DYRK2 maintains genome stability via neddylation of cullins in response to DNA damage

DNA is constantly exposed to endogenous and exogenous factors, such as ultraviolet light, oxidative stress, carcinogens and radiation. To maintain genome stability, cells construct cellular pathways, including pathways for sensing, signaling and repair of damaged DNA, which are collectively termed the DNA damage response (DDR) (Ciccia and Elledge, 2010; Jackson and Bartek, 2009). The DDR is orchestrated by kinases that belong to the phosphoinositide 3-kinase (PI3K)-like kinase family, namely ataxia-telangiectasia mutated (ATM), ATM- and Rad3-related (ATR) and DNA-dependent protein kinase catalytic subunit (DNA-PKcs; encoded by PRKDC). These kinases induce the phosphorylation of histone H2A.X on serine 139 (known as γH2A.X), which is a critical early step in the cellular response to double-strand breaks (DSBs) (Rogakou et al., 1998) and which activates DNA repair signaling factors, such as p53 (encoded by TP53), resulting in the cell cycle arrest in G1, senescence or apoptosis (Blackford and Jackson, 2017). More recently, abnormalities of DNA repair or sensing processes or both have been found to induce genome instability and to lead to micro-nuclei (Takahashi et al., 2017; 2018). Notably, genome instability induces several disease processes, such as carcinogenesis (Aparicio et al., 2014), and is a factor in neurodegenerative diseases (McKinnon, 2017) and aging (Wang and Lindahl, 2016).

For the DDR, post-translational modifications (PTMs) are essential regulators (Walsh et al., 2005). Ubiquitin and the small ubiquitin-like modifier proteins (SUMOs), which are ubiquitin-like proteins (UBLs), have been well studied as regulators for the cellular response to DSBs (Schwertman et al., 2016). It is also known that other UBLs, especially neural precursor cell-expressed developmentally down-regulated 8 (NEDD8), have a role in this process. Neddylation is a type of PTM that conjugates UBLs and NEDD8 onto substrates. In the process of neddylation, NEDD8 is first activated by the NEDD8-activating enzyme E1 (a heterodimer composed of NAE1 and ubiquitin-like modifier activating enzyme 3; UBA3); the activated NEDD8 is then transferred to NEDD8-conjugating enzyme E2 [UBC12 (also known as UBE2M) or UBE2F]. Finally, NEDD8-conjugating enzyme E2 interacts with the substrate-specific NEDD8-E3 ligase to conjugate NEDD8 to its target substrates, mainly cullins, which are a large family of multiunit E3 ubiquitin ligases that regulate degradation of ∼20% of proteasome-regulated proteins (Petroski and Deshaies, 2005; Soucy et al., 2009). In the process of DSB repair, NEDD8 is accumulated at DNA damage sites (Ma et al., 2013). Moreover, inhibition of neddylation hypersensitizes human cells to DNA damaging agents, such as mitomycin C, cisplatin and ionizing radiation (Brown and Jackson, 2015). Hence, clarifying the regulatory mechanism of neddylation could provide insight into the genome stability underlying the DDR.

Dual-specificity tyrosine-regulated kinases (DYRKs) are a family that belongs to the CMGC group, which includes cyclin-dependent kinases (CDKs), mitogen-activated protein kinase (MAPKs), glycogen synthase kinases (GSKs) and CDK-like kinases (CLKs) (Becker and Sippl, 2011). In human cancer cells, we have identified DYRK2 as a regulator of p53-induced apoptosis in response to DNA damage (Taira et al., 2007) and of G1/S transition (Taira et al., 2012). Numerous studies have demonstrated that DYRK2 is downregulated in various tumors, such as tumors of the breast (Mimoto et al., 2017), colon (Ito et al., 2017; Kumamoto et al., 2020), brain (Shen et al., 2017), liver (Yokoyama-Mashima et al., 2019), lung (Yamashita et al., 2009a,b) and prostate (Taira et al., 2012). Importantly, this low DYRK2 expression is correlated with a poor prognosis (Enomoto et al., 2014; Mimoto et al., 2017; Taira et al., 2012; Zhang et al., 2016; Yoshida and Yoshida, 2019), indicating the tumor-suppressive functions of DYRK2. More recently, in normal cells, but not in tumor cells, whole-genome RNA sequencing of primary mouse embryonic fibroblasts (MEFs) derived from Dyrk2^−/− mice has shown that these cells exhibit downregulation of genes related to cell division and the mitotic cell cycle checkpoint (Yoshida et al., 2020). These findings led us to speculate that DYRK2 plays other important roles in maintaining the homeostasis of normal cells and suppressing tumorigenesis.

In the present study, we investigated the role of DYRK2 in genome stability. Here, we demonstrate that DYRK2 is a novel regulator of neddylation and that is acts to maintains genome stability.

To investigate the involvement of DYRK2 in genome stability, we perform immuno-cytostaining for γH2A.X (phospho-histone variant H2A.X at Ser139), which is the earliest response marker of DSBs (Branzei and Foiani, 2008), in primary MEFs derived from wild-type and Dyrk2^−/− mice (Yoshida et al., 2020). Higher levels of γH2A.X foci were observed in Dyrk2^−/− MEFs (Fig. 1A,B). To validate whether this genome instability observed in Dyrk2^−/− MEFs is conserved in other cell types, we knocked out DYRK2 in immortalized human retinal pigment epithelia (hTRET-RPE1) cells by CRISPR/Cas9 technology (Katoh et al., 2017). Knockout of DYRK2 (DYRK2^−/−) in hTRET-RPE1 cells also caused a marked increase of γH2A.X foci (Fig. 1C,D).

DSBs trigger the activation of the ATM–cell cycle checkpoint kinase 2 (CHK2; also known as CHEK2) pathway, which is one of the major DDR pathways (Ahn et al., 2000, 2002). Deletion of DYRK2 showed no effects on total ATM and CHK2, but significantly promoted phosphorylation of ATM (Ser1981) and CHK2 (Thr68) (Fig. 1E; Fig. S1). Persistent activation of the DDR causes micro-nuclei to form (Takahashi et al., 2018). Knockout of DYRK2 significantly induced formation of micro-nuclei in hTRET-RPE1 cells (Fig. 1F,G) and also in Dyrk2^−/− MEFs (Fig. 1H,I). These data demonstrate that deletion of DYRK2 induces genome instability via persistent DSBs and activation of the ATM–CHK2 pathway.

The activation of the DDR orchestrates the detection and repair of DNA damage with transient cell cycle arrest to ensure maintenance of genome stability and also induce senescence or apoptosis (Jackson and Bartek, 2009). To investigate the effects of DYRK2 in cell cycle, hTRET-RPE1 cells were synchronized at G0 phase by serum starvation for 24 h and then re-addition of serum (Fig. 2A). Cells progressed to the S-G2-M phase over time after serum had been added back. A transient knockdown by means of siRNA against DYRK2 (siDYRK2), however, induced a delay in cell cycle re-entry (Fig. 2B,C). The protein level of G1/S markers [cyclin D1 and p27 (encoded by CDKN1B)] also showed that there was a delay in cell cycle re-entry in hTRET-RPE1 cells (Fig. 2D). Cells undergoing senescence show cell cycle arrest concurrently with metabolic changes, including production of senescence-associated β-galactosidase (SA-β-gal) (Coppé et al., 2008). Under serum starvation, the proportion of SA-β-gal-positive cells was significantly higher in DYRK2^−/− hTRET-RPE1 cells than in wild-type cells (Fig. 2E,F). Similar results were observed in Dyrk2^−/− MEFs (Fig. 2G). Taken together, these results demonstrate that depletion of DYRK2 induces G0/1 phase cell cycle arrest and initiation of cellular senescence.

To elucidate the molecular mechanisms of G0/1 phase cell cycle arrest and initiation of cellular senescence caused by genome instability, we focused on factors involved in cellular stress and cell cycle arrest. Especially, we investigated the relationship between DYRK2 and the DDR factor p53, which is activated via stabilization and translocation to nuclei by cellular stress, including DSBs (Branzei and Foiani, 2008; Hafner et al., 2019). We found that protein levels of p53 and its downstream factor, p21, were increased (Fig. 3A,B) and accumulated in nuclei in DYRK2^−/− hTRET-RPE1 (Fig. 3C). Gene expression of CDKN1A (encoding p21) was upregulated in DYRK2^−/− hTRET-RPE1 cells (Fig. S2A). Phosphorylation of p38 MAPKs, which respond to cellular stress and link to the cell cycle through senescence and differentiation (Yee et al., 2004), was also increased in DYRK2^−/− hTRET-RPE1 cells (Fig. 3A). Similar phenotypes were observed in Dyrk2^−/− MEFs (Fig. S2B,C) and a transient knockdown by means of siDYRK2 in hTRET-RPE1 cells (Fig. S2D). Under basal conditions, the protein level of p53 is maintained at low levels because it is degraded rapidly following ubiquitylation by the associated ubiquitin ligase, MDM2 (Shieh et al., 1997; Siliciano et al., 1997). Under stress conditions, however, p53 is stabilized by phosphorylation at Ser15, which prevents the recruitment of MDM2. The protein level of phospho-p53 (Ser15) was increased in DYRK2^−/− hTRET-RPE1 cells (Fig. 3D) indicating that stabilization of p53 is, at least in part, caused by inhibition of interaction with MDM2.

We subsequently confirmed whether p53 regulates p21 and p38 MAPKs. The expression of p21 is regulated at the transcription level by direct p53 binding to p53 response elements in the promoter of the p21 gene (Abbas and Dutta, 2009). Transient knockdown by means of siRNA against TP53 (siTP53) repressed protein levels of p21 and p38 MAPKs in DYRK2^−/− hTRET-RPE1 cells (Fig. 3E). In contrast, protein levels of p53 and p21 remained unchanged upon treatment with SB203580 (a p38 MAPK inhibitor) in DYRK2^−/− hTRET-RPE1 cells (Fig. S2E). Collectively, these data show that the deletion of DYRK2 induces p53 stabilization and activation to induce p21 transcription and p38 MAPK activation.

To investigate how DYRK2 suppresses persistent DSBs, we focused on protein neddylation, which plays important roles during DNA damage signaling (Ma et al., 2013; Wang et al., 2017). We found that protein levels of NAE1 and UBA3, which compose the E1 enzyme in this process (see Introduction), were decreased in DYRK2^−/− hTRET-RPE1 cells (Fig. 4A). Gene expression analysis demonstrated no changes in the gene expression levels of NAE1 and UBA3 in DYRK2^−/− hTRET-RPE1 cells (Fig. S3). These data indicate that the deletion of DYRK2 reduces the stabilization of NAE1 and UBA3 in a posttranslational manner. Notably, NEDD8–cullins (neddylated cullins) were suppressed in DYRK2^−/− hTRET-RPE1 cells (Fig. 4A). To verify that DYRK2 is involved in the stabilization of NAE1 and UBA3, we conducted a transient overexpression experiment with wild-type human DYRK2 or a DYRK2-K251R construct, which expresses a kinase dead mutant (Taira et al., 2012) in DYRK2^−/− hTRET-RPE1 cells by adenovirus infection (Yokoyama-Mashima et al., 2019; Yoshida et al., 2020). Overexpression of the wild-type DYRK2 construct, but not the DYRK2-K251R mutant, restored protein levels of NAE1, UBA3 and NEDD8–cullins (Fig. 4B).

To understand the mechanisms underlying the suppression of E1 ligases in DYRK2^−/− hTRET-RPE1 cells, we analyzed protein–protein interaction between DYRK2 and the E1 enzyme by immunoprecipitation. The result demonstrates that DYRK2 interacts with NAE1 but not with UBA3 (Fig. 4C). The DYRK2-K251R mutant also interacted with NAE1; however, this interaction was markedly less compared to that of wild-type DYRK2 (Fig. S4). Moreover, a transient knockdown by siDYRK2 induced polyubiquitylation of NAE1 (Fig. 4D). These data indicate that DYRK2 stabilizes the E1 enzyme in a kinase-activity-dependent manner by binding to NAE1 and suppressing its ubiquitylation.

Finally, we examined whether overexpression of NAE1 and UBA3, which is suppressed in DYRK2^−/− cells, restores genomic stability. In DYRK2^−/− hTRET-RPE1 cells, immunocytostaining for γH2A.X demonstrated that NAE1 and UBA3 double-positive cells had decreased levels of γH2A.X foci (Fig. 4E,F). Taken together, these findings indicate that DYRK2 maintains the level of neddylation by binding to the E1 enzyme and contributes to the maintenance of genomic stability.

The present study demonstrates that the loss of DYRK2 suppresses neddylation and causes persistent DSBs and DDRs, such as cell cycle arrest and cellular senescence, through the activation of p21 and p38 MAPKs via p53 (Fig. 5). Mechanistically, DYRK2 promotes neddylation via forming a complex with NAE1 and maintains the protein levels of the E1 enzyme (Fig. 5). Ultimately, NEDD8 conjugates mainly with cullins and activates cullin–RING ligases (CRLs), which regulates ∼20% of the degradation of proteasome-regulated proteins and promotes DDR (Petroski and Deshaies, 2005; Soucy et al., 2009).

Recently, MLN4924, which is a selective small-molecular inhibitor of the NEDD8-activating enzyme, has been found to suppress E1 enzyme activity and the neddylation of cullins (Brownell et al., 2010). Treatment with MLN4924 causes accumulation of CRL substrates and DSB formation, which consequently induce persistent activation of DDRs and cellular senescence (Milhollen et al., 2011; Jia et al., 2011). Similarly, our present data indicate that deletion of DYRK2 shows this typical phenotype when neddylation is suppressed by treatment with MLN4924. Taken together with the finding that overexpression of NAE1 and UBA3 restores genome stability in DYRK2^−/− cells, we conclude that DYRK2 contributes to genome stability via regulation of neddylation. To elucidate the function of DYRK2 in NAE1 stabilization, we focus on the ubiquitin-proteasome system. Deletion of DYRK2 promotes poly-ubiquitylation and decreases the protein level of NAE1, implying that DYRK2 stabilizes NAE1 through the inhibition of ubiquitylation and subsequent proteasomal degradation. Although mechanisms underlying the suppression of NAE1 ubiquitylation via interaction with DYRK2 in a kinase-dependent manner remain unclear, the present study is the first to demonstrate that the protein level of NAE1 is regulated by inhibition of polyubiquitylation.

To protect against a variety of stresses, cells keep cellular homeostasis via cell arrest, senescence, and apoptosis (Childs et al., 2014). We have reported that ATM phosphorylates DYRK2 at Thr33 and Thr369 under DNA damage and that DYRK2 subsequently phosphorylates p53 at Ser46 to induce apoptosis (Taira et al., 2007). This induction of apoptosis contributes to the elimination of cells with irreparable DNA damage (Coates et al., 2005). DYRK2 also directly interacts with ring finger protein 8 (RNF8), which catalyzes Lys63-linked ubiquitylation of histone H2A.X in response to genotoxic stress (Yamamoto et al., 2017). Interestingly, DYRK2 directly phosphorylates telomerase reverse transcriptase (TERT) and induces subsequent ubiquitylation and degradation as a scaffold for E3 isolated by the differential display (EDD; also known as UBR5)–DNA-damage binding protein 1 (DDB1)–Vpr-binding protein (VprBP; also known as DCAF1) (EDVP) E3 ligase complex (Jung et al., 2013). TERT, a component of telomerase, which adds a chromosomal end structure, contributes to genomic stability (Cech, 2004; Negrini et al., 2010). Taken together with the present finding that DYRK2 maintains genomic stability via neddylation, DYRK2 functions in cellular homeostasis in a multistep manner.

Numerous studies indicate that neddylation dysfunction causes several human diseases (Zhang et al., 2021; Shukla et al., 2020; Jiang and Jia, 2015). Especially in cancers, the components of CRLs are overexpressed or mutated, and many CRLs regulate the activity of proteins that act as tumor suppressors or oncoproteins (Watson et al., 2011). In TP53-deficient hTERT-RPE1 cells, suppression of neddylation has recently been found to promote oncogenic proliferation and play an important role in tumorigenesis (Drainas et al., 2020). In this context, DYRK2 induces apoptosis via phosphorylation of p53 (Taira et al., 2007) and suppresses the cell cycle via oncogenic factors c-Jun and c-Myc (Taira et al., 2012). Given the findings that DYRK2 is downregulated in various cancer cells, DYRK2 plays a tumor-suppressive function (Enomoto et al., 2014; Mimoto et al., 2017; Taira et al., 2012; Zhang et al., 2016; Yoshida and Yoshida, 2019). In the present study, loss of DYRK2 induced γH2A.X foci, indicating genome instability via persistent DSBs. Genome instability via abnormalities of DNA repair or sensing processes, or both, induces carcinogenesis (Deman and Van Larebeke, 2001; Kinzler and Vogelstein, 1997; Loeb, 1994). Although the present study does not provide direct evidence that deletion of DYRK2 causes tumorigenesis, genome stabilization via DYRK2-mediated neddylation might contribute to the suppression of carcinogenesis.

The present study is the first to demonstrate that DYRK2 regulates neddylation and is necessary for maintaining genome stability. Mechanistically, DYRK2 promotes neddylation by forming a complex with NAE1 and maintains NAE1 protein levels by suppressing polyubiquitylation. These findings support further understanding of the maintenance of genomic stability and the mechanisms of the carcinogenesis suppression via neddylation.

Full-length cDNA fragments of human DYRK2, NAE1, UBA3 and ubiquitin were amplified by PCR and cloned in frame into pcDNA3+FLAG, pcDNA3+HA and pEGFP-C1 (Takara Bio Inc., Otsu, Japan) using NEBuilder HiFi DNA Assembly Master Mix (New England Biolabs, Ipswich, MA, USA). The nucleotide sequences of the primers used are listed in Table S1.

Cell culture was performed as described in a previous study (Yoshida et al., 2020). Primary MEFs were generated from wild-type and Dyrk2^−/− mouse embryos at E13.5 (Yoshida et al., 2020). The MEFs and immortalized human retinal pigment epithelia cells (hTERT-RPE1, ATCC, Manassas, VA, USA) and the Lenti-X 293T cell line (Takara Bio Inc.) were cultured in Dulbecco's modified Eagle's medium (Nacalai Tesque, Kyoto, Japan) with 10% fetal bovine serum (Biowest, Nuaillé, France), 1% GultaMAX (Gibco, Gaithersburg, MD, USA), and 1% penicillin-streptomycin (Nacalai Tesque) at 37°C under 5% CO₂. For SB203580 stimulation, cells were treated with 0.5 µM SB203580 (Selleck Biotech, Tokyo, Japan) for 72 h after serum starvation. Transient knockdown was achieved with the Lipofectamine RNAiMAX transfection regent (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer's instructions with a final concentration of 20 nM siRNA (Table S2). For transient overexpression, transfection was performed with the reagent X-tremeGENE^TM9 (Merck KGaA, Darmstadt, Germany) for hTERT-RPE1 cells and with Polyethylenimine ‘Max’ (Polysciences, Warrington, PA, USA) for Lenti-X 293T cells. For MG-132 treatment, Lenti-X 293T cells were incubated with 5 µM MG-132 (Merck) for 4 h.

CRISPR/Cas9-mediated DYRK2 knockout hTERT-RPE1 cells were made using homology-independent repair performed according to a previous report (Katoh et al., 2017). Briefly, two independent single-guide RNA (sgRNA) sequences targeting the human DYRK2 gene (#1: 5′-CCTGGATCTGTCCGTGAGCG-3′) and (#2: 5′-GAGCCCGGTAAAAACGCGAC-3′) were designed and inserted into peSpCAS9(1.1)-2×sgRNA (Addgene plasmid #80768). The hTERT-RPE1 cells were transfected with the sgRNA vector and the pDonor-tBFP-NLS-Neo (Universal) donor knock-in vector (Addgene plasmid #80767) via a X-tremeGENE^TM9 transfection reagent (Roche Applied Science), selected by 600 μg/ml G418, and subcloned.

Adenovirus construction and infection were performed as described in a previous study (Yoshida et al., 2020). Briefly, Flag–DYRK2 and Flag–DYRK2-K251R (Mimoto et al., 2013; Taira et al., 2010; Yokoyama-Mashima et al., 2019) were expressed depending upon Cre expression.

For immunocytochemistry, cells were cultured on eight-well chamber slides (Thermo Fisher Scientific) coated with Poly-D-lysin (Sigma-Aldrich, St Louis, MO, USA). Cells were fixed and antigen-retrieved depending on the antibody. The primary antibody reaction was performed at an appropriate dilution (Table S2) in the presence of a blocking buffer at 4°C overnight. After immunoreactions, cells were incubated with secondary antibodies via Cy3- or Cy5-conjugated AffiniPure donkey anti-mouse and rabbit IgG (Jackson ImmunoResearch, West Grove, PA, USA). The cells were then washed and incubated with DAPI (Vector Laboratories, Burlingame, CA, USA). Immunofluorescence was observed under a BZ-X800 fluorescence microscope (Keyence, Osaka, Japan).

Cells were washed twice in chilled phosphate-buffered saline and lysed with a RIPA buffer containing inhibitors (1 mM PMSF, 10 µg/ml aprotinin, 1 µg/ml leupeptin, 1 µg/ml pepstatin A, 1 mM Na₃VO₄, 10 mM NaF and 1 mM dithiothreitol). Equal amounts of protein (5 µg) were separated by SDS-PAGE and transferred to PVDF membranes (Merck). Membranes were blocked with 0.1% casein/gelatin in Tris-buffered saline with Tween-20. Primary and secondary antibodies (Table S2) were reacted with in each blocking buffer. Signals were detected with a chemiluminescent reagent, ImmunoStar LD (Wako Pure Chemical Industries, Ltd., Osaka, Japan) or Western Lightning Plus ECL (PerkinElmer, Waltham, MA, USA). Signals were observed and band intensity was measured with a Fusion-Solo system (M&S Instruments, Tokyo, Japan). Full blot images for images in the figures are shown in Fig. S5.

Cell fractionation was performed with a cell fractionation kit (Abcam, Cambridge, UK) according to the manufacturer's instructions. Briefly, cells were resuspended in buffer A and permeabilized with detergent I. The cell suspension was then centrifuged at 10,000 g for 2 min. The resulting supernatant was collected as the cytosol fraction. The cytosol-depleted pellet was resuspended in buffer A and solubilized with detergent II. Following centrifugation at 10,000 g for 2 min, the supernatant was collected as the mitochondria-enriched fraction. The cytosol- and mitochondria-depleted pellet was resuspended in buffer A, and used as the nuclear fraction.

For immunoprecipitation, cells transfected with indicated plasmids were lysed with NP40 buffer (25 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM EDTA, 1.0% NP40 and 5% glycerol) containing inhibitors (1 mM phenylmethylsulphonyl fluoride, 10 µg/ml aprotinin, 1 µg/ml leupeptin, 1 µg/ml pepstatin A, 1 mM Na₃VO₄ and 10 mM NaF) on ice for 20 min, after which the lysates were centrifuged at 20,000 g for 20 min at 4°C. The supernatants were collected and Flag M2 affinity beads (Sigma-Aldrich) were added and incubated at 4°C for 90 min. The beads were collected by centrifugation at 2500 g for 1 min and washed and boiled with 2× SDS sample buffer at 95°C for 5 min.

Cells were fixed with 70% ethanol and stored at −20°C for no longer than 24 h. Residual ethanol was eliminated from the cells being centrifuged (300 g for 3 min) and washed twice with fluorescence-activated cell sorter (FACS) buffer (2% fetal bovine serum and 1 mM EDTA in phosphate-buffered saline). For cell cycle analysis of the G0/1 to S phases, cells were re-stimulated with a medium containing 600 ng/ml nocodazole (Tocris Bioscience, Bristol, UK) to arrest dividing cells in metaphase. Cells were stained with 2 µl/ml Hoechst 33342 (AdipoGen Life Sciences, San Diego, CA, USA) and incubated for 20 min at room temperature. Stained cells were analyzed with a MACSQuant analyzer (Miltenyi Biotec, Tokyo, Japan). Data were analyzed with the software application FlowJo (Tomy Digital Biology, Tokyo, Japan).

Immunocytostaining for senescence β-galactosidase (SA-β-gal) was performed with a Senescence β-Galactosidase Staining Kit (Cell Signaling Technology, Danvers, MA, USA) according to the manufacturer's introductions. To induce cellular senescence, cells were serum-starved for 72 h and then fixated with a fixative solution for 15 min. Cells were washed twice in HEPES buffer and incubated with β-gal staining solution at 37°C overnight in a dry incubator. Signals were observed under a BZ-X800 microscope (Keyence). To conduct FACS of β-gal-labeled cells, cells were fixed with 4% formaldehyde at room temperature for 10 min. Senescence detection was performed with the Cell Event™ Senescence Green Detection Kit (Invitrogen, Waltham, MA, USA) according to the manufacturer's introductions. Stained cells were analyzed with a MACSQuant analyzer (Miltenyi Biotec), and data were analyzed with FlowJo (Tomy Digital Biology).

Isolation of total RNAs was performed with the RNeasy Mini Kit (Qiagen, Germantown, MD, USA). Reverse transcripts were obtained with PrimeScript Reverse Transcriptase (Takara Bio Inc.) and subjected to quantitative PCR (qPCR) with the PikoReal 96 system (Thermo Fisher Scientific). Reactions were performed in KAPA SYBR FAST qPCR Master Mix (Nippon Genetics, Tokyo, Japan) that included 0.2 μM of a specific primer set for each gene (Table S1). Data were calculated with the comparative CT method (ΔCT method) to estimate the mRNA copy number relative to that of HPRT1 as an internal standard. The DNA sequence of the PCR product was confirmed with nucleotide sequencing (data not shown).

Each experiment was confirmed by at least three independent technical replicates per condition. Data are presented as the means±s.e.m. The Prism 7 software program (GraphPad, San Diego, CA, USA) was used for statistical analyses. Means between groups were compared with a paired two-tailed Student's t-test. Multiple intergroup differences were analyzed with one-way ANOVA followed by Tukey's multiple comparison test.

We would like to thank Dr. Yohei Katoh and Dr. Kazuhisa Nakayama at Kyoto University for technical supports to establish knockout cell lines by the CRISPR-Cas9 system. We also would like to thank Naoko Tago for technical supports of FACS analysis and Prof. Masao Okazaki for English proofreading.

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.259514.