A New Mechanistic Scenario for the Origin and Evolution of Vertebrate Cartilage

The appearance of cellular cartilage was a defining event in vertebrate evolution because it made possible the physical expansion of the vertebrate “new head”. Despite its central role in vertebrate evolution, the origin of cellular cartilage has been difficult to understand. This is largely due to a lack of informative evolutionary intermediates linking vertebrate cellular cartilage to the acellular cartilage of invertebrate chordates. The basal jawless vertebrate, lamprey, has long been considered key to understanding the evolution of vertebrate cartilage. However, histological analyses of the lamprey head skeleton suggest it is composed of modern cellular cartilage and a putatively unrelated connective tissue called mucocartilage, with no obvious transitional tissue. Here we take a molecular approach to better understand the evolutionary relationships between lamprey cellular cartilage, gnathostome cellular cartilage, and lamprey mucocartilage. We find that despite overt histological similarity, lamprey and gnathostome cellular cartilage utilize divergent gene regulatory networks (GRNs). While the gnathostome cellular cartilage GRN broadly incorporates Runx, Barx, and Alx transcription factors, lamprey cellular cartilage does not express Runx or Barx, and only deploys Alx genes in certain regions. Furthermore, we find that lamprey mucocartilage, despite its distinctive mesenchymal morphology, deploys every component of the gnathostome cartilage GRN, albeit in different domains. Based on these findings, and previous work, we propose a stepwise model for the evolution of vertebrate cellular cartilage in which the appearance of a generic neural crest-derived skeletal tissue was followed by a phase of skeletal tissue diversification in early agnathans. In the gnathostome lineage, a single type of rigid cellular cartilage became dominant, replacing other skeletal tissues and evolving via gene cooption to become the definitive cellular cartilage of modern jawed vertebrates.


Introduction
A defining feature of the craniate subphylum (vertebrates and hagfish) is the ''New Head'', which has been linked to the evolution of active predation in the vertebrate lineage [1]. The vertebrate head is a composite structure consisting of paired sense organs, an expanded central nervous system, peripheral ganglia, and a muscular, pumping pharynx. These components are physically supported and protected by cellular cartilage, which is replaced by bone in osteichthian vertebrates.
While it can be considered a vertebrate-specific morphological unit, most components of the vertebrate head have clear invertebrate antecedents. For instance, the basic organization of the vertebrate brain has deep roots in the neural tube of invertebrate chordates [2,3], while the paired sense organs and cranial ganglia appear to be derived from evolutionarily ancient cranial placodes and sensory cells [4,5,6]. Similarly, the vertebrate pharynx, though modified for pumping water over the gills and capturing prey, retains the basic respiratory and feeding functions of the pharynx in all deuterostomes.
Although much of the vertebrate head likely evolved via the reorganization and augmentation of simpler precursors, the origin of cellular cartilage is less clear. Invertebrates appear to lack any tissue displaying the combination of morphological and biochemical properties that defines vertebrate cellular cartilage. In addition, comprehensive analysis of gene expression in amphioxus, a basal chordate, suggests that no single invertebrate cell type coexpresses all, or most, of the genes needed to drive cellular cartilage formation [7]. Rather, individual components of the vertebrate cartilage gene regulatory network (GRN) are expressed in different tissues, most of which are mesoderm-derived.
To better understand the evolutionary relationships between agnathan skeletal tissues and gnathostome cartilage, we are analyzing the expression of gnathostome cartilage GRN components in the sea lamprey Petromyzon marinus. Here we describe the expression of lamprey homologs of three key regulators of gnathostome chondrogenesis, Runx, Barx and Alx/Cart1 from embryonic through larval stages when differentiated cellular cartilage is histologically discernable. Contrary to our expectations, we find that neither Runx nor Barx are co-expressed with SoxE genes in pharyngeal CNCC destined to form the cellular cartilage of the branchial basket. Rather, we see Runx expression in mesenchyme flanking the mouth, and Barx expression in the lower lip and mesenchyme medial to the pharyngeal mesoderm. Similarly, we see restricted expression of Alx in the dorsal and ventral aspects of the branchial basket and in the upper lip, but not in other SoxE and fibrillar collagen-expressing skeletal tissues. Given that Runx, Barx, and Alx are all required for gnathostome cellular cartilage development, our data show that lamprey and gnathostome cellular cartilage develop using divergent GRNs. Furthermore, combinatorial expression of lamprey SoxE, Runx, Alx, and Barx in different populations of presumptive mucocartilage suggest lamprey possess multiple skeletal tissues with genetic features of gnathostome cellular cartilage. Based on these findings, and previous work, we propose a scenario in which early agnathan vertebrates possessed an array of genetically and functionally distinct CNCC-derived skeletal tissues. In the gnathostome lineage, only one of these ancestral tissues was retained and evolved via gene cooption to become the cellular cartilage of modern gnathostomes.

Results and Discussion
SoxE expression, type A fibrillar collagen expression, and alcian blue staining support common evolutionary and developmental origins for mucocartilage and definitive cellular cartilage In gnathostomes, the SoxE transcription factor, Sox9, is a marker of CNCC and pre-chondrocytes [56]. Gnathostome Col2a1 is a molecular marker for nascent and differentiated cellular cartilage that is directly regulated by Sox9 [26,57,58,59]. Alcian blue binds the chondroitin sulfate proteoglycans (CSPGs, i.e. Aggrecan) secreted by differentiated cellular cartilage, and is considered diagnostic for this tissue [60]. Together, the combined expression of these molecules is used to monitor the development of cellular cartilage in gnathostomes. To obtain a more complete view of lamprey cartilage development, we examined SoxE expression, type A fibrillar collagen expression, and alcian blue reactivity at several time points during chondrogenesis in lamprey ( Figure 1).
We observed a lack of SoxE1 expression at st. 25, and strong expression in the forming branchial bars at st. 26.5, indicating that SoxE1 is temporarily down-regulated sometime between st. 23 and st. 25 and reactivated at st. 26 ( Figure 1A-G). At st. 28, expression of SoxE1 in the branchial basket was reduced ( Figure 1C,H), particularly in the middle of the vertical branchial bars where cartilage differentiation is first detectable by alcian blue staining [37]. Also at st. 28, we observed broad staining throughout the pharynx including the mesenchyme of the ventral pharynx and oral region. This staining continued until st. 30, when SoxE1 expression in the branchial bar cartilage was undetectable ( Figure 1D).
The loss of SoxE1 in the branchial bar chondrocytes at st. 30 was coincident with their differentiation as revealed by alcian blue staining ( Figure 1I). At 35 days post fertilization (approximately 13 days after st. 30), alcian blue staining revealed the deposition of CSPGs in the fully differentiated branchial basket as well as in the mucocartilage of the oral region (upper lip, first pharyngeal arch, lower lip) and ventral pharynx ( Figure 1E, J).
Expression of lamprey fibrillar collagen in the forming branchial basket at select time points (st. 25, st. 26) has been described [38,39,62]. We replicated and expanded upon these studies, looking at Col2a1a expression at key stages before, during, and after chondrogenesis. As previously reported, we observed broad Col2a1a expression in post-migratory cranial neural crest cells and somitic mesoderm at st. 25 ( Figure 1K). While SoxE1 expression is temporarily downregulated at this stage, cranial neural crest expression of SoxE1 and SoxE2 is seen earlier at st. 23 and st. 24, consistent with Col2a1 regulation by SoxE in CNCC as seen in gnathostomes. At st. 26.5, strong Col2a1a expression is seen in presumptive mucocartilage in the ventral pharynx, upper and lower lips, first pharyngeal arch, and in dorsal fin mesenchyme ( Figure 1L,M). Strong Col2a1a expression is also observed in a horizontal band of cells dorsal to the pharyngeal arches at the position of the nascent subchordal cartilage bars. Negligible expression was seen in the forming branchial bars at this stage. At st. 28 strong expression of Col2a1a was detected throughout the mucocartilage around the mouth, with weaker expression in branchial basket cartilage ( Figure 1N). Similar expression was seen later at 35 d.p.f. ( Figure 1O).
Our results show SoxE1 expression is high in post-migratory pharyngeal CNCC, where it is co-expressed with Col2a1, and is then downregulated as these cells differentiate into cellular cartilage. This sequence mirrors the down-regulation of Sox9 seen in differentiating gnathostome cartilages and supports conserved roles for SoxE in lamprey branchial basket cartilage and gnathostome cellular cartilage development as previously proposed [40,47].
In addition to the branchial basket, however, we also observed SoxE1 expression and alcian blue staining throughout the developing mucocartilage of the oral apparatus and ventral pharynx. Furthermore, we found that these tissues express high levels of fibrillar collagen [38,39] as originally suggested by Schaffer [51] and Johnels [42]. Due to its staining properties and mesenchymal morphology, mucocartilage is generally considered a derived form of connective tissue unrelated to cellular cartilage [55]. However, co-expression of fibrillar collagen, SoxE, and CSPGs in mucocartilage suggest some developmental and/or evolutionary relationship between mucocartilage and cellular cartilage. Hardisty [44] proposed that mucocartilage may represent an undifferentiated embryonic tissue that is carried over into the larval stage. An alternate hypothesis is that mucocartilage represents an evolutionary precursor to bona fide cellular cartilage. We decided to further investigate these possibilities by looking at the expression of three other key regulators of vertebrate chondrogenesis; Runx, Barx, and Alx/Cart1, in the lamprey head skeleton.
Runx genes are expressed in lamprey mucocartilage, but not in the branchial basket cartilage Runx genes are key regulators of both cartilage and bone development in jawed vertebrates. During embryogenesis all three gnathostome Runx paralogs, Runx1,2,3 [18] are expressed in presumptive chondrocytes, and Runx2 and Runx3 have been shown to be required for their differentiation into cellular cartilage [10,20,21]. After chondrogenesis, Runx genes regulate the formation of replacement cartilage and the ossification of endochondral bone [63]. Functional studies suggest Runx genes are regulated by both Sox9 and Barx genes and are downstream of these factors in the cartilage GRN [9,64].
We performed an exhaustive search of P. marinus pre-assembly genome and identified several contigs with similarity to the four most highly conserved deuterostome Runx exons. For three of these exons, we found two distinct aligning lamprey sequences encoding different proteins, suggesting the presence of two lamprey Runx genes ( Figure S1, S2). While it is possible that the preassembly P. marinus genome does not include all Runx genes, the level of coverage, and the presence of two Runx genes in hagfish, a related agnathan, suggest this is the full complement of lamprey Runx genes. These genes were designated RunxA and RunxB. RunxA expression was first observed at Tahara st. 25 in the presumptive cranial ganglia and mesenchyme flanking the mouth (Figure 2A). By st. 26.5, the onset of chondrogenesis, expression was seen in the pharyngeal endoderm, cranial ganglia, lateral oral mesenchyme, and intermediate first arch (Figure 2B,E,F). Spots of expression were also observed in the mesodermal core of the pharyngeal arches ( Figure 2E). Expression in the cranial ganglia, upper lip, first arch, pharyngeal endoderm and pharyngeal mesoderm was still detectable at st. 27, when cartilage is histologically identifiable by alcian blue staining ( Figure 2C,G). By st. 30, when most of the branchial basket cartilage has differentiated, RunxA was still detectable in the pharyngeal endoderm and cranial ganglia ( Figure 2D,H). RunxB deployment was much less extensive at early larval stages, marking bilateral spots in the anterior neural tube and cranial ganglia at st. 25 and 26.5 ( Figure 2I,J). At st. 27 additional expression was observed in the dorsal fin mesenchyme, around the heart, and in the endostyle ( Figure 2K). At st. 30 strong expression was observed in the dorsal fin and around the heart, with reduced expression in the brain ( Figure 2L). By 35 d.p.f. weak expression of both RunxA an RunxB was seen in the pharyngeal endoderm. ( Figure S3).
We then examined Runx, SoxE1, and Col2a1a expression in premetamorphic ammocoete larvae and recently metamorphosed juvenile lampreys, to determine if these genes are activated in cellular cartilage around the time of metamorphosis. While we did see expression of Col2a1a in the branchial bars ( Figure 3B), we observed no specific signal in any pharyngeal tissue at these stages with either the RunxA (not shown) or RunxB riboprobes ( Figure 3G). Similarly, SoxE1 expression above background levels was not seen in pre-or post-metamorphic branchial basket cartilage, though specific expression was seen in the gills, notochord, and a subset of cells in the spinal cord ( Figure 3C,H).
Taken together, we detected no lamprey RunxA or RunxB expression in branchial basket cartilage at any stage, either before or after metamorphosis. While it is formally possible that an unidentified lamprey Runx paralog is expressed in lamprey branchial basket cartilage, we view this as unlikely. The apparent lack of Runx expression in lamprey cartilage contrasts with gnathostomes where Sox9 and all three Runx genes are broadly co-expressed and interact extensively to regulate chondrogenesis. Assuming RunxA and RunxB represent the full complement of lamprey Runx genes, this difference implies either that Runx lost its ancestral function in cartilage development in the lamprey lineage, or that Runx genes were not part of the ancestral vertebrate cartilage GRN.
Recent work using qPCR on adult tissue suggests enrichment of Runx transcripts in the cartilage of adult hagfish, leading to speculation that Runx genes drive skeletogenesis in these agnathans [65]. Hagfish have historically been considered basal jawless craniates, but modern molecular phylogenies support their grouping with lampreys in a single agnathan clade [66]. If Runx is in fact required for chondrogenesis in hagfish, it would support the lineage-specific loss of Runx from the lamprey cartilage GRN.
Alternately, Runx expression in adult hagfish could reflect a function in adult tissue maintenance or metabolism unrelated to cartilage specification during development. In jawed vertebrates, Runx genes perform a wide range of functions in embryonic and adult tissues. Gnathostome Runx genes are downregulated in differentiated adult cartilage and upregulated in a variety of other tissues including skin, liver, intestine, thyroid and blood where they regulate cell division and stem cell quiescence [19,67,68,69]. Consistent with a general role in adult tissue maintenance, hagfish Runx transcripts were found in every hagfish tissue examined, with slightly higher levels in hard cartilage [65]. In either case, the lack of detectable enrichment of lamprey Runx mRNA in branchial basket cartilage indicates Runx genes are dispensable for the development of histologically discernable cellular cartilage. We propose this reflects the basal vertebrate state, but it could formally represent the presence of a derived chondrogenic GRN specific to lampreys. Runx expression during cartilage development in hagfish embryos would help distinguish between these two scenarios.
While lamprey Runx expression was not observed in the definitive cellular cartilage of the branchial basket, strong, localized RunxA expression was observed in subpopulations of nascent mucocartilage around the mouth and in the intermediate portion of the first arch. This restricted expression suggests lamprey Runx genes may be performing tissue-specific functions in the development of particular kinds of mucocartilage, rather than acting as general drivers of skeletogenesis as they are in gnathostomes.
Other domains of Runx expression included the brain, cranial ganglia, and pharyngeal endoderm in early larvae (st. 25-26.5) and the heart, and dorsal fin mesenchyme in late larvae (st. [27][28][29][30]. Expression of Runx genes is observed in the cranial ganglia of all gnathostomes examined, indicating a deeply conserved role for Runx in the development of these structures [18,21]. Expression in the dorsal fin may be related to a skeletogenic function as these cells also express SoxD and fibrillar collagen (Figure 1) [38], and react with with alcian blue ( Figure 1E).
An endodermal specification function for Runx genes appears to be evolutionarily ancient as gnathostomes, amphioxus, sea urchin, and C. elegans all express Runx in the nascent gut, and Runx is required for gut formation in C. elegans [7,67,70]. Runx expression in the pharyngeal endoderm of adult amphioxus has been interpreted as evidence of a rudimentary skeletogenic program in this tissue [65]. Given the pleiotropic nature of Runx function, and the pan-metazoan expression of Runx genes in endoderm, it seems equally likely that Runx genes are performing some other function in this tissue.
Lamprey Barx is expressed in lower lip mucocartilage and CNCC in the medial pharyngeal arches, but not in developing branchial bar cartilage Recent work has shown that Barx is required for the differentiation of CNCC-derived cartilage in the zebrafish head where it is downstream of FGF signaling and regulates Runx expression [9]. Similar broad expression of Barx in the CNCC of chick and mouse suggest this role is conserved among all gnathostomes [17,71]. In addition, functional studies in mouse have demonstrated an essential role for Barx in the chondrogenesis of the mesoderm-derived appendicular skeleton [27].
We isolated a lamprey Barx ortholog, and described its expression in the first pharyngeal arch at st. 26.5 in a previous study [72]. Here we detail lamprey Barx expression from embryonic through late larval stages in all tissues. Barx transcripts are first weakly detectable in a stream of CNCCs migrating into the region of the first pharyngeal arch at st. 24 (data not shown). At st. 25, Barx expression is seen in the lower lip and the forming pharyngeal arches (Figure 4A). At st. 26.5 and 27, strong Barx expression is seen in the lower lip, extending dorsally into the intermediate region of the first pharyngeal arch as previously described ( Figure 4B,F,C,G). Expression is also seen in presumptive CNCC-derived mesenchyme positioned medial to the pharyngeal arch mesoderm in the posterior arches ( Figure 4E). These domains of expression persist into st. 30 when chondrogenesis of the branchial bars is complete ( Figure 4D,H). Similar expression is still apparent at 35 d.p.f ( Figure S3). No Barx expression was seen in the pharynx of pre-metamorphic ammocoete larvae ( Figure 3E) or post-metamorphic juveniles (not shown).
As with Runx, we noted an absence of Barx expression in the laterally-positioned pharyngeal CNCCs which give rise to the definitive cellular cartilage of the gill bars. Thus, like Runx, Barx is likely not directly involved in the development of lamprey cellular cartilage, whereas it is required for definitive cartilage formation in gnathostomes.
Although Barx expression did not mark CNCC in the lateral pharynx, we did observe Barx expression in CNCC medial to the mesodermal core of each pharyngeal arch. These cells express other markers of CNCC including tfap2, Dlx, and SoxE, as well as fibrillar collagen [39,72]. However, they do not react with alcian blue or correspond to histologically recognizable mucocartilage. In lamprey, this region of the pharyngeal arches gives rise to the smooth muscle of the branchial veins and arteries [43], tissue which is CNCC-derived in gnathostomes. We speculate that lamprey Barx expression in the medial portion of the pharyngeal arches marks CNCC fated to form the branchial vasculature.

Lamprey Alx marks upper lip mesenchyme and portions of the branchial basket cartilage
Gnathostomes possess three Alx paralogs, Alx3, Alx4, and Cart1. The expression and function of these factors during cartilage development have been studied almost exclusively in mouse, where all three paralogs mark chondrogenic mesenchyme in the frontonasal mass, pharyngeal arches, and limb buds [25,73]. Consistent with their expression patterns, double knock-outs of Alx3 and Alx4 result in severe reductions in the craniofacial and appendicular skeletons [22]. Among invertebrate deuterostomes, Alx is required for skeletal development in sea urchin [74], and amphioxus Alx is expressed in pharyngeal mesoderm thought to secrete acellular cartilage [7], suggesting an ancient role for Alx genes in skeletogenesis.
We found multiple contigs corresponding to a single Alx gene in the P. marinus genome ( Figure S1, S2), suggesting it is the only lamprey Alx homolog. We then examined its expression in embryos and larvae during cartilage development. Lamprey Alx expression was first seen at st. 23.5 in a spot of mesenchyme in the forming upper lip (data not shown). This expression expanded and intensified to fill the mesenchyme of the upper lip at st. 25 ( Figure 5A). Additional expression was seen in the dorsal and ventral aspects of the forming pharyngeal arches and in the forming dorsal fin. By st. 26.5, expression was seen in the mesenchyme of the dorsal and ventral pharyngeal arches, upper lip, and dorsal fin ( Figure 5B,F). Expression was excluded from the lower lip mesenchyme and the intermediate portion of the pharyngeal arches. This expression resolved into distinct dorsal and ventral spots in each arch at st. 27 ( Figure 5C,G). By st. 30, Alx expression in the pharynx had diminished, but was still detectable in the first arch and subchordal chondrocytes ( Figure 5D,H). This expression pattern persisted until 35 d.p.f. (Figure S3), though no Alx expression was detectable in the pharynx of ammocoetes ( Figure 3D) or metamorphosed juveniles (not shown).
In gnathostomes, SoxE and Alx genes are co-expressed in CNCCderived chondrogenic mesenchyme [25]. While no regulatory relationship between the two factors has been demonstrated, Sox9 and Alx3/4 double mutant mice have similar hypoplastic skeletal phenotypes, suggesting they operate within the same cartilage  GRN [22,75]. Unlike their gnathostome counterparts, lamprey SoxE and Alx genes are not broadly co-expressed in nascent cellular cartilage. While lamprey SoxE marks the entire branchial basket cartilage, lamprey Alx is restricted to the dorsal and ventral aspects, and is excluded from the vertical branchial bar cartilage. Recent work has shown that the lamprey branchial basket consists of morphologically and molecularly distinct cell types arranged along its dorso-ventral (DV) axis [37,72]. In the central portion of the pharyngeal gill bars, vertically stacked branchial bar chondrocytes display a highly ordered discoidal ''stack-of-coins'' morphology. Dorsally, subchordal, parachordal and trabecular chondrocytes are disorganized and polygonal in shape. Ventrally, the horizontal hypobranchial bars have a semi-ordered pseudo-discoidal morphology. Alx expression in the pharynx corresponds to the forming subchordal and hypobranchial chondrocytes. Both subchordal and hypobranchial chondrocytes likely serve as rigid structural elements, in contrast to the central discoidal chondrocytes, which form flexible bars that bend and recoil during pharyngeal pumping. It is possible that Alx expression in the lamprey branchial basket identifies a particular kind of rigid structural cartilage homologous to the cellular cartilage that comprises the bulk of the gnathostome head skeleton.
Similarly, Alx expression in the upper lip mesenchyme suggests Alx may also specify a type of mucocartilage. Historical descriptions classify the skeletal tissue in the upper lip, lower lip, first pharyngeal arch, and ventral pharynx as a single kind of generic mucocartilage [42,43,51]. However, at the stages we examined, the skeletal tissue around the larval lamprey mouth appears compact, while the presumptive mucocartilage of the ventral pharynx forms a loose mesenchyme. Alx expression in the upper lip may confer some unique physical properties that distinguish this skeletal tissue from the loose mucocartilage in the ventral pharynx.
Like Runx, and Col2a1a, we also observed Alx transcripts in the dorsal fin mesenchyme, a tissue that reacts with alcian blue (Figure 1) and is at least partially neural crest-derived [45]. Coexpression of Runx and Alx is not seen in any skeletal tissue in the head, suggesting lamprey dorsal fin mesenchyme may represent a mucocartilage-like tissue unique to the trunk.
Summary of SoxE1, Col2a1a, CSPG, Runx, Barx, and Alx expression in lamprey skeletal tissues In sum, our results show expression of SoxE1, Col2a1a, and staining with alcian blue (a proxy for CSPG expression) in all mucocartilage and cellular cartilage of the lamprey head at early larval stages (st. 26.5-st. 30). During this period, RunxA expression was seen in a small subpopulation of mucocartilage flanking the mouth, while RunxB was observed in dorsal fin mesenchyme. Barx expression was observed in a portion of the mucocartilage in the lower lip/first arch, while Alx expression was detected in upper lip mucocartilage, two subpopulations of cellular cartilage, and dorsal fin mesenchyme. No significant overlap of RunxA, RunxB, Barx or Alx expression was seen cellular cartilage ( Figure 6), or any other skeletal tissue except in the intermediate first arch (Barx and Runx) the dorsal fin mesenchyme (RunxB and Alx) (Figure 7). However, RunxA, Barx and Alx were all expressed with SoxE1 and Col2a1a in various skeletal tissues and at some point between st. 26.5 and st. 30. In pre-metamorphic ammocoete larvae and post-metamorphic juvenile lampreys, only expression of Col2a1a was detectable above background levels in branchial basket cartilage.

A hypothetical scenario for the origin and evolution of vertebrate cartilage
We are taking a molecular approach to examine possible evolutionary relationships between invertebrate skeletal tissue, gnathostome cartilage, and the skeletal tissues of lamprey. Our data show that, despite similarities with gnathostome cellular cartilage, the cellular cartilage of the lamprey branchial basket deploys a rudimentary differentiation program that does not incorporate Runx or Barx. Furthermore, we find that the flexible ''stack of coins'' chondrocytes of the branchial basket do not express Alx genes and are thus molecularly distinct from the rigid polygonal subchordal and hypobranchial chondrocytes ( Figure 6). Finally, we find that lamprey mucocartilages, though morphologically very different from gnathostome cellular cartilage, express various combinations of genes involved in gnathostome chondrogenesis (Figures 6, 7).
Taken together, our data reveal an array of histologically and molecularly distinct skeletal tissue types in the larval lamprey head, each displaying a subset of the features seen in gnathostome cellular cartilage (Figure 7). This contrasts with the gnathostome condition in which all embryonic cellular cartilage appears to deploy a single tightly conserved chondrogenic GRN. Assuming that the lamprey head skeleton approximates the basal vertebrate condition, our results have two significant evolutionary implications. First, they suggest that the evolution of the gnathostome head skeleton involved a marked reduction in skeletal tissue diversity. This implies that the basal vertebrate state is the possession of multiple cartilage-like tissue types, while gnathostomes display a derived dependence on a single kind of cellular cartilage. Second, our results suggest that this loss of skeletal tissue Figure 7. The diversity of skeletal tissues in the larval lamprey head. Based on morphology and gene expression, we speculate that lamprey larvae may possess as many as 11 distinct types of skeletal tissue. This includes 3 kinds of cellular cartilage, 7 kinds of mucocartilage, and a mucocartilage-like skeletal tissue unique to the dorsal fin. The diagram shows the first 6 pharyngeal arches. The upper lip mesenchyme is derived from the premandibular cranial neural crest (pm). For reference, cellular morphology and gene expression in gnathostome cellular cartilage and joint tissue are shown in the shaded rows. doi:10.1371/journal.pone.0022474.g007 diversity was coincident with the consolidation of several rudimentary skeletogenic GRNs into a single chondrogenic GRN. Thus, during gnathostome evolution, gene programs operating primitively in different types of skeletal tissue became co-expressed in a single type of cellular cartilage.
We previously proposed that genetic cooption of mesodermal gene networks by the evolving cranial neural crest drove the origin of vertebrate cellular cartilage [7]. Our current results suggest similar cooption events occurred during the evolution of gnathostome cellular cartilage. Below we integrate our current and previous data and propose a stepwise mechanistic scenario for the origin and evolution of vertebrate cellular cartilage (Figure 8).
Step 1: The origin of skeletogenic CNCC Our current results, and previous work, reveal a set of core features shared by all skeletogenic neural crest cells, including gnathostome cellular cartilage, lamprey cellular cartilage, and lamprey mucocartilage. These include; 1) differentiation from a mesenchymal precursor cell, 2) persistent expression of Twist, Ets, Id, and SoxE after migration into the pharynx [39,47], 3) secretion of CSPGs [37], and 4) expression of fibrillar collagen. Notably, this list does not include expression of Runx or Barx, which our data show are not general markers of neural crest-derived skeletal tissue.
Using these criteria as a minimal definition of vertebrate neural crest-derived skeletal tissue, we can then ask if any protochordate tissue possesses all, or most, of these features. As has been shown previously, Twist, Ets, Id, Alx, and fibrillar collagen are coexpressed in the pharyngeal mesoderm of amphioxus larvae [7,50]. Furthermore, this tissue displays mesenchyme-like cell morphology and may give rise to acellular cartilage bars composed of CSPGs [7,50]. Thus, skeletogenic mesoderm with most attributes of skeletogenic neural crest cells, but lacking SoxE expression, was likely present in the first chordates.
The existence of migratory neuroblasts in a range of invertebrates [76,77,78] suggests that the vertebrate ancestor also possessed some kind of migratory neural tube cell. Assuming that co-expression of Twist, Ets, Id, Alx, and fibrillar collagen reflects a conserved GRN, it is possible that skeletogenic neural crest cells evolved via activation of this GRN in non-skeletogenic migratory neural tube cells. The lack of SoxE expression in amphioxus pharyngeal mesoderm further suggests this cooption involved the integration of SoxE into the evolving skeletogenic GRN. How these changes occurred is unclear. Aside from their function in skeletogenesis, vertebrate SoxE paralogs are required for the specification and differentiation of all neural crest cell lineages, as well as for the development of glia and the otic placode [79]. In addition, amphioxus SoxE is expressed throughout the nascent nervous system in early larvae, but not in any obviously skeletogenic tissues aside from a few cells in the notochord [80]. These data imply that the ancestral function of chordate SoxE was in neural development, rather than skeletogenesis, as has been proposed [65,81]. Furthermore, these results suggest that migratory neural tube cells could have acquired SoxE simply by maintaining SoxE expression after leaving the CNS. Based on these observations, we propose that skeletogenic neural crest cells evolved by activation of a mesodermal skeletogenic GRN in migratory neural tube cells expressing SoxE. (Figure 8A,B). A role for SoxE genes in skeletogenesis then evolved later in the vertebrate lineage as SoxE was integrated into this skeletogenic GRN.
The cartilage GRN in vertebrates is induced and maintained by signals secreted by the pharyngeal endoderm and ectoderm, including Hedgehogs and FGFs [82,83]. Provocatively, FGF and Hedgehog ligands are expressed in the amphioxus pharynx at larval stages when skeletogenesis is likely initiated [7,84]. We further speculate that the activation of mesodermal skeletogenic gene programs could have evolved as proto-CNCC cells gained responsiveness to these signals.
Based on Runx and SoxE co-expression in the foregut of adult amphioxus, it has been proposed that cooption of a endodermal skeletogenic GRN(s) incorporating these factors drove the appearance of skeletogenic CNCC [65,81]. However, expression of lamprey SoxE1 and Runx in non-skeletogenic pharyngeal endoderm and gills (Figure 2, Figure 3B) suggests these factors perform non-skeletogenic functions in chordate endoderm. Furthermore, our data showing that Runx is not expressed in lamprey cellular cartilage implies it is not a core feature of chordate skeletal development. Finally, the early role of SoxE genes in the initial formation of all neural crest cells, suggests its ancestral function was in neural crest cell specification rather than chondrogenesis [79]. Thus, cooption of Runx and SoxE from endoderm was likely not involved in the origin skeletogenic neural crest cells. Instead, our data support later recruitment of Runx by SoxE-expressing skeletogenic neural crest cells in the gnathostome lineage (see below).
Step 2: The diversification of CNCC-derived skeletal tissues in early jawless vertebrates The ''New Head'' hypothesis suggests that the original function of vertebrate cartilage was to provide structural support for pharyngeal pumping [1]. The concurrent expansion of the CNS, anterior sense organs, and cranial ganglia also likely required additional skeletal support, as did the novel structural demands on the oral region of early vertebrate predators. In the first vertebrates, we posit that these various functions were performed by a generic CNCC-derived skeletal tissue expressing Twist, Id, Ets, Alx, SoxE, and fibrillar collagen, and secreting CSPGs ( Figure 8B). Based on gene expression in lamprey, we further speculate that this generic CNCC-derived head skeleton then diversified, with different regions taking on specialized properties in response to particular structural demands. During this phase of vertebrate skeletal evolution, Runx, and Barx, were recruited to different parts of the CNCC-derived skeleton, and Alx, which was initially expressed in all skeletogenic CNCC, was lost from some regions. These changes in gene expression drove the establishment of several different skeletal tissue types with distinct physical properties ( Figure 8C). Among these were compact tissues composed of cells with polygonal and discoidal morphologies that would be histologically identifiable as proper ''cellular cartilage'', and several types of mesenchymal skeletal tissues in the head and dorsal fin. This primitive diversity of CNCC-derived skeletal cell types is thus preserved in lamprey, where Runx, Barx, and Alx genes mark different subsets of cartilage and mucocartilage around the mouth and in the pharynx (Figure 7). Localized expression of these factors would also require some level of underlying anteroposterior (AP) and dorsovental (DV) patterning in the oro-pharyngeal region. Consistent with this, recent work suggests sophisticated gnathostome-like AP and DV patterning systems were in place before the divergence of jawed and jawless vertebrates [72,85].
Step 3: The evolution of gnathostome cartilage While chondrogenic gene expression in lamprey identifies several molecularly distinct skeletal tissues (Figure 7), the ubiquitous coexpression of chrondrogenic GRN components in the gnathostome head skeleton suggests the presence of only one major cartilage type. If lamprey represents the basal state, then the evolution of the gnathostome head skeleton was associated with a reduction in skeletal tissue diversity. A hallmark of the lamprey head skeleton is the preponderance of soft, flexible skeletal tissues such as the mucocartilage of the oral region and ventral pharynx, and the discoidal ''soft'' cartilage of the branchial basket. This contrasts with the gnathostome condition, where uniformly rigid cartilage elements articulate at foci of soft joint tissue. Rigid ''hard'' cartilages have been described in lamprey based on their staining properties [42,43,51]. Interestingly, these cartilages have the polygonal morphology typical of most gnathostome cartilages and express Alx. We speculate that a similar type of specialized, polygonal, Alx-positive cartilage was present in the first vertebrates where it constituted a small portion of the head skeleton. In the gnathostome lineage, this rigid cartilage proliferated, becoming the dominant component of the head skeleton ( Figure 8D). The broad expression of Runx and Barx in the gnathostome head skeleton suggests this transition also involved the expanded expression of these factors. We posit that Runx and Barx may have initially conferred novel structural properties upon this tissue. The transition to rigid structural cartilage may have been driven by the novel physical requirements of holding and processing increasingly large prey items and supporting and protecting a large CNS and sense organs. Similar selective pressures may have driven the subsequent evolution of bone.
In addition to rigid cellular cartilage, gnathostomes also possess soft joint tissue between cartilaginous elements in the head. Like cellular cartilage, these cells are CNCC-derived and initially express SoxE, Twist, Ets, and Id and stain with alcian blue [86]. However, unlike, cellular cartilage, this tissue remains mesenchymal and does not express Barx and Runx [9,60]. Based on Figure 8. A hypothetical scenario for the origin and evolution of vertebrate cartilage. A) Based on the expression of SoxE genes throughout the neural tube of amphioxus larvae, and the presence of migratory neural cells in urochordates and other invertebrates, we posit that the ancestral pre-vertebrate chordate possessed migratory neural tube cells that expressed SoxE genes (light blue). These cells migrated into the pharynx and generated neural and/or pigment cells, but were not skeletogenic. The pharyngeal skeleton at this stage (pink) resembled that of amphioxus, consisting of acellular cartilage incorporating acid mucopolysaccharides and secreted by pharyngeal mesoderm expressing Twist, Ets, Id, Alx, CSPGs and fibrillar collagen. B) The origin of bona fide skeletogenic neural crest, and the first craniates, was marked by the activation of this mesodermal skeletogenic gene program in SoxE-expressing migratory neural tube cells (light blue). This ancestral form of cellular cartilage provided structural support for the evolving head. C) In the lineage leading to the common ancestor of modern jawed and jawless vertebrates, this generic skeletal tissue diversified into several subtypes with distinct molecular and structural properties via cooption of Runx and Barx and the selective loss of Alx expression. This included definitive ''cellular cartilage'' similar to that seen in gnathostomes and multiple mucocartilage subtypes as seen in lamprey. This basal condition is maintained in lamprey. D) In gnathostomes this heterogenous head skeleton was replaced with a skeleton consisting mainly of a single type of rigid cellular cartilage. Coincident with this shift was the consolidation of SoxE1, Runx, Barx, and Alx genes into a single gnathostome cellular cartilage gene network. Based on similar gene expression and cell morphology, the joint tissue seen in gnathostomes (light blue) may represent a basal form of skeletal tissue similar to mucocartilage in the ventral pharynx of lamprey. doi:10.1371/journal.pone.0022474.g008 morphology and Gdf5/6/7 expression we previously proposed an evolutionary relationship between gnathostome joint tissue and lamprey mucocartilage in the ventral pharynx [72]. Our current results showing that mucocartilage in the ventral pharynx does not express Runx and Barx lend support this hypothesis. We speculate that gnathostome joints may have evolved by the redeployment of a primitive mucocartilage-like skeletal tissue to the regions between rigid cellular cartilage condensations ( Figure 8D).

Ethics Statement
All methods were reviewed and approved by the University of Colorado, Boulder IACUC protocol 08-07-MED-02.
We performed an exhaustive search of the 5.9X coverage, preassembly Petromyzon marinus genome for homologs of gnathostome Runx, Barx and Alx genes by repeated BLAST [87] searching with gnathostome, amphioxus, and sea urchin protein sequences. We then designed exact-match PCR primers for the two Runx paralogs (RunxA and RunxB), and the single Barx and Alx genes. Runx and Barx exons were PCR amplified from adult lamprey genomic DNA according to standard methods. Alx was isolated from embryonic cDNA using the GeneRacer kit (Invitrogen). A SoxE1 fragment corresponding to nucleotides 211-619 of the published SoxE1 nucleotide sequence [47] was amplified from embryonic cDNA for use as a riboprobe template.
Embryos and early larvae for in situ hybridization were obtained from adult spawning phase sea lampreys (Petromyzon marinus) as previously described [88]. 5-10 cm pre-metamorphic ammocoete larvae, and newly metamorphosed juvenile lamprey were collected from streams feeding Lake Huron in the Fall with a backpack electroshocker and kept in chilled holding tanks until fixed in MEMFA (MOPS buffer, EGTA, MgSO 4 , Formaldehyde). Thick cross sections of 100-200 microns were then cut through the pharynx using a vibratome.