Insights from Amphioxus into the Evolution of Vertebrate Cartilage

Central to the story of vertebrate evolution is the origin of the vertebrate head, a problem difficult to approach using paleontology and comparative morphology due to a lack of unambiguous intermediate forms. Embryologically, much of the vertebrate head is derived from two ectodermal tissues, the neural crest and cranial placodes. Recent work in protochordates suggests the first chordates possessed migratory neural tube cells with some features of neural crest cells. However, it is unclear how and when these cells acquired the ability to form cellular cartilage, a cell type unique to vertebrates. It has been variously proposed that the neural crest acquired chondrogenic ability by recruiting proto-chondrogenic gene programs deployed in the neural tube, pharynx, and notochord. To test these hypotheses we examined the expression of 11 amphioxus orthologs of genes involved in neural crest chondrogenesis. Consistent with cellular cartilage as a vertebrate novelty, we find that no single amphioxus tissue co-expresses all or most of these genes. However, most are variously co-expressed in mesodermal derivatives. Our results suggest that neural crest-derived cartilage evolved by serial cooption of genes which functioned primitively in mesoderm.


INTRODUCTION
The transition from sessile filter feeding to active predation in the vertebrate lineage was made possible by the evolution of a robust head skeleton. Embryologically, most vertebrate craniofacial cartilages and all pharyngeal cartilages are derived from the neural crest [1], a migratory and multipotent cell population formed at the edges of the nascent central nervous system. Data from invertebrate chordates suggest that the neural crest evolved from a population of migratory neural tube cells with limited developmental potential [2,3]. Key to understanding the origins of the vertebrate head is understanding how these neural cells acquired the ability to form cellular cartilage.
Based on comparative morphology [4,5] and the fossil Haikouella [6] it has been proposed that the first cartilages in the vertebrate head were pharyngeal cartilages of neural crest origin. In modern vertebrates, several genes mark post-migratory cranial neural crest cells as they populate the pharynx and form cartilage. These genes can be classified into three groups based on their expression patterns and demonstrated regulatory interactions ( Figure 1A). The first set of genes is expressed broadly in neural crest cells during migration, and persists at high levels in postmigratory cranial neural crest. This group includes, but is not limited to, Sox9 [7] (SoxE), Sox5/6 [8] (SoxD), Twist1/2 [9], Id2/ 3 [10], and Ets1/2 [11] . All of these genes except Ets1/2 have also been shown to be necessary for the formation of neural crestderived cartilages. Expression of these factors precedes upregulation of several genes expressed in nascent chondrocytes and shown to be necessary for cartilage and bone differentiation. These genes include the transcription factors Barx1/2 [12], Cart1 [13], Alx3/ 4 [14], Bapx1 [15], and Runx1/2/3 [16] and the TGF-beta signaling molecule GDF5 [17]. A third group of genes is expressed in differentiated cartilage and include classical markers of vertebrate cartilage like Col2a1 [18] and the chondroitin sulfate-binding lecticans [19] (i.e. aggrecan). Also essential for the differentiation of neural crest-derived cartilage are two classes of signaling molecules, FGFs [20] and endothelins [21]. These factors are secreted from adjacent pharyngeal endoderm and overlying ectoderm and are necessary for both cartilage differentiation and patterning via Dlx, Msx, Hand2, Bapx1, and Gsc transcription factors [21].
Gene expression studies suggest that the evolution of neural crest migratory ability and multipotency involved the cooption of transcription factors from other cell types [2,10,22,23]. Genomic comparisons suggest this cooption also coincided with the evolution of new effector genes and signaling molecules [24] . While it is likely that the evolution neural crest-derived cartilage was driven by gene cooption it is unclear if this involved recruitment of individual genes into a novel gene regulatory network, or wholesale cooption of a pre-existing genetic program. Various chordate tissues have been proposed to represent evolutionary precursors of neural crest-derived cartilage, implying the utilization of pre-existing proto-chondrogenic gene networks. Cephalochordates possess skeletal elements reminiscent of vertebrate neural crest-derived pharyngeal cartilages. Though acellular, amphioxus pharyngeal gill bars are composed of fibrillar collagen [25,26] and chondroitin sulfate [27]. They are also positioned between the endoderm and ectoderm and function to support the pharynx and express the cranial neural crest marker Id [10]. Based on these similarities, it has been suggested that the genetic network operating in cranial neural crest was recruited from collagen-secreting pharyngeal mesoderm [10] and/or endoderm [26,28]. Like amphioxus gill bars, the notochords of urochordates, cephalochordates, and vertebrates also express fibrillar collagen [29,30]. This has lead to speculation that a gene network mediating filbrillar collagen expression in vertebrate head cartilage was coopted from the notochord [30]. It has also been proposed that vertebrate neural crest-derived chondrocytes evolved from central nervous system (CNS) cells with the ability to express fibrillar collagen [31].
We tested whether vertebrate-like proto-chondrogenic gene programs could be operating in amphioxus tissues with proposed evolutionary relationships to the neural crest, including pharyngeal mesoderm, pharyngeal endoderm, neural tube and notochord. To this end we isolated amphioxus orthologs of 11 vertebrate genes involved in neural crest chondrogenesis and analyzed their expression patterns in embryos and larvae (summarized in Figure 1B). We find that no amphioxus cell type co-expresses orthologs of all or most vertebrate chondrogenic network components, arguing against wholesale cooption of a protochondrogenic gene program from any single cell type. Instead, our data suggests piecemeal assembly of the vertebrate cartilage gene network via repeated cooption of genes which functioned primitively in the mesoderm of the pre-vertebrate chordate.

Identification of amphioxus neural crest and cartilage gene homologs
Using vertebrate protein sequences we BLAST searched an amphioxus EST database (Jr Kai Yu, unpublished results) for putative amphioxus orthologs of vertebrate cranial neural crest and cartilage genes. We identified amphioxus clones corresponding to vertebrate Twist1/2, Ets1/2, Alx3/Alx4/Cart1, Runx1/2/3, Bapx1, FGF8/17/18 genes and a single amphioxus class A fibrillar collagen (ColA) ( Table 1). Amphioxus genome release v1.0 (Joint We have classified genes in the network as cranial neural crest (CNC) markers, cartilage markers, or effector genes based on their expression, regulatory relationships, and biochemical functions. Among the CNC markers are Sox9 (SoxE), Sox5/6 (SoxD), Twist1/2 and Ets1/2 genes. All of these factors are expressed in post-migratory chondrogenic cranial neural crest [8,9,11,45,46]. SoxE, SoxD, and Twist1/2 have been shown to crossregulate, and to activate cartilage specifiers and effector genes. SoxE is required for expression of both SoxD and Twist1/2 in migrating CNC [8,34], while Twist1/2 is necessary for the continued expression of SoxE in postmigratory CNC [9]. Both SoxE and SoxD cooperate to directly activate the definitive cartilage differentiation marker Col2a1 in chondroblasts [47], while Twist1/2 is required for expression of the aristalless-related transcription factors Alx3/4 and Cart1 [9]. Ets1/2 expression overlaps temporally and spatially with SoxE, SoxD and Twist1/2, though functional relationships between it and the other network components have yet to be demonstrated [11]. In sea urchins, Ets1/2 and Alx3/4 orthologs are necessary for the formation of skeletogenic mesenchyme and are regulated by the same upstream factors, suggesting they cooperate in an evolutionarily ancient skeletogenic program [48,49]. As chondrogenesis begins, presumptive pharyngeal chondrocytes express genes grouped here as cartilage markers (Barx1/2, Alx3/4, Cart1, Runx1/2/3, Bapx1, and GDF5). These genes are expressed in differentiating CNC-derived chondrocytes [12][13][14]16,21,[50][51][52], are downstream of CNC specfiers and upstream of effector genes like Col2a1 and Aggrecan, Barx1 physically interacts with Sox9 to directly activate Collagen2a1 expression [37]. As indicated above, Alx3/4 and Cart1 are regulated by Twist1/2 [9]. Runx1/2/3 expression in chondrocytes is dependent on SoxE function [16]. In the pharynx, Bapx1 functions mainly to position the jaw joint by regulating expression of GDF5 [21,53]. In the mesoderm-derived axial skeleton, however, Bapx1 is expressed broadly and operates upstream of Sox9, Col2a1, and Runx1/2/3 [54,55]. Essential for maintenance and establishment of the chondrogenic subnetwork are signaling molecules of the FGF and Endothelin families which are secreted by surround pharyngeal endoderm and ectoderm [20,21] (not shown). These genes also mediate pharyngeal arch patterning by activating nested expression of various transcription factors in the nascent cartilages including Dlx and Msx genes, Gsc, and Hand2 [21,43](not shown). (B) The major expression domains of chondrogenic neural crest gene homologs in amphioxus neurulae and larvae. No single cell type expresses the complete set of vertebrate chondrogenic network genes, indicating the cranial neural crest cartilage program is a vertebrate novelty. Notably, most factors are expressed in mesodermal derivatives, suggesting neural crest-derived cartilage evolved via repeated cooption of primitively mesodermal genes. doi:10.1371/journal.pone.0000787.g001 Genome Institute) was searched for homologs of vertebrate Sox5/6 (SoxD), Barx1/2, and GDF5/6/7. Fragments of these genes were isolated by PCR. A single amphioxus SoxE cDNA was isolated by degenerate PCR and phage library screening. The amphioxus genome and EST database were searched exhaustively for potential amphioxus-specific duplicates of these genes, but none were found, indicating they are all present as single copies in the amphioxus genome. Putative orthology of each gene was suggested by BLAST searches of GenBank with amphioxus sequences and gene models. In each case, the highest identity hits were vertebrate or sea urchin homologs of the genes used to do the original searches. Orthology was further confirmed by phylogenetic analyses (Text S1, Figure S1, S2, S3, S4 and S5). Searches of the amphioxus genome with several vertebrate endothelin sequences yielded no similar sequences. Searches with vertebrate aggrecan sequences yielded two ESTs with high similarity to the c-terminal EGF-lectin modules of vertebrate lecticans.
Expression of the cranial neural crest marker orthologs, SoxE, SoxD, Twist, and Ets At neurula stages, amphioxus SoxE was observed in cells of the ventral notochord and medial neural plate (Figure 2A,B). In early larvae (24 h), SoxE expanded throughout the neural tube, but was lost from the ventral notochord ( Figure 2D,C). SoxE expression was not detectable in late larvae. SoxD expression was seen in the nascent notochord and medial somite of neurulae ( Figure 2E,F), then in the notochord, anterior gut, and cerebral vesicle of early larvae ( Figure 2G, H). Twist expression was seen in the lateral somites and notochord of neurulae ( Figure 2I,J), similar to Twist expression in the Chinese lancelet [32]. In early larvae, Twist transcripts were detected in the ventrolateral somites as they expanded to line the coelomic wall ( Figure 2K,L). In late larvae, Twist expression was observed in the mesoderm of the first forming pharyngeal arch and right gut diverticulum ( Figure 2Q,R). Ets expression was observed in the posterior gut and in the ventral part of the anterior somites at neurula stages ( Figure 2M,N). At early larval stages expression was seen in the posterior gut, ventrolateral mesoderm, and the anterior gut diverticulae ( Figure 2O,P). In late larvae, Ets was observed in the pharyngeal mesoderm of the first pharyngeal arch and the anterior gut diverticulae ( Figure 2S,T).
Expression of amphioxus orthologs of the chondrocyte markers Alx3/Alx4/Cart1, Runx, Barx, Bapx, and GDF5 Amphioxus Alx, the ortholog of vertebrate Alx3,Alx4, and Cart1, was expressed in the lateral somites and strongly in the right gut diverticulum at neural stages ( Figure 3A,B). In early larvae, Alx expression persisted in ventral somitic mesoderm and the right gut diverticulum ( Figure 3C,D). In late larvae, expression of Alx was seen in pharyngeal arch mesoderm and the right gut diverticulum ( Figure 3E,F). Barx expression was limited to a few ectodermal cells immediately caudal to the preoral pit at larval stages ( Figure 3G, H). Amphioxus Bapx was expressed in the medial somite at embryonic stages ( Figure 3I,J). In early larvae amphioxus Bapx marked a stripe of endoderm on the right side of the pharynx approximating the future location of the first gill slit ( Figure 3K,L). Amphioxus Runx expression was seen in the posterior gut of neurulae ( Figure 3M,N) and early larvae ( Figure 3O,P) and diffusely in the late larval ectoderm (not shown). No detectable expression of amphioxus GDF5/6/7 was observed in embryos or larvae up to day 4.5.

Expression of amphioxus fibrillar collagen and aggrecan-like genes
Expression of the amphioxus ortholog of the definitive vertebrate cartilage marker Col2a1 (Amphioxus ColA) in the embryonic ( Figure 4A,B) and larval ( Figure 4C,D) notochord and neural tube was similar to previous reports. However, we noted additional expression domains not previously described. Importantly, we saw strong expression of amphioxus ColA in the pharyngeal arch mesoderm of late larvae ( Figure 4E-G), consistent with fibrillar collagen expression in the adult pharyngeal skeleton. We also observed weak embryonic expression of amphioxus ColA in the paraxial mesoderm of early larvae ( Figure 4D). We did not observe expression of the aggrecan-like c-lectin domain clones Agc-like1 or Agc-like2 in amphioxus embryos or larvae.

Expression of amphioxus FGF8/17/18
We also identified an ortholog of vertebrate FGF8, a signaling molecule secreted from pharyngeal endoderm required for pharyngeal chondrogenesis in vertebrates. At late neurula stages, amphioxus FGF8/17/18 is expressed only in the cerebral vesicle (data not shown). In early larvae, expression was observed in two patches of pharyngeal endoderm ( Figure 4H,I). Similar expression was seen in late larvae, corresponding to points of contact between the pharyngeal endoderm and ectoderm around the first and second gill slits and club-shaped gland ( Figure 4J,K).

DISCUSSION
While gene expression data in itself is not evidence of gene regulatory relationships, the co-expression of non-secreted factors is a prerequisite for direct regulatory interactions. Thus, expression data can be used to falsify hypothetical gene regulatory interactions, such as those implicit in homology arguments. In this study we examined whether any tissue in the basal chordate amphioxus could be considered an evolutionary precursor, or latent homolog [33], of neural crest-derived cartilage. We reasoned that such a tissue should broadly co-express most factors required for neural crest chondrogenesis, including upstream transcriptional regulators and downstream markers of overt cartilage differentiation. As a starting point we focused on four tissues with proposed evolutionary relationships to neural crest-derived cartilage; the neural tube, pharyngeal mesoderm, pharyngeal endoderm, and notochord.

Expression of amphioxus cartilage marker orthologs in the neural tube
Based on the expression of fibrillar collagens in parts of the vertebrate CNS, it has been proposed that vertebrate cartilage evolved from CNS cells with the latent ability to express fibrillar collagen [31]. Implicit in this scenario is that CNS cells in the prevertebrate chordate co-expressed orthologs of fibrillar collagen and the transcription factors which regulate it in cartilage. To test this we looked for neural expression of fibrillar collagen and its (W) Diagram of cross section through first pharyngeal arch in 1.5d larva. In cephalochordate larvae, gill slits on opposite sides of the pharynx form asynchronously, with the right gill slits forming first. Thus, cross sections through the pharynx of amphioxus larvae reveal single gill bars rather than the symmetrical pharyngeal arches typical of analogous sections through vertebrate embryos. In U,V, and W, light blue is epidermal ectoderm, dark blue is the neural tube, brown is the notochord, yellow is endoderm, pink is somitic mesoderm, and red is pharyngeal arch mesoderm. doi:10.1371/journal.pone.0000787.g002 putative upstream regulators in amphioxus. We observed coexpression of amphioxus ColA and SoxE throughout the neural tube of early larvae with the exception of the cerebral vesicle, which lacks ColA expression. SoxE is then down-regulated before the mouth forms in late larvae, while ColA persists at least until the 2 gill slit stage. No other cartilage marker orthologs are broadly coexpressed with these two factors, though SoxD and Ets both label the larval cerebral vesicle. In vertebrate embryos SoxE genes (Sox8, Sox9 and Sox10) are required for initial specification of the neural crest [34]. Later in post-migratory cranial neural crest cells, SoxE genes regulate the chondrogenic program and cooperate with SoxD and Barx factors to directly activate Col2a1 [35][36][37]. Both these early and late functions of SoxE genes appear conserved to the base of the vertebrate lineage [38,39]. The lack of broad coexpression of amphioxus SoxE with SoxD and Barx, argues against the presence of a vertebrate-like proto-chondrogenic program in the CNS of the prevertebrate chordate ancestor. However, the tight temporal and spatial co-expression of SoxE and ColA is consistent with an ancient gene-regulatory relationship between these genes in neural tissue.

Expression of amphioxus cartilage marker orthologs in the pharyngeal mesoderm and endoderm
Electron microscopy and immunohistochemical analyses have revealed that, like vertebrate pharyngeal cartilages, the amphioxus gill bar skeleton is composed of fibrillar collagen and chondroitin sulfate [25][26][27]. This has lead to speculation that a vertebrate-like skeletogenic gene program operated in the pharyngeal endoderm [26,28] or mesoderm [10] of the pre-vertebrate chordate.
To evaluate these hypotheses we tested for broad co-expression of cartilage marker orthologs in the pharynx of amphioxus. In the pharyngeal mesoderm of larvae we observed coexpression of Twist, Ets, Alx, and a homolog of the differentiated cartilage marker Col2a1. In vertebrates, Twist1/2 and Ets1/2 genes are expressed at high levels in post-migratory pharyngeal neural crest during pharyngeal arch formation. Twist 1/2 has been shown to be necessary for these cells to form cartilage [9,11] and Ets1 regulates expression of integrins in chondrocytes [40]. The aristallessrelated cartilage markers Alx3, Alx4, and Cart-1 are similarly expressed in post-migratory pharyngeal neural crest and are required for chondrogenesis [13,41]. Previous studies have not reported pharyngeal ColA mRNA expression at larval stages which would account for the presence of collagen protein in the gill bars [29,30], though ColA transcripts are expressed broadly in pharyngeal ectoderm, endoderm, and mesoderm of adults [28]. We detect strong expression of ColA in larval pharyngeal mesoderm (Figure 4), suggesting amphioxus gill bars are initially mesodermal in origin. Thus, amphioxus pharyngeal mesoderm coexpresses orthologs of three transcription factors which regulate chondrogenesis in vertebrates, in addition to ColA and the cranial neural crest marker Id. While broad coexpression of these factors is suggestive of a rudimentary vertebrate-like chondrogenic program, amphioxus pharyngeal mesoderm does not express SoxE or SoxD, two factors essential for the formation of all vertebrate cartilages. This tissue also does not deploy orthologs of the vertebrate cartilage markers Runx1/2/3, GDF5, Barx1/2, or Bapx1. Furthermore, Twist1/2 and Alx3/4 mark lateral plate mesoderm in vertebrates, indicating their function in amphioxus pharyngeal mesoderm is not necessarily skeletogenic. Thus, while our data suggests some genes involved in vertebrate chondrogenic genes are expressed together with fibrillar collagen in amphioxus pharyngeal mesoderm, it is unclear if they function in a vertebrate-like chondrogenic gene network.
In the pharyngeal endoderm, we observed partially overlapping expression of SoxD, Bapx, and the signaling molecule FGF8/17/18. SoxD was expressed throughout the pharyngeal endoderm while Bapx mRNA was detected in a restricted domain approximating the region of the forming mouth. FGF8/17/18 was observed in patches of ventral endoderm near the forming gill slits. In vertebrates, FGF3 and FGF8 are expressed in pharyngeal endoderm where they function to induce cartilage. Conserved expression of amphioxus FGF8/17/18 in pharyngeal endoderm may indicate a conserved function in inducing pharyngeal skeletogenesis or patterning. However, the lack of broad coexpression of cartilage marker orthologs, including ColA, in amphioxus pharyngeal endoderm indicates this tissue does not deploy a vertebrate-type skeletogenic gene program at larval stages.
A proto-chondrogenic differentiation program does not operate in the amphioxus notochord Based on gross structural and biochemical similarities, and the expression of fibrillar collagen, it has been proposed that vertebrate cellular cartilage evolved by redeployment of a gene program which operated primitively in the notochord [38]. To address this possibility we assayed for coexpression of amphioxus orthologs of vertebrate cartilage markers in the notochord.
We observed expression of four transcription factors and ColA in the axial mesoderm of amphioxus. As previously reported, amphioxus ColA marks axial mesoderm until larval stages, mimicking Col2a1 expression in vertebrates [38]. This expression overlapped to a limited extent with expression of SoxE in ventral notochord cells. However, broad expression of amphioxus ColA throughout the axial mesoderm implies that amphioxus ColA expression in this tissue is not SoxE-dependent as it is in neural crest-derived cartilage. Twist and SoxD are also coexpressed with ColA in the axial mesoderm of neurulae. Twist marks strips of ventral and dorsal notochord cells and is downregulated before larval stages. SoxD is broadly coexpressed with ColA , but like Twist , it is downregulated in early larvae. Notochord expression of ColA also overlaps with weak expression of Ets in the anterior notochord at neurula stages. In vertebrates, SoxD genes are necessary for Col2a1 expression in notochord-derived cells [42]. Coexpression of amphioxus SoxD and ColA in the notochord may reflect an evolutionarily conserved regulatory relationship in axial mesoderm. Taken together, we find little evidence that a gene network resembling the neural crest chondrogenic program operates in the amphioxus notochord. However, it is possible that both the amphioxus notochord and cranial neural crest cells utilize SoxD genes to regulate fibrillar collagen expression.

Lecticans and endothelins are vertebrate novelties associated with the evolution of cartilage
In vertebrate cartilages, lecticans are the major chondroitin sulfatebinding proteins. We could not identify clear amphioxus orthologs of lecticans (i.e. aggrecan) in the amphioxus genome. We did isolate two EST clones similar to the c-terminal domain of vertebrate lecticans (Tab. 1), but neither was expressed in embryos or larvae. Histological and biochemical assays demonstrate that amphioxus gill bars contain acid mucopolysaccharides and chondroitin sulfate [27]. It is possible that genes structurally related to lecticans, but not strictly orthologous to them, bind chondroitin sulfate in amphioxus. In vertebrates, endothelins are secreted molecules which induce and pattern pharyngeal arch cartilages [43]. We did not find a clear homolog of vertebrate endothelins in the amphioxus genome. Like the lecticans, this class of genes may represent a vertebrate novelty associated with the evolution of cellular cartilage. Recent genomic comparisons confirm the absence of endothelins in protochordates and indicate that other families of signaling molecules associated with neural crest migration and differentiation are unique to vertebrates [24]. Thus, the evolution of chondrogenic neural crest is associated with the cooption of evolutionarily ancient transcriptional regulators, as well as the appearance of novel downstream effector genes and signaling molecules like lecticans and endothelins.
De novo assembly of the vertebrate CNC cartilage program via cooption of primitively mesodermal genes It is unknown how neural crest cells acquired the genetic machinery necessary to form cellular cartilage. One possibility is this occurred relatively rapidly by wholesale cooption of an ancient chondrogenic program. Alternately, chondrogenic ability could have evolved gradually in neural crest cells via piecemeal cooption of individual genetic components. We find that no embryonic or larval tissue in amphioxus co-expresses all or most cartilage network orthologs, supporting de novo assembly of the vertebrate chondrogenic neural crest gene program. Though it is possible that some cartilage network orthologs are re-deployed together after metamorphosis, we view this as unlikely since most amphioxus tissue types, including the primary gill bars, form during larval stages [44].
Assuming that the vertebrate chondrogenic gene network is unique to vertebrates, we asked what the primitive function of these genes may have been in the vertebrate ancestor. While expression patterns alone do little to inform the precise functions of genes, similar expression of orthologous genes across related phyla often reflects conserved functional relationships. As mentioned above, SoxE and ColA both mark the neural tube in amphioxus, suggesting a vertebrate-type regulatory relationship between these genes in neural tissue predates the evolution of vertebrate cartilage. In adult hemichordates, which lack a central nervous system, SoxE and ColA are also coexpressed in pharyngeal endoderm, indicating this regulatory cassette may have evolved before the origins of chordates.
In both sea urchins and vertebrates, Alx and Ets genes are expressed in skeletogenic mesenchyme. Expression of amphioxus Alx and Ets genes in pharyngeal mesoderm which gives rise to collagenous skeletal elements may reflect conservation of an ancient deuterostome skeletogenic gene program coopted by neural crest cells. Functional studies will reveal if these genes act to regulate gill bar formation in amphioxus and if they can be considered components of a primitive chordate skeletogenic gene program.
In addition to Alx and Ets, we found that most other amphioxus orthologs of vertebrate CNC and cartilage markers were expressed in mesodermal derivatives, while relatively few genes mark epidermal, neural, or endodermal cells ( Figure 1B). Similar expression of vertebrate chondrogenic neural crest markers in mesodermal derivatives suggests that most components of the vertebrate neural crest cartilage program operated primitively in mesoderm. This is consistent with the overlapping developmental potentials of cranial neural crest cells and mesoderm to generate connective tissue, muscle, and cartilage. In sum, our data suggests that the vertebrate chondrogenic program likely evolved via serial cooption of primitively mesodermal genes to neural crest cells in the first vertebrates.

MATERIALS AND METHODS
Vertebrate protein sequences were used to BLAST search an amphioxus EST database (Jr Kai Yu, unpublished results) for putative amphioxus orthologs. Access to library clones was kindly provided by Drs.J.K. Yu and Linda Holland. Degenerate PCR using primers against vertebrate Sox8, Sox9, and Sox10 (SoxE591: TACGAYTGGWCIYTNGTNCCIATGCC, Sox-E391:GGCTGRTAYTTRTAITCIGGRTRRTC) followed by phage library screening (library a gift of Jim Langeland) was used to isolate an amphioxus SoxE ortholog. Amphioxus genome release v1.0 (Joint Genome Institute) was searched for putative orthologs of vertebrate SoxD, Barx1, and GDF5. Fragments of these genes were isolated by PCR using diluted phage library and a vectorspecific primer (SoxD) or genomic DNA (Barx, GDF). Forward and reverse primer sequences were: SoxD59: CCCCACATCAAGCGGCCAATGAATG BarxEx159: TATAGCTGGTTGTGCCTCTTG BarxEx139: AACATTCTACACACTGCGACG BarxEx2 59: GGAGGAGTTTACAGAGAGTAAC BarxEx2 39: ACAAGTCTTGTTGTGACCTGTAC BarxEx3 59: GCAATTAGCCTACGGACAC BarxEx3 39: GCGTGTTCCGATTAGTACAG GDF5Ex1 59: GAAAGGGGTAGATTGATTTCTTTTC GDF5Ex1 39: TACAGCCTTGTCGACGAAC GDF5Ex2 59: ATTTTGAACAGCTGCCGGG GDF5Ex2 39: CTGGGGTTCATGGAGTTG For each amphioxus gene, in situ hybridization was performed on embryos and larvae ranging from 12 hour early neurula to 2.5d feeding larva (2-3 gill slits) as described previously. In the cases of Barx and GDF, in situ probes made against their amplified exonic sequences were pooled and used together. Embryos were embedded in 20% gelatin in Phosphate Buffered Saline and cryostat sectioned. Wholemount embryos in 40% glycerol/PBS, and sections, were photographed using a Zeiss AxioSkop 2 Plus. Images were processed using Adobe Photoshop.

SUPPORTING INFORMATION
Text S1