genes: green (Prph/Des), blue (DnaJB2/DnaJB6), orange (Tub1/Tub4) and black (other genes). Chromosome or scaffold is indicated for each species. (B) Possible evolutionary scenario for the insertion of the genomic fragment containing Lmbr1 and Rnf32 into the ancestral Shh/Dhh regulatory locus, some time between the two rounds of vertebrate WGD. Nnej1 may have been lost along with the insertion or in a different event.

Full size image

To determine the source of this discrepancy we then studied the chromosomal regions containing the single-copy ancestral Hh genes in invertebrates. We found that Nhej1 is widely linked to Hh in invertebrates (amphioxus, sea urchin, sea anemone and the coral Acropora, Figure 5A), demonstrating that Njeh1 was ancestrally linked to Hh, but has been lost in the Shh and Dhh regions. Interestingly, we found that neither Lmbr1 nor Rnf32 is linked to Hh in invertebrates; however, these genes are found linked to each other and with the same relative orientation in a different amphioxus genomic scaffold. Therefore, the simplest explanation for these data is that a small-scale rearrangement during vertebrate evolution introduced a chromosomal fragment including both Lmbr1 and Rnf32 into a (largely) intact Hh locus and likely removed another fragment containing a Nhej1 paralog, creating the novel arrangement observed in Shh and Dhh (the latter of which has apparently subsequently lost Rnf32). Furthermore, that Lmbr1 genes are found in the regions surrounding Shh and Dhh, but not in the third paralog Ihh, which in turn maintains the ancestral linkage to Nhej1 present in invertebrate species, suggests that this arrangement arose after the first round of genome duplication (giving rise to the Ihh locus and the ancestor of the Shh/Dhh locus), but before the second duplication that gave rise to the separate Shh and Dhh loci, thus providing a very precise time point for origin of the now-key bystander relationship of Shh and Lmbr1 and Rnf32: between the two ancestral vertebrate WGDs.

Discussion

Using comparative genomics and transgenesis in zebrafish we have investigated the evolution of the Hedgehog regulatory landscape within chordates. Despite remarkably conserved expression patterns during early embryonic development, we found little evidence for cis-regulatory conservation between the cephalochordate amphioxus and vertebrates, notwithstanding large conservation of non-coding regions within each lineage. In addition, we identified a vertebrate-specific genomic rearrangement, further differentiating the regulatory landscapes in both lineages.

Many cis-regulatory elements of Shh have been identified or defined by comparison of non-coding sequences among vertebrates4,5,14,15,18,19,20,21,23, suggesting that Shh regulation is largely conserved within the vertebrate lineage (or at least within tetrapods), with some elements, such as the limb enhancer MFCS1/ZRS, dating back to the origin of gnathostomes4,22,44 and others of all vertebrates40,45. Similarly, we found widespread conservation of non-coding sequences among the three studied amphioxus species, spanning ∼100 million years of cephalochordate evolution36,38, at a level comparable to other loci with well-known conserved expression patterns (e.g. the Hox cluster46). On the other hand, we found no conservation of non-coding sequences between cephalochordates and vertebrates. Although the large evolutionary distance between both lineages (at least 535–550 my47) has often rendered too large for identification of conserved non-coding sequences, some HCNRs and cryptic conserved regulatory elements have been indeed identified for some important developmental genes with conserved expression patterns41,42,48,49,50, suggesting that the regulation of Hedgehog loci is, at least, not particularly constrained over long evolutionary times compared to other genes with similarly crucial functions.

Further supporting the idea that the Hedgehog locus may have experienced large evolutionary divergence during chordate evolution, we also found no clear cases of cryptic conservation of regulatory elements in our transgenic assays. Only one out of the seven fragments (F) – spanning >30 Kbp of the amphioxus Hh locus – tested for enhancer activity in zebrafish drove reporter expression consistent with the endogenous zebrafish and amphioxus Hedgehog genes. This fragment overlaps ∼0.5 Kbp with fragment G (Figure 4A), partly including a highly conserved block described above, but not the short conserved stretch reported by Rétaux et al40. Since fragment G did not drive similar expression, the reported enhancer activity may therefore lay within the upstream half of the second intron (red in Figure 4) and not within the largest highly conserved region. Importantly, this sequence promotes expression to pharyngeal endoderm and developing branchial arches. The only Shh enhancer with similar activity described to date4, MRCS1 (Figure 1), does not lay within the orthologous intron, but far upstream, more than 500 Kbp away in mouse and close to the bystander Rnf32 and it is conserved only from mammals to reptiles. In addition, in both mouse and zebrafish, extensive probing of both intron sequences for enhancer activity15,18,22,51 did not show any equivalent enhancer in any of the two vertebrate species. This evolutionary divergence is also consistent with comparisons of cis-regulatory elements between mouse and zebrafish. Although some of the enhancers can be identified as orthologous by sequence similarity in the two vertebrate species, they hardly drive similar expression patterns when tested in reporter assays22. Therefore, Shh regulatory landscapes do not seem to be tightly constrained at the sequence level even within major vertebrate groups, despite the extensive expression pattern conservation observed across lineages.

However, it is important to note that several experimental limitations may lead to false negatives when probing sequences for regulatory activity. First, the amphioxus sequences are being tested in heterologous systems, not in their endogenous regulatory environments. Although amphioxus sequences have been extensively reported to be active in vertebrate systems3,41,42,48,49,50, it is still unclear how sensitive and reliable the heterologous approach is. Second, in the specific case of the various Shh floor plate enhancers both in mouse and in zebrafish, they have been shown to be often codependent and their activity enhanced in a cooperative or synergic way18,22,39. Therefore, the combination of different amphioxus sequences could also be necessary to drive significant reporter expression. Unfortunately, this issue is very difficult to evaluate without knowing were the specific regulatory elements reside in amphioxus. Third, it is also possible that other amphioxus enhancers lay in further upstream or downstream regions, or even within the neighboring gene Nhej1, as the previously reported endochondral bone Ihh enhancer26. Consistent with this possibility, several discrete HCNRs are detected within the two long introns of the amphioxus Nhej1 gene, comparing B. floridae and B. belcheri (Figure S4). Finally, only early developmental stages have been probed in this study and thus it is possible that shared regulatory inputs do exist for later stages of development; however, large conservation of expression patterns between Shh and amphioxus Hh is observed only at these early stages29,30.

Perhaps the most exciting finding of this study is the vertebrate-specific genomic novelty associated with the origin of the Shh-Rnf32-Lmbr1 genomic regulatory block. First, these results suggest that Dhh and Shh may be more phylogenetically related to each other than to Ihh, in contrast to previous phylogenetic analyses32,45,52, likely affected by the faster evolutionary rates of Dhh coding sequence. Second, this genomic novelty may be associated with a key novel expression domain of Shh. Shh is expressed in the limbs of all jawed vertebrates, including both bony and cartilaginous lineages24,44. The recruitment of Hedgehog signaling to these structures has been suggested as one of the crucial events for the origin of paired appendages, probably through the cooption of genetic programs that were already operating in the median fins44,53. Importantly, despite extensive searches for regulatory elements in different species, only one enhancer responsible for the limb expression of Shh has been identified to date, the MFCS1/ZRS enhancer, which is located within the fifth intron of the bystander gene Lmbr121 and is highly conserved across gnatostomate species4,14,20,44. Remarkably, our results demonstrate that the recruitment of the Lmbr1 gene into the Shh regulatory landscape to establish a new GRB – and seemingly replace the old one integrated by Nhej126 – occurred within the vertebrate lineage, though a genome rearrangement between the two rounds of WGD. Whether Lmbr1 already contained regulatory elements at the time of the genomic rearrangement or it simply provided the appropriate raw material for the evolution of the enhancer, this new syntenic configuration may have allowed the recruitment of Shh expression to the limbs. Although it may not be possible to confidently establish a causal relationship between the two evolutionary events, it suggests the exciting possibility that, in some cases, the remodeling of the genome architecture may underlay the evolution of gene regulation and the appearance of novel traits.

Methods

Genomic library screening and PCR-based cloning

We screened a Lambda Fix II/XhoI genomic library (Stratagene) of B. lanceolatum54 with [α-32P] dCTP-labeled probes by random-hexamer priming. Approximately 6×105 recombinant phages were screened at standard conditions (60°C)54. For the primary screening, we used a probe for each of the three B. lanceolatum exons (EU754743). This strategy allowed the identification of positive phages containing the first (λ5020A), the second (λ5010B) and the second and third exons together (λ5009BC) and neighboring non-coding regions (Figure 2). We performed a second screening using a probe designed at the 5′ of λ5020A that provided 15 Kbp upstream the ATG (λ5032). All phages were sequenced by randomly interspersed primer-binding sites technology using a Tn7 transposon-based system (GPS®-1 Genome Priming System, New England BioLabs) and specific ‘walking’ primers and the assembly was made by Phred, Phrap and Consed software55,56,57.

We next used a different strategy to clone further upstream B. lanceolatum genomic sequence. Taking advantage of the high polymorphism in the B. floridae amphioxus genome, we aligned the genomic sequences from the two Hh haplotypes (scaffolds 137 and 532) and selected blocks that had >99% conservation over long sequence stretches (300–600 bp). We then designed 2–3 forward and reverse primers spanning these regions and use them together in a single PCR reaction for each block using B. lanceolatum genomic DNA and low annealing temperature. We cloned and sequence the PCR products for each block using pCRII/TOPO vectors (Invitrogen). Then, between each block we designed B. lanceolatum specific primers and performed PCR reactions using iProof DNA polymerase (Promega) to amplify long fragments and cloned them. Using this strategy we cloned three new blocks, BS2-4 (Figure 2), that were sequenced using primers specifically designed for sequence walking. All primer sequences are available upon request. The whole assembled Hh locus from B. lanceolatum has been submitted to GeneBank (accession number JX034725).

Phylogenetic footprinting analyses

We used the following genomic sequences and annotations: (i) B. floridae, scaffold 532 combined with 137 when necessary, from Nhej (inclusive) to 2 Kbp downstream Hh (total ∼82 Kbp); (ii) access to unpublished B. belcheri genome sequence was kindly provided by Dr. Anlong Xu and the equivalent region to B. floridae was used (∼68.5 Kbp); for vertebrates, the regions including the bystander genes (Lmbr1/Lmbr1l, Rnf32 and/or Nhej1) to the Hedgehog paralogs were extracted from Ensembl (see below), together with prebuilt VISTA annotations. Annotations for the amphioxus genes were done by sequence conservation to the B. lanceolatum orthologs (BlHh, EU754743; BlNhej1, JX034724). Orthology relationships could be unambiguously determined by best reciprocal blasts (see also below for Lmbr1).

Phylogenetic footprinting was performed using the visualization tool mVISTA58 for multi-species alignments generated using the LAGAN software59 (visualization of alignments produced by AVID and Shuffled-LAGAN yielded similar results). For comparisons between the three amphioxus species we used high stringency conditions for peak calling in the VISTA plots (window size of 300 bp, minimum width of 300 bp and sequence identity threshold of 80%). For comparisons within vertebrate paralogs and between vertebrates and amphioxus, we used standard conditions (window size of 100 bp, minimum width of 100 bp and sequence identity threshold of 70%) for vertebrates and lower stringent conditions for amphioxus (window size of 50 bp, minimum width of 50 bp and sequence identity threshold of 60%). Usage of different reference genomes for alignment visualization (i.e. amphioxus, medaka or Xenopus) yielded similar results.

Synteny comparisons and genomic resources

We used the following genome resources to browse and search for orthologs of Hh, Nhej1, Lmbr1 and Rnf32: Trichoplax adhaerens Grell-BS-1999 v1.0, Nematostella vectensis v1.0, B. floridae v1.0, Ciona intestinalis v2.0 and v1.0, Daphnia pulex v1.0, Lottia gigantea v1.0 and Capitella teleta v1.0, at DOE Joint Genome Institute (JGI) webpage (http://genome.jgi-psf.org/euk_home.html) and of Strongylocentrotus purpuratus Build 2.1, Drosophila melanogaster Build Fb5.3, Homo sapiens Build GRCh37, Mus musculus Build 37.1, Gallus gallus v2.1, Anolis carolinensis AnoCar1.0, Xenopus tropicalis JGI v4.1 and Oryzias latipes at the Ensembl webpage (http://www.ensembl.org) and Saccoglossus kowalevskii 2008-Dec-09 scaffolds at HGSC Baylor College of Medicine webpage (http://blast.hgsc.bcm.tmc.edu) and the Acropora digitifera genome60. Paralogs for the different Shh, Dhh and Ihh neighbouring genes in vertebrates were obtained by the Ensembl paralog tool. Lamprey could not be included in the analyses because of the current incomplete and fragmentary genomic assembly, in particular for both Hh genes45.

Phylogenetic analyses of Lmbr1/Lmbr1l genes

We downloaded full protein sequences for Lmbr1 and Lmbr1l from H. sapiens, X. tropicalis and D. rerio, and Lmbr1 orthologs from the following invertebrates: B. floridae, B. belcheri, S. purpuratus, S. kowalevskii, L. gigantea, C teleta, D. melanogaster, D. pulex, N. vectensis from the sources mentioned above. Sequences for Apis mellifera and Tribolium castaneum were obtained through NCBI BLAST (http://blast.ncbi.nlm.nih.gov/Blast.cgi). In addition we used putative Lmbrd1 genes from H. sapiens, B. floridae, L. gigantea and N. vectensis as outgroups. Phylogenetic trees were generated by the Bayesian method with MrBayes 3.1.261,62 using two independent runs (each with four chains). Model selection using ProtTest63 (best model: CpRev+G), convergence determination, burn-in and consensus tree calculations were done as previously described64,65. In total, 3,000,000 generations were run, reaching convergence at generation 685,000; all trees prior convergence were discarded and the remaining ones were used to build the consensus tree (Figure S5). This tree shows that all investigated Lmbr1 genes are orthologs and that Lmbr1 and Lmbr1l in vertebrates arose from a vertebrate-specific duplication, most likely one of the two WGDs.

Transgenic analyses in zebrafish and in situ hybridization

We designed primers to amplify seven overlapping B. lanceolatum genomic fragments of 4–7.0 Kbp. PCRs were performed on B. lanceolatum genomic DNA or on the corresponding phage DNA extractions using iProof™ High-Fidelity DNA Polymerase (Bio-Rad). PCR products were cloned in pCR8GW/TOPO vector (Invitrogen) according to manufacturer. Sequence-verified clones were then transferred with the Gateway recombination system (Invitrogen) to the ZED vector66. The final transgenic constructs were purified using phenol-chlorophorm and normalized at 50 ng/ml in DEPC water prior to microinjection. For each construct, >100 injected embryos were assayed and GFP expression investigated at 24 and 48 hpf. RFP expression within the muscles observed 72 hpf served as a positive control for transgenesis. Two constructs showed consistent GFP expression in F0 (D, 28/130 injected embryos (22%) and F, 25/165 (15%)); of these, only F – which showed an expression pattern consistent with the endogenous shha gene – was raised to the next generation to obtain stable transgenic lines (F1 lines), in addition to fragments D and G, with negative GFP activity. For embryonic gene expression analysis of GFP driven by fragment F and endogenous shha by in situ hybridization, zebrafish embryos were fixed at different stages in 4% paraformaldehyde overnight at 4°C; and in situ hybridizations carried out as previously described67.