Human cells lacking CDC14A and CDC14B show differences in ciliogenesis but not in mitotic progression

CDC14 encodes an essential highly conserved dual specificity protein phosphatase that in budding yeast Saccharomyces cerevisiae promotes mitotic exit, the transition from mitosis into G1 phase, and cytokinesis (Taylor et al., 1997). During most of the cell cycle, S. cerevisiae Cdc14 is entrapped in the nucleolus by complex formation with the protein Net1 (also known as Cfi1) (Shou et al., 1999; Visintin et al., 1999). With anaphase onset, Cdc14 becomes released from the nucleolus in two waves. The Cdc14 early anaphase release (FEAR) pathway, consisting of the polo-like kinase Cdc5, the kinetochore protein Slk19, Spo12 and the cohesion regulator separase, triggers a partial release of Cdc14 that leads to the dephosphorylation of spindle associated proteins and of the kinase Cdc15 (Jaspersen and Morgan, 2000; Stegmeier et al., 2002). Cdc15 is a component of the mitotic exit network (MEN), a GTPase-driven signalling cascade that controls Cdc14 after FEAR (Pereira et al., 2000). MEN activation of Cdc14 then promotes mitotic exit and cytokinesis through the dephosphorylation of cyclin-dependent kinase (CDK1) sites in proteins that trigger cyclin B degradation or play a role in anaphase spindle formation and cytokinesis (Jaspersen et al., 1999; Palani et al., 2012; Raspelli et al., 2015; Bremmer et al., 2012). In addition, the fission yeast Cdc14 orthologue Clp1 (also known as Flp1) is entrapped in the nucleolus during interphase; however, it is released by prophase, and thus earlier in the cell cycle than Cdc14 (Cueille et al., 2001). As for S. cerevisiae Cdc14, fission yeast Clp1 controls cytokinesis (Trautmann et al., 2001). An additional cell cycle function of Clp1/Flp1 is at the spindle pole body (SPB) where Clp1/Flp1 participates in an activation loop triggering Cdc25 activation and mitotic entry (Chan et al., 2017; Esteban et al., 2004; Wolfe and Gould, 2004).

The human genome encodes for three CDC14 genes, named human CDC14A, CDC14B and CDC14C (Mailand et al., 2002; Mocciaro and Schiebel, 2010). CDC14C is a near pseudogene of CDC14B and only differs in a small number of bases from CDC14B cDNA (Rosso et al., 2008). CDC14A and CDC14B vary by a 54-amino-acid long N-terminal extension in CDC14B that is important for the targeting of this protein to the nucleolus during interphase (Kaiser et al., 2002). Functions for the human CDC14 phosphatases that have been suggested are DNA damage control, cell cycle regulation, regulation of the primary cilia and the actin cytoskeleton (Mocciaro and Schiebel, 2010; Uddin et al., 2018). Loss of CDC14A function in humans is associated with deafness and infertility, indicating an important role of this phosphatase in the inner ear and testis (Imtiaz et al., 2018). A systematic study on substrates that are dephosphorylated by CDC14A identified components of the actin cytoskeleton, such as the protein eplin (also known as LIMA1) (Chen et al., 2017, 2016). CDC14B substrates are not as well characterized.

Do human CDC14 paralogues have cell cycle functions, as this is the case for yeast CDC14 phosphatases? Based on a gene knockout (KO) for human CDC14B it was concluded that it is dispensable for chromosome segregation and mitotic exit (Berdougo et al., 2008). However, it was never tested whether functional redundancy between the three CDC14 paralogues masks cell cycle functions. Here, we show that CDC14C is not expressed in human retina epithelial cells immortalized with hTERT (RPE1) cells. RPE1 cells with genomic inactivation (knockout; KO) of the CDC14A, CDC14B and CDC14A CDC14B show no mitotic or other severe cell cycle defects. However, knockout of CDC14A or CDC14B impacts ciliogenesis of cycling cells, although in different ways. These data suggest that the functions of CDC14 phosphatases diverged during evolution from essential cell cycle regulation of mitosis in budding yeast to more divergent functions centred on the actin cytoskeleton and cilia in higher eukaryotes.

Redundancy between CDC14 paralogues could mask functions upon inactivation of only one CDC14 gene, for example in cell cycle regulation. A prerequisite for redundancy is the co-expression of CDC14 genes. In order to judge this, we first analysed transcription of CDC14 genes in tissues and cell lines using published databases. This indicated that CDC14C is only transcribed in testis and brain but not in other organs, whereas CDC14A and CDC14B are expressed in almost all tissues (Brawand et al., 2011) (Fig. 1A). In line with this finding, RNA sequencing using the non-transformed telomerase-immortalized cell line RPE1 derived from retina did not identify CDC14C transcripts, while CDC14A and CDC14B mRNAs were detected (Fig. 1B (Santaguida et al., 2015)). To confirm that the CDC14C gene is not expressed in RPE1 cells, we analysed CDC14C mRNA by RT-PCR and DNA sequencing. This analysis was complicated by the fact that the DNA sequence of the chromosomal copy of CDC14C is very similar to the cDNA of CDC14B (Fig. 1C, red stars indicate differences). To avoid amplifying the chromosomal CDC14B, we choose specific primers that amplified the cDNA of CDC14B and CDC14C, and genomic CDC14C, but not genomic CDC14B due to the large size (26.348 bp) of this genome-specific CDC14B PCR product (Fig. 1C, see panel I). To judge expression of CDC14B and CDC14C and to exclude contamination with genomic CDC14C DNA, we treated the mRNA before the RT-PCR with and without DNaseI. Without DNaseI incubation, the RT-PCR combined with DNA sequencing identified the CDC14B and CDC14C signatures (Fig. 1C, see panel II). With DNaseI treatment, however, only the CDC14B signature was detected (Fig. 1C, see panel III) suggesting that the CDC14C PCR product in the absence of DNaseI arose from genomic DNA. To exclude that CDC14C expression is upregulated by the loss of CDC14A and CDC14B, we analysed CDC14C transcripts in CDC14A and CDC14B double-KO (hereafter CDC14A/B KO) cells. Only CDC14B but not CDC14C mRNA was detected in this experiment (Fig. 1C, see panel IV). This indicates that CDC14C is not expressed in cycling RPE1 cells while CDC14A and CDC14B are expressed.

The main function of budding yeast Cdc14 is in the regulation of anaphase events and mitotic exit (Mocciaro and Schiebel, 2010). In mammalian cells, this function of CDC14 is mainly taken over by the phosphatases PP1 and PP2 (Cundell et al., 2013; Holder et al., 2019). However, because of the lack of a systematic study on cell cycle progression on cells with impaired CDC14A and CDC14B function, it is still possible that CDC14 phosphatases control aspects of mitosis. Because of the lack of CDC14C expression in RPE1 cells (Fig. 1B,C), we focused the analysis on CDC14A and CDC14B. To avoid misinterpretations from incomplete siRNA depletions, we inactivated CDC14A and CDC14B by genomic gene deletion (see Materials and Methods). We addressed mitotic functions of CDC14A and CDC14B by analysing transition times of mitotic phases from live-cell imaging data of wild-type (WT), CDC14A KO (AKO6), CDC14B KO (BKO10) and CDC14A/B KO (BAKO3) RPE1 cell lines expressing the chromatin marker H2B-mNG. We did not observe changes in the timing of the transition in the timing of the transition from prometaphase to metaphase, metaphase to anaphase onset and prometaphase to anaphase onset in any of the deletion cell lines compared to the WT control (Fig. 2A,B). In addition, analysis of the timing of actin ring contraction did not indicate a role of the CDC14A and CDC14B phosphatases in the regulation of cytokinesis (Fig. 2C,D), in contrast to what was observed previously in Caenorhabditis elegans embryos (Gruneberg et al., 2002). Finally, we analysed mitotic exit by arresting cells in prometaphase using the microtubule depolymerizing drug nocodazole, followed by nocodazole washout and analysis of progression of cells into G1 (Fig. 2E, top). No difference between WT and CDC14 KO cells was observed for mitotic exit (Fig. 2E, bottom). Thus, in human RPE1 cells CDC14A and CDC14B phosphatases do not have obvious mitotic functions.

Because of the lack of mitotic defects, we asked whether CDC14A and CDC14B have a more general role in the cell growth and division. We first tested whether cell proliferation was affected when cells lacked CDC14A or CDC14B. This was done with the MTT [3-(4,5-dimethylthiazole-2-yl)-2,5-diphenyltetrazolium bromide] assay, which measures the activity of the enzyme NAD(P)H-dependent cellular oxidoreductase and therefore reflects fitness and proliferation of cells (Mosmann, 1983). Analysis of the MTT profiles of RPE1 cells with single and double-deletion of CDC14A and CDC14B did not indicate a proliferation difference of the CDC14 mutants in comparison to WT RPE1 (Fig. 3A). To measure the cell doubling more directly, we covalently labelled intracellular proteins with a pulse of the fluorescent cell division tracker dye carboxyfluorescein succinimidyl ester (CFSE) (Chung et al., 2017). The decrease in CFSE intensity due to growth and cell division was monitored by flow cytometry. The CFSE signal decreased similarly for WT and CDC14 (AKO6, BKO10, BAKO3) cells (Fig. 3B). The protein Ki-67 (also known as MKI67) is a marker for cell proliferation (Gerdes et al., 1983). A similar percentage of WT and CDC14 KO cells were Ki-67 positive (Fig. 3C) suggesting that most cells are dividing. This is consistent with the results from the MTT and CFSE assays, and together indicates that CDC14A and CDC14B are not important for growth and division of RPE1 cells.

In order to detect cell cycle differences between WT and CDC14 KO RPE1 cells, we analysed cycling RPE1 cells by flow cytometry using propidium iodide and phospho-histone H3 antibodies as a mitosis marker. This did not indicate major differences in the cell cycle distribution between WT and CDC14 KO cells (Fig. 3D–F).

Taken together, loss of CDC14A and CDC14B does not significantly alter the cell proliferation rate, cell cycle distribution and the mitotic index, which is consistent with our finding that mitotic progression and mitotic exit is undisturbed (Fig. 2).

Despite high conservation, human CDC14 phosphatases appear to play a role in more divergent biological processes when compared to the mitotic functions in yeast. Loss of function analysis of CDC14A and CDC14B in mice and DT40 cells suggests both CDC14 paralogues have a role of in DNA repair (Lin et al., 2015; Mocciaro et al., 2010; Wei et al., 2011). In addition, depletion of CDC14A and CDC14B in zebrafish suggests a role in ciliogenesis (Clement et al., 2011, 2012). We previously observed longer cilia in serum starved CDC14 KO RPE1 cells (Uddin et al., 2018).

To test whether deletion of CDC14 genes in RPE1 cells leads to the accumulation of DNA damage and affects ciliogenesis, we analysed cycling CDC14 KO cells for both phenotypes using appropriate markers. CDC14A and CDC14B single- and double-KO cells showed an increase in the number of nuclear phospho-H2AX (S139) foci (Fig. S1A,B), which appear in response to double-strand breaks (Rogakou et al., 1998). Moreover, loss of CDC14A and CDC14B affected ciliogenesis in RPE1 KO cells. Approximately 6% of cycling WT and CDC14A KO cells and 15% of CDC14B KO and CDC14A/B KO cells carried cilia (Fig. 4A,C). The average cilia length of WT and CDC14B KO (BKO10) cells was 1.6 µm. The length in CDC14A KO (AKO6) and CDC14A/B KO (BAKO3) cells was with 2 µm significantly longer (Fig. 4B,C).

Replicating RPE1 cells are able to assemble a primary cilium in G1 phase (Spalluto et al., 2013). To test whether this is also the case in the RPE1 CDC14 KO cell lines, we arrested the cell cycle of RPE1 cells in G1 using the CDK4–cyclin D inhibitor palbociclib (Trotter and Hagan, 2020) in the presence of serum and measured formation and properties of cilia. In these conditions, 20% of the RPE1 WT and CDC14A KO cells and 30% of the CDC14B KO and CDC14A/B KO cells formed cilia (Fig. 4D,E), which is 2–4-times more frequently than in cycling cells (Fig. 4A). As observed for cycling cells, the cilia from CDC14A KO and CDC14A/B KO cells were longer than those from WT and CDC14B KO cells (Fig. 4F). Importantly, after the release from the G1 block by palbociclib washout CDC14B KO and CDC14A/B KO cells disassembled their cilia three times faster than WT and CDC14A KO cells (Fig. 4G).

Complementation of CDC14 KO cells by rescue constructs is complicated by the fact that CDC14 overexpression has a strong impact on cell cycle progression, centrosome duplication and cell division (Kaiser et al., 2002; Mailand et al., 2002; Uddin et al., 2018). To confirm the DNA damage and cilia phenotype of CDC14A and CDC14B KO cells, we used independently constructed CDC14 KO RPE1 cell lines (AKO7, BKO17 and BAKO20) that showed the same accumulation of phospho-H2AX foci (Fig. S2A) and the same difference in cilia number, length and disassembly (Fig. S2B–D) than the first cohort of cell lines. In summary, the lack of CDC14A and CDC14B leads to the accumulation of DNA damage and impacts cilia formation in G1 phase and their subsequent disassembly in later cell cycle phases.

The molecular cause of the accumulation of phospho-H2AX foci and the altered ciliogenesis regulation in CDC14 KO RPE1 cells is currently unclear and needs further investigation. Previously, we have shown that in serum starved cells, the phospho-regulation of actin-binding protein drebrin by the cyclin-dependent kinase CDK5 and the phosphatase CDC14A contributes to cilia length regulation (Maskey et al., 2015; Uddin et al., 2018). Furthermore, the CDK10 and cyclin M protein kinase, whose deficiency is linked to STAR syndrome, have been reported to negatively regulate ciliogenesis (Guen et al., 2018, 2016). It is therefore conceivable that CDC14A or CDC14B also counteract members of the CDK kinase family in some process of ciliogenesis during interphase.

hTERT-RPE1 cells were cultured in Gibco DMEM/F-12 (Gibco) medium supplemented with 10% FBS, 1% L-glutamine and 1% penicillin/streptomycin at 37°C with 5% CO₂. Cells were routinely tested for mycoplasma contamination. RPE1 cells were treated for 18 h with 100 nM palbociclib (Tocris cat. #4786) to arrest cells in G1. To arrest cells in prometaphase, cells were first arrested in G1 with palbociclib for 18 h, followed by palbociclib washout and treatment with 330 nM nocodazole for 24 h.

RPE1 CDC14B, CDC14A single KO and CDC14A/B KO cell lines were generated previously by zinc finger nuclease-based approaches and confirmed by southern blotting, RT-PCR and sequencing (Chen et al., 2017, 2016; Uddin et al., 2015). In brief, exon 9 of CDC14A was targeted to generate a deletion of 77 amino acids (203–279 aa) around the catalytic cysteine residue. CDC14B KO cells were generated by Cre-recombinase-mediated removal of the selection marker leading to incorporation of pre-mature stop codons in all three frames immediately followed by a single loxP site (34 bp) at exon 4 upstream of the catalytic site. CDC14A KO was carried out on top of Cre-infected CDC14B KO cells to obtain double-KO cells.

H2B–mNeonGreen-expressing cell lines were generated via retroviral gene transfer according to manufacturer's protocol (Clontech) using the construct pQCXIZ-H2B-mNeonGreen. Stable cells constitutively expressing the construct were enriched by fluorescence-activated cell sorting using a BD FACS Aria III instrument (Beckton Dickinson).

Total RNA was isolated from cells using the RNeasy Mini Kit (Qiagen, Hilden, Germany) following the manufacturer's instructions. RT-PCR analysis were carried out using SuperScript^® III One-Step RT-PCR System with Platinum^® Taq DNA Polymerase (Invitrogen). Sequencing was carried out after gel purification of the PCR products. PCR primer sequences used to amplify the sequence flanking exon 2 and exon 7 of CDC14B/C were 5'-GCCATTCTCTACAGCAG-3′ and 5'-GGAAGCCATACTGCATTGC-3′. The results of our sequence analysis were compared with publicly available RNA sequencing data banks (Brawand et al., 2011; Santaguida et al., 2015). The utilized datasets are available at NCBI GEO under accession codes GSE30352 and GSE60570, respectively.

Cells were counted three times and 3×10⁴ cells were resuspended in 8 ml cell medium. 750 cells were seeded into 96-well in quadruplicates for each time point and grown up to 7 days. Every 24 h, cell viability/proliferation was assayed by replacing cell medium with medium containing freshly prepared MTT (5 mg/ml MTT in PBS diluted 1:10 in cell medium). After 3.5 h at 37°C, MTT medium was carefully removed and formazan crystals were dissolved in 100 μl 2-propanol for 2 h while shaking. MTT reduction was quantified at an absorbance of 595 nm using the M1000 plate reader (Tecan).

For CFSE staining, a CellTrace™ CFSE Cell Proliferation Kit (Thermo Fisher) was used according to manufacturer's protocol. Stained cells (5×10⁴) were seeded in six-well plates and grown for up to 4 days before fixation with 4% PFA in PBS for 10 min at room temperature. The CFSE intensity of fixed cells was assessed by flow cytometry (BD Canto II).

For analysis of cell cycle distribution and mitotic index, cells were washed once with PBS and fixed with 70% ethanol in phosphate-buffered saline (PBS) at −20°C for 3 h. Fixed cells were permeabilized with 0.25% Triton X-100 for 15 min on ice before staining with anti-phospho-histone H3 (Ser10) (rabbit, 1:1000, Cell Signaling #3377S) overnight at 4°C. Then, cells were washed with PBS containing 1% BSA and stained with fluorescently labelled (Alexa Fluor 488) anti-rabbit-IgG secondary antibody (Invitrogen, A-21206, 1:500) for 30 min at room temperature. Cells were washed with PBS and treated with RNase A (100 μg/ml) at 37°C in dark for 30 min. Immediately before flow cytometric analysis (BD Canto), propidium iodide staining solution (25 μg/ml) was added.

Cells were seeded on coverslips and allowed to attach overnight before treatment. Depending on the primary antibody, cells were fixed with 4% PFA in PBS (10 min at room temperature) or with methanol (5 min at −20°C). Fixed cells were permeabilized with 0.1% Triton X-100 in PBS for 10 min at room temperature and incubated with 10% FBS for 1 h at room temperature to block non-specific antibody binding. Next, cells were incubated with primary antibody for 1 h. Then, cells were incubated with the secondary antibody and the DNA dye Hoechst 33342 (1:2000) for 30 min. Coverslips were mounted with Moviol and dried overnight. Images were acquired on a DeltaVision RT system (Applied Precision) equipped with an Olympus IX71 microscope. The following antibodies were used in immunofluorescence microscopy (IF) experiments: H2AX^pS139 (rabbit, 1:1000, Cell Signaling #9718), Ki-67 (mouse, 1:100, Santa Cruz Biotechnology sc-23900), C-Nap1 (goat, 1:1000, Panic et al., 2015), ARL13B (rabbit, 1:200, Proteintech 17711-1-AP), Alexa Fluor 488/555/647 (mouse/rabbit/sheep/goat, 1:500, Invitrogen).

Live-cell imaging was performed using a DeltaVision RT system (Applied Precision) with an Olympus IX71 microscope at 37°C. Cells were seeded in Ibidi glassware and cultured in DMEM/F-12 complete medium with HEPES and without Phenol Red (Gibco). If applicable, cell medium was supplemented with 100 nM SiR-actin (Lukinavičius et al., 2014) (Spirochrome) and labelled for 6 h before imaging. Time-lapse images were collected at 5 min intervals.

If not otherwise stated, results were confirmed by three independent experiments (n=3) and are expressed as mean±s.d. Prism 8 software (GraphPad) was used for statistical significance analysis including two-tailed unpaired Student's t-test and one-way or two-way ANOVA followed by Tukey's multiple comparison test (ns, P>0.05; *P≤0.05; **P≤0.01; ***P≤0.001; ****P≤0.0001).

We thank Dr Monika Langlotz (Heidelberg) for the FACS experiments and Dr Iain Hagan (Manchester) for communicating data before publication. Some of the content from Fig. 1 in this paper formed part of Borhan Uddin's PhD thesis in the Combined Faculties for the Natural Sciences and for Mathematics at the Ruperto-Carola University of Heidelberg in 2018.

The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.255950.reviewer-comments.pdf