Evidence for the prepattern/cooption model of vertebrate jaw evolution

Abstract

The appearance of jaws was a turning point in vertebrate evolution because it allowed primitive vertebrates to capture and process large, motile prey. The vertebrate jaw consists of separate dorsal and ventral skeletal elements connected by a joint. How this structure evolved from the unjointed gill bar of a jawless ancestor is an unresolved question in vertebrate evolution. To understand the developmental bases of this evolutionary transition, we examined the expression of 12 genes involved in vertebrate pharyngeal patterning in the modern jawless fish lamprey. We find nested expression of Dlx genes, as well as combinatorial expression of Msx, Hand and Gsc genes along the dorso-ventral (DV) axis of the lamprey pharynx, indicating gnathostome-type pharyngeal patterning evolved before the appearance of the jaw. In addition, we find that Bapx and Gdf5/6/7, key regulators of joint formation in gnathostomes, are not expressed in the lamprey first arch, whereas Barx, which is absent from the intermediate first arch in gnathostomes, marks this domain in lamprey. Taken together, these data support a new scenario for jaw evolution in which incorporation of Bapx and Gdf5/6/7 into a preexisting DV patterning program drove the evolution of the jaw by altering the identity of intermediate first-arch chondrocytes. We present this “Pre-pattern/Cooption” model as an alternative to current models linking the evolution of the jaw to the de novo appearance of sophisticated pharyngeal DV patterning.

The evolution of vertebrates from invertebrate chordates involved a shift from a lifestyle of passive filter feeding to one of active predation (1). This transition was marked by major changes to the structure of the pharynx, including the evolution of a robust pharyngeal skeleton composed of cellular cartilage. In the first vertebrates, this skeleton was simple, consisting of unjointed cartilaginous rods. In the lineage leading to gnathostome (jawed) vertebrates, these rods were modified to form distinct cartilages connected by specialized joint tissue. This arrangement is most conspicuous in the first arch, where articulating dorsal and ventral cartilages form the jaw. How the unjointed gill bars of the first vertebrates gave rise to the jointed gnathostome pharyngeal skeleton and jaws is an outstanding question in vertebrate evolution.

The gnathostome pharyngeal skeleton is derived exclusively from cranial neural crest cells (CNCCs) that migrate ventrally from the anterior neural tube into the pharynx. Once in the pharynx, these cells receive a variety of signals from surrounding tissues that regulate CNCC proliferation and pattern the nascent head skeleton (2, 3). Along the dorso-ventral (DV) axis, these signals drive the expression of several target genes, including the transcription factors Hand, Msx, Dlx2,3,5,6, Goosecoid (Gsc), Barx, and Bapx, and the secreted proteins Gdf5 and Chordin. Hand and Msx are both expressed in nascent ventral cartilages (4, 5), whereas Dlx genes display nested expression patterns, with Dlx3 expressed in an intermediate domain, Dlx5,6 expression extending more dorsally, and Dlx2 ubiquitously marking all DV domains (6–8). Gsc and Barx are expressed in discontiguous ventral and dorsal domains flanking the nascent joints in the first and second arches, whereas Bapx, Gdf5, and Chordin mark the nascent jaw joint in the intermediate domain of the first arch (5, 9).

Genetic perturbations in mouse, chick, and zebrafish have demonstrated that these genes interact with each other to determine the composition and shape of skeletal elements. In ventral arch skeletal primordia, a loss of Hand results in a loss of ventral arch cartilages, a loss of ventral Gsc expression, and a ventral expansion of Bapx (5). Loss of Bapx results in loss of Gdf and Chordin expression and elimination of the jaw joint itself (5), whereas combinatorial losses of Dlx3,5, and 6 result in various homeotic transformations in which ventral cartilages take on more dorsal identities (7, 8, 10).

Although many of the core genetic interactions that pattern the gnathostome pharyngeal skeleton and jaw have been described, it is unknown when these mechanisms first evolved in the vertebrate lineage or how they may have been altered during jaw evolution. An evolutionarily informative model for understanding the genetic changes underlying jaw evolution is the agnathan vertebrate lamprey. Lampreys are the most basal vertebrates amenable to experimental manipulations at embryonic stages. Like all vertebrates, lamprey has a pharyngeal skeleton made wholly or partially of neural crest-derived cellular cartilage (11–13). However, unlike jawed vertebrates, the lamprey pharyngeal skeleton consists of cartilage bars that fuse to form an unjointed branchial basket (Fig. S1). This unjointed arrangement is similar to that seen in the fossil agnathans (14, 15), and is thought to reflect the ancestral vertebrate condition.

Recent work suggests that the pharyngeal skeletons of agnathans and gnathostomes are patterned along the antero-posterior (AP) axis by largely conserved mechanisms. In both groups, cranial neural crest cells (CNCCs) destined to form the head skeleton migrate as three streams (the trigeminal, hyoid, and common branchial) at stereotyped AP positions (16, 17). Lamprey also displays gnathostome-like AP nesting of Hox expression in the pharynx, including a Hox-free first arch, a Hox2-expressing second arch, and Hox2-8-positive respiratory arches (18). Finally, in both lamprey and gnathostomes, the first arch and premandibular skeletons appear to be patterned by foci of FGF8 and BMP signaling (19).

Although the core mechanisms for patterning the pharynx along the AP axis are conserved between jawed and jawless vertebrates, morphology and gene expression suggest differences in DV patterning between the two groups. In larval and adult lampreys, the branchial basket is virtually symmetrical along the DV axis, with the morphology of its dorsal half mirroring that of its ventral half (19). In embryos, expression of lamprey Dlx genes has been reported throughout the DV extent of the pharyngeal arches (20–22), rather than in the nested pattern seen in gnathostomes (7). Intriguingly, loss of Dlx nesting in mouse results in agnathan-like DV symmetry in the first-arch skeleton (7, 10). These observations have led to speculation that the lamprey pharynx lacks significant DV patterning and that the evolution of such patterning was instrumental in the evolution of the jaw (7, 19, 20, 22).

Here we test the hypothesis that jaw evolution was driven by the evolution of sophisticated pharyngeal DV pattern, by examining the expression of lamprey homologs of gnathostome pharyngeal patterning genes. We observe nested Dlx gene expression in the lamprey head skeleton, as well as DV-restricted expression of Hand, Msx, and Gsc, revealing an unexpectedly high level of gnathostome-like DV pattern. In addition, we find that Bapx, a master regulator of jaw joint formation in gnathostomes, and Gdf5, its downstream target, are not expressed in the lamprey intermediate first arch, whereas Barx, which is excluded from the intermediate first arch in gnathostomes, marks this domain in lamprey. Taken together, these data support a model for gnathostome evolution in which the appearance of the jaw joint was driven by incorporation of Bapx and Gdf5 into a sophisticated ancestral DV-patterning program. Based on gene expression in zebrafish mutants lacking jaw joints, and cell morphology in the lamprey head skeleton, we further propose that these cooptions potentiated jaw joint evolution by altering the fate of first-arch chondrocytes.

Results

Cloning and Expression of Lamprey MsxA and MsxB.

A low-stringency library screen for Msx genes was performed using an arrayed embryonic Petromyzon marinus cDNA library (23). Two Msx paralogs were isolated. One corresponded to a previously described Lethenteron japonicum Msx gene, LjMsxA. The other was a previously uncharacterized Msx designated MsxB (Fig. S2). Like LjMsxA, P. marinus MsxA expression was most prominent in the upper lip ectoderm and mesenchyme at stage (st.) 25–28, with minor expression in the lower lip (Fig. S3). MsxB expression was first observed in the dorsal neural tube at neurula stages. Expression in the ventral pharynx began around st. 23 and expanded into the upper and lower lips at st. 25. This expression persisted until st. 26.5 (Fig. 1 A, B, and M and Fig. S3), when an addition dorsal pharyngeal domain was observed (Fig. 1 A and M). Sectioning revealed ventral pharyngeal expression was in mesenchyme surrounding the endostyle, the ventral-most aspect of the pharyngeal bars, and ventral ectoderm. Dorsal pharyngeal staining was also in pharyngeal mesenchyme/CNCC (Fig. 1M).

Fig. 1.

Expression of MsxB, Hand, and Dlx genes in the larval lamprey head. All larvae are Tahara st. 26.5 and are shown in side view (A, C, E, G, I, and K), sagittal section (B, D, F, H, J, and L), and transverse section (M) (refer to Fig. S1 for a detailed description of larval lamprey pharyngeal morphology). (A, C, and M) MsxB expression in mesenchyme of the upper lip, lower lip, and ventral and dorsal pharynx. (C, D, and M) Hand expression in the ventral pharyngeal mesenchyme. (E, F, and M) DlxA expression around the mouth and in the pharyngeal arches. Expression in the lower lip is restricted to the posterior edge of the mouth. (G, H, and M) DlxB expression in the mesenchyme of the lower lip, pharyngeal arches, and surrounding the endostyle. Expression in the upper lip is reduced. (I and J) DlxC expression in ectoderm and mesenchyme of the upper and lower lips and in the pharyngeal arches. (K, L, and M) DlxD expression in the upper and lower lip, the pharyngeal arches, and overlying the endostyle. (M) Side-by-side comparison of Dlx, Ms,x and Hand expression along the DV axis in the pharynx at st. 26.5. Transverse sections through the dotted lines in A, C, E, G, and K. Sections were level matched to the fifth pharyngeal arch using the posterior limit of the endostyle as an internal reference point. Orange box serves as guide and bounds the pharyngeal cavity across all sections. Blue boxes define the DV extent of gene expression. DV expression of DlxC is similar to DlxD and is omitted for simplicity. Combinations of Dlx, Msx, and Hand genes mark four distinct DV domains. The ventral domain (v) expresses DlxB, MsxB, and Hand. Ventral intermediate (vi) domain expresses DlxA,B,C,D, and MsxB. The dorsal intermediate (di) domain expresses DlxA,B,C,D. Dorsal domain (d) expresses DlxB,C,D, and MsxB.

Cloning and Expression of Lamprey Hand.

A single exon with high similarity to vertebrate Hand genes was identified in the preassembly P. marinus genome (Fig. S2) and used to design RACE primers and isolate a 5′ RACE product. Hand expression was first observed at st. 23 in the ventral pharynx. This pattern persisted until st. 26.5 (Fig. 1 C, D, and M). Sectioning showed that this expression was in mesenchyme underlying and flanking the endostyle. A medial stripe of Hand expression was also observed in the mesenchyme of the lower lip (ventral first arch) at st. 26.5 (Fig. S3).

Cloning and Expression of Lamprey DlxA, DlxB, DlxC, and DlxD.

A low-stringency library screen for Dlx genes identified four Dlxs corresponding to the previously described P. marinus and L. japonicum paralogs, DlxA, DlxB, DlxC, and DlxD (20) (Fig. S4). Expression of all Dlxs in presumptive CNCC at st. 23–25 (Piavis st. 14) was essentially as previously described (Fig. S5). At st. 26.5, we observed clear differences in Dlx paralog expression along the DV axis of the pharynx and around the mouth (Fig. 1 E–M and Fig. S6). DlxA expression was observed in the lower two-thirds of the posterior gill bars and was excluded from the mesenchyme around the endostyle (Fig. 1 E, F, and M and Fig. S6). In the oral region, DlxA marked the upper lip, the distal tip of the lower lip (ventral first arch) and the first-arch gill bar (Fig. 1 E, F, and M and Figs. S3 and S6). DlxB was expressed broadly throughout the pharyngeal mesenchyme/CNCC including the mesenchyme around the endostyle (Fig. 1 G, H, and M and Fig. S6). In the oral region, DlxB was strongest in the lower lip and was down-regulated in the medial portion of the upper lip (Fig. 1 G, H, and M and Figs. S3 and S6). DlxC expression was broader than DlxA but less extensive than DlxB, with expression throughout the DV extent of the gill bars and absent from the mesenchyme around the endostyle (Fig. 1 I and J and Fig. S6). In the oral region, strong DlxC expression was seen in the ectoderm and mesenchyme of the upper and lower lips (Fig. 1 I and J and Figs. S3 and S6). The DV extent of DlxD expression in the pharyngeal arches and around the mouth was similar to that of DlxC (Fig. 1 K, L, and M and Figs. S3 and S6). When considered in the context of Msx and Hand expression, unique combinations of Dlx, MsxB, and Hand were observed, marking four distinct DV domains of pharyngeal mesenchyme/CNCC: a DlxB⁺/Msx⁺/Hand⁺ ventral domain, a DlxA,B,C,D⁺/MsxB⁺ ventral–intermediate domain, a DlxA,B,C,D⁺ dorsal–intermediate domain, and a DlxB,C,D⁺/MsxB⁺ dorsal domain (Fig. 1M) in the posterior arches and around the mouth (Fig. S3).

Cloning and Expression of Lamprey Goosecoid.

A single lamprey Goosecoid (Gsc) homolog was identified in the P. marinus EST database (Fig. S4). The cognate cDNA clone was isolated and used to make a riboprobe. Gsc expression was first observed in patches of upper lip mesenchyme and in pharyngeal endoderm at st. 25. By st. 26.5, Gsc expression in the upper lip mesenchyme had expanded and endodermal expression had resolved into a stripe of expression in the first pharyngeal pouch (Fig. 2 A and B and Fig. S3). Discontiguous expression of Gsc was seen in the first-arch mesenchyme, with transcripts detected in the gill bar and the distal tip of the lower lip (Fig. 2 A and B).

Fig. 2.

Expression of lamprey Gsc, Barx, Bapx, and Gdf5/6/7b in the embryonic and larval head. All larvae are Tahara st. 26.5 and are shown in side view (A, C, E, and G) and sagittal section (B, D, F, and H). (A and B) Gsc expression in the mesenchyme of the upper lip, first-arch gill bar (arrow), and distal lower lip (arrowhead). (C and D) Barx expression in the mesenchyme of the lower lip, extending dorsally into the ventral portion of the first gill bar (arrowhead). (E and F) Bapx expression in the epidermis of the ventral pharynx (arrowhead). (G and H) Gdf5/6/7b expression in the upper and lower lip ectoderm, upper lip mesenchyme, pharyngeal endoderm, ventral ectoderm ,and ventral mesenchyme surrounding the endostyle (arrowhead).

Cloning and Expression of Lamprey Barx.

Two exons of a single lamprey Barx gene were identified by searching the preassembly P. marinus genome using gnathostome Barx sequences (Fig. S4). These exons were amplified directly from genomic DNA and used to generate riboprobes. Lamprey Barx expression was first detected in two bilateral stripes in the lower lip at st. 25. By st. 26.5, Barx transcripts were detected in the presumptive CNCC medial to the mesoderm in every pharyngeal arch and strongly in the lower lip and velum (Fig. 2 C and D and Fig. S3). Parasagittal sections revealed a contiguous domain of expression in the first arch extending from the lower lip to the ventral portion of the gill bar (Fig. 2D).

Cloning and Expression of Lamprey Bapx.

A single Bapx exon was identified in the preassembly P. marinus genome, and was used to design RACE primers and to isolate 5′ and 3′ RACE products (Fig. S7). Bapx expression was seen in the paraxial mesoderm, presumptive pronephros, and pharyngeal endoderm from st. 21–25 (Fig. S8). Beginning at st. 24, signal was observed in bilateral patches of ventral pharyngeal ectoderm (Fig. S8). This staining persisted until st. 26.5 (Fig. 2 E and F and Fig. S3).

Cloning and Expression of Lamprey Gdf5/6/7a and Gdf5/6/7b.

The preassembly lamprey genome was searched for homologs of gnathostome Gdf5, a marker of joints in gnathostomes. Two 500- to 600-bp exons were identified, amplified directly from genomic DNA, and used to synthesize riboprobes and to screen an arrayed embryonic DNA for full-length clones. Phylogenetic analysis grouped the first with gnathostome Gdf7 but with low confidence. The second was placed as an outgroup to gnathostome Gdf5 and Gdf6 paralogs at moderate bootstrap values (Fig. S7). To reflect the uncertain phylogeny, we designated the genes Gdf5/6/7a and Gdf5/6/7b. Expression of lamprey Gdf5/6/7a was seen throughout the head epidermis from st. 21 to st. 26.5 (Fig. S3). Gdf5/6/7b expression was first seen in the mesenchyme and ectoderm of the ventral pharynx and in two stripes of ectoderm flanking the preoral region at st. 23. These ectoderm patches fused anteriorly at st. 25–26.5, strongly marking the upper lip (Fig. 2 G and H and Fig. S3). Pharyngeal expression of Gdf5/6/7b persisted until st. 26.5 in the ventral ectoderm and mesenchyme flanking the endostyle (Fig. 2 G and H).

Discussion

Combinatorial Expression of Dlx, Msx, and Hand Define Four Distinct DV Domains in both Lamprey and Gnathostomes.

Deep conservation of early neural crest developmental mechanisms (11, 12, 23) implies that jaw evolution was driven by later changes to the patterning of postmigratory cranial neural crest cells (CNCCs) and/or skeletogenic mesoderm. In gnathostomes, extracellular signals pattern the nascent pharyngeal skeleton by activating combinatorial expression of various target genes at specific DV levels. These factors then cooperate to confer positional identities upon nascent skeletal elements, determining their final shape and composition (2, 24). We examined the expression of lamprey homologs of the ventrally expressed patterning genes Msx and Hand in the lamprey pharynx. Consistent with a conserved role in specifying ventral CNCC, we observed lamprey MsxB in the ventral aspect of the pharyngeal bars and Hand/MsxB coexpression in the lower lip and mesenchyme surrounding the endostyle (Fig. 1).

Nested expression of Dlx2,3,5,6 has been shown to confer DV identities upon pharyngeal CNCC in both mouse and zebrafish (6, 7, 25), suggesting that this mechanism evolved before the divergence of lobe-finned and ray-finned fish. Previous reports describe ubiquitous expression of four Dlx paralogs in the pharynx of P. marinus and the Japanese lamprey Lethenteron japonicum at early larval stages (Tahara st. 25/Piavis st. 14, 12–13 d postfertilization), before the pharyngeal arches have fully formed. These results led to speculation that a pharyngeal Dlx code is a gnathostome novelty associated with the evolution of the jaw (7, 19, 20, 22). Because of the clear DV restriction of Hand and Msx in later larvae, we decided to revisit lamprey Dlx gene expression at st. 26.5 (15–16 d postfertilization). At this stage, pharyngeal arch formation is complete and pharyngeal CNCCs have condensed into prechondrogenic rods. Recent work has shown that the full Dlx code does not resolve until similar late-pharyngula stages in gnathostomes (8, 26). As with Msx and Hand, we used a high-stringency in situ hybridization protocol and level-matched transverse sections to accurately compare the extent of Dlx paralog expression along the DV axis. With these methods, we found clear differences in the expression of P. marinus Dlx paralogs, with different combinations of Dlxs marking distinct DV domains in the upper lip, first arch, and posterior arches (Fig. 1). Notably, these results differ from previous reports of uniform Dlx expression throughout the pharyngeal CNCC in P. marinus and L. japonicum. Although staging and methodological differences likely explain these discrepancies, it is formally possible that they reflect independent evolution of nested Dlx expression in P. marinus, or the loss of such patterning in L. japonicum. Despite a recent divergence time of 10–30 million years ago for the two species (27), differences in gene expression between them have been suggested (18, 28).

In jawed vertebrates, combinatorial expression of Dlxs, Msx, and Hand define four distinct DV domains of pharyngeal CNCC. Recent work in zebrafish has refined this map (8), revealing a Dlx2⁺/Msx⁺/Hand⁺ ventral domain, a Dlx2,3,5,6⁺ ventral-intermediate domain, a Dlx2,5,6⁺ dorsal-intermediate domain, and a Dlx2⁺ dorsal domain. This code diverges slightly in the first arch, with expression of Dlx5,6 extending into the ventral domain. We found that, as in gnathostomes, combinatorial expression of Msx, Hand, and Dlxs in lamprey also defines four DV domains. These include a DlxB⁺/Msx⁺/Hand⁺ ventral domain, a DlxA,B,C,D⁺/MsxB⁺ ventral-intermediate domain, a DlxA,B,C,D⁺ dorsal-intermediate domain, and a DlxB,C,D⁺/MsxB⁺ dorsal domain. Furthermore, as in gnathostomes, lamprey displays alterations to this code in the first arch, with the expression of intermediate Dlx paralogs (DlxC,D) extending ventrally. Although strikingly gnathostome-like, we also noted agnathan-specific divergence of the pharyngeal Dlx/Msx/Hand code. In gnathostomes, only Dlx2 marks the dorsal-most domain, whereas three Dlx genes and Msx mark this domain in lamprey. In addition, DlxA expression in the lamprey first arch has extended dorsally and become discontiguous, marking separate domains in the dorsal and ventral first arch (first gill bar and lower lip, respectively). Despite these differences, the most parsimonious explanation for the existence of a gnathostome-type pharyngeal Dlx/Msx/Hand code in lamprey is that such patterning was present in the common ancestor of jawed and jawless vertebrates. This ancestral pattern then diverged in one or both lineages during the half-billion years separating lamprey and gnathostomes.

Conservation of a gnathostome-type pharyngeal Dlx/Msx/Hand code in lamprey requires that some of the gene duplications that produced the gnathostome Dlx paralogs occurred in the first vertebrates. The six gnathostome Dlx genes were created by tandem duplication of an ancestral Dlx gene, followed by two whole-genome duplications and gene loss. These events produced three pairs of linked genes (Dlx1,2, Dlx3,4, and Dlx5,6), with each pair including members of two Dlx subfamilies (Dlx1,4,6 and Dlx2,3,5). Phylogenetic analyses (20, 22, 29) (Fig. S4) group lamprey DlxA,B,C within the Dlx2,3,5 subfamily, but fail to support placement of DlxD within either gnathostome Dlx subfamily. Neighbor-joining analysis provides some support for a DlxB/Dlx2 clade (Fig. S4) that is consistent with the broad Dlx2-like expression observed for lamprey DlxB. However, there is no support for one-to-one orthology between any other lamprey and gnathostome Dlx. Taken together, the tenuous phylogenetic positions of lamprey Dlxs leave open the precise relationships between agnathan and gnathostome Dlx paralogs. This may be due to the rapid evolution of lamprey genes relative to their gnathostome homologs (27). Assembly of the lamprey genome and identification of conserved vertebrate synteny groups should help to establish the timing of the duplications that generated the gnathostome and lamprey Dlx paralogs.

DV Patterning in the Lamprey Pharynx Correlates with Different Skeletal Cell Types.

In gnathostomes, combinatorial expression of Dlx, Msx, and Hand is thought to perform two related patterning functions. Initially, it confers positional identities on pharyngeal CNCC, instructing cells at intermediate levels to differentiate into joint tissue, and those at dorsal and ventral levels to form cartilage. Later, it instructs nascent cartilages to take on level-specific morphologies (8). The lamprey head skeleton lacks both joint tissue and individual cartilages with level-specific morphologies. This lack of overt patterning raises the question of why the lamprey pharynx displays such sophisticated patterns of gene expression along its DV axis. Recent and historical accounts suggest that, despite DV symmetry in the branchial basket, the larval lamprey head skeleton varies in cellular composition along its DV axis. Schaffer (30) and others (31, 32) describe a skeletal tissue consisting of mesenchymal cells embedded in a stiff extracellular matrix termed “mucocartilage” in the ventral-most pharynx (Fig. S1). This tissue surrounds the endostyle and sits ventral to the definitive cartilage of the branchial basket. Modern dye-labeling (11) and gene expression (12) suggests that mucocartilage mesenchyme is derived from CNCC. Recent work has also revealed differences in the morphology of branchial basket chondrocytes (33), with those in the gill bars displaying a highly ordered “stack of coins” morphology, those in the dorsal-most horizontal (subchordal) cartilage bars appearing polygonal in shape, and those in the ventral-most horizontal (hypobranchial) cartilage bars forming irregular stacks. Furthermore, these populations likely form from separate dorsal, ventral, and intermediate chondrogenic condensations (34). Thus, the larval lamprey pharyngeal skeleton consists of at least four distinct CNCC-derived skeletal cell types positioned at specific DV levels: mesenchymal muco-cartilage, hypobranchial chondrocytes, gill bar chondrocytes, and subchordal chondrocytes. Interestingly, the positions of these cell types correspond well to four domains of combinatorial Dlx, MsxB, and Hand expression identified in this study (Fig. 3). These data suggest that the ancestral function of pharyngeal DV patterning may have been to specify skeletal cell types and morphologies at different DV levels within the pharyngeal arches. This system was then used to position joint tissue and to regulate the shape of nascent cartilages in the gnathostome lineage.

Fig. 3.

Comparison of DV patterning gene expression in the developing zebrafish and lamprey pharyngeal skeletons. (A) Ventral view of a 48hpf zebrafish embryo showing Bapx expression flanking the mouth in the forming jaw joints. (B) Lamprey Bapx expression is never observed in the nascent head skeleton. Dotted line marks the posterior edge of the mouth in both A and B. (C) DV extent of Gdf5, Bapx, Barx, Gsc, Dlx2,3,5,6, Msx, and Hand in the pharyngeal arches of the zebrafish. First-arch–specific patterning is depicted by the colored bars to the left of the diagram. Asterisk marks the jaw joint. (D) DV extent of Barx, Gsc, DlxA,B,C,D, Msx, Hand, and Gdf5/6/7b in the pharyngeal arches of the sea lamprey. First-arch–specific patterning is depicted by the colored bars to the left of the diagram. In both zebrafish and lamprey, combinatorial expression of Dlx, Msx, and Hand genes defines four domains of CNCC in the pharynx; ventral (v), ventral-intermediate (vi), dorsal-intermediate (di), and dorsal (d), suggesting sophisticated DV patterning is a basal feature of vertebrate head skeleton development. Although lamprey possesses a region topologically homologous to the jaw joint (asterisk), as indicated by similar first-arch–specific expression of Dlx, Msx, Hand, and Gsc, first-arch expression of Gdf5 and Bapx appears to be unique to gnathostomes, suggesting that recruitment of these genes to the first arch was associated with the evolution of the jaw. Combinatorial expression of lamprey Dlx, Msx, and Hand in the posterior pharynx corresponds to four distinct skeletal cell types, namely, mucocartilage (yellow), hypobranchial chondrocytes (green), “stack-of-coins” gill bar chondrocytes (light blue), and subchordal chondrocytes (dark blue). Pm, premandibular mesenchyme.

First-Arch Expression of the Jaw Joint Regulators Bapx and Gdf5/6/7 Is Unique to Gnathostomes.

Like the posterior arches, the gnathostome first arch is patterned by combinatorial expression of Dlx, Msx, and Hand transcription factors. Gnathostome-like expression of these genes in lamprey suggests that this patterning is an ancestral feature of vertebrate pharyngeal arch development. Superimposed upon this Dlx/Msx/Hand DV code, the gnathostome first arch has additional patterning distinguishing it from the posterior arches. This includes discontiguous expression of Gsc and Barx in the dorsal and ventral aspects of the first arch, focal expression of Bapx and Gdf5 in the intermediate domain that gives rise to the jaw joint, and ventral expression of Dlx5,6 (5, 9). These patterning add-ons constitute potential novelties associated with specialization of the first arch into the jaw.

To test when these features of first-arch patterning evolved, we examined expression of Gsc, Barx, Gdf5, and Bapx in lamprey larvae. The lamprey first arch is part of a highly specialized oral skeleton derived from both premandibular and first arch CNCC (35) (Figs. S1 and S3). Although very different from the gnathostome jaw, recent work has defined the contribution of the first arch to the lamprey oral skeleton, facilitating comparisons between the lamprey and gnathostome first arch (19, 35, 36). We found that, as in gnathostomes, lamprey Gsc marks discontiguous dorsal and ventral domains in the presumptive first arch. These domains overlap with ventral Dlx, Msx, and Hand expression in the lower lip and dorsal Dlx expression in the pharyngeal gill bar, and flank a Gsc-free intermediate domain between the lower lip and first gill bar. Taken together, these data suggest topological homology between the intermediate first arch in lamprey and the gnathostome jaw joint, and support evolution of first-arch–specific patterning early in the vertebrate lineage (Fig. 3).

In contrast to Gsc, we did not find gnathostome-like expression of Barx, Bapx, or Gdf5/6/7 in the lamprey first arch. Instead, we observed contiguous expression of Barx in the lower lip and intermediate first arch, and no Gdf5/6/7 or Bapx expression in the presumptive CNCC of any pharyngeal arch. These findings are particularly provocative, as Bapx loss-of-function in gnathostomes leads to a loss of Gdf5 expression and fusion of the maxillary and mandibular cartilages (5), whereas a loss of Gdf5/6/7 function results in the widespread loss of joints in the axial and appendicular skeletons (37). In addition, zebrafish furinA and plc3ßc mutants, which lack intermediate first arch Bapx expression, have fused first-arch cartilages, and display contiguous first-arch Barx expression (9), reminiscent of the agnathan condition. These results suggest that a gain of Bapx and Gdf5/6/7 expression in the gnathostome first arch was associated with the evolution of the jaw joint.

It is unclear what the consequences of Bapx and Gdf5/6/7 cooption to the intermediate first arch may have been. Bapx labels intermediate first arch CNCCs that form the jaw joint in nonmammalian gnathostomes (5) and the homologous joint between the malleus and incus in mammals (38). Bapx knock-down in zebrafish causes a loss of the jaw joint and jaw joint-specific expression of Gdf5 and Chordin in zebrafish (5). However, first-arch Bapx expression in zebrafish is broader than the joint, including nascent chondrocytes in the upper and lower jaws (5). Furthermore, in mouse, Bapx knock-out causes a reduction in the size of the inner ear bones but does not effect the malleus-incus joint or joint-specific expression of Gdf5/6(38). Bapx knock-down also does not effect the formation joints in the posterior pharyngeal arches or elsewhere in the axial or appendicular skeletons of fish or mouse (5, 39). Taken together these data suggest that Bapx likely performs a first-arch–specific patterning function rather than a joint-specification function. In this context, it is unlikely Bapx cooption alone was a driving force in the evolution of the jaw joint; rather, novel expression of Bapx in the intermediate first arch likely served to reinforce or refine first-arch DV pattern in early gnathostomes.

In contrast to Bapx, members of the Gdf5/6/7 paralogy group are general markers of joint tissue in gnathostomes (5, 40, 41). Loss of Gdf5 and/or Gdf6 results in the loss of joints as well as in reduced growth of bones and cartilages, suggesting that Gdf5/6/7 plays roles in both joint specification and cartilage growth (40–42). In lamprey, Gdf5/6/7b is expressed in patches of pharyngeal ectoderm and endoderm, as well as in the mucocartilage of the ventral pharynx (Fig. 2). Interestingly, classical and modern descriptions of mucocartilage suggest it shares several features with the Gdf5/6/7-expressing hyaline cartilage found in gnathostome joints (31). Both tissues consist of scattered secretory cells embedded in a matrix of hyaluronic acid, chondroitin sulfate, and microfibrils, and both are surrounded by a collagenous perichondrium. It is possible that novel expression of Gdf5/6/7 may have driven the formation of mucocartilage-like joint tissue in the intermediate first arch of the gnathostome ancestor.

Prepattern/Cooption Model of Vertebrate Jaw Evolution.

Aspects of pharyngeal arch patterning conserved in agnathans and gnathostomes can reasonably be expected to have been present in their last common ancestor. Based on similar expression of Dlx, Msx, and Hand in the pharynx, we propose that the last common ancestor of jawed and jawless vertebrates deployed these genes to pattern its pharynx along the DV axis. This pattern was sophisticated, dividing the pharyngeal CNCC into at least four domains specified by various combinations of Dlx, Msx, and Hand genes, and giving rise to distinct types of skeletal tissue (Fig. 3). This basal pattern was likely present in all pharyngeal arches, although the first arch displayed some additional pattern including discontiguous domains of Gsc expression, and ventral expansion of some Dlx paralogs. From this ancestral state, the lamprey and gnathostome patterns diverged, with changes to the DV boundaries of particular Dlx and Msx genes in one or both lineages. In gnathostomes, Bapx was coopted to the intermediate first-arch domain, disrupting ancestrally contiguous Barx expression and augmenting or refining the ancestral DV pattern. Subsequent recruitment of Gdf5/6/7 to this domain then drove the evolution of the jaw joint, possibly by conferring mucocartilage-like properties upon nascent first-arch chondrocytes.

Our model is based on the assumptions that features present only in gnathostomes constitute gnathostome novelties, whereas similarities between lamprey and gnathostomes reflect shared, derived features of vertebrates. However, it is also possible that gnathostome-specific features such as first-arch Bapx expression are the result of evolutionary loss in the lamprey lineage. Similarly, putatively conserved features such as nested Dlx expression could formally reflect parallel evolution from an ancestor lacking pharyngeal DV pattern. Critical tests of these alternate scenarios will be the expression of pharyngeal DV patterning genes in elasmobranchs and hagfish. Recent molecular phylogenies suggest that hagfish are derived agnathan vertebrates related to lampreys (29, 43), and progress has been made in obtaining hagfish embryos (44). Gnathostome-like expression of Dlx, Msx, and Hand in hagfish would confirm that a pharyngeal Dlx/Msx/Hand code is a basal feature of vertebrate development. First-arch expression of Bapx and Gdf5/6/7 in elasmobranchs, but not agnathans, would suggest that these aspects of first-arch patterning are indeed gnathostome synapomorphies, as predicted by our model.

Materials and Methods

Low and moderate stringency library screens for Msx, Dlx, and Gdf5/6/7 genes were performed as previously described (23). Barx, Gdf5/6/7a, and Gdf5/6/7b exons were PCR amplified from adult lamprey genomic DNA according to standard methods. All other genes were isolated from embryonic cDNA using the GeneRacer kit (Invitrogen). In situ hybridizations were performed as previously described (16) with some modifications. For details, see SI Materials and Methods.