The ubiquitin proteasome system is required for cell proliferation of the lens epithelium and for differentiation of lens fiber cells in zebrafish

During development, the lens placode delaminates from the epidermal ectoderm and forms the lens vesicle. In the lens vesicle, posterior epithelial cells differentiate into lens fiber cells, whereas anterior epithelial cells are retained as proliferative lens progenitor cells (Bassnett et al., 1999; Lovicu and McAvoy, 2005). In zebrafish, a lens mass without a vesicle delaminates, and the cells reorganize into the central core and epithelial cells (Greiling and Clark, 2009). At the interface between the anterior lens epithelium and posterior lens fiber core, called the equator, epithelial cells differentiate into lens fiber cells, which elongate towards both anterior and posterior poles of the lens sphere and cover the old lens fiber core. In addition, lens fiber cells lose subcellular organelles, including the nucleus (Bassnett and Mataic, 1997), Golgi apparatus, endoplasmic reticulum (Bassnett, 1995) and mitochondria (Bassnett and Beebe, 1992). Previous studies have revealed regulatory factors involved in lens fiber cell differentiation. The transcription factor Pax6 is essential for lens placode induction (Ashery-Padan et al., 2000). Pax6 cooperates with several transcription factors, including Prox1 (Wigle et al., 1999), Pitx3 (Semina et al., 1998) and Foxe3 (Medina-Martinez and Jamrich, 2007). Foxe3 functions as a central factor in this regulatory network (Chow and Lang, 2001; Medina-Martinez and Jamrich, 2007) and regulates the expression of terminal differentiation genes, including that which encodes DNase II-like acid DNase (DLAD; Dnase2b). However, it is unclear how lens morphogenesis and organelle loss are related to these transcription factors.

The ubiquitin proteasome system (UPS) is important for the removal of unnecessary proteins and regulates normal cellular functions through the degradation of a variety of regulatory proteins (Nakayama and Nakayama, 2006; O'Connell and Harper, 2007). A cascade of enzymes, including ubiquitin-activating (E1), ubiquitin-conjugating (E2) and ubiquitin ligase (E3) enzymes, conjugate polyubiquitin to lysine residues in target proteins. Polyubiquitylated proteins are degraded by the 26S proteasome (Finley, 2009). The 26S proteasome is a ~2500 kDa complex consisting of one 20S core particle (CP) and two 19S regulatory particles (RPs) (Murata et al., 2009). Previous studies using in vitro cultures showed that the UPS is required for the proliferation of lens epithelial cells (Guo et al., 2006), mitochondrial degradation during lens fiber differentiation (Zandy and Bassnett, 2007), TGFβ-mediated lens fiber cell differentiation (Wu et al., 2007) and posterior capsular opacification (Hosler et al., 2006). It was reported that the UPS cooperates with a chaperone system to control protein quality (Bian et al., 2008; Bornheim et al., 2008). However, there have been a limited number of in vivo studies on the role of the UPS in lens fiber differentiation.

In zebrafish, lens development is similar to that of other vertebrates except that there is no formation of a lens vesicle. After 24 hours post-fertilization (hpf), differentiation of primary lens fiber cells begins, followed by formation of the anterior epithelium and the differentiation of secondary fibers from elongating epithelial cells near the equator. At 72 hpf, the lens grows to form a multilayered fiber core (Dahm et al., 2007; Greiling and Clark, 2009; Soules and Link, 2005). Genetic approaches have identified zebrafish mutants with defects in lens development (Fadool et al., 1997; Gross et al., 2005; Neuhauss et al., 1999; Vihtelic and Hyde, 2002). In this study, we describe a zebrafish mutant, volvox (vov), in which lens epithelial cells fail to maintain cell proliferation. Furthermore, lens fiber cells fail to fully elongate, and the number, shape and position of their nuclei are abnormal. The vov mutant gene encodes Psmd6, the RP subunit of the 26S proteasome, such that the mutation results in a reduction in proteasome activity. We found that an E3 ubiquitin ligase anaphase-promoting complex/cyclosome (APC/C) is implicated in vov-mediated lens defects. These data suggest that the UPS is required for cell proliferation of lens epithelium and for the differentiation of lens fiber cells in zebrafish.

Zebrafish (Danio rerio) were maintained according to standard procedures (Westerfield, 1995). The RIKEN-wako (RW) and WIK strains were used as wild types for mutagenesis and mapping, respectively. Mutagenesis was performed as previously described (Masai et al., 2003). A vov mutant allele, vov^rw619, and a p53 mutant allele, tp53^M214K (Berghmans et al., 2005), were used. SAGFF168A was identified as a zebrafish Tol2 transposon-mediated transgenic line that expresses a modified Gal4 in lens fiber cells (see Fig. S1 in the supplementary material) (Asakawa et al., 2008). The zebrafish Cecyil (cell cycle illuminated) transgenic line was used (see Fig. S1 in the supplementary material) (Sugiyama et al., 2009).

C57BL/6J mice were used. Embryos were fixed with 4% paraformaldehyde (PFA), dehydrated and embedded in paraffin wax.

Immunostaining of cryosections and paraffin sections, plastic sectioning and in situ hybridization were carried out as described previously (Imai et al., 2006; Masai et al., 2003). Paraffin sectioned slides were pretreated with heat (121°C, 5 minutes, in 10 mM citrate buffer pH 6.0). Nuclear staining was carried out using 1 μg/ml 4′,6-diamidino-2-phenylindole (DAPI; Sigma) or 1 nM TOPRO3 (Molecular Probes). Images were scanned under a LSM510 (Carl Zeiss) or FV1000 (Olympus) confocal laser-scanning microscope. Antibodies are described in Table 1. Terminal deoxynucleotide transferase-mediated dUTP nick-end labeling (TUNEL) was carried out using an In Situ Cell Death Detection Kit (Roche).

Embryos were soaked in water containing 10 mM 5-bromo-2′-deoxyuridine (BrdU) or 1 mM ethynyl uridine (EdU) on ice for 15 minutes, rinsed, incubated for 30 minutes in water at 28.5°C and fixed with 4% PFA. Anti-BrdU labeling was carried out as described previously (Yamaguchi et al., 2005). EdU signals were detected using the Click-iT EdU Alexa Fluor Imaging Kit (Invitrogen).

Donor embryos were labeled at the one-cell stage with a 3-5% 1:1 mixture of FITC-dextran and biotin-dextran (Molecular Probes). Donor cells (5-10) were transplanted to host embryos at the late blastula stage. Host embryos carrying donor cells in the lens were selected by observing FITC fluorescence at 24 hpf, fixed with 4% PFA at 72 hpf and cryosectioned. Filamentous actin (F-actin) was stained using 0.1 μM Rhodamine-conjugated phalloidin (Molecular Probes).

Seventy-two hpf embryos were used for western blotting as described previously (Imai et al., 2006). An antibody against the N-terminal peptide sequence of zebrafish Psmd6 (amino acids 166-179) was generated using a synthetic peptide. Signals were developed using horseradish peroxidase (HRP)-conjugated secondary antibodies and Immobilon Western HRP substrate (Millipore). Luminescence was quantified using a Fuji LAS 4000 image analyzer (Fuji Photo Film). Antibodies are described in Table 1.

GFP and monomeric Kusabira-Orange (mKO) were fused in frame to the N-terminus of the coding region of zebrafish psmd6 (GenBank NM_200291) and lamin A (GenBank AF397016), respectively, and subcloned into the pCS2 vector (Rupp et al., 1994).

MOs (Gene Tools) targeted against psmd6 (psmd6-MO), psmc2-MO, fzr1-ATG-MO, fzr1-splicing-MO and their five-mismatch (5mis) MOs are described in Table S1 in the supplementary material. Because injection of fzr1-5misMO causes severe arrest of embryonic development during gastrulation, we confirmed that co-injection of fzr1 RNA suppresses fzr1 morphant defects (see Fig. S2 in the supplementary material).

The numbers of lens fiber cell nuclei were counted on DAPI-labeled sections. The average and s.d. were determined using six lens images from three different embryos for each combination of genotypes, injections and stages. The percentage of lens fiber cell nuclei positioned in the anterior half of the lens sphere relative to the total number of lens fiber cell nuclei was calculated. Average and s.d. were determined using images of three sectioned lenses from different embryos. The maximum and minimum thickness of lens fiber cell nuclei were measured from images of lens sections and the ratio of the maximum to the minimum thickness was calculated for individual lens fiber cell nuclei using one image for each combination of genotypes, injections and stages. The length of individual lens fibers was measured on three-dimensional images, which were constructed by scanning GFP driven by the SAGFF168A transgenes. Averages were determined using nine lens fibers from two different wild-type embryos, eight lens fibers from two different psmd6 morphant embryos and ten lens fibers from three different fzr1 morphant embryos. Probabilities were determined by Student's t-test.

Total RNA was prepared from ten wild-type and ten vov mutant embryos at 72 hpf. Using an RNA LA PCR Kit (Takara Shuzo), psmd6 and β-actin cDNA fragments were amplified from the same amount of wild-type and vov mutant cDNA in a series of PCR amplification cycles using gene-specific primers (see Table S2 in the supplementary material).

We performed mutagenesis of zebrafish and identified a mutant, vov, which has a small lens core surrounded by less transparent, more irregularly shaped cells at 72 hpf (Fig. 1A-C). Cell death occurred near the ciliary marginal zone (CMZ) in the neural retina in the vov mutant (Fig. 1E; see Fig. S4H,J,L in the supplementary material). In the wild type, nuclei of lens fiber cells are flat and located in the posterior marginal region of the lens sphere at 72 hpf (Fig. 1F), whereas round-shaped nuclei were located not only in the posterior but also anterior marginal region surrounding a small transparent lens core in the vov mutant (Fig. 1G). We examined the expression of molecular markers at 72 hpf. Pax6 (Macdonald and Wilson, 1997) and foxe3 (Shi et al., 2006) are expressed in the lens epithelium of zebrafish (Fig. 1H; see Fig. S3A-A″,E in the supplementary material). Both markers were observed in the lens epithelium of the vov mutant (Fig. 1I; see Fig. S3B-B″,F in the supplementary material). Proliferating cell nuclear antigen (Pcna) is a marker of the S phase and is normally expressed strongly in the lens epithelium and weakly in four or five cells that pass over the region adjacent to the retinal CMZ (Fig. 1J); these latter cells also express the early differentiation marker Prox1 (Glasgow and Tomarev, 1998) (see Fig. S3C-C‴ in the supplementary material). In the vov mutant, Pcna was expressed in the lens epithelium and in the cells located at the posterior edge of the lens sphere (Fig. 1K), and the latter mostly expressed Prox1 (see Fig. S3D-D‴ in the supplementary material), suggesting that lens fiber differentiation is initiated in the vov mutant. We examined cell proliferation in the lens epithelium by labeling with anti-phosphorylated histone H3 (pH3) antibody and BrdU incorporation. In the vov mutant, almost all lens epithelial cells were BrdU negative (Fig. 1L,M). The number of pH3-positive cells in the lens epithelium (n=0.26±0.14 per sectioned lens) was lower in vov mutants than in wild types (n=0.47±0.19), although the difference was not significant (P=0.198, t-test). These data suggest that the lens epithelium is specified normally in terms of the expression of Pax6 and foxe3, but that cell proliferation is slower or arrested in the vov mutant.

Next, we examined the structural integrity of the lens epithelium. Zonula occludens-1 (ZO-1; Tjp1 – Zebrafish Information Network) is a component of the tight and adherens junctions (Fanning and Anderson, 2009) and is localized at the apical surface of both lens epithelium and differentiating lens fiber cells (Kiener et al., 2007) (Fig. 1H; see Fig. S3A-A″ in the supplementary material). In the vov mutant, ZO-1 signals were observed beneath the lens epithelium, although they were weak and discontinuous at some points (Fig. 1I; see Fig. S3B-B″ in the supplementary material). Furthermore, aggregated ZO-1 signals were observed in the posterior region of the lens sphere in the vov mutant (Fig. 1I). Since the lens epithelium seemed to be maintained as a single sheet, discontinuous and aggregated ZO-1 signals in the vov mutant might be due to the disorganization of lens fiber cells. alpha A crystallin mRNA is expressed in differentiating lens fiber cells in wild-type zebrafish (see Fig. S3G in the supplementary material) (Kurita et al., 2003) and was similarly expressed beneath the lens epithelium in the vov mutant (see Fig. S3H in the supplementary material). Aquaporin-0 (AQP0; Mipa/b – Zebrafish Information Network) (Shiels and Bassnett, 1996) is expressed in newly differentiating lens fiber cells in the wild type, which surround the AQP0-negative old lens fiber core (Fig. 1N). In the vov mutant, AQP0 expression was observed in lens cells, which were not tightly packed and surrounded a small AQP0-negative core (Fig. 1O). Lamin B1 demarcates nuclear membranes in zebrafish (Collas, 1999). After cell cycle exit, lens fiber nuclei become flat and are arranged along the posterior rim of the lens sphere in the wild type (Fig. 1P), whereas lens nuclei were not flat and were located in the marginal region of the entire lens in the vov mutant (Fig. 1Q). These data suggest that lens fiber cell differentiation is initiated but nuclear position and shape are disorganized in the vov mutant (Fig. 1R).

We examined apoptosis by TUNEL. There was no TUNEL signal in the wild-type lens at 48, 72 and 96 hpf (see Fig. S4A,C,E in the supplementary material), suggesting that TUNEL does not detect lens denucleation in zebrafish by 96 hpf. In the vov mutant, there was no TUNEL signal at 48 hpf (see Fig. S4B in the supplementary material), but dot-like TUNEL signals were observed around the central lens fiber core at 72 hpf (see Fig. S4D in the supplementary material), and strong TUNEL signals were observed in cells surrounding the small lens fiber core at 96 hpf (see Fig. S4F in the supplementary material). These data suggest that lens nuclei are degraded in the vov mutant after 72 hpf, but possibly not by a normal process of denucleation.

We examined cell proliferation in lens epithelium by EdU incorporation. Here we defined Pax6-positive lens cells as epithelial cells with the margin associated with the retinal CMZ (Fig. 1H; Fig. 2A, red cells). In wild-type epithelium, 30-40% of cells were labeled with EdU from 36 to 72 hpf (Fig. 2B-E), whereas the percentage of EdU incorporation decreased at 48 hpf (Fig. 2F,G) and was almost zero at 60 and 72 hpf (Fig. 2H,I) in the vov mutant. These data suggest that cell proliferation becomes slower or arrested and ceases after 60 hpf in the vov mutant (Fig. 2L). Next, we examined the number of nuclei of lens fiber cells. We defined Pax6-negative lens cells as lens fiber cells (Fig. 2A, green cells) and counted their nuclear number on lens sections cut along the optic axis. In wild types, the number of lens fiber nuclei increased and peaked at 48 hpf, maintained a plateau from 48 to 60 hpf, and decreased at 72 hpf (Fig. 2J,M). The plateau between 48 and 60 hpf suggests that the number of newly generated lens cells is nearly equal to the number of lens nuclei degraded by denucleation. In the vov mutant, the number of lens fiber cell nuclei was lower than in wild type at 36 and 48 hpf (Fig. 2K,M), suggesting a reduced generation rate of lens fiber cells. In contrast to the wild type, the number of lens fiber cell nuclei continued to increase after 48 hpf and reached wild-type levels at 60 hpf in the vov mutant (Fig. 2M). The increase of lens fiber cell nuclei between 48 and 60 hpf in the vov mutant suggests that the number of lens fiber cells generated is higher than the number of lens nuclei degraded by denucleation. From 60 to 72 hpf, the number of lens fiber nuclei decreased in the vov mutant (Fig. 2M). Because there was no production of lens fiber cells after 60 hpf in the vov mutant (Fig. 2L), this reduction of lens nuclei seems to be due to the loss of lens fiber cell nuclei, which may be detected by TUNEL (see Fig. S4D,F in the supplementary material).

We next examined the position of lens fiber cell nuclei. Sectioned lenses cut along the optic axis were divided into anterior and posterior halves (Fig. 2A). The percentage of lens fiber nuclei positioned in the anterior half relative to the total number of lens fiber cell nuclei was defined as the index of nuclear anterior localization (Ina) (Fig. 2J,K,N). In wild types, Ina was ~30% at 36 and 48 hpf, decreased to 20% at 60 hpf and 3% at 72 hpf (Fig. 2N), suggesting that 97% of lens fiber cell nuclei were positioned in the posterior half of the lens sphere at 72 hpf. However, in the vov mutant, Ina was greater than 30% until 60 hpf and decreased to 15% at 72 hpf, which is higher than that of wild types. These observations suggest that a significant number of lens nuclei are positioned in the anterior half of the lens sphere in the vov mutant.

Next, we examined the shape of lens fiber cell nuclei. To determine their flatness, we measured the maximum and minimum thickness of lens fiber cell nuclei and calculated the ratio of the maximum to the minimum thickness as the index of nucleus flatness (Inf) (Fig. 2A). In wild types, Inf varied from 1.0 to 2.0 at 24 hpf, from 1.0 to 4.5 at 36 hpf, and from 1.0 to greater than 5.0 after 48 hpf (Fig. 2O), indicating that the maximum value of Inf increases as development proceeds. Lens nuclei with a higher value of Inf were more posteriorly positioned in the lens sphere (Fig. 2J, lower panels). In a majority of lens fiber cell nuclei of the vov mutant, Inf was less than 3.0 at all stages examined (Fig. 2K,O), suggesting that the nuclei of lens fiber cells fail to flatten. Electron microscopy analyses confirmed the defects in nuclear position and shape of lens fiber cells in the vov mutant (see Fig. S5 in the supplementary material). Furthermore, we found that retinal apoptosis in the vov mutant was inhibited by a genetic mutation of the tumor suppressor p53 (tp53) (Berghmans et al., 2005) (see Fig. S6DL in the supplementary material). However, the proliferation of lens epithelial cells and the nuclear position and shape of lens fiber cells were affected in the vov; p53 double mutant (see Fig. S6H,V,W in the supplementary material). It seems likely that retinal apoptosis is not a primary cause of vov-mediated lens defects (see Fig. S6 in the supplementary material).

To visualize individual lens fiber cells, we used a zebrafish transgenic line SAGFF168A; UAS:EGFP, in which GFP is expressed in differentiating lens fiber cells. At 50 hpf, GFP-positive lens fiber cells elongate toward the anterior and posterior poles of the lens sphere (Fig. 3A). At 72 hpf, most lens fiber cells express GFP and their apical tips converge at one point at the anterior pole of the wild-type lens sphere (Fig. 3C). In the vov mutant, lens fiber cells were shorter than in the wild type at 50 hpf (Fig. 3B), and markedly smaller at 72 hpf (Fig. 3D), suggesting that the vov mutant lens fiber cells do not fully elongate. Actin filament bundles develop during the differentiation of lens fiber cells (Bassnett et al., 1999; Lee et al., 2000). Filamentous actin (F-actin) was densely labeled with phalloidin in the wild-type lens fiber region (Fig. 3E), whereas F-actin density was sparse in the vov mutant (Fig. 3F).

Next, we carried out cell transplantation at the blastula stage. When wild-type donor cells were incorporated into the wild-type recipient lens, the donor lens cells differentiated into flat and elongated lens fiber cells (Fig. 3G), their nuclei were observed in the posterior half of the lens sphere at 72 hpf (Fig. 3H), and F-actin accumulated in differentiating lens fiber cells (Fig. 3I). When the vov mutant donor cells were incorporated into the wild-type recipient lens, the vov mutant donor cells seemed to differentiate into lens fiber cells that were thicker than wild-type donor cells (Fig. 3J), and their nuclei were frequently positioned in the anterior half of the lens sphere (n=7 from three lenses) (Fig. 3K,M,M′, yellow arrowheads). F-actin density was decreased in the vov mutant donor lens fiber cells (n=5 from three lenses) (Fig. 3L,N,N′, white arrows). These data suggest that the vov mutation cell-autonomously compromises the elongation of lens fiber cells, their nuclear position and F-actin assembly.

We mapped the genetic locus of the vov mutation between two polymorphic markers, zK22H2 and zK251M8, at 39.3 cM on chromosome 11 (see Fig. S7A in the supplementary material), and sequenced cDNAs of candidate genes. We found that the vov gene encodes Psmd6, a RP subunit of the proteasome. In the vov mutant genome, a G-to-A substitution occurred at the splice acceptor site prior to the seventh exon of psmd6, resulting in a single base shift of the splice acceptor site (Fig. 4A, top). This one-base deletion of cDNA causes a frameshift in the coding region, resulting in the loss of the proteasome-COP9-initiation factor (PCI) domain, which is required for proteasome activity (Sha et al., 2007) (Fig. 4A, bottom). Next, we carried out RNA-mediated rescue and MO-mediated knockdown experiments. Injection of GFP-tagged psmd6 RNA into vov mutant embryos suppressed retinal cell death (see Fig. S8 in the supplementary material) and lens defects at 72 hpf (Fig. 4B,C; see Fig. S7C,G in the supplementary material). The injection of psmd6-MO induced the vov-like lens defects (Fig. 4B,C; see Fig. S7E,I in the supplementary material). These data suggest that the vov mutant gene encodes Psmd6.

We next examined psmd6 mRNA expression. psmd6 mRNA was ubiquitously expressed at the eight-cell stage, suggesting maternal expression (Fig. 4D). The expression was observed ubiquitously at 20 hpf (Fig. 4E), and was prominent in the head at 48 (Fig. 4F) and 72 (Fig. 4H) hpf. psmd6 mRNA was expressed in the anterior lens epithelium and retinal CMZ (Fig. 4G,I), suggesting that psmd6 mRNA is expressed in proliferating cells and early differentiating cells in both the lens and retina. Semi-quantitative PCR revealed that the mRNA level in the vov mutant was similar to that of the wild type (Fig. 4J). Western blotting, using an anti-zebrafish Psmd6 antibody, revealed a single band of 44 kDa, which is identical to the molecular weight of Psmd6 (Fig. 4K). This band decreased in both the vov mutant and psmd6 morphant at 72 hpf (Fig. 4K), suggesting that the level of Psmd6 protein is reduced in both cases.

A yeast homolog of psmd6, rpn7, is required for proteasome activity (Isono et al., 2007; Isono et al., 2004; Sha et al., 2007). We examined the level of polyubiquitylation in the vov mutant and psmd6 morphant by western blotting using the anti-polyubiquitylated protein antibody FK2 (Fujimuro et al., 1994). Polyubiquitylated proteins accumulated in both the vov mutant and the psmd6 morphant at 72 hpf (Fig. 4L), suggesting that proteasome-dependent protein degradation is compromised in both cases. Furthermore, we injected an MO for psmc2 (rpt1 in yeast), which encodes another RP subunit of the proteasome, into wild-type embryos. The psmc2 morphant showed an increase in polyubiquitylation at 72 hpf (Fig. 4L) and lens defects similar to those of the vov mutant and psmd6 morphant (Fig. 4B,C; see Fig. S7J-N in the supplementary material). These data suggest that the reduction of proteasome activity correlates with lens defects in the vov mutant.

Using the anti-polyubiquitylated protein antibody FK2, we examined the spatiotemporal pattern of polyubiquitylation in the zebrafish lens. At 24 hpf, signals were detected in nuclei of wild-type lens cells (Fig. 5A). At 48 hpf, signals were weaker in nuclei of lens epithelial cells, but were maintained at high levels in the nuclei of differentiating lens fiber cells (Fig. 5B). At 72 hpf, signals were detected in nuclei of differentiating lens fiber cells (Fig. 5C, arrowheads). In the vov mutant, the expression levels of polyubiquitylated proteins increased (Fig. 5D-F). The antibody signals nearly reached saturation level in the vov mutant, and signals were high in lens fiber cell nuclei (Fig. 5E). These observations suggest that polyubiquitylated proteins accumulate in the nucleus of zebrafish lens fiber cells. We also found that polyubiquitylated proteins accumulated in the nuclei of lens epithelium and lens fiber cells in the mouse (Fig. 6A,B).

We examined the localization of Psmd6 in the zebrafish lens using the anti-Psmd6 antibody used for western blotting (Fig. 4K). However, we were unable to detect specific signals by immunohistolabeling (data not shown). Because the GFP-tagged Psmd6 suppressed lens defects in the vov mutant (Fig. 4B,C), the GFP-tagged Psmd6 was considered to be functional. Thus, we injected RNA encoding the GFP-tagged Psmd6 into wild-type embryos and examined its expression pattern. GFP signals were detected throughout the lens cells at 24 hpf (Fig. 7A), but localization was restricted to nuclei of lens fiber cells at 33 (Fig. 7B) and 48 (Fig. 7C) hpf. The intensity of GFP in the lens epithelial nuclei was weaker than that of lens fiber cells at 33 and 48 hpf. These localization patterns were similar to those of polyubiquitylated proteins (Fig. 5A-C).

To confirm whether the nuclear localization of GFP-tagged Psmd6 is related to proteasome function, we carried out two sets of experiments. First, we examined the subcellular localization of GFP-tagged Psmd6 carrying an amino acid substitution, L289D, which compromises the assembly of Psmd6 into the proteasome complex (Sha et al., 2007). GFP-tagged Psmd6 with the L289D mutation was localized to the nuclei of lens fiber cells less efficiently than GFP-tagged wild-type Psmd6 (Fig. 7D-F). Second, we examined the localization of GFP-tagged Psmd6 in the psmc2 morphant lens. GFP-tagged Psmd6 was not localized efficiently to lens fiber nuclei in the psmc2 morphant (Fig. 7G-I). These data suggest that the nuclear localization of GFP-tagged Psmd6 depends on proteasome function.

APC/C regulates the G2/M phase transition in the cell cycle (Pesin and Orr-Weaver, 2008; Peters, 2006) and the morphogenesis of differentiating neurons (Konishi et al., 2004; Stegmuller et al., 2008). APC/C inhibition causes a decrease in the number of lens fibers in the mouse (Wu et al., 2007). We focused on one of the APC/C targets, Geminin (McGarry and Kirschner, 1998). To monitor the APC/C-dependent degradation of Geminin, we used Cecyil transgenic zebrafish, which express monomeric Kusabira-Orange 2-tagged zebrafish Cdt1 N-terminal 190 amino acid domain (mKO2-zCdt1) and monomeric Azami-Green-tagged zebrafish Geminin N-terminal 100 amino acid domain (mAG-zGem) (Sugiyama et al., 2009). In the Cecyil transgenic line, mAG-zGem expression was detected only in lens epithelial cells at 48 hpf (Fig. 8A), whereas mKO2-zCdt1 expression was detected in lens fiber cells (Fig. 8A′). In the psmd6 morphants, mAG-zGem was detected ectopically in lens fiber cells, which were mKO2-zCdt1 positive (Fig. 8B′), suggesting that APC/C-dependent protein degradation is compromised in the psmd6 morphant. Fzr1 (Cdh1) is a coactivator of APC/C (Pesin and Orr-Weaver, 2008; Peters, 2006). We examined the lens phenotypes of zebrafish fzr1 morphants. In the fzr1 morphant lens, mAG-zGem was retained in mKO2-zCdt1-positive lens fiber cells (Fig. 8C′,D′). Furthermore, rounded nuclei of lens fiber cells were positioned throughout the entire region of the lens sphere (Fig. 8E). Similar to the vov mutant, cell proliferation was affected, but lens epithelium and lens fiber cells were specified in terms of Pax6 and AQP0 expression in the fzr1 morphant lens (see Fig. S9 in the supplementary material). The maximum value of Inf was lower in the severe fzr1 morphants than in wild types (Fig. 8H). Furthermore, Ina was higher in the fzr-1 morphants than in wild types (Fig. 8I). There was no TUNEL signal in the fzr1 morphant lens (Fig. 8F,G). Finally, lens fiber cells did not elongate fully in the fzr1 morphant (Fig. 8J-M).

Proliferation and differentiation are tightly controlled in the vertebrate lens (Lovicu and McAvoy, 2005). It has been reported that the application of proteasome inhibitors suppresses the proliferation of lens epithelial cells (Guo et al., 2006; Liu et al., 2006). In vov mutants, lens epithelial cells are specified normally in terms of the expression of pax6 and foxe3 but fail to maintain cell proliferation. Our findings provide in vivo evidence that the UPS is required for cell cycle control in the vertebrate lens epithelium. In the vov mutant, the number of lens epithelial cells in the S phase is reduced at 48 hpf and they are almost absent after 60 hpf. However, pH3-positive lens epithelial cells and Prox1-positive early differentiating lens fiber cells appear in the vov mutant at 72 hpf, although the number of pH3-positive cells is lower in the vov mutant than in the wild type. How is cell cycle control affected in the vov mutant? The first possibility is that lens epithelial cells are arrested in G1 phase and do not enter S phase after 60 hpf. In this case, lens epithelial cells that exit from the S phase by 60 hpf may continue to undergo the G2/M phase, and some of them may differentiate into lens fiber cells at 72 hpf. The second possibility is that cell cycle arrest or stalling occurs in the intra-S, G2 or late M phase. Further studies will be necessary to address this issue.

Another important question is whether developmental defects outside the lens contribute to the vov lens phenotypes. It has been reported that proliferation of lens epithelial cells depends on growth factors (Lovicu and McAvoy, 2005), and as such the vov-mediated defects in cell cycle control might be caused by extrinsic factors that emanate from tissues outside the lens. In this study, we found that retinal cell death is suppressed in the vov; p53 double mutant, but that the p53 mutation does not suppress cell proliferation defects in the vov mutant. It seems unlikely that retinal apoptosis is a primary cause of the vov-mediated lens defects.

In this study, we found that nuclear and cell morphogenesis are affected and F-actin density is sparse in the vov mutant lens fiber cells. Actin filaments are integral components of the plasma membrane-associated cytoskeleton. Stabilization and remodeling of the membrane cytoskeleton are important for lens fiber differentiation (Lee et al., 2000). The disruption of Rho GTPase activity impairs the organization of actin filaments and lens fiber cell elongation (Maddala et al., 2008). Recent proteome analysis has revealed that ubiquitin is one of the most abundant membrane-associated components of mouse lens fiber cells (Bassnett et al., 2009). These observations suggest that UPS is required for the morphogenesis of lens fiber cells, probably through the regulation of actin filament assembly. It was reported that TGFβ signaling is required for lens fiber differentiation and that the introduction of a dominant-negative TGFβ receptor in the lens results in the inhibition of cell migration and in a failure to assemble F-actin (de Iongh et al., 2001). In the mouse lens, TGFβ signaling promotes APC/C-dependent degradation of the transcriptional co-repressor Sno, which inhibits the transcription of Cdk inhibitors (Wu et al., 2007), suggesting that APC/C mediates the TGFβ-induced cell cycle exit of lens epithelial cells. We found that knockdown of the APC/C regulator Fzr1 induces the nuclear and cell morphogenesis defect of lens fiber cells. These findings suggest that APC/C regulates not only the cell cycle, but also nuclear and cell morphogenesis during lens fiber cell differentiation.

Denucleation is a characteristic feature of lens fiber differentiation. The degradation of DNA of lens cells depends on DLAD (Nishimoto et al., 2003), suggesting that DNA degradation occurs in an acidic organelle, possibly in the lysosome. This raised the possibility that cellular organelles are degraded by autophagy in the cell's own lysosomes during lens fiber cell differentiation (Nagata, 2005; Nakahara et al., 2007). However, organelle degradation occurs normally in the lens of autophagy-deficient Atg5 knockout mice (Matsui et al., 2006). In this study, we found that the stage-dependent increases and decreases in lens fiber cell nuclei are altered in the vov mutant. A number of lens fiber cell nuclei maintained a plateau between 48 and 60 hpf in the wild type, suggesting a balance between the generation of lens fiber cells and the denucleation of lens fiber nuclei. In the vov mutant, the number of lens fiber cell nuclei continued to increase from 48 to 60 hpf, suggesting that the degradation of lens fiber cell nuclei is affected in the vov mutant. Furthermore, the fraction of lens fiber cell nuclei located in the anterior half of the lens sphere decreased in the wild type, suggesting that most of the lens nuclei, which are initially positioned in the anterior half region, are degraded. However, a significant number of lens fiber nuclei are retained in the anterior half region in the vov mutant. This also supports the idea that denucleation is affected in the vov mutant lens fiber cells.

How is the UPS involved in the denucleation processes? Polyubiquitylated proteins are localized in the nucleus of lens fiber cells in zebrafish and mice. GFP-tagged Psmd6 shows a similar spatial localization pattern to the polyubiquitylated proteins, which seems to depend on proteasomal functions. UPS components are localized in the nuclei of bovine differentiating lens fiber cells (Girao et al., 2005). These observations suggest that a high level of polyubiquitylation occurs in nuclei of lens fiber cells among vertebrates. It is possible that the UPS directly dismantles the nuclear membrane and chromatin from DNA during lens cell denucleation. In the future, it will be important to elucidate the roles of the UPS in lens denucleation.

We thank Dr Atsushi Miyawaki (RIKEN BSI, Japan) for providing the zebrafish Cecyil transgenic fish. This work was supported by a Grant-in-Aid from the Ministry of Education, Culture, Sports, Science and Technology to F.I. and a grant from OIST to I.M.