Chemokine signaling links cell-cycle progression and cilia formation for left–right symmetry breaking

Zebrafish dorsal forerunner cells (DFCs) undergo vigorous proliferation during epiboly and then exit the cell cycle to generate Kupffer’s vesicle (KV), a ciliated organ necessary for establishing left–right (L–R) asymmetry. DFC proliferation defects are often accompanied by impaired cilia elongation in KV, but the functional and molecular interaction between cell-cycle progression and cilia formation remains unknown. Here, we show that chemokine receptor Cxcr4a is required for L–R laterality by controlling DFC proliferation and KV ciliogenesis. Functional analysis revealed that Cxcr4a accelerates G1/S transition in DFCs and stabilizes forkhead box j1a (Foxj1a), a master regulator of motile cilia, by stimulating Cyclin D1 expression through extracellular regulated MAP kinase (ERK) 1/2 signaling. Mechanistically, Cyclin D1–cyclin-dependent kinase (CDK) 4/6 drives G1/S transition during DFC proliferation and phosphorylates Foxj1a, thereby disrupting its association with proteasome 26S subunit, non-ATPase 4b (Psmd4b), a 19S regulatory subunit. This prevents the ubiquitin (Ub)-independent proteasomal degradation of Foxj1a. Our study uncovers a role for Cxcr4 signaling in L–R patterning and provides fundamental insights into the molecular linkage between cell-cycle progression and ciliogenesis.


Introduction
Vertebrates exhibit striking left-right (L-R) asymmetries in the structure and position of their cardiovascular and gastrointestinal systems. Initially, early embryos develop symmetrically along the prospective body midline. This embryonic symmetry is broken during somite stages when an asymmetric fluid flow is generated by motile cilia within the L-R organizer (LRO), a transient structure located at the posterior end of the notochord [1]. Specifically, in zebrafish, PLOS  the ciliated LRO is referred to as Kupffer's vesicle (KV), which forms from dorsal forerunner cells (DFCs), a group of superficial cells in the organizer region of the gastrula [2,3]. It has been well established that the architecture of KV cells and asymmetric KV cilia generate a counterclockwise nodal flow. This leads to the asymmetrical expression of early laterality genes, including nodal-related southpaw (spaw) and paired-like homeodomain 2c (pitx2c) in the left lateral plate mesoderm (LPM), and ultimately the establishment of L-R asymmetric patterning [4]. The origin of L-R asymmetry is conserved across many vertebrates, and defects in the establishment of these asymmetries can result in a broad spectrum of birth defects, often including congenital heart malformations [5,6]. The progression of cells through the G1 and S phases of the cell cycle is tightly controlled by the sequential activation of a family of serine-threonine kinases known as the cyclin-dependent kinases (CDKs). CDK4 and its homolog CDK6 are activated by D-type cyclins in early to mid-G1 phase, whereas CDK2 is activated by E-and A-type cyclins during the late G1 and S phases, respectively [7]. Recent evidence indicates that cell-cycle dynamics have emerged as a key regulator of stem cell fate decisions [8][9][10]. Specifically, Cyclin D proteins have been shown to activate CDK4/6, which restricts the activity of Sma-and Mad-related Protein (Smad) 2/3 in late G1 phase and results in a switch from endoderm to neuroectoderm potential in human pluripotent stem cells [11]. The G1 cyclin proteins together with their associated CDKs also play essential, direct roles in the maintenance of cell stemness and in the regulation of cell fate specification in mouse embryonic stem (ES) cells by phosphorylation and stabilization of the core pluripotency factors, Nanog homeobox (Nanog), SRY-box transcription factor (Sox) 2, and Octamer-binding protein 4 (Oct4) [12]. In addition, CDK4/Cyclin D1 overexpression has been shown to prevent G1 lengthening and functions to inhibit neurogenesis in mouse embryos [13]. In zebrafish, DFCs vigorously proliferate and collectively migrate towards the vegetal pole during epiboly stages. They then cluster and differentiate into polarized epithelial cells of KV [4,14,15]. Interestingly, DFC proliferation defects are often accompanied by impaired cilia elongation in KV [16][17][18], indicating a possible connection between cell-cycle events and KV cilia formation. However, the underlying mechanism remains poorly understood.
Foxj1, a forkhead-domain-containing transcription factor that is expressed in various ciliated tissues, has been associated with motile cilia formation and L-R axis development in mammals [19,20]. In zebrafish, two foxj1 paralogs have been identified, including forkhead box j1a (foxj1a) and foxj1b [21]. foxj1a has been shown to be highly expressed in the DFCs toward the end of gastrulation and plays a primary role in KV ciliogenesis, while foxj1b is expressed in the otic vesicle, where it has been shown to regulate motile cilia formation [21,22]. The expression level of foxj1a transcripts has been shown to be regulated by the Hedgehog, wingless/integrated (Wnt)/β-catenin, and fibroblast growth factor (FGF) signaling pathways [17,21,23]. Ligand of numb-protein X 2b (Lnx2b)-a Really Interesting New Gene (RING) domain containing E3 ubiquitin (Ub) ligase, which is specifically expressed in the migratory DFCs and developing KV-plays a critical role in the establishment of L-R laterality. This indicates the involvement of protein ubiquitination in the determination of L-R asymmetry [24]. However, whether the function of Foxj1 protein in KV ciliogenesis is regulated by Ub modification remains unknown.
Chemokines are small (8)(9)(10)(11)(12)(13)(14), vertebrate-specific proteins that can be categorized into four subgroups according to the presence and position of conserved cysteine residues (C, CC, CXC, and CX3C) [25]. Among chemokines of the CXC class, the stromal-cell-derived factor 1 (SDF-1/CXCL12) and its receptor CXCR4, which were first identified because of their primary role in leukocyte homing, have been implicated in the regulation of cell adhesion and migration during embryonic development [25][26][27]. Interestingly, in zebrafish, two Cxcl12 ligands and two Cxcr4 receptors were found to be expressed across a wide range of cell types and developmental stages and were found to act as discrete pairs to direct cell migration [26]. Cxcl12a-Cxcr4b signaling controls processes such as the directional migration of primordial germ cells, the collective migration of the lateral line primordium, and the formation of the trunk lymphatic network [28][29][30]. On the other hand, the Cxcl12b-Cxcr4a axis has been shown to play a role in endodermal morphogenesis, vascular system patterning, and the migration and prechondrogenic condensation of cranial neural crest cells [31][32][33][34]. It has been shown previously that cxcr4a and cxcr4b possess mutually exclusive expression patterns in the majority of cell lineages [35]. For example, cxcr4a, but not cxcr4b, is expressed in the primordium of KV [35]. While cxcr4b expression reveals an asymmetric pattern in habenular neurogenesis, the cxcr4b mutant odysseus displays no obvious phenotype in L-R epithalamic asymmetry [36]. These observations bring into question whether the signaling cascades initiated by Cxcl12b and Cxcr4a play a role in the establishment of L-R asymmetry.
Here, we provide evidence suggesting that the Cxcl12b-Cxcr4a axis is essential for L-R asymmetric development. Cxcr4a um20 mutants were found to exhibit poor DFC proliferation and abnormal KV cilia formation. Specifically, depletion of cxcr4a in DFCs was found to lead to a significant decrease in ERK1/2 signal activation, which was essential for the expression of cyclin D1. Subsequent biochemical and functional approaches demonstrated that Cyclin D1-CDK4/6 functions to accelerate the G1/S transition to promote DFC proliferation and stabilize Foxj1a for cilia formation. Mechanistically, CDK4 phosphorylates Foxj1a at T102 and then disrupts its association with Psmd4b, which in turn prevents the Ub-independent proteasomal degradation of Foxj1a protein. Therefore, Cxcl12b-Cxcr4a chemokine signaling links cellcycle progression and cilia formation for L-R symmetry breaking via regulating Cyclin D1 expression.

Cxcl12b-Cxcr4a axis is required for L-R laterality
cxcr4a was previously found to be expressed in KV progenitors at the end of gastrulation [35]. To address the detailed expression patterns of cxcr4a, whole-mount in situ hybridization (WISH) was carried out during early zebrafish embryogenesis. As shown in S1A Fig, cxcr4a expression was observed in endoderm cells and migrating DFCs throughout gastrulation. At early somite stages, concomitant with the onset of anterior neural plate expression, cxcr4a was also found to be activated in the developing KV cells besides the central lumen (S1B Fig). Therefore, we hypothesized that cxcr4a might play a critical role in KV organogenesis and L-R asymmetric patterning.
We then set out to test whether cxcr4a is required for L-R development. cxcr4a um20 mutant carries a null mutation in the cxcr4a gene, which causes an in-frame 5-amino-acid deletion within the second transmembrane domain of the Cxcr4a receptor that eliminates its membrane localization [34]. To generate homozygous cxcr4a mutants, about 200 of the offspring from heterozygous cxcr4a um20 parents were bred and maintained to sexual maturity, but only a few of homozygous mutants were identified, suggesting most of them could not survive up to adulthood. As previously reported, the maternal-zygotic cxcr4a (MZcxcr4a) mutant embryos generated by incrossing homozygous cxcr4a um20 adults exhibited distinct defects in lateral dorsal aorta formation (S1C Fig) [34]. Hereinafter, if not specially mentioned, the terms cxcr4a um20 mutant or cxcr4a um20 embryo refers to the MZcxcr4a mutant. In order to analyze the laterality information in cxcr4a um20 mutants, we examined cardiac development by WISH against cardiac myosin light chain 2 (cmlc2). At 48 hours postfertilization (hpf), the majority of wild-type embryos showed a heart tube looping to the right (D-loop) (Fig 1A and 1B). Cxcl12b-Cxcr4a signaling axis is essential for L-R asymmetric development. (A-D) WT embryos and cxcr4a um20 mutants injected with or without 300 pg cxcr4a mRNA at the 256-cell stage were examined for cardiac looping and liver laterality at 48 hpf by WISH against cmlc2 (A) and hhex (C). Embryos with different phenotypes are shown in ventral (A) or dorsal view (C), and the ratios are shown in (B) and (D). Underlying data can be found in S1 Data. (E-H) Embryo ratios with different expression patterns of cmlc2 and hhex at 48 hpf in WT embryos injected with 8 ng cxcr4a MO (4a MO) at the 256-cell stage (E and F) and cxcl12b mutants (G and H). Underlying data can be found in S1 Data. (I-N) cxcr4a deficiency alters Nodal gene expression pattern. Representative images of spaw and pitx2c expression in cxcr4a mutants (I and M) and morphants (K). All embryos are shown in dorsal views with anterior on the top. Ratios of embryos are shown in (J), (L), and (N). Underlying data can be found in S1 Data. However, the heart localization was mildly affected in cxcr4a um20 mutants, of which 21% showed a "no-looping" or reversed "left-looping" heart (L-loop) (Fig 1A and 1B). Moreover, about 30% of cxcr4a um20 embryos also exhibited disrupted liver position, as revealed by hematopoietically expressed homeobox (hhex) expression (Fig 1C and 1D). In addition, these L-R laterality defects were well rescued by injection of 300 pg cxcr4a mRNA into embryonic yolk at the midblastula stage (256-cell stage) (Fig 1B and 1D; S3A and S3B Fig), which would lead to a specific overexpression of cxcr4a in DFC/KV cells [37]. Conversely, we did not observe any rescue effect with the overexpression of cxcr4a um20 , which contains the 5-amino-acid in-frame deletion (S2A and S2B Fig). Thus, cxcr4a is important for organ laterality in zebrafish embryos. Interestingly, when MZcxcr4a mutant adult females were mated to wild-type male fish, the resultant embryos (maternal cxcr4a mutants) displayed normal asymmetry of the heart and liver (S3A and S3B Fig). In contrast, zygotic cxcr4a um20 homozygous mutants, which were identified from cxcr4a um20 heterozygous fish crosses by genotyping, had similar L-R defects as MZcxcr4a embryos (S3C and S3D Fig). Therefore, these data suggest that the zygotic, but not maternal, cxcr4a function is required for L-R asymmetric development.
Because cxcr4a depletion would impair endoderm cell migration during gastrulation and cause bilateral duplication of endodermal organs such as the liver [31], we injected a previously validated morpholino (MO) that targets cxcr4a (4a MO) at the midblastula stage to specifically block cxcr4a activity in DFC/KV cells. In comparison to injection with a standard control MO (cMO), injection of 4a MO led to similar laterality abnormalities as observed in cxcr4a um20 mutants (Fig 1E and 1F). This suggests that the organ localization defects are not secondary effects of impaired endoderm migration. Therefore, the cxcr4a expression in DFC/KV cells is required for L-R laterality. In addition, the deficiency of cxcl12b in cxcl12b mu100 mutants [38] was also found to result in laterality defects (Fig 1G and 1H), indicating that the Cxcl12b-Cxcr4a signaling pathway is critical for L-R symmetry breaking.
Because organ laterality is regulated by evolutionally conserved asymmetric L-R gene expression in vertebrates, we next examined the expression patterns of spaw and its downstream gene pitx2c [1,39]. At the late somite stages, we observed spaw and pitx2c expression in the left LPM in wild-type embryos, whereas expression of these genes was found to be bilateral or absent in cxcr4a um20 mutants and DFC 4a MO embryos (Fig 1I-1N). Interestingly, the bilateral expression domain of spaw in a subset of cxcr4a um20 mutants was located in the more posterior region in the LPM (Fig 1I), indicating a delay in the anterior spreading of spaw expression.
Collectively, these results demonstrate a sustaining expression of cxcr4a in the DFC/KV cells and implicate a crucial role of Cxcl12b-Cxcr4a chemokine signaling in L-R laterality determination.

Ablation of cxcr4a compromises KV organogenesis and ciliogenesis
To determine whether a loss of cxcr4a alters KV morphogenesis, we first examined the formation of DFC clusters during gastrulation in cxcr4a um20 mutants carrying the transgenic DFC/KV reporter sox17:green fluorescent protein (GFP). We observed that in both wild-type embryos and cxcr4a um20 mutants, the GFP-positive DFCs were maintained as a cohesive group and migrated towards the vegetal pole during mid to late gastrulation (S4A Fig). Meanwhile, in comparison to control embryos, cxcr4a um20 mutants exhibited a normal expression pattern of sox17 transcripts in the DFC clusters (S4B Fig). Based on these observations, we concluded that cxcr4a is unnecessary for the specification, clustering, and collective migration of DFCs.
We then examined the morphology of KV in live embryos at the 10-somite stage, at which point KV was well formed [4]. Wild-type embryos exhibited a normal button-like KV at the terminus of the notochord, as observed under bright-field microscopy (Fig 2A). In contrast, a large majority of cxcr4a um20 mutants displayed a smaller or tiny/absent KV (Fig 2A and 2B). Consistently, similar defects in KV morphology were found in cxcl12b mu100 mutants (Fig 2A  and 2B). In order to monitor the dynamic changes during KV formation, we carried out in vivo time-lapse image analysis on cxcr4a-deficient transgene (Tg) sox17:GFP embryos from the 1-to 6-somite stages. DFCs were found to have already rearranged into a single rosette concurrent with the formation of the preliminary lumen in 1-somite-stage wild-type and cxcr4a um20 mutant embryos ( Fig 2C). However, the GFP-positive KV appeared to be dramatically smaller in cxcr4a um20 embryos in comparison to control animals from the 3-to 6-somite stages ( Fig 2C). We next looked into the apical-basal polarity of KV epithelial cells, which is critical for the correct establishment of L-R asymmetry [40]. Immunostaining experiments revealed that the distributions of the basal-lateral marker E-cadherin and the apical marker Light micrographs at the 10-somite stage showed smaller or even absent KVs in cxcr4a um20 or cxcl12b mu100 mutants. Scale bar, 200 μm. Embryo ratios with different KV sizes were shown in (B). Underlying data can be found in S1 Data. (C) Time-lapse confocal images from the 1-somite stage to the 6-somite stage showed the dynamic formation of KV in WT and cxcr4a-deficient Tg(sox17:GFP) embryos. Scale bar, 20 μm. The ratios of affected embryos were indicated. (D-F) Fluorescent immunostaining of KV using anti-GFP and anti-α-Tubulin antibodies at the 10-somite stage in WT embryos and cxcr4a um20 mutants. The boxed areas in the images were presented at higher magnification in the relevant insets. Scale bar, 20 μm. Cilia average number and length were quantified from three independent experiments, and the group values were expressed as the mean ± SD (E and F). Student t test, � P < 0.05, �� P < 0.01. Underlying data can be found in S1 Data. (G) Fluorescent bead tracking experiments showed that fluorescent beads moved in a persistent counterclockwise fashion in WT embryos but had no directional flow in cxcr4a um20 mutants. White spots, yellow spots, and curves mark the start points, the end points, and the tracks of bead movements, respectively. GFP, green fluorescent protein; KV, Kupffer's vesicle; sox, SRY-box transcription factor; Tg, transgene; WT, wild-type; α-Tubulin, acetylated tubulin. atypical protein kinase C (aPKC) in KV epithelial cells of 10-somite-stage cxcr4a um20 embryos were correct (S4C and S4D Fig). These observations suggest that cxcr4a is critical for organ size control but is not required for epithelial cell polarization during KV organogenesis.
Monocilia in KV are known to generate a counterclockwise fluid flow, which creates the asymmetrical signals required to break L-R symmetry [4]. We next analyzed KV cilia formation by probing acetylated tubulin (α-Tubulin) in the cxcr4a um20 embryos at the 10-somite stage. We found that in comparison with control embryos, cxcr4a um20 mutants exhibited a significant decrease in cilia number and a steady reduction in cilia length (Fig 2D-2F). As expected, cxcl12b mu100 mutant embryos showed similar abnormalities in ciliogenesis (S5A- S5C Fig). We then sought to determine whether the KV directional fluid flow was altered in cxcr4a um20 mutants. Fluorescent beads were injected into KVs at the 6-somite stage, and the movements of the beads were tracked at the 10-somite stage. The fluorescent beads moved in a persistent counterclockwise fashion in wild-type embryos, whereas they exhibited no directional flow in cxcr4a-deficient embryos (Fig 2G; S1 and S2 Movies). Interestingly, cxcr4a um20 mutants exhibited no obvious defects in the KV cilia motility as revealed by high-speed video microscopy (S3 and S4 Movies), indicating that the impaired fluid flow in KVs is primarily resulted from the reduction in the number and length of motile cilia. Therefore, these results indicated that cxcr4a is indispensable for KV ciliogenesis and cilia-driven fluid flow.

Absence of cxcr4a attenuates G1/S transition and zFoxj1a protein expression
Our studies suggest that cxcr4a deficiency leads to smaller KV size as well as fewer KV cilia. Interestingly, the majority of the KV epithelial cells in cxcr4a mutants exhibited an intact apical-basal polarity and formed notably shortened cilia (Fig 2D; S4C and S4D Fig). This suggests that the altered KV cilia numbers may be caused by defects in cell proliferation. To address this issue, we first examined the proliferation profile of DFC/KV cells by performing bromodeoxyuridine (BrdU) incorporation assays in Tg(sox17:GFP) embryos during gastrulation and early somite stages. Consistent with previous reports [16], approximately 70% of DFCs were positively stained with BrdU at the midgastrulation stage, whereas very few BrdU-positive cells were observed in the developing KV at the bud and the 6-somite stages (Fig 3A and 3B), suggesting that vigorous proliferation occurs in DFCs during epiboly stages and then declines at the end of gastrulation. Impressively, we found a dramatic decrease of the BrdU-positive DFC number in cxcr4a-deficient embryos at midgastrulation stage ( Fig 3C and 3D), indicating a crucial requirement of cxcr4a in DFC proliferation.
We next examined the detailed effects of cxcr4a deficiency on DFC cycle progression in Tg (sox17:GFP;EF1α:mKO2-zCdt1(1/190)) double transgenic embryos, a model in which cells in G1 phase exhibit red nuclear fluorescence [41]. Because the G1/S transition is very short in cells that undergo rapid mitotic cycles [41], we were unable to identify any mKO2-zCdt1-positive DFCs in both control embryos and cxcr4a um20 mutants during gastrulation. Interestingly, while the majority of cells in the developing KV enter into a quiescent state at the end of gastrulation ( Fig 3A and 3B), we observed no or only few KV progenitors with mKO2-zCdt1 fluorescence in wild-type embryos at the bud and 3-somite stages (Fig 3E-3H). These results, combined with the previous observation that the mKO2-zCdt1 signal was highlighted in differentiated cells-including postmitotic neurons and muscle cells from the 10-somite stage [41]-imply that there exist limitations in the ability to dissect cell-cycle behavior using this fluorescent indicator in early zebrafish embryos. However, cxcr4a um20 mutant embryos showed a marked proportion of KV progenitors with robust mKO2-zCdt1 expression, indicating impaired G1/S transition and an apparent lengthening of the G1 phase (Fig 3E-3H). . Underlying data can be found in S1 Data. (C-D) BrdU incorporation experiments showed reduced proliferating DFCs in cxcr4a um20 mutants at the 75% epiboly stage. Scale bar, 20 μm. Statistical data from three independent experiments are shown in (D). Student t test, ��� P < 0.001. Underlying data can be found in S1 Data. (E-I) Depletion of cxcr4a inhibits the G1/S transition in DFCs. Representative confocal sections of WT and cxcr4a-deficient Tg(sox17:GFP;EF1α:mKO2-zCdt1 (1/190)) embryos at the bud and 3-somite stages are shown in (E) and (G). Dorsal views with anterior to the top. Scale bar, 20 μm. The percentage of mKO2-positive KV cells were quantified from three independent experiments (F and H). The average numbers of KV cells in WT embryos and cxcr4a um20 mutants at the 3-somite stage are shown in (I). The significance of differences compared with the WT group was analyzed with the Student t test, �� p < 0.01; ��� p < 0.001. Underlying data can be found in S1 Data. (J-K) cxcr4a deficiency down-regulates zFoxj1a protein expression levels. WT and cxcr4a-deficient Tg(sox17:GFP) embryos were harvested at the 75% epiboly and bud stages and then subjected to western blot analysis (J) and immunostaining (K) with the indicated antibodies. Scale bar, 20 μm. BrdU, bromodeoxyuridine; DFC, dorsal forerunner cell; Ef1α, eukaryotic translation elongation factor 1α; ep, epiboly; Foxj1a, forkhead box j1a; GFP, green fluorescent protein; KV, Kupffer's vesicle; mKO2, monomeric Kusabira Orange2; sox, SRY-box transcription factor; Tg, transgene; WT, wild-type; zCdt1, zebrafish chromatin licensing and DNA replication factor 1; zFoxj1a, zebrafish Foxj1a; α-Tubulin, acetylated tubulin.
When cxcr4a in DFCs was knocked down, similar defects in cell-cycle progression were found (S6A- S6D Fig). Besides, consistent with the smaller KVs in cxcr4a-deficient embryos during early somite stages (Fig 2C), we found that the number of KV cells declined about 50% in cxcr4a um20 mutants at the 3-somite stage compared with that of wild-type controls (Fig 3G and  3I). Nevertheless, our results suggest that Cxcr4a-mediated chemokine signaling is responsible for driving DFC proliferation by accelerating the G1/S transition. This provides an explanation for the smaller KV size that was observed in cxcr4a um20 mutants.
Because zebrafish Foxj1a (zFoxj1a) is a master regulator of KV ciliogenesis [21,22], we then sought to determine whether zFoxj1a expression was affected in cxcr4a um20 mutants. In situ hybridization analysis demonstrated normal expression levels of foxj1a transcripts in cxcr4a um20 embryos during gastrulation (S7 Fig). To examine the changes of zFoxj1a protein expression in cxcr4a um20 embryos, we first validated the reliability of an antibody against human FOXJ1 in foxj1a morphants. Western blotting revealed a specific protein band corresponding to zFoxj1a protein (about 70 kDa) in the lysates from 75%-epiboly-stage embryos injected with cMO, which was drastically reduced in the lysates from foxj1a morphants (S8A  (Fig 3J and 3K). These analyses provide strong evidence that Cxcr4a signaling is responsible for controlling KV ciliogenesis through regulation of zFoxj1a protein expression at the post-transcription level.
Among the cell-cycle-regulatory genes, Cyclin D1 expression is known to be specifically activated by the Cxcr4a-ERK1/2 cascade to promote cell proliferation [42,43,46]. Consistent with these previous studies, double fluorescence in situ hybridization analyses indicated a dramatic reduction in cyclin D1 expression in cxcr4a-deficient DFCs (Fig 4F). In addition, the introduction of caMEK mRNA counteracted the cxcr4a depletion effects on cyclin D1 expression ( Fig 4F). Cell-cycle progression from G1 to the S phase is governed by CDK4/6 and CDK2, which are activated by D-type and E-or A-type cyclins, respectively [7]. To investigate the connections between cell-cycle progression and organ laterality, Tg(mKO2-zCdt1(1/190)) embryos were treated with a selective CDK4/6 inhibitor PD0332991 and a specific CDK2 (A) ERK1/2 phosphorylation levels were dramatically decreased in cxcr4a um20 mutants. WT and cxcr4a-deficient Tg(sox17:GFP) embryos were harvested at the 75% epiboly stage and subjected to immunostaining for p-ERK1/2 (red) and GFP (green). All embryos are shown in dorsal views with anterior to the top. Scale bar, 20 μm. (B-E) caMEK mRNA overexpression in DFCs rescued L-R patterning defects in cxcr4a um20 mutants. Different types of heart looping and liver laterality at 48 hpf in cxcr4a um20 mutants following midblastula injection of different caMEK mRNA doses were visualized by cmlc2 and hhex expression (B and C). The ratios are shown in (D) and (E). Underlying data can be found in S1 Data. (F) Cxcr4a-deficient Tg(sox17:GFP) embryos were injected with 60 pg caMEK mRNA at the 256-cell stage and then harvested at the 75% epiboly stage for fluorescence in situ hybridization experiments with cyclin D1 (red) and GFP (green) probes. Dorsal views with anterior to the left. Scale bar, 20 μm. (G-H) WT embryos were treated with 0.5 μM PD0332991 or 0.2 μM CY202 from the shield stage to bud stage and then analyzed for L-R patterning defects at 48 hpf by in situ hybridizations with cmlc2 and hhex probes. The proportion of treated embryos exhibiting each type of heart looping and liver laterality is shown in (G) and (H). Underlying data can be found in inhibitor CYC202 from the shield stage [47,48]. Both PD0332991 and CYC202 treatments induced a remarkable increase of the number of cells stagnated at G1 stage in the resulting embryos, indicating the high efficiency of these chemicals in the zebrafish embryos (S10 Fig). Interestingly, upon exposure of wild-type embryos to PD0332991 from the shield stage to the bud stage, the resulting animals exhibited L-R defects similar to those observed in cxcr4a um20 mutants (Fig 4G and 4H). At the same time, embryos treated with CYC202 showed no significant changes in laterality development (Fig 4G and 4H). These results suggest that Cyclin D1-CDK4/6 complexes play critical roles in DFC proliferation and L-R asymmetric development. Importantly, reintroduction of cyclin D1 into DFCs was found to relieve the inhibition of cell proliferation, the reduction of KV size, and the defects of organ laterality in cxcr4a um20 mutants (Fig 4I-4M). Collectively, our data indicate that the Cxcr4a-ERK1/2 cascade functions in DFC proliferation through regulation of Cyclin D1 expression during zebrafish L-R development.

CDK4 and its kinase activity is required for Foxj1 protein stabilization
Because the injection of cyclin D1 mRNA into DFCs was found to rescue the L-R defects in cxcr4a um20 mutants (Fig 4L and 4M), we hypothesized that Cyclin D1 acts downstream of Cxcr4a signaling to control both cell proliferation and cilia formation. In support of this hypothesis, upon injection of 300 pg cyclin D1 mRNA into midblastula-stage cxcr4a um20 embryos, we observed that both the number and the length of KV cilia were restored (Fig 5A-5C), indicating a role of Cyclin D1 in ciliogenesis. It is surprising that the DFC-specific delivery of cyclin D1 mRNA was also able to restore the expression of endogenous zFoxj1a protein in cxcr4a um20 embryos ( Fig 5D).
We next aimed to understand whether Foxj1 protein stability is regulated by Cyclin D1-CDK4/6 complexes. As depicted in Fig 5E, Cyclin D1 or CDK4 overexpression in HEK293T cells notably increased zFoxj1a expression. This increase in exogenously expressed zFoxj1a was even more apparent when Cyclin D1-CDK4 complexes were ectopically expressed ( Fig  5E). We noted that Cyclin D1-CDK4 complexes also showed similar effects on the expression levels of mouse Foxj1 (mFoxj1) (Fig 5F). In combination with our observation of unchanged foxj1a transcript expression in cxcr4a um20 mutants, these results suggest that Cyclin D1-CDK4 complexes may play crucial roles in the prevention of Foxj1a protein degradation. Indeed, we observed that the treatment of MG132, a proteasome inhibitor, but not NH 4 Cl, a lysosome inhibitor, or 3-methyladenine (3-MA), a well-characterized inhibitor of autophagy, dramatically stabilized zFoxj1a protein (Fig 5G). In contrast, blocking endogenous CDK activity with PD0332991 treatment was found to result in a clear reduction of zFoxj1a expression (Fig 5H). In addition, the PD0332991-treatment-induced zFoxj1a turnover was completely suppressed by cotreatment with MG132 ( Fig 5H). Therefore, Cyclin D1-CDK4 complexes contribute to Foxj1 protein stabilization. To further determine whether CDK4 kinase activity is important for Foxj1 protein stability, we generated a S1 Data. (I-J) Reintroduction of cyclin D1 into DFCs relieves DFC proliferation defects in cxcr4a um20 mutants. Cxcr4a-deficient Tg(sox17:GFP) embryos were injected with or without 300 pg cyclin D1 mRNA at the 256-cell stage, followed by coimmunostaining with anti-BrdU (red) and anti-GFP (green) antibodies at the 75% epiboly stage. Representative images are shown in (I), and the percentage of BrdU-positive DFCs is indicated in (J). Scale bar, 20 μm. Student t test, ��� P < 0.001. Underlying data can be found in S1 Data. (K-M) The reduced ratios of embryos affected in KV size (K) and cmlc2 (L) or hhex (M) expression show that DFC-specific overexpression of cyclin D1 rescued the defects of KV formation (K) and L-R patterning (L and M) in cxcr4a mutants. Underlying data can be found in S1 Data. BrdU, bromodeoxyuridine; caMEK, constitutively activated version of MEK; cmlc2, cardiac myosin light chain 2; DFC, dorsal forerunner cell; ERK, extracellular regulated MAP kinase; GFP, green fluorescent protein; hhex, hematopoietically expressed homeobox; hpf, hours postfertilization; KV, Kupffer's vesicle; L-R, left-right; ns, no significant difference; p-ERK1/2, phosphorylated ERK1/2; sox, SRY-box transcription factor; Tg, transgene; WT, wild-type. https://doi.org/10.1371/journal.pbio.3000203.g004 Cxcr4a guides cell proliferation and ciliogenesis zebrafish kinase-deficient mutant denoted CDK4-K38M, in which the ATP-binding site (Lys-38) in the catalytic subunit was mutated to a methionine residue [49]. As shown in Fig 5I, ectopic expression of CDK4-K38M had no effect on the zFoxj1a expression, and PD0332991 treatment eliminated CDK4-mediated protein stabilization. These data suggest that CDK4 stabilizes Foxj1 through its kinase activity. . KV cells were labeled with antibodies against GFP (green), and cilia were visualized by α-Tubulin immunofluorescence (red). The boxed areas in the images are presented at higher magnification in the relevant insets. Scale bar, 20 μm. Cilia number (B) and length (C) were analyzed. Student t test, � P < 0.05, �� P < 0.01. Note that when cyclin D1 mRNA was injected into cxcr4a-deficient DFCs, both the KV cilia number and length were restored. Underlying data can be found in S1 Data. (D) Western blots of total lysates from 75%-epiboly-stage WT embryos and cxcr4a um20 mutants injected with or without 300 pg HA-cyclin D1 mRNA at the 256-cell stage. Tubulin was used as the loading control. Endogenous zFoxj1a protein levels in cxcr4a um20 mutants was rescued by DFC-specific overexpression of HA-cyclin D1. Cxcr4a guides cell proliferation and ciliogenesis

CDK4 physically interacts with and directly phosphorylates Foxj1
To understand whether Foxj1 is a substrate of CDK4 kinase, we first examined the possible interaction between these two proteins. HeLa cells were transiently transfected with Flagtagged zFoxj1a and Myc-tagged CDK4. Immunofluorescence staining experiments revealed a colocalization of overexpressed zFoxj1a and CDK4 in the nuclear aggregates (Fig 6A), suggesting a potential interaction between these two proteins. Because zFoxj1a was observed to be localized exclusively to the nucleus even when coexpressed with CDK4 (Fig 6A), we excluded the effects of CDK4 on the subcellular distribution of zFoxj1a. We next examined the association between zFoxj1 and CDK4 in wild-type embryos and HEK293T cells. Coimmunoprecipitation experiments demonstrated that overexpressed CDK4 interacted with endogenous and ectopically expressed zFoxj1a (Fig 6B and 6C). Interestingly, the kinase-deficient form of CDK4 failed to associate with zFoxj1a ( Fig 6C). In order to test whether CDK4 interacts directly with Foxj1 protein, we carried out an in vitro binding assay using purified proteins. As depicted in Fig 6D, Myc-CDK4 protein purified from bacterial cells was able to bind to glutathione S-transferase (GST)-zFoxj1a but not GST proteins. Collectively, these results demonstrate that CDK4 interacts directly with zFoxj1a protein.
It is well established that the CDK families of serine/threonine protein kinases phosphorylate substrates containing the consensus amino-acid sequence (S/T)PXR/K [50]. Because Cyclin D1-CDK4 complexes contribute to the stabilization of both mouse and zebrafish Foxj1 (Fig 5E and 5F), we hypothesized that CDK4 is involved in the phosphorylation of Foxj1 within conserved classic substrate motifs. Interestingly, we found that zFoxj1a contains a potential CDK phosphorylation motif "TPGK" at the N-terminal region, which is highly conserved in vertebrates (Fig 6E). To investigate whether CDK4 phosphorylates Foxj1 protein, a phospho-threonine-proline antibody was used to enrich CDK substrates from whole-cell lysates, and the presence of phosphorylated Foxj1 was examined by western blot. With these experiments, we found that CDK4 or Cyclin D1-CDK4 complexes could effectively phosphorylate zFoxj1a and mFoxj1 (Fig 6F; S11A Fig). As expected, CDK4-K38M almost lost the ability to induce zFoxj1a phosphorylation (Fig 6F). These results clearly indicate that Foxj1 can be phosphorylated by CDK4.
We then aimed to determine whether the threonine 102 residue (T102) within the putative conserved substrate motif of zFoxj1a is a major CDK phosphorylation site. Excitingly, we observed that Cyclin D1-CDK4 complexes significantly promoted wild-type zFoxj1a phosphorylation, which was nearly abolished in T102 mutant, an unphosphorylated form of zFoxj1a (Fig 6G). Similarly, in vitro phosphorylation assays showed that purified CDK4 or Cyclin D1-CDK4 complexes resulted in distinct phosphorylation events when incubated with recombinant wild-type zFoxj1a protein but not T102 mutant (Fig 6H). In addition, we observed an increased expression of wild-type zFoxj1a, but not its T102 mutant, in HEK293T cells upon coexpression with CDK4 (Fig 6I). Similarly, CDK4 was able to control mFoxj1 stabilization through phosphorylation of the T87 residue, located in the conserved substrate motif (S11B and S11C Fig). Taken together, we showed that CDK4 directly phosphorylates Foxj1 to suppress its degradation.

zFoxj1a undergoes Ub-independent proteasomal degradation via a direct interaction with Psmd4b
Previous studies have suggested a primary role of the Ub-proteasome system in the elimination of abnormal proteins and selective destruction of regulatory proteins [51,52]. To explore whether ubiquitination is required for Foxj1 degradation, we examined the effect of Ub K48R/ G76A overexpression on zFoxj1a degradation. If zFoxj1a degradation was determined to be phosphorylates zFoxj1a at T102. HEK293T cells were transfected with the indicated plasmids. CDK substrates were immunoprecipitated using a phospho-threonine-proline antibody and blotted with anti-Flag antibody to detect phosphorylated zFoxj1a or zFoxj1a-T102A. zFoxj1a-T102A is an unphosphorylated form of zFoxj1a. Note that WT zFoxj1a could be phosphorylated by CDK4 but not by the CDK4-K38M mutant (F). CDK4-mediated phosphorylation was nearly abolished in zFoxj1a-T102A (G). (H) In vitro kinase assays revealed that CDK4 phosphorylates WT zFoxj1a but not the zFoxj1a-T102A mutant. zFoxj1a and zFoxj1a-T102A proteins were purified from bacterial cells and incubated with recombinant Cyclin D1 and CDK4 proteins in the presence or absence of ATP. Phosphorylation of zFoxj1a and zFoxj1a-T102A was detected by western blot using a phospho-threonineproline antibody, and input proteins were examined by Coomassie blue staining. (I) Ectopic CDK4 expression is unable to stabilize zFoxj1a-T102 mutant. Lysates from HEK293T cells transfected with the indicated plasmids were subjected to immunoblotting. CDK, cyclin-dependent kinase; Foxj1a, forkhead box j1a; GST, glutathione S-transferase; HA, hemagglutinin; HEK, human embryonic kidney; IgG, immunoglobulin G; IP, immunoprecipitation; T102, threonine 102; WB, western blot; WT, wild-type; zFoxj1a, zebrafish Foxj1a; α-Tubulin, acetylated tubulin. ubiquitylation-dependent, we would expect zFoxj1a to be stabilized upon overexpression of Ub K48R/G76A, which serves as a dominant negative inhibitor of poly-Ub chain formation [53,54]. Indeed, the presence of Ub K48R/G76A, but not wild-type Ub, efficiently inhibited the turnover of β-catenin (S12A Fig), which would be phosphorylated by glycogen synthase kinase-3β and destined for Ub-mediated degradation [55]. Unexpectedly, overexpression of Ub K48R/G76A was unable to promote the stabilization of Flag-tagged zFoxj1a (S12A Fig). Because the Flag epitope contains two lysine residues, we generated a lysine-free version of zFoxj1a, termed HA-zFoxj1a-K20R, by replacing the Flag tag with the HA epitope and mutating all 20 lysine residues within zFoxj1a protein to arginine residues. We found that both the wild-type and lysineless zFoxj1a were significantly stabilized with CDK4 overexpression (S12B Fig). Therefore, zFoxj1a is able to be degraded independently of ubiquitylation.
Increasing evidence suggests that a list of proteins that directly interact with proteasomal subunits are thought to be degraded through a Ub-independent degradation mechanism [56]. It has been demonstrated that the 19S regulatory subunit regulatory particle non-ATPase 10 (Rpn10) plays a critical role in the recognition of Ub-independent substrates [57][58][59]. Therefore, we examined whether zFoxj1a binds to zebrafish Psmd4b, the ortholog of mammalian Rpn10. Indeed, overexpressed or endogenous zFoxj1a was found to interact with Psmd4b ( Fig  7A and 7B). Consistent with these observations, an in vitro binding assay revealed a direct binding between purified zFoxj1a and Psmd4b (Fig 7C).
To further unveil how CDK4 and its kinase activity regulate Foxj1 stabilization, we examined whether CDK4 functions to control the interaction between Psmd4b and zFoxj1a. As expected, ectopic expression of wild-type CDK4, but not its kinase-deficient mutant, dramatically suppressed the association of Psmd4b with zFoxj1a or zFoxj1a-K20R (Fig 7D and 7E). In contrast, the binding ability of the unphosphorylated form of zFoxj1a for Psmd4b was retained with CDK4 coexpression (Fig 7F). Interestingly, the phospho-mimicking mutant of zFoxj1a (zFoxj1a-T102D) completely lost its ability to bind Psmd4b (Fig 7G), suggesting that CDK4-mediated phosphorylation of zFoxj1a at T102 eliminates its affinity for Psmd4b. In addition, Psmd4b overexpression reduced the expression levels of zFoxj1a and its lysine-free mutant (Fig 7H). When HEK293T cells were transfected with same amount of plasmid DNA to express wild-type zFoxj1a and its T102A and T102D mutants, respectively, we detected the highest expression level of zFoxj1a-T102D ( Fig 7I). However, T102D mutant could not be further stabilized by CDK4 overexpression (Fig 7I). Taken together, these results indicate that CDK4 phosphorylates and stabilizes zFoxj1a by disrupting its association with Psmd4b.
We next addressed the developmental relevance of CDK4-induced zFoxj1a stabilization. Cxcr4a-deficient DFCs exhibited defective cyclin D1 expression and impaired G1/S transition (Figs 3E-3H and 4F), implying a dysregulated activation of the CDK4/6 kinases. DFC-specific overexpression of zFoxj1a-T102D, but not wild-type or zFoxj1a-T102A, was found to restore the length of KV cilia in cxcr4a um20 mutants (Fig 7J and 7K). Because zFoxj1a is not involved in DFC proliferation [21,22], it was reasonable to find that the decrease in the number of cilia was not alleviated (S13 Fig). Overall, our results support a model in which Cxcl12b/Cxcr4a signaling activates ERK1/2, which then promotes Cyclin D1 expression. This, in turn, activates CDK4/6 kinase activity in DFCs. These activated G1 CDKs drive G1/S transition during DFC proliferation and promote zFoxj1a stability by phosphorylation to support KV ciliogenesis at later stages (Fig 7L).
In addition, it is surprising that overexpression of wild-type zFoxj1a could not rescue the length of KV cilia in cxcr4a um20 mutants because such overexpression should be able to bypass the effects of protein degradation. The FoxP family of transcription factors can bind as homoor heterodimers, and this interaction is essential for FoxP family members to regulate their targets [60,61]. These previous observations enlightened us to explore whether Foxj1 proteins The presence of Flag-Psmd4b in the protein complex that was pulled down by glutathione agarose was assessed by western blot, and input proteins were examined by Coomassie blue staining. (D-E) CDK4 kinase activity is required for its inhibitory effect on the association between Psmd4b and zFoxj1a. HEK293T cells transfected with the indicated plasmids were treated with MG132 for 5 hours prior to harvest for immunoprecipitation. Note that ectopically expressed CDK4 efficiently disrupted the association of Psmd4b with zFoxj1a (D) or its lysine-free mutant zFoxj1a-exist as a dimer. To test this, HEK293T cells were transfected with plasmids expressing Flag-zFoxj1a and HA-zFoxj1a. Coimmunoprecipitation experiments revealed a physical interaction between zFoxj1a proteins tagged with different epitopes (S14A and S14B Fig). Interestingly, compared with wild-type proteins, the unphosphorylated form of zFoxj1a showed a muchreduced affinity to each other, while the phospho-mimicking mutants exhibited a stronger binding ability to form complexes (S14C Fig). Moreover, whole-cell lysates containing overexpressed Flag-tagged wild-type zFoxj1a or its T102A or T102D mutants were separated on SDSand native PAGEs, respectively. Most of the wild-type as well as the phospho-mimicking proteins gave rise to a band about 130 kD, corresponding to zFoxj1a dimers in native PAGEs, implying that the wild-type zFoxj1a proteins might be phosphorylated at T102 to avoid degradation and form homodimers (S14D Fig). Indeed, the dimerization of the unphosphorylated zFoxj1a mutants was obviously decreased (S14D Fig). These results implicate that the phosphorylation at T102 site of zFoxj1a is important for its dimerization. Based on these observations, we supposed that the insufficiency of Cyclin D1-CDK4/6 activity in cxcr4a um20 mutants would lead to impaired dimerization of overexpressed wild-type zFoxj1a proteins, which might eventually deprive them of their transcriptional activity.

Discussion
In zebrafish embryos, DFCs undergo mitotic proliferation during epiboly and then exit the cell cycle, giving rise to epithelial cells that assemble cilia in the mature KV organ [16]. Cell-cycle defects in DFCs are often accompanied by an alteration in KV cilia elongation, raising the issue of whether there exists a feasible link between the cell cycle and cilia formation [16][17][18]. In this study, our experiments resolve this issue by demonstrating that Cxcr4 signaling is required for DFC proliferation and KV ciliogenesis by promoting Cyclin D1 expression. Specifically, we found that Cyclin D1-CDK4/6 accelerates the G1/S transition in DFCs while also facilitating cilia formation via stabilization of zFoxj1a. Ciliary dynamics appear to be precisely coordinated with cell-cycle progression [62]. It has been suggested previously that cell quiescence is essential for the formation of mouse nodal cilia [63]. Indeed, we observed that proliferating DFCs enter into a quiescent state upon differentiation into ciliated epithelial KV cells. Interestingly, our data indicate that during epiboly stages, Cxcr4a-signal-induced expression of Cyclin D1 functions to regulate DFC proliferation and zFoxj1a stability, which is important K20R (E), while overexpression of the CDK4 kinase-deficient mutant CDK4-K38M had no effect on their interaction (D). (F-G) Psmd4b has a much lower affinity for zFoxj1a-T102D. zFoxj1a-T102A is an unphosphorylated form, and zFoxj1a-T102D is a phospho-mimicking mutant of zFoxj1a. Note that Psmd4b was particularly associated with zFoxj1a and zFoxj1a-T102A (F) but was unable to bind zFoxj1a-T102D (G). The association between Psmd4b and zFoxj1a-T102A was unaffected by CDK4 overexpression (F). (H) Psmd4b overexpression induces a dramatic reduction in zFoxj1a expression. Note that a similar reduction in the expression level of the lysine-free mutant zFoxj1a-K20R was observed upon Psmd4b overexpression. (I) Comparison of protein stability in WT zFoxj1a and its mutants. HEK293T cells were transfected with the indicated constructs and harvested for western blot analysis. In comparison to WT zFoxj1a and its T102A mutant, the zFoxj1a-T102D mutant exhibited greater stability but could not be further stabilized by CDK4 overexpression. (J-K) DFC-specific overexpression of zFoxj1a-T102D in cxcr4a um20 mutants restores KV cilia length. Confocal images of 10-somite-stage cxcr4a-deficient embryos were injected with 200 pg of WT zfoxj1a or zfoxj1a-T102A or zfoxj1a-T102D mRNA at the 256-cell stage. The resulting embryos were harvested at the 10-somite stage for immunostaining using an antibody against α-Tubulin (J). The boxed areas in the images are presented at higher magnification in the relevant insets. Scale bar, 20 μm. Statistical data for cilia length are shown in panel K. Student t test, � P < 0.05, �� P < 0.01. Underlying data can be found in S1 Data. (L) A schematic diagram showing the regulatory mechanism of Cxcl12/Cxcr4a signaling in DFC proliferation and cilia formation. CDK, cyclin-dependent kinase; DFC, dorsal forerunner cell; ERK, extracellular regulated MAP kinase; Fox1ja, forkhead box j1a; GST, glutathione S-transferase; HA, hemagglutinin; HEK, human embryonic kidney; IgG, immunoglobulin G; IP, immunoprecipitation; KV, Kupffer's vesicle; ns, no significant difference; Psmd4b, proteasome 26S subunit, non-ATPase 4b; T102, threonine 102; Ub, ubiquitin; WT, wild-type; zFoxj1a, zebrafish Foxj1a; α-Tubulin, acetylated tubulin. https://doi.org/10.1371/journal.pbio.3000203.g007 for the ciliogenesis of quiescent KV cells. Therefore, the rapid cell-cycle progression of DFCs during epiboly stages is not only required for the generation of enough cells to construct KV but also plays a critical role in reserving sufficient levels of zFoxj1a protein to support subsequent cilia formation. Because Wnt/β-catenin signaling has been reported to play a role in both DFC proliferation and KV cilia elongation [17,18], it is interesting to consider whether this signaling pathway contributes to zFoxj1a stabilization via regulation of cell-cycle progression during the establishment of L-R asymmetry.
Most cxcr4a or cxcl12b mutants display a severe KV morphogenesis phenotype, a reduction in cilia length, and abnormal fluid flow in KV, but these embryos exhibit a relatively low penetrance in the situs abnormalities at later stages. Moreover, the mutated Cxcr4a has no residual activity because of its inability to rescue the laterality defects in cxcr4a um20 embryos. It has been well established that the Nodal-Pitx2 pathway (the Spaw-Pitx2c pathway in zebrafish) plays a central and conserved role in L-R symmetry breaking [1]. Recently, a zebrafish mutant defective for the spaw gene has been identified [64]. Surprisingly, in 70% of the spaw mutant embryos, cardiac looping occurs in the correct dextral direction. That is to say, only about 30% of the spaw mutants showed a "no-looping" or reversed "left-looping" heart [64]. This previous study proposed a Nodal-and KV-function-independent mechanism that drives asymmetric heart looping in zebrafish embryos [64]. In our study, about 20% of the cxcr4a or cxcl12b mutants showed a "no-looping" or reversed "left-looping" heart, which is slightly lower than that of spaw mutants. Therefore, Cxcr4a signaling primarily functions in regulating KV formation and ciliogenesis but may have little role in fulfilling the Nodal-and KV-function-independent mechanism. In addition, a higher percentage of spaw mutants (about 50%) exhibited the L-R asymmetry defects in digestive system laterality [64]. However, a relative lower percentage of cxcr4a or cxcl12b mutants (about 30%) displayed liver laterality defects. Similarly, it has been reported that cxcr4a um20 mutants showed a partial penetrance of circulation defects by day 2 of development [34]. These observations suggest that there could be some unknown mechanisms to partially compensate for the function of cxcr4a at later developmental stages.
It has been reported previously that G1 cyclins function together with their associated CDKs to phosphorylate a variety of transcription factors, including Smad2/3 and pluripotency factors, to control ES cell differentiation [11,12]. A systematic screen for CDK4/6 substrates identified fox family transcription factor forkhead box M1 (FOXM1) as a critical phosphorylation target [65]. CDK4/6 stabilize and activate FOXM1 by phosphorylation at multiple sites to protect cancer cells from senescence [65]. In contrast, CDK2 reduces DNA-damage-induced cell death by phosphorylation of forkhead box O1 (FOXO1) at Ser249, resulting in cytoplasmic localization of FOXO1 [66]. In this study, we show that CDK4 directly interacts with and phosphorylates zFoxj1a at a conserved "TPGK" motif within the N-terminal region. Phosphorylation at T102 was not found to alter the subcellular distribution of zFoxj1a but was shown to promote its stabilization. Therefore, the functional interaction between CDK4 and zFoxj1a provides a mechanism by which cilia development is facilitated. Because CDK4 also stabilizes mFoxj1 through phosphorylation of T87 within the substrate motif, it is likely that the molecular linkage between cell-cycle progression and ciliogenesis is conserved among vertebrates.
Previous studies have shown that individual knockout of either Cdk4 or Cdk6 in mice did not reveal any obvious developmental defects [67,68]. In contrast, embryos defective for Cdk4 and Cdk6 die during the late stages of embryonic development because of severe anemia [68]. Cdk2 knockout mice were viable and developed normally [69]. However, compound mutation of Cdk4 and Cdk2 results in impaired proliferation and heart growth [70]. These results indicate a putative redundant or compensatory role of these G1 phase Cdk genes. The complicated interactions between these Cdk genes also prevent us from exploring the in vivo functions of zebrafish cdk4 in DFCs or KV cells. Nevertheless, we found that wild-type embryos treated with the CDK4/6 chemical inhibitor PD0332991 exhibited L-R defects. Moreover, overexpression of Cyclin D1 in cxcr4a mutants ideally rescued the defects in KV cell proliferation and cilia formation. These observations suggest an important function of CDK4/6 in connecting cell-cycle progression with cilia formation.
The majority of proteosomal protein degradation relies on Ub conjugation. However, there are increasing numbers of examples of proteasomal degradation that occur without prior ubiquitination [71,72]. Our study reveals that overexpression of the dominant negative Ub has no effect on zFoxj1a stabilization. However, wild-type and lysineless zFoxj1a are found to be similarly stabilized by ectopic CDK4 expression. Therefore, zFoxj1a is targeted for proteasomal degradation in a Ub-independent manner. Intriguingly, E3 Ub ligases, including mahogunin RING finger 1 (MGRN1) and Lnx2b, have been reported to play a role in L-R laterality specification in rodents and zebrafish [24,73], suggesting a role of the Ub-proteasome system in the modulation of protein turnover during L-R body patterning. However, because of the fact that L-R symmetry breaking occurs within a short time window during vertebrate embryonic development [4,74], the accelerated and economical regulation of protein degradation may be essential. Because Ub-independent degradation does not require the enzymatic cascade of Ub conjugation, it would be more efficient to alter the concentration of zFoxj1a protein levels via Ub-independent proteasomal degradation during L-R asymmetric development. Interestingly, a recent study has demonstrated that Foxj1 is rapidly turned over by the Ub-proteasome system in mouse primary ependymal cells [75]. Therefore, Foxj1 is a protein with a short half-life that undergoes proteasomal degradation via Ub-dependent or independent pathways dependent on the cellular context.
Several proteins have been reported to interact with the 19S regulatory subunit Rpn10 via their Ub-like (UBL) domains [57][58][59]. Interestingly, while lacking a UBL domain, zFoxj1a interacts directly with Psmd4b, the zebrafish ortholog of mammalian Rpn10. Our study demonstrates that CDK4 phosphorylates and stabilizes zFoxj1a by disrupting its association with Psmd4b. Similarly, the Ub-independent proteasomal degradation of Yeast phosphatidic acid phosphohydrolase 1 (Pah1) has also shown to be governed by its phosphorylation state [76]. Therefore, this may represent a general mechanism by which protein kinase-mediated phosphorylation plays a critical role in the protection of their substrates from Ub-independent proteasomal degradation.

Ethics statement
Our zebrafish experiments were all approved and carried out in accordance with the Animal Care Committee at the Institute of Zoology, Chinese Academy of Sciences (permission number: IOZ-13048).

WISH
WISH was performed using the NBT-BCIP substrate following standard procedures. For twocolor fluorescence in situ hybridization, anti-digoxigenin-POD (11633716001; Roche, Basel, Switzerland) and anti-fluorescein-POD (11426346910, Roche) were used as primary antibodies to detect digoxigenin-labeled sox17 probes and fluorescein-labeled cyclin D1 probes, respectively. Fluorescence in situ hybridization was then carried out using the Perkin Elmer TSA fluorescein system (NEL701A001KT; Waltham, MA, USA) according to the manufacturer's instructions.

Cell lines and transfection
HEK293T and HeLa cell lines (American Tissue Culture Collection [ATCC], Manassas, VA, USA) were cultured in DMEM medium supplemented with 10% FBS in a 37˚C humidified incubator in a 5% CO 2 environment. Cell transfections were carried out using Lipofectamine 2000 (11668019; Invitrogen, Carlsbad, CA, USA) following the manufacturer's instructions.

Immunostaining and confocal microscope
Embryos were fixed in 4% paraformaldehyde overnight. Fixed embryos were then rinsed with PBST for a total of 4 times every 5 minutes. Embryos were then blocked at room temperature for 1 hour in 10% heat-inactivated goat serum and then stained with the following affinity- was used to stain nuclei. The stained embryos were then embedded with 2% low melting agarose and imaged using a Nikon A1R+ confocal microscope (Nikon, Tokyo, Japan) with identical settings.

Fluorescent bead tracking
Fluorescent red beads of 1 uM diameter (1:500; 18660-5, Polysciences, Warrington, PA, USA) were injected into KV of embryos at the 6-somite stage. The resulting embryos were then embedded in 2% low-melting agarose at the 10-somite stage for confocal imaging. Beads tracking videos and images were processed using Image Pro 6.0.

High-speed cilia video microscopy
For recording cilia motility in KV, we embedded 10-somite-stage embryos in 3% methylcellulose on glass cover slides. Cilia movements were recorded with a high-speed camera (Motion-BLITZ EoSens mini1; Mikrotron, Unterschleiβheim, Germany) mounted on a Leica Sp8 confocal microscope with a 100× oil objective in bright field. Movies were recorded at a rate of 500 frames per second, and playbacks were at 25 frames per second.

Antibodies and immunoprecipitation assays
For immunoblotting experiments, we used the following affinity-purified antibodies: anti-Flag For coimmunoprecipitation assays, embryos or HEK293T cells were harvested and lysed with TNE lysis buffer (10mM Tris-HCl [pH 7.5], 150 mM NaCl, 2 mM EDTA, and 0.5% Nonidet P-40) containing a protease inhibitor mixture. Lysates were incubated with anti-Flag-agarose beads (A2220, Sigma-Aldrich) or protein A-Sepharose beads (101041, Invitrogen) and anti-phospho-threonine-proline antibody (1:5,000; 9391, Cell Signaling Technologies) at 4˚C for 4 hours. Beads were washed 4 times with TNE buffer. Bound proteins were then separated by SDS-PAGE and visualized by western blots.

BrdU incorporation assay
Half an hour before the embryos were fixed for BrdU labeling experiments, they were dechorionated and incubated with 10 mM BrdU solution (B5002, Sigma-Aldrich) for 20 min at 4˚C. Then, these embryos were allowed to develop to desired stages in Holtfreter's solution at 28.5˚C. Incorporated BrdUs and GFP were detected using anti-BrdU (1:1,000; ab6326, Abcam) and rabbit anti-GFP antibodies (1:1,000; A-11122, Invitrogen) by wholemount immunostaining.

In vitro GST pull-down
GST fusion proteins were expressed in Escherichia coli strain BL21 and purified using Glutathione-Sepharose 4B beads (71024800-GE, GE Healthcare, Chicago, IL, USA). GST-Myc-CDK4, GST-HA-cyclinD1, and GST-Flag-psmd4b were treated with Thrombin (1:1,000; T4648, Sigma-Aldrich) to cleave their GST tags. For in vitro binding assays, GST-Foxj1a proteins were immobilized by Glutathione-Sepharose 4B beads and incubated with the indicated purified proteins at 4˚C for 3 hours. Following washing, the bound proteins were separated with SDS-PAGE and analyzed by western blots.

Statistical analysis
Cilia number and length were measured using ImageJ software. All results were expressed as the mean ± SD. Differences between control and treated groups were analyzed using the unpaired two-tailed Student t test. Results were considered statistically significant at P < 0.05.  (1/190)) embryos were treated with 0.5 μM PD0332991 or 0.2 μM CYC202 from the shield stage to the 10-somite stage. Then, these embryos were harvested for in vivo confocal imaging. Note that both PD0332991 and CYC202 treatments induced a remarkable increase of the number of mKO2-zCdt1-positive cells. Scale bar, 200 μm. CDK, cyclin-dependent kinase; EF1α, eukaryotic translation elongation factor 1α; mKO2, monomeric Kusabira Orange2; Tg, transgene; zCdt1, zebrafish chromatin licensing and DNA replication factor 1. (TIF) S11 Fig. CDK4 phosphorylates and stabilizes mFoxj1. (A-B) HEK293T cells were transfected with the indicated plasmids and then harvested for immunoprecipitation with a phospho-threonine-proline antibody. Phosphorylation of mFoxj1 (A) and its T87A mutant (B) was detected by western blot. Note that the CDK4-mediated phosphorylation of mFoxj1 was clearly decreased in the T87A mutant. (C) Western blots of total lysates from HEK293T cells transfected with the indicated plasmids. Note that CDK4 overexpression could stabilizes wildtype mFoxj1 but not the T87A mutant. CDK, cyclin-dependent kinase; HEK, human embryonic kidney; Foxj1, forkhead box j1a; mFoxj1, mouse Foxj1. (TIF) S12 Fig. Proteasomal degradation of zFoxj1a is independent of Ub modification. (A) Overexpression of Ub K48R/G76A was unable to stabilize zFoxj1a. Flag-tagged β-catenin and zFoxj1a were coexpressed with wild-type Ub or Ub K48R/G76A, a dominant negative inhibitor of chain formation and degradation. Cell extracts were immunoblotted with the indicated antibodies. (B) CDK4 overexpression similarly promoted the expression of wild-type zFoxj1a and its lysineless mutant K20R. CDK, cyclin-dependent kinase; Foxj1a, forkhead box j1a; Ub, ubiquitin; zFoxj1a, zebrafish Foxj1a. (TIF) S13 Fig. The decrease in the number of cilia in cxcr4a um20 mutants was not alleviated by DFC-specific overexpression of wild-type zFoxj1a and its T102A and T102D mutants, respectively. cxcr4a-deficient embryos were injected with 200 pg of wild-type zfoxj1a or zfoxj1a-T102A or zfoxj1a-T102D mRNA at the 256-cell stage. The resulting embryos were harvested at the 10-somite stage for immunostaining using an antibody against α-Tubulin. Cilia number was quantitatively analyzed using ImageJ software. Student t test, �� P < 0.01. Underlying data can be found in S1 Data. DFC, dorsal forerunner cell; Foxj1a, forkhead box j1a; ns, no significant difference; T102, threonine 102; zFoxj1a, zebrafish Foxj1a; α-Tubulin, acetylated tubulin. (TIF) S14 Fig. Phosphorylation of T102 is essential for the dimerization of zFoxj1a proteins. (A-C) In vivo self-association of zFoxj1a. HEK293T cells were transfected with indicated plasmids encoding differently tagged wild-type zFoxj1a or its mutants. Lysates were immunoprecipitated with anti-Flag or anti-HA antibodies and then immunoblotted with indicated antibodies. In panel C, HEK293T cells were treated with MG132 for 5 hours prior to harvest for immunoprecipitation. (D) zFoxj1a forms homodimers. HEK293T cells were transfected with indicated plasmids encoding Flag-tagged wild-type zFoxj1a or its mutants. Lysates were then separated on SDS-and native PAGEs, respectively. Foxj1a, forkhead box j1a; HA, hemagglutinin; HEK, human embryonic kidney; T102, threonine 102; zFoxj1a, zebrafish Foxj1a. (TIF) S1 Raw Images. The uncropped blots for the westerns and immunoprecipitations. (PDF) S1 Movie. Wild-type embryos, normal KV flow. Wild-type Tg(sox17:GFP) embryos were injected at the 6-somite stage with fluorescent beads and imaged using a Nikon A1R+ confocal microscope at the 10-somite stage. Dorsal view with anterior to the top. GFP, green fluorescent protein; KV, Kupffer's vesicle; sox, SRY-box transcription factor; Tg, transgene.