Two Sox9 Genes on Duplicated Zebrafish Chromosomes

Developmental Biology 231, 149–163 (2001) doi:10.1006/dbio.2000.0129, available online at http://www.idealibrary.com on View metadata, citation and similar papers at core.ac.uk brought to you by CORE Two Sox9 Genes on Duplicated Zebraﬁsh provided by Elsevier - Publisher Connector Chromosomes: Expression of Similar Transcription Activators in Distinct Sites

Evelyn F.-L. Chiang,* Chin-I Pai,† Mary Wyatt,* Yi-Lin Yan,‡ John Postlethwait,‡ and Bon-chu Chung*,†,1 *Institute of Molecular Biology, Academia Sinica, Nankang, Taipei, Taiwan, Republic of China; †Graduate Institute of Biochemistry, Yang-Ming University, Shi-Pai, Taipei, Taiwan, Republic of China; and ‡Institute of Neuroscience, University of Oregon, Eugene, Oregon 97403-1254

Sox9 is a transcription factor required for cartilage formation and testis determination in mammals. We have cloned from zebrafish two sox9 genes, termed sox9a and sox9b. Gene phylogenies showed that both genes are orthologous to tetrapod SOX9 genes. Genetic mapping showed that these two loci reside on chromosome segments that were apparently duplicated in a large-scale genomic duplication event in ray fin fish phylogeny. Both Sox9a and Sox9b proteins bind to the HMG consensus DNA sequences in vitro. We tested different domains for transactivation potential and identified a potential activation domain located in the middle of both Sox9a and Sox9b. During embryogenesis, sox9a and sox9b expression patterns are distinct but overlap in some regions of the brain, head skeleton, and fins. Expression of sox9a/b correlates well with that of col2a1 in chondrogenic elements. In the adults, sox9a is expressed in many tissues including brain, muscle, fin, and testis, whereas sox9b expression is restricted to previtellogenic oocytes of the ovary. This expression pattern predicts that sox9a and sox9b may have unique functions in some specific tissues during development. The role of gene duplication for the evolution of developmental gene function is discussed. © 2001 Academic Press Key Words: evolution; skeleton; chondrogenesis; gonad; sex determination; gene duplication; transcription factor; activation domain; testis; ovary.

INTRODUCTION developmental pathways. For example, Sox1, Sox2, Sox3, and Sox11 are expressed in the developing nervous system Sox protein family members all contain the SRY-like (Collignon et al., 1996; de Martino et al., 2000; Uwanogho HMG-box, a 79-amino acid domain that binds and bends et al., 1995). Sox4 acts as a transcriptional activator in T DNA. The index gene for the family, SRY, is the only factor lymphocytes (van de Wetering et al., 1993). on the mammalian Y chromosome that is sufficient and Human SOX9 is 509 amino acids long, with an necessary to cause an individual to develop a male pheno- N-terminal HMG DNA-binding domain and a C-terminal type (Koopman et al., 1991). More than 20 SOX cDNAs transactivation domain (Sudbeck et al., 1996; Wagner et al., have been cloned and partially characterized. The family is 1994). The SOX9 HMG domain binds to the AGAA- subdivided into seven subgroups, A–G, based on amino acid CAATGG sequence and the HMG consensus AACAAAG similarity within their HMG domains (Wright et al., 1993). sequence (Bell et al., 1997; Harley et al., 1992; Mertin et al., Many of these genes are known to be expressed tissue- 1999; Sudbeck et al., 1996). These sequences are present in specifically and act as transcriptional factors in different the regulatory region of the human type II collagen gene, COL2A1. In transgenic mice, their mutation abolishes 1 To whom correspondence should be addressed at Institute of SOX9 binding and chondrocyte-specific expression of a Molecular Biology, Academia Sinica, Nankang, Taipei, Taiwan COL2A1-driven reporter gene whereas expression of exog- 11529, Republic of China. Fax: 886-2-2782 6085. E-mail: enous Sox9 transactivates both the reporter gene and the [email protected]. endogenous Col2a1 gene expression (Bell et al., 1997).

These studies demonstrate that mouse Sox9 functions as a Accession Nos. AF277096 and AF277097), expressed sequenced tags transcription activator through binding to specific se- (ESTs) representing sox9a (Accession No. AW184648) and sox9b quences. (Accession Nos. AI883804 and AW343046) were identified by the In situ hybridization of mouse embryos showed strong Washington University EST project (http://www.genetics.wustl.edu/ Sox9 expression in mesenchymal condensations before and fish_lab/frank/cgi-bin/fish/home_est.cgi). An additional EST (AW153579, fi23c10) is highly similar to vertebrate Sox8 genes; during chondrogenesis, and the expression pattern is very although full-length sequence of this clone is not yet available, similar to that of the type II collagen gene, Col2a1 (Ng et alignment of the available sequence shows that fi23c10 is a gene al., 1997; Wright et al., 1995). Patients with SOX9 defi- different from either sox9a or sox9b. Our unpublished data showed ciency suffer from campomelic dysplasia, a disease with that this sequence maps to a position near sox9b, on LG3. The severe skeletal malformation and sometimes male sex fi23c10 locus is closely related in sequence to SOX8, on human reversal (Foster et al., 1994; Wagner et al., 1994). Thus, the chromosome Hsa 16p13.3, and it maps next to hbaa1, whose function of Sox9 in skeletal formation correlates well with human orthologue HBA is also located at Hsa16p13.3 (data not its specific expression in the developing cartilage and the shown). disease symptoms caused by its deficiency. During gonad differentiation, Sox9 is up-regulated in RNA Analysis testes and down-regulated in ovaries (Kent et al., 1996; Reverse transcription was performed using the Superscript pre-

Morais da Silva et al., 1996; Moreno Mendoza et al., 1999). amplification system (Gibco BRL) with 0.5 ␮g of oligo(dT)12–18 and This expression pattern, along with male sex reversal in 5 ␮g of each total RNA in a 20-␮l reaction as described (Guo et al., SOX9-deficient patients, implies that SOX9 is important in 1993). The cDNA product (1 ␮l) was used in PCR with sox9a or the male developmental pathway, but the function of Sox9 actin primers for 25–30 cycles at 94°C for 40 s, 60°C for 50 s, and in sex differentiation is still not clear. 72°C for 70 s. The sox9a primers (CGGTGAAGAACGGCCA- To examine the degree of evolutionary conservation of GAGC and CTGTAGAGTCAGCAATGGGT) amplified cDNA SOX9 gene function in vertebrate development, we have from nt 714 to 1472. The sox9b primers (TCGGGATCGGACAGC- GAGAC and GAGCACCGCTCCAAACACTG) amplified cDNA cloned and characterized two sox9 genes from zebrafish. from nt 254 to 1087. The actin primers (Hwang et al., 1997) Genetic mapping experiments showed that the sox9a and (TCACACCTTCTACAACGAGCTGCG and GAAGCTGTAGC- sox9b genes are located in duplicated chromosomal regions. CTCTCTCGGTCAG) generated a 340-bp fragment. PCR products Phylogenetic analyses suggest that both genes, being ortho- were analyzed on 1.2% agarose gels, blotted, and hybridized with logues of the tetrapod SOX9 gene, arose during a whole sox9a or sox9b cDNA probe. genome duplication event hypothesized to have occurred at the base of teleost radiation (Amores et al., 1998; Postle- Phylogenetic Analysis thwait et al., 1998). In zebrafish, both genes are expressed in To identify sequences similar to zebrafish Sox9b protein, we used chondrogenic cell lineages, but sox9a is expressed in the the tblastx algorithm in a sequence similarity search (http:// sox9b testis and in the ovary. Despite different expression www.ncbi.nlm.nih.gov/blast/blast.cgi). Sequences showing the high- sites, both proteins have similar DNA-binding and transac- est levels of sequence similarity were downloaded from GenBank tivation properties as shown by our molecular analyses. (http://www3.ncbi.nlm.nih.gov/htbin-post/Entrez/query?dbϭn& formϭ0) and imported into CLUSTALX (Julie Thompson and Fran- cois Jeanmougin, ftp://ftp-igbmc.u-strasbg.fr/pub/ClustalX). MATERIALS AND METHODS Sequence alignments are available on request. Neighbor-joining trees were drawn with NJPlot (http://pbil.univ-lyon1.fr/software/ Fish Stocks njplot.html), and maximum likelihood and parsimony trees were constructed using PAUP* (Phylogenetic Analysis Using Parsi- Zebrafish (Danio rerio, AB strain) were maintained as described mony) (Sinauer Associates, Sunderland, MA). The statistical ro- (Westerfield, 1995). Developmental age is given as hours postfer- bustness of each node in the tree was estimated by bootstrapping tilization (hpf) at 28.5°C. analysis (Efron and Gong, 1983; Felsenstein, 1985; Swofford et al., 1996). Bootstrapping involves resampling the sequence data, with Isolation of sox9 cDNA replacement, to create a series of samples with the same size as the original data, and constructing new phylogenetic trees from each Zebrafish sox9a and sox9b were obtained by hybridizing a cDNA sample. We ran 1000 bootstrap analyses for neighbor-joining trees, library established from 1-month-old zebrafish (Clontech Labora- and 100 for maximum-likelihood and parsimony trees, and the tories, Inc) with a 318-bp DNA containing the HMG domain (nt number given in the figure is the number of times the indicated 587–904) of mouse Sox9. Filter hybridization was performed in node was obtained. 50% formamide at 37°C followed by washes with 0.5ϫ SSC, 0.1% We follow the rules of gene and protein nomenclature for each SDS at 42°C. Ten positive clones were subcloned into PCR II vector individual species when formally established. Zebrafish nomencla- for DNA sequencing, data base search, sequence alignment, and ture (sox9a for the gene, Sox9a for the protein) is found at http:// evolutionary analysis using the GCG computer analysis program. zfish.uoregon.edu/zf_info/nomen.html, mouse (Sox9, gene; SOX9, Two clones, designated sox9a and sox9b, have high homology to protein) at http://www.informatics.jax.org/support/nomen/, and hu- mouse Sox9 throughout their entire sequence. The remaining man (SOX9, gene; SOX9, protein) at http://www2.ncbi.nlm.nih.gov/ positive clones represent other sox family members. After we had LocusLink/LLnomen.html. Referring to vertebrate genes in general, cloned, sequenced, and characterized sox9a and sox9b (GenBank we use the human nomenclature.

In Situ Hybridization Gel Mobility-Shift Assays (EMSAs) The riboprobes for in situ hybridization were synthesized from The 30-bp COL2A1 enhancer sequence that binds SOX9 (called linearized plasmids by in vitro transcription using UTP labeled col2c2 probe) was previously described (Bell et al., 1997; Ng et al., with digoxigenin or fluorescein. The sox9a probe was generated 1997). The 26-bp HMG-binding probe corresponds to a fragment of from the pGM-sox9a/EcoRV-linearized template using T7 RNA the CD⑀ enhancer that contains a consensus binding site for HMG polymerase, and the sox9b probe was from the pBluescriptKSϩ- box proteins as described previously (Harley et al., 1992; Lefebvre sox9b/StuI-linearized template. To increase hybridization specific- et al., 1997). The unrelated sequence (kuc1), CTAGACAAGGT- ity, the sequence containing the 3Ј-untranslated region (3Ј-UTR) CATCATCTAG, corresponds to a fragment that contains a con- and part of the C-terminal coding region excluding the HMG box sensus binding site for SF-1 protein (Guo et al., 1994). In vitro- was chosen as a probe for the respective sox9 cDNAs. Thus sox9a synthesized proteins were generated using the Promega in vitro probe is 698 bp in length containing 557 bp of coding region and 141 TNT system. The probes were made by labeling the complemen- bp of 3Ј-UTR. The sox9b probe is 945 bp in length containing 404 tary oligonucleotides with [␥-32P]ATP using T4 polynucleotide bp of coding region and 541 bp of 3Ј-UTR. The col2a1 probe was kinase. Electrophoretic mobility-shift analysis was performed us- generated as previously described (Yan et al., 1995). ing 3 ␮l of translation products, 0.01-pmol probe, 1 pmol competi- Tissues or embryos were fixed overnight in 4% paraformalde- tor oligos, following the published procedure (Guo et al., 1994). hyde at 4°C before processing for hybridization with digoxigenin- labeled probes as described (Hu et al., 1999; Oxtoby and Jowett, 1993). When necessary, the hybridized samples were cryosectioned RESULTS according to the established procedure (Korzh et al., 1993). Cloning and Characterization of Zebrafish Sox9 Homologues Genetic Mapping To search for zebrafish homologues of Sox9, we used a For gene mapping, single-strand conformation polymorphism 318-bp fragment containing the HMG box of mouse Sox9 to analysis (SSCP) (Fo¨rnzler et al., 1998) was performed on DNA from screen a cDNA library constructed from 1-month-old ze- 96 F2 progeny from a linkage map cross with over 650 PCR-based brafish larvae. From 10 positive clones, we selected 2 clones markers (Johnson et al., 1996; Knapik et al., 1996, 1998; Postle- that showed the highest sequence similarity to mouse Sox9 thwait et al., 1994, 1998). The strain distribution patterns were analyzed using MapManager (http://mcbio.med.buffalo.edu/ for further characterization. The sox9a (AF277096) and mapmgr.html) and maps were constructed with MapMaker (Lander sox9b (AF277097) clones were 1790 and 1916 bp long and et al., 1987). For SSCP analysis, genomic DNAs from C32 and SJD should encode proteins of 462 and 407 amino acids, respec- parental strains were amplified using primers specific to the tively (Fig. 1A). sequence of the 3Ј-untranslated region of sox9a and sox9b genes. Sequence alignment (Fig. 1B) reveals that the HMG The sox9a mapping primers, AGACCATGAGCACGCCTGAAT domain (domain B) is the most conserved, with more than and GTCTTTCCCATCATGCACTGAACG, amplified a 238-bp 96% sequence identity throughout vertebrates. Sequence ϩ ϩ fragment from 1547 to 1784. The sox9b mapping primers, comparison of their HMG domains predicts that zebrafish CTCTGCCCGCTCACATCCATACTC and AGCGCCACTGCA- Sox9 falls into the E subfamily of the SOX proteins together GATTAGATTGAA, amplified a 235-bp fragment (ϩ1461 to ϩ1695). with SOX 8, 9, and 10 (data not shown) (Denny et al., 1992; In each PCR for mapping, one of the primers was end-labeled using [␥-32P]ATP for detection by autoradiography on acrylamide gels. The Laudet et al., 1993). Sequence homology among SOX9s strain distribution patterns for mapped loci that are not given in this from different species extends beyond the HMG domain. paper are available at http://www.neuro.uoregon.edu/postle/ Domains A, C, and E have an overall 70–85% sequence mydoc.html. identity among all known vertebrate SOX9 proteins. Do- main D is almost nonexistent in zebrafish. It is present as a short stretch of residues without the characteristic Pro- and Transfection Gln-rich (P/Q) region. It is called domain D for operational The Gal4 DNA-binding domain was cloned into pCDNA3 at the purposes. The Gln- and Ser-rich (Q/S) domain E is variable HindIII-BamHI sites to form pGal4-DBD (kindly provided by M. in length. It is shortest in zebrafish Sox9b. Hu). Various fragments of sox9 cDNA were then cloned into pGal4-DBD to form a Gal4-DBD-Sox9 fusion protein. The sox9 cDNA fragments were amplified by PCR before cloning. Phylogenetic Analysis of Vertebrate SOX9 Genes COS-1 cells in 6-cm plates were transfected by the Lipofectamin To determine the relationships of zebrafish sox9 genes to (Gibco) method with 0.2 ␮g pCMV-␤Gal (as an internal control), 1 each other and to those of other vertebrates, we constructed ␮ g pG5E1bCAT (containing 5 Gal4-binding sites in front of the phylogenetic trees of amino acid sequences and of nucleic CAT reporter gene), and varying amounts of effector plasmids as acid sequences. In the absence of SOX9 sequence data from described in Fig. 10. CAT activities were calculated after calibra- suitable nonvertebrate outgroups, such as a Cephalochor- tion with internal control. Standard deviations were calculated from data derived from three separate experiments. The amount of date or a Urochordate, the tree was rooted on vertebrate Sox9 protein expression after transfection was determined by SOX10 and SOX8 sequences, which have considerable simi- Western blot analysis against the Gal4-DBD antibody (1:10,000, larity to SOX9 over nearly all of the molecule. The results Santa Cruz Biotechnology, Inc.) after equal amounts of lysates were showed Danio rerio Sox9 fits strongly into the SOX9 clade loaded on gels. of vertebrate SOX genes and SOX proteins, in a separate

FIG. 2. Phylogenetic and mapping analysis. (A) Neighbor-joining phylogenetic tree of vertebrate SOX9 proteins. Numbers on branches are the number of times, in a thousand runs, that the two clades branched as sisters. The results show, with very high FIG. 1. (A) Sequence alignment of zebrafish Sox9a and Sox9b. bootstrap support, that zebrafish has two loci in the SOX9 clade. Identical amino acids are connected by a line, whereas similar The accession numbers for the sequences are as follows: SOX9 amino acids are connected by dots. Gaps are represented as periods. human (P48436); Sox9 Pig (O18896); SOX9 mouse (S52469); Sox9 The HMG box is underlined. (B) Comparative Analysis of Sox9 chicken (AB012236); Sox9 alligator (AAD17974); Sox9 Rana frog sequence identity in various domains. Chicken and fish Sox9 (BAA95427); Sox9a (zebrafish, AF277096); Sox9 trout (AB006448); proteins lack the P/Q-rich region. Numbers give amino acid Sox9b zebrafish (AF277097); SOX10 human (P56693); SOX10 positions. Sequence of chicken Sox9 is incomplete at the mouse (Q04888); Sox10 chicken (CAB65026); SOX8 human N-terminus. The percent amino acid identities of different domains (AF226675); SOX8 mouse (AF191325); Sox8 chicken (AF228664). relative to zebrafish Sox9a are given as percentages. The accession The marker length corresponds to a 10% sequence difference. (B) Sox9 numbers for the sequences are as follows: human SOX9 Maximum-likelihood phylogenetic tree of vertebrate genes. Sox (HSSOX9MRN), chicken Sox9 (GGU12533), zebrafish Sox9a Coding sequences of vertebrate genes were analyzed with (AF277096), rainbow trout Sox9 (AB006448), zebrafish Sox9b PAUP* using the maximum-likelihood algorithm, with bootstrap- (AF277097). ping. Numbers on branches are the number of times, in a hundred runs, that the two clades branched as sisters. Nucleotide sequences from the ATG to the stop codon were aligned by clustalW. The accession numbers for the sequences are as follows: mouse Sox8 (AF191325); chicken Sox9 (AB012236); chicken Sox8 (AF228664); clade from Sox10 and Sox8 proteins (Fig. 2A). The position zebrafish sox9b (AF277097); rana frog Sox9a (AB035887); sox9a of Sox9a in the protein phylogeny is as expected if sox9a is zebrafish (AF277096); pig Sox9 (AF029696); human SOX8 an orthologue of human SOX9. Sox9b appeared as a sister (AF226675); human SOX10 (NM_006941); human SOX9 (Z46629); group to other vertebrate SOX9 proteins, although still trout SoxP1 (D83256); mouse Sox10 (AF047389); alligator Sox9 within the SOX9 clade, with very high bootstrap support, (AF106572); trout Sox9 (AB006448); drosophila sox100B 1000 of 1000 bootstrap runs. Sox9b appears to have evolved (AJ251580).

Copyright © 2001 by Academic Press. All rights of reproduction in any form reserved. Zebrafish sox9a and sox9b Genes 153 more rapidly than other vertebrate SOX9 proteins as judged by branch lengths. This long branch may account for the protein appearing as a sister group to other vertebrate Sox9 proteins rather than as just a sister group to zebrafish Sox9a. Note that the rainbow trout Sox9 protein branches with the zebrafish Sox9a protein, and not the Sox9b protein. This result would be expected if the duplication of sox9 genes occurred before the divergence of zebrafish and trout lineages, and the trout orthologue of sox9b has either been lost or exists but has not yet been identified. Further understanding of the phylogenetic relationships of zebrafish sox9 genes comes from trees that compare nucleotide sequences. Figure 2B shows the maximum- likelihood tree for the coding portions of SOX genes from several vertebrates, with the tree rooted on the Drosophila gene sox100B, which, in a BLASTX search brings up vertebrate SOX10 genes as best hits. In this analysis, sox9b branches as a sister group of teleost sox9a genes (zebrafish sox9a and trout Sox9), with very strong bootstrap support (99 of 100 runs), well within the vertebrate Sox9 clade, which itself branches with strong support away from vertebrate Sox8 and Sox10 clades. These results were con- FIG. 3. Genetic mapping of sox9 genes. (A) The sox9b gene maps firmed by nucleotide trees that are constructed based on to the lower end of Linkage Group 3 (LG3), and sox9a maps to the parsimony and distance analyses (data not shown). The upper portion of LG12. (B) Conserved syntenies in the regions phylogenetic analysis of vertebrate Sox genes strongly sup- surrounding the sox9 genes. A comparison of LG3 to LG12 shows ports the conclusion that sox9a and sox9b are orthologues that several genes are present as duplicate copies on the two of human SOX9, and not of SOX8 or SOX10. An additional chromosomes. Gene phylogenies show that both copies of the two and independent data set to infer phylogenetic relationships genes are orthologues of the mammalian gene shown on the same is available from comparative gene mapping experiments. row. These mammalian orthologues are syntenic on human (Hsa) chromosome 17 and mouse (Mmu) chromosome 11. A number of other genes that do not appear to be duplicated in zebrafish are also on these two mammalian chromosomes. Linkage Mapping of Zebrafish sox9 Genes

To discover the chromosomal location of zebrafish sox9 genes, we performed mapping experiments. We found that rest of the precursors of LG3 and LG12, and resulted from a sox9b mapped to the lower tip of LG3, and sox9a mapped to large-scale chromosome duplication event. the upper arm of LG12 (Fig. 3A). Figure 3B compares some of the loci on zebrafish LG3 and LG12 with their ortho- Different Expression of sox9a and sox9b in Adult logues on human (Hsa) and mouse (Mmu) chromosomes. At Zebrafish Tissues least five loci in addition to sox9a/sox9b are present in duplicate copies on LG3 and LG12: hoxba/hoxbb (Amores We first examined sox9a and sox9b expression by RT- et al., 1998), col1a1a/col1a1b (Woods et al., 2000), dlx8/ PCR analysis using total RNAs prepared from various dlx7 (Postelthwait et al., 1998; Stock et al., 1996), rara2a/ tissues collected from sexually mature zebrafish. PCR with rara2b (Stachel and Kushner, GenBank Accession Nos. actin primers showed that equal amounts of RNA were L03398 and L03399, 1993), and hbae1/hbae4 (Chan et al., used in all reactions. The PCR products were hybridized 1997). These five loci also reside on the corresponding with specific sox9a or sox9b cDNA probes to confirm mammalian chromosome segments, Hsa17 and Mmu11. gene-specific detection. Plasmids containing sox9a or sox9b The mammalian orthologues of several other loci on LG3 cDNA were also used as control templates for PCR and and LG12 also reside on Hsa17 and Mmu11. In addition, hybridization specificity. Transcripts from sox9a were ex- human SOX9 and mouse Sox9 genes reside on these same pressed in many tissues including brain, pectoral fin, kid- mammalian chromosomes. We conclude that parts of LG3 ney, muscle, and testis (Fig. 4). Northern blot analysis and LG12 are orthologous to the mammalian chromosome confirmed the expression of a 2-kb transcript in pectoral fin, segments that contain SOX9. This provides strong evidence testis, and brain (data not shown). Zebrafish sox9b, on the that sox9a and sox9b are both orthologues of mammalian contrary, was detected only in the ovary of adult tissues in SOX9. these experiments. We conclude that zebrafish sox9a and sox9b are duplicate The expression of sox9a in testis and sox9b in ovary poses copies of a single SOX9 gene that was duplicated with the an intriguing question, as chicken and mouse Sox9 have

detected in the head region of the embryo from the late segmentation stage. In 24 hpf embryo, the transcripts were found in the region along the boundary of the diencephalon and telencephalon, the epiphysis, the ventral midbrain, and hindbrain (Figs. 6A–D). Differences between the two expression patterns are apparent after careful examination. Expression of sox9b along the diencephalon and telencephalon boundary is restricted to the ventral region, and is missing in the midbrain-hindbrain boundary. Double staining followed by cryosection reveals that sox9a and sox9b expression overlaps mainly in the cells around the ventricle throughout the brain, and in certain dorsal clusters of cells in the hindbrain (arrows in Figs. 6E–G). The significance of sox9a and sox9b expression in the CNS is not yet clear. The main difference between the two genes at this stage is the expression of sox9a in the cranial mesenchyme in FIG. 4. Expression of zebrafish sox9a and sox9b in different adult addition to CNS (Fig. 6C). The cranial mesenchyme lying tissues by RT-PCR. Total RNA from different tissues was used for ventral and ventro-lateral to the neural tube contributes to RT-PCR analysis followed by hybridization to the sox9a or sox9b the formation of the neurocranium and pharyngeal arches cDNA probe. Plasmids bearing sox9a and sox9b were used as (Kuratani et al., 1997; Schilling and Kimmel, 1994). Using control templates to monitor PCR fidelity and hybridization speci- col2a1 as a marker for mesenchyme of the neurocranium ficity. Actin primers were used to determine whether equal and pharyngeal arches (Yan et al., 1995), we confirmed that amounts of RNA were used for each PCR. sox9a is indeed coexpressed with col2a1 in the neurocranium mesenchyme (ms) beneath the neural tube in the hindbrain sections (Figs. 6H–J). Yet, in the precondensed mesenchyme of the pharyngeal arch primordia, only sox9a both been shown to be expressed in testis but not in ovary is expressed (arrowheads in Fig. 6H). The sox9b expression (Kent et al., 1996). To investigate this phenomenon further, in the mesenchyme of segmented pharyngeal arch primor- we performed in situ hybridization on adult tissues. The dia only begins just prior to hatching (Fig. 7B). By this stage, sox9a transcript was observed in whole mounts of adult both sox9a and sox9b transcripts were detected abundantly testis but not ovary. Conversely, the sox9b probe detected in the head with similar but distinct expression patterns transcripts in whole mounts of adult ovary, but not in testis (Figs. 7A, B). (Fig. 5, negative data not shown). Sectioning and histologi- The elements of zebrafish cranio-facial cartilage begin to cal examination showed that sox9a was expressed in Sertoli condense and differentiate between Day 2 and Day 3 cells at the periphery of cysts in the testis (Fig. 5B). This (Kimmel et al., 1995). Expression of sox9a and sox9b result was confirmed by simultaneous hybridization of the persists in mesenchymal condensations during chondrogen- testis sections with a vasa probe (data not shown), which esis in almost all elements of the pharyngeal arch and detects germ cells but not the mesodermal components of neurocranium. These arch elements include the palato- the gonads (Komiya et al., 1994; Yoon et al., 1997). quadrate (pq) and Meckel’s cartilage (mc) of the mandibular In contrast to sox9a, sox9b transcripts were mainly arch, the ceratohyal (ch) and hyosymplectic (hs) of the hyoid detected in the stage I oocyte, at the primary growth or arch, the ceratobranchials (cb) of the branchial arch, and the perinucleolus stage of ovary development, by whole mount trabeculae (tb) and ethmoid plate (ep) of the neurocranium in situ hybridization followed by sectioning (Figs. 5C and (Figs. 7C, D). The level of sox9b expression in all these D). The staging of oocytes follows the nomenclature of developing elements, however, is much lower. The col2a1 Selman et al. (1993). Hybridization signals became diluted gene is also expressed strongly at this stage and overlaps as oocytes grew to stage II, the cortical alveolus stage, and with sox9a expression, but sox9a expression is more re- were further diluted in stage III, or the vitellogenic oocytes. stricted (Figs. 7E, F). The expression of both sox9a and sox9b genes in the cranio-facial skeleton is progressively down-regulated when chondrogenesis is complete (data not Overlapping Expression of sox9a and sox9b in the shown). Brain and Developing Cranio-Facial Skeleton It is intriguing to find that zebrafish sox9a and sox9b are expressed in distinct adult tissues. Since murine Sox9 plays Overlapping Expression of sox9a and sox9b in the an important role during embryogenesis, we set out to Developing Fin and Sensory Organs examine expression of zebrafish sox9a and sox9b during embryonic development. In addition to expression in the head skeleton, sox9a and Similar expression patterns of the two genes were first sox9b are both expressed in the developing pectoral fin

(Figs. 8A–D). Fish pectoral fins are appendages homolo- Sox9a and Sox9b Can Bind DNA Specifically gous to mammalian forelimbs. Coexpression of sox9a and sox9b first appears at the pharyngula stage in the pair of The experiments described above showed that sox9a and fin rudiments lying on the top of the yolk and facing the sox9b are expressed in distinct tissues in adults, but in trunk (Figs. 8A, B). Gene expression is restricted to the similar regions during skeletal differentiation of zebrafish mesenchyme and not detected in the apical ectodermal larvae. In addition, our phylogenetic analysis showed that ridge of the fin buds (Figs. 8B–D). As cell condensation the structures of Sox9a and Sox9b are somewhat divergent from tetrapod SOX9 proteins, and that Sox9b especially was proceeds, the sox9a- and sox9b-expressing region at the evolving more rapidly than other vertebrate SOX9 proteins. innermost precartilage core elongates from the base of It was thus important to determine whether these proteins the appendage where the fin endoskeleton will form (Figs. retained SOX9 function. 8C, D). This expression is down-regulated once differen- We first used electrophoretic mobility-shift assays to test tiation of cartilage in the fin is complete (data not whether zebrafish Sox9a and Sox9b proteins bind DNA in a shown). sequence-specific manner. The test proteins were trans- We also detected sox9 expression in sensory organs of the lated in vitro from cDNA for sox9a, sox9b, mouse Sox9, or sox9b zebrafish embryo. The retina of the eye expressed at the vector pSG5 alone as a negative control (Fig. 9). Sox9a, the pharyngula stage (Figs. 6D–F); cells in the developing ear Sox9b, and mouse SOX9 can all bind to a 30-bp col2c2 (otic vesicle) also showed different degrees of sox9a and oligonucleotide derived from the human COL2A1 enhancer sox9b expression (Figs. 6C, D, 7A, B). Using zebrafish to form a retarded band in the gel. The unbound probe ran col2a1 as a marker (Yan et al., 1995), we confirmed that off the gel in this set of experiments. This binding is expression of sox9a is in the epithelium of the otic vesicle sequence specific, because the band can be competed away (Figs. 6G, J). The patterns of sox9a, sox9b, and col2a1 by unlabeled col2c2 oligo, but not by the unrelated oligo expression in the developing otic vesicle, however, are not kuc1 (Fig. 9B). Similarly, when the consensus sequence identical (Figs. 6C, D, G, J, 7A, B). recognized by HMG-binding proteins was used as a probe, a Our in situ hybridization has demonstrated that the specific complex formed between this oligo and each of expression patterns of zebrafish sox9a and sox9b genes Sox9a, Sox9b, or mouse SOX9 proteins (Fig. 9C). This partially overlap in the skeletal and nonskeletal regions. HMG-binding oligo and the col2c2 oligo can both compete With the notable exception of ovary sox9b expression, their effectively for the binding, indicating that both binding combined expressions mirror that of mouse Sox9, indicat- sites are recognized by the same protein. These results ing that sox9 gene functions have been preserved through- suggest that zebrafish Sox9a and Sox9b can recognize both out evolution. HMG-binding sequences and col2c2 sequences in vitro.

FIG. 5. In situ hybridization shows transcripts of sox9a in zebrafish testis and sox9b in the ovary. (A, C) Whole-mount (B, D) sections of testis and ovary. Transcripts of sox9a (A, B) were detected in the zebrafish testis. Scattered cells across the mature testis express sox9a (arrows in A and B). Cryosection revealed that these cells are Sertoli cells (arrows B), which are irregularly shaped and lie close to the edge of the large membrane-bound cyst-like structures in the testis. Spermatogenesis proceeds synchronously within each cyst. Different stages of spermatogenesis can be seen in different cysts. Large and round shaped germ cells are early stage spermatocytes (cy). These spermatocytes decrease in size as spermatogenesis progresses and develop into spermatids (t). Arrowheads in B indicate the edge of one of the cysts. Transcripts of sox9b (C, D) were evident in the previtellogenic oocytes of the ovary. I, follicles at the primary growth stage (stage I); II, follicles at the cortical alveolus stage (stage II); III, stage III vitellogenic follicles. FIG. 6. Expression patterns of sox9a and sox9b in the head during the pharyngula stage. (A, B) Lateral view and (C, D) dorsal view of the 24 h head. In (A) and (B), both sox9a and sox9b transcripts were detected in the epiphysis (ep), forebrain, and the ventral midbrain. Expression of sox9a in the diencephalon (d) of the forebrain extends dorsally, whereas sox9b expression is mainly at the ventral diencephalon. t, telencephalon. Only sox9a transcripts are detected in the midbrain-hindbrain boundary (MHB), pharyngeal pouches (arrowheads in C), and the head mesenchyme (ms) in the caudal hindbrain. Transcripts of sox9b are detected in the eye (e) close to the midline and in six stripes of the hindbrain, with the highest expression in the four posterior rhombomeres (D). Otic vesicle (ov) marks the position of the caudal hindbrain. (E–G). Double staining with sox9a (red) and sox9b (blue) at 34 h. Both whole mount (E) and cryosections (F, G) show that both genes are expressed in cells around the ventricle (arrow in F) and in two areas (arrow in G) in the dorsal hindbrain. In the head mesenchyme (ms) and segmented pharyngeal arch primordia, which are shown in the whole mount (as arrowheads in E) or in sections (labeled as 1st, 2nd, 3rd), only sox9a transcripts were detected. As in panel D, the retina (r), but not the lens (l) of the eye (e), retains sox9b expression. (H–J) Double staining with sox9a (red) and col2a1 (blue) of the head. The precondensed mesenchyme in the segmented pharyngeal arches express only sox9a, not col2a1 (arrowheads in H and 2nd in I). Transverse sections through the rostral hindbrain (I) and at the level of the ear (J) confirm that the transcripts of the two genes overlap in the head mesenchyme (ms) which lie beneath or to the side of the neural tube. Expression of col2a1 in the epithelium (ep) of otic vesicle is much stronger than that of sox9a. Only a few epithelial cells are positive for both sox9a and col2a1 staining (arrow in J). Expression of col2a1 was also detected in the midline structures, including the floor plate (fp) and notochord (nd), but not in CNS.

Both Zebrafish Sox9 Proteins Can Activate Transcription The major difference between SOX9 proteins from zebrafish and other vertebrates is that zebrafish Sox9a and Sox9b have shorter C-terminal domains (Fig. 1B). The C-terminal domain of mouse and human SOX9 acts as a transactivation domain (McDowall et al., 1999; Ng et al., 1997; Sudbeck et al., 1996). It is therefore important to test if zebrafish Sox9a and Sox9b still have the activation function. We fused cDNA encoding amino acids 184–462 of Sox9a or amino acids 168–407 of Sox9b, comprising domains C, D, and E, with the DNA- binding domain (DBD) of the Gal4 protein. These fusion constructs were used as effectors in cotransfection experiments together with the CAT reporter construct that contains five recognition sites for Gal4-DBD (Fig. 10A). As shown in Fig. 10B, the C-terminal half of mouse SOX9 and zebrafish Sox9a and Sox9b can activate transcription in a dose- dependent manner. The transactivation potential appears to be proportional to the length of the C-terminal region; mouse SOX9 has the longest C-terminal domain and highest activation potential, with Sox9a second, and Sox9b third, though still retaining substantial transactivation potential. The full- length proteins (Sox9a-F, Sox9b-F, and mSOX9-F) did not activate transcription, perhaps because the HMG domains might interfere with the DNA-binding domain of Gal4. These results demonstrate that zebrafish Sox9a and Sox9b bind DNA specifically and activate transcription in vitro, suggesting both proteins are probably transcription activators in vivo. The function of the transactivation domain of Sox9a and Sox9b was further dissected. Subdomains C, D, and E were fused to Gal4-DBD either separately or in combination so FIG. 9. Binding of Sox9a and Sox9b to specific DNA sequences. (A) their potential to transactivate the reporter gene could be The sequences of oligo used in the protein–DNA-binding experiment. further tested (Fig. 11). While the longest domain (amino The sequence of the known binding site for SOX9 is underlined. acids 184–462 of Sox9a) had the highest transactivation Electrophoretic mobility-shift assay was performed by incubating 32 function, the C domain (amino acids 184–307 of Sox9a and P-labeled col2c2 (B) or HMG (C) probe with in vitro translation amino acids 168–293 of Sox9b) also activated transcription, products derived from plasmids encoding no protein (pSG5), mouse whether transfected alone or in combination with other SOX9, or zebrafish Sox9a or Sox9b followed by electrophoresis. A 100-fold excess of unlabeled oligo was added as competitor. “Ϫ” refers subdomains. Domains D and E combined or alone did not to samples with no competitors. Zebrafish Sox9a and Sox9b recognize appear to be sufficient to activate transcription. The immu- the col2c2 motif and the HMG consensus sequence as effectively as noblot at the right of Fig. 11 shows that all the fusion does mouse SOX9. The protein–DNA complex can be competed away proteins were expressed in high amounts after transfection. by unlabeled col2c2, but not by the unrelated kuc1. The unbound Our result suggested that domain C of zebrafish Sox9a probe has run off the gel in this set of experiments.

FIG. 7. Expression of sox9a and sox9b in the developing pharyngeal arches. (A, B) Lateral view of the 45 h embryo. Both sox9a and sox9b were detected strongly in the pectoral fin buds (pf), most segments of pharyngeal arch primordia, such as mandibular (m), hyoid (h), and ceratobranchial (cb) arches, CNS, and otic vesicles (ov). Expression patterns of the two genes in the brain and otic vesicle are not identical. (C, D) Ventral view of head at 72 h. Differentiating cartilage elements of pharyngeal arches express both sox9a and sox9b; the level of sox9b expression is lower and concentrated in the ceratobranchial. Other abbreviations: cb, ceratobranchials; ch, ceratohyal; mc, Meckel’s cartilage; pq, palatoquadrate; tb, trabeculae. (E, F) Double staining of 76 h head with col2a1 in blue and sox9a in red (E), or col2a1 in red and sox9a in blue (F). Expressions of the two genes overlap in most cartilaginous tissues in the cranio-facial and pectoral fin skeletons. The distribution of col2a1 in these elements is less restricted than that of sox9a. FIG. 8. Expression of sox9a and sox9b in the developing pectoral fins. (A, B) Double staining of sox9a (red) and sox9b (blue) at the early fin bud stage. (A) Dorsal view of rostral trunk shows that both genes are expressed in the developing fin bud with broader sox9a expression. (B) Lateral view. Both transcripts are restricted to the mesoderm of the fin bud. (C, D) Dorsal view of rostral trunk at 60h. Both sox9a and sox9b transcripts are expressed in precartilaginous mesenchyme of the elongated pectoral fins but not in ectodermal fin fold (ff).

Copyright © 2001 by Academic Press. All rights of reproduction in any form reserved. Zebraﬁsh sox9a and sox9b Genes 159 and Sox9b could be the major activation domain, whereas the C-terminal domain E only plays a minor role, presum- ably due to its short length. Apart from a stretch of conserved Ser/Gln/Pro/Thr (amino acids 235–246 of Sox9a), the sequence of domain C does not resemble any other known transactivation domains.

DISCUSSION

Two Orthologues of Mammalian SOX9 in Zebrafish We have cloned two zebrafish sox9 genes, which encode transcription activators with DNA-binding and transcription activation domains. Phylogenetic analyses of both protein and DNA sequences by the neighbor-joining, maximum-likelihood, and parsimony methods all group sox9a and sox9b in the vertebrate Sox9 clade, distinct from Sox8 and Sox10. Furthermore, sox9 genes are located on duplicated zebrafish chromosome segments, in syntenic regions conserved in mammalian genomes. This provides strong evidence that sox9a and sox9b are both orthologues of mammalian SOX9. We conclude that the sox9a/sox9b gene pair arose by duplication events that involved large chromosome sections, consistent with their origin in a genome duplication event in ray-finned fish (Amores et al., 1998; Postlethwait et al., 1998). There appear to be many examples of duplication in zebrafish similar to that of sox9, including cyp19 (Chiang et al., 2001; Chiang et al., 2000), dlx (Ellies et al., 1997; Stock et al., 1996), msx (Ekker et al., 1997), hedgehog (Ekker et al., 1995), and nkx (Lee et al., 1996). FIG. 10. The C-terminal domains of mSOX9 and zebrafish Sox9a, Expression of Zebrafish sox9 in Skeletogenesis and and Sox9b can activate transcription. (A) Structure of expression CNS plasmids. The Gal4 DNA-binding domain (DBD) was linked to different lengths of the C-terminal domain or to full-length SOX9 The preservation of the sox9a/sox9b duplicates in ze- protein. The numbers below each box represent residue numbers. brafish raises an interesting question regarding their func- The reporter contains the chloramphenicol acetyltransferase (CAT) tional relationship. One way to investigate this is to look at gene driven by a minimal TATA box and 5 copies of the Gal4 their expression patterns. Our in situ hybridization demon- recognition sequence. (B) Reporter gene activity. Increasing strated expression of zebrafish sox9a and sox9b genes before amounts of effector DNA were cotransfected with 1 ␮gof ␮ ␤ and during prechondrogenic mesenchymal condensation pG5E1bCAT and 0.5 g of CMV- -gal into COS-1 cells, and CAT correlating well to the situation in mouse. Coexpression of activities scores after normalization with internal control. sox9a and sox9b in the pharyngeal arch and pectoral fin mesenchyme indicates the possibility of overlapping functions of the duplicates in the developing cranio-facial and dium of the pharyngeal arch and pectoral fin. This suggests appendicular skeletons. However, the lower level of sox9b a sox9 function in committing undifferentiated mesenchy- expression in the developing cranio-facial elements may mal cells to their appropriate chondrogenic pathways before imply a gradual loss of function for sox9b in this domain initiating col2a1 expression. during evolution. Coexpression of zebrafish sox9a, sox9b, and col2a1 starts Mouse Sox9 is known to coexpress with Col2a1 to at the end of Day 2 when the mesenchyme begins to regulate its expression during chondrocyte differentiation condense and differentiate into cartilage. This is in agree- and maturation (Bell et al., 1997; Lefebvre et al., 1997; Ng et ment with a second function of the sox9 duplicates in the al., 1997). Evidence also implicates SOX9 involvement in regulation of col2a1 expression in the cartilaginous ele- mesenchymal condensation (Bi et al., 1999). In zebrafish, ments. The expressions of sox9a and sox9b quickly dimin- expressions of sox9a and sox9b are significantly ahead of ish once col2a1 expression reaches its peak, similar to the col2a1 in the prechondrogenic mesenchyme in the primor- observation in mouse.

FIG. 11. Transactivation domain analysis of Sox9a and Sox9b. (A) Structure of expression plasmids. The Gal4 DNA-binding domain (DBD) was linked to different lengths of the C-terminal domain or full-length Sox9 protein. The numbers above each box represent residue numbers. The reporter contains the chloramphenicol acetyltransferase (CAT) gene driven by a minimal TATA box and 5 copies of the Gal4 recognition sequence. (B) CAT activity of the reporter pG5E1bCAT cotransfected with different effector plasmids after normalization. (C) Western blot of 20 ␮g protein from each cell lysate after transfection probed with anti-Gal4 antibody. Effector proteins could be detected in all transfections.

We detected similar patterns of sox9a and sox9b expres- pathway may facilitate gene function ablation for further sion in the developing ear and brain, but different cell types elucidation of their roles in the developmental processes. express each gene. The expression is mainly concentrated in the ventral forebrain, the cells around the ventricle, and Differential Expressions of sox9a and sox9b in the in certain groups of dorsal cells across the hindbrain. Mouse Gonad Sox9 is expressed in similar regions of the CNS (Ng et al., 1997). The significance of Sox9 expression in the CNS and The expression pattern of the duplicates in mature ze- sensory organs is not yet understood, but some human brafish gonads is surprising. In mouse and chicken, Sox9 campomelic dysplasia (CD) patients with SOX9 mutations expression is down-regulated in the differentiating gonad exhibit deafness and mental retardation (Houston et al., just before ovary formation, but the expression in Sertoli 1983), suggesting that SOX9 may regulate genes in these cells of the developing testis persists throughout adulthood tissues to initiate pathways involved in neuronal differen- (Kent et al., 1996; Morais da Silva et al., 1996; Moreno tiation. Mendoza et al., 1999). This male-favored expression pattern The overlapping expression patterns of sox9a and sox9b at the stage of testis formation was also observed in fresh- in the CNS and skeletal lineage suggest a possible func- water and marine turtles (Moreno Mendoza et al., 1999; tional redundancy in these domains. SOX9 mutations in Spotila et al., 1998), suggesting that the functional impor- CD patients show a haploid-deficiency phenotype with tance of SOX9 for male differentiation is conserved among severe defects in skeletal formation (Foster et al., 1994; tetrapods. In zebrafish, we detected sox9a transcripts in Kwok et al., 1996; Wagner et al., 1994). The lethality of Sertoli cells of the testis, while abundant sox9b expression Sox9 heterozygosity makes it difficult to study the null was seen in the ovary. function of mouse Sox9 by gene targeting (Bi et al., 1999). The different expression patterns of duplicates sox9a/ The presence of two zebrafish sox9 genes and their poten- sox9b in fish gonads suggest that sox9a retained its func- tial functional redundancy in the same developmental tion in the testis while sox9b possibly acquired a “new”

Copyright © 2001 by Academic Press. All rights of reproduction in any form reserved. Zebrafish sox9a and sox9b Genes 161 function in zebrafish ovary during evolution. However, we ACKNOWLEDGMENTS cannot rule out the possibility that this split expression pattern of sox9a/sox9b may merely mimic an ancestral role We thank Peter Koopman for the gift of mouse Sox9 cDNA and of a single SOX9 which functioned in both testis and ovary. antibody, Jye-I Wang for initial fish maintenance, and Vladimir In support of this, the expression of SOX9 in the differen- Korzh for helpful suggestions. This work was supported by Grants tiated ovary has also been reported in other species, includ- NSC89-2311-B-001-099 and NSC88-2611-B-001-008-B24 from the ing human and frog (Hanley et al., 2000; Takase et al., National Science Council as well as by Academia Sinica, Republic 2000). of China (B-cC), by NIH Grant R01RR10715 and by NSF Grant IBN-9728587 (J.H.P. and Y.Y.). We thank the National Institutes of Health 1-G20-RR11724), National Science Foundation (STI- 9602828), M. J. Murdock Charitable Trust (96127:JVZ:02/27/97), Similar DNA-Binding and Transactivation and W. M. Keck Foundation (961582) for support of the University Properties of the Two Genes of Oregon Zebrafish Facility.

Our results showed that sox9a and sox9b are both detected in chondrogenic cells, but are expressed in different REFERENCES gonads. Do they have different activities in the cell? There have been many precedents of two transcription factor isoforms exerting opposite functions as transcription acti- Amores, A., Force, A., Yan, Y.-L., Wang, Y.-L., Fritz, A., Amemiya, C., Lynch, M., Prince, V., Ho, R., Ekker, M., and Postlethwait, J. vator and repressor. FosB/FosB2, Ad4BP/ELP, MyoD/Id, and (1998). Zebrafish hox clusters and vertebrate genome evolution. LAP/LIP are a few of these examples (Benezra et al., 1990; Science 282, 1711–1714. Descombes and Schibler, 1991; Morohashi et al., 1994; Yen Bell, D. M., Leung, K. K., Wheatley, S. C., Ng, L. J., Zhou, S., Ling, et al., 1991). The C-terminal transactivation domain of K. W., Sham, M. H., Koopman, P., Tam, P. P., and Cheah, K. S. Sox9b is very short, posing the question of whether it could (1997). SOX9 directly regulates the type-II collagen gene. Nature function as a transcription activator. We show that the Genet. 16, 174–178. C-terminal region of Sox9a and Sox9b can activate tran- Benezra, R., Davis, R. L., Lockshon, D., Turner, D. L., and Wein- scription in a dose-dependent manner. Therefore, Sox9a and traub, H. (1990). The protein Id: A negative regulator of helix- loop-helix DNA binding proteins. Cell 61, 49–59. Sox9b are both functional transcription activators in our Bi, W., Deng, J. M., Zhang, Z., Behringer, R. R., and de Crombrug- cell culture system. ghe, B. (1999). Sox9 is required for cartilage formation. Nature We dissected Sox9 into five domains and presented evi- Genet. 22, 85–89. dence indicating that domain C of Sox9a and Sox9b pos- Chan, F. Y., Robinson, J., Brownlie, A., Shivdasani, R. A., Donovan, sesses transactivation function. Human SOX9 has a major A., Brugnara, C., Kim, J., Lau, B. C., Witkowska, H. E., and Zon, transactivation domain at the carboxyl terminus and a L. I. (1997). Characterization of adult alpha- and beta-globin weaker one at amino acids 249–316, a region spanning the genes in the zebrafish. Blood 89, 688–700. C and D domains (McDowall et al., 1999). The internal Chiang, E. F.-L., Tong, S.-K., Yan, Y.-L., Hsiao, P.-H., Guiguen, Y., region of mouse SOX9 also has a weak activation function, Postlethwait, J., and Chung, B.-c. (2001). Two functional Cyp19 (P450 aromatase) genes on duplicated zebrafish chromosomes re although this phenomenon has not been carefully analyzed expressed in ovary or brain. Mol. Biol. Evol., in press. (Ng et al., 1997). It appears that maximal transactivation of Chiang, E. F.-L., Yan, Y.-L., Tong, S.-K., Hsiao, P.-H., Guiguen, Y., SOX9 requires the actions of both C-terminal and internal Postlethwait, J., and Chung, B.-c. (2000). Characterization of domains (McDowall et al., 1999). Because the E domains of duplicated zebrafish cyp19 genes. J. Exp. Zool., in press. zebrafish Sox9a and of Sox9b are short and weak in trans- Collignon, J., Sockanathan, S., Hacker, A., Cohen-Tannoudji, M., activation activity, the internal domain C apparently be- Norris, D., Rastan, S., Stevanovic, M., Goodfellow, P. N., and comes more important for transactivation. Nevertheless, Lovell-Badge, R. (1996). A comparison of the properties of Sox-3 zebrafish Sox9a and Sox9b in common with human SOX9 with sry and two related genes, Sox-1 and Sox-2. Development 122, 509–520. need both internal and C-terminal domains for full activa- de Martino, S., Yan, Y.-L., Jowett, T., Postlethwait, J. H., Varga, tion function. Z. M., Ashworth, A., and Austin, C. A. (2000). Expression of Sox9a and Sox9b are functional transcription activators sox11 gene duplicates in zebrafish suggests the reciprocal loss of that exert their physiological functions at the site of their ancestral gene expression patterns in development. Dev. Dyn. expression. Sox9a may have a role in testis, while Sox9b 217, 279–292. may be important for ovary development. In chondrogenic Denny, P., Swift, S., Brand, N., Dabhade, N., Barton, P., and cells, which express both sox9a and sox9b, functional Ashworth, A. (1992). A conserved family of genes related to the redundancy may exist to ensure the proper control of gene testis determining gene, SRY. Nucleic Acids Res. 20, 2887. Descombes, P., and Schibler, U. (1991). A liver-enriched transcrip- expression during cartilage differentiation, or the two genes tional activator protein, LAP, and a transcriptional inhibitory may play complementary roles. Gene duplication followed protein, LIP, are translated from the same mRNA. Cell 67, by differential expression of the duplicated gene may have 569–679. contributed to more flexible gene regulation and function in Efron, B., and Gong, G. (1983). A leisurely look at the bootstrap, the zebrafish development (Force et al., 1999). jacknife, and cross-validation. Am. Stat. 37, 36–48.

Ekker, M., Akimenko, M. A., Allende, M. L., Smith, R., Drouin, G., Kimmel, C. B., Ballard, W. W., Kimmel, S. R., Ullmann, B., and Langille, R. M., Weinberg, E. S., and Westerfield, M. (1997). Schilling, T. F. (1995). Stages of embryonic development of the Relationships among msx gene structure and function in ze- zebrafish. Dev. Dyn. 203, 253–310. brafish and other vertebrates. Mol. Biol. Evol. 14, 1008–1022. Knapik, E. W., Goodman, A., Atkinson, O. S., Roberts, C. T., Ekker, S. C., Ungar, A. R., Greenstein, P., von Kessler, D. P., Porter, Shiozawa, M., Sim, C. U., Weksler Zangen, S., Trolliet, M. R., J. A., Moon, R. T., and Beachy, P. A. (1995). Patterning activities Futrell, C., Innes, B. A., Koike, G., McLaughlin, M. G., Pierre, L., of vertebrate hedgehog proteins in the developing eye and brain. Simon, J. S., Vilallonga, E., Roy, M., Chiang, P. W., Fishman, Curr. Biol. 5, 944–955. M. C., Driever, W., and Jacob, H. J. (1996). A reference cross DNA Ellies, D. L., Langille, R. M., Martin, C. C., Akimenko, M. A., and panel for zebrafish (Danio rerio) anchored with. Development Ekker, M. (1997). Specific craniofacial cartilage dysmorphogen- 123, 451–460. esis coincides with a loss of dlx gene expression in retinoic Knapik, E. W., Goodman, A., Ekker, M., Chevrette, M., Delgado, J., acid-treated zebrafish embryos. Mech. Dev. 61, 23–36. Neuhauss, S., Shimoda, N., Driever, W., Fishman, M. C., and Felsenstein, J. (1985). Confidence limits on phylogenies: An ap- Jacob, H. J. (1998). A microsatellite genetic linkage map for proach using the bootstrap. Evolution 39, 783–791. zebrafish (Danio rerio). Nature Genet. 18, 338–343. Force, A., Lynch, M., Pickett, F. B., Amores, A., Yan, Y. L., and Komiya, T., Itoh, K., Ikenishi, K., and Furusawa, M. (1994). Isola- Postlethwait, J. (1999). Preservation of duplicate genes by tion and characterization of a novel gene of the DEAD box complementary, degenerative mutations. Genetics 151, 1531– protein family which is specifically expressed in germ cells of 1545. Xenopus laevis. Dev. Biol. 162, 354–363. Fo¨rnzler, D., Her, H., Knapik, E. W., Clark, M., Lehrach, H., Koopman, P., Gubbay, J., Vivian, N., Goodfellow, P., and Lovell- Postlethwait, J. H., Zon, L. I., and Beier, D. R. (1998). Gene Badge, R. (1991). Male development of chromosomally female mapping in zebrafish using single-strand conformation polymor- mice transgenic for Sry. Nature 351, 117–121. phism (SSCP) analysis. Genomics 51, 216–222. Korzh, V., Edlund, T., and Thor, S. (1993). Zebrafish primary Foster, J. W., Dominguez Steglich, M. A., Guioli, S., Kowk, G., neurons initiate expression of the LIM homeodomain protein Weller, P. A., Stevanovic, M., Weissenbach, J., Mansour, S., Isl-1 at the end of gastrulation. Development 118, 417–425. Young, I. D., Goodfellow, P. N., et al. (1994). Campomelic Kuratani, S., Matsuo, I., and Aizawa, S. (1997). Developmental dysplasia and autosomal sex reversal caused by mutations in an patterning and evolution of the mammalian viscerocranium: SRY-related gene. Nature 372, 525–530. Genetic insights into comparative morphology. Dev. Dyn. 209, Guo, I.-C., Huang, C., and Chung, B.-c. (1993). Differential regula- 139–155. tion of the CYP11A1 (P450scc) and ferredoxin genes in adrenal Kwok, C., Goodfellow, P. N., and Hawkins, J. R. (1996). Evidence to and placental cells. DNA Cell Biol. 12, 849–860. exclude SOX9 as a candidate gene for XY sex reversal without Guo, I.-C., Tsai, H.-M., and Chung, B.-c. (1994). Actions of two skeletal malformation. J. Med. Genet. 33, 800–801. different cAMP-responsive sequences and an adrenal enhancer of Lander, E. S., Green, P., Abrahamson, J., Barlow, A., Daly, M. J., human CYP11A1 (P450scc) gene in adrenal Y1 and placental Lincoln, S. E., and Newburg, L. (1987). MapMaker: An interactive JEG-3 cells. J. Biol. Chem. 269, 6362–6369. computer package for constructing primary genetic linkage maps Hanley, N. A., Hagan, D. M., Clement-Jones, M., Ball, S. G., of experimental and natural populations. Genomics 1, 174–181. Strachan, T., Salas-Cortes, L., McElreavey, K., Lindsay, S., Rob- Laudet, V., Stehelin, D., and Clevers, H. (1993). Ancestry and son, S., Bullen, P., Ostrer, H., and Wilson, D. I. (2000). SRY, diversity of the HMG box superfamily. Nucleic Acids Res. 21, SOX9, and DAX1 expression patterns during human sex determination and gonadal development. Mech. Dev. 91, 403–407. 2493–2501. Harley, V. R., Jackson, D. I., Hextall, P. J., Hawkins, J. R., Berkovitz, Lee, K. H., Xu, Q., and Breitbart, R. E. (1996). A new tinman-related G. D., Sockanathan, S., Lovell Badge, R., and Goodfellow, P. N. gene, nkx2.7, anticipates the expression of nkx2.5 and nkx2.3 in (1992). DNA binding activity of recombinant SRY from normal zebrafish heart and pharyngeal endoderm. Dev. Biol. 180, 722– males and XY females. Science 255, 453–456. 731. Houston, C. S., Opitz, J. M., Spranger, J. W., Macpherson, R. I., Lefebvre, V., Huang, W., Harley, V. R., Goodfellow, P. N., and de Reed, M. H., Gilbert, E. F., Herrmann, J., and Schinzel, A. (1983). Crombrugghe, B. (1997). SOX9 is a potent activator of the The campomelic syndrome: Review, report of 17 cases, and chondrocyte-specific enhancer of the pro alpha1(II) collagen gene. follow-up on the currently 17-year-old boy first reported by Mol. Cell. Biol. 17, 2336–2346. Maroteaux et al in 1971. Am. J. Med. Genet. 15, 3–28. McDowall, S., Argentaro, A., Ranganathan, S., Weller, P., Mertin, Hu, M.-C., Huang, Y.-Y., Hsu, N.-C., Li, H., and Chung, B.-c. (1999). S., Mansour, S., Tolmie, J., and Harley, V. (1999). Functional and Tissue-specific, hormonal and developmental regulation of structural studies of wild type SOX9 and mutations causing CYP11A1/LacZ expression in transgenic mice leads to adreno- campomelic dysplasia. J. Biol. Chem. 274, 24023–24030. cortical zone characterization. Endocrinology 140, 5609–5618. Mertin, S., McDowall, S. G., and Harley, V. R. (1999). The DNA- Hwang, S. P. L., Tsou, M. F., Lin, Y. C., and Liu, C. H. (1997). The binding specificity of SOX9 and other SOX proteins. Nucleic zebrafish BMP4 gene: Sequence analysis and expression pattern Acids Res. 27, 1359–1364. during embryonic development. DNA Cell Biol. 16, 1003–1011. Morais da Silva, S., Hacker, A., Harley, V., Goodfellow, P., Swain, Johnson, S. L., Gates, M. A., Johnson, M., Talbot, W. S., Horne, S., A., and Lovell Badge, R. (1996). Sox9 expression during gonadal Baik, K., Rude, S., Wong, J. R., and Postlethwait, J. H. (1996). development implies a conserved role for the gene in testis Centromere-linkage analysis and the consolidation of the ze- differentiation in mammals and birds. Nature Genet. 14, 62–68. brafish genetic map. Genetics 142, 1277–1288. Moreno Mendoza, N., Harley, V. R., and Merchant Larios, H. Kent, J., Wheatley, S. C., Andrews, J. E., Sinclair, A. H., and (1999). Differential expression of SOX9 in gonads of the sea turtle Koopman, P. (1996). A male-specific role for SOX9 in vertebrate Lepidochelys olivacea at male- or female-promoting tempera- sex determination. Development 122, 2813–2822. tures. J. Exp. Zool. 284, 705–710.

Morohashi, K.-i., Lida, H., Nomura, M., Hatano, O., Honda, S.-i., sion during gonadal development in the frog Rana rugosa. FEBS Tsukiyama, T., Niwa, O., Hara, T., Takakusu, A., Shibata, Y., Lett. 466, 249–254. and Omura, T. (1994). Functional difference between Ad4BP and Uwanogho, D., Rex, M., Cartwright, E. J., Pearl, G., Healy, C., ELP, and their distributions in steroidogenic tissues. Mol. Endo- Scotting, P. J., and Sharpe, P. T. (1995). Embryonic expression of crinol. 8, 643–653. the chicken Sox2, Sox3 and Sox11 genes suggests. Mech. Dev. 49, Ng, L. J., Wheatley, S., Muscat, G. E., Conway Campbell, J., Bowles, 23–36. J., Wright, E., Bell, D. M., Tam, P. P., Cheah, K. S., and Koopman, van de Wetering, M., Oosterwegel, M., van Norren, K., and Clevers, P. (1997). SOX9 binds DNA, activates transcription, and coex- H. (1993). Sox-4, an Sry-like HMG box protein, is a transcrip- presses with type II collagen during chondrogenesis in the tional activator in lymphocytes. EMBO J. 12, 3847–3854. mouse. Dev. Biol. 183, 108–121. Wagner, T., Wirth, J., Meyer, J., Zabel, B., Held, M., Zimmer, J., Oxtoby, E., and Jowett, T. (1993). Cloning of the zebrafish krox-20 Pasantes, J., Bricarelli, F. D., Keutel, J., Hustert, E., Wolf, U., gene (krx-20) and its expression during hindbrain development. Tommerup, N., Schempp, W., and Shcerer, G. (1994). Autosomal Nucleic Acids Res. 21, 1087–1095. sex reversal and compomelic dysplasia are caused by mutations Postlethwait, J. H., Johnson, S. L., Midson, C. N., Talbot, W. S., in and around the SRY-related gene SOX9. Cell 79, 1111–1120. Gates, M., Ballinger, E. W., Africa, D., Andrews, R., Carl, T., Westerfield, M. (1995). “The Zebrafish Book: A Guide for the Eisen, J. S., et al. (1994). A genetic linkage map for the zebrafish. Laboratory Use of Zebrafish (Danio rerio).” Univ. of Oregon Science 264, 699–703. Press, Eugene. Postlethwait, J. H., Yan, Y. L., Gates, M. A., Horne, S., Amores, A., Woods, I. G., Kelly, P. D., Chu, F., Ngo-Hazelett, P., Yan, Y.-L., Brownlie, A., Donovan, A., Egan, E. S., Force, A., Gong, Z., Huang, H., Postlethwait, J. H., and Talbot, W. S. (2000). A Goutel, C., Fritz, A., Kelsh, R., Knapik, E., Liao, E., Paw, B., comparative map of the zebrafish genome. Genome Res., in Ransom, D., Singer, A., Thomson, M., Abduljabbar, T. S., Yelick, press. P., Beier, D., Joly, J. S., Larhammar, D., Talbot, W. S., et al. (1998). Vertebrate genome evolution and the zebrafish gene map. Nature Wright, E., Hargrave, M. R., Christiansen, J., Cooper, L., Kun, J., Genet. 18, 345–349. Evans, T., Gangadharan, U., Greenfield, A., and Koopman, P. Schilling, T. F., and Kimmel, C. B. (1994). Segment and cell type (1995). The Sry-related gene Sox9 is expressed during chondro- lineage restrictions during pharyngeal arch development in the genesis in mouse embryos. Nature Genet. 9, 15–20. zebrafish embryo. Development 120, 483–494. Wright, E. M., Snopek, B., and Koopman, P. (1993). Seven new Selman, K., Wallace, R. A., Sarka, A., and Qi, X. (1993). Stages of members of the Sox gene family expressed during mouse devel- oocyte development in the zebrafish, Brachydanio rerio. J. Mor- opment. Nucleic Acids Res. 21, 744. phol. 218, 203–224. Yan, Y. L., Hatta, K., Riggleman, B., and Postlethwait, J. H. (1995). Spotila, L. D., Spotila, J. R., and Hall, S. E. (1998). Sequence and Expression of a type II collagen gene in the zebrafish embryonic expression analysis of WT1 and Sox9 in the red-eared slider axis. Dev. Dyn. 203, 363–376. turtle, Trachemys scripta. J. Exp. Zool. 281, 417–427. Yen, J., Wisdom, R. M., Tratner, I., and Verma, I. M. (1991). An Stock, D. W., Ellies, D. L., Zhao, Z., Ekker, M., Ruddle, F. H., and alternative spliced form of FosB is a negative regulator of tran- Weiss, K. M. (1996). The evolution of the vertebrate Dlx gene scriptional activation and transformation by Fos proteins. Proc. family. Proc. Natl. Acad. Sci. USA 93, 10858–10863. Natl. Acad. Sci. USA 88, 5077–5081. Sudbeck, P., Schmitz, M. L., Baeuerle, P. A., and Scherer, G. (1996). Yoon, C., Kawakami, K., and Hopkins, N. (1997). Zebrafish vasa Sex reversal by loss of the C-terminal transactivation domain of homologue RNA is localized to the cleavage planes of 2- and human SOX9. Nature Genet. 13, 230–232. 4-cell-stage embryos and is expressed in the primordial germ Swofford, D. L., Olsen, G. J., Waddell, P. J., and Hillis, D. M. (1996). cells. Development 124, 3157–3165. Phylogenetic Inference. In “Molecular Systematics” (D. M. Hil- lis, C. Moritz, and B. K. Mable, Eds.), pp. 407–514. Sinauer, Received for publication June 27, 2000 Sunderland, MA. Revised November 27, 2000 Takase, M., Noguchi, S., and Nakamura, M. (2000). Two Sox9 Accepted November 27, 2000 messenger RNA isoforms: Isolation of cDNAs and their expres- Published online January 26, 2001