Tubgcp3 Is Required for Retinal Progenitor Cell Proliferation During Zebrafish Development

The centrosomal protein γ-tubulin complex protein 3 (Tubgcp3/GCP3) is required for the assembly of γ-tubulin small complexes (γ-TuSCs) and γ-tubulin ring complexes (γ-TuRCs), which play critical roles in mitotic spindle formation during mitosis. However, its function in vertebrate embryonic development is unknown. Here, we generated the zebrafish tubgcp3 mutants using the CRISPR/Cas9 system and found that the tubgcp3 mutants exhibited the small eye phenotype. Tubgcp3 is required for the cell cycle progression of retinal progenitor cells (RPCs), and its depletion caused cell cycle arrest in the mitotic (M) phase. The M-phase arrested RPCs exhibited aberrant monopolar spindles and abnormal distributed centrioles and γ-tubulin. Moreover, these RPCs underwent apoptosis finally. Our study provides the in vivo model for the functional study of Tubgcp3 and sheds light on the roles of centrosomal γ-tubulin complexes in vertebrate development.


INTRODUCTION
Centrosome, the major microtubule-organizing center (MTOC) in vertebrate cells, provides a major site for microtubule (MT) nucleation and plays key roles in bipolar spindle assembly during mitosis (Kellogg et al., 1994). A typical centrosome consists of a pair of centrioles surrounded by the pericentriolar matrix (PCM) (Bornens, 2002). Many proteins, including γ-tubulin (GCP1), γ-tubulin complex proteins (GCPs) and a large number of other centrosome-associated proteins, localize to PCM and are involved in the formation of mitotic spindles. Together with other GCPs, γ-tubulin forms two distinctly sized complexes: the γ-tubulin small complex (γ-TuSC) and the γ-tubulin ring complex (γ-TuRC). The γ-TuSC is a heterotetramer consisting of two copies of γ-tubulin and one copy each of GCP2 and GCP3/Tubgcp3. Seven γ-TuSCs with GCP4, GCP5, GCP6 and other accessory proteins assemble into the γ-TuRC, which facilitates MT nucleation by capping the minus ends of MTs and protecting them from depolymerization (Zheng et al., 1995). Centrosomes have critical roles in brain development, and mutations in genes encoding for centrosome-associated proteins have been shown to be genetically linked to neurodevelopmental disorders (Novorol et al., 2013;Chavali et al., 2014;Morris-Rosendahl and Kaindl, 2015;Buchwalter et al., 2016).
Many studies have shown that γ-TuSC is involved in MT nucleation and mitotic spindle assembly, and every component of γ-TuSC is indispensable for cell cycle progression (Oakley et al., 1990;Geissler et al., 1996;Knop et al., 1997;Yuba-Kubo et al., 2005;Xiong and Oakley, 2009;Pouchucq et al., 2018). In Saccharomyces cerevisiae, cells overexpressing the wild-type Spc98p/GCP3 or carrying the temperature-sensitive allele spc98-1 arrest in mitosis with a defective spindle (Geissler et al., 1996). The duplication and separation of spindle pole body (SPB; yeast centrosome) are not affected in these cells. Disruption of gcpC/GCP3 in Aspergillus nidulans results in absent functional spindles and defective mitosis (Xiong and Oakley, 2009). In Drosophila mutants for disks-degenerate 4 (dd4, which encodes GCP3), cells exhibit reduced density of spindle microtubules and delayed cell cycle progression from mitosis (Barbosa et al., 2000). Centriole duplication and separation are defective in some of the dd4 mutant cells. In addition, γ-tubulin is missing from the spindle poles and becomes dispersed throughout the cell. Mitotic arrest has been observed in Hela, T98, and U87MG cell lines after depletion of GCP3 using siRNA (Draberova et al., 2015;Farache et al., 2016;Cota et al., 2017). However, Mikule et al. (2007) found that cells harboring wild-type p53 (RPE-1, BJ-1, HME-1, and HCT-116) arrest in G1 phase in a p53-dependent manner after being transfected with siRNA against GCP3. Given that U87MG cells expressing wild-type p53 show mitotic arrest, it remains unclear which types of cell cycle defects will happen in the GCP3-deleted cells with wild-type p53 gene. These different cell cycle defects may occur in a cell type-specific manner as these studies were carried out using different cell lines. Moreover, there have been no functional studies of GCP3 in vivo in a vertebrate system.
Among vertebrate models, the zebrafish provides many unique advantages over other rodents for gene functional study during early embryonic development. For example, owing to its external fertilization and rapid development, zebrafish early embryos can be easily visualized and manipulated. In addition, maternal gene products, synthesized during oogenesis and supplied to the egg, play essential roles in the earliest stages of zebrafish embryonic development (Pelegri, 2003). With disruption of any of the maternal-effect genes, the embryos continue to develop until the maternal supply is exhausted, which facilitates the functional study of these genes at relatively late developmental stages. The zebrafish retina is part of the central nervous system (CNS), and its neuroanatomy is well characterized. The ciliary marginal zone (CMZ), a proliferative region located at the periphery of the retina, provides an excellent model for the study of neurodevelopment (Novorol et al., 2013). The retina grows continuously throughout life and almost all retina neurogenesis comes from CMZ after the embryogenesis of retina is completed at 60 hpf (Marcus et al., 1999). The CMZ contains retinal stem cells (RSCs) and retinal progenitor cells (RPCs), exhibiting a peripheral-tocentral arrangement pattern. The RSCs were located nearest to the periphery, the proliferative RPCs resided in the middle, and the post-mitotic RPC cells were positioned at the most central of the CMZ (Wehman et al., 2005;Cerveny et al., 2010;Valdivia et al., 2016;Wan et al., 2016). In CMZ, cell proliferation and differentiation are precisely coordinated for the growth of zebrafish eyes. Many cell proliferation and differentiation-associated genes have been studied using the zebrafish model (Wehman et al., 2005;Cerveny et al., 2010;Valdivia et al., 2016).
In this study, we explored the in vivo function of tubgcp3 using the zebrafish model. We found that Tubgcp3 is essential for the development of zebrafish retina. Knockout of the tubgcp3 gene resulted in the small eye phenotype exhibiting CMZ defects due to the abnormal cell cycle progression. Depletion of Tubgcp3 in RPCs caused mitotic arrest and apoptosis. Our findings reveal the critical roles of GCP3 in cell cycle progression, providing insights into the function of its associated complexes, γ-TuSC and γ-TuRC, in development, and establish a vertebrate model for further study.

Zebrafish Maintenance
Zebrafish (Danio rerio) were maintained and bred under standard conditions as previously described (Westerfield, 2000). Stages of embryonic development were determined according to their morphology (Westerfield, 2000). Embryos were treated with 0.003% 1-phenyl-2-thiourea (PTU) in egg water to prevent the production of pigment. The transgenic line Tg(HuC:GFP) was used in this study, in which the zebrafish brain is labeled with green fluorescent protein (GFP) (Park et al., 2000). All animal experiments were approved by the Institutional Animal Care and Use Committee, Fudan University.

Generation of tubgcp3 Mutants by the CRISPR/Cas9 System and Phenotypic Analysis
Disruption of tubgcp3 transcript variant X1 (GenBank accession number XM_005170995) was performed using CRISPR/Cas9 technology. Cas9 nuclease target sites were designed using the ZiFiT Targeter online software. The target site in the exon 4 was selected and the corresponding sequence was 5 -GGTCCTCACAGAGGCTGAGCAGG-3 . Cas9 mRNA and guide RNA (gRNA) were synthesized using the mMESSAGE mMACHINE T7 Transcription Kit (AM1344, Invitrogen, United States) and MAXIscript T7 Transcription Kit (AM1322, Invitrogen, United States), respectively. Then, 300 pg Cas9 mRNA and 30 pg gRNA were co-injected into the one-cell stage zebrafish embryos to knockout the tubgcp3 gene. The PCR primers for genotyping were as follows: forward (5 -ATTACGCAGAGGACCAAGA-3 ) and reverse (5 -AAAATAGGATTTCATACAGGAACCCG-3 ). The tubgcp3 mutants were identified at 3 dpf by the developmental defects of the head and the arch. Heterozygous animals were incrossed to generate embryos for experimental analyses. Homozygous mutants were used as the experimental group and siblings as the control in this study.
For morphological measurements, embryos were anesthetized with 0.16 mg/mL Tricaine (A5040, Sigma-Aldrich) and embedded in 3% methylcellulose (M0387, Sigma-Aldrich). Body length was measured laterally in the rostral-caudal axis. Eye size was measured laterally along the longest diameter of the eye. Brain size was measured by calculating the dorsal area of the brain using the transgenic line Tg(HuC:GFP) (Park et al., 2000).
For cartilage staining, 5 dpf embryos were fixed in Bouin's solution overnight. The fixed embryos were washed three times in 70% ethanol/0.1% ammonia, then washed two more times in 95% ethanol. The embryos were transferred into Alcian blue solution (80% ethanol/20% glacial acetic acid, 0.1% Alcian blue) and stained overnight. After staining, the embryos were gradually rehydrated in a decreasing ethanol series (60%, 40%, and 20%) and washed twice in phosphate-buffered saline (PBS). The embryos were incubated in 3% hydrogen peroxide/1% potassium hydroxide until the pigmentation was removed. After that, the samples were incubated in 0.05% trypsin in 30% sodium tetraborate until cleared. Then, the embryos were transferred into 70% glycerol for imaging.

In situ Hybridization Analyses
In situ hybridization was performed on 10 µm cryosections of zebrafish eyes. Briefly, zebrafish embryos were fixed in 4% paraformaldehyde overnight at 4 • C. The next day, the fixed embryos were washed three times with PBS and dehydrated in 30% sucrose. Then, the embryos were embedded in OCT (Invitrogen) before being sectioned at 10 µm with a Leica cryostat (Leica CM1860, Leica Microsystems, Germany). For whole-mount in situ hybridization, the steps were performed as previously described (Thisse and Thisse, 2008).

HE Staining
For Hematoxylin and Eosin (HE) staining, embryos were dehydrated using a graded ethanol series (50, 70, 90, 95, and 100%) before being transferred to xylene. Then, the embryos were embedded in paraffin wax and sectioned at 5 µm using a Leica Cryostat (Leica RM2235, Leica). HE staining was performed according to standard protocols.
To detect cell death, a TUNEL assay was performed on sections using the In Situ Cell Death Detection Kit (Roche) according to the manufactures' instruction.

Senescence-Associated β-Galactosidase (SA-β-gal) Assay
To detect cellular senescence, β-galactosidase assay was performed using the Senescence Cells Histochemical Staining Kit (CS0030, Sigma-Aldrich). Zebrafish embryos (5 dpf) were fixed in 0.2% glutaraldehyde at RT for 2 h. The fixed embryos were washed three times in PBS before being embedded in OCT (Invitrogen). Then, the embryos were sectioned at 10 µm for the assay.

Imaging
The images of in situ hybridization, HE staining and SA-β-gal assay were captured using a Nikon ECLIPSE Ni microscope (Nikon) or Olympus microscope (IX83, Olympus). Immunostaining images were acquired using a Zeiss Axio Observer.Z1 microscope (Zeiss) or Nikon A1 confocal microscope (Nikon). Bright-field images of whole mount zebrafish embryos were taken using a Leica microscope (Leica M205FA, Leica).

Statistical Analyses
To calculate cell proliferation and cell death in zebrafish retina, sections of the most central portion of the retina were selected for the analysis. The cell number was manually counted using ImageJ software within a defined region in the nasal part of the retina including the nasal CMZ (140 µm × 100 µm). Proliferating cells in embryo tails was counted within a defined region including ∼400 µm of the tails. Statistical analyses were performed with unpaired Student's t-test using GraphPad Prism software. A P-value < 0.05 was considered statistically significant ( * P < 0.05, * * P < 0.01). Data are expressed as mean + SEM.

RESULTS
The Expression Pattern of Zebrafish tubgcp3 During Early Development To analyze the roles of tubgcp3 during zebrafish early development, we first examined the expression pattern of tubgcp3 by RT-PCR ( Figure 1A) and whole-mount in situ hybridization (WISH) (Figures 1B-J). The RT-PCR analyses showed tubgcp3 transcripts were maternally deposited and continually expressed from 2-cell stage to 5 days post fertilization (dpf) during early embryonic development ( Figure 1A). The expression of tubgcp3 mRNA first decreased and then increased from 2-cell stage to 24 hpf, indicating the degradation of maternally deposited tubgcp3 transcripts and the activation of zygotic transcripts during the maternal-to-zygotic transition (MZT). Consistent with RT-PCR results, WISH analyses showed that tubgcp3 mRNA was detected at all developmental stages from cleavage stage to 3 dpf compared with the control group (Figures 1B-J,N,O). Furthermore, we found that the expression of tubgcp3 was enriched in the head and eye as well as weak expression in other tissues of the embryos from 1 to 3 dpf (Figures 1B-J).
To further characterize its expression in zebrafish retina, in situ hybridization (ISH) was performed on cryosections of the embryo eyes. In comparison with the sense probe as the control (Figure 1P), tubgcp3 was expressed throughout the retina with a gradual enrichment at CMZ (Figures 1K-M). We also detected the expression of tubgcp2 and γ-tubulin, which encodes another two subunits of γTuSC. Their expression was similar to that of tubgcp3 (Supplementary Figures S1A-H). In addition, the expression of pcna, a proliferation marker, was confined in the CMZ from 3 to 5 dpf (Supplementary Figures S1I-L). These data indicate that tubgcp3 may have a role in cell proliferation during zebrafish retinal development.

Generation of Zebrafish tubgcp3 Mutants and Morphological Analysis
To investigate the roles of tubgcp3 during zebrafish embryonic development, we established tubgcp3 knockout zebrafish lines using the CRISPR/Cas9 system. The CRISPR/Cas9 target site was designed in the exon 4 of the tubgcp3 gene. Finally, two mutant alleles with 5-base pair (bp) deletion and 11-bp insertion were identified (Figure 2A and data not shown). Since the two mutant lines exhibited the same phenotypes, we used the 5-bp deletion line for further research. The 5-bp deletion was predicted to lead to a premature termination in Tubgcp3 protein translation and yield a 99-amino acid (aa)-truncated Tubgcp3 protein absent of the two conversed GRIP domains ( Figure 2B). tubgcp3 transcripts were detected in the tubgcp3 mutant retina (Supplementary Figure S2A), which suggested that the mutated tubgcp3 mRNA escaped non-sense-mediated mRNA decay (NMD). In addition, sequencing analyses confirmed that the 5-bp deletion transcripts existed in tubgcp3 mutants (Supplementary Figure S2B). Furthermore, the full-length Tubgcp3 protein was significantly decreased in mutant embryos ( Figure 2C). The residual full-length Tubgcp3 in tubgcp3 mutants maybe attributed to the maternal supplied Tubgcp3 that was unusually stable in its associated complexes or depleted at a very low rate in embryos. All these results indicate a loss-of-function mutation in tubgcp3 gene.
The tubgcp3 mutant embryos were indistinguishable from their wild-type siblings before 2 dpf. From 3 to 5 dpf, the mutants exhibited a progressed MCPH (microcephaly)-like phenotype with reduced brain and eye size (Figures 2D,F).
We determined the small brain phenotype using transgenic Tg(HuC:GFP), in which GFP expression is controlled by the promoter of neuronal gene HuC (Park et al., 2000). We found that the brain size was reduced about 10% in the tubgcp3 mutants compared to the wild-type siblings at 3 dpf ( Figure 2G). The brain size reduction in tubgcp3 mutants became more significant (about 24% reduction) when measured at 5 dpf ( Figure 2G). Notably, the eyes became smaller in the tubgcp3 mutants when compared with wild-type siblings from 3 dpf to 5 dpf (Figures 2E,F). Other defects including dorsal tail curvature (Figure 2D), uninflated swim bladder (Figure 2D), body length reduction (Supplementary Figure S3A) and jaw malformation (Supplementary Figure S3B) were also visible in tubgcp3 mutants. The developmental defects were gradually more serious, and the mutants died around 10-14 dpf.

The tubgcp3 Mutants Exhibit Abnormal CMZ
In order to investigate the eye defects in tubgcp3 mutants in detail, we performed Hematoxylin and Eosin (HE) staining on transverse eye paraffin sections. The tubgcp3 mutant retina showed defective CMZ with disorderly cells when compared to wild-type siblings at 3 dpf (Figures 3A,B,E,F). Notably, these abnormal cells disappeared from the CMZ at 5 dpf (Figures 3C,D,G,H). However, the laminar structure of central retina appeared normal in tubgcp3 mutants at 3 dpf and 5 dpf (Figures 3I-L), including the ganglion cell layer (GCL), inner plexiform layer (IPL), inner nuclear layer (INL), outer plexiform layer (OPL), outer nuclear layer (ONL) and retinal pigment epithelium (RPE).
Ciliary marginal zone is responsible for adding new neurons to the continuously growing retina in zebrafish larvae, which consists of retinal stem cells (RSCs) and retinal progenitor cells (RPCs) (Wehman et al., 2005;Cerveny et al., 2010;Valdivia et al., 2016;Wan et al., 2016). vsx2 encodes a homeodomain transcription factor, which is expressed in CMZ cells, Müller glia and a subpopulation of bipolar cells (Vitorino et al., 2009). The expression region of vsx2 was reduced in tubgcp3 mutant CMZ from 3 to 5 dpf (Figures 4A-D), which was consistent with the development defects detected by HE staining (Figures 3E-H). col15a1b expressed at the most periphery of CMZ, where it was considered to be a stem-cell niche (the location of RSCs) (Pujic et al., 2006;Cerveny et al., 2010;Gonzalez-Nunez et al., 2010;Valdivia et al., 2016). Its expression seems to be unaffected in the tubgcp3 mutants (Supplementary Figure S4A). ccnd1, a marker for proliferating cells, encodes Cyclin D1 (CCND1) which is required for G1-S transition (Cerveny et al., 2010;Valdivia et al., 2016). The expression of ccnd1 was significantly reduced in tubgcp3 mutant CMZ (Figures 4E-H). atoh7 encodes ATH5, a member of bHLH transcription factor, and is involved in the differentiation of ganglion cells (Masai et al., 2000;Cerveny et al., 2010;Valdivia et al., 2016). cdkn1c encodes the p57 cip/kip , a cyclin-dependent kinase inhibitor, which is required for many retinal cells to exit from the cell cycle before their further differentiation (Ohnuma et al., 1999;Shkumatava and Neumann, 2005;Cerveny et al., 2010;Valdivia et al., 2016). The expression of atoh7 and cdkn1c was significantly reduced in tubgcp3 mutant CMZ at 3 dpf and almost disappeared at 5 dpf (Figures 4I-P). These results reveal that the tubgcp3 mutants suffer from a significant reduction in proliferating and differentiating in CMZ cells.
In addition, different retinal cell types, including Müller glia cells (expressing vsx2) (Figures 4A-D), a subset of bipolar cells (expressing vsx2) (Figures 4A-D), cones (labeled with anti-ZPR1 for double cone photoreceptors and anti-M-opsin for cone outer segments) (Figures 4Q-T and Supplementary Figure S4B) and rods (labeled with anti-4D2 for rod outer segments) (Supplementary Figure S4B), appeared normal in the tubgcp3 mutants. Moreover, cilia in photoreceptor cells (labeled with anti-acetylated α-tubulin) and retinal laminar structures, including the inner segment, inter plexiform layer (IPL), the outer limiting membrane (OLM) and the outer plexiform layer (OPL) (labeled with phalloidin) (Supplementary Figure S4B and data not shown), were also normal in tubgcp3 mutants. These results suggest that these differentiated cells and structures are unaffected in the central reitna of the tubgcp3 mutants.

Loss of Tubgcp3 in CMZ Cells Causes M-Phase Arrest
Given the reduction of cell proliferation and differentiation in the tubgcp3 mutant CMZ (Figures 4E-P), we first examined whether cell cycle progression was defective in mutant retinae using BrdU incorporation as an S phase marker and Phospho-Histone H3 (PH3) as a mitotic marker. Considering the length of the cell cycle in the CMZ at early larval stages is about 6 to 8 h (Li et al., 2000;Wehman et al., 2007), zebrafish embryos were incubated with BrdU from 66 to 72 hpf and then subjected to section and cell cycle analysis. Cycling RPCs would be labeled by BrdU except those that had passed S phase and would exit cell cycle and undergo differentiation.
We compared expression patterns of BrdU and PH3 between the tubgcp3 mutants and wild-type siblings. In wild-type siblings, the entire CMZ was labeled with BrdU with about three PH3+ cells scattered along the CMZ (Figures 5A,C,E,G). In the tubgcp3 mutants, there was a significantly increased number of PH3+ cells in CMZ and its adjacent regions (Figures 5B,F,G). We noticed that a portion of these PH3+ cells were BrdU negative ( Figure 5F). In contrast, there were no PH3+ BrdU-cells in wild-type siblings ( Figure 5E). In addition, BrdU+ cells were significantly reduced in mutant CMZ (Figures 5D,F,G). The PH3+ BrdU-cells in tubgcp3 mutants had passed S phase when we performed BrdU treatment, and they then went into mitosis as they were stained by PH3. However, these cells were still PH3+ after the 6 h BrdU treatment, which indicates that they were still in the mitotic (M) phase. Since G2 and M phase are shorter than 6 h, the results suggest that these RPCs were arrested in M-phase. Moreover, there were also more BrdU+ PH3+ cells in the tubgcp3 mutants than in wild-type siblings (Figures 5E,F,G), suggesting that some of the cycling RPCs were arrested in M-phase after they passed the S phase. When whole-mount immunostaining was carried out under the same BrdU treatment condition, we observed similar results that a large number of PH3+ BrdU-cells existed in the tubgcp3 mutant tails while they absent in wild-type siblings (Supplementary   Figures S5A,B). These results suggest that Tubgcp3 is required for RPCs to go through mitosis.
Given the mitotic arrested cells detected in the tubgcp3 mutant retinae, we analyzed the spindle formation by immunostaining using anti-α-tubulin and anti-PH3 antibodies. In wild-type siblings, bipolar mitotic spindles were formed in mitotic cells (Figures 5J,L). In contrast, abnormal monopolar spindles were observed in the mitotic arrested cells in tubgcp3 mutant retina (Figures 5K,M,N). The microtubules arrayed radially in the center and the condensed chromosomes located at the periphery in these cells (Figures 5I,K,M). These results suggest that depletion of Tubgcp3 causes a defect in mitotic spindle formation in RPCs, resulting in mitotic arrest in these cells.

Tubgcp3 Deficiency Impairs Centrioles Distribution
Since centriole duplication and segregation are essential for bipolar spindle formation during mitosis, we examined whether these processes were affected in the tubgcp3 mutant retinal cells using anti-centrin (a marker for centriole) and anti-PH3. In wild-type siblings, the mitotic cells had a pair of centrioles (two centrioles per centrosome) at each spindle pole (Figure 5O). In the tubgcp3 mutants, the M-phase arrested cells had four centriole dots like the wild-type siblings (Figures 5P,Q). However, the distribution of centrioles was abnormal. They were located at the center of the cell (Figure 5P) or randomly scattered in the cell ( Figure 5Q). These data suggest that the centrioles duplicate normally but fail to separate correctly in mitotic arrested RPCs in the tubgcp3 mutants.
Previous work has reported that Tubgcp3 interacts with γ-tubulin to form γ-TuRC complexes, which are located at the centrosome during mitosis (Pereira et al., 1998;Kollman et al., 2010; Figure 5U). We next detected whether the distribution of γ-tubulin was affected after Tubgcp3 depletion. Our immunostaining results showed that γ-tubulin located at each spindle pole during mitosis in wild-type siblings ( Figure 5R). In contrast, the distribution of γ-tubulin was abnormal in the mitotic arrested RPCs in tubgcp3 mutants. Most of the arrested cells that displayed γ-tubulin focus in the center of the cells (Figure 5S). Others showed scattered γ-tubulin staining patterns ( Figure 5T). In order to test whether the interaction between Tubgcp3 and γ-tubulin is conserved in zebrafish, we performed co-Immunoprecipitation (co-IP) assays using HEK293 cells by co-transfecting Myc-tagged zebrafish Tubgcp3 and GFP-tagged zebrafish γ-tubulin. Co-IP assays showed that Tubgcp3 binds to γ-tubulin through its C-terminal domain, which is consistent with previous studies (Figure 5V). Taken together, our data suggest that depletion of Tubgcp3 does not affect centriole duplication, but results in abnormal distribution of centrioles and γ-tubulin.

Tupgcp3-Deficient Cells Undergo Apoptosis and Senescence
Given our HE staining results showed disappearance of cells in the tubgcp3 mutant CMZ from 3 to 5 dpf (Figures 3A-H), we next investigated whether the mitotic arrest RPCs underwent apoptosis using TUNEL assay. There were almost no TUNEL+ cells in wild-type siblings (Figures 6C,E,G,I). In contrast, TUNEL+ cells were significantly increased in tubgcp3 mutant CMZ and the adjacent regions (Figures 6D,F,H,I) at 3 dpf. However, the TUNEL+ cells and PH3+ cells did not co-localize in the mutants (Figures 6F,H). This may be due to the fact that PH3 expression was absent in late-stage apoptotic cells (Huang et al., 2006). Next, we performed immunostaining using γ-H2AX, an early stage apoptotic marker (Rogakou et al., 2000), together with PH3. There was a significant increase in the number of γ-H2AX + cells in the tubgcp3 mutant retina at 3 dpf ( Figures 6M,O,Q,R), compared to wild-type siblings (Figures 6L,N,P,R). Moreover, some of the γ-H2AX+ cells overlapped with PH3+ cells (Figures 6O,Q). These results suggest that the M-phase arrested RPCs undergo apoptosis in the tubgcp3 mutants.
In addition, RSCs [marked by col15a1b (Pujic et al., 2006;Cerveny et al., 2010;Gonzalez-Nunez et al., 2010;Valdivia et al., 2016)] still could be detected at the most peripheral CMZ (close to the lens) in the tubgcp3 mutant retina at 5 dpf (Supplementary Figure S4A). Apoptosis signal was also absent in this region at 3 dpf (Figures 6H,Q) and 5 dpf (Supplementary Figure S6). We next want to detect cell proliferation at the peripheral CMZ. Embryos were incubated with BrdU for 24 h before being fixed at 5 dpf for the assay. In this region, BrdU+ cells were significantly decreased in the tubgcp3 mutants compared to wild-type siblings (Figures 6S-T",Y). As the proliferation of RSCs occurs at the peripheral CMZ, we speculated that RSC proliferation might be affected in the tubgcp3 mutant CMZ. Previous studies have reported that centrosome dysfunction caused by centrosomal protein depletion could induce cellular senescence (Manning and Kumar, 2010;Schmidt et al., 2010;Hossain and Tsang, 2013). Consistent with these reports, we found that the expression of senescence-associated β-galactosidase (SA-β-gal), a widely used senescent marker (Dimri et al., 1995), was significantly increased at the peripheral CMZ in the tubgcp3 mutant CMZ compared to the wild-type siblings (Figures 6U-X). These results indicate that depletion of Tubgcp3 caused cellular senescence at the peripheral CMZ that contains RSCs.

DISCUSSION
In this study, we generated the zebrafish tubgcp3 mutant and investigated the function of tubgcp3 in eye development. Our data reveal that Tubgcp3 plays important roles in cell cycle progression in the CMZ of the retina. Depletion of Tubgcp3 in RPCs results in monopolar spindle formation and impairs the distribution of centrioles and γ-tubulin, causing M-phase arrest and further apoptosis. In addition, depletion of Tubgcp3 results in cell proliferation defects and senescence at the peripheral CMZ. These defects led to the small eye phenotype in the tubgcp3 mutants. To our knowledge, this is the first in vivo model for the functional study of Tubgcp3. Our findings also provide some clues for studying the roles of its associated complexes, γ-TuSC and γ-TuRC, in vertebrate development.
γ-TuSC, composed of Tubgcp3, Tubgcp2, and γ-tubulin, is an essential complex for microtubule nucleation and spindle assembly during mitosis. Each component of γ-TuSC is likely essential for all cell proliferation in the developing zebrafish. Recently, Pouchucq et al. (2018) reported that morpholino (MO)-mediated tubg1/γ-tubulin knockdown caused zebrafish embryo development arrest at the mid-gastrula stage. The early developmental defects are most probably attributable to the fact that MO was designed against the translation start site of tubg1 mRNA, which blocks the translation of both the maternal and zygotic tubg1 mRNA (Pouchucq et al., 2018). Although cell cycle defects and increased apoptosis were observed after depletion of γ-tubulin (Pouchucq et al., 2018), study on the function of γ-tubulin in later stages of development has been limited. In contrast, the tubgcp3 mutants can survive longer than (R) Bar chart analyses depicting quantification of γ-H2AX+ and PH3+ cells in wild-type sibling and the tubgcp3 mutant retinae. Data are mean + SEM from 30 sections for each group. Student's t-test: * * P < 0.01. (S-T") ISH and immunostaining analysis of cell proliferation at the extreme periphery of CMZ at 5 dpf. Zebrafish embryos were incubated in BrdU for 24 h before collected at 5 dpf for the double staining assay. In the wild-type sibling, there are many BrdU+ cells in col15a1b-labeled region (S,S',S"). In contrast, BrdU+ cells are significantly decreased in this region in the tubgcp3 mutant CMZ (T,T',T"). (U-X) Senescence-associated β-galactosidase (SA-β-gal) staining exhibiting increased β-galactosidase activity at the CMZ in the tubgcp3 mutant (V) compared to the wild-type sibling (U). Nuclei are stained with DAPI (W,X). Insets indicate high-magnification images of the peripheral edge of CMZ in rectangles in (U-X). (Y) Bar chart analyses depicting quantification of BrdU+ cells in the col15a1b-labeled region in wild-type sibling and the tubgcp3 mutant CMZ. Data are mean + SEM from 36 sections for each group. Student's t-test: * * P < 0.01. Scale bars: 20 µm (A-H); 20 µm (J-Q); 20 µm (S-T"); 50 µm (U-X).
10 days due to the maternal deposition of tubgcp3 gene products.
The maternal effect provides a possibility to study its function at relatively late developmental stages. In the tubgcp3 mutant, the differentiated cells seem to be unaffected (Figures 3I-L). However, the CMZ exhibits obvious defects from 3 to 5 dpf (Figures 3E-H). This is associated with the relatively higher expression of tubgcp3 in CMZ compared to other retinal regions (Figures 1L,M). In CMZ, new neurons are produced for the continuous growth of zebrafish eyes after the embryogenesis of retina completed at 60 hpf. The tissue-specific defects in CMZ are in accordance with the rapid depletion of the maternal store in the regions of high proliferation. This also correlates with the function of γ-TuSC in cell proliferation. Despite the fact that many studies have been carried out among different organisms to understand the function of Tubgcp3 in γ-TuRC assembly and centrosome function (Geissler et al., 1996;Barbosa et al., 2000;Mikule et al., 2007;Xiong and Oakley, 2009;Farache et al., 2016;Cota et al., 2017), its roles in vertebrate development remain unclear. In vitro studies using human cell lines show that GCP3/Tubgcp3 is required for cell cycle progression (Mikule et al., 2007;Draberova et al., 2015;Cota et al., 2017). However, the cell cycle defects caused by GCP3 depletion are conflicting. Draberova et al. (2015) reported that the wild-type p53 U87MG cells and mutant p53 T98 cells arrested in M-phase after depletion of GCP3 using siRNA. GCP3-depleted HeLa cells also arrested in the M-phase (Cota et al., 2017). In the tubgcp3 mutants, we found that RPCs arrested in M-phase in the CMZ of the retina (Figures 5B,D,F,G), which is consistent with previous studies. Moreover, the mitotic arrest RPCs underwent apoptosis (Figures 6A-R). However, Mikule et al. (2007) found that some GCP3-depleted cell lines, such as RPE-1, BJ-1, HME-1, and HCT-116, did not arrest in M-phase, but arrested in G1 phase in a p53-dependent manner. In our study, we found that cell proliferation was affected at the peripheral CMZ in the tubgcp3 mutant retina. These cells might be RSCs, which failed to progress into S phase (Figures 6S-T",Y). They did not arrest in M-phase as PH3+ cells did not accumulate in the mutant peripheral CMZ at 5 dpf (Supplementary Figure S7). It is not feasible for us to isolate RSCs for fluorescence-activated cell sorter analysis (FACS) due to the absence of available transgenic zebrafish lines expressing RSC-specific reporters in our lab. Whether depletion of Tubgcp3 in RSCs causes G1 arrest still needs further study. To study whether p53 is involved in these processes will be an important extension to the functional study of Tubgcp3. It has been reported that centrosome dysfunction could induce cellular senescence (Manning and Kumar, 2010;Schmidt et al., 2010;Hossain and Tsang, 2013). Consistent with these reports, we observed that some cells underwent senescence at the extreme periphery of CMZ in tubgcp3 mutants. Since CMZ cells are involved in the growth of zebrafish eye, these defects together caused the small eye phenotype in the tubgcp3 mutants.
Autosomal recessive primary microcephaly (MCPH) is a neurodevelopmental disorder characterized by markedly reduced brain size. The patients exhibited significantly reduced number of neural progenitor cells caused by cell proliferation defects, increased cell death and the disruption of the balance between symmetric and asymmetric division (Barbelanne and Tsang, 2014;Buchwalter et al., 2016;O'Neill et al., 2018). Currently, at least nine centrosome-related genes have been genetically linked to microcephaly disorders (Megraw et al., 2011;Novorol et al., 2013;Chavali et al., 2014;Morris-Rosendahl and Kaindl, 2015;Buchwalter et al., 2016). Mutations in tubgcp4 and tubgcp6 have been reported to be associated with MCPH (Buchwalter et al., 2016). They encode TUBGCP4 and TUBGCP6, respectively, two components of γ-TuRC (Zheng et al., 1995). In addition, the MCPH models have been established using zebrafish retinal neuroepithelium by depletion of centrosomal proteins, including STIL, ASPM, WDR62, and ODF2 (Novorol et al., 2013). STIL localizes to the procentriolar cartwheel region and plays important roles in centriole duplication (Arquint and Nigg, 2014). ASPM localizes at the spindle poles and is involved in the organization of spindle poles (Tungadi et al., 2017). WDR62, which also localizes to the spindle poles, is required for mitotic entry of neural stem cells (Ramdas Nair et al., 2016). ODF2, a mother centriole subdistal appendage protein, is indispensable for the assembly of the mother centriole (Ishikawa et al., 2005). In these models, zebrafish embryos display the MCPH phenotype with significantly reduced head and eye size (Novorol et al., 2013). After knockout of tubgcp3 gene, the zebrafish embryos exhibit the similar MCPH phenotype (Figures 2E-G). M-phase cell cycle arrest (Figures 5A-G) and increased apoptosis are also observed (Figures 6A-R) in tubgcp3 mutant retinae, which is consistent with these MCPH models (Novorol et al., 2013). Moreover, cell proliferation defect and senescence were also observed at the peripheral CMZ (stem cell niche) (Figures 6S-V,Y). Senescent cells were also observed in the brain of the tubgcp3 mutants (Supplementary Figure  S8). Given the MCPH phenotype observed in tubgcp3 mutants (Figures 2E-G) and the requirement of Tubgcp3 in γ-TuRC assembly and centrosome function (Geissler et al., 1996;Barbosa et al., 2000;Mikule et al., 2007;Xiong and Oakley, 2009;Farache et al., 2016;Cota et al., 2017), we speculate that premature senescence may provide a possible new view to explain the mechanism of MCPH. Further studies to characterize the function of other centrosomal proteins will be necessary to fully understand the relationship between centrosome and microcephaly disorders.

ETHICS STATEMENT
This study was carried out according to the Guide for the Care and Use of Laboratory Animals from the National Institutes of Health. The experimental protocol was approved by the Animal Care and Use Committee of Fudan University.

AUTHOR CONTRIBUTIONS
TZ conceived and directed the project. GL carried out the experiments and discovered the roles of Tubgcp3 in cell cycle progression during zebrafish retinal development. TZ, DJ, and GL prepared the figures and wrote the manuscript.

FUNDING
This work was supported by grants from the National Science Foundation of China (31530044, 31671514, and 31471357).