αA-crystallin expression prevents γ-crystallin insolubility and cataract formation in the zebrafish cloche mutant lens

Cataracts, the loss of eye lens transparency, are the leading cause of blindness in humans (Resnikoff et al.,2004), with about 17 million cases worldwide per year(Foster and Resnikoff, 2005). Understanding the molecular mechanisms responsible for cataract formation could be useful for designing treatments other than surgical intervention,which is presently the only available recourse. Genetic studies in human and rodents have implicated crystallin defects as major contributory causes of cataract formation (Graw,2004). Crystallins are a large family of proteins that constitute 90% of the soluble protein in lens fiber cells(Bloemendal et al., 2004; Graw and Loster, 2003). There are three classes of these proteins, including α-, β- andγ-crystallins. Each crystallin is expressed in the lens with spatial and temporal expression patterns characteristic for each class. Mutations in humanγ C-crystallin (CRYGC) or γD-crystallin(CRYGD) are associated with protein misfolding and protein aggregation within the lens (Graw,2004). Whereas β- and γ-crystallins are structural proteins, αA-crystallin is a member of the small heat shock protein family and is also known as HSPB4 (Franck et al., 2004). αA-crystallin acts as a molecular chaperone that protects proteins from misfolding(Horwitz, 2003). Defects in this gene are associated with autosomal dominant congenital cataracts(Litt et al., 1998).

Zebrafish embryos are excellent models for studying the genes that regulate embryonic development (Talbot and Hopkins,2000; Thisse and Zon,2002). Zebrafish embryos are small, can be maintained on a large scale and develop rapidly within days. Unlike mice, the embryos are extracorporeal and transparent, allowing ready visualization of organs during development. The transparency and relatively large eyes make zebrafish embryos a particularly excellent model for studying eye development and ocular diseases (Glass and Dahm,2004; Malicki,2000). Functional vision is developed by 3 days post-fertilization(dpf) (Easter and Nicola,1996). Chemical mutagenesis screening has identified mutants with an opaque lens, e.g. lop(Vihtelic et al., 2005b). However, the cause of the opacity has not been established.

Forward genetics have been used to identify zebrafish genes involved in normal development and genetic disorders, and to elucidate their function. For example the cloche mutant lacks blood cells and blood vessels(Qian et al., 2005; Stainier et al., 1995; Sumanas et al., 2005; Weber et al., 2005). However,a limitation of a strictly genetic approach is the inability to study protein function. Proteomics is an alternative approach that identifies proteins,measures their abundance, detects post-translational modifications, and provides sequence data for cloning purposes. To date there have been very few attempts to analyze zebrafish protein profiles by proteomics. We chose to study protein defects in cloche as it is not clear what the genetic defect is. Surprisingly, we noted that cloche not only had defects in hematopoiesis and vasculogenesis, but also had lens opacity, a characteristic of cataracts.

We report here that γ-crystallin protein is insoluble in cloche lens, resulting in lens opacity. Furthermore, the cloche lens is deficient in αA-crystallin, a chaperone protein that keeps proteins soluble. Overexpression of exogenousα A-crystallin (cryaa - Zebrafish Information Network)rescues to a great degree the cloche lens phenotype; for example,γ-crystallin protein becomes more soluble, the lens becomes more transparent in vivo and lens fiber cell differentiation is induced. It was concluded that αA-crystallin expression is required for normal lens development and that these results are significant as they show, possibly for the first time, that cataract formation can be prevented in vivo.

Anti-newt γ-crystallin guinea pig antiserum and anti-mouseαA-crystallin rabbit antiserum were kindly obtained from Dr M. Okamoto(Nagoya, Japan) and from Dr J. J. Liang (Harvard Medical School, Boston, MA). Horseradish peroxidase (HRP)-conjugated anti-guinea pig antibody was purchased from Jackson ImmunoResearch Laboratories. HRP-conjugated anti-mouse antibody was purchased from Amersham Biosciences UK.

All animal protocols were approved by the Children's Hospital Institutional Animal Care and Use Committee. Zebrafish were maintained as described(Westerfield, 2000), and staged as described (Kimmel et al.,1995). Two independent cloche alleles, cloche^m39 and cloche^s5, were used in this study. Embryo age is defined as the number of days post-fertilization(dpf). Euthanasia of zebrafish was accomplished by treatment with tricaine(Sigma-Aldrich) before protein or RNA extraction. The homozygotes were separated based on visual criteria such as cardiac edema, enlarged hearts and lack of blood cell circulation.

Living embryos in embryo media including tricaine were analyzed by stereomicroscopy (Olympus). For confocal microscopic analysis, embryos were embedded in 3% methyl cellulose in embryo media including tricaine. Stacks of confocal micrographs were recorded from anesthetized zebrafish lens using a Leica TCS SP2 (Leica Microsystems) confocal laser-scanning microscope fitted to a DM IRE 2 inverted microscope with a 488 nm argon ion laser. Serial sections were obtained at 1 μm intervals under the reflected light mode. Images recorded under transmitted light were used as a general locator for the quantitative analysis of selected reflected light images. Reflected light was analyzed by the digital image analysis program (Leica Confocal Software and ImageJ).

Approximately 100 zebrafish embryos were transferred into 1.5 ml microcentrifuge tubes. Centrifugation at 15,000 g for 5 minutes resulted in two phases. The lower phase contained the yolk, whereas the upper phase contained zebrafish without yolk. The bottom of the tube was pierced with a 26 G needle to remove the bottom yolk layer. To obtain total embryo extracts containing soluble protein and cell membrane-associated proteins, embryos were washed with ice-cold PBS three times and lysed in 8 M urea, 4% CHAPS and 60 mM dithiothreitol by sonication for 20 seconds on ice,and the lysates were left on ice for an additional 30 minutes. To avoid protein degradation, proteinase inhibitor (Roche) and phosphatase inhibitor 1 and 2 (Sigma) were added to embryo lysates. Precipitates were removed by centrifugation at 15,000 g for 20 minutes at 4°C. To prepare soluble and insoluble protein, zebrafish embryos (2.5 dpf) were homogenized in detergent-free 20 mM Tris-HCl (pH 7.5). Soluble and insoluble fractions were separated by centrifugation.

Total embryo extracts were analyzed by two-dimensional gel electrophoresis. A BioRad protein IEF cell was used for the first dimension. Embryo extracts(80 μg) were applied to immobilized pH gradient gel strips, 11 cm, linear(pH 3-10). IEF was carried out at 20°C for 2 hours at 250 V (linear ramp),and for 5.5 hours at 8000 V (linear ramp). The second dimension was a 10-20%gradient SDS-PAGE for 1.5 hours at 150 V. Proteins were visualized by SYPRO Ruby fluorescent staining (BioRad). The gels were photographed and the intensities of gel spots were analyzed by a digital image analysis program(ImageJ). The intensities of spots were normalized relative to wild-type samples being 100%. Gel spots were excised using a spot cutter (The Proteome Works, BioRad). Obtaining sufficient amounts of protein for sequencing (20 pmol) required excision from two or three spots. Samples were analyzed by electrospray tandem mass spectrometry. These spectra were correlated with known protein sequences using the Sequest algorithm and other programs developed in the Harvard Microchemistry Lab(Chittum et al., 1998; Eng et al., 1994).

Total embryo extracts, or extracts fractionated into soluble and insoluble protein, were analyzed by western blot. Two μg of embryo extracts were analyzed by 12% SDS-PAGE. The membranes were incubated with anti-mouseαA-crystallin antisera (1/2000 dilution) or anti-newt γ-crystallin antisera (1:2000 dilution) followed by HRP-conjugated anti-guinea pig IgG(Jackson ImmunoResearch Laboratories, 1:5000) or anti-rabbit IgG (Amersham Pharmacia Biotech; 1: 10,000), and visualization with enhanced chemiluminescence (Perkin Elmer Life Science). The specificity of each antibody was confirmed by western blotting of lysates prepared by transiently transfecting Chinese Hamster Ovarian cells (American Type Culture Collection)with expression constructs of zebrafishα A-(Runkle et al.,2002) (GenBank Accession Number, NM 152950),α B1-(Posner et al.,1999; Smith et al.,2006) (GenBank Accession Number, NM 131157),β B1-(Chen et al.,2001) (GenBank Accession Number, NM 173231) andγ -crystallin genes. Cell maintenance and gene transfection procedures were as described previously(Goishi et al., 2003). Protein bands were analyzed by the digital image analysis program (ImageJ). Results were normalized by setting the densitometer of the wild-type samples as 100%.

Sixteen peptide sequences from tandem mass spectrometry were analyzed in a BLAST search. Twelve of the peptide sequences were completely identical to several EST clones. Oligonucleotide primers for PCR were generated based on matched EST clone sequences. Using the SMARTRACE cDNA Amplification Kit (BD bioscience), 5′ and 3′ zebrafish γ-crystallin cDNA ends were obtained by RT-PCR from 48 hpf embryo total RNA.

Sense primers for 3′RACE are follows: CryG-S1,5′-GGCGACTGTGGTGACTTCTCCTCCTACA-3′; CryG-S2,5′-CATGAAGGTCACCTTCTTTGAGGA-3′; CryG-S3,5′-GCAAACCATGATGGGCAAAGGTC-3′; CryG-S4,5′-AGTAACCATGATGGGCAAGGTCAT-3′; CryG-S5,5′-CGCAACCATGATGGGCAAGGTCGTT-3′, CryG-S6,5′-CGCAAACATGATGGGCAAGGTCGCC-3′; CryG-orf1-s,5′-AACATGAAGGTCACCTTCTTTGAGGACA-3′.

Antisense primers for 5′RACE are follows: CryG-As1,5′-CCGAAGTGCCACATCCTGCCTCTGTAGT-3′; CryG-As2,5′-GTACCAAGAGTCCATGATACGCCTCA-3′; CryG-orf1-AS,5′-CTAGTACCAAGAGTCCATGATACGCCTCA-3′.

PCR products for γ-crystallin genes were subcloned into the pCRII/TOPO vector (Invitrogen). γ-crystallin open reading frames were generated by RT-PCR. PCR products were subcloned into the pCRII/TOPO vector and the expression vector pcDNA3.1/V5-His/TOPO (Invitrogen). An EST database BLAST search with mammalian αB-crystallin protein sequences revealed a second zebrafish αB-crystallin gene that was cloned by PCR. The second αB-crystallin gene was amplified using the following primers: CRY-ABb-S,5′-TTGCAGAAGAGGCCCAGACTCA-3′; CRY-ABb-As-stop,5′-CACAAACATTTGGCCGTCAGTAGG-3′.

Whole-mount in situ hybridization was performed as previously described(Bovenkamp et al., 2004; Lee et al., 2002). Briefly,crystallin cDNAs were used as a template to synthesize digoxigenin-labeled sense and antisense riboprobes. Embryos were fixed in 4% paraformaldehyde overnight and then permeabilized by proteinase K treatment prior to hybridization. After color was detected with alkaline phosphatase substrates,the embryos were mounted in glycerol and photographed. Paraformaldehyde fixed,paraffin-embedded embryos were used for histology (Hematoxylin and Eosin).

Two μg of total RNA from 2 dpf, 3 dpf and 4 dpf zebrafish were isolated and reverse transcribed by using SuperScript II reverse transcriptase(Invitrogen) according to the manufacturer's suggestions. Zebrafish elongation factor 1α (ef1a; GenBank Accession Number,NM 131263) was used as a loading control.

Each sample was amplified with the following primers.α A-crystallin: CRYAA-S,5′-TATGGATATTGCGATCCAACACCCTT-3′; CRYAA-AS,5′-AGAGCCTGAGTTGCTCTTGTCCTCG-3′. αB1-crystallin:CRYAB-S, 5′-TGTCAGCTCCAAACTAAGGTACAATG-3′; CRYAB-AS,5′-TGCAGCTCTGGACAGGTTCAT-3′. βB1-crystallin:CRY-BB1-S, 5′-CACCATGTCTCAGACCGCCAAA-3′; CRY-BB1-As,5′-CTGTTTAGCGGCGGTCATGTTGA-3′. αBb-crystallin(GenBank Accession Number AY939876): CRY-ABb-S; CRY-ABb-As,5′-CTTCTGAGGGCCAGCGACGGT-3′. γ-crystallin: Cry-G-S1 and Cry-G-As1. flk1 (GenBank Accession Number, NM 131472):Flk1-1098S, 5′-TGCCATCGAACCAGAAAGACCAAGAG-3′; Flk1-1299As,5′-TGCCCTTGTCCTCCATTGTCACATTA-3′. ef1a: Ef1-S,5′-GCCCCTGCCAATGTA-3′; Ef1-As,5′-GGGCTTGCCAGGG-AC-3′.

Zebrafish αA- and γ-crystallin cDNA were subcloned into pXT7, which was obtained from Dr S. Sokol (Beth Israel,Boston). To generate capped mRNA, plasmids were linearized by SalI and transcribed with T7 RNA polymerase using the mMESSAGE mMACHINE (Ambion)according to the manufacturer's instructions. For microinjection of RNA into zebrafish, 100 pg of synthesized mRNA were injected into the one-to four-cell stage embryos obtained from crosses between cloche heterozygotes. Theα A-crystallin promoter-EGFP(Kurita et al., 2003) was injected (100 pg) into one-to four-cell stage embryos. The injected embryos were examined under a fluorescence microscope (Olympus).

The Mann-Whitney U test was used to test for differences between two groups. Accepted level of significance for all tests was P<0.05. Box-and- whisker plots were generated to show the levels of reflectance intensity. The line in the box corresponds to the median value. The lower and upper whiskers correspond to 10th and 90th percentiles,respectively.

Embryos were fixed in 4% paraformaldehyde/PBS (pH 7.4) for 2 hours at room temperature, incubated in 30% sucrose/PBS overnight, and embedded in Neg50 frozen section medium (Richard-Allen Scientific). Sections of 10 μm section were prepared, rehydrated and blocked in 0.1% Tween20, 10% goat serum and 0.5%TritonX-100 in PBS. Immunohistochemistry was carried out with an antibody zpr-1 that detects zebrafish photoreceptors, a gift from Dr J. Malicki(Massachusetts Eye and Ear Infirmary, Boston)(Pujic and Malicki, 2001). Analysis was carried out by confocal microscopy.

cloche mutant and wild-type zebrafish embryos extracts were compared by proteomics to identify possible altered protein profiles that might reveal clues about the mechanisms contributing to the clochephenotype. Two dimensional (2D) gel analysis of total extracts, as described in Material and methods, showed consistently that one spot (20 kDa, pI of 8.8)was strongly reduced by about 75% in cloche compared with wild-type embryos (Fig. 1A). Tandem mass spectrometry (MS/MS) and subsequent gene cloning identified this altered zebrafish embryonic protein to be γ-crystallin(Fig. 1B). Eight differentγ-crystallin cDNAs were isolated from 2 dpf embryos and the encoded proteins were 95-100% identical at the amino acid level (GenBank Accession Numbers: DQ294939, DQ294940, DQ288267, DQ294941, DQ294942, DQ294943, DQ298448,DQ298449). One of these cDNAs was designated as embryonic zebrafishγ -crystallin M2 type1 (γEM2-1-crystallin,GenBank Accession Number, DQ294939) (Fig. 1B, top, EM2-1). This zebrafish cDNA encodes a protein of 174 putative amino acids with two β/γ-crystallin domains, the same as in other species. It is 78% homologous in amino acid sequence to a very recently described zebrafish γ-crystallin M2 gene(crygm2b; GenBank Accession Number, NM001020783) that had been cloned from adult lens tissue (Vihtelic et al.,2005a; Wistow et al.,2005) (Fig. 1B,bottom, M2b). However, our crystallin gene appears to be embryonic and was found only in embryonic cDNA databases. We also isolated a secondα B-crystallin gene, which we calledα Bb-crystallin (cryabb; GenBank Accession Number,AY939876). This gene is identical to αB2-cystallin (GenBank Accession Number; DQ113417) (Smith et al.,2006) and similar to a previously reportedα B1-crystallin (GenBank Accession Number, NM 131157)(Posner et al., 1999).

Crystallins are expressed mostly in lens and constitute 80-90% of lens protein (Bloemendal et al.,2004; Graw and Loster,2003). The spatiotemporal expression pattern ofγ -crystallin was analyzed by whole-mount in situ hybridization using our cloned zebrafish γ-crystallin EM2-1 cDNA. At 2 dpf,γ-crystallin was expressed entirely in the lens(Fig. 1C). However, unlike the protein, there was no significant difference between cloche and wild-type embryo γ-crystallin gene expression.

There are several classes of crystallins, including α-,β - and γ-crystallin, and each class has several different genes (Chen et al.,2001; Posner et al.,1999; Runkle et al.,2002; Vihtelic et al.,2005a; Wistow et al.,2005; Smith et al.,2006). The expression levels of αA-,α Bb-, βB1- and γ-crystallin were analyzed by semi-quantitative RT-PCR (Fig. 2A). This analysis showed that αA-crystallin gene expression was strongly downregulated in cloche, whereas gene expression of the other crystallins was not altered. The clochemutant is defective in hematopoiesis and blood vessel development(Qian et al., 2005; Stainier et al., 1995; Sumanas et al., 2005; Weber et al., 2005). As a control to confirm the cloche phenotype, the RNA levels of flk1, a VEGF receptor expressed by endothelial cells, were diminished in cloche. As a loading control, the levels of ef1a, which are known to remain constant from 8 hours to 72 hours of zebrafish development(Qian et al., 2005; Xu et al., 2003), did not change.

To show that αA-crystallin mRNA levels decrease in vivo as well, an αA-crystallin promoter-EGFP construct was injected into cloche embryos (n=470; homozygous=120, wild-type siblings=350) (Fig. 2B). The specificity and reliability of this had been described previously(Kurita et al., 2003). At 2.5 dpf, the wild-type siblings showed strong and uniform fluorescent intensity in the lens and nowhere else (Fig. 2B, top). However, in cloche there was only faint fluorescence intensity in the lens and no expression in the lens core at 2.5 dpf (Fig. 2B, bottom) and even at 5 dpf (not shown).

Western blotting of total extract with specific anti-mouseαA-crystallin antibody showed an 85% reduction in the levels ofαA-crystallin protein (Fig. 2C) consistent with the RNA results. A time course revealed that the level of αA-crystallin mRNA in cloche were diminished and these low mRNA levels remained constant from 2-5 dpf(Fig. 2D). However, the levels of γ-crystallin were the same developmentally in both wild type and cloche (Fig. 2D).

Crystallin defects are associated with cataracts, which are characterized by light scattering opacity of the lens(Bloemendal et al., 2004; Graw, 2004). Zebrafish embryos are transparent and thus the gross morphology of the eye and lens can be examined readily by bright-field microscopy of living zebrafish embryos. At 2.5 dpf, lenses of wild-type embryo siblings are completely transparent (0 opaque/300 embryos, 0% opacity) (Fig. 3A, top), whereas the cloche embryo lenses are cloudy with light scattering in 428 of 633 embryos (68% opacity)(Fig. 3A, bottom). By 3 dpf,the cloudy lens phenotype was more apparent, ranging from 84% (533/633)opacity in one analysis to 100% opacity (34/34) in another. In another approach, lenses were analyzed by stereomicroscopy with highly oblique(80° off-axis) illumination. There was no light scattering in wild-type lens, whereas strong reflectance was observed in the cloche lens (not shown).

During normal lens development, the nuclei of lens fiber cells that express crystallins are readily detected by histology at 2 dpf, diminished by 3 dpf and absent by 4 dpf, at which point the lens fiber cells had become terminally differentiated and the lens was transparent(Fig. 3B, top panels). By contrast, in the cloche embryo, lens fiber cell nuclei were still detectable at 4 dpf (Fig. 3B,bottom panels) and even at 5 dpf (not shown). The presence of nuclei in the cloche lens was confirmed by staining with Hoechst 33342 which visualizes nuclei (not shown). A quantitative analysis shows an 85-90% loss of wild-type nuclei by 3 dpf but persistent appearance of nuclei in cloche at 2-4 dpf (Fig. 3C, n=14). The persistence of nuclei in the cloche mutant suggests that this phenotype was not due to delayed development.

Lenses were also analyzed by confocal microscopy to measure lens transparency in a non-invasive and more quantitative manner. At 2-4 dpf,wild-type sibling lens were totally transparent(Fig. 3D, top panels), whereas the cloche lenses displayed intensive light scattering in the lens core (Fig. 3D, bottom panels). The intensity of reflectance in the live cloche embryo lens increased with time. The median value of intensity was 37-fold higher in clochelens compared with the wild-type lens at 4 dpf (P=0.0209)(Fig. 3E). Another allele of a zebrafish cloche mutant, cloche^s5(Brown et al., 2000; Qian et al., 2005; Weber et al., 2005), was also analyzed for the presence of cataracts. As in cloche^m39,the lens was opaque (not shown). However, confocal microscope analysis showed a quantitative high degree of reflectance in cloche^s5comparable with cloche^m39(Fig. 3F,G).

cloche is a mutant defective in blood vessel formation. However,lack of blood vessels does not appear to contribute to the cataract phenotype. When wild-type embryos were treated with VEGFR2 kinase inhibitor, blood vessels were not formed, but neither were the cataracts (not shown).

Besides the lenses other parts of the zebrafish eye were examined, in particular the retina, by Dr L. Smith (Children's Hospital, Boston, MA, USA)and Dr J. Malicki (Massachusetts Eye and Ear Infirmary, Boston, MA, USA) who specialize in retinal disease (Smith,2003; Malicki,2000). Immunohistochemistry with a monoclonal antibody zpr-1 that detects photoreceptors (Larison and Bremiller, 1990; Pujic and Malicki, 2001) demonstrated that the photoreceptors (green)develop normally in cloche even at 5 dpf. The thickness of the retinal layers at 3.5 dpf and 5 dpf were similar to wild-type retina (see Fig. S1 in the supplementary material).

αA-crystallin is a small heat shock protein that acts as a chaperone,inhibits aggregation of several target proteins and maintains the solubility of crystallin proteins in vitro (Dahlman et al., 2005; Horwitz,2003). Accordingly, zebrafish αA-crystallin mRNA was injected into zebrafish embryos at the one- to four-cell stage and the effects on opacity were analyzed by bright-field microscopy, confocal microscopy and histology (Fig. 4). Wild-type embryo lenses were clear (0% opaque, n=300)(Fig. 4A, lane 1). Uninjected cloche embryos had a cloudy and opaque lens (84% opaque, n=633), as measured by bright-field microscopy(Fig. 4A, panel 2). However, cloche embryos injected with αA-crystallin mRNA showed a much lower frequency of opacity (19.8% opaque, n=96) at 4 dpf(Fig. 4A, panel 3). Conversely,overexpression of γ-crystallin, which has no chaperone activity, in cloche embryos had no protective effect on lens opacity(100% opaque, n=35) (Fig. 4A, panel 4).

The efficacy of rescue was also measured by confocal microscopy(Fig. 4B). There was no reflectance in wild-type embryos (Fig. 4B, panel 1) and a great degree of reflectance in clochelens (Fig. 4B, panel 2). However, overexpression of αA-crystallin resulted in a significant decrease in reflectance. The median lens reflectance(n=4) of cloche lens was 70 units but whenα A-crystallin was overexpressed the median reflectance decreased to 41 units (n=4, P=0.0433, Fig. 4B, panel 3). However, as a control for possible non-specific effects resulting from mRNA injections,overexpression of γ-crystallin had no protective effect(Fig. 4B, panel 4). The median reflectance value was 85 units, more than cloche but not statistically significant (n=4). The differences between overexpression of αA-crystallin compared with overexpression of γ-crystallin were substantial and significant(P=0.0209).

Histology indicated that overexpression of αA-crystallin (4 dpf) also rescued the lens fiber nuclei phenotype(Fig. 4C). Unlike cloche at 4 dpf (Fig. 4C, panel 2) nuclei were not detected when αA-crystallin was overexpressed (Fig. 4C, panel 3) and the lens appeared more like the wild-type lens at 4 dpf. (Fig. 4C, panel 1). However, overexpression of γ-crystallin in clocheembryos had no protective effect on lens nuclei(Fig. 4C, panel 4).

The increased transparency as a result of overexpression ofα A-crystallin could be due to its preventing γ-crystallin from becoming insoluble. Accordingly, the solubility of γ-crystallin protein was analyzed in cloche with and withoutα A-crystallin overexpression(Fig. 5). Protein was extracted from 2.5 dpf zebrafish embryos in detergent-free 20 mM Tris-HCl (pH 7.5),centrifuged to separate soluble and insoluble fractions, and analyzed by western blot. Of the total protein, 83% of wild-type embryo γ-crystallin was soluble and 17% was insoluble, whereas 20% of clocheγ-crystallin was soluble and 80% was insoluble(Fig. 5A). However, whenα A-crystallin was overexpressed in cloche, there was a significant increase in soluble γ-crystallin levels, with over 60% of the protein being soluble (Fig. 5B). These results show that αA-crystallin rescues the insolubility phenotype and indicates that the lower levels of clocheγ-crystallin shown by 2D gel electrophoresis(Fig. 1A) were due to protein insolubility.

Similar results were found in the cloche^s5 allele (see Fig. S2 in the supplementary material). αA-crystallin was not expressed at the protein and mRNA levels, γ-crystallin was insoluble, and the cataract phenotype was rescued by overexpression ofα A-crystallin mRNA.

The zebrafish cloche mutant, noted for defects in hematopoiesis and vascular development, has surprisingly an additional phenotypic characteristic, cataract formation. The cataracts are characterized by insoluble, aggregated protein, a high degree of reflectance, and a lack of lens fiber cell differentiation. We propose an in vivo mechanism for cataract formation in which αA-crystallin, a chaperone for γ-crystallin, is not expressed, the result being insolubility of γ-crystallin, and lack of terminally differentiated mature lens fiber cells. The net results of these defects are lens opacity and loss of transparency.

The evidence for this mechanism combines protein and zebrafish analysis. cloche γ-crystallin is expressed at the same mRNA levels as wild-type γ-crystallin and is localized to the lens in a normal manner as shown by in situ hybridization. However, the cloche γ-crystallin protein is mostly insoluble as shown by 2D-gel electrophoresis and western blot. Thus, there appears to be aγ-crystallin protein defect in cloche. However,α A-crystallin mRNA expression, promoter reporter activity in the lens and protein levels are all greatly diminished. αA-crystallin is not only a structural lens protein but is also a functional chaperone that prevents protein denaturation in vitro(Dahlman et al., 2005; Horwitz, 2003). Evidence thatαA-crystallin acts as a chaperone is that the overexpression ofα A-crystallin in cloche mutants prevented cataract formation in 80.2% of the embryos as measured by bright-field microscopy. Quantitative analysis in the living embryo lens by confocal microscopy showed that reflectance was reduced by 41%. Biochemical analysis revealed that the insolubility of γ-crystallin was reduced by 60%. In addition,overexpression of αA-crystallin induced lens fiber cell denucleation as shown by histological staining, contributing to an increase in lens transparency. As a control for crystallin specificities,γ -crystallin overexpression had no such protective effects. These results in vivo are consistent with previous reports that zebrafishαA-crystallin functions as a chaperone for target substrates in vitro(Dahlman et al., 2005).

Our results are consistent with a number of reports in mammals that crystallin defects are associated with cataracts. For example, in mice,targeted deletion of αA-crystallin leads to cataracts(Brady et al., 1997). Missense mutations in αA-crystallin have been found in mice with cataracts (Chang et al., 1999; Graw et al., 2001). Mouseα A/αB crystallin double knockouts show that bothα-crystallin genes are necessary for proper fiber cell formation(Boyle et al., 2003). At least 20 missense, nonsense, deletion and insertion mutations in mouseγ -crystallin have been identified(Graw, 2004). In humans,missense mutations in αA-crystallin have been found in human autosomal dominant congenital cataracts(Litt et al., 1998). A nonsense mutation in αA-crystallin is associated with autosomal recessive congenital cataracts (Pras et al., 2000). There are also human γ-crystallindefects associated with cataracts (Graw,2004). Missense, nonsense and insertion mutations associated with cataracts have been found in γC-crystallin andγ D-crystallin (Graw,2004). These studies have provided good evidence linking crystallin defects to mammalian cataracts. However, a mechanism for cataract formation and prevention of cataracts in the mammalian lens in vivo has been lacking so far.

There is the possibility that the cataract phenotype is a non-specific consequence of developmental delay. However, this is unlikely for several reasons. One is that αA-crystallin expression is very diminished compared with wild type and these low mRNA and promoter levels remain constant from 2-5 dpf. The relatively high levels ofγ -crystallin do not change in this time period. These results suggest that γ-crystallin will be irreversibly insoluble throughout development. Furthermore, during normal lens development, the nuclei of lens fiber cells that express crystallins were readily detected by histology at 2 dpf, greatly diminished by 3 dpf and absent by 4 dpf, at which point the lens fiber cells had become terminally differentiated and the lens was transparent. However, in the cloche embryo, lens fiber cell nuclei were still detectable at 5 dpf. The persistent presence of nuclei and concomitant lack of transparency in the developing cloche mutant is inconsistent with delayed development. In addition, we find the same lens cataract phenotype and rescue by αA-crystallin mRNA in two different alleles of zebrafish cloche, cloche^m39 and cloche^s5 (see Fig. S2 in the supplementary material), that have different genetic alterations, making it unlikely that cataract formation in cloche is non-specific. A degree of specificity for the lens phenotype is indicated in that the photoreceptors in cloche retina appear structurally normal in development at 3.5-5 dpf (see Fig. S1 in the supplementary material). Finally, we examined a zebrafish mutant, crash&burn, that is developmentally delayed because of loss of bmyb function, resulting in decreased cyclin B1 expression and mitotic arrest (Shepard et al.,2005; Stern et al.,2005). In these mutants, there are no cataracts and the lenses are totally transparent (Fig. S3 in the supplementary material). Thus,developmental delay is insufficient to induce cataracts. Taken together, these results suggest that the cloche lens phenotype represents a specific progressive pathological disorder rather than a delay in physiological development.

Zebrafish cloche was originally described as a mutant that had defects in hematopoiesis and blood vessel development(Stainier et al., 1995). Whether there is a link between the cataract phenotype and this blood/blood vessel phenotype is not clear, the lens being avascular. VEGF is a major regulator of angiogenesis. VEGF acting through VEGF receptor tyrosine kinases is necessary for axial vessel and intersegmental vessel development(Goishi and Klagsbrun, 2004). These interactions can be disrupted by administration of VEGFR2 kinase inhibitors or morpholino antisense oligos (MO) to zebrafish embryos(Chan et al., 2002; Lee et al., 2002; Nasevicius et al., 2000). However, neither a VEGF receptor kinase inhibitor nor an anti-VEGF MO induced a cataract phenotype in wild-type zebrafish embryos, even at 5 dpf. Another plausible candidate could be FGF2, one of the first angiogenesis factors to be identified (Shing et al.,1984). Endothelial cells in culture respond to FGF2 by migrating and proliferating mainly via FGF receptor 1 (FGFR1). In addition to a role in angiogenesis, FGFR1 is expressed by hematopoietic/endothelial precursor cells(Magnusson et al., 2005) and FGF signaling regulates hematopoietic development(Faloon et al., 2000). FGF also initiates lens fiber cell differentiation in mice(Lovicu and McAvoy, 2005). Whether FGF defects are linked to both loss of blood/blood vessels and cataracts remains unclear. Our preliminary data showed that FGFR kinase inhibitor did not induce a cataract phenotype in wild-type zebrafish lens.

There are several cloche alleles, one being cloche^m39, which is a spontaneous mutant, and another being cloche^s5, which is chemically induced mutant(Brown et al., 2000; Qian et al., 2005; Weber et al., 2005). The original studies carried out here have used cloche^m39embryos. However, cloche^s5 also shows diminishedα A-crystallin levels, insoluble γ-crystallin protein, and evidence of cataracts. In both alleles, overexpression ofα A-crystallin in cloche rescued the cataract phenotype.

Our results are significant as they may constitute the first demonstration of blocking cataract formation in vivo. There may be some practical consequences of using zebrafish as a cataract model. Zebrafish are amenable to large-scale, systematic screens to identify small molecules that can suppress disease phenotypes (Peterson and Fishman,2004; Zon and Peterson,2005). For example, the zebrafish mutation gridlockdisrupts aortic blood flow in a region and manner akin to aortic coarctation in the human. A zebrafish small molecule screen has been used to discover a class of compounds that suppressed the coarctation phenotype(Peterson et al., 2004). The zebrafish lens is very visible, transparent and relatively large. It is possible to administer drugs to zebrafish in their swim medium(Peterson and Fishman, 2004; Zon and Peterson, 2005); thus,zebrafish cloche may be useful for screening anti-cataract drugs.

Finally, our results demonstrate the value of proteomics in detecting protein alterations in zebrafish mutants. Most studies in zebrafish are based on genetic analysis. However, gene expression and mRNA abundance do not necessarily correlate with protein abundance. Another major limitation of mRNA-based approaches is the inability to detect post-translational modifications that may determine protein function. In our study,γ -crystallin mRNA was normal but the protein was not soluble. We propose that zebrafish proteomics will provide a new dimension in zebrafish research.

The authors thank Drs D. R. Bielenberg, A. J. Davidson, M. R. Freeman and B. H. Paw for discussion of this work; Drs J. Malicki (Massachusetts Eye and Ear Infirmary, Boston, MA) and L. E. Smith (Children's Hospital, Boston, MA)for analysis of cloche retina photoreceptors; Drs K. A. Dooley and G. J. Weber for advice on cloche maintenance; A. Lord for technical assistance for crash&burn maintenance; Drs M. Okamoto (Nagoya,Japan) and J. J. Liang (Harvard Medical School, Boston, MA) for providing antibodies; D. Zurakowski for advice on statistical analysis; and M. Mang, E. Pravda and R. Sanchez for technical assistance. This work was supported by the National Institutes of Health Grants CA37392 and CA45448 (to M.K.) and DK55381-06 (to L.I.Z.). L.I.Z. is supported by the Howard Hughes Medical Institute. The confocal microscopy studies were supported by CA45448(R.R.).