Cell cycle progression and transmitotic apoptosis resistance promote escape from extrinsic apoptosis

Healthy development and physiological organ functions rely on appropriately balanced cell proliferation and cell death, and consequently dysregulations in this balance manifest in both proliferative and degenerative diseases (Hanahan and Weinberg, 2011; Zhivotovsky and Orrenius, 2010). Proliferation signalling drives progression through the cell cycle and accumulation of cell mass, whereas apoptosis, the major form of programmed cell death, eliminates cells by triggering proteolytic cascades in which caspases act as cell death initiators and executors. Two main apoptosis pathways activate caspases. The extrinsic pathway is triggered by death ligands binding to their respective cell surface receptor(s), whereas the intrinsic, mitochondrial pathway is initiated in response to cytotoxic stress (Taylor et al., 2008). The extrinsic pathway can be triggered by immune cells to induce apoptosis in transformed and potentially harmful cells (Nagata and Tanaka, 2017), and the death ligand TRAIL (also known as TNFSF10) now serves as a template for the development of novel anti-cancer therapeutics and treatment strategies, with second-generation ligands recently having entered clinical trials (Phillips et al., 2021; Von Karstedt et al., 2017).

Following ligand binding, activated death receptors cluster and form the death-inducing signalling complex (DISC), the activation platform for initiator caspase-8. Active caspase-8 in select cell types sufficiently cleave and activate effector caspase-3, resulting in type I apoptosis (Scaffidi et al., 1998). Typically, however, the presence of the inactive caspase-8 homologue cFLIP (also known as CFLAR) as well as the caspase-3 inhibitor XIAP prevent this type of apoptosis. Instead, the majority of cells rely on the amplification of the apoptotic signal via the mitochondrial pathway (i.e. type II signalling; Danish et al., 2018; Jost et al., 2009; Wilson et al., 2009). Here, caspase-8 cleaves the BH3-only protein Bid, which then translocates to the mitochondria and stimulates Bax- and Bak (also known as BAX and BAK1, respectively)-dependent mitochondrial outer membrane permeabilisation (MOMP), a process that is inhibited by anti-apoptotic Bcl-2 family members such as Bcl-2, Bcl-xL (encoded by BCL2L1) and Mcl-1 (Kalkavan and Green, 2018). MOMP results in the release of cytochrome c and Smac (also known as DIABLO) from the mitochondria, driving efficient caspase-3 activation and apoptosis execution (Tait and Green, 2010).

Whereas prolonged cell cycle arrest, for example by cytotoxic drugs such as doxorubicin, CPT-11, bortezomib, simvastatin or the anti-microtubule agent taxol, evidently induces stress responses that sensitize cells to extrinsic apoptosis (Bullenkamp et al., 2014; Ehrhardt et al., 2013; Jin et al., 2002; Kim et al., 2008; Ray et al., 2007), it so far remains unsolved whether extrinsic apoptosis susceptibility is modulated during normal cell cycle progression. This may not be surprising, since combined and parallel monitoring of externally unperturbed cell cycle progression and apoptosis signalling is challenging. For example, bulk analyses of chemically synchronized cell populations provide little temporal resolution and suffer from interfering with normal cell cycle progression (Cooper, 2003). Single-cell time-lapse analyses, instead, so far remain restricted to studies conducted in the presence of apoptosis sensitizers, such as the translation inhibitor cycloheximide (Rehm et al., 2009; Spencer et al., 2009). Given that cell cycle progression heavily relies on concerted protein production and degradation, and since short-lived modulators of extrinsic apoptosis responsiveness, such as cFLIP, are rapidly lost upon translation inhibition (Poukkula et al., 2005), the question of whether cell cycle progression modulates extrinsic apoptosis signal transduction remains largely unaddressed so far.

Here, we therefore combined mathematical modelling and live-cell imaging to quantitatively study extrinsic apoptosis in TRAIL-susceptible cell lines during unperturbed cell cycle progression, with a specific focus on investigating whether and how cell cycle progression influences how and when cells commit to apoptotic cell death. Interestingly, we found that cells progressing towards mitosis delay apoptotic signal transduction, which was attributable to an intrinsic, cell cycle-associated upregulation of Mcl-1 and, to a lesser extent, Bcl-xL. These cells transition through mitosis with activated caspase-8 and frequently escape cell death, but show signs of sublethal DNA damage. Importantly, Mcl-1-targeting BH3-mimetics suffice to prevent unwanted transmitotic escape from extrinsic apoptosis.

To analyse whether cell cycle progression influences the timing of apoptosis execution in response to TRAIL treatment, we employed a combined approach of mathematical modelling and experimental observation of cell divisions and cell death events. To non-invasively monitor cell cycle phases, we generated NCI-H460 and HCT-116 cells stably expressing the fluorescently tagged degron of geminin (Kuritz et al., 2017; Sakaue-Sawano et al., 2008), which intrinsically labels cells throughout S, G₂ and M phases, but not in G₁ (Fig. S1A). Treatment with a translationally relevant TRAIL receptor (TRAIL-R) agonist (Fc-scTRAIL) (Hutt et al., 2017) dose-dependently reduced cell viability, accompanied by cleavage of caspase-8 and caspase-3, as well as PARP (Fig. S1B,C). Addition of the pan-caspase inhibitor QVD-OPh averted cell death, demonstrating that cell death proceeded entirely by apoptosis (Fig. S1D). Following TRAIL addition, we microscopically recorded whether cells executed apoptosis prior to cell division or instead underwent mitosis before executing apoptosis. These data were then compared to expected patterns of apoptosis execution obtained from a mathematical cell population model in which progression towards mitosis and the timing of apoptosis were implemented as independent processes (Fig. 1A) (Imig et al., 2020). Although modelling predicted that ∼20% of the NCI-H460 and HCT-116 cell populations would be expected to divide before executing apoptosis, significantly more cells underwent cell division prior to apoptosis in these experiments (Fig. 1B). Closer examination of the imaging data indeed revealed that geminin-positive cells in S/G₂ phases frequently first underwent mitosis, with daughter cells dying soon after cytokinesis (Fig. 1C; Fig. S1E and Movie 1). Correspondingly, when both model and experimental analyses were constrained to cells in S/G₂/M phases at the time of TRAIL addition, the mismatch in the timing between apoptosis execution and mitosis was even more pronounced (Fig. 1D). The comparison of mathematical predictions and experimental observations of cell death and cell division events therefore suggests that cell cycle progression might delay or decelerate the kinetics of apoptosis signal transduction, and that such delays would be expected to manifest predominantly during the S/G₂/M phases of the cell cycle.

We next experimentally studied whether signal transduction kinetics towards apoptosis execution indeed differed between cells exposed to TRAIL at different times of the cell cycle and, if this is the case, at which point during cell cycle progression the decision is made to delay apoptosis in favour of mitosis. To this end, we monitored untreated cells for at least 20 h, so that cell divisions could be observed, and the age of each individual cell was known at a resolution of minutes at the time of TRAIL addition. In parallel, we tracked the durations of cell cycle phases based on geminin intensity and cell morphology, as well as the times from TRAIL addition to apoptosis execution (Fig. 2A). At the population level, NCI-H460 and HCT-116 cells underwent apoptosis as soon as 1 h following TRAIL addition, with a median t_death of ∼3 h (2.8 h and 3.7 h, respectively) (Fig. 2B,C, orange). When separating cell populations based on the cell cycle phase in which they were exposed to TRAIL (Fig. 2A), we noted that cells exposed to TRAIL in S/G₂/M phases required significantly longer to die than cells exposed to TRAIL in G₁ phase (Fig. 2B,C). The delays observed in the S/G₂/M populations could be ascribed to cells dying after mitosis (t_death calculated from f0 and f1 cell generations) (Fig. 2B,C). Similar results were obtained in HeLa and HT1080 cells treated with TRAIL (Fig. 2D,E), in NCI-H460 cells treated with FasL (also known as FASLG) as an alternative death ligand (Fig. S2A), and in two primary cell isolates obtained from melanoma metastases (Fig. 2F).

We next sought to identify when during the S/G₂/M phases cells decide to delay apoptosis in favour of mitosis. Following data normalisation (Imig et al., 2020), we plotted the t_death of individual cells against their cell cycle position. This analysis revealed that cells notably delaying apoptosis execution following TRAIL exposure had completed approximately mid-to-late S phase (Fig. 2G,H).

Taken together, these results demonstrate that many cells that would have had sufficient time to die prior to mitosis instead undergo cell division and delay extrinsic apoptosis, a behaviour that manifests once cells have entered mid to late S phase.

We next studied at which point during signal transduction the propagation of extrinsic apoptotic signalling is delayed. Surface amounts of TRAIL-R1 and -2 (TNFRSF10A and TNFRSF10B) continuously increased from G₁ throughout S/G₂ and M phases in both NCI-H460 and HCT-116 cells, with M phase amounts approximately twice as high as in G₁ (Fig. S2B,C), as would be expected for proteins whose steady state amounts correlate with cell growth (Lin and Amir, 2018). TRAIL binding correlated with these changes (Fig. S2D,E). These results therefore exclude that delays in apoptosis signalling emanate from lack of receptors or impaired ligand binding in later phases of the cell cycle. Next, we studied caspase-8 processing and activity in cells exposed to TRAIL at different times of the cell cycle. Since the activation of downstream caspases interferes with measurements of initiator caspase processing and activities (McStay et al., 2008), we used HCT-116 cells deficient for Bax and Bak expression [hereafter denoted HCT-116 (Bax/Bak)^−/−] (Fig. S3A), which are unable to undergo MOMP and apoptosis execution, in our initial analyses. HCT-116 (Bax/Bak)^−/− cells sorted by geminin fluorescence for G₁ or G₂/M phases similarly processed pro-caspase-8 into p43/41 and p18 subunits (Fig. 3A). We next studied whether caspase-8 processing results in notable activity during S/G₂/M phases. To this end, we introduced a FRET probe carrying the optimal recognition motif (IETD) for caspase-8-dependent proteolysis (Hellwig et al., 2008) into HCT-116 (Bax/Bak)^−/− cells or NCI-H460/Bcl-2 cells (which overexpress Bcl-2; Fig. S3B). Time-lapse imaging revealed comparable caspase-8 activities in all stages of the cell cycle, and that cells can enter and pass through mitosis with activated caspase-8 (Fig. 3B; Fig. S3C,D). Corresponding to these observations, cFLIP, the first substrate directly cleaved by caspase-8 within the DISC, is equally cleaved in G₁ or G₂/M cells (Fig. 3C). Cleavage of Bid, the substrate of caspase-8 that links caspase-8 activity to the mitochondrial apoptosis pathway and which is essential for TRAIL-induced viability loss in NCI-H460 and HCT-116 cells (Fig. 3D), was indistinguishable between cells in G₁ or G₂/M phases (Fig. 3E).

Caspase-8 additionally cleaves and activates effector pro-caspase-3. In cells relying on the mitochondrial apoptosis pathway, X-linked inhibitor of apoptosis protein (XIAP) suppresses full autoproteolytic maturation of the large caspase-3 subunit (Deveraux et al., 1997; Jost et al., 2009). Consequently, in the absence of MOMP and amplification of apoptosis signalling, caspase-3 processing is limited and only the initial cleavage product of the large caspase-3 subunit can be detected. Indeed, this was seen when comparing caspase-3 processing between HCT-116 (Bax/Bak)^−/− cells, parental HCT-116 cells and NCI-H460 cells (Fig. 3F). Similar patterns of caspase-3 processing were observed in Bid-depleted (shBid) cells (Fig. 3G). Unfortunately, parental cells induced to undergo apoptosis were too fragile to be sorted into G₁ and G₂/M subpopulations, but instead could be fixed and analysed by flow cytometry. Staining HCT-116 (Bax/Bak)^−/− cells for DNA and processed caspase-3, revealed identical pro-caspase-3 processing in G₁, S and G₂/M phases (Fig. 3Hi). In MOMP-competent HCT-116 cells, instead, additional peaks for cells with fully cleaved caspase-3 pools could be detected. Interestingly, these were restricted to cells in G₁ and S phases, indicating that cells in G₂/M phases suppress full processing of caspase-3 (Fig. 3Hii). NCI-H460 cells similarly limited caspase-3 processing in G₂/M phases, indicative of evading the post-MOMP phase of apoptosis execution (Fig. 3Hiii). When quantifying the kinetics of pro-caspase-3 cleavage events in the distinct subpopulations, we noted that initial pro-caspase-3 cleavage occurred independently of the cell cycle in both NCI-H460 and HCT-116 cells (Fig. 3I). Pre- and post-MOMP caspase-3 processing followed very similar kinetics in cells from G₁ and S phases, whereas the post-MOMP subpopulation remained very low in cells analysed in G₂/M phases (Fig. 3I).

Taken together, we conclude that caspase-8 is activated independently of the cell cycle phase, that caspase-8 remains active throughout mitosis and that activity can be transmitted to daughter cells. The natural caspase-8 substrates cFLIP, Bid and pro-caspase-3 can be processed in all phases of the cell cycle. However, cells analysed in G₂/M phases can prevent full caspase-3 processing, suggesting that MOMP is not triggered in these subpopulations.

Given that cells are capable of evading MOMP and that daughter cells inherit caspase-8 activity following TRAIL treatment, we studied whether such caspase-8 activities inevitably result in apoptosis execution after mitosis. Even though we observed cells dying after mitosis (Fig. 2), we also noted that substantial numbers of cells remained viable after long treatment times (Fig. S1B). As expected, cells dying after mitosis lost their mitochondrial membrane potential, a surrogate marker for MOMP, just prior to displaying apoptotic phenotypes (Fig. 4A,B). Moreover, we occasionally observed that sibling cells may experience distinct cell fates after mitosis, with mitochondrial depolarization and apoptosis observed in one cell but not in the other (Fig. 4B). It is therefore conceivable that cells might escape apoptosis induction subsequent to caspase-8 activation. We consequently studied whether caspase activities could be observed in cells not showing overt apoptotic phenotypes following TRAIL treatment and mitosis. As a reference, TRAIL treatment caused dramatic caspase-dependent FRET disruption in non-dividing, apoptosis-competent NCI-H460 and HCT116 cells expressing the IETD FRET probe (Fig. 4C). Analysing cells passing through mitosis and subsequently either dying or failing to show an apoptotic phenotype revealed notable differences. Cells dying after mitosis displayed FRET loss that was similar to that in non-dividing cells, whereas cells that escaped apoptosis cleaved significantly less FRET probe (Fig. 4D). Importantly, FRET efficiency in apoptosis-averting cells was significantly different from that in untreated control cells, indicating that caspases were activated in sublethal amounts (Fig. 4D). Apparently, many of these cells suffer DNA damage, as detected by phosphorylation of H2A.X (pH2A.X; also known as γH2AX), a well-established marker for DNA double strand breaks (Fig. 4E,F). Addition of QVD-OPh or siRNA-mediated silencing of caspase-activated DNase (CAD; also known as DFFB), the major executioner of apoptotic DNA fragmentation, suppressed pH2A.X signals, demonstrating that DNA damage was caspase dependent (Fig. 4F,G).

Overall, these experiments therefore indicate that caspase-8 activity, which can persist through mitosis, does not inevitably induce cell death in daughter cells. Cells surviving TRAIL treatment, however, display overt signs of caspase-dependent DNA damage.

Cells delaying or escaping from TRAIL-induced apoptosis do so by averting MOMP in late stages of the cell cycle. We therefore studied which molecular mechanisms might confer this increased MOMP resistance. The anti-apoptotic Bcl-2 family members Mcl-1 and Bcl-xL, key regulators of MOMP sensitivity, have previously been reported to prevent or suppress apoptosis during extended mitotic arrest induced by anti-mitotic drugs such as paclitaxel and taxol (Bah et al., 2014; Harley et al., 2010; Haschka et al., 2015; Shi et al., 2011; Sloss et al., 2016). First, we therefore determined absolute expression quantities for both proteins in NCI-H460 and HCT-116 cells, in comparison to HeLa cells for which nanomolar expression amounts were reported previously (Lindner et al., 2013). Both cell lines expressed higher levels of Bcl-xL than Mcl-1, and NCI-H460 cells expressed ∼9-fold higher amounts of Mcl-1 than HCT-116 cells (Fig. 5A). The combined expression of Bcl-xL and Mcl-1 was, however, comparable between both cell lines. Studying expression of both proteins across the cell cycle indicated that both cell lines substantially elevated Mcl-1 in M phase (i.e. above the expected doubling throughout the cell cycle; Lin and Amir, 2018; Soltani et al., 2016), as determined by flow cytometry (Fig. 5B). This finding is in line with the Mcl-1 accumulation observed across the cell cycle in synchronised U2OS cell populations (Harley et al., 2010). Higher than expected amounts of Bcl-xL were observed only in HCT-116 cells (Fig. 5B). Immunoblotting independently confirmed the upregulation of Mcl-1 accompanied by increased Noxa (also known as PMAIP1) in later stages of the cell cycle (Fig. 5C). Importantly, TRAIL treatment did not change amounts of Mcl-1 or Bcl-xL (Fig. S4A). A more detailed kinetic analysis of Bcl-xL and Mcl-1 expression across the cell cycle likewise confirmed the increase in Mcl-1, beginning in mid-to-end S phase and accumulating substantially in late G₂/M phases (Fig. 5D,E).

We next studied which processes might contribute to the accumulation of Mcl-1 in late cell cycle stages. In contrast to cyclin B1 (CCNB1) mRNA, Mcl-1 mRNA did not accumulate in G₂/M phases, so that transcriptional upregulation could be excluded (Fig. 5F). After short periods of proteasome inhibition induced by bortezomib (BTZ) treatment, Mcl-1 accumulated in cells from both G₁ and S/G₂/M phases (Fig. 5G). In reverse, Mcl-1 amounts decreased in both groups of cells when inhibiting protein translation through cycloheximide (CHX) treatment (Fig. 5G). Mcl-1 is therefore translated and degraded in early and in later stages of the cell cycle, so that a combination of alterations in both translation and/or degradation rates would contribute to a net increase in Mcl-1 amounts prior to mitosis. Multi-domain Bcl-2 family members continuously shuttle between mitochondrial membranes and the cytosol (Edlich et al., 2011). Given that proteasomal degradation is limited to the soluble protein fraction, increased Mcl-1 stability could also be achieved by a higher membrane association during later cell cycle stages. Studying subcellular distributions indeed indicated a significant redistribution of Mcl-1 towards the mitochondrial fraction in later cell cycle stages, which was not observed for Bcl-xL (Fig. 5H). Overall, these data suggest that the pronounced accumulation of Mcl-1 towards mitosis and its redistribution towards mitochondrial membranes could contribute to delaying or preventing apoptosis in cells transitioning through mitosis.

To assess whether anti-apoptotic Bcl-2 family members interfere with apoptosis signal transduction during later stages of the cell cycle, we next studied cells in which Bcl-2, Bcl-xL or Mcl-1 were pharmacologically inhibited. In NCI-H460 cells, Bcl-2 inhibition with ABT-199 or Bcl-xL inhibition by WEHI-539 moderately reduced t_death times in cells passing through mitosis, whereas Mcl-1 inhibition by S63845 was significantly more effective (Fig. 6A,B compare to Fig. 2G). A similar shortening of t_death times was observed in HCT-116 (Fig. 6A), HeLa and HT1080 (Fig. S5A,B) cells upon Mcl-1 inhibition. Additionally, treatment with the Bcl-xL inhibitor WEHI-539 also prevented prolonged t_death times in HCT-116 (Fig. 6A), HeLa or HT1080 (Fig. S5A,B) cells passing through mitosis, possibly due to their high dependency on Bcl-xL expression (Fig. 5A). As expected, Bcl-2 family antagonists also slightly affected Fc-scTRAIL-induced death kinetics in cells that did not progress through mitosis (Fig. S5C,D). Since Mcl-1 inhibition averts delays in apoptosis signalling in cells preparing to enter mitosis, it would be expected that fewer cells overall are able to progress through mitosis under these treatment conditions. Indeed, the frequency of mitotic events observed for TRAIL-treated NCI-H460 and HCT-116 cells was significantly reduced when co-treated with S63845 (Fig. 6C).

Importantly, none of the inhibitors reduced cell viability or induced notable apoptosis in absence of TRAIL treatment (Fig. S5E,F). Using A-1210477 as an alternative Mcl-1 inhibitor gave similar findings (Fig. S5G,H). Similar to what was seen upon pharmacological Mcl-1 inhibition, depleting Mcl-1 expression by siRNA also abolished prolonged t_death times and reduced the frequency of cell divisions following TRAIL treatment (Fig. 6D–F). In line with these findings, flow cytometric studies revealed that TRAIL-treated cells no longer suppress caspase-3 processing in G₂/M phases (Fig. 6G,H, compare to Fig. 3H,I). Instead, the kinetics of caspase-3 processing were largely identical in cells from G₁, S and G₂/M phases.

Together, these findings demonstrate that Mcl-1 confers transient resistance to TRAIL-induced extrinsic apoptosis in cells entering and passing through mitosis, and that this resistance can be antagonized by inhibiting or depleting Mcl-1. Consequently, the combination of TRAIL and Mcl-1 inhibition almost completely abrogated long-term survival of NCI-H460, HCT-116 (Fig. 6I,J), HeLa or HT1080 (Fig. S5I,J) cells, with HCT-116 cell survival also markedly reduced by co-treatment with Bcl-xL inhibitor (Fig. 6I).

Here, we showed that cells progressing towards and passing through mitosis delay TRAIL-induced extrinsic apoptosis. Importantly, cells transition through mitosis with active caspase-8, and are protected from mitochondrial apoptosis, but nevertheless accumulate caspase-dependent sublethal DNA damage. Upregulation of Mcl-1 and, to a lesser extent, Bcl-xL confers transmitotic apoptosis resistance, with BH3-mimetics targeting these proteins prevent unwanted escape from extrinsic apoptosis.

To obtain insight into the relationship between extrinsic apoptosis signalling and cell cycle progression, we employed fluorescence-based time-lapse imaging of non-synchronized cell populations. This circumvented the limited temporal resolution and accuracy of traditional chemical synchronization (Cooper, 2003). Furthermore, cell cycle arrest for synchronization purposes not only would have interfered with normal cell cycle progression but could have further confounded the analyses by inducing stress responses that modify apoptosis sensitivities in distinct cell cycle phases.

We found that cells delay or prevent apoptosis execution when proceeding towards cell division, with the decision on whether to favour mitosis over apoptosis being made approximately half way through S phase. Decision making therefore substantially precedes entry into mitosis and gives rise to prolonged periods of increased apoptosis resistance, which also explains why cell death events are more frequently observed during G₁ phase. The cell-to-cell signalling heterogeneity during the pre-MOMP phase so far was solely attributed to protein expression noise across cell populations or to variation in survival pathway activities (Baskar et al., 2019; Spencer et al., 2009), whereas here we show that cell cycle progression adds an additional and significant non-random layer of cell death regulation. Since we observed similar delays in extrinsic apoptosis signalling when stimulating NCI-H460 cells with FasL, transmitotic apoptosis resistance seems independent of the choice of death ligand.

Mcl-1 is the primary factor conferring transmitotic resistance to extrinsic apoptosis. A key role for Mcl-1 in delaying or suppressing cell death has been reported previously in conditions of mitotic arrest (Allan et al., 2018; Harley et al., 2010; Haschka et al., 2015; Sloss et al., 2016; Wertz et al., 2011), and various E3 ligases have been reported that could contribute to eliminating Mcl-1 (Senichkin et al., 2020). In our setting, where mitosis proceeds unperturbed, cells rapidly deplete Mcl-1 upon entry into G₁ phase, with many daughter cells regaining the competency to die due to their inherited caspase-8 activity. Whether one or several of the known Mcl-1 E3 ligases is implicated in rapid Mcl-1 depletion upon regular exit from mitosis remains to be studied.

We found that Mcl-1 steadily accumulates from mid S phase on, matching trends reported for synchronized cell populations (Harley et al., 2010), but also to additionally sharply increase at the very end of the cell cycle. Apoptosis resistance must be very high during this period, since we never observed apoptosis execution during mitosis itself, despite analysing hundreds of cells. Since transcription is largely blocked upon entry into mitosis, Mcl-1 accumulation could arise from continued translation during M phase, possibly combined with suppressed degradation, continuing trends we observed when analysing the pooled fraction of cells in G₂/M phases. Indeed, Mcl-1 mRNA continues to be translated during mitosis, as shown in other studies (Sloss et al., 2016; Zhou et al., 2013). Besides translation and proteasomal degradation contributing to defining Mcl-1 amounts throughout the cell cycle, we observed a significant redistribution of Mcl-1 towards mitochondrial membranes in later cell cycle phases. This offers a possibly underappreciated mechanism by which Mcl-1 could be depleted from the degradable soluble fractions and which thereby could contribute to protein stabilisation. Bcl-xL likewise accumulates across the cell cycle but lacks the sharp increase in its amounts during mitosis. While Bcl-xL can reduce apoptosis sensitivity during mitotic arrest by paclitaxel in breast cancer cells (Bah et al., 2014), in our cell line models Mcl-1 inhibition more consistently avoided transmitotic apoptosis resistance. Mcl-1 inhibition likewise outperforms Bcl-xL inhibition in apoptosis induction by mitotic driver drugs, such as aurora kinase inhibitors (Bennett et al., 2016). The third major anti-apoptotic Bcl-2 family member, Bcl-2 itself, is not regulated by cell cycle progression (Harley et al., 2010) and instead is apparently inactivated by phosphorylation prior to/upon entry into mitosis (Terrano et al., 2010). Similar activity-suppressing phosphorylation has also been reported for Bcl-xL (Bah et al., 2014; Basu and Haldar, 2003; Terrano et al., 2010). Overall, this indicates that in particular during mitosis, apoptosis resistance would be expected to strongly rely on Mcl-1. Earlier studies have reported the suppression of caspase-8 activation by Ser-387 phosphorylation in cells arrested in mitosis (Matthess et al., 2010, 2014), whereas we observed that caspase-8 can be activated and remains active in all phases of the cell cycle. The phosphorylation-dependent inhibition of caspase-8 activation therefore might be restricted to cells arrested in mitosis rather than manifesting prominently during regular cell cycle progression or might represent a layer of regulation that does not prominently display in the cellular model systems studied here. Likewise, additional regulatory steps downstream of MOMP, such as the inactivation of caspase-9 by phosphorylation during mitosis (Allan and Clarke, 2007), could further contribute to establishing transiently higher apoptosis resistance.

A more general question arises regarding why cells would need to raise their apoptosis resistance in preparation for and during mitosis. During S phase, cells expend substantial amounts of energy on synthesis and growth processes, which in themselves might impose increased basal stress against which cells protect themselves. Alternatively or in addition, proliferation might be associated with mild or moderate intrinsic pro-apoptotic activities that are not supposed to lead to cell death. Indeed, various apoptotic proteins have been ascribed roles that are independent of cell death. For example, cellular remodelling during development and cell fate determination of stem cells, as well as the regulation of cytoskeletal dynamics partially depend on non-apoptotic caspase signalling (Nakajima and Kuranaga, 2017). Interestingly, RIPK1-containing complexes form during mitosis to activate caspase-8, which contributes to orderly chromosome segregation, without these cells showing signs of apoptosis (Liccardi et al., 2019). Furthermore, caspase-8 activity seems to be required to overcome the G₂/M checkpoint by p53 destabilization (Müller et al., 2020). This would imply that apoptosis thresholds must be elevated during mitosis to withstand unwanted apoptosis. In line with this, cells in which Mcl-1 and Bcl-xL are pharmacologically inhibited spontaneously and preferentially die during mitosis (Bennett et al., 2016). Evasion from extrinsic apoptosis thus would arise as an unwanted side effect, which however could have substantial implications for the efficacy by which the immune system can clear transformed cells and, by extension, the risk to relapse from immune stimulatory or death ligand-based therapy (Brown et al., 2018; Seidel et al., 2018; Von Karstedt et al., 2017). Inhibitors of Mcl-1 and Bcl-xL might therefore have a space as co-treatments in such scenarios, especially since Mcl-1 is highly amplified in various cancers (Beroukhim et al., 2010). Results from a clinically relevant TRAIL variant, used in combination with ABT-199 as a Bcl-2 inhibitor, and active clinical studies (Phillips et al., 2021; NCT03082209), could also serve as frameworks in which similar Mcl-1 and Bcl-xL antagonists could be studied. Clinically relevant scenarios might also include treatments with MAPK inhibitors, where Mcl-1 amounts appear to co-determine response variabilities in isogenic cell populations (Inde et al., 2020). Moreover, the Mcl-1 dependency of malignant neoplasms generally might be elevated, since recent CRISPR-Cas9 screening identified Mcl-1 and, with a lower score, Bcl-xL as attractive priority targets in colorectal cancers (Behan et al., 2019).

We noted caspase-dependent DNA damage in cells escaping TRAIL-induced apoptosis by passing through mitosis. DNA damage in TRAIL-surviving cells has been previously described (Lovric and Hawkins, 2010; Miles and Hawkins, 2017), and such damage could arise from moderate activation of caspase-3 either through direct cleavage by caspase-8 and/or the inefficient activation of MOMP, resulting in minority MOMP, a process of inefficient MOMP induction that can contribute to further cell transformation in response to Bcl-2 family antagonism and to inflammatory cytokine production in infection scenarios (Brokatzky et al., 2019; Ichim et al., 2015). Given that caspase-8 can also cleave DEVD sequences, such as found in PARP (Boldin et al., 1996; McStay et al., 2008), it cannot be excluded that caspase-8 itself likewise contributes to DNA damage.

Taken together, our study shows that regular cell cycle progression modulates signal transduction during extrinsic apoptosis, with Mcl-1 modulating cellular decision making between death, proliferation and survival from inefficient apoptosis. These findings not only extend our understanding of the temporal regulation of extrinsic apoptosis susceptibility but also are relevant for better understanding cellular mechanisms contributing to escape from apoptosis induction.

Cells were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum (FBS) and kept at 37°C in 5% CO₂ humidified atmosphere. HCT-116 cells were obtained from Interlab Cell Line Collection (Italy). NCI-H460, HeLa and HT1080 cells were obtained from ATCC (LGC Standards GmbH, Germany). Cell isolates from melanoma metastases were kindly provided by Dagmar Kulms (Department of Dermatology, University of Dresden, Germany) and obtained as part of routine resections at University Hospital Dresden, under the auspices of the local Ethics Committee (ethical approval number EK335082018). Platinum-E (Plat-E) cells were kindly provided by Monilola Olayioye, University of Stuttgart, Germany. Cell lines were routinely tested for mycoplasma infection and verified by multiplex human cell line authentication test (Multiplexion GmbH, Germany). NCI-H460/geminin and NCI-H460/Bcl-2 cells were described previously (Danish et al., 2018; Kuritz et al., 2017), HCT-116 (Bax/Bak)^−/− cells were kindly provided by Richard J. Youle (National Institutes of Health, USA).

Primary antibodies were as follows: mouse anti-β-Actin (8H10D10, Cell Signaling #3700, 1:2000), rabbit anti-Bak (Cell Signaling #3814, 1:1000), rabbit anti-Bax (Cell Signaling #2772, 1:1000), rabbit anti-Bcl-xL [Cell Signaling #2764, 1:1000 western blotting (WB), 1:200 flow cytometry (FC) and immunofluoresence (IF)], rabbit anti-Bid (Cell Signaling #2002, 1:1000), rabbit anti-Bim (Cell Signaling #2933, 1:1000), mouse anti-caspase-8 (1C12, Cell Signaling #9746, 1:1000), rabbit anti-caspase-3 (Cell Signaling #9662, 1:1000), rabbit anti-COX IV (3E11, Cell Signaling #4850, 1:1000), rabbit anti-cyclin B1 (D5C10, Cell Signaling #12231, 1:1000 WB, 1:200 FC), rabbit anti-FLIP (D5J1E, Cell Signaling #56343, 1:1000), mouse anti-GAPDH (D4C6R, Cell Signaling #97166, 1:2000), rabbit anti-Mcl-1 (D2W9E, Cell Signaling #94296, 1:1000 WB, 1:50 FC, 1:400 IF), mouse anti-Noxa (F-3, Santa Cruz sc-515840, 1:500), mouse anti-PARP (4C10-5, BD Biosciences 556494, 1:1000), rabbit anti-phospho-histone H2A.X (Ser139) (20E3, Cell Signaling #9718, 1:1000 WB, 1:400 IF), rabbit anti-phospho-histone H3 (Ser10) Pacific Blue™ Conjugate (D2C8, Cell Signaling #8552, 1:50 FC), mouse anti-α-tubulin (DM1A, Cell Signaling #3873, 1:2000), rabbit anti-cleaved caspase-3 AF647 Conjugate (Asp175) (D3E9, Cell Signaling #9602, 1:50 FC), mouse anti-TRAIL-R1 (R&D Systems MAB347, 5 µg/ml FC), mouse anti-TRAIL-R2 (R&D Systems MAB6311, 5 µg/ml FC), mouse IgG1 κ isotype control (BD Biosciences 554121, 5 µg/ml FC), mouse IgG2a κ isotype control (BD Biosciences 554126, 5 µg/ml FC), mouse IgG2b κ isotype control (BD Biosciences 555740, 5 µg/ml FC), mouse anti-FLAG-PE conjugate (BioLegend #637310, 1 µg/ml FC), rabbit (DA1E) mAb IgG XP® isotype control (Cell Signaling #3900, 0.5 µg/ml FC), rabbit IgG isotype control Alexa Fluor 647 conjugate (Cell Signaling #3452, 0.5 µg/ml FC). Secondary antibodies were goat anti-rabbit IgG (H+L) secondary antibody, Alexa Fluor 647 (Thermo Fisher Scientific, #A-21244, 1:500 IF), Alexa Fluor 647-conjugated goat anti-mouse IgG (Dianova #1036-31, 1:500 FC), IRDye®680RD goat anti-rabbit/mouse IgG (LI-COR Biosciences, 1:10,000), IRDye®800CW goat-anti-rabbit/mouse IgG (LI-COR Biosciences, 1:10,000) and peroxidase-conjugated goat anti-rabbit/mouse IgG (Dianova #111-035-144/115-035-062, 1:10,000).

Fc-scTRAIL was produced as described previously (Hutt et al., 2017), FasL-Fc was kindly provided by Harald Wajant (University Hospital Würzburg, Germany) and Annexin V–GFP was produced in-house. DMSO was purchased from Carl Roth, ABT-199 was from Active Biochem, A-1210477 and Q-VD-Oph from Selleckchem, S63845 and WEHI-539 from APExBIO and bortezomib (PS-341) from UBPBio. TMRM, MitoTracker™ Red CMXRos and DAPI were obtained from Invitrogen, propidium iodide (PI) and cycloheximide (CHX) were purchased from Sigma. Puromycin and G-418 were obtained from AppliChem GmbH.

Cells were treated as indicated and viability was determined by Crystal Violet staining after 24 h of treatment. Absorbance was measured in a micro plate reader at λ=550 nm and normalized to the value of control cells. Alternatively, cells were harvested, stained with Annexin V-GFP and PI and viability was analysed by flow cytometry (MACSQuant, Miltenyi Biotec, Germany). To analyse the long-term survival, cells were treated as indicated for 24 h. Cells were washed to remove dead cells and surviving cells were cultivated in fresh medium for 7 days. Viability was then determined by Crystal Violet staining.

Plasmid pEFpuro-hGem(1/110) was obtained by subcloning cDNA encoding mAG-hGem(1/110) (Sakaue-Sawano et al., 2008) into pEF-pGKpuro (kindly provided by Andreas Strasser, WEHI, Australia). The plasmid pSCAT8 encoding the caspase-8 FRET probe CFP-IETD-Venus was described previously (Hellwig et al., 2008), and a SCAT8 variant in a pIRESpuro3 backbone was obtained by subcloning the CFP-IETD-Venus sequence. Cells were transfected using Lipofectamine®LTX (Thermo Fisher Scientific) according to the manufacturer's protocol. At 24 h post transfection, cells were cultured in medium supplemented with appropriate selection antibiotics and fluorescent clones were isolated or enriched after 2 weeks.

The pQCXIPN/ecoR plasmid encoding the receptor for ecotropic murine retroviruses was a kind gift of Monilola Olayioye, University Stuttgart, Germany (Herr et al., 2018). Cells were transfected using Lipofectamine^®LTX. At 24 h post transfection, cells were incubated with supernatant from Plat-E cells transfected with pSuper RV or pSUPER RV Bid (#2) (kindly provided by Stephen Tait, University of Glasgow, UK) (Tait et al., 2007). Cells were then selected in medium with puromycin and analyzed for Bid expression.

Silencer®Select Negative Control siRNA #2 (4390846, Thermo Fisher Scientific), Silencer®Select siRNA targeting Mcl-1 (s8583) or Silencer®Select siRNAs targeting CAD (s4060, s223407) were transfected using Lipofectamine®RNAiMAX (Thermo Fisher Scientific) according to the manufacturer's protocol.

Cells were lysed in solubilization buffer [50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% (v/v) TritonX-100 plus Complete Protease Inhibitors (Roche) and 1 mM DTT], incubated on ice for 10 min and centrifuged at 16,100 g and 4°C. Protein concentrations were determined by Bradford assay (Bio-Rad). Proteins were separated by SDS-PAGE (4-12% NuPAGE® Novex Bis-Tris gels, Thermo Fisher Scientific) and transferred to nitrocellulose membranes (iBlot® Gel Transfer Stacks, Thermo Fisher Scientific). Membranes were then blocked with 1% blocking reagent (Roche) or 5% BSA in TBS containing 0.05% Tween-20 (TBST) and incubated with primary antibodies followed by POD-conjugated secondary antibodies, and detection was performed using Pierce ECL Western Blotting Substrate (Thermo Fisher Scientific) using an Amersham Imager 600 (GE Healthcare). Alternatively, membranes were incubated with IRDye® secondary antibodies and detection was performed using an Odyssey Imaging system (Li-COR Biosciences).

Total RNA was extracted from cells using the NucleoSpin RNA kit (Machery & Nagel). Synthesis of cDNA and amplification of BCL2L1 (Bcl-xL), MCL1 (Mcl-1), CCNB1 (Cyclin B1) or RPLP0 (60S acidic ribosomal protein P0) was performed using enzyme mix (Power SYBR™ Green RNA-to-CT™ 1-Step kit) and the following primers: Bcl-xL-for, 5′-AGGCGGATTTGAATCTCTTTC-3′; Bcl-xL-rev, 5′-CCCGGTTGCTCTGAGACATT-3′; Mcl-1-for, 5′-AACTGGGGCAGGATTGTGAC-3′; Mcl-1-rev, 5′-CAAACCCATCCCAGCCTCTT-3′; CCNB1-for, 5′-ACCTGTGTCAGGCTTTCTCTG-3′; CCNB1-rev, 5′-CTGACTGCTTGCTCTTCCTCA-3′; RPLP0-for, 5′-CTCTGCATTCTCGCTTCCTGGAG-3′; RPLP0-rev, 5′-CAGATGGATCAGCCAAGAAGG-3′. Relative expression was calculated using the ΔΔCt method and was plotted as fold of G1 mRNA amounts.

Cells were harvested, fixed, permeabilised and incubated with Alexa Fluor 647-labelled caspase-3 antibody or isotype control antibody for 60 min, followed by DNA staining (1 µg/ml DAPI) for 10 min. Cells were then washed and analysed by flow cytometry (MACSQuant). Cleaved caspase-3 subpopulations in distinct cell cycle phases were quantified by multi-experiment mixture model analysis based on three lognormal distributions, using the MATLAB Toolbox MEMO (Geissen et al., 2016). Mixture model parameters, comprising subpopulation mean and variance, and subpopulation weights for all time points and cell cycle stages were derived by minimizing the negative logarithm of the likelihood to observe the experimental data.

For live-cell imaging, cells were plated on 35 mm glass-bottom dishes (CellView Cell Culture Dish, Greiner Bio One) in Phenol Red-free RPMI 1640 containing 10% FBS. Images were acquired for 20 h in 15 min intervals at 37°C and 5% CO₂ on a Zeiss Cell Observer microscope equipped with an Axiocam MRm CCD camera, a Plan-Apochromat 20×/0.8 objective. Fluorescence of mAG-hGeminin(1/110) was acquired with a 470 nm LED module combined with a 62 HE filter set (Carl Zeiss). Then, medium containing Fc-scTRAIL alone or in combination with DMSO or the indicated inhibitors was added and cells were imaged for another 10 h. Randomly chosen cells were tracked and the time of division, onset of geminin expression, subsequent division and t_death of cells were recorded. Cells were defined as apoptotic upon onset of membrane blebbing.

FRET-based imaging was performed on a Zeiss Cell Observer SD Spinning Disk microscope, equipped with an Axiocam 503 mono CCD camera. For ratiometric FRET imaging, CFP and FRET channels were acquired with a 445 nm diode excitation laser combined with 485/30 nm (CFP) and 562/45 nm (FRET) emission filters. Venus was excited with a 514 nm laser combined with a 562/45 nm emission filter. Images were acquired in 2×2 binning mode at 37°C and 5% CO₂ every 15 min for 20 h. Cells were then treated with Fc-scTRAIL and imaged for another 8–10 h. Subsequently, Venus was bleached in the final frame with a 515 nm bleaching laser using a UGA42-Firefly illumination system (Rapp Optoelectronics, Germany). The CFP:YFP emission ratio upon background subtraction over time, the FRET efficiency of the final frame and the FRET probe cleavage were calculated using Zen blue 2.3 software (Zeiss). To record mitochondrial membrane potentials, cells were stained with 30 nM TMRM for 30 min at 37°C and imaged (561 nm excitation laser combined with a 600/50 nm emission filter) in 3 min intervals for 6 h post treatment. Image analysis was performed with Zen Blue 2.3 software (Zeiss). Cell divisions and the time from Fc-scTRAIL addition to apoptotic cell death were analysed in parallel. For pH2A.X staining, cells grown on coverslips were treated for 24 h, dead cells were removed by washing, and surviving cells were fixed, permeabilised and immunostained for pH2A.X followed by DAPI-based nuclear staining. Images were acquired on a Zeiss Axio Observer SD Spinning Disk microscope using 405, 488, and 638 nm excitation with a PlanApochromat 63×/1.4 NA oil objective and an Axiocam 503 Mono CCD camera. Images were processed with the ZEN software (Zeiss).

To analyse Mcl-1 or Bcl-xL subcellular localization, cells grown on coverslips were incubated with MitoTracker Red CMXRos (100 nM) for 30 min before fixation, permeabilisation and immunostaining for Mcl-1 or Bcl-xL. Nuclei were counterstained using DAPI. Images were acquired on a Zeiss Axio Observer SD Spinning Disk microscope equipped with a PlanApochromat 40×/1.4 NA oil objective and an Axiocam 503 Mono CCD camera. Geminin was excited with a 488 nm diode laser using a 525/50 nm emission filter, the Alexa Fluor 647 dye was excited with a 638 nm diode laser using the 690/50 nm emission filter and DAPI was excited with a 405 nm diode laser using the 450/50 nm emission filter. MitoTracker Red was excited with a 561 nm diode laser using a 575/50 nm emission filter. Maximum intensity projections of the acquired z-stacks were calculated and used for analysis with Cell Profiler 3.1.9 (McQuin et al., 2018). In brief, masks for nuclei, mitochondria and cytoplasm (cell mask excluding mitochondria) were segmented, mean fluorescence intensities were measured and the ratio for non-mitochondria to mitochondria was plotted for geminin-negative (G₁)/DAPI-low or geminin-positive (late S/G₂)/DAPI-high cells.

The mathematical model serving as a reference for independent progression of apoptosis and cell cycle signalling was implemented in MATLAB (The MathWorks, UK). Detailed information on model development and parameterization are reported elsewhere (Imig et al., 2020).

For intracellular protein staining, cells were fixed with 4% paraformaldehyde, washed with PBS supplemented with 3% (w/v) FBS, and permeabilised with FACS^TM permeabilizing solution 2 (BD Biosciences, Germany). Cells were incubated with primary antibodies diluted in PBS with 3% FBS for 60 min on ice followed by two washing steps. Cells were then incubated with Alexa Fluor 647-conjugated secondary antibody diluted in PBS with 3% FBS. After washing, cells were incubated with phospho-Histone H3 (Ser10) Pacific Blue™ for 30 min and subsequently analysed by flow cytometry.

For DNA labelling using EdU incorporation, cells were incubated in medium containing 10 µM EdU for 60 min and subsequently fixed, permeabilized and stained according the recommended protocol (Click-iT™ EdU AF488 Flow Cytometry assay kit, Thermo Fisher Scientific). Flow cytometry data was processed using custom scripts in MATLAB (The MathWorks, UK).

To analyse the effect of bortezomib (BTZ) or cycloheximide (CHX) on Mcl-1 levels, cells were treated with DMSO, BTZ (25 µg/ml) or CHX (5 µg/ml) for 1.5 h before fixation. For staining of surface TRAIL-R1 and TRAIL-R2, cells were resuspended in PBA (PBS with 0.25% BSA and 0.02% NaN₃) containing primary antibody or corresponding isotype control and incubated for 45 min on ice. After washing, cells were incubated with Alexa Fluor 647-conjugated secondary antibody in PBA for 45 min on ice. Cells were fixed, permeabilised and incubated with phospho-histone H3 (Ser10) Pacific Blue™ for 30 min, washed again and analysed by flow cytometry (MACSQuant). To analyse the binding of Fc-scTRAIL, cells were incubated with 0.06 nM Fc-scTRAIL for 15 min, washed and incubated with anti-FLAG–PE conjugate on ice for 30 min. Cells were washed and analysed by flow cytometry.

To analyse protein amounts in G₁ (geminin-negative) or G₂/M (representing the highest 20% of geminin-positive cells) phases, cells were harvested, pelleted and washed with PBS supplemented with 10% FBS. Cells (2×10⁶/ml) were then separated according their mAG-geminin intensity (excitation 494 nm/emission 519 nm) using a FACSAria™III cell sorter (BD Biosciences) at 4°C. Sorted subpopulations were centrifuged and analysed by Western blotting or used for RNA isolation and quantitative RT-PCR analysis.

Statistical analysis was performed using PRISM 7.05 (GraphPad Software).

The authors acknowledge Drs. D. Kulms (University of Dresden), M. Olayioye (University of Stuttgart), S. Tait (University of Glasgow), H. Wajant (University of Würzburg) and R. J. Youle (NIH Bethesda) for providing crucial materials. The authors thank Beate Budai and Nathalie Peters for expert technical assistance.

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.258966.