Identification of chaperonin CCTγ subunit as a determinant of retinotectal development by whole-genome subtraction cloning from zebrafish no tectal neuron mutant

Zebrafish (Danio rerio) is an excellent model organism with which to investigate gene function in vertebrates because external fertilization,transparent embryos and random mutagenesis techniques facilitate phenotype-driven forward genetic analyses(Eisen, 1996). Large-scale screening of zebrafish mutagenized by N-ethyl-N-nitrosourea(ENU) identified ∼2400 mutations affecting development and behavior(Driever et al., 1996; Haffter et al., 1996). However, molecular cloning of ENU-induced mutant genes will require positional cloning and/or candidate gene approaches(Postlethwait and Talbot,1997), because ENU induces predominantly substitution mutations(Mullins et al., 1994; Solnica-Krezel et al., 1994). Thus, availability of polymorphic markers will be rate-limiting. In terms of cloning, insertion mutagenesis with pseudotype retroviruses is a straightforward approach to obtain mutated genes(Gaiano et al., 1996). Hopkins and colleagues isolated more than 500 insertional mutants(Golling et al., 2002), but the efficiency of mutagenesis was lower than ENU mutagenesis(Amsterdam et al., 1999).

To investigate the molecular mechanisms underlying neural network formation, we developed a highly efficient mutagenesis procedure using 4,5′,8-trimethylpsoralen (TMP) (Ando and Mishina, 1998). In a pilot screen, we isolated and characterized two mutant lines with abnormalities in the nervous system. The no tectal neuron (ntn) mutation impaired the development of tectal neurons and eyes, whereas the edawakare mutation affected the arborization of the trigeminal ganglion and Rohon-Beard sensory neurons. TMP is a DNA crosslinking agent that can frequently induce small deletions in combination with UV irradiation in Escherichia coli and Caenorhabditis elegans (Sladek et al., 1989; Yandell et al.,1994; Liu et al.,1999). A potential advantage of the method would be that one could isolate mutagenized genes directly by whole-genome subtraction using TMP-induced deletions as molecular tags. Representational difference analysis(RDA) is a powerful subtraction method of the entire genome(Lisitsyn and Wigler,1993).

We have applied RDA to characterize the genomic region of the TMP-induced ntn mutation (Ando and Mishina,1998). Successful isolation of tightly linked polymorphic markers by the whole-genome subtraction method led to the construction of genetic and physical maps of the zebrafish genomic region responsible for retinotectal development. RT-PCR analysis of transcripts from the ntn region identified a 143 bp deletion in the cct3 gene encoding the γsubunit of chaperonin containing TCP1 (CCT, also called the TCP1 ring complex or TriC). Injection of antisense cct3 morpholino oligonucleotides into zebrafish embryos induced characteristic ntn phenotypes,including the degeneration of retinal ganglion cells and tectal neurons. Furthermore, injection of cct3 mRNA successfully rescued ntnmutant embryos. These results revealed that chaperonin CCTγ controls specifically retinotectal development in zebrafish. Our results open a novel TMP mutagenesis-RDA cloning strategy of zebrafish forward genetics characterized by high efficiency and rapid cloning.

Zebrafish of the AB strain were raised and kept under standard conditions at ∼28°C (Mullins et al.,1994). The ntn mutant line jt5 was isolated by TMP mutagenesis previously (Ando and Mishina, 1998). The ntn mutant embryos at 3 days postfertilization (dpf) showed small eyes and turbid tectum. Mutation carriers were identified by random intercrosses and were then outcrossed to wild-type zebrafish of the AB strain. Mutant embryos were obtained by crossing heterozygous fish. The transgenic zebrafish line carrying the nicotinic acetylcholine receptor β3 (nAChRβ3) gene promoter-driven enhanced green fluorescent protein (EGFP) expression vector (PAR-EGFP) has been established previously (Tokuoka et al.,2002). Embryos were raised at 28.5°C in embryo rearing solution (ERS) (Easter and Nicola,1996). For microscopic observation, 0.2 mM phenylthiocarbamide was added to ERS at 12 hours postfertilization (hpf) to prevent melanocyte pigmentation (Westerfield,1995).

Genomic DNAs were prepared from pools of 40 mutant embryos and 40 wild-type siblings and RDA was performed essentially as described(Lisitsyn et al., 1994; Lisitsyn and Wigler, 1995). Genomic DNAs (1 μg of each) were digested with BglII, EcoRI, HindIII, SpeI and XbaI, and were PCR-amplified to generate amplicons. Primer sequences were designed according to the protocol (Lisitsyn and Wigler,1995) with some modifications on their cohesive end sequences compatible for the restriction endonuclease(Table 1). The iterative hybridization-amplification step was repeated three times for BglII, EcoRI, SpeI or XbaI amplicons and four times for HindIII amplicons. The resulting RDA products were digested with the corresponding restriction enzymes, agarose gel-isolated and cloned into pBluescript II SK(+) (Stratagene). Genomic DNA with respective restriction enzymes and amplicon (5 μg of each) were electrophoresed in 2% agarose gels and transferred to Hybond-N⁺ nylon membranes (Amersham). Membranes were hybridized with RDA products labeled using a random primed DNA labeling kit (Roche) as probes.

Genomic DNA was extracted from 93 pools of five ntn mutant embryos obtained from crosses of heterozygous fish. One μg genomic DNA from each pool was digested with BglII, EcoRI, HindIII, SpeI or XbaI. Digested DNAs were PCR-amplified to generate amplicons as above. Southern hybridization analyses of amplicons with RDA products were performed.

Genomic sequences flanking the restriction fragment length polymorphism(RFLP) sites of RDAjt5B430, RDAjt5B460 and RDAjt5E340 were cloned by vectorette-PCR using a GenomeWalker kit (Clontech), and primers were designed to identify cleaved amplified polymorphic sequences (CAPS): RDAjt5B430 RFLP site, 5′-TGGCGGTTCATTCTGCTGTGCGAC-3′ and 5′-CGACAAGACTTTTGTCAGGTAG-3′; RDAjt5B460 RFLP site,5′-CAATACCGGCAACTTTCAAC-3′ and 5′-CAAGGACAAGAAATCATGCC-3′; and RDAjt5E340 RFLP site,5′-GTCAAAATGCTCACTATACTAACTGCTGTC-3′ and 5′-AGTTTCGGCTTGGTTACGGAATCTC-3′. Mapping of YD1 and YH2 loci was carried out by PCRs using primers flanking the deletions; YD1 locus,5′-GACAGTGGAAATGCGGCTAT-3′ and 5′-TACCCATGTCTTCTGCGTAG-3′; YH2 locus,5′-GGCCAGAGTTTACATAGGGGT-3′ and 5′-GGTTTTTGCTGTGTCTGCCTG-3′. Radiation hybrid (RH) mapping was performed on the Goodfellow zebrafish T51 RH panel (Research Genetics)(Geisler et al., 1999) using the primers for the RDAjt5E340 RFLP site and YD1 locus.

YAC clones D04128 and H0145 were obtained by PCR screening of a zebrafish YAC library (Resource Center/Primary Databank, Germany) with the primers for RDAjt5E340 RFLP site and their terminal sequences were determined as described(Zhong et al., 1998). PAC 10J03 and BAC 18M9 clones were isolated from zebrafish BAC and PAC libraries(Genome Systems) using ³²P-labeled RDAjt5B430 and RDAjt5E340 as probes, respectively. The inserts of these clones were sized by pulsed-field gel electrophoresis using CHEF-Mapper (BioRad). For YAC clonality analysis,the blot was probed with the ³²P-labeled YAC arm pRML plasmid.

Zebrafish genomic DNA in YAC D04128 clone purified by pulsed-field gel electrophoresis was biotinylated using a random primed DNA labeling kit(Roche). A random primed zebrafish cDNA library was synthesized using RNA from zebrafish embryos at 36 hpf as templates (SuperScript Choice System,Invitrogen) and PCR-amplified after ligation to adapters 5′-TAGTCCGAATTCAAGCAAGAGCAGA-3′ and 5′-CTCTTGCTTGAATCGGACTA-3′. After preincubation with 2 μg sonicated zebrafish genomic DNA and 1 μg HaeIII-digested yeast genomic DNA, 1 μg preamplified cDNA was hybridized with 100 ng biotinylated zebrafish genomic DNA in YAC D04128 clone as described(Del Mastro and Lovett, 1996). The sequences of the 5′ and 3′ regions of the cct3 cDNA were obtained by 5′- and 3′-rapid amplification of cDNA ends(RACE) using a SMART cDNA kit (Clontech), respectively. The entire coding sequence of cct3 cDNA was cloned into pCRII vector (Invitrogen) by RT-PCR with primers, 5′-TGTCCGGTACCGGTGATCTAAC-3′ and 5′-AAATGGATTTCTGATGAGAACGTTGT-3′ to yield pCRII-CCT3. The deletion mutation in the cct3 mRNA was identified by sequencing RT-PCR products of mRNAs from ∼200 ntn homozygous embryos and ∼650 wild-type siblings at 50 hpf according to the protocol of the SuperScript Choice System (Invitrogen). The mutation in the cct3 gene was confirmed by PCR on 100 ng genomic DNAs from ∼300 wild-type siblings and∼100 ntn embryos with primers flanking the 143 bp deletion,5′-GCCATGCAAGTGTGTCGTAATG-3′ and 5′-CTCAGAGAAGTGAGCACACGAATG-3′. Genotyping of embryos was performed using the same primer set.

The entire coding sequence of the zebrafish brn3b/pou4f1 was obtained by RT-PCR using primers 5′-CGGTCGCAAATATGATGATG-3′ and 5′-ATGATTCCACATCCCCTTTG-3′, and was cloned into pCRII vector to yield pCRII-BRN3B (GenBank Accession Number AB122025). We carried out whole-mount in situ hybridization with antisense RNA probes prepared with a DIG RNA labeling kit (Roche), paraffin-embedding and sectioning of whole embryos as described previously (Mori et al., 1994; Jowett,1999). Deparaffinized sections were counterstained with 0.5%Methyl Green for 10 minutes. Probes for ath5/lakritz, dlx2, hlx1, krox20,myod, ntl, pax2a/noi (previously pax2.1), shh and zash1 mRNAs were as described previously(Masai et al., 2000; Akimenko et al., 1994; Seo et al., 1999; Oxtoby and Jowett, 1993; Weinberg et al., 1996; Schulte-Merker et al., 1994; Krauss et al., 1991; Krauss et al., 1993; Allende and Weinberg,1994).

Whole-mount immunostaining of zebrafish embryos with anti-acetylated tubulin antibody (Sigma) were carried out as previously described(Hammerschmidt et al., 1996)except that Alexa 488 anti-mouse IgG antibody was used as secondary antibody. Immunostaining of cryosectioned embryos using a monoclonal antibody zn5(Oregon Monoclonal Bank) was performed as described(Masai et al., 2003). Terminal deoxynucleotide transferase-mediated dUTP nick-end labeling (TUNEL) was carried out in cryosections according to the manufacturer's protocol (ApopTag Apoptosis Detection Kit; Serologicals Corporation).

Dechorionated embryos were soaked in ERS containing 5 μg/ml Acridine Orange (Sigma) or 100 μM Bodipy-ceramide (Fl C5, Molecular Probes) and 2%DMSO for 30 minutes in the dark. The embryos were washed, anesthetized by 0.02% 3-aminobenzoic acid ethyl ester (tricaine, Sigma) and embedded in low-melting temperature agarose gels during microscopic observation. The eyes and tectum of Bodipy-ceramide-stained embryos were scanned by confocal microscopy.

The fluorescein-tagged morpholino oligonucleotide complementary to the nucleotide residues –16 to +9 of the zebrafish cct3 mRNA(nucleotide residues are numbered from the putative translational initiation codon, GenBank Accession Number AF506209)(Golling et al., 2002) was obtained from GeneTools. The antisense or control oligonucleotide at a concentration of 4 μg/μl in 1× Danieau buffer(Nasevicius and Ekker, 2000)was injected into the yolk of one- to four-cell stage wild-type or transgenic embryos carrying PAR-EGFP using a microinjector (IM-300, Narishige). Distribution of oligonucleotides in embryos was monitored under fluorescent microscopy. Fluorescence of EGFP was much stronger and easily distinguishable from that of fluorescein. Fluorescence of EGFP-labeled retinal ganglion cell(RGC) axons was observed as described(Tokuoka et al., 2002).

The 1.8 kb EcoRV-SpeI fragment from pCRII-CCT3 was cloned into the StuI and XbaI sites of pCS2+ vector(Turner and Weintraub, 1994)to yield pCS-CCT3. Capped cct3 mRNA was prepared from 1 μg pCS-CCT3 linearized with NotI using an mMessage mMachine kit(Ambion). We injected 110-270 pg mRNA into the cytoplasm of embryos produced by crossing heterozygous fish.

Zebrafish no tectal neuron (ntn) mutants isolated by TMP mutagenesis showed degeneration of tectal neurons and eyes(Ando and Mishina, 1998). The ntn^jt5 allele was of recessive inheritance and fully penetrant. If TMP mutagenesis induces deletions in the zebrafish genome as in E. coli and C. elegans(Sladek et al., 1989; Yandell et al., 1994; Liu et al., 1999), it would be possible to directly clone deleted DNA segments by subtraction between pooled wild-type and mutant genomes. Alternatively, such whole-genome subtraction may allow us to obtain DNA segments tightly linked to the ntn locus if there were enough RFLPs in the AB strain genome. We subtracted the mutant genome from the wild-type sibling genome by the RDA method(Lisitsyn and Wigler, 1993). Zebrafish genomic DNA was digested with various 6 bp recognizing restriction endonucleases, and five enzymes that produced DNA fragments with short mean lengths appropriate for RDA were selected. We prepared PCR-amplified DNA fragments (amplicons) of genomic DNA from pools of 40 ntn mutant embryos and 40 wild-type siblings using each of BglII, EcoRI, HindIII, SpeI and XbaI restriction enzymes. Whole-genome subtraction was carried out using the amplicons from the wild-type genome as the tester and those from the ntn mutants as the driver. After three or four rounds of the subtractive hybridization amplification, difference products became detectable as clearly seen bands in agarose gels from respective amplicons(Fig. 1A). We thus successfully obtained six RDA products from the BglII amplicons designated as RDAjt5B430 (B430), RDAjt5B440 (B440), RDAjt5B460 (B460), RDAjt5B470 (B470),RDAjt5B610 (B610) and RDAjt5B630 (B630); five products from the EcoRI amplicons designated as RDAjt5E230 (E230), RDAjt5E340 (E340), RDAjt5E420(E420), RDAjt5E550 (E550) and RDAjt5E560 (E560); two products from HindIII amplicons designated as RDAjt5H400 (H400) and RDAjt5H480(H480); one product from the SpeI amplicons designated as RDAjt5S610(S610); and five products from the XbaI amplicons designated as RDAjt5X300 (X300), RDAjt5X350 (X350), RDAjt5X430 (X430), RDAjt5X520 (X520) and RDAjt5X600 (X600). Southern blot hybridization analysis of genomic DNA from ntn mutants and wild-type siblings showed that each RDA product hybridized with two fragments in the wild-type genome, but with only the larger one in the mutant genome digested by the corresponding restriction enzymes (Fig. 1B). Two hybridizing fragments were also present in the genome of the AB strain. These results suggest that all the RDA products resulted from RFLPs between the wild-type and ntn mutant genomes, rather than deletions in the mutant genome. We thus obtained polymorphic markers linked to the ntn locus by genetically directed RDA.

We then genetically selected polymorphic markers close to the ntnlocus by Southern blot hybridization analysis of amplicons prepared from genomic DNA of 465 ntn mutant embryos. Genomic DNAs isolated from 93 pools of five mutant embryos were digested with the five restriction endonucleases used for RDA and amplicons were prepared. Blots of 93 amplicons were hybridized with the RDA products (Fig. 1C). The numbers of hybridization positive amplicons in the 93 amplicons were one for B460, H480, X300, X350 and X430; two for B440, B470,B610, B630, E230, E420, E560, S610 and X520; four for H400; six for X600; and eleven for E550. The genotyping of 93 mutant embryo pools revealed the positions of the polymorphic markers relative to the ntn locus(Fig. 1D). No recombinants among the ntn, B430 and E340 loci were found in 465 mutant embryos. The genetic mapping localized the ntn locus between the B440/E230/E560 and B460/H480/X300/X350/X430 loci.

We further determined the genotypes of 432 ntn mutant embryos for the B430 and E340 loci using CAPS markers. There was one recombination event between the ntn and B430 loci, the calculated genetic distance being 0.12 cM (95% confidence interval, 0.00-0.64 cM). However, we found no recombination events between the ntn and E340 loci. Thus, among isolated polymorphic RDA markers, E340 was the marker closest to the ntn locus.

We next screened zebrafish genomic libraries with E340 as a probe and obtained two YAC clones and one PAC clone(Fig. 1E). YAC H0145 contained a terminal sequence of BAC clone 18M9 obtained by screening with B430 as a probe. By comparing the 2.4 kb terminal sequences of YAC D04128 and the 2.9 kb terminal sequences of YAC H0145 with the corresponding genomic sequences of wild-type and ntn mutant embryos, we found 28 bp and 107 bp deletions in the ntn genome, designated as YD1 and YH2, respectively(Fig. 1E). Genotyping of 432 ntn mutant embryos by PCR using primers flanking the deletions in respective YAC ends identified one recombination event out of 864 meioses between the ntn and YH2 loci and one between the ntn and YD1 loci. We detected ten recombinations between the ntn and B460 loci by genotyping 432 ntn mutants using B460 CAPS marker. Thus, the relative order of markers around the ntn locus on the chromosome was B460-YH2-ntn/E340-YD1-B430 (Fig. 1E). These analyses localized the ntn locus within a 0.24 cM region between YH2 and YD1 markers.

Using YAC D04128 as a probe, we isolated ∼300 cDNA clones by screening a cDNA library prepared from zebrafish embryos at 36 hpf. Sequence analysis suggested that these clones encoded at least three genes. The first group of 26 cDNA clones and zebrafish ESTs (EST269466, fb13f04, fc26d03, fc72f03 and fe18d09) from the database of the Washington University zebrafish EST project encoded the γ subunit of zebrafish chaperonin CCT, which shared 87%,86%, 85%, 70% and 58% amino acid sequence identities with the Xenopus laevis, mouse, human, Drosophila melanogaster and yeast counterparts, respectively (Chen et al.,1994; Kubota et al.,1994; Dunn and Mercola,1996; Walkley et al.,1996a; Walkley et al.,1996b; Walkley and Malik,1996). The second group of five cDNA clones encoded a putative protein that had 69% amino acid sequence identity with human C18orf1, a transmembrane protein with a LDL receptor type A domain(Yoshikawa et al., 1998). A putative protein encoded by the third group of seven cDNAs showed 32% amino acid sequence identity with mouse semaphorin 6C precursor(Kikuchi et al., 1999). One cDNA contained a single open reading frame, but there found no proteins homologous to the putative protein. The rest of the cDNA clones contained(CA)n dinucleotide repetitive sequences of various length and were not characterized further.

By RT-PCR analysis of mRNAs, we found a 143 bp deletion in the CCT γsubunit gene (cct3) transcript from the mutant embryos(Fig. 2A,B). The deletion was present in the cct3 gene of mutant embryos, but not in the wild-type gene (Fig. 2C). The γsubunit of zebrafish CCT consisted of 543 amino acids and had five regions common to the CCT subunit proteins (Kim et al., 1994), the N- and C-terminal equatorial ATPase domains, two intermediate domains and the apical substrate-binding domain(Fig. 2A). The cct3transcript in the mutant embryos completely lacked a putative ATPase motif and the reading frame was shifted by the deletion. There were no mutations within the coding sequences of three other candidate gene transcripts.

We studied the expression of the cct3 gene by in situ hybridization analysis (Fig. 2D). No transcripts were detectable in one-cell, shield and 90%epiboly stage embryos at 0.2, 6 and 9 hpf. At 12 hpf, strong hybridization signals for the cct3 mRNA appeared in the entire embryos. The cct3 mRNA was detected throughout the body at 24 hpf, and the brain,eyes and somites showed strong hybridization signals. At 36 hpf, hybridization signals for the cct3 mRNA were observed in the entire brain and somites.

The characteristic phenotypes of ntn mutant embryos were selective impairment of development of the eyes and tectum at ∼40 hpf(Ando and Mishina, 1998). At later stages (4 dpf), ntn mutants can be macroscopically distinguished from their wild-type siblings as having small eyes and turbidity in the developing tectum (Fig. 3A). There appeared no other abnormalities even at this stage except for small pectoral fins and some underdeveloped jaw skeletons in ntn mutant embryos.

At 36 hpf, Acridine Orange staining(Abrams et al., 1993; Furutani-Seiki et al., 1996)showed strong signals selectively in the tectum of mutant embryos, suggesting apoptotic cell death (Fig. 3B). In fact, staining with Bodipy-ceramide, which was freely distributed in the interstitium between cells and delineated cellular membrane(Cooper et al., 1999),revealed patches of cell-free space probably reminiscent of engulfed dead cells in the tectum of mutant embryos at 46 hpf(Fig. 3C). Furthermore, the tectum of the mutant embryos at 48 hpf was hardly stained by antibody against acetylated tubulin (Fig. 3D), a marker protein of mature neurites (Chitnis and Kuwada, 1990), as described previously(Ando and Mishina, 1998). Neurites in the retinal ganglion cell layer were poorly developed and the bundle of the optic nerve was faint in ntn mutants(Fig. 3E). However, the staining patterns of other neurons such as axons of the anterior commissure neurons, hindbrain neurons, trigeminal ganglion cells and dorsal longitudinal fascicles were comparable between wild-type and mutant embryos(Fig. 3D,E).

To examine whether the mutation in the cct3 gene was responsible for the ntn phenotypes, we injected an antisense morpholino oligonucleotide complementary to the cct3 mRNA sequence encompassing the translation start codon into the wild-type embryos. The injected embryos showed small eyes at 4 dpf (Fig. 3F) and Acridine Orange staining in the tectum at 36 hpf(Fig. 3G). Bodipy-ceramide staining revealed loss of tectal cells(Fig. 3H). Anti-acetylated tubulin antibody immunostaining demonstrated that the formation of tectal neuropil and optic nerve was impaired in the embryos injected with an antisense morpholino oligonucleotide (Fig. 3I,J). Injection of a control morpholino oligonucleotide with the inverted antisense sequence exerted little effect on the development of the tectal and retinal neurons (Fig. 3F-J). These results suggest that suppression of CCTγexpression by the antisense morpholino oligonucleotide induces the characteristic phenotypes of ntn mutant embryos.

We next injected wild-type cct3 mRNA into one-cell stage embryos derived from crosses of heterozygous (+/–) fish to examine whether the cct3 gene could rescue ntn mutant embryos. The injected embryos were stained with Acridine Orange for testing the ntn mutant phenotypes (Fig. 3K). Among 77 embryos injected with cct3 mRNA, only four embryos (5%) showed strong Acridine Orange staining in the tectum at 36 hpf and 73 embryos (95%) appeared normal, which deviated significantly from recessive inheritance(χ² test, P<0.001). The genotyping of embryos with wild-type phenotypes revealed that 11 of 73 embryos were cct3^-/- at the ntn locus. The genotype at the ntn locus of 20 embryos was cct3^+/+ and that of remaining 42 embryos was cct3^+/–. Among 137 mock-injected embryos, 37 embryos (27%) showed the ntn mutant phenotypes (Fig. 3K) and 100 embryos (73%) exhibited the wild-type phenotypes, which was consistent with a recessive mode of inheritance (χ² test, P=0.70). In 100 mock-injected embryos showing wild-type phenotypes, there were no embryos with cct3^-/- genotype at the ntn locus. All the embryos with wild-type phenotypes had genotypes of either cct3^+/+ or cct3^+/– at the ntn locus. The development of the eyes at 3 dpf was normal in the embryos that did not show an increase in Acridine Orange staining in the tectum. These results suggest that injection of cct3 mRNA rescued ntn mutant embryos.

The gene encoding nAChRβ3 is an early differentiation marker of RGCs(Matter-Sadzinski et al.,2001). To examine the differentiation of RGCs, we crossed the heterozygous ntn fish with a transgenic zebrafish carrying the nAChRβ3 gene promoter-directed EGFP expression vector(Tokuoka et al., 2002). Crossing of doubly transgenic fish with heterozygous ntn fish yielded homozygous mutant embryos with EGFP-labeled RGCs. Expression of EGFP signals in ntn embryos injected with cct3 mRNA indicated the restoration of development of RGCs (Fig. 3L). In control mock-injected embryos, no EGFP signals appeared in the retina. Genotypes of ntn embryos injected with cct3 mRNA showing wild-type phenotypes were confirmed to be cct3^-/-at the ntn locus by PCR using the primers flanking the deletion in the cct3 gene (Fig. 3M).

Chaperones play an important role in folding of many proteins and CCT is a member of two major chaperone systems implicated in cytoplasmic protein folding in eukaryotes (Bukau and Horwich,1998; Hartl and Hayer-Hartl,2002). By 12 hpf, the strong expression of the cct3 mRNA started in entire zebrafish embryos and continued thereafter(Fig. 2D). In the segmentation period (10-24 hpf), a variety of morphogenetic movements occur: the somites develop, the rudiments of the primary organs become visible and the overall body length of the embryo increases very rapidly(Kimmel et al., 1995). Thus,we investigated whether there were any changes in body patterning or neurogenesis in ntn mutant embryos at 30 hpf. Two zebrafish homologs of the neurogenic gene achaete-scute, zash1a (asha –Zebrafish Information Network) and zash1b, were strongly expressed in the entire neural retina (Fig. 4A), and in the midbrain and hindbrain(Fig. 4B) of both wild-type and ntn embryos at 30 hpf, respectively. Between wild-type and ntn embryos, there were no differences in the expression patterns of the homeobox gene hlx1 (dbx1a – Zebrafish Information Network) in the midbrain (Fig. 4C), dlx2 in the telencephalon and diencephalon(Fig. 4D), krox20(egr2b – Zebrafish Information Network) in rhombomeres 3 and 5(Fig. 4E), and pax2ain the midbrain-hindbrain boundary (Fig. 4G). The expression patterns of dlx2 in the pharyngeal arches (Fig. 4D), shhin midline structures (Fig. 4F), pax2a in the optic stalk, otic vesicle and pronephric duct (Fig. 4G), ntl in the notochord (Fig. 4H) and myod in the myotomes(Fig. 4I) were also comparable between wild-type and mutant embryos. Thus, the development of zebrafish embryos appeared to have proceeded normally until 30 hpf without the CCTγ subunit.

To examine the role of the cct3 gene more precisely, we analyzed retinotectal development in ntn mutant embryos by Bodipy-ceremide and TUNEL staining and by expression patterns of differentiation markers. Zebrafish neural retina consists of proliferating neuroepithelial cells at 27 hpf (Hu and Easter, 1999). The differentiation of RGCs began in the ventronasal region of the retina at 28 hpf, then increasingly spread over the central region of the retina until 37 hpf (Hu and Easter, 1999; Malicki, 1999; Schmitt and Dowling, 1999; Tokuoka et al., 2002). Bodipy-ceremide staining showed that retinal neuroepithelial cells with elongated shapes were arranged radially and expanding in both wild-type and mutant embryos at 27 hpf and 36 hpf (Fig. 5A). No differences were detectable in cellular morphology and alignment between wild-type and mutant retinae at these stages. At 48 hpf, the number of retinal cells increased and their radial alignment became less prominent in wild-type embryos (Fig. 5A). However, the staining revealed a loss of significant numbers of cells in the inner region of the retina in mutant embryos, although differentiation of the lens vesicle to the lens and onionskin organization of the lens proceeded normally. Consistent with the results of Bodipy-ceremide staining, there were no significant differences in TUNEL staining between wild-type and ntn mutant embryos at 30 hpf and 36 hpf(Fig. 5B). At 48 hpf, clusters of TUNEL-positive cells were found in the peripheral region of the mutant retina, suggesting the occurrence of apoptotic cell death. However, the ciliary marginal zone where neuroepithelial cells proliferated was TUNEL negative. These results suggest that the ntn mutation affected postmitotic developments rather than the proliferation of neuroepithelial cells.

RGCs are the first postmitotic neurons to be born in the zebrafish retina(Hu and Easter, 1999). We thus examined retinal development by immunostaining with monoclonal antibody zn5,an RGC differentiation marker (Trevarrow et al., 1990). The zn5 antibody staining showed strong signals in the central region of the wild-type retina at 48 hpf, whereas little staining was detectable in the ntn mutant retina, indicating the failure of RGC differentiation (Fig. 5C). Disturbance of retinal lamination was observed in mutant embryos at 72 hpf(data not shown). To further investigate the development of RGCs, we examined the expression of the nAChRβ3 gene, an early RGC differentiation marker(Matter-Sadzinski et al.,2001), by crossing ntn heterozygotes with transgenic zebrafish carrying the nAChRβ3 gene promoter-driven EGFP transgene(Tokuoka et al., 2002). In wild-type embryos, RGCs with EGFP signals appeared in the ventronasal region of the retina at 30 hpf, spread in the central region of the retina at 36 hpf and expanded in the inner retina to form the ganglion cell layer at 48 hpf(Fig. 5D). Retinal axons crossed the midline at 36 hpf and were extending towards the tectum at 48 hpf. In the mutant embryos, however, little EGFP signal was detectable in RGCs at 30 hpf, 36 hpf and 48 hpf (Fig. 5D). By contrast, EGFP signals in the trigeminal ganglion cells,Rohon-Beard sensory neurons and pituitary gland of mutant embryos were as strong as those of the wild-type embryos. The dorsal longitudinal fascicles and the axon fasciculation of trigeminal ganglion cells were comparable between wild-type and mutant embryos. Immunostaining with antibodies against acetylated tubulin, a general differentiation marker of neurons(Chitnis and Kuwada, 1990),showed that a small number of RGCs in the central region of the wild-type retina extended their axons at 30 hpf and 36 hpf(Fig. 5E). Anti-acetylated tubulin antibody stained RGC axons expanded in the inner retina and formed thick bundles at 48 hpf. However, little immunoreactivity to acetylated tubulin was detectable in the central region of the ntn mutant retina. At 36 hpf and 48 hpf, there were faint signals in the central region of the mutant retina, indicating poor development of RGCs. We next examined the expression of the proneural basic helix-loop-helix transcription factor atonal homolog 5 (ath5/lakritz; atoh7 – Zebrafish Information Network) as a marker for retinal neurogenesis by in situ hybridization. The expression pattern of ath5 in the retina was comparable between wild-type and mutant embryos at 33 hpf(Fig. 5F). Furthermore, a zebrafish homolog of the POU transcription factor brn3b(pou4f1 – Zebrafish Information Network), a gene essential for the differentiation of RGCs (Xiang,1998), showed similar expression patterns in the wild-type and mutant retinae at 36 hpf and 48 hpf (Fig. 5G). These results suggest that ntn mutation of the cct3 gene exerted little effect on the commitment of retinal neuroepithelial cells to postmitotic retinal neurons but severely impaired the differentiation of retinal neuroepithelial cells to RGCs. Apoptosis of retinal cells observed after the impairment of RGC differentiation may be secondary effects of the ntn mutation.

We also analyzed the development of tectal neurons by staining with Bodipy-ceremide (Fig. 6A). Zebrafish tectal cells proliferated over the whole extent of the tectal plate at 24 hpf and many tectal precursor cells turned into postmitotic cells forming the central differentiated zone by 48 hpf, while cells in the peripheral marginal zone still remained proliferative(Wullimann and Knipp, 2000). Large and round cells representing mitotically active cells(Cooper et al., 1999) were found along the edge of the tectum in wild-type embryos at 28 hpf and 36 hpf. During development from 28 hpf to 36 hpf, the total volume of the tectum remained relatively constant, whereas each tectal neuron precursor became smaller in wild-type embryos. The ventricle between the tectum and the posterior tectal membrane became less prominent in wild-type embryos from 36 hpf to 40 hpf. There were no detectable abnormalities in the organization of the tectal neuroepithelium, alignment and mitotic cell images of tectal precursor neurons in ntn embryos at 28 hpf and 36 hpf. At 40 hpf,however, staining revealed cell-free spaces in the central zone of the tectum in mutant embryos. Consistently, TUNEL staining signals appeared in the tectum of mutant embryos at 40 hpf, but not at 30 hpf and 36 hpf(Fig. 6B). Immunostaining with anti-acetylated tubulin visualized the axons of trigeminal ganglion cells extending along the epidermis over the tectum and there were no significant differences in the immunostaining patterns of the tectum between wild-type and mutant embryos at 30 hpf and 36 hpf (Fig. 6C). At 40 hpf, immunostaining showed the formation of tectal neuropil by vigorous neurite extension of tectal neurons in wild-type embryos,but there was little staining in mutant embryos. However, the expression pattern of brn3b in tectal cells was comparable between wild-type and mutant embryos at 48 hpf (Fig. 6D), indicating the presence of tectal neurons. These results suggest that the ntn mutation of the cct3 gene exerted little effect on the production of tectal cells but suppressed their differentiation to form tectal neuropil.

We developed TMP mutagenesis procedures in zebrafish in an attempt to facilitate the cloning of the mutated genes(Ando and Mishina, 1998). TMP,in combination with UV irradiation, forms interstrand crosslinks with DNA double helices and frequently results in deletions, possibly via incomplete nucleotide excision and recombination repair in E. coli and C. elegans (Cimino et al.,1985; Sladek et al.,1989; Yandell et al.,1994). Subtraction of the mutant genome from the wild-type genome should yield deleted genes or tightly linked RFLP markers. The ability to directly clone the mutated gene by the whole-genome subtraction method of Lisitsyn and Wigler (Lisitsyn and Wigler,1993) crucially depends upon the size of a deletion and the presence of two restriction enzyme sites within the deleted segment. Because the sizes of deletions were 0.1 to 15 kb in C. elegans(Yandell et al., 1994; Jansen et al., 1997; Liu et al., 1999), we selected five restriction enzymes suitable for PCR-based amplicon preparations from zebrafish genomic DNA to increase the probability of success. Three to four rounds of subtraction between ntn and wild-type littermate genomes digested with respective restriction enzymes yielded one to six RDA products. We thus successfully isolated polymorphic markers tightly linked to the ntn locus, although none of the RDA products was deleted in the mutant genome. These markers enabled us to obtain YAC clones carrying the ntn locus and to identify a deletion in the chaperonin cct3gene as the cause of ntn mutation.

Our results demonstrate that the combination of TMP mutagenesis and genetically directed RDA provides a highly efficient and rapid cloning strategy for zebrafish forward genetics. Using RDA products from a TMP-induced zebrafish mutant, vibrato, with defects in the spontaneous contraction and touch response, we also successfully constructed a high-resolution physical map of a genomic region of 720 kb containing the mutant locus (Sato and Mishina,2003). The whole-genome subtraction method will be also applicable to zebrafish mutants induced by ENU. In addition to the 143 bp deletion in the cct3 gene, there were two small deletions in ∼5 kb ntngenomic region, indicating successful TMP deletion mutagenesis in zebrafish. We also found larger deletions in the genome of edawakare^jt10 mutant zebrafish obtained by TMP mutagenesis(T. Morita, unpublished). Thus, direct selection by RDA of the mutated genes from TMP-induced mutant fish would be feasible depending on the sizes of deletions induced. In C. elegans, deletion sizes were dependent on TMP concentrations (Gengyo-Ando and Mitani, 2000).

In the present investigation, we identified the γ subunit of chaperonin CCT as an essential regulator of retinotectal development in zebrafish by whole-genome subtraction cloning from TMP-induced ntnmutants. Induction of ntn phenotypes by injection of cct3antisense morpholino oligonucleotide into wild-type embryos and rescue of ntn mutants by injection of wild-type cct3 mRNA clearly showed that the impaired retinotectal development in the ntn mutant fish was caused by the mutation in the chaperonin cct3 gene. The ntn^jt5 mutation appears to be null because the deletion causes frameshift of translation. Available information suggests the presence of a single gene for the CCTγ subunit as well as the α, δ,ϵ, ζ, η and θ subunits in the zebrafish genome(http://www.ensembl.org/Danio_rerio/). A cct3 zebrafish mutant was on the list of retroviral insertion mutants but no characterization was reported(Golling et al., 2002).

Molecular chaperones play an important role in folding of many proteins and CCT is a member of two major chaperone systems implied in cytoplasmic protein folding in eukaryotes (Bukau and Horwich,1998; Hartl and Hayer-Hartl,2002). In mammalian cells, ∼15-20% of newly synthesized proteins transiently bind to Hsp70 and ∼9-15% of them interact with CCT(Thulasiraman et al., 1999). Pharmacological inhibition of Hsp90, which cooperates with Hsp70 in folding of signal-transduction proteins (Young et al., 2001), was lethal and affected the development of various organs in zebrafish embryos (Lele et al.,1999). In addition, compromised Hsp90 activity in Drosophila and Arabidopsis caused a wide array of morphological variations, suggesting that Hsp90 acts as a capacitor for evolution (Rutherford and Landquist,1998; Queitsh et al., 2002). Thus, it is surprising that the impairment of CCTγ caused defects specifically in coordinate retinotectal development of zebrafish. The causal relationship between CCTγ defect and the degeneration of retinal and tectal cells implies the importance of chaperones in neurodegenerative diseases(Slavotinek and Biesecker,2001). CCT is a large cylindrical complex composed of eight different subunits providing physically defined compartments inside which a complete protein or a protein domain can fold while being sequestered from the cytosol (Kubota et al., 1995; Llorca et al., 1999). It is possible that the defect of the γ subunit in ntn mutants can be compensated by other CCT subunits to form functional CCT complex that assist folding of many proteins except for those specifically dependent on theγ subunit. The finding that the binding of actin to CCT is both subunit specific and geometry dependent (Vinh and Drubin, 1994; Llorca et al.,1999) may be consistent with this view. Major substrates of CCT are tubulin and actin in mammalian and yeast cells(Stoldt et al., 1996; Thulasiraman et al., 1999). In fact, CCT is essential for mitosis and growth in budding yeast Saccharomyces cerevisiae and conditional mutations in individual CCT subunit genes affect biogenesis of tubulin and/or actin(Chen et al., 1994; Stoldt et al., 1996). However,there was no detectable expression of the cct3 mRNA in zebrafish embryos from one-cell to 90%-epiboly stages when vigorous cell proliferation and gastrulation took place. The cct3 mRNA was strongly expressed at 12 hpf in the entire embryo and sustained thereafter, but the defects in development of ntn mutant embryos became detectable only at ∼30 hpf and specifically in the retinotectal system. Anti-acetylated tubulin immunostaining in most of neurons other than RGCs and tectal neuropil was comparable between wild-type and ntn mutant embryos. It is known that axonogenesis also involves actin biogenesis and polymerization(Chien et al., 1993). Thus, it is unlikely that the ntn mutation directly impaired the actin and/or tubulin biogenesis. Transducin α requires CCT activity for folding (Farr et al., 1993). However, unlike ntn mutants, zebrafish transducinα mutants showed morphologically normal retina(Brockerhoff et al., 2003).

One may speculate that the effect of the ntn mutation of the CCTγ gene on the retinotectal development is rather nonspecific, as CCT complex should assist folding of many proteins and zebrafish mutants affecting both retina and tectum were frequently found in large-scale screens(Abdelilah et al., 1996; Furutani-Seiki et al., 1996). However, the specificity of the ntn phenotypes is threefold. First,there were no detectable abnormalities in body patterning and neurogenesis in ntn mutant embryos at 30 hpf, despite the fact that the strong expression of the cct3 mRNA in the entire embryos started by 12 hpf and that very active developmental changes occurred in the segmentation period(10-24 hpf), including a variety of morphogentic movements, the development of somites and primary organ rudiments and rapid increase in overall body length of the embryo (Kimmel et al.,1995). Second, ntn phenotypes appeared specifically in the retina and tectum at ∼2 dpf. At later stages (4 dpf), however,underdevelopment of pectoral fins and some jaw skeletons were noted in addition to small eyes and turbid tectum. Such abnormalities may be caused secondarily or may represent nonspecific effects. Third, a specific step in RGC differentiation is impaired in ntn mutants. The cellular organization of the retinal neuroepithelium at 27 and 36 hpf suggested that the formation of eye primordium, the proliferation of retinal cell progenitors and retinal patterning proceeded normally in ntn mutant embryos. The expression patterns of transcription factors ath5 and brn3b,which are essential for the development and maintenance of RGCs(Erkman et al., 1996; Xiang, 1998; Brown et al., 2001; Kay et al., 2001; Matter-Sadzinski et al.,2001), were indistinguishable between wild-type and ntnmutant embryos, but those of early and late differentiation markers of RGCs,nAChRβ3 and zn5, were diminished in mutant embryos. Immunostaining of acetylated tubulin also revealed the impairment of RGC axon extension and optic nerve formation. Thus, ntn mutation of the cct3 gene exerted little effect on the commitment of retinal neuroepithelial cells to postmitotic retinal neurons but severely impaired the differentiation of retinal neuroepithelial cells to RGCs. Similarly, the expression of brn3b was normal in the tectum of ntn mutants, but tectal neuropil formation was abolished. These results suggest that the γsubunit of chaperonin CCT plays an essential role in retinotectal development.

We are grateful to T. Morita and T. Sato for helpful advice on RDA procedures; I. Masai and M. Take-uchi for advice on the analysis of retinal development; T. Kaneko and S. Fukuoka for technical assistance; and K. Kinomoto for help in zebrafish breeding. We thank A. Chitnis for dlx2cDNA; T. Jowett for krox20 cDNA; S. Krauss for shh cDNA; I. Masai for ath5 cDNA; S. Shulte-Merker for ntl cDNA; H. Takeda for hlx1, myod and pax2a cDNAs; and E. Weinberg for zash1a and zash1b cDNAs. We also thank H. Mori and T. Yoshida for critical reading of the manuscript. We are grateful to M. Ohtsuka for his encouragement and support. This work was supported in part by research grants from the Japan Science and Technology Agency and the Ministry of Education, Culture, Sports, Science and Technology of Japan.