Autophagy Governs Protumorigenic Effects of Mitotic Slippage–induced Senescence

This article is featured in Highlights of This Issue, p. 1615

Antimitotic drugs are used extensively in the treatment of a variety of malignancies (1). The most commonly utilized class of antimitotic drugs are the microtubule poisons. These drugs interfere with cellular proliferation by disrupting microtubules (MT), thereby inducing a mitotic arrest that culminates in mitotic cell death (MCD; ref. 2). Unfortunately, the clinical success of MT-targeting drugs is often impeded by chemoresistance and disease relapse. Thus far, the majority of studies addressing mechanisms underlying therapy resistance have focused on tubulin mutations and drug efflux pumps, and have yet to yield improved outcomes for patients. Hence, a better understanding of possible cellular pathways and alternative molecular mechanisms underlying escape from cell death are critical to address tumor progression following MT-targeting drug treatment.

An alternative outcome to MCD following prolonged mitotic arrest is mitotic slippage. Here, cells prematurely exit from mitosis and enter a pseudo-G₁ phase without proper chromosome segregation and cytokinesis. Postslippage cells are often characterized as tetraploid multinucleated cells due to nuclear envelope assembly around condensed scattered chromosomes (3). Cells tend to undergo mitotic slippage rather than MCD when cyclin B1 degradation precedes proapoptotic signal accumulation during prolonged mitotic arrest, as posited by the prevailing “competing networks-threshold” model (4). There can be several cell fates following mitotic slippage. One possible outcome is cell death postslippage, that fulfils the cytotoxic goal of therapy in addition to MCD (5). Cells can also continue to proliferate as genomically unstable cells (5), thereby constituting a potential source of chemoresistance. In addition, cells have been shown to arrest in the next interphase postslippage and eventually enter cellular senescence (5). Indeed, cellular senescence has emerged as a major outcome of a variety of chemotherapies in a process known as therapy-induced senescence (6).

Classically, senescence is considered to be a barrier against tumorigenesis, as it restricts cell proliferation (7). This is mediated by the senescent cell secretome, known as the senescence-associated secretory phenotype (SASP), which consists of a variety of cytokines, chemokines, growth factors, and matrix metalloproteases (8). In addition to reinforcing stable growth arrest via both autocrine and paracrine signaling, SASP factors also promote immunosurveillance of senescent cells, leading to tumor remission (9). In this way, senescence serves as an important tumor-suppressive mechanism. However, senescent cells also possess oncogenic potential via paracrine effects of the SASP. SASP factors have been shown to engender protumorigenic effects such as cellular motility (invasion, migration, and metastasis), epithelial–mesenchymal transition (EMT), proliferation, angiogenesis, as well as inflammation in neighboring cells (8).

Molecular mechanisms underlying senescence following mitotic slippage and their consequential cell fate significance for MT-targeting therapies have not been well-explored. Here, we examine the senescent cell fate following nocodazole (MT-destabilizing agent) or paclitaxel (MT-stabilizing agent)-induced mitotic slippage. Our results demonstrate that mitotic slippage–induced senescence could confer paracrine tumorigenic effects via the SASP. We sought to improve therapeutic outcomes and identified ER stress–triggered autophagy as an effector of mitotic slippage–induced senescence. We found that inhibition of autophagy via pharmacologic means or silencing of autophagy-associated genes could override senescence, leading to cell death upon antimitotic drug treatment. Importantly, combination treatment of tumors in mice with often-used paclitaxel and autophagy inhibitor chloroquine showed significant tumor growth arrest, underscoring autophagic inhibition as a potential strategy of tackling resistance. We further found that xenograft tumors with wild-type p53 exhibited a superior response toward combinatorial therapy compared with tumors with p53 knockout. This suggests that sensitivity toward this treatment is dictated by p53 status, which could serve as a potential biomarker in predicting clinical response.

U2OS, HCT116, MCF7, MDA-MB-231, PanC1, MIA PaCa2, HEK293T, and hTERT-RPE-1 cells were purchased from ATCC at the start of the project in 2014. All cell lines were cultured in DMEM supplemented with 10% FBS (Hyclone GE), with the exception of hTERT-RPE-1 (DMEM-F12). Cells were freshly thawed monthly and Mycoplasma testing was performed using EZ-PCR Mycoplasma Test Kit (Biological Industries). Reagents used in this study are included as Supplementary Tables S2 and S3.

Cells were transfected with siRNA using Lipofectamine RNAiMAX (Invitrogen) according to manufacturer's instructions. Cells were treated with drugs for 16 hours after transfection. siRNA targeting scrambled sequence (#D-001810-10-05) and ATG5 (#M-004374-04) were purchased from Dharmacon and used according to the manufacturer's instructions.

Cells were stained using the senescence β-galactosidase (SA-β-gal) Staining Kit (#9860; Cell Signaling Technology) according to manufacturer's instructions.

Cells were incubated with BrdU (BrdU Labeling and Detection Kit, Roche) for 16 hours and visualized according to the manufacturer's instructions.

Cells were lysed in RIPA buffer (Pierce) containing Protease/Phosphatase Inhibitor Cocktail (Thermo Scientific) and equal amounts of total protein were subjected to SDS-PAGE.

After fixation in ice-cold 70% ethanol overnight, cells were stained with propidium iodide, and incubated with phospho-Histone H3 antibody conjugated to Alexa Fluor 488 (Cell Signaling Technology). Samples were analyzed using Accuri C6 flow cytometer (BD Biosciences).

Cells were fixed in 4% paraformaldehyde in PBS for 10 minutes at room temperature. Antibody incubation and staining were performed as described previously (10).

Cells were imaged on a Nikon inverted fluorescence microscope at multiple sites every 30 minutes for 48–72 hours using a Plan Apo 40× objective.

Cells were treated with drugs for two days followed by culture in drug-free media containing 0.5% FBS for additional two days. Culture media were then centrifuged at 5,000 × g, filtered through 0.22-μm pore filter (Pall Corporation) and mixed with media containing 40% FBS in a proportion of 3:1 to generate conditioned media (CM) containing 10% FBS.

Cells were treated with drugs for 72 hours and reseeded at 5,000 cells per 6-well, followed by culture for 10 days with fresh media supplemented every other day. Colonies were stained with 0.05% crystal violet (Sigma). Number of colonies was analyzed using GelCount (Oxford Optronix) according to manufacturer's instructions.

Cells were seeded at 4,000 cells per 96-well and treated with drugs for 72 hours. Cell viability was assessed using CellTiter 96 One Solution Proliferation Assays Kit (G3580; Promega) according to manufacturer's instructions.

Total RNA was extracted using RNeasy plus Mini Kit (QIAGEN) according to manufacturer's instructions. cDNA was reverse transcribed using iScript RT Supermix (Bio-Rad) and subjected to SYBR Kit (#4472942; Life Technologies). Relative expression values of each gene were normalized to GAPDH expression. Primers used are in Supplementary Table S4.

Forty-one analytes from Human Cytokine Panel 1 (Merck Millipore) were measured as per manufacturer's instructions. Plates were washed using Tecan Hydrospeed Washer (Tecan) and read with Flexmap 3D system (Luminex Corp). Data were analyzed using Bio-Plex manager 6.0 (Bio-Rad) with a 5-parameter curve-fitting algorithm applied for standard curve calculations.

For scratch wound migration assay, cells were incubated with CM for two days. At 90% confluency, the cell monolayer was scratched and the wound closure rate tracked. For cell invasion assay, cells expressing H2B-GFP were incubated with CM for two days, plated on top surface of transwell filter chambers precoated or uncoated with Matrigel (BD Biosciences), and the percentage of invasive cells determined after 24 hours. For choroid angiogenesis assay, segments of the peripheral choroid layer from eyes of P3 mice were incubated with CM 1:3 diluted with EGM2 media (Lonza) over four days and imaged under phase contrast.

U2OS H2B-GFP cells were incubated with CM for two days and stained with DiI (Vybrant, Life Technologies) before injection into zebrafish embryos. Embryos were imaged 48 hours after injection to determine the metastatic tumor foci position relative to injection site.

All animal studies were approved by the Institutional Animal Care and Use Committee (IACUC no. 151103) of the Biological Resource Centre and were carried out under the policies from the Animal Facility Centre of the Agency for Science, Technology and Research (A*STAR, Singapore). HCT116 colon cancer cells were injected subcutaneously into both dorsal flanks of female BALB/c athymic nude mice (nu/nu) (InVivos; n = 5 mice in each treatment group). When tumor volume reached around 300 mm³, mice were injected intraperitoneally with respective treatment every three days for two weeks. Tumor volume was measured every three days from start of treatment. Mice were sacrificed at the end of the treatment schedule and all tumors were harvested for weight measurement, IHC, and immunofluorescence.

This study was conducted in accordance with recognized ethical guidelines (Singapore Guideline for Good Clinical Practice, Declaration of Helsinki and Belmont Report) and approved by Nanyang Technological University Institutional Review Board (IRB-2018-04-009). Written informed consents were obtained from participating patients. Biopsies were collected using core needle biopsy procedure and were snap-frozen in liquid nitrogen immediately after aspiration. Frozen biopsies were thawed slowly, fixed, and processed accordingly for IHC.

After fixation, sections were incubated with primary antibodies overnight at 4°C, followed by incubation with biotinlyated secondary antibodies (Vector Laboratories). Sections were then incubated in Avidin:Biotinylated Enzyme Complex (ABC; Vector Laboratories) for 30 minutes, developed with 3,3′-diaminobenzidine (DAB) substrate (Vector Laboratories), and nuclei counterstained with hematoxylin. Slides were mounted with Fluka Eukitt quick-hardening mounting medium (Sigma-Aldrich).

RPE-1 cells were cultured in SILAC DMEM/F12 media, containing light or heavy arginine and lysine [“R0K0 DMEM/F12” with ¹²C₆-l-arginine (Sigma-Aldrich) and ¹²C₆-l-lysine (Sigma-Aldrich) or “R10K8 DMEM/F12” with ¹³C₆-l-arginine (Cambridge Isotope) and ¹³C₆-l-lysine (Cambridge Isotope)] and supplemented with dialyzed FBS. SILAC labeling was performed as described previously (11). After lysis in RIPA buffer (Thermo Scientific) with Protease/Phosphatase Inhibitor Cocktail (Thermo Scientific), equal amounts of protein from R0K0 and R10K8 were mixed and boiled with loading buffer at 95°C for 5 minutes. Protein complexes were separated by one-dimensional 4%–12% NuPage Novex Bis–Tris Gel (Invitrogen), stained with Colloidal Blue Staining Kit (Invitrogen), and subjected to in-gel digestion as described previously (11).

Additional Materials and Methods can be found under Supplementary Methods.

To study the fate of cells postslippage, we chose mitotic slippage–prone cells, namely osteosarcoma U2OS cells, colorectal carcinoma HCT116 cells, and nontransformed telomerase-immortalized hTERT-RPE-1 cells (hereafter referred to as RPE-1). Mitotic slippage, instead of MCD, was previously found to be the preferred pathway in these cells upon treatment with MT poisons (4). To achieve maximum homogeneity in the cell population, U2OS and RPE-1 cells were first synchronized before treatment with the microtubule-depolymerizing drug, nocodazole for a prolonged period of 72 hours to allow cells to undergo mitotic slippage (experimental scheme in Supplementary Fig. S1A). Cells were monitored by time-lapse microscopy from the start of nocodazole treatment (t = 0 hours). U2OS and RPE-1 cells predominantly displayed a “rounded up” morphology, typical of mitotic arrest, at 16 hours post-nocodazole treatment (Supplementary Fig. S1A). By 72 hours post-nocodazole treatment, 70.9% of U2OS cells and 87.8% of RPE-1 cells were observed to have undergone blebbing and reflattening without cytokinesis (characteristics of mitotic slippage). This was further confirmed by time-course experiments in synchronized nocodazole-treated U2OS cells that showed decreased levels of mitotic cyclin B1, mitotic marker phosphorylated Histone H3, and spindle assembly checkpoint markers BubR1 and Bub1 by immunoblotting (Supplementary Fig. S1B). In addition, a mitigated decline in levels of antiapoptotic protein Mcl-1 and corresponding increase in cell death marker cleaved PARP was detected. An accumulation of multinucleated tetraploid G1 U2OS cells was also observed, as shown by 4C-G1 population by flow cytometry and multiple nuclei by phase-contrast microscopy (Supplementary Fig. S1C and S1D). A similar increase in multinucleation was seen in nocodazole-treated HCT116 cells postslippage (Supplementary Fig. S1E).

Because cells following prolonged nocodazole treatment displayed an enlarged, flattened morphology reminiscent of senescence, we stained cells for the SA-β-gal. As shown in Fig. 1A, at day 3 post-nocodazole treatment, about 45% of U2OS cells and 70% of RPE-1 cells stained positive for SA-β-gal, indicating entry into senescence. Negligible BrdU staining in multinucleated cells compared with control indicating a lack of cell proliferation further supporting this finding (Supplementary Fig. S1F). To confirm this as a bona fide senescence phenotype, nocodazole was removed and cells were cultured in fresh media for an additional 6 days (experimental scheme in Fig. 1A). A progressive increase in the number of cells that entered senescence was observed (Fig. 1A), confirming that cells underwent a stable cell-cycle arrest. Loss of lamin B1 has been associated with senescence (12). Our data show a decrease in lamin B1 levels after day 3 (Fig. 1B). Increased cyclin-dependent kinase inhibitor p21, tumor suppressors p53 and total retinoblastoma protein (pRb), and hypophosphorylation of pRb at Ser780 (13) compared with control further confirmed the senescence phenotype at day 3 (Fig. 1B). Interestingly, we observed a decline in p53 levels from day 6 onwards even though the senescence phenotype persisted. This is consistent with a report showing downregulation of p53 to be crucial for induction of SASP (14). By inference, because p21 is part of the p53–p21 senescence pathway where p53 activates p21, this could also explain the corresponding decrease observed for p21 (15). Taken together, our results confirmed that cells postslippage entered senescence following G₁ arrest.

To determine whether antimitotic drugs broadly induce senescence, cells were treated with the MT-stabilizing drug paclitaxel (PTX), Aurora kinase B inhibitor ZM447439 (ZM), or kinesin-related Eg5 inhibitor Monastrol (Mon). SA-β-gal staining showed discernible differences in the extent of senescence, as paclitaxel and ZM significantly increased the percentage of SA-β-gal–positive cells while Mon-treated cells hardly displayed signs of senescence (Supplementary Fig. S2A). Increased SA-β-gal staining was also observed for nocodazole- and paclitaxel-treated HCT116 cells (Supplementary Fig. S2B). Interestingly, treatments that resulted in senescence, namely paclitaxel and ZM, showed high degrees of multinucleation, whereas Mon-treated cells were mostly mononucleated (Supplementary Fig. S2C). This could imply that postslippage multinucleated cells may have enhanced proclivity to enter senescence. Indeed, a previous report demonstrated that tetraploid cells with irregular-shaped nuclei progressively developed senescence following Aurora kinase B inhibition (16).

To determine whether mitotic slippage–induced senescence could be observed in vivo, we treated HCT116 xenograft mice with either nocodazole or paclitaxel. Hematoxylin and eosin staining of tissue sections revealed a discernible increase in multinucleated cells in the nocodazole- or paclitaxel-treated xenografts compared with vehicle-treated control (Supplementary Fig. S3A). In addition, the majority of cells that stained positive for SA-β-gal were observed to be multinucleated (Supplementary Fig. S3A), suggesting possible correlation between multinucleation and senescence. Nocodazole- or paclitaxel-treated xenografts also showed a significant increase in both p21 mRNA and protein levels compared with vehicle-treated control (Supplementary Fig. S3B and S3C). In addition, tissue biopsies from invasive ductal breast carcinoma from patients treated with paclitaxel showed increased populations of cells in the humoral region that stained positive for p21 expression as compared with biopsies obtain from patients before treatment (Supplementary Fig. S3D).

SASP factors secreted by cells which have undergone senescence due to replicative exhaustion, oncogene activation, or irradiation can modulate the tissue microenvironment to stimulate tumor progression (8). As a first step to ascertain potential tumorigenic role for SASP following mitotic slippage, we performed gene expression microarray analysis on U2OS cells treated with either DMSO (control) or nocodazole for 48 hours. Microarray data confirmed SASP factor expression, including cytokines and chemokines IL1α, IL1β, CXCL8, and CCL3 among the upregulated genes (Supplementary Fig. S4A). This was further confirmed at the protein level with stable isotope labeling with amino acids in cell culture (SILAC) followed by high-resolution mass spectrometry (MS; Supplementary Fig. S4B; ref. 11). Notably, mass spectrometric analysis revealed upregulation of SASP factors including IL1β and IL8, and senescence-associated histone H3.3 variant (ref. 17; Supplementary Fig. S4C; Supplementary Table S1). In addition, both quantitative real-time PCR (qRT-PCR) and Luminex assays confirmed upregulated expression and secretion of various SASP factors in nocodazole-treated U2OS cells at 72 hours postslippage compared with DMSO-treated control cells (Fig. 1C).

The observed increase in SASP factor secretion prompted us to test whether SASP factors could mediate tumorigenic phenotypes such as migration, invasion, and angiogenesis. To this end, we prepared conditioned media (CM) from RPE-1 and U2OS cells treated with nocodazole for 48 hours for use in various phenotypic assays (experimental scheme in Fig. 1D). To assess migratory capability, RPE-1 and U2OS cells expressing chromatin marker H2B-GFP exposed to CM from their respective postslippage parental cells were subjected to the scratch wound healing assay. As shown in Fig. 1E, postslippage CM (nocodazole CM) promoted wound closure at a rate faster than control CM (DMSO CM), indicating increased induction of migration. Notably, cell proliferation assays by BrdU labeling and cell count experiments demonstrated that factors secreted did not alter rate of cell proliferation in CM-exposed cells (Supplementary Fig. S5A and S5B), indicating increased wound closure was not influenced by proliferation. To evaluate invasive capability, U2OS H2B-GFP cells incubated with either control or postslippage CM were used in transwell invasion assays. An increased number of cells exposed to postslippage CM invaded the bottom of the filter chamber compared with control CM (Fig. 1F). An upregulation of invasion and migration-related markers fibronectin and MMP9 in U2OS cells exposed to postslippage CM compared with control CM was also observed (Fig. 1G), confirming migratory and invasive capabilities. To test for angiogenic capability, we used the choroid angiogenesis assay, an ex vivo model of angiogenesis. Increased induction of vascular sprouting from choroid explants incubated with postslippage CM compared with control CM was observed four days after incubation (Fig. 1H), demonstrating angiogenic capability conferred by factors from postslippage cells. Taken together, we conclude that postslippage SASP proteins confer paracrine protumorigenic potential.

We next wished to investigate the molecular determinants that induce senescence in postslippage cells. Mass spectrophotometric analysis of U2OS cells treated with nocodazole for 48 hours, revealed four autophagy-related proteins, namely MAP1A, MAP1B, MAP1LC3B (also known as LC3B), and GABARAP to be downregulated postslippage compared with control (Supplementary Fig. S4C; Supplementary Table S1). Because all four of these proteins serve as autophagic substrates, their downregulation suggests increased autophagic activity. Autophagy is a catabolic process in which cytoplasmic constituents are targeted for removal or recycling in autophagosomes that fuse with the lysosome (18). MAP1A and MAP1B are microtubule-associated proteins, which are precursor polypeptides that undergo proteolytic processing to generate heavy and light chain subunits such as MAP1LC3A and MAP1LC3B (19). MAP1LC3B and GABARAP belong to the ATG8 orthologs of mammalian cells (19). Consistent with upregulated autophagy postslippage, time-lapse microscopy revealed increased autophagic marker GFP-LC3 punctate foci in U2OS cells postslippage (t = 48 hours), indicating autophagosome accumulation (Fig. 2A).

To further evaluate autophagic flux, we assessed conversion of the nonlipidated form LC3B-I to the lipidated autophagosome-associated form LC3B-II (hereafter LC3-I and LC3-II, respectively), as well as degradation of the ubiquitin-binding autophagic adaptor protein p62/SQSTM1. To ensure that autophagy flux was assessed only in adherent postslippage cells destined to enter senescence, and that quantitation was not affected by cell death postslippage, cells were washed 36 hours post nocodazole treatment to remove apoptotic cells. As shown by phosphorylated Histone H3 levels in Fig. 2B and C, HCT116 and U2OS cells were in mitosis (M) at t = 16 hours and postslippage (PS) from t = 24–36 hours. Increased autophagy flux postslippage was observed as shown by an increase in LC3-II isoforms at t = 24–36 hours compared with t = 16 hours (Fig. 2B and C). To confirm this, U2OS cells were treated with chemical inhibitor of autophagy Bafilomycin A1 (Baf A1), which blocked lysosome acidification and prevented autophagosome clearance. Increased LC3-II was observed at t = 36 and 48 hours (Fig. 2C) compared with nocodazole control, suggesting LC3-II accumulation postslippage to be due to autophagosome accumulation and not impairment of downstream autophagic processes such as autophagosome–lysosome fusion or lysosomal degradation. Concomitantly, a decrease in p62, which can be blocked by Baf A1, further confirmed active autophagy (Fig. 2B and C). LC3-II accumulation in postslippage cells was also blocked by knockdown of autophagy by stable expression of short hairpin RNA targeting ATG5 (shATG5) required for autophagy elongation (ref. 20; Fig. 2D).

Autophagic induction occurs via a number of pathways that finally converge on regulation of the AMPK/ULK/mTOR axis, which integrates growth factor and nutrient signals to regulate cellular metabolism and maintain energy homeostasis (21). Phosphorylation of AMP-activated protein kinase (AMPK) on Thr172 activates autophagy by directly activating Unc-51 Like Autophagy Activating Kinase 1 (ULK1) through phosphorylation of Ser317 and Ser777 (21, 22). In contrast, high activity of the mTOR negatively regulates autophagy by preventing ULK1 activation via ULK1 Ser757 phosphorylation and disruption of ULK1 and AMPK interaction (21). To test whether cells postslippage invoked the AMPK/ULK/mTOR axis for autophagy induction, nocodazole-treated U2OS cells were subjected to immunoblotting. We observed that in mitotic cells 16 hours post-nocodazole treatment, both mTOR and AMPK were activated as shown by increased phosphorylation of the mTOR substrate p70 S6K, conversion of total ULK1 to phosphorylated Ser757 ULK1 as well as increased phosphorylated Thr172 AMPK (Fig. 2E). This correlated with enhanced autophagy, and was consistent with an observed increase in LC3-II (Fig. 2E). As it was previously reported that autophagy can be activated in a mTOR-independent manner (23, 24), our findings suggest presence of autophagic activity despite mTOR activation during mitosis. On the other hand, diminished levels of phosphorylated-p70 S6k indicating decreased mTOR activity with concomitant decrease in ULK1 phosphorylated on Ser757 residue was observed in postslippage cells (Fig. 2E). These findings together with the observed increase for phosphorylated Thr172-AMPK and LC3-II levels, strongly suggested that the AMPK/ULK/mTOR axis promotes autophagy induction postslippage.

We next sought to understand the molecular mechanism by which mitotic slippage could lead to AMPK activation and autophagy induction. A recent report showed that aneuploidy caused substantial endoplasmic reticulum (ER) proteotoxic stress and unfolded protein response (UPR) activation in cells that underwent cell death postslippage (25). We therefore examined the expression of UPR-related protein kinase RNA-like ER kinase (PERK) and inositol-requiring protein 1 (IRE1) in U2OS cells after nocodazole and paclitaxel treatment. Upon accumulation of misfolded or unfolded proteins in the ER, PERK is activated through dimerisation and trans-autophosphorylation on multiple residues including Thr980 (26). Activated PERK phosphorylates eukaryotic translation initiator factor 2α (eIF2α) on Ser51, which then attenuates global protein synthesis. Phosphorylated eIF2α leads to an increase in transcription factor CCAAT/enhancer binding protein (C/EBP) homologous protein (CHOP) at both transcriptional and translational level. Our data revealed that nocodazole and paclitaxel treatment increased the phosphorylation of PERK at Thr980 and eIF2α at Ser51 postslippage compared with control (t = 0 hours, start of treatment; Fig. 3A). An increase in CHOP at both protein and mRNA levels was also observed postslippage compared with control (Fig. 3A and B). An alternative signaling branch of the UPR is the IRE1–XBP1 pathway (27). Phosphorylation of IRE1 leads to cleavage of XBP1 mRNA and generates a translational frameshift which acts as a potent transcriptional activator (26). Increased IRE1 phosphorylation at Ser724 and XBP1 mRNA splicing was observed postslippage compared with DMSO-treated control (Fig. 3C). Thus, in addition to cell death postslippage, our results indicate induction of ER stress and UPR induction with activation of PERK and IRE1 UPR pathways in postslippage senescent cells as well.

To investigate if either of these pathways regulate autophagy, PERK or IRE1 activity was inhibited with PERK inhibitor GSK2656157 or IRE1 inhibitor 4μ8C. GSK2656157 reduces both total and phosphorylated PERK levels, while 4μ8C blocks substrate access to the active site of IRE1 and selectively inactivates both XBP1 splicing and IRE1-mediated mRNA degradation. Interestingly, while inhibition of IRE1 did not prevent the accumulation of LC3-II, PERK inhibition decreased LC3-II levels in postslippage cells to that of cycling cells (Fig. 3D and E), revealing that autophagy induction postslippage occurred through PERK activation. As described previously (Fig. 2), autophagy was affected via AMPK activation, which is downstream of PERK (28). We find that PERK inhibition led to decreased AMPK phosphorylation (Fig. 3D), further confirming that the increase in LC3-II was predominantly induced by the PERK–AMPK axis of the ER stress response in postslippage cells.

Because autophagy is active postslippage, we sought to investigate whether the senescence cell fate could be influenced by autophagic inhibition. Inhibition of autophagy by Baf A1 or transfection with shATG5 resulted in senescence bypass as observed by decreased SA-β-gal staining (Fig. 4A; Supplementary Fig. S6A). Concomitantly, a substantial increase in cells entering S-phase was suggested by increased BrdU labeling upon inhibition of autophagy (Fig. 4B). Transfection of siATG5 clearly suppressed the increase in LC3-II indicating autophagy was indeed blocked (Fig. 4C). This unequivocally confirmed that postslippage senescence was dependent on autophagy. Time-lapse microscopy revealed that autophagy inhibition in U2OS cells treated with nocodazole or paclitaxel increased cell death postslippage (Fig. 4D) and reduced cell viability (Fig. 4E). Consistent with this, clonogenic assays showed reduction of long-term cell survival following autophagic inhibition (Fig. 4F). Autophagy has been described to mediate degradation of lamin B1 (29). Interestingly, nocodazole-treated RPE-1 cells pretreated with Baf A1 prevented lamin B1 degradation at day 3 and day 6 (Supplementary Fig. S6B and S6C). Notably, these differences were not due to changes at the transcriptional levels as our RT-PCR results excluded this possibility (Supplementary Fig. S6D).

To determine whether these findings could be extrapolated to other antimitotic drugs, we assayed cells treated with nocodazole, paclitaxel, Mon and ZM for the apoptotic marker cleaved PARP following ATG5 knockdown. Interestingly, the increase in cell death was restricted to MT-targeting drugs nocodazole and paclitaxel (Fig. 4G). In addition, autophagy inhibition increased DNA damage (as determined by γH2AX foci and 53BP1 protein levels) compared with control (Fig. 4G–I). Enhanced replication stress shown by increased RPA32 phosphorylation at 72 hours post-nocodazole treatment compared with control was also observed (Fig. 4I). Taken together, our findings indicate that autophagy modulates cell fate upon antimitotic drug treatment and that inhibition of autophagy engenders cells to bypass senescence postslippage and enter S phase. These cells then undergo increased replication stress and DNA damage, culminating in cell death.

A recent study reported that autophagy plays a role in promoting MCD during prolonged mitotic arrest (30), suggesting that entry into mitotic slippage per se could potentially be inhibited by autophagy. We therefore tracked individual cell fate (MCD vs mitotic slippage) in U2OS cells stably transfected with shATG5 treated with either nocodazole or paclitaxel for 48 hours using time-lapse microscopy. Intriguingly, there was no significant difference in the proportion of cells undergoing MCD or slippage upon autophagic inhibition compared with control (Supplementary Fig. S7A). Consistently, the duration from mitosis to mitotic slippage (Supplementary Fig. S7B), cyclin B1 degradation, and dephosphorylation of BubR1 and Histone H3 were not affected by autophagy inhibition postslippage (24–36 hours post-nocodazole treatment; Supplementary Fig. S7C and S7D). Our results thus indicate that autophagy modulates cell fate specifically after escape from mitotic arrest through mitotic slippage.

We observed a discernible decrease in expression of SASP cytokines CXCL3, IL1β, IL6, IL8, CCL7, and PDGF-AB/BB and increased expression of BMP2 postslippage following autophagy inhibition compared with control (Fig. 5A; Supplementary Fig. S8A and S8B). This suggests that autophagy enables senescence and consequently modulates SASP following mitotic slippage. Notably, IL1β regulatory control via the autophagy axis was observed to be more prominent posttranscriptionally as shown in Supplementary Fig. S8B (compare with mRNA level in Supplementary Fig. S8A). In addition, SASP-induced tumorigenic function was attenuated following autophagic inhibition (Fig. 5). Postslippage CM promoted increased vascular sprouting compared with control CM from cycling cells in choroid explants, whereas autophagic inhibition significantly inhibited sprouting activity (Fig. 5B). In addition, transwell assays using cells incubated with CM from postslippage U2OS cells expressing shATG5 showed a reduction of these cells capable of invasion compared with control CM (Fig. 5C). To assess whether cytokines were required for SASP-mediated paracrine signaling functions in vitro, we used CM from IL1β and IL8-deficient postslippage cells (Supplementary Fig. S8C) for gene expression analysis of invasion and migration-related markers. Depletion of either IL1β or IL8 or both cytokines from the CM resulted in a significant decrease in invasiveness compared with control CM (Supplementary Fig. S8D). In addition, decreased fibronectin expression was observed in all conditions (Supplementary Fig. S8E), whereas a decrease in MMP-9 was only observed upon double knockdown of IL1β and IL8 (Supplementary Fig. S8E). Supplementation of IL1β or IL8 to postslippage CM collected under autophagy inhibition conditions proved sufficient to rescue the reduction in invasiveness (Supplementary Fig. S8F). Taken together, this suggests that autophagy-enabled production of IL1β and IL8 is crucial for postslippage senescent cells to promote paracrine tumorigenic progression.

To examine whether the in vitro migratory and invasive reduction following autophagic inhibition could be extrapolated to metastatic inhibition in vivo, we used the zebrafish model of malignancy to monitor migration. The transparency of the zebrafish embryo provides the unique ability to visualize in vivo migratory changes of tumor cells in real time. U2OS H2B-GFP–expressing cells were incubated with CM from either cycling or postslippage cells expressing shATG5 or control for two days before injection into zebrafish embryos. H2B-GFP–expressing cells incubated with postslippage CM were observed to migrate out from the injection site and metastasize further than those with control CM (Fig. 5D and E). This metastatic potential was significantly reduced following autophagic inhibition via shATG5 (Fig. 5D and E). These observations further support the role of autophagy-dependent SASP secretion in paracrine migration and invasion in vivo.

As p53 is a well-known regulator of cellular senescence (31), we hypothesized that p53 status could be an important predictor of response to combinatorial treatment of nocodazole or paclitaxel with autophagy inhibition. Increased DNA damage (as measured by γH2AX protein levels) and a corresponding increase in cell death (by cell viability assays and cleaved PARP) were detected in U2OS cells postslippage with depleted p53 (shp53) compared with control (Supplementary Fig. S9A and S9B). In addition, we observed that HCT116 cells with wild-type (WT) p53 showed increased cell death upon combination treatment compared with p53-deficient cells (Fig. 6A). Cancer cells lines with dominant-negative p53 mutations, namely MDA-MB-231 breast cancer cells with p53 R280K, PanC1 pancreatic cancer cells with p53 R273H and PaCa2 pancreatic cancer cells with p53 R248H also showed less sensitivity to combination treatment compared with MCF7 breast carcinoma cells with WT p53 (Supplementary Fig. S10A). This suggests that cells with intact p53 might respond more favorably to combinatorial treatment.

To test whether these findings could be extrapolated to an in vivo model, we treated mice xenografted with HCT116 cells (p53^+/+ or p53^−/−) with combination treatment of either nocodazole or paclitaxel and autophagic inhibitor chloroquine (CQ). Combination treatment (nocodazole + CQ or nocodazole + CQ) compared with control nocodazole or paclitaxel alone significantly decreased tumor growth and final tumor weight (Fig. 6B–D), particularly in mice xenografts with WT p53, suggesting that tumors with intact p53 were most sensitive to combination treatment. Molecular assessment of tumor sections derived from mice xenografts showed more robust LC3 and p62 puncta accumulation in the combination drug-treated mice compared with nocodazole or paclitaxel treatment alone (Supplementary Fig. S10B and S10C), indicating efficient autophagic inhibition.

In conclusion, we propose the model outlined in Fig. 6E. In response to the prolonged mitotic arrest induced by antimitotic drugs, cells undergo either mitotic cell death or mitotic slippage. In cells postslippage, autophagy is induced through ER stress- and UPR-mediated regulation of the AMPK/mTOR/ULK1 axis. This contributes to senescence and SASP production, which confers protumorigenic potential in a paracrine manner. Conversely, following autophagy inhibition, senescence is bypassed and postslippage cells undergo death in a p53-dependent manner.

The success of antimitotic therapies, often used as first-line treatment of several malignancies (1), is limited due to acquired resistance. The fate of cells following mitotic slippage, albeit representing a route of escape from mitotic arrest and mitotic cell death, has not been extensively studied. Here, we show that multinucleated tetraploid cells that accumulate postslippage can undergo senescence and drive paracrine tumorigenic effects both in vitro and in vivo in an autophagy-dependent manner.

Although prior seminal work on the autophagy-senescence connection by Young and colleagues (32) showed that autophagic inhibition delayed the senescence phenotype following oncogene activation, our work demonstrates that inhibition of autophagy postslippage results in senescence bypass and accelerated cell death. The most compelling evidence for senescence bypass after autophagy inhibition was increased entry into S-phase, the induction of DNA damage and replication stress, and reduction in cell viability in postslippage cells collaterally.

Autophagy has also been implicated as a potential inhibitory regulator of senescence (33). How does one reconcile the fact that autophagy can both activate and inhibit cellular senescence? Kang and colleagues (34) provide an explanation for these conflicting results by describing the differential regulation of senescence via selective versus general autophagy. This is achieved by transcription factor GATA4, which the authors established to be a senescence regulator (34). GATA4 bound to autophagy adaptor p62 is degraded by selective autophagy under nonsenescent conditions. This selective autophagy is suppressed upon senescence induction where GATA4 is stabilized due to decreased GATA4–p62 interaction. GATA4 stability triggers a series of events which ultimately leads to activation of NFκB for SASP production, thus facilitating senescence (34). In addition, autophagy has been described to mediate degradation of lamin B1 (29). The loss of lamin B1, as described before (12) and also shown in our study, is a senescence-associated marker. Inhibition of autophagy or LC3-lamin B1 interaction prevented oncogene-induced lamin B1 loss and attenuated senescence (29), and we observed similar block in lamin B1 degradation upon autophagy inhibition. It will be worth exploring whether LC3–lamin B1 interaction and GATA4 contribute to autophagy-induced senescence in our context.

Our findings demonstrate that autophagy is induced during mitotic slippage through the PERK–AMPK axis of the ER stress response. Inhibition of ER stress decreased the rate of LC3-II conversion by preventing AMPK phosphorylation. Although some studies indicate that ER stress prevents cellular senescence by upregulating autophagy (35), there is compelling evidence to support the opposite. Using a mouse model of B-cell lymphoma treated with the DNA-damaging agent cyclophosphamide, the Schmitt and colleagues' group (36) recently reported that senescent cells rely on autophagy to cope with SASP-coupled proteotoxic stress. This proteotoxic stress can lead to transcriptional activation of UPR genes and autophagy induction (25), rendering the senescent cells sensitive to autophagic inhibition. In addition, autophagy might further be maintained as part of the innate immune response against cytoplasmic chromatin fragments associated with senescence through the cGAS–STING pathway (37), as cGAS directly interacts with Beclin-1 (38) to clear cytosolic DNA in an autophagy-dependent manner. In support of this, the cGAS and/or STING pathway has been reported by multiple groups to be involved in induction of senescence and/or SASP (37, 39).

Postslippage SASP factors conferred paracrine protumorigenic phenotypic effects such as migration, invasion, and vascularization on neighboring cells. Proliferation of cells incubated with CM from postslippage cells was unaffected. As it has been shown that SASP composition and quantity is dependent on cell type and mode of senescence induction (8), this suggests that postslippage SASP may serve to play a “secondary” protumorigenic role in cells where migration, invasion, and angiogenesis are uncoupled from cell proliferation thereby leading to the development of a more malignant phenotype in cancer cells (40). It is possible that once neighboring cells become transformed and stimulated to proliferate, postslippage SASP then plays a role to further enhance the tumorigenic capabilities of these cells, demonstrating that mitotic slippage–induced senescence could serve as a conduit for malignant transformation and antimitotic therapy resistance.

In contrast, multiple studies have reported the role of autophagy in promoting the elimination of tumor cells through modulation of the inflammatory response (41). The autophagy-dependent inflammatory response is bidirectional and context-dependent. To enhance efficacy of antimitotic therapies, one might not only aim to eliminate protumorigenic SASP, but may also provoke desirable immunogenic cell death and clearance by the induction of antitumorigenic SASP. Therefore, it will be important that future studies test the impact of mitotic slippage–induced SASP expression on the tumor microenvironment and host immune system in a spontaneous tumor model.

In addition, targeting of senescent cells seems not only relevant for tumor cells and their microenvironment, but also for noncancerous cells. The Campisi and colleagues' group (9) recently showed that cytotoxic drugs induce senescence in normal cells thereby contributing to the systemic side effects of chemotherapy. Hence, several approaches targeting or bypassing senescence during chemotherapy to reduce the likelihood of acquired resistance and chemotoxicity by redirecting cells toward apoptosis have been attempted (42). In addition, the conviction that senescence is a truly irreversible process has been challenged as stem-like aggressive tumor-initiating cells have emerged from senescent cells (43, 44). Interestingly, senescence revertants (without stem-like properties) have recently been shown to contain a subset of senescence-activated genes that can contribute to aggressiveness of the revertants (45).

Our results establish a key mechanism underlying cell fate after mitotic slippage and suggest a novel strategy combining cytotoxic drugs with autophagy inhibition to tackle undesired effects of mitotic slippage–induced senescence. It is conceivable that senescent cells rely on autophagy as a nutrient source to support cellular metabolism and survival, as well as for autophagy-dependent clearance of targeted substrates. This dependence could render them selectively vulnerable to autophagic inhibition. Our in vivo zebrafish and mice xenograft tumor models showed tumor growth arrest upon synergistic autophagy inhibition and antimitotic drug treatment. The success of drug combination treatment was dependent on the status of p53, where tumors with wild-type p53 show greater sensitivity to combinatorial treatment (46). This lends support to the idea that p53 status is an important factor influencing therapeutic success and could be used as a potential biomarker for patient stratification, with exclusion of patients carrying p53-mutant tumours. Currently, the autophagy inhibitors CQ and hydroxychloroquine (HCQ), clinically approved for treatment of malaria, and additional newly discovered, more potent autophagy-modulating compounds are showing promising results in clinical trials for the safe use to overcome chemotherapeutic resistance (47, 48).

Our studies also point to regimens targeting senescence and SASP as potential combinatorial strategies with antimitotic drugs. Although our current work focuses on MT-targeting drugs, we anticipate other classes of chemotherapeutic drugs to have similar roles in engendering non-cell–autonomous tumor progression. Hence, it will be of interest to identify and delineate signaling molecules and pathways leading to SASP and protumorigenic function postslippage. Future studies will focus on nodal points in the “slippage-autophagy-senescence-SASP-tumor progression” axis as these could potentially serve as beneficial therapeutic targets in combination with antimitotic drug therapy. Interestingly, induction of cellular senescence is often correlated with polyploidy (49). It has been posited that polyploid tumors with elevated genomic content tend to become resistant to chemotherapy due to rapid tumor evolution (49). In addition, near-tetraploid cancer cells have recently been shown to exhibit enhanced invasiveness (50). It will be of interest to determine whether high ploidy status per se is able to contribute to SASP-induced tumor progression.

No potential conflicts of interest were disclosed.

Conception and design: B. Cheng, K. Guo, X. Wang, J.E. Connolly, K.C. Crasta

Development of methodology: A.X.F. Wong, B. Cheng, K. Guo, K.J. Lim, J.B.K. Khoo, B.T. Chua, I. Sinha, J.E. Connolly, K.C. Crasta

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Jakhar, M.N.H. Luijten, A.X.F. Wong, B. Cheng, K. Guo, S.P. Neo, B. Au, M. Kulkarni, K.J. Lim, J. Maimaiti, H.C. Chong, E.H. Lim, T.B.K. Tan, K.W. Ong, Y. Sim, J.B.K. Khoo, I. Sinha, J.E. Connolly, J. Gunaratne, K.C. Crasta

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Jakhar, M.N.H. Luijten, A.X.F. Wong, B. Cheng, K. Guo, M. Kulkarni, K.J. Lim, H.C. Chong, B.T. Chua, I. Sinha, X. Wang, J.E. Connolly, J. Gunaratne, K.C. Crasta

Writing, review, and/or revision of the manuscript: R. Jakhar, M.N.H. Luijten, A.X.F. Wong, B. Cheng, K.J. Lim, I. Sinha, J.E. Connolly, J. Gunaratne, K.C. Crasta

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.X.F. Wong, B. Cheng, K. Guo, S.P. Neo, J. Maimaiti, J.T.S. Ho, J.E. Connolly, J. Gunaratne, K.C. Crasta

Study supervision: J.E. Connolly, K.C. Crasta

Other (contributed to mass spectrometry analysis): S.P. Neo, J. Gunaratne

Other (performed biopsies to obtain material): J.S.L. Wong

This research is supported by the National Research Foundation, Prime Minister's Office, Singapore, under its NRF Fellowship Programme (NRF Award no. NRF-NRFF2013-10), Nanyang Assistant Professorship Grant, Nanyang Technological University and the Singapore Ministry of Education Academic Research Fund Tier 1 (Grant no.:2015-T1-002-046-01; provided to K. Crasta). We are grateful to Sixun Chen and other members of the Crasta Lab for valuable comments and help with manuscript preparation. We thank AMPL for histology services, and G.L. Siok and X.L.R. Hai for technical support in MS analysis.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.