KIF24 depletion induces clustering of supernumerary centrosomes in PDAC cells

Depletion of the centrosomal kinesin KIF24, known to restrain the assembly of primary cilia, suppresses multipolar spindle formation by clustering centrosomes in centrosome-amplified PDAC cells.


Introduction
The centrosome comprises centrioles and a pericentriolar matrix (PCM). The two cylinder-like centrioles in the G1 phase are duplicated through the S-G2 phase, and the two-paired centrioles ensure bipolar spindle formation during mitosis. As new centrioles are duplicated from the older centriole, the differentially aged centrioles in the G1 phase are termed the older mother and the younger daughter centrioles. The mother centriole, characterized by the distal and sub-distal appendages, nucleates the primary cilium during the interphase, generally in the G0 phase. This single hair-like organelle is ubiquitously expressed in mammalian cells containing specific receptors and channels to receive multiple signaling molecules. Given that both the spindle and primary cilium share the centriole for their assembly, they are thought to be incompatible with each other in normal somatic cells Sánchez & Dynlacht, 2016).
Contrary to normal cells, the numbers of centrosomes and primary cilia are anomalous in many cancer types. The number of primary cilia decreases in various cancers, which probably causes aberrant signal transduction and cell cycle progression (Eguether & Hahne, 2018;Liu et al, 2018). Supernumerary centrosomes are also observed in cancers that potentially nucleate multipolar spindles during mitosis, thereby leading to cell death. Despite this, the cancer cells can often avoid the detrimental multipolar spindles by forming pseudo bipolar spindles in which the multiple centrosomes are clustered into one spindle to allow bipolar separation, a process termed centrosome clustering (Funk et al, 2016;Milunović-Jevtić et al, 2016;Levine & Holland, 2018). Pseudo bipolar mitosis not only allows completion of cell division and survival of the daughter cells but also occasionally induces certain tumor cell hallmarks, such as chromosome segregation errors, aneuploidies, and invasion (Ganem et al, 2009).
Pancreatic ductal adenocarcinoma (PDAC) accounts for over 90% of pancreatic cancers, with a 5-yr survival rate of less than 10% (Adamska et al, 2017). Similar to other cancers, primary cilia are decreased in PDAC lesions and cell lines (Seeley et al, 2009;Schimmack et al, 2016). Most PDAC samples harbor actively mutated oncogenic KRAS, and KRAS signaling is known to inhibit primary ciliogenesis in PDAC (Seeley et al, 2009). We have previously revealed that KRAS and a histone deacetylase HDAC2 suppress primary ciliation by regulating Aurora A kinase (AurA) expression in PDAC cells . Furthermore, the suppression of ciliogenesis has been reported to promote the transformation of normal pancreatic ductal cells into cancer cells (Deng et al, 2018). Centrosomal amplification and clustering also occur in PDAC (Sato et al, 1999;Zhu et al, 2005;Mittal et al, 2015;Ansari et al, 2018); however, the molecular mechanisms underlying the centrosome clustering remain largely unclear.
Kinesin family member 24 (KIF24) belongs to the Kinesin-13 subfamily with a kinesin domain in its middle region (Miki et al, 2005). KIF24 localizes to centrioles and interacts with two centriolar proteins, CP110 and CEP97, which prevent the assembly of the primary cilia in cycling mammalian cells . KIF24 likewise suppresses cilia formation through the dual roles in which KIF24 recruits CP110 to the mother centriole and antagonizes the extension of the ciliary axoneme with its microtubule (MT)-depolymerizing activity in non-transformed cells The indicated Panc1 cells were cultured and immunoblotted as described in Fig 1A. (E) The indicated Panc1 cells were cultured and immunostained as described in Fig 1B. The percentage of . NEK2 phosphorylates KIF24, which enhances MT-depolymerizing activity (Kim et al, 2015). MPP9 also forms a complex with CP110-CEP97-KIF24 to suppress ciliogenesis (Huang et al, 2018).
In this study, we generated KIF24-depleted PDAC cells to assess whether the forced expression of primary cilia restrained the proliferation of PDAC cells. Unexpectedly, we found that KIF24depleted cells exhibited vigorous proliferation compared with control cells. KIF24 depletion was found to suppress the formation of multipolar spindles by clustering excess centrosomes and to improve mitotic progression in PDAC cells. On the other hand, the elimination of primary cilia in KIF24-depleted cells failed to affect both the proliferation and centrosome clustering. Mechanistically, NEK2-mediated phosphorylation and MT-depolymerizing activity were dispensable for the mitotic function of KIF24. The inhibition of HSET/KIFC1, which promotes the formation of centrosome clusters (Kleylein-Sohn et al, 2012), induced the de-clustering of centrosomes in KIF24-depleted cells. Moreover, KIF24 depletion specifically blocked the formation of multipolar spindles and induced hyper-proliferation in PDAC cells harboring supernumerary centrosomes. These results represent a novel function of KIF24 in the centrosome clustering, irrespective of primary ciliation in centrosome-amplified PDAC cells.

KIF24 depletion induces primary cilia in PDAC cells
To examine whether KIF24 contributes to the suppression of primary ciliogenesis in PDAC cells, we used Panc1 cells, which were originally isolated from a PDAC patient and assemble primary cilia with low frequency (Lieber et al, 1975;Nielsen et al, 2008). Because siRNA-mediated acute knockdown of KIF24 induced the formation of primary cilia in Panc1 cells (Fig S1A and B), KIF24-mutated Panc1 cells were subsequently generated using the CRISPR/Cas9 method. The sequencing analysis indicated four distinct mutations in Kif24mutated cells (named Kif24-3 cells), leading to premature stop codons in three alleles (allele A-C) but an amino acid deletion in one allele (allele D) (Fig S1C), suggesting that KIF24 was not completely knocked out in Kif24-3 Panc1 cells. A rescue clone was further established in which ectopic KIF24 was stably expressed in Kif24-3 cells (Kif24-3_KIF24), and control clones harboring the empty vector (EV) (Panc1_EV and Kif24-3_EV) were also generated. Western blotting analysis showed KIF24 expression in Panc1_EV and Kif24-3_KIF24 cells but a substantial decrease in KIF24 in Kif24-3_EV cells (Fig 1A), indicating that KIF24 is drastically depleted in Kif24-3 cells. The assembly of primary cilia was assessed in these cells by immunofluorescence experiments using an anti-ARL13B antibody that specifically recognizes the ciliary membrane. Kif24-3 cells formed significantly more primary cilia than Panc1_EV and Kif24-3_KIF24 cells in both serum-fed (FBS+) and -deprived (FBS−) media ( Fig 1B and C), suggesting that KIF24 suppresses primary ciliogenesis in Panc1 cells. In addition, Panc1 cells stably expressing shRNA targeting KIF24 were generated (Figs 1D and S1D). The percentage of cells with primary cilia was significantly increased in two individual shKIF24-expressing clones compared with shControl cells (Fig 1E). Furthermore, siRNA-mediated silencing of KIF24 in another PDAC cell line, CFPAC1, also induced ciliation (Fig S1A and B). These results collectively indicate that KIF24 negatively controls the assembly of primary cilia in PDAC cells, as expected from previous reports Kim et al, 2015).

Loss of KIF24 enhances the proliferation of Panc1 cells in vitro and in vivo
The proliferation of KIF24-depleted cells was evaluated in vitro. As primary cilia appear to inhibit cell division, KIF24-depleted cells were predicted to exhibit slower growth because of an increase in primary cilia. However, to our surprise, both KIF24-mutated and KIF24-knockdown (KD) cells grew more vigorously than control cells (Fig 2A and B). Moreover, nuclear expression of Ki67, a marker of proliferating cells, was up-regulated in the KIF24-depleted cells ( Fig 2C). These results suggest that KIF24 depletion promotes the proliferation of Panc1 cells.
Next, tumor formation in KIF24-mutated Panc1 cells was evaluated in a mouse xenograft model. Kif24-3-tumors were larger than tumors derived from parental cells at early stages (4 and 6 wk after injection) ( Fig S2A). However, the differences in the tumor volume were not significant at later stages (8-14 wk after injection). Similarly, the excised Kif24-3 tumors weighed moderately heavier than the WT tumors (14 wk after injection) (Fig S2B and C). These results suggest that KIF24 depletion tends to accelerate tumor formation of Panc1 cells in vivo. Subsequently, the expression of primary cilia was examined in tumor slices. Consistent with previous results of in vitro cultivation (Fig 1B and C), more primary cilia were observed in the slices of the Kif24-3 tumors than in the WT tumors (Fig S2D and E). Altogether, these results suggest that KIF24 depletion in Panc1 cells induces modest enhancement of tumorigenesis and primary ciliogenesis in vivo.

Primary cilia are irrespective of over-proliferation in KIF24-mutated cells
To test whether primary cilia are related to the hyper-proliferation of KIF24-depleted cells, an intraflagellar transport (IFT) protein IFT88, which is essential for cilia formation, was stably knocked down ( Fig 2D). If forced ciliation by KIF24 loss is linked to the overproliferation of Panc1 cells, de-ciliation might affect the overproliferation. In contrast to the substantial decrease in primary ciliation (Fig 2E), cell growth was not affected by silencing of IFT88 in WT and Kif24-3 cells ( Fig 2F). To further confirm the primary ciliaindependent hyper-proliferation of Kif24-3 cells, cells were treated with chloral hydrate (ClHy), which is known to exclude primary cilia ciliated cells was determined. The average of three to four independent experiments is shown; >250 cells were scored each time. (C, E) All data are shown as mean ± SD. Two-tailed t test. **P < 0.01; *P < 0.05. Source data are available for this figure.  from the cell surface (Ho et al, 2013;Kobayashi et al, 2020). ClHy treatment significantly reduced ciliation but failed to impact growth in WT and Kif24-3 cells (Fig 2G and H). These results strongly indicate that KIF24 depletion induces over-proliferation of Panc1 cells independent of primary cilium assembly. As Hedgehog (Hh) signaling depends on primary cilia and is relevant in PDAC (Morris et al, 2010), we examined Hh signaling in KIF24-depleted cells. Ciliary localization of G-protein coupled receptor GPR161 and its reduction upon Smo agonist (SAG) stimulation were equally observed in Kif24-3 cells and WT or rescue cells (Fig S3A and B). These results suggest that primary cilia induced by KIF24 depletion are functional. Next, the expression of genes targeted by the Hh pathway was assessed using quantitative PCR (qPCR). Although the expression of patched 1/PTCH1 was elevated in Kif24-3 cells, the expression levels of the other genes were nearly similar in analyzed cells (Fig S3C), suggesting that the overproliferation of KIF24-depleted cells was not because of enhanced Hh signaling.

KIF24 depletion induces the centrosome clustering
KIF24 is known to localize to the centrosome not only during the interphase but also during mitosis in mammalian cells . We specifically detected KIF24 at the centrosome in Panc1 cells during mitosis (Fig S4A and B) and therefore focused on mitotic events in Kif24-3 cells. Immunofluorescence experiments to visualize mitotic morphologies showed that~20% of WT (Panc1_EV) cells form multipolar spindles (Fig 3A and B), which is consistent with a previous study showing that Panc1 cells frequently harbor amplified centrosomes and assemble multipolar spindles (Difilippantonio et al, 2009). In contrast, the proportion of mitotic cells with multipolar spindles was significantly decreased in Kif24-3_EV cells compared with Panc1_EV and Kif24-3_KIF24 cells (Fig 3A  and B). Next, we stained cells with antibodies recognizing centrin and γ−tubulin to detect centrioles and centrosomes, respectively. Remarkably, this experiment revealed a decrease in multipolar cells with a concurrent increase in pseudo bipolar cells in which more than two centrin dots were observed on one side of the spindle (=γ-tubulin dot[s]) in mitotic Kif24-3 cells (Fig 3C and D), indicating that the supernumerary centrosomes were tightly clustered in KIF24-mutated cells. We also observed more missegregated chromosomes in the anaphase in Kif24-3 cells (Fig 3E  and F), which is typically induced by pseudo bipolar spindles, supporting that the occurrence of mitotic cells with pseudo bipolar spindles was increased by KIF24 loss. Moreover, shKif24-Panc1 cells displayed more pseudo bipolar, rather than multipolar, spindles ( Fig 3G). In contrast, the number of cells with increased γ-tubulin dots (>2 γ-tubulin dots) was not altered in Kif24-3 cells (Fig S5A). In addition, the percentages of mitotic cells with overamplified centrioles (>4 centrin dots) or fragmented centrosomes (γ-tubulin foci without centrin) were comparable among the analyzed cells (Fig S5B and C). Collectively, these results suggest that KIF24 depletion specifically evokes centrosome clustering.
Subsequently, immunofluorescence studies were conducted in cells treated with shIFT88 or ClHy. Neither IFT88-KD nor ClHy treatment affected the number of cells with multipolar and pseudo bipolar spindles in Kif24-3 cells (Fig 3H and I). These data suggest that the centrosome clustering occurs irrespective of primary cilia in KIF24-mutated cells, which is consistent with the overproliferation of these cells.
NEK2-mediated phosphorylation and MT-depolymerizing activity are dispensable for the mitotic function of KIF24 The amino acid residues KEC (positions 483-485) are conserved in the Kinesin-13 family of proteins and are important for the MTdepolymerizing activity of KIF24 . In addition, the NEK2-mediated phosphorylation of Thr 622 and Ser 623 was found to enhance the MT-depolymerizing activity of KIF24 (Kim et al, 2015). KIF24/KEC483-485AAA (KEC)-or KIF24/TS622, 623AA (TS)-expressing Kif24-3 cells were generated to elucidate whether the modification and activation of KIF24 are involved in spindle morphology ( Fig 4A). Although KIF24/KEC and KIF24/TS failed to suppress the assembly of primary cilia (Fig 4B), these mutants significantly decreased pseudo bipolar formation and concomitantly increased multipolar formation in KIF24-depleted cells (Fig 4C). These results suggest that the MT depolymerizing activity and NEK2-mediated phosphorylation of KIF24 are dispensable for its mitotic function and further support our idea that centrosome clustering occurs independently of primary cilium assembly in KIF24 depleted cells.

HSET/KIFC1 inhibition suppresses centrosome clustering caused by KIF24 depletion
The mitotic kinesin HSET/KIFC1 promotes centrosome clustering (Kleylein-Sohn et al, 2012), and treatment with the allosteric HSET inhibitor CW069 provokes multipolar spindles in centrosomeamplified cells (Watts et al, 2013). These findings prompted us to examine whether HSET inhibition suppresses pseudo bipolar formation in KIF24-depleted cells. CW069 treatment clearly induced formation of multipolar spindles instead of pseudo bipolar assemblies in Kif24-3 cells, and the ratios were comparable in Pan-c1_EV and Kif24-3_KIF24 cells (Fig 5A). These results suggest that KIF24 functions upstream of HSET in the centrosome clustering cascade. On the other hand, HSET on the mitotic spindle microtubules was unaffected in KIF24-depleted cells (Fig 5B). KIF24 at spindles also remained unchanged by CW069 treatment (Fig 5C). These results suggest that KIF24 is unrelated to HSET localization and vice versa in mitosis.
shown; >250 cells were scored each time.  CP110 expression was also evaluated because the silencing of KIF24 decreases CP110 protein levels in normal diploid RPE1 cells but not in cancerous U2OS cells . We found that CP110 levels were not altered in Kif24-3 cells (Fig S5D), suggesting that, like in U2OS cells, CP110-dependent centriolar events are not affected by KIF24 depletion in Panc1 cancer cells. The centrosomal protein CENPJ/CPAP is involved in centrosome clustering in cancer cells (Mariappan et al, 2019) and antagonizes CP110 during the extension of centriole and cilia (Schmidt et al, 2009;Wu & Tang, 2012). Therefore, we examined CPAP foci at the centrosome in KIF24-depleted cells, but no alteration was detected (Fig S5E and F). These results suggest that KIF24 affects centrosome clustering independently of CP110 and CPAP.

KIF24 depletion promotes mitotic progression in Panc1 cells
Live-cell imaging of mitotic progression was subsequently performed in KIF24-mutated cells. To visualize the chromosome dynamics during mitosis, histone H2B-mCherry-stably expressing Panc1 or Kif24-3 cells were generated. Time-lapse imaging revealed that the mitotic duration from nuclear envelope breakdown to the anaphase onset was slightly but significantly shortened in Kif24-3 cells (Fig 6A and B). Moreover, the ratio of cells that entered mitosis but failed to divide was largely decreased in Kif24-3 cells (Fig 6A and  C). These results suggest that KIF24 depletion improves the mitotic progression in centrosome-amplified Panc1 cells, probably leading to their accelerated proliferation.

KIF24 depletion suppresses multipolar spindle formation and enhances cell growth specifically in centrosome-amplified PDAC cells
To test whether KIF24 controls centrosome clustering in other PDAC cells, shKIF24-expressing MiaPaCa2 or Hs766t cells were generated ( Fig S1C). A previous report indicated that centrosomes are amplified in Hs766t cells but not in MiaPaCa2 cells (Difilippantonio et al, 2009). These PDAC cells rarely assembled primary cilia even after KIF24 silencing, suggesting that these cells have essentially lost the ability to ciliate (Fig 7A). In MiaPaCa2 cells, both control and KIF24depleted cells assembled multipolar spindles with only~5% frequency (Fig 7B and C). In contrast, multipolar spindles were detected in~27% of control Hs766t cells, and this frequency was considerably reduced by KIF24 depletion (Fig 7B and C). Concomitant with multipolar spindle formation, KIF24-depleted Hs766t cells grew more vigorously than control Hs766t cells, in sharp contrast with MiaPaCa2 cells (Fig 7D). On the other hand, the number of cells with increased γ-tubulin dots was comparable between control and KIF24-depleted PDAC cells (Fig S5A). These results suggest that KIF24 induces multipolar spindle formation and thereby slow growth, specifically in centrosome-amplified PDAC cells.

Discussion
This study analyzed PDAC cells with supernumerary centrosomes revealing that KIF24 suppresses the clustering of excess centrosomes during mitosis and the assembly of primary cilia. NEK2-mediated phosphorylation and MT-depolymerizing activity were dispensable for the mitotic role of KIF24, unlike the ciliary role, indicating that distinct regulatory systems govern the dual functions of KIF24. These mitotic and ciliary phenotypes in cells depleted of KIF24 are thought to exert reciprocal effects on cell proliferation. KIF24-depleted PDAC cells overgrew in vitro, in which primary cilia are rarely assembled (~10% in Kif24-3 cells, Fig 1C), probably because of improved mitosis. The tumor development also occurred earlier with KIF24-depleted cells than with control cells in vivo (4-6 wk after injection), suggesting that the mitotic effect of KIF24 depletion promotes the tumor onset. In contrast, Kif24-3 tumors were not significantly larger than those derived from parental cells in the late stages (8-14 wk after injection). As more primary cilia were assembled in the excised tumors than in the cultured cells (~25% in Kif24-3 tumor slices, Fig S2E), it is plausible that these organelles opposed the long-term in vivo growth of Kif24-3 cells. Alternatively, aneuploidy provoked by chromosome mis-segregation in Kif24-3 cells may result in heterogeneous populations of cells with diverse proliferative profiles during long-term tumor growth. This notion is supported by data showing that Kif24-3 tumors exhibited larger variation than wild-type tumors. On the other hand, as the transient assembly of primary cilia can promote cell proliferation by activating the Hh pathway in medulloblastoma cells , we cannot exclude the possibility that primary cilia induced by KIF24 loss positively affects tumor growth in vivo, where implanted tissue is presumably exposed to Hh ligands. In that sense, an earlier onset of Kif24-3 tumors may be attributed to increased number of primary cilia. A clinical report showing that expression of primary cilia is correlated with poor prognosis of PDAC patients also suggests that primary cilia could promote growth in PDAC in vivo (Emoto et al, 2014). Given that primary cilia appear to be dynamically assembled and disassembled even in KIF24-depleted cells, perturbation of their disassembly may lead to different outcomes in PDAC growth.
Several MT-binding proteins are known to be involved in the regulation of multipolar spindle formation in cancer cells but not in PDAC cells. HSET promotes the clustering of amplified centrosomes in breast cancer and melanoma cells (Kleylein-Sohn et al, 2012). HSET interacts with a centrosomal protein CEP215 during its operation (Chavali et al, 2016). IFT-B proteins, which are essential for cilia assembly, are also associated with HSET to promote centrosome clustering in several centrosome-amplified cells (Vitre et al, 2020), whereas we did not detect any alterations of centrosome clustering in IFT88depleted Panc1 cells. Contrary to CEP215 and IFT-B proteins, KIF24 is unlikely to directly associate with HSET but probably acts upstream of HSET in PDAC cells based on our data. Another kinesin, KIF18A, facilitates the proliferation of colorectal and breast cancer cells through the regulation of centrosome fragment formation (Marquis et al, 2021). The association of CPAP with MT is required for centrosome clustering in breast and lung cancer cells (Mariappan et al, 2019). CPAP antagonizes CP110 in centriole elongation and is required for ciliogenesis (Schmidt et al, 2009;Wu & Tang, 2012). KIF24 cooperatively interacts with CP110 to suppress primary cilia formation . Although the expression of CP110 and CPAP was not affected by KIF24 depletion, future studies are needed to examine whether KIF24-CP110 is involved in the functionality of CPAP.
Although we showed that primary cilium assembly has no bearing on centrosome clustering in PDAC cells, the possibility that primary cilia or their remnants after resorption influence mitosis cannot be excluded. Pharmacological inhibition of mitotic kinases, Aurora A or PLK1, induces mitotic primary cilia in normal mouse IMCD3 cells (Bowler et al, 2019). It was recently reported that NEK2 knockout causes the expression of primary cilia or their remnants in mitotic RPE1 cells (Viol et al, 2020). As described above, NEK2 phosphorylates and activates KIF24 to disassemble primary cilia in the G2-M phase (Kim et al, 2015). Indeed, mitotic primary cilia in KIF24-mutated Panc1 cells were found; however, we did not observe detectable alterations in centrosome clustering (data not shown). It will be interesting to determine whether and how primary cilia influence various mitotic events in cancer cells.
Centrosome clustering is an attractive target for cancer therapy because it is frequently observed in cancer cells but not in normal cells. Based on our study, potentiation of the mitotic activity of KIF24 could be a valid intervention strategy for PDAC, although we need to consider that loss of primary cilia by KIF24 might aggravate this cancer at present. Future studies clarifying the mechanistic details of KIF24 in spindle formation will represent a promising therapeutic target to develop novel centrosome de-clustering drugs for PDAC.
Plasmid transfection into Panc1 and Lenti-X 293T cells was performed using Lipofectamine 2000 (Invitrogen) and PEI Max (Polysciences) according to the manufacturer's instructions, respectively.

Generation of Kif24-3 cells
The PX459-KIF24 plasmid was transfected into Panc1 cells using Lipofectamine 2000. Transfected cells were cultured in medium containing 5 μg/ml puromycin (Nacalai Tesque) for 72 h and singly plated into 96-well plates. Genomic DNA was extracted from surviving cells using QuickExtract DNA Solution 1.0 (Epicentre), and amplified PCR products using the primers listed in Table S1 were sub-cloned into pGEM-T Easy (Promega). The purified plasmid DNA was sequenced using M13 primers.

Immunofluorescence microscopy
Cultured cells were fixed with cold methanol for 5 min, 4% paraformaldehyde (Nacalai Tesque)/PBS for 10 min, or 3.7% formalin (Nacalai Tesque)/PBS for 10 min. After permeabilization with 0.2% Triton X-100/PBS for 10 min, the slides were blocked with 5% BSA/ PBS before incubation with the primary antibodies. Primary and secondary antibodies were diluted to the desired concentrations using 5% BSA/PBS. As secondary antibodies, we used Alexa Fluor 488-or Alexa Fluor 594-conjugated donkey anti-mouse, anti-rabbit, or anti-goat IgG (Invitrogen). The cells were stained with Hoechst 33342 to visualize DNA. Slides mounted with PermaFluor Mounting Medium (Thermo Fisher Scientific) were observed and imaged using Axio Observer with a 63× lens.
Tumors were fixed with 3.7% formalin/PBS at 4°C for 12 h; sequentially equilibrated with 10%, 20%, and 30% sucrose/PBS; and embedded with OCT compound (Sakura Finetek) at −80°C. The frozen tumors were sliced into 10 μm-thick sections using Cryostat NX70 (Thermo Fisher Scientific) and mounted on MAS-coated slide (Matsunami). The mounted sections were fixed with Acetone for 15 min, soaked into boiled water for 15 min to retrieve antigens, and permeabilized with 0.2% Triton X-100/PBS for 10 min. After permeabilization, procedures were same as those used for the cultured cells. Fluorescence was quantified using ImageJ/Fiji software (Kobayashi et al, 2020).

qPCR
Total RNA was isolated from cultured cells using Sepasol (Nacalai Tesque), and following reverse transcription reaction was performed using a ReverTra Ace qPCR RT kit (TOYOBO). qPCR was performed using THUNDERBIRD SYBR qPCR mix (TOYOBO) in a LightCycler96 (Roche). All reactions were conducted according to the manufacturer's instructions. The primers are listed in Table S1.

Live cell imaging
H2B-mCherry-expressing Panc1 or Kif24-3 cells were seeded into a CELLview Cell Culture Dish (Greiner) and treated with 2 mM thymidine for 40 h. At 9 h after thymidine washout, the cells were imaged using an LSM710 (Zeiss) operated by Zen software (Zeiss). Six Z-stack images were acquired every 5 min for 8 h using a 20× objective lens (Zeiss).
Xenograft 2 × 10 6 Panc1 cells in PBS were subcutaneously injected into 6-wkold female nude mouse (Nihon SLC). Tumors were measured each week using a caliper, and their volumes were calculated using the formula: length × width × height × 0.5. After 14 wk, the tumors were excised and weighed. All experiments were approved by the NAIST Animal Committee and conducted in accordance with the guideline of the NAIST animal facility.

Statistical analysis
The statistical significance of the differences was determined using two-tailed t test. The figure legends indicate the number of independent replicates conducted and the number of cells analyzed in each replicate. Differences were considered as significant when P < 0.05.