ATM- and ATR-Induced Primary Ciliogenesis Promotes Cisplatin Resistance in Pancreatic Ductal Adenocarcinoma

Pancreatic ductal adenocarcinoma (PDAC) is one of the most lethal cancers because of its late diagnosis and chemoresistance. Primary cilia, the cellular antennae, are observed in most human cells to maintain development and differentiation. Primary cilia are gradually lost during the progression of pancreatic cancer and are eventually absent in PDAC. However, recent study showed that primary cilia regrowth contributes to the development of diverse kinase inhibitor resistance in lung cancer. We elucidated the role of regrowth primary ciliogenesis in PDAC chemoresistance and uncovered the underlying molecular mechanism. We showed that cisplatin-resistant PDAC regrew primary cilia. Additionally, genetic or pharmacological disruption of primary cilia sensitized PDAC to cisplatin treatment. Mechanistically, ataxia telangiectasia mutated (ATM) and ATM and RAD3-related (ATR), tumor suppressors that initiate DNA damage responses, promoted the excessive formation of centriolar satellites (EFoCS) and autophagy activation. Disruption of EFoCS and autophagy inhibited primary ciliogenesis, sensitizing PDAC cells to cisplatin treatment. Collectively, our ndings revealed an unexpected interplay among DNA damage response, and

(PI3KK) superfamily, initiate the DNA damage response for cell cycle arrest, damage repair, or even apoptosis when damage is irreversible. Although ATM and ATR are long considered tumor suppressors, inhibition of ATM or ATR signaling ameliorates drug sensitivity, particularly in radioresistant and chemoresistant cancer cells [6,7]. Thus, ATM and ATR inhibitors have been developed to improve chemotherapeutic e ciency in cancer therapy [8,9].
Primary cilia are detected in most human cells; they are cellular antennae responsible for receiving and transducing environmental signals into cells. Thus, precise regulation of primary cilia is required for proper development and differentiation [10][11][12][13]. Loss of primary cilia occurs in the early stage of oncogenic transformation and tumorigenesis [14,15]. During the development of pancreatic tumors, the numbers of primary cilia decrease gradually from normal pancreatic ductal epithelium to pancreatic intraepithelial neoplasia (PanIN) [16]. Eventually, cilia are scarcely detected in PDAC. Although the loss of primary cilia is one of the hallmarks of PDAC, a recent study showed that the regrowth of primary cilia contributes to the development of diverse kinase inhibitor resistance in lung cancer [17]. Thus, primary cilia may play a role in modulating drug resistance.
Centriolar satellites are nonmembranous, electron-dense granules scattered around centrosomes. Because of a substantial overlap between the protein compositions and functions of satellites with centrosomes and cilia, centriolar satellites are considered parts of the centrosome or cilium complexes [18]. Pericentriolar material-1 (PCM1) acts as a scaffold protein to organize other components of satellites [19]. Centriolar satellites are moved dynamically along the microtubules by the dynein/dynactin complex. Following DNA damage, excessive formation of centriolar satellites (EFoCS) around centrosomes induces centrosome ampli cation [20]. Additionally, removing several components, such as CEP131, CEP290, and PCM1, from centriolar satellites promotes primary ciliogenesis [21]. However, the depletion of several centriolar satellite genes inhibits primary ciliogenesis [19,22]. Thus, the proper dynamics of centriolar satellites maintain primary ciliogenesis.
Autophagy is also linked to primary cilia formation. Inhibition of autophagy alleviates serum deprivationinduced primary cilia. Additionally, autophagy participates in chemotherapeutic resistance [23]. Autophagy maintains energy homeostasis by degrading and recycling macromolecules. During metabolic stress, AMP-activated protein kinase (AMPK) and UNC-51-like kinase-1 (ULK1) are activated for autophagy initiation [24]. ATG7-mediated lipidation of LC3 promotes the conversion of LC3-I to LC3-II for autophagosome formation [25]. After fusion with lysosomes, autolysosomes are formed, followed by degradation of the contents of autophagosomes. Although autophagy has been studied extensively, the interplay among the DNA damage response, autophagy, and chemoresistance is unclear.
PDAC cells were devoid of primary cilia. Here, we showed that following chemotherapeutic drug treatment, PDAC cells regrew primary cilia, leading to chemoresistance. Mechanistically, ATM and ATR induced EFoCS and activated autophagy, two events that cooperatively contributed to primary ciliogenesis. Thus, we have uncovered the novel function and underlying molecular mechanism by which primary cilia contribute to the development of chemoresistance in pancreatic ductal adenocarcinoma.

Results
Primary cilia contribute to cisplatin resistance in PDAC.
Primary cilia contribute to kinase inhibitor resistance in lung cancer; however, the underlying molecular mechanism remains unclear. Loss of primary cilia was observed in pancreatic ductal adenocarcinoma (PDAC); thus, we investigated whether primary cilia contribute to chemoresistance in PDAC and how primary ciliogenesis is regulated. To test our hypothesis, PANC-1 cells were treated with several chemotherapeutic drugs, and primary cilia were examined. Primary cilia were scarcely observed in PANC-1 cells. However, when cells were treated with cisplatin (CPT), gemcitabine (GEM), and etoposide (ETO), the proportion of cells with primary cilia increased, while paclitaxel (Taxol) did not affect primary ciliogenesis (Figs. 1A-B and S1A-B). Additionally, the cells grew primary cilia in a dose-and timedependent manner (Figs. 1C-E and S2A-C). A marker of ciliary axoneme, acetylated tubulin, was also increased following CPT and GEM treatment (Figs. 1F and S2D), suggesting that chemotherapeutic drugs induced primary cilia formation. The expression of ciliary genes, including KIF7, SCLT1, IFT43, RNF38, TOPORS, and C5orf30, increased signi cantly following CPT, GEM, and ETO treatment ( Fig. S3A-C). The structure of the primary cilium was further examined in detail by immuno uorescence staining. CEP164 (a distal appendage marker; Fig. 1G), acetylated tubulin (an axoneme marker; Fig. 1G-I), ARL13b (a ciliary membrane marker; Fig. 1H), and IFT88 (an intra agellar transporter marker; Fig. 1I) were detected in CPTinduced primary cilia, suggesting that these cilia contained essential ciliary components. Next, we tested whether CPT-resistant PANC-1 cells grew primary cilia. In parental cells, seeding cells at a low density of 5x10 4 did not induce primary ciliogenesis following CPT treatment, and the ability of PANC-1 cells to grow primary cilia increased in a density-dependent manner (Fig. 1J). Importantly, CPT-resistant PANC-1 cells at a density of 5x10 4 grew primary cilia (Fig. 1K), and these cilia contained essential ciliary components ( Fig. S4A-C). Thus, chemoresistant PANC-1 cells grew primary cilia.
Next, we tested whether primary cilia contributed to chemoresistance in PANC-1 cells. PANC-1 cells were transfected with siRNA against IFT88 for 72 h, followed by the examination of primary cilia and cell viability. The IFT88 abundance was reduced e ciently ( Fig. 2A), and IFT88 depletion inhibited CPT-or GEM-induced primary ciliogenesis (Figs. 2B and S5A). We then examined the role of primary cilia in chemoresistance. IFT88 knockdown did not affect cell viability but sensitized cells to CPT or GEM treatment (Figs. 2C-D and S5B-C). Importantly, IFT88 depletion reduced cell viability in CPT-resistant PANC-1 cells (Fig. 2E). To further con rm this nding, ve different shRNA sequences against IFT88 (shIFT88#1-5) were delivered into cells by lentivirus infection. All ve sequences reduced IFT88 expression, and shIFT88#3 had the highest knockdown e ciency (Fig. S5D). IFT88 depletion with shIFT88#3 reduced primary ciliogenesis (Fig. S5E) and sensitized cells to either CPT or GEM treatment ( Fig. S5F-H). This nding was further strengthened by treating cells with roscovitine, a known primary cilia inhibitor [26]. Following CPT or GEM treatment, roscovitine e ciently alleviated primary cilia formation (Figs. 2F and S5I). Additionally, roscovitine sensitized cells to CPT treatment (Figs. 2G) and reduced cell viability in CPT-resistant PANC-1 cells (Fig. 2H). The data suggest that primary cilia contribute to chemotherapeutic resistance in PANC-1 cells.
Next, the initiation of primary ciliogenesis was examined. Removal of CP110 from the distal end of the mother centriole initiates primary ciliogenesis following serum starvation [27]. Typically, two CP110 proteins cap the distal ends of both mother and daughter centrioles (Fig. S6A), and CP110 was displaced from the mother centriole for ciliogenesis during serum starvation (Fig. S6B). Interestingly, following CPT treatment, CP110 was removed from the mother centriole for ciliogenesis, but several CP110-aggregated puncta were scattered around the primary cilia (Fig. S6C). The data suggest that CP110 removal for ciliogenesis is triggered by serum starvation and CPT treatment; however, aberrant accumulation of CP110 around the cilia occurs in CPT-treated PANC-1 cells.
ATM and ATR contribute to primary ciliogenesis for cisplatin resistance.
Genotoxic stresses contribute to primary cilia formation via the DNA damage response in retinal pigmented epithelium [29]. Next, we tested whether the DNA damage response participated in primary ciliogenesis and chemoresistance in PANC-1 cells. CPT induced DNA damage, as shown by increased γ-H2AX (Figs. 4A and S8A). The DNA damage responses were then examined. CPT activated DNA damage signaling, including ATM, ATR, and DNA-PK (Figs. 4B). Depletion of ATM and ATR by transfecting cells with speci c siRNA alleviated CPT-induced primary cilia formation ( Fig. 4C-F). However, depletion of DNA-PK did not affect CPT-induced primary ciliogenesis ( Fig. S8B-C). Thus, we checked the effect of ATM and ATR on the chemosensitivity of PANC-1 cells. Depletion of either ATM or ATR had modest or no effects on cell viability but sensitized cells to CPT treatment ( Fig. 4G-J). Inhibition of ATM by its selective inhibitor Ku55933 also sensitized cells to CPT treatment (Fig. S8D). Importantly, depletion of ATM or ATR reduced cell viability in CPT-resistant PANC-1 cells (Fig. 4K). This nding was further con rmed by treating CPTresistant PANC-1 cells with ATM and ATR inhibitors (Fig. S8E). Thus, ATM-and ATR-induced primary cilia contribute to chemoresistance in PANC-1 cells.
The downstream effectors were then examined. CHEK1, CHEK2, AKT, and p53 were activated following CPT treatment (Fig. 5A-D). We then examined whether activation of these effectors contributed to primary ciliogenesis. Inhibition of CHEK1 and p53 did not affect CPT-induced ciliogenesis ( Fig. S9A-B); however, CPT-induced primary cilia were reduced signi cantly when CHEK2 and AKT were inactivated ( Fig. S9C-D).
To further con rm these ndings, the effectors were depleted by speci c siRNA and then primary ciliogenesis was examined. Surprisingly, the depletion of CHEK1, CHEK2, or AKT did not inhibit CPTinduced primary ciliogenesis (Fig. 5E-J), implying that other signaling pathways may play important roles in triggering primary cilia formation following CPT treatment.
Autophagy contributes to primary ciliogenesis.
Autophagy contributes to primary cilia formation [29,30]. Thus, we speculated whether autophagy contributed to CPT-induced primary ciliogenesis. CPT treatment activated autophagy, as evidenced by increased LC3 puncta (Fig. 6A) and the LC3 II to I ratio (Fig. 6B). To further con rm this nding, the initiation of autophagy was examined. Increased phosphorylation of AMPK at Thr172 (active site) and decreased phosphorylation of ULK1 at Ser757 (inhibitory site) were observed following CPT treatment ( Fig. 6C-D). Pharmacological inhibition of autophagy by treating cells with chloroquine (CQ), ba lomycin A1 (BafA1), or ULK1 inhibitor (ULK1i) alleviated primary cilia formation, implying that autophagy contributed to ciliogenesis ( Fig. S10A-C). To further con rm our nding, ATG7 was depleted by infecting cells with lentivirus containing different shRNA sequences against ATG7. ATG7 was e ciently depleted by shATG7#3 (Fig. 6E), and CPT-induced ciliogenesis was reduced signi cantly in ATG7-de cient cells (Fig. 6F). ATG7 depletion had no effect on cell viability but sensitized cells to CPT treatment ( Fig. 6G-H), and this nding was further supported by treating cells with chloroquine (Fig. S10D). Furthermore, treating cells with AMPK or ULK1 inhibitor reduced cell viability in CPT-resistant PANC-1 cells (Fig. 6I). Thus, autophagy promotes CPT resistance in PANC-1 cells.
Next, we examined whether ATM and ATR regulated primary ciliogenesis and chemoresistance via autophagy and excessive formation of centriolar satellites. Depletion of ATM or ATR alleviated CPTinduced AMPK activation, suggesting that autophagy was regulated by ATM and ATR activation (Fig. 7A-B). Next, we examined the excessive formation of centriolar satellites, which was reduced in ATM-or ATRde cient cells (Fig. 7C-D). Additionally, the depletion of ATM or ATR also reduced the levels of CPTincreased PCM1 (Fig. 7E), and these data were further supported by treating cells with caffeine, a PI3KK paninhibitor (Fig. 7F). Taken together, ATM and ATR trigger excessive formation of centriolar satellites and autophagy cooperatively for primary ciliogenesis, leading to chemoresistance. Discussion PDAC cells were devoid of primary cilia. Here, we showed that following chemotherapeutic drug treatment, PDAC cells regrew primary cilia. We uncovered the novel function of ATM and ATR in regulating primary cilia formation, at least partly contributing to cisplatin resistance in PDAC. We further demonstrated that activated ATM and ATR promoted primary ciliogenesis via excessive formation of centriolar satellites (EFoCS) and autophagy (Fig. S11). Thus, our study unraveled the interplay among the DNA damage response, primary cilia, and chemoresistance.
Most cells in our body grow primary cilia. However, the loss of primary cilia has been observed during tumorigenesis, including colon, breast, and pancreatic cancers [14,31,32]. Surprisingly, recent studies have shown that primary cilia are linked to chemoresistance in lung and breast cancers. During the development of diverse kinase inhibitor resistance in lung cancer, primary cilia can regrow [17]. Additionally, sonic hedgehog signaling facilitates the stemness of mammary tumor-initiating cells via primary cilia [33]. Thus, when cancer cells develop more malignant chemoresistance and stemness phenotypes, primary cilia reappear. In our study, cisplatin-resistant PDAC cells grew primary cilia.
Importantly, genetic or pharmacological disruption of these cilia sensitized resistant PDAC cells to cisplatin treatment, supporting that primary cilia contributed to chemoresistance. It remains unclear how primary cilia contribute to chemoresistance. Previous study showed that primary cilia maintained DNA damage responses when exposing to genotoxic stresses [34]. DNA damage responses maintain cell survival and genome integrity by repairing damaged genome [4]. We that speculated that primary cilia accelerated DNA damage responses to maintain cell survival and repair damaged DNA in CPT-resistant PANC-1 cells. Disruption of primary cilia attenuated DNA damage response, leading to robust DNA damage and sensitizing cells to cisplatin. Thus, our study strengthened the role of primary cilia in contributing to the development of chemoresistance in pancreatic cancers. Investigating whether these cilia-harboring cells show stemness and what signaling pathway contributes to these events will be important issues in the future.
ATM and ATR are tumor suppressors that maintain genome integrity by regulating DNA damage responses [35]. In addition to acting on the damaged genome, ATM and ATR are also correlated with chemoresistance. For example, ATM is hyperactivated following chemotherapy, thus maximizing the DNA damage response to maintain the survival of drug-resistant leukemic cells [36]. Additionally, ATMdependent activation of transglutaminase 2 and NF-κB signaling promotes a doxorubicin-resistant phenotype in breast cancers. Interestingly, using 3D-reconstituted basement membrane breast and lung cancer cell culture models, ATR is activated, promoting cisplatin resistance by activating translesion DNA synthesis modulation. Furthermore, in high-grade serous ovarian cancer, multiple resistance mechanisms to cisplatin have been revealed [37]. However, inhibition of PARP and ATR simultaneously increased DNA damage and sensitized high-grade serous ovarian cancer cells to chemotherapy [34]. Thus, targeting ATM and ATR is considered a novel approach to overcome drug resistance. In our study, we also demonstrated that activation of ATM and ATR contributed to cisplatin resistance in PDAC. Aside from their canonical roles in maintaining genome integrity, we showed that ATM and ATR promoted cisplatin resistance, at least partly by inducing primary ciliogenesis. Either ATM-or ATR-mediated nuclear or ciliary events cooperatively contribute to chemoresistance in PDAC. Thus, targeting the inhibition of ATM and ATR can be used as adjuvant therapy to offer clinically important distinctions in treating patients with PDAC. Centriolar satellites are associated with primary ciliogenesis. Orofaciodigital syndrome type I (OFD1), a component of centriolar satellites, is removed from centriolar satellites by autophagy for primary cilia formation during serum starvation [38]. However, following etoposide treatment, OFD1 is displaced from centriolar satellites, and inhibition of autophagy does not reverse this phenotype, suggesting that etoposide-induced OFD1 removal is independent of autophagy [29]. Additionally, in response to other cellular stresses, such as UV radiation, heat shock, and transcription stresses, p38 is activated, displacing PCM1, AZI1, and CEP290 from centriolar satellites for primary cilia formation [21]. These data suggest that the displacement or degradation of centriolar satellite components promotes primary cilia formation following metabolic or genomic stresses. However, excessive formation of centriolar satellites (EFoCS) has been observed in response to DNA damage, and EFoCS is required to induce centrosome ampli cation [20]. Thus, under different cellular stresses, some components of centriolar satellites may displace from or aggregate to centriolar satellites around the centrosome/basal body. Here, we showed that in PDAC, cisplatin induced primary ciliogenesis and EFoCS. Importantly, PCM1, a centriolar satellite scaffold, was also upregulated. Depletion of PCM1 inhibited primary ciliogenesis. More importantly, the disruption of dynactin to suppress EFoCS alleviated cisplatin-induced primary cilia, suggesting that EFoCS contributed to primary cilia formation. Thus, different cellular stresses lead to distinct centriolar satellite dynamics for primary cilia formation. How centriolar satellite dynamics affect primary cilia remains unclear. Centriolar satellites are required for primary cilia because the depletion of several satellite components inhibits primary ciliogenesis. However, following metabolic or genomic stresses, primary cilia grew, but disruption of centriolar satellites was observed, suggesting that the role of centriolar satellites in primary ciliogenesis is stress dependent. How centriolar satellite dynamics affect primary cilia formation must be clari ed in the future.

Conclusion
In this study, we uncovered the novel role of ATM and ATR in cisplatin resistance in PDAC, at least partly by inducing primary cilia formation. ATM-and ATR-induced EFoCS and autophagy cooperatively promote primary cilia. Thus, ATM and ATR not only prevent cancer cell death by maintaining genome integrity but also facilitate primary cilia, leading to cisplatin resistance in PDAC.

Cell culture
The human pancreatic ductal adenocarcinoma (PANC-1) cell line was maintained in Roswell Park Memorial Institute (RPMI)-1640 medium containing 10% fetal bovine serum (FBS) and 1% sodium pyruvate. Stable knockdown of IFT88 (shIFT88#3) was generated in PANC-1 cells via the lentiviral delivery of shRNA against IFT88 (#3; clone TRCN0000141713; RNAi core lab of Genomics Research Center, Academia Sinica, Taipei, Taiwan). Stable cells were selected with puromycin (1 μg/ml). To establish cisplatin-resistant pancreatic cancer cells, PANC1 cells were initially treated with a low dose of cisplatin (0.5 μM) for one month and then the dose of cisplatin was gradually increased to 1, 2 and 4 μM (each dose for one month of incubation). The culture medium of RPMI-1640 with cisplatin was changed every two days. CPT-resistant PANC-1 cells were maintained in 4 μM cisplatin-containing culture medium. All the cell lines were cultured at 37 °C in a humidi ed atmosphere of 5% CO 2 . These cells were regularly examined for mycoplasma contamination by immuno uorescence staining and immunoblotting assays according to published methods [39]. Immuno uorescence microscopy Cells were xed with ice-cold methanol at −20 °C for 5 min. After washing with PBS, the cells were blocked with 5% BSA for 1 hour at room temperature, followed by incubation with primary antibodies at 4°C for 12 hours. Next, the cells were washed with PBS three times. After that, the cells were incubated with uorescein isothiocyanate-conjugated or Cy3-conjugated secondary antibodies at room temperature for one hour in the dark. The nuclei were stained with 6-diamino-2-phenylindole (DAPI; 0.1 μg/ml) simultaneously. After washing with PBS three times, the coverslips were overlaid and mounted on glass slides in 50% glycerol. Prepared cells were observed using an Axio imager M2 uorescence microscope (Zeiss, Switzerland) and captured using ZEN pro software (Zeiss, Switzerland). 3D images of excessive formation of centriolar satellites were generated using Imaris software (Zurich, Switzerland).

RNA interference (RNAi)
IFT88, CEP164, PCM1, p150 glued , ATM, ATR, DNA-PKcs, CHEK1, CHEK2 and AKT were depleted in human PANC-1 and cisplatin-resistant PANC-1 cells using annealed siRNAs with the following target sequences:  inhibitors (Roche, Mannhein, Germany) on ice for 10 minutes. Next, the lysates were centrifuged at 13,300 rpm for 10 minutes at 4 °C. The cell lysates were collected and quanti ed using the Bradford protein quantity assay (Bio-Rad, Hercules, CA, USA). The quanti ed lysates were mixed with sample buffer and heated at 100 °C for 20 minutes. Next, the prepared samples were loaded on gels and separated by SDS-PAGE. After gel separation, the samples were transferred to PVDF membranes at 20 V for 720 min in a cold room. After washing with TBST, the membranes were blocked with 3% BSA in TBST at room temperature for 1 hour. Next, the membranes were incubated with primary antibodies at 4 °C for 12 hours.
Following extensive washing with TBST three times, the membranes were incubated with HRPconjugated secondary antibodies at room temperature for 1 hour. After washing with TBST three times, the signals were detected by ECL™ Detection Reagents.

Statistical analysis
All the results were presented as means ± S.D. from at least three independent experiments, and more than 500 cells were counted in each individual group. The error bars in bar plots represent the standard error of the mean from at least three independent experiments. Differences between two groups were compared using unpaired two-tailed t-tests and ANOVA for multigroup comparisons, for which a P value < 0.05 was considered statistically signi cant.