Replication stress induces mitotic death through parallel pathways regulated by WAPL and telomere deprotection

Mitotic catastrophe is a broad descriptor encompassing unclear mechanisms of cell death. Here we investigate replication stress-driven mitotic catastrophe in human cells and identify that replication stress principally induces mitotic death signalled through two independent pathways. In p53-compromised cells we find that lethal replication stress confers WAPL-dependent centromere cohesion defects that maintain spindle assembly checkpoint-dependent mitotic arrest in the same cell cycle. Mitotic arrest then drives cohesion fatigue and triggers mitotic death through a primary pathway of BAX/BAK-dependent apoptosis. Simultaneously, a secondary mitotic death pathway is engaged through non-canonical telomere deprotection, regulated by TRF2, Aurora B and ATM. Additionally, we find that suppressing mitotic death in replication stressed cells results in distinct cellular outcomes depending upon how cell death is averted. These data demonstrate how replication stress-induced mitotic catastrophe signals cell death with implications for cancer treatment and cancer genome evolution.

Inducing genotoxic stress is a common mechanism of action for frontline chemotherapeutics.
Overwhelming a cell's capacity to cope with exogenous DNA damage results in lethality through activation of regulated cell death pathways such as apoptosis 1 . Cell death in response to genotoxic stress is more pronounced in actively dividing cells, which imparts the clinical efficacy of genotoxic agents on proliferative cancers. Consistent with the link between proliferation and lethality, many genotoxic agents are associated with cell death during mitosis 2 . Pathways of mitotic death fall under the broad classification of "mitotic catastrophe" 3 .
Mitotic catastrophe is a regulated onco-suppressive mechanism that responds to aberrant mitoses by removing damaged cells from the cycling population 4 . This can occur through cell death or permanent growth arrest. Whilst mitotic catastrophe has been described for some time, there is no clear understanding of the pathway(s) regulating mitotic death 3 . Nevertheless, evidence suggests that mitotic catastrophe is a mechanism of cell death during chemotherapeutic intervention, and that developing resistance to mitotic death would impart a selective advantage to neoplastic cells 5

.
Many genotoxic agents directly or indirectly induce "replication stress" 6 . Replication stress is broadly defined as the slowing or inhibition of DNA replication. This includes endogenous stress originating from oncogene expression, and repetitive and/or secondary DNA structures, or exogenous stress from pharmacological agents that induce DNA damage, reduce nucleotide production, or inhibit DNA polymerases 6,7 . Difficulties arising during DNA replication often manifest as mitotic abnormalities, including chromosome segregation errors, anaphase-or ultra-fine bridges, micronuclei, and the passage of replication stress-induced DNA damage through mitosis [8][9][10][11] . While replication stress is implicated as a driver of mitotic catastrophe, mechanisms of replication stress induced cell death remain unclear.
Here we investigated the cellular response of human primary and cancer cells to pharmacologically induced DNA replication stress with a specific focus on cell lethality. We observed that high dosages of replication stress inducing drugs resulted in mitotic bypass and proliferative arrest without lethality in p53-competent primary fibroblasts. Conversely, in p53compromised fibroblasts and cancer cells, lethal replication stress induces a striking outcome of spindle assembly checkpoint (SAC)-dependent mitotic arrest and cell death principally conferred during mitosis. We found this mitotic death resulted from activation of two independent pathways engaged during mitotic arrest. A primary pathway regulated by wings apart-like protein homologue (WAPL) that drove cohesion fatigue and BAX/BAK-mediated intrinsic apoptosis, and a secondary pathway signalled through non-canonical telomere deprotection. Together, these data reveal novel functions for WAPL and telomeres in sensing replication-stress induced damage and executing mitotic death, with implications for cancer therapy and cancer genome evolution.

Replication stress induces SAC-dependent mitotic arrest and mitotic death
To visualize replication stress induced lethality, we treated human cells with escalating doses of Aphidicolin (APH) or Hydroxyurea (HU) and visualized the outcomes with live-cell imaging (Supplementary Movie 1). APH inhibits family B DNA polymerases, whereas HU inhibits ribonucleotide reductase to limit the nucleoside pool available for nascent DNA synthesis. Cultures were visualized after drug treatment with differential interference contrast (DIC) microscopy every six minutes for up to sixty-five hours (Fig. 1a). Qualitative results indicated a strong mitotic arrest and mitotic death phenotype. We therefore measured mitotic duration from nuclear envelope breakdown to mitotic exit or mitotic death, and classified mitotic outcomes as normal, death, multipolar cell division (mitosis resulting in more than two daughter cells), or mitotic slippage (transition from mitosis to interphase without cell division) (Fig. 1b).
In primary IMR90 fibroblasts, APH or HU treatment induced a marked reduction in mitotic entry and proliferative arrest ( Fig. 1c-e, Supplementary Fig. 1a). In contrast, inhibiting p53 and Rb in IMR90 cells through exogenous expression of HPV16 E6 and E7 (IMR90 E6E7) resulted in a significant increases in mitotic duration and mitotic death with escalating dosages of APH and HU (Fig. 1c, d). Two-dimensional data representation revealed that mitoses which died arose after 20 hours of APH or HU treatment, and that mitotic death correlated with increased mitotic duration (Fig. 1f, Supplementary Fig. 1a). IMR90-hTERT cultures expressing the dominant negative p53DD allele also depicted similar mitotic arrest and mitotic death with delayed kinetics in response to lethal APH, indicating that p53 inhibition is required for replication stress-induced mitotic death in IMR90 cells ( Supplementary Fig.1b, c).
p53-compromised cancer cells also exhibited mitotic arrest and mitotic death with lethal replication stress. HT1080 6TG are a p53 mutant derivative of the HT1080 fibrosarcoma cell line. Treating HT1080 6TG cultures with escalating concentrations of APH and HU revealed concomitant significant increases in mitotic duration and mitotic death (Fig. 1g, h). Mitotic events resulting in death started 20 hours after 1 µM APH, or 30 hours after 500 µM HU treatment, and correlated with increased mitotic duration (Fig. 1i, Supplementary Fig. 1d).
HeLa cervical carcinoma and p53-null Saos-2 osteosarcoma cells also exhibited increased mitotic duration and mitotic death when treated with lethal dosages of APH ( Supplementary   Fig. 1e, f).
Correlation between mitotic duration and death suggested that mitotic arrest drives replication stress lethality. The SAC is regulated by MPS1 kinase and arrests mitosis until tension is established across the mitotic spindle 12 . We tested SAC involvement in replication stressinduced mitotic arrest by performing live cell imaging of HT1080 6TG cultures treated with APH or HU, and the MPS1 inhibitor Reversine 13 . Reversine suppressed mitotic arrest and death, consistent with mitotic arrest being a key determinant of replication stress lethality ( Fig.   1j-l, Supplementary Fig. 1g). Additionally, rescuing mitotic death with Reversine conferred an increase in multipolar cell division in APH treated cells and mitotic slippage in HU treated cultures (Fig. 1k).

Replication stress induces mitotic death in the same cell cycle
Mitotic death in multiple p53-compromised cell lines required twenty or more hours of APH or HU treatment. To determine if replication stress induced lethality occurred the same or subsequent cell cycle, we created fluorescent, ubiquitination-based cell cycle indicator (FUCCI) expressing HT1080 6TG cultures 14 (Fig. 2a). HT1080 6TG-FUCCI cells were treated with APH or DMSO and visualized with DIC and fluorescent live cell imaging every six minutes for up to sixty hours (Supplementary Movie 2). Cells were scored for G1 and S/G2 duration, respectively, by mCherry-hCdt1(30/120) and mVenus-hGeminin(1/110) stability.
Asynchronous HT1080 6TG-FUCCI cells displayed a stratified response to APH treatment characterized by cell cycle phase at the time of drug administration. Cells in S/G2 phase when APH was added to the media, and which remained in S/G2 for < 20 hours with APH, displayed little to no mitotic arrest or cell death phenotype (Fig. 2b, c). Conversely, cells in S/G2 phase when APH was added to the media, and which remained in S/G2 for > 20 hours with APH, displayed common mitotic arrest and mitotic death in the same cell cycle (Fig 2b, c). APH did not impact G1 duration (Fig. 2d). However, when cells exited G1 in the presence of APH they displayed a subsequent extended S/G2, mitotic arrest and mitotic death all in the same cell cycle (Fig. 2b, d). Notably, 85% of cell death events in HT1080 6TG-FUCCI cells observed in the first cell cycle with APH treatment occurred during mitosis (Fig. 2e). Cumulatively, these data indicate lethal replication stress administered early in S-phase predominantly kills cells in the immediately following mitosis.
To determine if persistent APH treatment during mitotic arrest impacts mitotic death, we imaged HT1080 6TG-FUCCI cells during APH treatment and following washout of the drug from the growth media. Cells were cultured with 1 µM APH for 30 hours before media removal and replacement with culture media containing DMSO or fresh 1 µM APH ( Supplementary   Fig. 2a, b). Notably, APH washout failed to rescue lethality in cells already arrested in mitosis.
Additionally, for two hours after APH removal, cells continued to enter mitosis, arrest and die.
Mitotic arrest and death thus did not result from persistent APH treatment during cell division.
Indicating instead that an outcome induced by lethal replication stress, prior to mitotic entry, confers the subsequent arrest and mitotic death.

Replication stress induces mitotic bypass in p53 competent fibroblasts
To determine how replication stress impacted cell cycle progression in p53-competent primary fibroblasts, we live-cell imaged IMR90 FUCCI cultures. APH treatment of IMR90-FUCCI cells induced an extended S/G2, mitotic bypass, and growth arrest in presumably tetraploid G1phase cells (Supplementary Fig. 2c). This is consistent with p53-activation from a genomic DNA damage response (DDR) being sufficient to induce mitotic bypass and proliferative arrest 15,16 . We note APH treated IMR90 FUCCI cells often displayed what appeared as a sustained G1-arrest without S-phase entry ( Supplementary Fig. 2c), likely reflecting early Sphase arrest when mCherry-hCdt1(30/120) remains stable. p53 thus prevents mitotic entry, and therefore mitotic death, in replication stress stressed IMR90 cultures. For the remainder of this study we focused on mitotic death in p53-compromised cells.

Distinct types of mitotic death are induced by replication stress
The specific contribution of cell death pathways in mitotic catastrophe remains poorly understood 3 . BAX and BAK are BCL-2 family proteins essential for apoptosis through their role in mitochondrial outer membrane permeabilization 17,18 . Using CRISPR-Cas9 we generated clonal HeLa and HT1080 6TG BAX and BAK double knock out (BAX BAK DKO) cell lines ( Supplementary Fig. 3a). Parental and BAX BAK DKO cells were treated with APH and visualized with live cell imaging. APH induced mitotic arrest in BAX BAK DKO cultures, with individual mitotic events exhibiting a longer duration mitotic arrest than observed in parental cells (Fig 3a, b, Supplementary Fig. 3b-d). Of note, BAX BAK DKO rescued most, but not all, mitotic death in APH treated cultures at the cost of increased multipolar cell division and mitotic slippage (Fig. 3c, Supplementary Fig. 3e).
Observation of apoptosis-dependent and -independent mitotic death suggested the potential for multiple mitotic death initiating events occurring simultaneously in response to replication stress. To assess this possibility, we imaged live cultures of H2B-mCherry expressing HT1080 6TG cultures treated with a lethal dose of APH. These experiments revealed two chromosome phenotypes associated with replication stress-induced mitotic death (Fig. 3d, Movie S3). "Type 1" mitotic death was defined by a dispersion of chromosomal material immediately preceding lethality, whereas "Type 2" mitotic death was defined by the collapse or condensation of the genetic material without chromosome dispersion (Fig. 3d). Type 1 mitotic death was the dominant outcome, contributing 69.5 ± 1.9% (mean ± s.e.m.) of total cell death events, while Type 2 mitotic death accounted for 23.5 ± 1.2% of cell lethality. The remaining 7.0 ± 2.3% of cell death occurred during interphase (Fig. 3e).

Lethal replication stress induces cohesion fatigue
We characterized the chromosome phenotypes associated with mitotic death by preparing cytogenetic chromosome preparations from APH or HU treated HT1080 6TG cultures (Fig. 4a,   b). This revealed a major phenotype of aberrant chromosome cohesion induced by replication stress (Fig. 4b, c, Supplementary Fig. 4b). We sub-classified the cohesion abnormalities as a "mild" phenotype defined by visible separation of sister centromeres; a "moderate" phenotype of cohesion loss between sister centromeres and the adjacent chromosomal region, with cohesion remaining at distal chromosome arms; and a "severe" phenotype of complete sister chromatid separation (Fig. 4b). Replication stress also induced entangled chromosomes consistent with condensation defects (Supplementary Fig. 4a) 19 .
The mild and moderate cohesion phenotypes initiated from the centromeres (Fig. 4b), suggesting involvement of microtubule pulling forces driving chromatid separation. We tested this by treating cells with a lethal dose of APH or HU in combination with Colcemid or Taxol to inhibit microtubule dynamics, or Reversine to silence the SAC (Fig. 4a). In APH and HU treated cells, Taxol, Colcemid, and Reversine all suppressed the moderate and severe cohesion phenotype (Fig. 4c, Supplementary Fig. 4b). These observations are consistent with "cohesion fatigue", a phenomena where sustained microtubule pulling forces exerted during mitotic arrest drive unscheduled chromosome segregation prior to anaphase onset 20,21 . We interpret these data to indicate the moderate and severe conditions represent a phenotypic continuum of cohesion fatigue as sister chromatids are progressively separated during mitotic arrest.
Conversely, the mild cohesion phenotype was not impacted by Reversine, Taxol or Colcemid.
Indicating instead that the mild phenotype results from replication stress, but not mitotic arrest nor microtubule-dependent forces. While chromosome entanglement occurred in all replication stress conditions (Fig. 4c, Supplementary Fig. 4a, b), the complexity of entangled chromosomes prevented classification of cohesion status and was thus excluded from subsequent quantitative analysis.

Type 1 mitotic death is WAPL and BAX/BAK dependent
The Cohesin antagonist WAPL regulates opening of the Cohesin DNA exit gate to enable dynamic interaction between Cohesin and chromatin 22,23 . WAPL depletion increases Cohesin retention on chromatin and suppresses cohesion fatigue 20,23,24 . To probe for WAPL involvement in the replication stress-induced cohesion phenotypes, we siRNA depleted WAPL and treated cultures with a lethal dose of APH or HU (Fig. 4d, e). Chromosome cohesion phenotypes were measured in cytogenetic chromosome preparations, and mitotic duration and outcome by live-cell imaging ( Fig. 4f-i, Supplementary Fig. 5a-f).
We found that WAPL depletion suppressed the mild, moderate and severe sister chromatid cohesion phenotypes induced by APH and HU in HT1080 6TG cells ( Fig. 4f-g, Supplementary   Fig. 5a). Additionally, WAPL depletion substantially reduced mitotic duration and rescued mitotic death in HT1080 6TG cultures treated with 1 µM APH or 350 µM HU (Fig 4h- Fig. 4b-c). This was not an artefact of reduced proliferation, as APH and HU treated cells continued to enter mitosis in WAPL depleted cultures ( Supplementary Fig. 5d).
WAPL depletion also significantly reduced mitotic duration and mitotic death in 1 µM APH treated HeLa cultures at the cost of a significant increase in multipolar cell divisions ( Supplementary Fig. 5e, f).
As WAPL depletion and BAX BAK DKO both rescued most of the mitotic death induced by replication stress, we reasoned that WAPL and BAX/BAK functioned in the same pathway. To test this, we created HT1080 6TG H2B-mCherry BAX BAK DKO cells ( Supplementary Fig.   5g). We then treated HT1080 6TG H2B-mCherry, WAPL depleted HT1080 6TG H2B-mCherry, or HT1080 6TG H2B-mCherry BAX BAK DKO cells with 1 µM APH and visualized mitotic outcomes with live cell imaging ( Fig. 4j-k, Supplementary Fig. 5h). WAPL depletion completely suppressed Type 1 mitotic death with no significant effect on Type 2 mitotic or interphase death (Fig. 4k). Similarly, BAX BAK DKO also largely suppressed Type 1 mitotic death with no significant effect on Type 2 mitotic or interphase death (Fig. 4k). Together, these data indicate that Type 1 mitotic death is WAPL and BAX/BAK dependent.
Additionally, we quantified how suppressing mitotic death in HT1080 6TG H2B-mCherry cells impacted genome stability. In the absence of replication stress, WAPL depletion induced a slight but significant increase in chromosome segregation errors (6.0% ± 2.4 of mitoses resulting in micronuclei or multipolar cell division, mean ± s.e.m., Supplementary Fig. 5h).

Lethal replication stress induces non-canonical mitotic telomere deprotection
To address potential mechanisms contributing to the minor Type 2 mitotic death pathway, we assayed DDR activation on cytocentrifuged mitotic chromatin following treatment with lethal dosages of APH or HU (Fig. 5a, b). Surprisingly, we observed similar numbers of genomic and telomeric γ-H2AX foci on mitotic chromosomes from APH and HU treated cells (Fig. 5c).
Telomeres represent a minor proportion of the total genome, suggesting induction of a telomere-specific phenotype with lethal replication stress. One possibility is that exogenous replication stress disproportionally impacts the difficult to replicate repetitive G-rich telomeric DNA sequence 25-27 . Replication stress in vertebrate telomeres is attenuated through the telomere-specific TRF1 protein 27 .
To investigate potential contribution of telomere replication stress to mitotic lethality, we overexpressed mCherry-TRF1 (mCherry-TRF1 OE ) in HT1080 6TG cells (Fig. 5d). Surprisingly, mCherry-TRF1 OE failed to reduce the number of mitotic telomere DDR foci observed with APH treatment (Fig. 5e, f). mCherry-TRF1 OE also failed to impact mitotic duration or confer a survival advantage to cells treated with 1 µM APH (Fig. 5g, h). Indicating that the mitotic outcomes observed with lethal APH treatment are independent of replication stress within the telomeric DNA.
Alternatively, the mitotic telomere DDR accompanying lethal replication stress could arise through non-canonical telomere deprotection ( Supplementary Fig. 6a) 28 . Canonical telomere function is to mediate chromosome end protection and regulate cell aging as a function of telomere length 29 . However, telomere-specific DDR activation also occurs through a noncanonical pathway during prolonged mitotic arrest 28 . Mitotic arrest-dependent telomere deprotection is regulated by the Aurora B kinase and the telomere-specific TRF2 protein 28 .
During mitotic arrest in human cells, TRF2 dissociates from chromosome ends resulting in a telomere macromolecular structural change from telomere-loops (t-loops) to linear telomeres 28,30 . This exposes the chromosome end as an ATM substrate to activate the mitotic telomere DDR 30 .
To test for mitotic arrest-dependent telomere deprotection we treated HT1080 6TG cells with lethal dosages of APH or HU in combination with the SAC inhibitor Reversine, the Aurora B inhibitor Hesperadin, or the ATM inhibitor KU-55933 ( Fig. 5i,j) 31,32 . Consistent with noncanonical mitotic telomere deprotection, the telomere DDR induced by lethal replication stress was suppressed by inhibiting Aurora B, ATM, or mitotic arrest (Fig 5j, Supplementary Fig.   6b). Reversine and Hesperadin did not reduce genomic DDR foci induced by APH or HU, indicating the reduced number of telomeric γ-H2AX foci did not result from global DDR suppression ( Fig. 5i, Supplementary Fig. 6b). KU-55933 did reduce genomic DDR foci induced by APH or HU, consistent with the global role for ATM in DDR signalling.
Additionally, no telomere shortening occurred with the telomere DDR in 1 µM APH treated cells, (Fig. 5k), consistent with telomere length independent non-canonical deprotection 28 . We note that mitotic telomere deprotection occurred in parallel with cohesion fatigue as we observed DDR-positive telomeres in mitoses with completely separated sister chromatids ( Supplementary Fig. S6c).

Mitotic telomere deprotection contributes to replication stress lethality
Mitotic telomere deprotection is regulated by TRF2 28 . While TRF2 deletion results in the complete loss of telomere protection and end-to-end chromosome fusions 33  also, respectively, sensitize and suppress non-canonical mitotic telomere deprotection 28 .
To determine if mitotic telomere deprotection impacts replication stress-induced mitotic death, we created HT1080 6TG TRF2 sh-F and HT1080 6TG TRF2 OE cell lines (Fig. 6a). Cell line characterization identified that TRF2 OE and TRF2 sh-F did not impact genomic DNA replication rates, or cohesion fatigue induced by replication stress (Supplementary Fig. 7a-c).
Additionally, no telomere fusions occurred with TRF2 sh-F as a potential source of mitotic arrest 36 ( Supplementary Fig. 7d-f).
Consistent with previous observations, TRF2 sh-F induced a mitotic telomere-specific DDR in the absence of replication stress 34 . However, the telomere DDR in TRF2 sh-F cells was significantly amplified in 1 µM APH treated cells (Fig. 6c, e). Conversely, TRF2 OE suppressed mitotic telomere deprotection in response to replication stress (Fig. 6d, e). TRF2 sh-F and TRF2 OE had an insignificant or very modest impact on mitotic duration with 1 µM APH ( Supplementary Fig. 7g). However, TRF2 sh-F significantly reduced the duration of mitotic arrest until cell death in 1 µM APH treated cultures (Fig. 6f). Additionally, TRF2 sh-F induced a significant increase in mitotic death in 1 µM APH treated cells, while TRF2 OE conferred a significant suppression of mitotic death under the same conditions (Fig. 6g). Cumulatively the data are consistent with mitotic telomere deprotection contributing to a minor proportion of mitotic death in response to lethal replication stress.

DISCUSSION
Here we identify that cell death induced by replication stress occurs predominantly in mitosis through parallel pathways regulated by WAPL and telomere deprotection. Further, we find that suppressing mitotic death enables replication stress to drive genome instability through diverse mechanisms depending on how the cell death is averted. As replication stress is a principle driver of genome instability in oncogenesis 9 , and many frontline chemotherapeutics directly or indirectly induce replication stress 6 , our findings provide insight into mechanisms of cell death, drug resistance, and genome evolution during cancer therapy.
Using extended duration live-cell imaging we identified that lethal replication stress induces SAC-dependent mitotic arrest and mitotic death in diverse p53-compromised human cell lines.
In asynchronous cultures, cells in G1 at the time of APH administration, or cells in S/G2 that took greater than 20 hours with APH to enter mitosis, had the highest incidence of mitotic death. We interpret this result to indicate that replication stress encountered in early S-phase conferred the greatest probability for subsequent mitotic lethality. Notably, lethal replication stress drove mitotic death in the same cell cycle, suggesting mitotic death is an active process to immediately eliminate compromised cells in the following mitosis. We anticipate this prevents accumulation of genetically unstable tissue by preventing proliferation of cells following excessive replication stress.
Replication stress is repaired during mitosis through a mechanism dependent POLD3, a component of DNA polymerase δ which is inhibited by APH 37-39 . However, APH washout failed to rescue the arrest or lethality in cells already in mitosis, or cells entering mitosis in the following two hours. Mitotic arrest is therefore not a result of inhibiting mitotic DNA synthesis.
Instead SAC maintenance resulted from passage of a phenotype induced by replication stress into mitosis. We suggest this is linked to the minor cohesion phenotype. In contrast to the moderate and severe phenotypes, the mild cohesion phenotype was dependent on replication stress but not mitotic arrest nor microtubule dynamics. Failure to establish or maintain cohesion at sister centromeres is expected to impact spindle tension and maintain the SAC 40,41 .
Consistent with this interpretation, we found that WAPL depletion simultaneously rescued both the minor cohesion phenotype and mitotic arrest in APH or HU treated cells.
Our observation of two types of mitotic death induced by replication stress reconciles conflicting reports on cell death mechanisms during mitotic catastrophe, by indicating that replication stress simultaneously engages multiple cell death pathways. The predominant Type 1 mechanism was defined by WAPL-dependent cohesion fatigue and BAX/BAK mediated intrinsic apoptosis. Cytogenetically, we observed most chromosomes in a mitotic cell typically displayed similar cohesion phenotypes, suggesting cohesion fatigue progresses rather uniformly during replication stress-induced mitotic arrest. We were unable to identify how cohesion fatigue signals apoptosis. However, one intriguing possibility is that sudden loss of spindle tension induces signalling to activate the apoptotic cascade. Because cohesion fatigue requires microtubule pulling forces, replication stress-induced mitotic death may differ mechanistically from cell death following treatment with mitotic poisons that inhibit microtubule dynamics.
A significant finding of our study is that WAPL plays a major role mediating replication stresslethality. We anticipate WAPL function extends beyond simply driving cohesion fatigue.
Cohesin dynamically interacts with chromatin and stalled replication forks 42 , and inhibiting WAPL-dependent Cohesin mobilization impacts cell viability in budding yeast treated with replication stress inducing drugs 43 . WAPL also regulates three-dimensional genome architecture though chromatin looping 44-47 , and chromatin looping organizes replication origins to facilitate DNA replication 48 . Consistent with a more expansive role for WAPL in regulating replication stress-induced lethality, we observed that WAPL depletion suppressed both the minor cohesion phenotype and mitotic arrest in APH or HU treated cells. This indicates WAPL regulates centromeric cohesion in response to replication stress to maintain SAC activation in the following mitosis. Future studies will focus on determining how WAPL cooperates with replication stress to confer mitotic arrest, potentially through interaction with Cohesin regulators at centromeres and via its control over chromatin architecture.
We also identified that lethal replication stress induced telomere-specific DDR signalling.
Surprisingly, the mitotic telomere DDR induced by replication stress was resistant to overexpression of the telomere replication regulator TRF1. Instead, the telomere DDR resulted from activation of non-canonical mitotic telomere deprotection. Telomeres sequester the chromosome terminus within a t-loop, which when opened in a linear conformation exposes the DNA end as an ATM activating substrate 30 . We conclude human telomeres signal mitotic abnormalities induced by genomic DNA replication stress through the active alteration of telomere structure to activate DDR signalling. This discovery expands the role for telomeres in regulating genome stability to include the selective removal of cells with excessive genomic DNA replication stress from the cycling population.
Several lines of evidence connect mitotic telomere deprotection to Type 2 mitotic death and indicate that mitotic telomere deprotection and cohesion fatigue progress independently during mitotic arrest. Specifically, WAPL depletion imparted no significant impact on Type 2 mitotic death, indicating that Type 2 cell death was independent of cohesion fatigue. In agreement, augmenting telomere protection by TRF2 overexpression or depletion affected mitotic death in replication stressed cells without impacting cohesion phenotypes. Further, the subtle effect on mitotic death by altering TRF2 expression is consistent with telomeres contributing to the minor Type 2 pathway. Our interpretation is that cohesion fatigue and telomere deprotection occur in parallel during mitotic arrest. This is supported by our observation of numerous DDRpositive telomeres in cytocentrifuged mitotic spreads with completely separated sister chromatids. We suggest mitotic death occurs when either cohesion fatigue or telomere deprotection reaches a signalling threshold within an individual mitosis. Amplifying telomere deprotection therefore shortens the mitotic duration to cell death in replication stressed cells as the aggregate signalling from deprotected telomeres increases. Conversely, suppressing telomere deprotection does not extend mitotic duration when cohesion fatigue remained active to induce lethality.
In response to lethal replication stress, we found that suppressing apoptosis resulted in mitotic slippage, whereas WAPL depletion substantially elevated chromosome segregation errors.
Both of these outcomes rescued cell death at the cost of supernumerary chromosomes.
Normally, supernumerary chromosomes induce a p53-dependent growth arrest 49 . However, p53 inhibition was requisite for mitotic entry in replication stressed cells. We expect mutations that enable p53 compromised cells to avoid mitotic death will impart a substantial selective advantage during replication stress. Specifically, such mutations would enable cells to avoid cell death while simultaneously driving chromosome segregation errors to promote tumour heterogeneity and cancer evolution 50 . Presumably, such mutations would also drive resistance to replication stress-inducing chemotherapeutics. While WAPL mutations are observed in cancers, they occur at a lower frequency than mutations in other Cohesin subunits 51 . Notably, the STAG2 Cohesin subunit is commonly mutated in human malignancies, and some patient derived STAG2 mutations confer reduced interaction between the STAG2 and WAPL proteins 51 . It will be interesting to determine in the future if mutations in WAPL or its regulatory partners confer resistance to mitotic death and if this has potential clinical value in cancer stratification.

Cell culture and treatments
HT1080 6TG cells were a kind gift from Eric Stanbridge (University of California, Irvine) and Saos-2 were provided by Roger Reddel (CMRI). IMR90 cells were purchased from ATCC and

Live cell imaging and analysis
Differential interference contrast (DIC) microscopy was used to visualize mitotic duration and outcome. These experiments were performed on a ZEISS Cell Observer inverted wide field microscope, with 20x 0.8 NA air objective, at 37 ⁰C, 10% CO2 and atmospheric oxygen. Images were captured every six minutes for a duration up to sixty-five hours using an Axiocam 506 monochromatic camera (ZEISS) and Zen software (ZEISS). FUCCI live cell imaging was conducted on the same instrument, using the same imaging duration and protocol, with the addition of a ZEISS HXP 120C mercury short-arc lamp and compatible filter cubes to obtain fluorescent images. Quantitative live cell imaging of mitotic chromosome dynamics in HT1080 6TG H2B-mCherry cultures and derivatives was done using either combined DIC and fluorescent imaging on the ZEISS Cell Observer described above with a 40x 0.95 NA plan-Apochromat air objective; or on a Zeiss Cell Observer SD spinning disk confocal microscope imaged using a 561nm, 50mW solid state excitation laser and appropriate filter sets with a 40x,

NA oil objective using an Evolve Delta camera (Photometrics). For these experiments, cells
were cultured at 37⁰ C, 10% CO2 and atmospheric oxygen and images captured every three minutes for sixtry hours. For all movies, mitotic duration was scored by eye and calculated from nuclear envelope breakdown until cytokinesis or mitotic death. FUCCI Movies were scored by eye, for G1 (red) and S/G2 (green) by color, and for mitotic entry, duration and outcome by cell morphology. Chromosome dynamics in HT1080 6TG-mCherry H2B were scored by eye. All live cell imaging analysis was done using Zen software.

Visualization of chromatid cohesion and mitotic DDR activation
For cohesion analysis, standard methanol and acetic acid fixed cytogenetic chromosome spreads were prepared and stained with fluorescent in situ hybridization (FISH) with a telomere (Alexa488-OO-ccctaaccctaaccctaa), and centromere (Texas Red-OO-aaactagacagaagcatt), peptide nucleic acid (PNA) probes (Panagene), as described elsewhere 53 . The mitotic telomere DDR was visualized using the "metaphase-TIF assay" as described previously 53  and DDR(-) chromosome ends, telomere signals were captured at fixed exposure time as described above, and images were analysed using Imaris software version 8.2 (BitPlane).
Images were segmented using the mitotic chromosomes as a region of interest, and the telomeres were detected as spots using the spot detection wizard. Telomere colocalization with γ-H2AX foci were identified using the co-localisation tool. The sum fluorescence intensity was determined for each telomere in the γ-H2AX colocalised (DDR+) and the γ-H2AX excluded (DDR-) cohorts.

Chromatin fibre analysis
Replication rates were measured using chromatin fiber analysis as described elsewhere 54 .
Briefly, unsynchronized cells are sequentially pulse-labelled with 100 μM of thymidine

Western blotting
Preparation of whole cell extracts and western blots were done as described elsewhere 30 and luminescence was visualized on an LAS 4000 Imager (Fujifilm).    For all panels, * p < 0.05, ** p < 0.01.          , c)       and IMR90 E6E7 treated with Hydroxyurea (HU) (data corresponding to Fig. 1c, d). Data are presented as described in Fig. 1d