Gene expression patterns underlying proximal–distal skeletal segmentation in late-stage zebrafish, Danio rerio


  • Patricia L. Crotwell,

    1. Department of Biology, University of South Dakota, Vermillion, South Dakota
    Current affiliation:
    1. Department of Medical and Molecular Genetics, Indiana University School of Medicine, Indianapolis, IN 46202
    Search for more papers by this author
  • Paula M. Mabee

    Corresponding author
    1. Department of Biology, University of South Dakota, Vermillion, South Dakota
    • Department of Biology, University of South Dakota, Vermillion, South Dakota 57069
    Search for more papers by this author


Timing and pattern of expression of ten candidate segmentation genes or gene pairs were reviewed or examined in developing median fins of late-stage zebrafish, Danio rerio. We found a general correspondence in timing and pattern of expression between zebrafish fin radial segmentation and tetrapod joint development, suggesting that molecular mechanisms underlying radial segmentation have been conserved over 400 million years of evolution. Gene co-expression during segmentation (5.5–6.5 mm SL) is similar between tetrapods and zebrafish: bmp2b, bmp4, chordin, and gdf5 in interradial mesenchyme and ZS; bapx1, col2a1, noggin3, and sox9a in chondrocytes. Surprisingly, wnt9a is not expressed in the developing median fins, though wnt9b is detected. In contrast to all other candidate segmentation genes we examined, bapx1 is not expressed in the caudal fin, which does not segment. Together, these data suggest a scenario of gene interactions underlying radial segmentation based on the patterns and timing of candidate gene expression. Developmental Dynamics 236:3111–3128, 2007 © 2007 Wiley-Liss, Inc.


The skeleton of the vertebrate limb is one of the most intensely studied systems in developmental and comparative evolutionary biology. Numerous developmental studies of the vertebrate limb have resulted in increasingly well-characterized genetic interactions that are involved in the developing skeleton. The synovial joints of the limb are of special significance from a biomedical standpoint, and mutant phenotypes and functional studies have led to hypotheses about key genes and their involvement in the processes of mesenchymal condensation, cartilage cell growth and differentiation, segmentation of cartilage rods (digits), and synovial joint cavitation. In contrast, functional studies of zebrafish fin development generally have focused on either early pectoral fin development, or on fin ray regeneration (Geraudie et al.,1993; Akimenko et al.,1995; van Eeden et al.,1996). Few studies in either teleost or tetrapod vertebrates have focused on the evolutionary origin of the developmental mechanisms underlying segmentation.

Similarity in the developmental pattern of segmentation in tetrapod limb cartilages and fish median fin cartilages (Fig. 1), and the earlier evolution of median fins compared with paired fins (Fig. 2) led us to propose that the genetic mechanisms underlying the joints in vertebrate limbs were derived from those underlying the segmenting cartilages in median fins (Crotwell et al.,2001; Mabee et al.,2002). Using the genes involved in tetrapod joint formation as key candidates (Fig. 3), we examined their expression patterns in the segmenting median fin cartilages of zebrafish. We report here the expression of these genes in comparison to tetrapod animal models, the chick and mouse. Our results provide a baseline for development of functional tests that will allow direct comparison of fin radial segmentation to vertebrate limb skeletal segmentation.

Figure 1.

A–D: Cleared and cartilage-stained skeletal preparations showing the progression of segmentation in zebrafish anal fin radials (A–C; adapted from Bird and Mabee,2003) and an 18 days postfertilization mouse embryo (D). A: 5.4 mm. B: 5.6 mm. C: 6.5 mm, proximal and distal radials are identified; asterisks mark the zones of segmentation. D: The mouse left forehand is shown here in a dorsal view with anterior at bottom. Two joints in digit four are marked with asterisks. Scale bars = 0.5 mm.

Figure 2.

Skeletal and fin features mapped on a phylogenetic tree for craniates (Coates,1994; Janvier,1996; Coates and Cohn,1998).

Figure 3.

Interactions of 21 genes and one hydrocarbon ring (retinoic acid) hypothesized to be involved in joint development, compiled from zebrafish, flounder, Xenopus, chick, and mouse (Supplementary Table, which can be viewed at The blue core represents putative interactions of the 10 candidate segmentation genes we examined in zebrafish, and these were derived primarily from mouse and chick studies. Not every gene whose expression and/or putative function suggests a role in joint development is included. Arrows indicate up-regulation of the target gene by the effector. T-lines indicate down-regulation or inhibition of the target gene by the effector.


It is well known that the paired limbs of tetrapods evolved from the paired fins of fishes (Coates,1994). Although tetrapod forelimbs have skeletal homologues within the pectoral fins of basal actinopterygian fishes such as sturgeons and paddlefishes (Mabee,2000), the pectoral fins of teleost fishes are more highly derived, and the common skeletal elements are lost (Fig. 2). It is less well known that median fins evolved before paired fins (Coates,1994; Janvier,1996) (Fig. 2), and the median fins may be considered iterative homologues of the paired limbs. Furthermore, it has been shown that the molecular mechanisms that underlie shark paired and median fin positioning and development have been conserved in tetrapod paired appendage positioning and development (Freitas et al.,2006; Dahn et al.,2007), although the paired fins are derived from lateral plate mesoderm and the median fins from sclerotome (Freitas et al.,2006). This evolutionary sequence suggests that the patterning/segmentation of the paired fins/limbs may have been “co-opted” (True and Carroll,2002) from that of the segmented median fins (Fig. 2). That is, paired fins may have co-opted their genetic mechanisms for segmentation from median fins at their origin (Ahlberg,1992; Mabee et al.,2002; Freitas et al.,2006). Thus, the segmentation mechanisms in the median fins might be expected to share fundamental aspects of genetic architecture with the well-studied limbs.

The synovial joints of tetrapod limbs develop by means of a well-regulated and consistent series of cellular differentiation events beginning with condensations of prechondrogenic mesenchyme (Mitrovic,1977,1978). The cells in these condensations differentiate into chondrocytes, forming the cartilaginous anlagen of the skeletal elements. This anlagen grows, branches, and then segments into separate pieces connected by joints (Mitrovic,1978; Fig. 1D). In the feet, for example, a metatarsal, proximal phalanx, and medial phalanx, and the joints between them, arise from a single condensation called a digital ray (Dalgleish,1964). The first indication of joint development in the digital rays is the formation of the interzone, a specialized region of high cell density within the condensation (Haines,1947). The process of cavitation or the initiation of joint formation may involve two closely linked events: first the loss of cohesion within the cartilaginous precursor at the presumptive joint line, which allows its separation to form distinct articulating elements, and second, movement of these elements apart to produce a fluid-filled space, the synovial cavity (Mitrovic,1977,1978). The morphological changes that occur at sites of presumptive joint formation are accompanied by characteristic changes in staining and expression patterns of cartilage cell markers (Archer et al.,2003; Pacifici et al.,2006).

Teleosts such as zebrafish have three median fins: dorsal, anal, and caudal (Fig. 4). The fins consist of repeated structures: the endoskeletal supports, i.e., the radials of the dorsal and anal fins and hypurals of the caudal fin, and the exoskeletal fin rays. The anal and dorsal fins develop and are patterned similarly in fishes (Mabee et al.,2002). The anal fin begins development slightly before the dorsal fin in zebrafish. Matrix-positive chondrocytes are present in mesenchymal condensations within the anal and dorsal fins by 5.1 mm NL and 5.5 mm SL, respectively (Bird and Mabee,2003; Fig. 1A; see the Experimental Procedures section for NL and SL terminology). This stage of development occurs approximately 10–12 days postfertilization (dpf; Appendix A, Cubbage and Mabee,1996). Generally, two to three radials form in the central region of the fin, and they are added anteriorly and posteriorly to form the adult complement. The radials are composed of a large proximal radial and a small distal radial that articulates with the external fin ray. The proximal and distal radials develop from a single rod of cartilage that segments in its development. Superficially at least, this segmentation resembles that observed in the tetrapod limb, for example, in the segmentation of phalanges from single rods (Hartmann and Tabin,2001; Sanz-Ezquerro and Tickle,2003; Satoh et al.,2005). Similar to tetrapod joints, the entire cartilaginous radial is heavily stained by Alcian blue, but as the zone of segmentation (ZS) develops (Fig. 1C), Alcian blue is no longer taken up by the cells between the proximal and distal radial, indicating a conversion to a nonchondrogenic (likely mesenchymal) cell type (Bird and Mabee,2003). The ZS persists throughout the life of the fish, although segmentation is complete by ∼7 mm (∼18 dpf, Cubbage and Mabee,1996). The fin rays (Fig. 4) begin development concurrent with radial segmentation.

Figure 4.

A diagram of the posterior axial skeleton of the zebrafish (adapted from Bird and Mabee,2003). Dorsal, anal, and caudal fins are denoted and the endoskeletal fin supports are colored blue. The dorsal and anal fin supports consist of proximal and distal radials (arrowheads). The caudal fin supports do not segment. Fin rays (arrows), ribs, and neural and hemal spines are also labeled.

The caudal fin is the first median fin to develop in most fishes, including zebrafish. The zebrafish caudal fin is internally supported by modified hemal spines and hypurals (Fig. 4). The supports begin formation at 3.8 mm NL (∼4 dpf, Cubbage and Mabee,1996), and all are present by 5.6 mm (Bird and Mabee,2003). The caudal fin rays begin development at 4.8 mm NL. Because the hypurals do not segment, they served as an internal control for the segmenting cartilages (radials) of the dorsal and anal fins.


We prioritized a subset of 10 candidate genes and gene interactions (Fig. 3, blue core) to compare the timing and pattern of gene expression in zebrafish median fins with that described for tetrapod limbs (Crotwell,2005). These 10 genes are not expressed at precisely the same time during limb development, but are known to up- and/or down-regulate each other in a dynamic pattern in tetrapods. We separated these genes into two subsets based on the timing and pattern of gene expression.

Subset 1 consists of genes expressed in tetrapod digit joints or exclusively in mesenchyme; these are Bmp2 (Macias et al.,1997), Bmp4 (Francis et al.,1994; Francis-West et al.,1999b), Chordin (Francis-West et al.,1999b), Gdf5 (Francis-West et al.,1999a; Merino et al.,1999; Storm and Kingsley,1999), Shh (Riddle et al.,1993), and Wnt9a (Hartmann and Tabin,2001; Guo et al.,2004).

Tetrapod Bmp2 and Sox9 exist as gene pairs in zebrafish; the gene pairs are a result of the duplication of a single ancestral gene that is retained in tetrapod vertebrates (Chiang et al.,2001b; Van de Peer et al.,2002). In contrast, a sequence analysis (Qian et al.,2003) showed that the joint-inducing gene Wnt14 (Hartmann and Tabin,2001) and the highly similar (in sequence) Wnt15 (Bergstein et al.,1997), are, in fact, orthologues of a single hagfish and shark wnt9 (Sidow,1992). Tetrapod Wnt14 and Wnt15 have been renamed Wnt9a and Wnt9b, respectively (Qian et al.,2003). Wnt9a is expressed in the developing limbs of both chick and mouse, but Wnt9b is not (Qian et al.,2003). Zebrafish have two wnt9 genes (perhaps more, although we did not discover any), and sequence analysis suggests the presence of both wnt9a and wnt9b in the zebrafish genome (Crotwell,2005). Because gene pairs may diverge in function and expression after the duplication event (Massingham et al.,2001; Li et al.,2005), it is necessary to examine both genes in zebrafish to determine whether both, either, or neither is correlated with the role of the tetrapod gene(s).

We previously compared expression of bmp2a and bmp2b in a developmental fin series in zebrafish and found that bmp2b expression mirrors that of Bmp2 expression in developing joints of tetrapods (Crotwell et al.,2004). In the developing chick limb, Bmp2 is expressed in interdigital regions (Francis et al.,1994; Duprez et al.,1996), and later in a horizontal stripe in the developing joint (Macias et al.,1997; Francis-West et al.,1999b). In addition to a stripe of expression in the dorsal and anal fin ZS, the fin rays and hypertrophic chondrocytes of all three median fins also expressed bmp2b (Crotwell et al.,2004). Expression of bmp2a is restricted to fin rays, which tetrapods lost with their fins (Fig. 2; Crotwell et al.,2004).

Bmp4 is expressed early in mesenchyme surrounding chondrogenic digits and is later restricted to the perichondrium (Jones et al.,1991; Francis et al.,1994; Duprez et al.,1996; Zou et al.,1997). In Xenopus, Bmp4 expression also is high in the interdigital region and later in the joint regions and in peripheral digit cartilages (Satoh et al.,2005). Zebrafish have one copy of bmp4 (Martinez-Barbera et al.,1997; Nikaido et al.,1997), but expression had not been examined past 2 dpf.

Chordin is a Bmp antagonist whose expression is positively regulated by Bapx1, Bmp4, and Wnt9a (Fig. 3). Expression patterns of Chordin during skeletal development in vertebrates suggest a role for this gene in patterning the skeleton. In zebrafish, chordin expression is localized to the developing jaw joint (56 hpf) (Miller et al.,2003). During chick digit development, Chordin is expressed both in the joint and around the epiphyses of the digit cartilages, although expression is absent from the cartilage itself (Francis-West et al.,1999b). Later, Chordin expression is restricted to the joint (Francis-West et al.,1999b). In mouse, Chordin is expressed in the developing joint, at the distal tips of the digits, and at high levels in perichondria; it is absent in the cartilage condensations themselves (Scott et al.,1999,2000). Of interest, Bmp1 and Tll-1 (Tolloid-like 1) proteins, which cleave Chordin, are expressed in virtually identical patterns as Chordin (Scott et al.,1999). In mouse and chick, Chordin is expressed at times and in patterns similar to the Bmp family members it antagonizes, thus it is temporally and spatially situated to function in skeletal development and patterning (Francis-West et al.,1999b; Scott et al.,1999,2000).

At least 10 different zebrafish chordin mutations have been identified (ZFIN,2007) and all disrupt dorsoventral patterning of the embryo to at least some extent (Schulte-Merker et al.,1997; Fisher and Halpern,1999). Despite disruption of early patterning, some chordin-mutant zebrafish do survive to adulthood, although they exhibit a range of axial skeletal abnormalities (Fisher and Halpern,1999). In the anal fin, the proximal radials are in some cases duplicated, giving rise to a disorganized adult anal fin skeleton, although segmentation appears to be normal (Fisher and Halpern,1999). Additionally, several chordin mutants rescued by RNA injection at the single cell stage possessed normal axial skeletons as adults, suggesting that chordin expression during gastrulation is sufficient for normal adult fin development (Fisher and Halpern,1999).

The growth and differentiation factor Gdf5 is closely related to the bone morphogenetic proteins within the Tgf-β gene superfamily, and it is a critical factor in the growth and differentiation of the cartilaginous phalanges of tetrapods (Storm and Kingsley,1999; Coleman and Tuan,2003). Joint formation generally will not progress if Gdf5 or its receptor is mutated or absent (Francis-West et al.,1999a; Storm and Kingsley,1999; Baur et al.,2000; Yi et al.,2000). Gdf5/CDMP-1 (human cartilage-derived morphogenetic protein-1) mutations are responsible for the skeletal abnormalities (phalangeal segmentation defects, shortening of the long bones of the limbs) observed in the mouse mutation brachypodism (Storm et al.,1994) and in human Hunter-Thompson type chondrodysplasia (Thomas et al.,1996). We previously described that zebrafish gdf5 is first expressed in the mesenchyme between the earliest developing radials (Crotwell et al.,2001). During radial segmentation, gdf5 expression extends into the ZS and is continuous with the anterior expression band of the adjacent posterior radial. After segmentation is complete (by 7.1 mm), expression of gdf5 ceases (Crotwell et al.,2001).

Hedgehog, like other members of the Hedgehog family, is a cell signaling glycoprotein (Motoyama et al.,1998; Britto et al.,2000). Hedgehog proteins are involved in a diverse array of embryological processes, including development and patterning of the limbs, eyes, and central nervous system of vertebrates (Ekker et al.,1995; Vortkamp et al.,1996; Endo et al.,1997; Sbrogna et al.,2003). Studies of the role of Sonic (Shh) in tetrapod limb development, particularly, are numerous. Shh is expressed in the posteriorly located ZPA (zone of polarizing activity) during early limb development (Riddle et al.,1993). Some functional studies (Chiang et al.,2001a; Kraus et al.,2001; Lettice et al.,2003; Sanz-Ezquerro and Tickle,2003) suggest that Shh is involved in patterning the digit field (i.e., changing the number of digits or phalanges), but not digit patterning itself (i.e., joint development). However, flounder exposed to the morphogen retinoic acid (RA), a regulator of Shh (Fig. 3), exhibited loss (or failure to form) of the distal fin radials along with severe malformations of head and pectoral fin cartilages (Suzuki et al.,1999,2000,2003).

Wnts are secreted glycoproteins that are known to be critical for developmental signaling resulting in cell polarity, proliferation, migration, and cell fate specification (Taipale and Beachy,2001). Wnt9a is one of only two genes suggested to induce the molecular and morphological signs of joint formation (Hartmann and Tabin,2001; Guo et al.,2004; Spater et al.,2006b). The second is β-catenin (Guo et al.,2004), which may not be necessary to induce joint development, but instead to maintain it (Spater et al.,2006a).

Subset 2 consists of genes that are expressed in mesenchymal condensations and/or cartilage early, then are down-regulated in the joint; these are Bapx1 (Church et al.,2005), Col2a1 (Bell et al.,1997; Ng et al.,1997), Noggin (Francis-West et al.,1999b), and Sox9 (Hartmann and Tabin,2001; Zehentner et al.,2002).

Bapx1 (also known as Nkx3.2) is a member of the NK family of transcription factors. The expression of bapx1 in zebrafish was investigated for a role in radial segmentation because of its known role in jaw joint formation and patterning in zebrafish and chick (Miller et al.,2003; Wilson and Tucker,2004) and potential role in human skeletal disorders (Tribioli and Lufkin,1997; Yoshiura and Murray,1997). In the limbs, Bapx1 is expressed broadly in newly differentiated chondrocytes and later becomes restricted to proliferative chondrocytes as the cells of the endoskeleton become organized into layers (Church et al.,2005). Bapx1 expression is down-regulated in the joints and another gene, Barx1, is expressed in the joint region in a complementary manner (that is, where Barx1 levels are high, Bapx1 levels are low and vice versa; Church et al.,2005).

Type II collagen is the primary structural protein of cartilage and is found in extracellular matrix (Mendler et al.,1989; Poole et al.,1989), where it is surrounded by large proteoglycans aggregated with hyaluronan (Spicer and Tien,2004). Col2a1 expression marks proliferating and prehypertrophic chondrocytes and its expression is down-regulated both during joint development, and as chondrocytes hypertrophy before ossification (Craig et al.,1987; Ng et al.,1993; Lui et al.,1995; Nalin et al.,1995; Wilson and Tucker,2004). Thus, in the joint, Col2a1 expression and type II collagen protein are absent (Bland and Ashhurst,1996,2001). Col2a1 has been widely used as a cartilage-specific marker to which expression patterns of other joint-associated genes may be compared (Lele and Krone,1997; Storm and Kingsley,1999; Sachdev et al.,2001).

Noggin is a secreted 222 (frog) to 232 (human, mouse) amino acid (McMahon et al.,1998; Balemans and Van Hul,2002), ∼32 kDa glycoprotein (Zimmerman et al.,1996). Three noggin homologues are present in zebrafish (Fürthauer et al.,1999), in contrast to the single-copy noggin gene present in tetrapods. Zebrafish noggin3 is considered the most likely candidate for a skeletogenic role (M. Fürthauer, personal communication). In mouse and chick limbs, Noggin is expressed throughout the early condensing digits and later is down-regulated in the joint region (Brunet et al.,1998; Merino et al.,1998; Francis-West et al.,1999b; Zehentner et al.,2002). Noggin protein is capable of directly binding to Bmp2, Bmp4, Bmp7, and Gdf5, thus preventing them from binding to their receptors (Zimmerman et al.,1996; Merino et al.,1999). Noggin is essential for normal joint development: in human and mouse, if the gene coding for Noggin is absent or mutated, joints fail to form (Brunet et al.,1998; Gong et al.,1999).

Sox9 is a member of a large family of transcription factors that share a conserved ∼80 amino acid motif called the SRY-like HMG domain (Healy et al.,1999). Tetrapod vertebrates posses one Sox9 gene, whereas zebrafish, medaka, and sticklebacks have two, sox9a and sox9b (Cresko et al.,2003; Kluver et al.,2005; Yan et al.,2005). These genes are expressed in distinct and overlapping domains in early development (Cresko et al.,2003; Kluver et al.,2005; Yan et al.,2005). In the developing limbs, Sox9 is expressed in chondrocytes in chick (Healy et al.,1999; Chimal-Monroy et al.,2003), mouse (Ng et al.,1997; Bi et al.,1999; Zehentner et al.,2002), and Xenopus (Satoh et al.,2005). Expression is down-regulated in the forming joints (Ng et al.,1997; Healy et al.,1999; Satoh et al.,2005).


Subset 1: Genes Expressed in Tetrapod Digit Joints or Exclusively in Mesenchyme: Bmp4, Chordin, Shh, and Wnt9a


In situ analysis showed expression of zebrafish bmp4 in the mesenchyme between the developing radials of the dorsal and anal fins (Fig. 5). Expression of bmp4 was not detected in the ZS, although it was maintained, albeit at lower levels, in the interradial mesenchyme of presegmentation radials (Fig. 5B,C). After segmentation was complete, expression of bmp4 was no longer detected between the radials, although it was maintained in the fin rays (Fig. 5D). Caudal fin expression was observed in fin rays and in developing bone surrounding each caudal fin support (Fig. 5F). Additional domains of expression were observed in mesenchyme dorsal to the notochord, and in the developing centra (Fig. 5H). Reverse transcriptase-polymerase chain reaction (RT-PCR) analysis confirmed the presence of a predicted ∼700-bp fragment (Martinez-Barbera et al.,1997) of bmp4 in the caudal fin rays, caudal hypurals, and anal fin tissue (Crotwell,2005).

Figure 5.

Expression of bmp4 in the median fins and axial skeleton of zebrafish. Solid arrows mark fin rays, arrowheads mark radials, and hypurals are numbered. Black patches are melanophores. Anterior is to the left and dorsal to the top. A: In the same individual as H, bmp4 expression is not detected in the anal fin (5.4 mm). B: At a slightly later stage, bmp4 expression is observed in the interradial mesenchyme (5.7 mm). C: Expression of bmp4 is not detected in the ZS (*) between the proximal (top row of arrowheads) and distal (bottom row of arrowheads) radials, although it is maintained in the interradial mesenchyme (∧) between more posteriorly located, less-developed radials (6.0 mm). D: When radial segmentation is complete, expression of bmp4 is detected only in the fin rays (arrows; 6.6 mm). E: In the same individual as H, bmp4 expression is not detected in the caudal fin (5.4 mm). F: At a slightly later stage, expression is observed in the fin rays and ossifying perichondrium surrounding the hypurals, and very strongly in mesenchyme (hollow arrowhead) dorsal to the notochord (n) of the caudal fin (5.7 mm). G: An open arrow marks expression in the dorsal retina (Chin et al.,1997), which served as a positive control (26 hpf). H: Expression of bmp4 in the developing axial skeleton (centra, ∧) is not as robust as the dorsal retina positive control, but is clearly detectable (5.4 mm). Scale bars = 0.1 mm.


Expression of chordin was detectable by in situ hybridization using antisense probe (Fig. 6), but the levels of expression were lower than those of other studied genes. Expression in gastrula stage (6 hours postfertilization [hpf]) embryos served as the positive control (Miller-Bertoglio et al.,1997; Schulte-Merker et al.,1997; Fig. 6A). chordin was expressed in the interradial mesenchyme in early dorsal and anal fin development (Fig. 6D). Later, expression was observed in the forming fin rays (Fig. 6E) and in the ZS between the distal and proximal radials (Fig. 6F). chordin expression in interradial mesenchyme was low or not detectable when ZS expression was apparent (Fig. 6E,F). In the caudal fin, chordin was expressed in mesenchyme distal to the hypurals (Fig. 6G). chordin was strongly expressed in the mesenchyme between the forming neural and hemal arches (Fig. 6H). RT-PCR analysis confirmed the presence of chordin mRNA in caudal fin rays, caudal hypurals, and anal fin tissue (Crotwell,2005).

Figure 6.

In situ hybridization showing expression of chordin during median fin development in zebrafish. Arrows mark fin rays (except in A,B), arrowheads mark radials, and hypurals are numbered. Black patches are melanophores. Anterior is to the left and dorsal to the top, except in A and B, which are dorsal views (after Miller-Bertoglio et al.,1997). A: Strong chordin expression in future dorsal cells (arrow) during the shield stage (Miller-Bertoglio et al.,1997) serves as a positive control (∼6 hours postfertilization [hpf]). B,C: Sense probe does not detect chordin early (∼6 hpf, B) or late (5.6 mm, C), but does provide relative levels of background staining. D: The chondrocytes of the forming radials (arrowheads) do not express chordin, but the gene can be detected in the mesenchyme between the radials. Expression is not strong, but the stain is distinctly above background levels (5.6 mm). E: As segmentation progresses and fin rays are formed, expression of chordin is observed in the fin rays (arrows), and in the zone of segmentation (ZS; distal radials not present in this plane of section, but note expression at arrowhead tips, which point at the distal edges of the proximal radials) (5.6 mm). F: Expression of chordin in the ZS (asterisks) is clearly visible (5.8 mm). G: Mesenchyme distal to the hypurals expresses chordin (see hypural 5), although not as strongly as we have observed for other genes (6.2 mm). H: The highest levels of chordin expression are detected in the mesenchyme (∧) between the forming posterior neural (shown) and hemal (not shown) arches. Note absence of expression in the arch cartilages (cells to the right of the patches of expression) themselves. Scale bars = 0.1 mm in A,B, 0.05 mm in C,H.


We examined a size series of transgenic zebrafish that expressed gfp (green fluorescent protein) under the control of the shh promoter (shh:gfp+; Shkumatava et al.,2004). Roughly half of the 30 transgenic zebrafish embryos (generously donated by Rolf Karlstrom, U. Mass. Amherst) were gfp+ (compare A,B in Fig. 7). Eight of the gfp+ embryos were examined at 3 dpf; they suffered high mortality after exposure to the confocal laser. We reared the remaining larvae to appropriate fin development stages and were able to collect confocal data from five (4.7 mm NL, 5.6 mm, 6.4 mm, 6.9 mm, and 7.0 mm).

Figure 7.

Expression of shh in median fins was assessed in shh:gfp (green fluorescent protein) transgenic zebrafish (Shkumatava et al.,2004) by means of confocal microscopy. Black patches are melanophores. Anterior is to the left and dorsal to the top. A: After prescreening for presence or absence of gfp, gfp-negative embryos were used to provide baseline levels of autofluorescence. B: Expression of shh as evidenced by gfp in the ventral diencephalon and the floor plate of the neural tube (not shown) were as expected (Ekker et al.,1995). C: Expression in the floor plate of the neural tube (not shown in C, but see F) is maintained for an extensive period of time as a new domain of expression is observed in the caudal fin. Both mesenchyme in the fin and fin rays themselves show strong expression. D: This is observed more clearly in the anal fin. Ray expression is distinct, and anal fin mesenchyme shows high levels of shh. E: Although it appears that the cartilage cells of the radials (arrowheads) express shh, this is an artifact caused by bleed-over of the high levels of mesenchyme fluorescence. F: Absence of shh in the hypurals is apparent, while expression in the distal caudal mesenchyme, caudal fin rays (arrows), and floor plate of the neural tube (or perhaps dorsal notochord, n, see Ekker et al.,1995) is maintained. G: The pattern of expression in the anal fin after segmentation is complete shows expression in the mesenchyme in the shape of an H, with the hollow portions of the H consisting of the distal (bottom row of arrowheads) and proximal (top row of arrowheads) radials. The alternating angled light and dark lines in G and H are an artifact of the confocal microscope. H: Although interradial shh expression is maintained after radial segmentation (radials not in plane of section here, but see G), the highest levels of shh expression are observed in the distal and still developing fin rays (arrows), which are not in the plane of section in G.

We observed shh expression in the mesenchyme surrounding the developing fin radials/hypurals and in the developing fin rays (Fig. 7). We did not observe this expression in wild-type or shh:gfp−/− embryos and larvae (Fig. 7A and data not shown). As would be predicted from gene expression data (Ekker et al.,1995), central nervous system shh expression was high at positive control stages (Fig. 7B). Strong shh expression was observed in the mesenchyme between and distal to the dorsal and anal fin radials and in the developing fin rays (Fig. 7D). Although it appeared that shh was expressed in the radial cartilages (Fig. 7D,E), this effect was caused by the very high levels of mesenchyme fluorescence bleeding over into the radials. Expression was maintained in the mesenchyme surrounding the radials (Fig. 7G) after segmentation was complete. The strongest expression was observed in the still forming distal ray tips (Fig. 7H).

In the developing caudal fin, both mesenchyme and fin rays showed high levels of fluorescence (Fig. 7C). The floor plate of the neural tube/dorsal notochord, faintly visible in Fig. 7C, also expressed very high levels of shh:gfp. The mesenchyme and fin rays fluoresced, but the hypurals did not (Fig. 7F). Notochord/neural tube expression was maintained for a long period of time (Fig. 7F), at least until 7.0 mm. RT-PCR confirmed the presence of shh mRNA in caudal fin rays, caudal hypurals, and anal fin tissue; however, we were unable to detect shh expression in those tissues by means of in situ hybridization (Crotwell,2005).


Full-length cDNA sequences of zebrafish wnt9a and wnt9b were shared by M. Walker and C. Kimmel (Institute of Neuroscience, U. Oregon), and we used those and our previously cloned wnt9a fragment (GenBank accession no. DQ490071) to design RT-PCR primers for zebrafish. Negative RT-PCR data indicated that wnt9a was not expressed in the median fins of zebrafish (Fig. 8). In contrast, wnt9b was amplified by means of RT-PCR from the same caudal fin ray, caudal hypural, and anal fin tissue used for wnt9a RT-PCR (Fig. 9). However, wnt9b expression was not detected in dorsal, anal, or caudal fin cartilages by in situ hybridization (Crotwell,2005).

Figure 8.

Reverse transcriptase-polymerase chain reaction (RT-PCR) analysis of wnt9a (top panel) in median fins of zebrafish (5.6–6.3 mm). Columns 1 and 2 contain 1-kb and 100-bp ladders, respectively, with 1-kb and 500-bp bands marked at left. Column 9 shows RT-PCR results with no template added to the reaction mix (negative control). Columns 6–8, with 36, 50, and 76 hours postfertilization (hpf) whole embryos, and the bottom panel, with constitutively expressed ef-1α (Nordnes et al.,1994), were used as positive controls. Columns 3–5 contain the RT-PCR reaction product for caudal fin rays, caudal hypurals, and anal fin tissue. Note the absence of wnt9a RNA in both caudal fin rays (column 3) and caudal hypurals (column 4), and lack of amplification of the expected ∼440-bp fragment in anal fin tissue (column 5) that is observed in 36, 50, and 76 hpf embryos. Lanes 5–8 exhibit several bands between 1 and 1.5 kb. These could represent tissue-specific splice variants or nonspecific primer binding. No additional ef- 1α bands were observed (not shown).

Figure 9.

Reverse transcriptase-polymerase chain reaction (RT-PCR) analysis of wnt9b (top panel) in median fins of zebrafish (5.6–6.3 mm). Column 1 contains a 100-bp ladder with the 500-bp band marked at left. Column 6 shows RT-PCR results with no template added to the reaction mix (negative control). Columns 2, with 50 hours postfertilization (hpf) whole embryos, and the bottom panel, with constitutively expressed ef-1α (Nordnes et al.,1994), were used as positive controls. Columns 3–5 contain the RT-PCR reaction product for caudal fin rays, caudal hypurals, and anal fin tissue. Note the presence of wnt9b RNA in caudal fin rays (column 3), caudal hypurals (column 4), and anal fin tissue (column 5). No additional wnt9b or ef-1α bands were detected (not shown).

Subset 2: Genes Expressed in Mesenchymal Condensations and/or Cartilage Early; and Down-regulated in the Tetrapod Joint: Bapx1, Col2a1, Noggin, and Sox9


The dorsal and anal fin radials expressed bapx1from ∼10 dpf (5.2 mm) in the distal chondrocytes and possibly in the distal mesenchyme of the first developing anal fin radials (Fig. 10A). This early bapx1 expression was observed only in distal cells, not in proximal radials or throughout the unsegmented radials as observed for sox9a, noggin, and col2a1. As radials segmented, the distal-only radial cartilage expression pattern became more striking, with expression high in distal portions of proximal radials and low, if detected at all, in the newly formed distal radials (Fig. 10B, 5.6 mm). There was no expression in the ZS nor in fin rays (Fig. 10D,E). At the end of the first 2 weeks of zebrafish development, bapx1 expression remained very strong in the distal portion of the proximal radials but absent in the ZS. Expression then commenced in the centers of the distal radials (Fig. 10C) and expanded. At ∼1 month postfertilization, the dorsal and anal fins were fully patterned and ossifying; bapx1 expression in the radials was no longer detected (Fig. 10F).

Figure 10.

Sagittal sections showing expression of bapx1 in the median fins, head, and trunk of zebrafish. In all panels, anterior is to the left and dorsal to the top. Black patches are melanophores or delimit one of the eyes (G). Arrows mark fin rays, solid arrowheads mark radials, and caret marks mark neural arches. The brown fiber in the top center of C and blue dot in hypural 2 (I) are artifacts. A: In newly forming anal fin radials (5.2 mm), bapx1 expression is observed in the distal chondrocytes and perichondrial cells. B: As radial development progresses (5.6 mm), distal chondrocytes of the proximal radials (top row of arrowheads) continue to express bapx1. bapx1 is not expressed in the ZS (*), the newly formed distal radials (bottom row of arrowheads), or the newly developing fin rays (arrows). C: By 6.4 mm, the distal radials (bottom row of arrowheads) begin to exhibit bapx1 expression, although not at the levels observed in the distal portions of the proximal radials (top row of arrowheads). The ZS (*) does not express the gene. D: The distinct expression pattern of bapx1 is clearly apparent in this section (6.7 mm). Note expression in the distal chondrocytes of the proximal radials, lack of expression in the proximal chondrocytes of the proximal radials and the ZS, and an increase in expression in the distal radials. E: In this 6.8-mm specimen, distal radial bapx1 expression is stronger (bottom row of arrowheads); fin rays do not express the gene (arrows). F: ∼9.0 mm SL: bapx1 expression is no longer detectable in the dorsal and anal fin supports and axial skeleton (9.4 mm). G: At 4.5 mm NL, bapx1 expression in the anterior notochord (nc) is strong. Additional expression domains are observed in the developing posterior articular (aa) and ventral quadrate (q) bones, which together form the jaw joint. Expression is not observed in the anterior anguloarticular or dorsal quadrate, or in the hyoid arch (ha). H: Tissue surrounding the posterior notochord strongly expresses bapx1 (4.5 mm NL). The parhypural (p) and hypurals 1–3 are present, but do not express the gene. I: Posterior notochord expression remains strong at 5.2 mm. All five hypurals are present (2–5 shown); neither hypurals nor caudal fin rays (arrows) express bapx1. J: Neural and hemal arches (∧) and spines (not visible in this plane of section) and the posterior notochord express bapx1 at 6.7 mm. Note continued lack of expression in the fin rays (arrows) and caudal fin supports (curly brackets, inset). Scale bars = 0.5 mm in D,G,J, 0.05 mm in all others.

In the caudal fin, bapx1 was not expressed in hypural mesenchyme or in the newly forming parhypural, hypurals, or fin rays (Fig. 10H). Expression was high in the anterior and posterior notochord (Fig. 10G,H) and in developing centra and neural and hemal arches and spines. Posterior notochord expression was maintained as the last forming hypurals (4–5) developed (Fig. 10I). The general pattern of bapx1 expression in the posterior notochord, centra, and neural and hemal arches and spines was maintained for approximately three weeks (Fig. 10J). Throughout this period of development, bapx1 expression was never detected in the caudal fin hypurals or rays (Fig. 10J, arrows and inset).

In zebrafish, bapx1 expression is first observed in the jaw joint at 49.5 hpf (Miller et al.,2003); this served as our positive control. We observed that the expression persists through ∼6 dpf in the posterior articular and ventral quadrate cartilages (Fig. 10G).


Cartilage cells, perichondrial cells, and interradial mesenchyme of the dorsal and anal fins expressed col2a1 (Fig. 11). Expression was high in the cartilaginous radials and slightly lower in mesenchyme; the entire fin field was stained (Fig. 11A). This pattern was maintained as the fins developed (Fig. 11B). Where ossification began (centrally in radials), expression ceased. Expression was maintained in what would remain for several more days as an un-ossified band in the distal portion of the proximal radial (Fig. 11C, arrowheads). Alcian blue preparations showed that larger specimens (∼20 mm) retained cartilage in a band within the distal portion of the proximal radial. Expression of col2a1 was thus retained in a pattern that mirrors Alcian blue staining. In the caudal fin, expression was strong in the parhypural and hypural 1 at 4.0 mm NL (Fig. 11D). By 4.5 mm NL, some mesenchymal cells distal to the developing hypural cartilages and parts of the developing hemal and neural arches expressed col2a1. Expression was high in distal portions of the hypurals and was slightly lower in mesenchyme and in the middle portion of hypurals (Fig. 11E). Gene expression was absent from proximal chondrocytes and from mesenchyme between the proximal portions of the hypurals; these were presumably no longer proliferating. The proximal portions of posteriorly located hemal arches expressed col2a1. Epurals, part of the caudal fin complex dorsal to the notochord, were developing at this size and they expressed col2a1 as well. As caudal fin development progressed, col2a1 expression became more distally restricted (Fig. 11F). By 18.0 mm, expression remained in only a narrow distal band in each hypural (not shown).

Figure 11.

Sagittal sections showing expression of the type II collagen gene in the median fins of zebrafish. In all panels, anterior is to the left and dorsal to the top. Black patches are melanophores. A: Early expression in the anal fin field (5.1 mm). Both prechondrogenic mesenchyme and radial chondrocytes (arrowheads) express type II collagen. B: Type II collagen is expressed throughout the fin field of both dorsal and anal fins (5.7 mm). Stain of mature chondrocytes in the radials is turquoise in color while interradial mesenchyme stains purple (this is loosely correlated to relative gene expression levels, with higher levels appearing as turquoise; R. Harland, personal communication). C: After radial segmentation is complete, type II collagen expression remains in the distal radials and the bands of cartilage that remain unossified in the distal portion of the proximal radials (arrowheads; 12.2 mm). This pattern is observed in the dorsal fin as well (data not shown). A few cells in the most proximal portion of the proximal radials (top arrowheads) also express type II collagen. D: Strong expression in mesenchymal condensations of the parhypural (p) and hypural one (1; 4.0 mm NL). E: Stain of mature chondrocytes in the hypurals is turquoise in color while inter-hypural mesenchyme stains purple (5.7 mm). A developing hemal arch expresses type II collagen as well (∧). Fin rays do not express the gene (arrows). F: Type II collagen expression becomes increasingly distally restricted as the hypurals continue to grow to adult size (11.3 mm). G: Sense probe does not exhibit any staining. Scale bars = 0.1 mm in A,D,G, 0.5 mm in B,C.


Expression of noggin3 in condensing compact mesenchyme in the dorsal (4.7 mm NL) and anal fins (4.4 mm NL) was observed well before condensations could be detected by Alcian blue staining (Fig. 12A). Differentiated chondrocytes of the radials and the surrounding perichondrium expressed noggin3, while interradial mesenchyme cells never did (Fig. 12B). As segmentation progressed, noggin3 expression was down-regulated in the ZS (Fig. 12C). When segmentation of a radial was complete, noggin3 expression was completely absent in the ZS. Also at this stage, the central region of the proximal radial exhibited significantly less noggin3 transcript than the proliferative proximal tip and the broad distal end (Fig. 12C). Given that, in zebrafish, ossification is initiated in the center of the proximal radial at this stage (Bird and Mabee,2003), it is likely that the observed decrease in noggin3 expression was correlated with a shift from chondrocyte proliferation to hypertrophy before ossification. By 8.8 mm (Fig. 12D), noggin3 expression was restricted to chondrocytes in the centers of the distal radials and centers of distal portions of the proximal radials. At 9.2 mm, only small areas of single cell expression in the distal portions of the proximal radials of the anal fin still expressed noggin3 (not shown).

Figure 12.

Expression of noggin3 in the median fins of zebrafish. Anterior is to the left and dorsal is to the top. Black patches are melanophores. An “n” marks the notochord in the caudal fin (E–G). Arrows mark noggin3 expression in radials/hypurals, an arrowhead and p mark the parhypural. Hypurals are numbered in the order of their development (h1–h5). A: Initial expression of noggin3 in the dorsal and anal fins occurs in a series of pre-cartilage condensations that mark the sites of future anal radial development (4.5 mm NL). B: By 5.1 mm NL, radial chondrocytes are distinct and express noggin3 along the full length of each forming radial; surrounding perichondrial cells also express noggin3. C: When segmentation of the radial into proximal and distal pieces occurs (arrow pairs, 6.0 mm), noggin3 is down-regulated in the ZS (*) and in chondrocytes that are no longer proliferative. Both the distal portion of the proximal radial and the distal radial remain proliferative and strongly express noggin3 (arrow pairs). A small domain of proliferation and noggin3 expression remains at the proximal tip of the proximal radial (∧). D: By 8.8 mm, expression is detected in only a few chondrocytes (arrows) in the distal portion of the proximal radial and the distal radial; this ceases by ∼9.2 mm (not shown). E: Before ∼4 days postfertilization (dpf), no expression is observed in the caudal fin in the region below the notochord that will give rise to the caudal fin hypurals and parhypural (3.6 mm NL). F: Presumptive parhypural (p) and hypural (h1) condensations express noggin3 in a slightly older specimen (3.7 mm NL). G: Expression of noggin3 in the caudal fin becomes restricted to the posterior, proliferative chondrocytes of the hypurals (5.8 mm), and ceases at ∼9.5 mm (not shown). H: No cells are labeled by the noggin3 sense probe control (5.8 mm). Scale bars = 0.05 mm.

Caudal fin expression of noggin3 was first detected in condensations of the parhypural and hypural 1 at 3.7 mm NL (Fig. 12F); expression in this region was not observed in precondensation stages (Fig. 12E). The remaining four hypurals are added posteriorly, and noggin3 was expressed throughout their condensations (similar to Fig. 12B). As development progressed, noggin3 expression became increasingly restricted to distal hypural chondrocytes and uncondensed distal mesenchyme (Fig. 12G) and by ∼9.5 mm, expression in the caudal complex ceased (not shown). Sense probe exhibited no staining (Fig. 12H).

sox9a and sox9b.

RT-PCR revealed the presence of a predicted ∼760 bp sox9a fragment (Chiang et al.,2001b) in the caudal fin rays, caudal fin hypurals, and anal fin tissue from zebrafish larvae 5.6–6.3 mm (Fig. 13, top panel). In contrast, while a ∼830 bp sox9b fragment (Chiang et al.,2001b) was amplified by RT-PCR in the 50 hpf whole embryos used as a positive control (Fig. 13, column 2), it was not detected in the caudal fin rays, caudal hypurals, or anal fin tissue (Fig. 13, middle panel). Thus, sox9a was the prime candidate gene to examine by means of in situ hybridization.

Figure 13.

Reverse transcriptase-polymerase chain reaction (RT-PCR) analysis of sox9a (top panel) and sox9b (middle panel) in median fins of zebrafish (5.6–6.3 mm). Column 1 contains a 100-bp ladder with appropriately sized bands marked at left. Column 2, with 50 hours postfertilization (hpf) whole embryos (Chiang et al.,2001b), and the bottom panel, with constitutively expressed ef-1α (Nordnes et al.,1994), were used as positive controls. Columns 3–5 show presence and relative levels of amplified RNA for caudal fin rays, caudal hypural tissue and anal fin tissue, respectively. While sox9a is strongly expressed in the sampled median fin tissues, sox9b is not detected at all, although it is present in 50 hpf whole embryos, as expected (Chiang et al.,2001b). Column 6 shows RT-PCR results with no template added to the reaction mix (negative control).

In the anal and dorsal fins, sox9a expression was first detected at 4.3 mm NL in two to three patches in positions that corresponded to the future location of the first two to three radials (Fig. 14A). The expression was diffuse, and the cells expressing sox9a lacked cuboidal chondrocyte morphology. As development progressed, cells rounded up, exhibited cuboidal chondrocyte morphology, and continued to express sox9a (Fig. 14B,C). Mesenchymal expression was maintained and was very strong in “balls” of cells distal to the as yet unsegmented radials, and in the interradial mesenchyme along the distal third of the radials (Fig. 14B). As radial segmentation occurred, sox9a expression was maintained in the proximal radials and in the newly formed distal radials (Fig. 14D,E, arrowheads), but it was down-regulated in the ZS (see * in Fig. 14D,E). Fin rays formed as radials segmented, but sox9a was not detected in them (not shown). As fin development progressed, sox9a expression remained strong in the chondrocytes of the proximal and distal radials, but was largely lost in the distal and interradial mesenchyme, and remained absent in the ZS and fin rays (Fig. 14F,G).

Figure 14.

A–L: Expression of sox9a in a developmental fin series (A–J; 4.0 mm NL, 11.5 mm) and head and early pectoral fin (K,L; positive controls).A: Mesenchymal condensations that will give rise to the first two to three anal fin radials express sox9a before cartilage differentiation (4.3 mm NL). B: The stacked chondrocytes of the dorsal fin radials express sox9a (bottom row of arrowheads). Expression is very strong in “balls” of distal mesenchyme (top row arrowheads) that presumably will be incorporated into developing radials (5.3 mm). C: Stacked chondrocytes of the anal fin radials strongly express sox9a, as does distal mesenchyme (not shown in this plane of section, but see B; 5.4 mm). D,E: In both the anal (D) and dorsal (E) fins, as distal radials (D, bottom arrowheads; E, top arrowheads) segment from proximal radials (D, top arrowheads; E, bottom arrowheads), cells in the zone of segmentation (ZS; see *) exhibit down-regulation of sox9a expression. Expression in anal and dorsal fin rays is not detected (fin rays not visible in these planes of section). F: Segmentation is complete. Chondrocytes in distal and proximal radials (arrowheads) continue to express sox9a, while it is not detected in the ZS (*) or in fin rays (arrows; 6.4 mm). G: As radial ossification progresses, expression of sox9a in dorsal and anal fins tapers off in zebrafish > 1 month postfertilization (11.5 mm). H: Below the pre-flexion notochord (nc), hemal spine 2 (hs2), the parhypural (p), and hypurals 1–3 are present and express the gene. Additionally, mesenchyme anterior to hemal spine 2, posterior to hypural 3, and distal to each caudal cartilage condensation strongly expresses sox9a (4.0 mm NL). I: Caudal fin expression is maintained in distal mesenchyme and distal proliferative chondrocytes of the hypurals (1–5), parhypural (p), and hemal spine 2 (hs2), and is observed in proximal chondrocytes of the parhypural and hemal spine 2 (arrowheads). Intermediate regions of the caudal fin supports have begun to ossify (Bird and Mabee,2003) and do not express sox9a. Expression in the caudal fin rays is not detected by in situ hybridization (arrows; 5.9 mm). J: In the ossifying caudal fin, expression is no longer detected in hypurals, although it is observed in the developing uroneural (uro) and epural (e; 11.5 mm). K: Expression of sox9a is observed in all cranial cartilages: shown here are the dentary (d), premaxilla (pm), palatoquadrate (pq), and branchial arches (hollow arrowheads). Additional domains of expression are seen in anterior notochord (nc) and in neural cells of the hindbrain (hb) and below the frontals (f) (4.0 mm NL). L: Chondrocytes in early pectoral fin (P1) strongly express sox9a (positive control; Yan et al.,2002). Scale bars = 0.1 mm.

In the caudal fin, sox9a was expressed in the chondrocytes of the hypurals, parhypural, and hemal spines and in mesenchyme distal to each cartilage condensation by 4.0 mm NL (Fig. 14H). Additionally, distinct patches of mesenchyme anterior to hemal spine 2 and posterior to hypural 3 (Fig. 14H, arrowheads) strongly expressed sox9a; these mesenchymal condensations corresponded to the future locations of hemal spine 3 and hypural 4, respectively. The caudal fin hypurals, which do not segment, continued to express sox9a in distal, proliferative chondrocytes (Fig. 14I). Additionally, mesenchyme distal to the more recently formed hypurals (3–5) maintained sox9a expression. The forming epural (not shown) and the proximal portions of the parhypural and hemal spine 2 (see arrowheads in Fig. 14I) expressed sox9a as well. The caudal fin rays did not appear to express sox9a. By ∼11.0 mm (or ∼1 month postfertilization), sox9a was down-regulated in almost all fin tissue (Fig. 14G) and was maintained only in the distal-most end of the epural and in the developing uroneural (Fig. 14J).

Expression of sox9a was observed in all of the cranial cartilages present in sections including, for example, the branchial and mandibular arches (Fig. 14K) and the epiphyseal bar and basioccipital (data not shown). Specific regions of the central nervous system expressed sox9a: for example, neural cells underlying the frontals and others in the hindbrain region (Fig. 14K). Expression also was high in the anterior tip of the notochord (Fig. 14K) and in cartilage cells of the pectoral fins (Fig. 14L) of 4–10 dpf larvae; pectoral fin and cranial cartilage expression served as positive controls (as per Yan et al.,2002,2005) when entire larvae were embedded and sectioned.


There is considerable correspondence in timing (Fig. 15) and pattern (Table 1) of expression during segmentation between tetrapod limb joint and zebrafish median fin candidate segmentation genes, and this finding supports the hypothesis that the genetic programs underlying median fin development may have been co-opted in the evolution of paired fins. In the process of segmentation of the radials, bmp4, chordin, and gdf5 are co-expressed in the interradial mesenchyme and bmp2b (Crotwell et al.,2004), chordin, and gdf5 (Crotwell et al.,2001) are co-expressed in the ZS, similar to tetrapod patterns (Table 1). Additionally, bapx1, col2a1, noggin3, and sox9a are co-expressed in the radial chondrocytes: they are down-regulated (col2a1, noggin3, sox9a) or never expressed (bapx1) in the ZS (Table 1). shh is broadly expressed in fin mesenchyme and rays, but is not expressed in chondrocytes. wnt9b, but not wnt9a, is detected in zebrafish median fin radials by RT-PCR, but not by in situ hybridization.

Figure 15.

Summary of onset and duration of gene expression in the dorsal and anal fins. Question marks indicate uncertainty either because expression is weak (chordin), or data were collected from an incomplete size series (shh) or reverse transcriptase-polymerase chain reaction (RT-PCR) only (wnt9, dashed line represents wnt9b in zebrafish, vs. wnt9a in tetrapods and paddlefish [paddlefish data not shown]). On the x axis are three cartoons representing, from left to right, the first appearance of anal fin radial condensations, first segmentation of anal fin radials, and the size at which anal fin segmentation is complete. Onset data are derived from the anal fin, which develops slightly before initiation of dorsal fin development. The earliest expression frequently was observed in the caudal fin, but because it does not segment, those data are not included here. Offset points are derived from combined anal and dorsal fin data. Data for gdf5 and bmp2b have been previously reported (Crotwell et al.,2001,2004). Expression of col2a1 is maintained until at least 18 mm, marked by a + on the chart.

Table 1. Comparison of Expression Patterns in Zebrafish Radials vs. Chick or Mouse Digits, and Zebrafish Interradial Mesenchyme and Zones of Segmentation (IR&ZS) vs. Chick or Mouse Interdigital Mesenchyme and Zones of Segmentation (ID&ZS)
 Zebrafish radialsChick digitsMouse digitsZebrafish IR&ZSChick ID&ZSMouse ID&ZS

Differences between zebrafish and chick/mouse gene expression were observed for bapx1, shh, and wnt9a. Zebrafish bapx1 and chick Bapx1 patterns are similar (Church et al.,2005; Crotwell,2005). However, it is unclear whether mouse digit Bapx1 expression is restricted to cartilage cells at the distal edge of each phalanx, or if the expression extends into the developing joint (see Fig. 7J,K and associated text in Tribioli et al.,1997). Church and coauthors (2005) describe the difficulty of detecting cartilage Bapx1 expression in cartilage in whole-mount chick in situ hybridizations, particularly at later stages of development. They performed hybridizations on sectioned tissue to describe Bapx1 expression in the developing chick limbs. It may be that the expression data presented by Tribioli et al. (1997) represent an incomplete picture of mouse limb Bapx1 expression, given that the data were collected from whole-mount specimens. The small amount of data collected from human embryos indicates an expression pattern similar to that depicted for chick (Tribioli and Lufkin,1997).

While the importance of Shh to vertebrate limb development cannot be underestimated, several functional studies (Chiang et al.,2001a; Kraus et al.,2001; Lettice et al.,2003; Sanz-Ezquerro and Tickle,2003) suggest that Shh is involved in patterning the digit field (i.e., changing the number of digits or phalanges), but not digit patterning itself (i.e., joint development). We found that, in zebrafish, shh is expressed in the mesenchyme surrounding the developing cartilages, a pattern that is predicted by previous studies showing that Shh acts in two ways. First, and relevant to our hypothesis, Shh is expressed in and defines the region of the limb bud that is competent to form digits (and we predict, radials; Drossopoulou et al.,2000). Second, Shh induces and maintains Bmp2 expression, and it is Bmp2 that subsequently patterns the digits (Drossopoulou et al.,2000) and perhaps radials. Thus, broad limb bud (or fin) regions are set up early, and subsequent segmentation and patterning are distinct from that earlier process (Richardson et al.,2004). A similar uncoupling of regionalization and segmentation occurs in the vertebral column (Richardson et al.,2004).

The absence of wnt9a in zebrafish median fins leads to the possibility that, in the evolution of wnt9 in teleosts, a segmentation function of wnt9a has been co-opted by wnt9b. This hypothesis is supported by the expression of wnt9a in regions of skeletal segmentation in the paddlefish paired fins (Crotwell,2005). However, to our knowledge, a wnt9b gene has not been cloned from paddlefish or other basal ray-finned (actinopterygian) fishes (Fig. 2); thus, this hypothesis awaits further testing. Additionally, future work on the expression pattern of wnt9b in zebrafish and wnt9a in the paddlefish in comparison to tetrapods (Hartmann and Tabin,2001; Guo et al.,2004) may provide critical insight into the difference between zebrafish and chick/mouse that we observed (Table 1). Such comparative research on homology and function of the Wnt gene family is an avenue of considerable interest.

Together, our data, despite the minor differences described above, suggest that the molecular mechanisms underlying skeletal segmentation have been conserved over 400 million years of evolution. Given that the median fins evolved before the paired fins (Fig. 2), and the precise onset of segmentation is unclear (although skeletal segmentation first evolved in the gnathostomes), it may be that a single segmentation gene network is widely used by different regions of the skeleton. Certainly, it is possible that a similar segmentation network could be independently assembled to pattern later evolving structures (i.e., digits), although de novo network construction appears to be rare after an initial network has been assembled (Ganfornina and Sanchez,1999).

We propose that radial development is at least a three phase process. First, the number and position of radials is determined, possibly by means of hox gene patterning (Mabee et al.,2002) as has been described for shark (Freitas et al.,2006). This likely happens before any visible sign of patterning, such as mesenchymal condensations. Next, condensation and chondrogenesis proceed, and shh, sox9a, col2a1, as well as bmps and their inhibitors, are involved in the processes of cartilage cell differentiation and proliferation. Finally, the patterning of the radial itself, i.e., segmentation, occurs. Changes in cell adhesion and dedifferentiation of chondrocytes to mesenchymal cells or other cell types in or adjacent to the ZS must occur in concert with a mechanism that determines the specific location of the ZS (Spater et al.,2006a; Iwamoto et al.,2007). This finding might be a threshold phenomenon related to the thickened end of the unsegmented radial, or it might be related to the distal coexpression of gdf5, chordin, and bmp4 in the interradial mesenchyme adjacent to the ZS.

The zebrafish caudal fin essentially serves as an internal control for hypotheses of segmentation, because the hypurals do not segment. We found that the caudal hypurals, which do not segment, do not express bapx1. This finding serves as additional support for a role in segmentation for bapx1. All other examined genes were expressed in all three (caudal, anal, dorsal) median fins. If bapx1 is necessary for segmentation, misexpressing bapx1 in the caudal fins would be predicted to induce hypural segmentation. Similarly, inhibiting bapx1 in the segmenting dorsal and anal fins would be predicted to inhibit their segmentation. This strategy is a potentially useful direction for future research.

Finding similarity in expression of a single gene would have been insufficient to support our hypothesis that similar (or the same) molecular mechanisms generally underlie vertebrate skeletal segmentation. However, given the similarity of expression of col2a1, gdf5, noggin3, chordin, bapx1, bmp2b, bmp4, and sox9a, in both zebrafish median fin development and tetrapod limb development, it is likely that a segmentation network has been conserved between the axial skeleton (median fins) and the appendicular skeleton (limbs). This may extend to the visceral skeleton as well, given the expression and function of two of those same genes (e.g., bapx1, gdf5) in the pharyngeal arches of zebrafish (Miller et al.,2003), the chick jaw (Wilson and Tucker,2004), and mouse middle ear (Tucker et al.,2004). Details regarding the appearance of the first joint are unclear, but segmentation was widespread in early gnathostomes (Janvier,1996), and it seems likely that the molecular mechanisms underlying joint development (Miller et al.,2003) and skeletal segmentation in general have been conserved.

The complexity of gene regulation and interactions in skeletogenesis is widely acknowledged (Koyama et al.,1995). Investigation of the ancestral homologue of joint formation is likely to be simpler, and may help dissect out the newly evolved mechanisms involved in such processes as joint cavitation. The vertebrate limb is one of the classic anatomic and evolutionary systems, and it is an increasingly well-understood model system of comparative developmental genetics (Coates,1994; Shubin et al.,1997; Coates and Cohn,1998; Arthur et al.,1999; Herault et al.,1999; Tabin et al.,1999). This background combined with recent work on fin and limb positioning mechanisms (Freitas et al.,2006) and our data on molecular mechanisms underlying skeletal segmentation of the fish skeleton together provide an excellent foundation from which to answer long-standing and broadly significant questions concerning the origin and evolution of the basic processes underlying fin and limb development and segment/joint formation.



Fish length, not age, is used to describe development of older specimens herein, as length is the best predictor of stage of skeletal development (Cubbage and Mabee,1996; Mabee and Noordsy,2004). Notochord length (NL) is measured from the anterior tip of the upper jaw to the posterior tip of the notochord; after the notochord flexes, standard length (SL) is measured from the anterior tip of the upper jaw to the posterior-most edge of the hypural complex of the caudal fin (Ahlstrom and Moser,1976). All lengths reported are standard length unless indicated otherwise.

In Situ Hybridization to Histological Sections

Zebrafish embryos were obtained and maintained as described in Westerfield (2000). Developmental age is given as hours postfertilization (hpf), days postfertilization (dpf), or size in millimeters. In specimens larger than 5.5 mm, the trunk was dissected from the cranium, shoulder girdle, and pectoral fins and embedded in paraffin for collection of serial sagittal sections (7–8 μm) on silane- or aminoalkylsilane-coated slides (Polysciences, Warrington, PA). In specimens smaller than 5.5 mm, the entire larva was embedded. Slides were hybridized with 50–150 ng of digoxigenin-labeled RNA probes in 100 μl of hybridization buffer according to standard protocol (Strähle et al.,1994). After hybridization, slides were examined using Zeiss Axioplan2 and Wild M5 microscopes. Images were captured using a Dage MTI DC-330 camera and Scion image (1.62c) capturing software and were processed using Adobe Photoshop 5.0. Original, un-retouched images were made available to reviewers upon request. In all in situ images, unless stated otherwise, anterior is to the left, and dorsal is to the top.


For use as positive controls, total RNA was extracted from ∼50 each whole 36, 50, and 76 hpf zebrafish embryos using Qiagen's RNeasy Mini Kit (catalog no. 74104) according to the manufacturer's protocol. Caudal fin ray, caudal hypural, and anal fin tissue was obtained from eight anesthetized, measured, zebrafish larvae between 5.6 and 6.3 mm and was immediately transferred to Ambion's RNAlater RNA Stabilization Solution (catalog no. 7024). Both the caudal hypural tissue and the anal fin tissue consisted of cartilage plus proximal fin rays, which could not be separated from the cartilages during dissection. Tissue types were combined in labeled vials containing fresh RNAlater; vials were stored at 4°C until RNA extraction (<2 weeks).

Total RNA was extracted from each tissue type using Ambion's RNAqueous-Micro Kit (catalog no. 1927) according to the manufacturer's protocol. Detection and amplification of the selected genes in each tissue type was performed using Invitrogen's SuperScript III One-Step RT-PCR System (Cat. No. 12574-026), following the protocol suggested by the manufacturer. Briefly, after cDNA synthesis at 55°C for 30 min, 40 cycles of 94°C for 15 sec, the appropriate annealing temperature for 30 sec, and 68°C for 1 min were used to detect and amplify each gene. Primer sequences were obtained from previously published work or were designed from DNA sequences using the DNASTAR primer design program (Windows 32 PrimerSelect v. 4.05).

Our zebrafish wnt9a fragment was cloned from zebrafish genomic DNA. Four degenerate oligonucleotides were prepared (by C. Hartmann, currently of the Research Institute of Molecular Pathology, Vienna) from conserved DNA sequences within exon IV of human, mouse, and chick Wnt9a and were used for PCR amplification. The forward (5′-GTGGGCAGCACYACCAAYGARGC-3′ and 5′-TGCACSGTSCGRA-CSTGCTGG-3′) and reverse (5′-GCYTGCAYTCCACRTAGCAGCACC-3′ and 5′-CARCGSAC-CTGGCAYTGGCAHGG-3′) oligonucleotides were used to produce several fragments, the largest of which was a ∼390-bp fragment whose sequence is posted on GenBank, accession DQ490071.

Confocal Microscopy

Confocal microscopy was used to examine shh expression in a shh:gfp (green fluorescent protein driven by the shh promoter) transgenic zebrafish line. Brightfield and fluorescent images were collected as z-stacks and as single plane-of-focus .tiff files using the helium neon green channel of an Olympus BX61 confocal microscope (Flouview v. 3.3 software). After the image files were transferred from the confocal computer to a personal computer (Dell Dimension 4550, Windows XP), we used the public domain image processing software program ImageJ ( to open the z-stack files. Appropriate image slices from the z-stack files were selected to illustrate shh expression patterns. The original z-stack and .tiff files are available upon request.


The zebrafish wnt9a fragment we used initially was cloned from genomic DNA under the direction of Christine Hartmann using primers she designed; our fragment sequence can be found on GenBank under accession no. DQ490071. We thank Macie Walker and Chuck Kimmel (University of Oregon) for sharing their zebrafish wnt9a and wnt9b full-length sequences with us. We are grateful for plasmid gifts from H. Bauer, S. Fisher, M. Fürthauer, M. Hammerschmidt, C. Kimmel, C. Miller, J. Postlethwait, D. Raible, D. Stock, N. Ueno, M. Walker, E. Weinberg, S. Wise, and Y. Yan. We thank Francis Day (confocal microscopy) and Rolf Karlstrom (shh transgenic zebrafish). Funds for this research were provided in part by grants in aid of research to P.L.C. from Sigma Xi, the Society of Integrative and Comparative Biology, and the University of South Dakota. Additional funds and equipment were obtained from grants to P.M.M. from the National Science Foundation/EPSCoR and the State of South Dakota.