dolphin corresponds to cyp26b1. (A-D) Live zebrafish larvae at 120 hpf. Dorsal views of head (A,B) and on pectoral fins (pf; C,D). wt, wild type; dol, dolphin allele ti230g. (E)The dol-bearing interval on chromosome 7. (F) Interval-spanning BAC contig, with indicated recombinations in 4500 meioses and showing the location of cyp26b1 gene. (G) Schematic of wild-type and mutant zebrafish Cyp26b1 proteins. (H) Sequencing profiles of cyp26b1 exon 3-intron junction from genomic DNA of wild type (+/+), heterozygous (+/-) and homozygous (-/-) ti230g mutants. (I) Schematic of exon 3-intron junction of cyp26b1. The mutated G in the splice-donor site of the ti230g allele is in red, the internal GT used in the mutant is in bold, and the 7 bp insert in mutant cDNA is in light gray. (J) Sequencing profiles of cyp26b1 exon 3-exon 4 junction from cDNA of wild type and ti230g mutants. The inserted sequence is underlined.

cyp26b1 is expressed in condensing chondrocytes and perichondrium. Stainings of wild-type zebrafish at the stages indicated in the upper right corners and with the in situ RNA probes or antibodies indicated in the lower right corners. fli1a-positive neural crest cells in D were visualized by anti-GFP immunostaining of a Tg(fli1a:EGFP)y1 transgenic embryo (Isogai et al., 2003). (A,B,D,E,H,I) Lateral views; (C,L-Q) dorsal views; (F,J) longitudinal sections; (G,K) transverse sections. (A-D) cyp26b1 is expressed close to, but not within, postmigratory cranial neural crest (CNC) cells. Arrows in A,B point to dlx2a-positive, cyp26b1-negative cells that according to lineage-tracing data are likely to give rise to the ethmoid plate (e). Arrows in C,D point to two cyp26b1 domains close to the pharyngeal arch-forming CNC. (E-H) cyp26b1 is expressed in chondrogenic mesenchymal condensations. (F,G,H) Magnifications of regions indicated in E. In F,G, pharyngeal endoderm is counterstained with zn5 antibody; in H, neural crest derivatives are stained for sox9a transcripts; double-positive domains in pharyngeal arch (pa) condensations are marked. (I-K) cyp26b1 is expressed in perichondrial cells around the pharyngeal arches (J), the ethmoid plate (K), and along the entire length of the ceratohyal (ch) and the ceratobranchials (cb) (I; inset shows magnification of one cb). (L-Q) By contrast, osx (L), opn (M) and col10a1 (N) are restricted to perichondrial cells around the ossifying, Alizarin Red (alR)-positive region of the ceratohyal (O). In col10a1-positive cells, cyp26b1 levels are lower than in the adjacent col10a1-negative perichondrium (P,Q; arrows point to cells double positive for cyp26b1 and col10a1). spt, subpallial telencephalon; vt, ventral thalamus.

cyp26b1 is expressed in osteoblasts and their precursors. Stainings of wild-type zebrafish at the stages indicated in the upper right corners and with the in situ RNA probes indicated in the lower right corners. (A-F) Confocal sections of double fluorescent in situ hybridizations and Alizarin Red stainings (alR), counterstained with DAPI (blue). Cells with weak cyp26b1 and strong col10a1 expression are indicated with white arrows, cells with strong cyp26b1 but absent col10a1 expression with red arrows, and cells with strong col10a1 but absent cyp26b1 expression, which most likely represent fully mature/active osteoblasts, with green arrows. See text for details. For single-channel images of B,E, see Fig. S4 in the supplementary material. (G-I) cyp26b1 displays uniform expression in perichordal cells in anterior regions of the notochord (n) (G; left panel shows longitudinal section; right panel shows transverse section; counterstained with Eosin), and metameric expression in the trunk (H,I; lateral views). In H, cyp26b1-positive cells are (still) underneath the notochord (arrowhead) and others are (already) in perichordal positions (arrows), whereas in I all cells are perichordal (arrows). Positive cells dorsal of the notochord in H most likely represent ventral spinal cord neurons (scn). (J-O) col10a1 and opn show a similar expression pattern to cyp26b1 (L-N) and transient coexpression with cyp26b1 (J,K) at intersomitic borders, coincident with the anterior borders of the the Alizarin Red-positive vertebral bodies (O). Arrows in L-N point to positive cells, arrowheads to borders of somites 5-8. (P) Numbers of perichordal cyp26b1-, col10a1- and opn-positive cells at different developmental time points. Ten fish were evaluated per condition; standard errors are indicated.

cyp26b1 mutants and morphants display deficiencies in midline cartilages of the neurocranium and visceral skeleton. All panels show ventral views of zebrafish head regions. (A-D) col2a1 in situ hybridization at indicated stages. (E-N) Alcian Blue stainings of cartilaginous craniofacial elements at 120 hpf. Pharyngeal arches are numbered (1, mandibulare; 2, hyoid; 3-7, branchial/gill arches 1-5). (E-G) Overviews of visceral skeleton. (H-K) Magnified views of ceratohyals (H,I) or pharyngeal arches 4-6 (J,K). Arrows in H,I point to ceratohyal (ch) attachment in midline. (L-N) Flat-mounts of neurocranium, revealing the absence of medial ethmoid (e) and anterior basicranial commissure (abc) in mutant and morphant. anc, chondrocytes of anterior neurocranium; bb, basibranchial; bh, basihyal; cb, ceratobranchials; m, Meckel's cartilage; n, notochord; pq, palatoquadrate; t, trabeculae cranii.

cyp26b1 mutants and morphants display increased ossification of endochondral and intramembranous craniofacial bones. (A-C) Lateral views and (D-F) ventral views of zebrafish larval heads after staining of ossified matrix with Alizarin Red at indicated ages. Insets in A-C show hyomandibula (hm; dorsal element of arch 2) of larvae of same genotype stained with Alcian Blue at 120 hpf. Mutant and morphant show an opercle (op) of increased size, whereas the hyomandibula fails to ossify, although its cartilage model is properly formed (insets). A similar combination of gain of opercle and loss of hyomandibula ossification has previously been described for endothelin mutants (Kimmel et al., 2003), possibly reflecting a morphogenetic effect of signaling to pattern ossification along the dorsoventral axis of the second arch and its associated elements. (D-F) Endochondral ossification within the ceratohyal (ch) is much more advanced in the mutant (E), comparable to the situation in a wild-type sibling 2 days later (F).

cyp26b1 mutants and morphants display precocious and increased ossification of vertebral centra in the developing vertebral column. (A-K) Alizarin Red stainings of zebrafish larvae of genotype indicated in the upper right corners and at stages indicated in the lower right corners. (A-F,I-K) Lateral views; (G,H) transverse sections. Arrows in A indicate time course of centra ossification, starting at centra 3/4. In contrast to the segmented ossification of the wild-type larva (A,D,I,K), the cyp26b1 mutant (C,J) and morphant (F) show uniform and caudally extended perichordal ossification. See text for details. bop, basioccipital articulatory process.

Retinoic acid (RA) deficiency reverts, whereas RA excess phenocopies, the skeletal alterations of cyp26b1 mutants, with a pattern of axial hyperossification different from that caused by Bmp2b overexpression. Genotypes are indicated in upper right corners, treatments in lower left corners. (A-D) Flat-mounts of Alcian Blue-stained neurocrania after treatment from 24-50 hpf; 120 hpf, dorsal views. Note the absence of the medial ethmoid plate (e) in the RA-treated wild-type zebrafish (B), and the recovery of this structure in the DEAB-treated cyp26b1 mutant (D). (E-H) Alizarin Red-stained heads after treatment from 24-50 hpf (F) or from 96-180 hpf (G,H); 180 hpf, dorsal views. Late (G), but not early (F), RA treatment phenocopies hyperossification of craniofacial bones, whereas late DEAB treatment reverses the mutant phenotype and causes delayed ossification (H). (I-T) Alizarin Red-stained centra after treatment with RA (J,K,N), R115866 (L), DEAB (Q,R) or heat shock (hs; O,P) at indicated developmental stages, or after aldh1a MO injection (T); 180 hpf(I-L,O-T) or 360 hpf (M,N; vertebrae numbers indicated); lateral views. In I-P, the early-specifiying centra at the level of somites 3-6 are indicated by a bar. Inset in P shows precocious and unsegmented perichordal mineralization at somite levels 18-26 of the same bmp2b transgenic animal. See text for details. cb5, ceratobranchial 5 (with teeth); cl, cleithrum; den, dentary; max, maxilla; ps, parasphenoid; see also Figs 4, 5, 6.

Loss of cyp26b1 and gain of Bmp signaling have different effects on the number and/or activity of osteoblasts. (A-K) In situ hybridization of zebrafish larvae of genotype indicated in upper right corners and with probes and at stages indicated at lower right corners. (A-I) Lateral views; (J,K) dorsal views. (A) Entire head; (B-E) opercle; (F-K) trunk at level of somites 6-10. In A, stronger cyp26b1 expression is seen in all craniofacial skeletogenic elements of the cyp26b1 mutant, but not in the dorsal brain. In G, perichordal opn-positive cells of the cyp26b1 mutant have largely given up their metameric distribution. (L) Average increase in the number of axial cyp26b1-(at 96 hpf), opn- or col10a1-positive cells (at 144 hpf) in the trunk/tail region of cyp26b1 mutants and heat-shocked Tg(hsp70:bmp2b) transgenic fish. Control wild-type (wt) siblings were set to a value of 1. Ten fish were counted per condition; standard errors are indicated. n, notochord.

Similar effects of sa0002 and ti230 mutations and rescue of ti230 mutant by inducible application of wild-type cyp26b1. (A-H) Alcian Blue staining of craniofacial cartilaginous elements at 120 hpf (A-D) and Alizarin Red staining of bone matrix in axial skeleton at 180 hpf (E-H). (A,E) Wild-type siblings; (B,F) ti230g mutants; (C,H) sa0002 mutants. The two alleles show craniofacial and axial skeletal defects of undistinguishable strengths. (D,H) ti230 mutants after reintroduction of wild-type Cyp26b1 at 96 hpf, leading to a loss of vertebral ossification (H), whereas patterning of midline craniofacial cartilage, which occurs earlier (24-50 hpf) (compare with Fig. 7A-D), remains unaltered (D). For cyp26b1 overexpression, the heat-inducible construct pTol2-hse-GTP/cyp26b1 was generated by cloning the cyp26b1 cDNA into the bicistronic vector pSGH2 (Bajoghli et al., 2004), which in addition to cyp26b1 drives expression of GFP under the control of heat-shock elements (hse). Subsequently, the cassette was ligated into vector pT2AL200R150G, which contains Tol2 recognition sites to allow early genomic integration and widespread expression (Urasaki et al., 2006). pTol2-hse-GTP/cyp26b1 plasmid DNA was co-injected with transposase mRNA (Kawakami et al., 2004) into the cytoplasm of one-cell stage embryos from a ti230/+ intercross. For transgene activation, injected embryos were transferred from 28°C to 39°C for 30 minutes at 96 hpf. (I-L) Cyp26b1 carrying the sa0002 or ti230 mutation is biologically inactive. For activity tests, the cDNA of human CYB26B1 was amplified by RT-PCR and cloned into plasmid pCS2+. Nonsense mutations at nucleotides 136 (AAGrTAG) or 697 (TACrTAA) were introduced by PCR-based site-specific in vitro mutagenesis, yielding truncated proteins corresponding to those encoded by the zebrafish sa0002 and ti230g alleles, respectively. Plasmids were linearized with NotI, mRNAs synthesized with the Message Machine Kit (Ambion, TX) and co-injected with fluorescein dextrane (Molecular Probes) into one blastomere of wild-type embryos at the two-cell stage, as described previously (Kudoh et al., 2002). At the 90% epiboly stage (9 hpf), embryos with unilateral fluorescence were selected and fixed for in situ hybridization analysis. Whole-mount in situ hybridization for the anterior neural marker otx2 (Li et al., 1994) and the posterior neural marker hoxb1b (Alexandre et al., 1996); 90% epiboly stage, dorsal views, anterior up. To identify the injected side, some embryos were counterstained with fluorescein-coupled anti-fluorescein antibody (Molecular Probes, 1:200; not shown). (I) Uninjected sibling (n=90/90). (J) Embryo injected with wild-type CYP26B1 mRNA on right side (n=25/27). (K) Embryo injected with ti230 CYP26B1 mRNA (n=30/30). (L) Embryo injected with sa0002 CYP26B1 mRNA (n=35/35). Wild-type CYP26B1 causes loss of hoxb1b expression on the injected side, indicative of an anteriorizing effect caused by Cyp26b1-mediated RA inhibition (Kudoh et al., 2002). By contrast, neither of the mutant versions displayed any effect, suggesting that the encoded truncated proteins are inactive.

Expression pattern of cyp26b1 in comparison to cyp26a1, cyp26c1 and aldh1a, and in pectoral fin buds. (A-I) Wild-type zebrafish after whole-mount in situ hybridizations at stages indicated in upper right corners with probes indicated in lower right corners. (A-C,E) Lateral views; (D) ventral view; (F-I) ventrolateral views. (A-C) cyp26a1, cyp26b1 and cyp26c1 show distinct, but partially overlapping expression patterns. In B, unique expression of cyp26b1 in the condensation of the ethmoid plate is indicated by an arrow. (D-E) aldh1a is expressed adjacent to the eye vesicles (red arrows in D,E), lateral of the cyp26b1 expression domain in the midline of the developing neurocranium (blue arrow in D). (F,G) aldh1a is expressed in discrete domains lateral of the cyp26b1 expression domains in the branchial arches; the midline is indicated by arrowheads. Together, this points to the possible existence of a mediolateral RA gradient in the craniofacial system. (H,I) Similar to the earlier exclusion of cyp26b1 expression from chondrogenic neural crest cells of the craniofacial system (see Fig. 2), cyp26b1 is expressed in a distal crescent of the early pectoral fin bud, while the sox9a-positive chondrocyte precursors (Wada et al., 2005) are positioned more proximally (H). A corresponding distribution has been described for mouse Cyp26b1 in the developing limb buds (Yashiro et al., 2004), suggesting that similar proximodistal patterning processes might occur in both zebrafish and mouse, accounting for the profound limb defects in mouse Cyp26b1 mutants (Yashiro et al., 2004), and the moderate pectoral fin defects in zebrafish cyp26b1 mutants (see Fig. 1C,D). e, ethmoid plate; pf, pectoral fin; pp, pharyngeal pouches.

Expression pattern of cyp26b1 in craniofacial bone elements in comparison to opn and osx. Stainings of wild-type animals at stages indicated in upper right corners, and with in situ RNA probes indicated in lower right corners. (A-C) cyp26b1 and opn are coexpressed in the opercle. The opn-positive cells in A display weak cyp26b1 expression that is not visible in this image (but compare with Fig. 3A); the opercle visible in B is overstained, showing similar signals in strongly and weakly expressing cells. In C, cells with cyp26b1 and opn coexpression are indicated with black arrow, cells positive for cyp261 or opn mRNA only with blue or red arrows, respectively. (D-G) Confocal sections of double fluorescent in situ hybridizations, counterstained with DAPI to visualize nuclei (blue). In left panels, col10a1 expression is in green; in middle panels, osx, opn or cyp26b1 expression is in red; in right-hand panels, merged green and red channels, superimposed by DAPI staining. (D-F) col101a1, osx and opn are coexpressed in the opercle. At 72 hpf, osteoblasts in central versus peripheral regions of the elements display slightly different signatures. opn expression is highest in regions close to the bony core of the element (E; compare with Fig. 3C), col10a1 expression is strongest in intermediate positions (E), whereas osx is strongest in peripheral regions (D). This suggests that opn is preferentially made by the most mature/most active osteoblasts, whereas osx marks more immature osteoblasts. cyp26b1 seems to be most strongly expressed in even less advanced osteoblasts. At 72 hpf, it displays strong expression at the dorsal base of the element (see Fig. 3A), which at that time is negative for osx, col10a1 and opn (indicated by white arrows in D,E). By contrast, col10a1 and opn are expressed in these dorsal cells at 120 hpf (indicated by white arrow in F). Also note that the location of osteoblasts generally shifts to more peripheral positions as the central bony core of the element grows (compare E and F with Fig. 3C,F). (G) cyp26b1 and col10a1 are coexpressed in cells adjacent to the bony core of the cleithrum. However, cyp26b1 levels in these central cells is lower than in more peripheral, col10a1-negative cells, which most likely represent immature and/or inactive osteoblasts.

Brain patterning and cranial neural crest migration is unaffected in cyp26b1 mutants. (A-L) In situ hybridizations of cyp26b1 mutants (dol) and wild-type siblings (WT) at stages indicated in upper right corners, and with probes indicated in lower right corners; dorsal views on head regions. (A,B) Mutant shows normal expression patterns of six3b, a marker for forebrain (Kobayashi et al., 1998), pax2a, a marker for the midbrain-hindbrain boundary region (Krauss et al., 1991), and egr2b (formerly called krox20), a marker for hindbrain rhomobomeres 3 and 5 (Oxtoby and Jowett, 1993), indicating that brain patterning is unaffected, including normal sizing of the forebrain primordium. Similarly, expression patterns of hoxb1a, a marker for rhombomere 4 (Amores et al., 1998; Rohrschneider et al., 2007) (C,D), and hoxb3a, a marker for rhomobomeres 5 and 6 (Amores et al., 1998; Hogan et al., 2004) (E,F) are normal, indicating that cyp26b1 is dispensable for hindbrain patterning. Furthermore, cyp26b1 mutants display normal expression of hoxb3a in r5/r6-derived neural crest, and of dlx2b, a marker for all cranial neural crest cells (Akimenko et al., 1994), indicating that neural crest patterning and migration are unaffected by the mutation. Normal expression pattern of isl1 (Inoue et al., 1994) further indicates that branchiomotor neurons, derivatives of the hindbrain, are specified and patterned normally in cyp26b1 mutants (I,J). IV, trochlear nucleus; V, trigeminal nucleus; VII, facial nucleus; X, vagal nucleus). (K,L) Normal expression of tfap2b (Knight et al., 2005) in tectum, hindbrain and spinal cord of cyp26b1 mutant. fb, forebrain; hb, hindbrain; mhb, midbrain-hindbrain boundary; nc, neural crest; ot, optic tectum; r, rhomobomere; ret, retina; sc, spinal cord.

RA treatments during late larval stages cause fusions of vertebral bodies. Alizarin Red (alR) stainings of vertebrae at 9 dpf (A,B) or 18 dpf (C-F). (A,C,E) Untreated siblings; numbers indicate anterior-posterior positions of shown vertebrae along the vertebral column. (B,D,E) Corresponding regions of animals treated with RA from 6-9 dpf (B), or from 15-18 dpf (D,F). RA treatments at these late developmental stages cause widespread fusions of centra, including centra 3-6, which ossify first and which have become insensitive to Bmp2 at 4 dpf (compare with Fig. 7). c, centrum; ha, hemal arch; na, neural arch.