PP2ACdc55 (the form of protein phosphatase 2A containing Cdc55) regulates cell cycle progression by reversing cyclin-dependent kinase (CDK)- and polo-like kinase (Cdc5)-dependent phosphorylation events. In S. cerevisiae, Cdk1 phosphorylates securin (Pds1), which facilitates Pds1 binding and inhibits separase (Esp1). During anaphase, Esp1 cleaves the cohesin subunit Scc1 and promotes spindle elongation. Here, we show that PP2ACdc55 directly dephosphorylates Pds1 both in vivo and in vitro. Pds1 hyperphosphorylation in a cdc55 deletion mutant enhanced the Pds1–Esp1 interaction, which played a positive role in Pds1 nuclear accumulation and in spindle elongation. We also show that nuclear PP2ACdc55 plays a role during replication stress to inhibit spindle elongation. This pathway acted independently of the known Mec1, Swe1 or spindle assembly checkpoint (SAC) checkpoint pathways. We propose a model where Pds1 dephosphorylation by PP2ACdc55 disrupts the Pds1–Esp1 protein interaction and inhibits Pds1 nuclear accumulation, which prevents spindle elongation, a process that is elevated during replication stress.
Cell cycle progression is regulated by phosphorylation and dephosphorylation of cyclin-dependent kinase (CDK) substrates. In S. cerevisiae, there are two phosphatases that counteract the sole CDK (Cdk1), PP2A and Cdc14 (Godfrey et al., 2017). PP2A is a heterotrimer consisting of an ‘A’ scaffolding unit (Tpd3), a ‘B’ regulatory unit (either Cdc55 or Rts1), and a ‘C’ catalytic unit (Pph21 and Pph22) (Sneddon et al., 1990). A third catalytic protein, Pph3, has also been proposed to associate with PP2A (Ronne et al., 1991). PP2ACdc55 and PP2ARts1 (i.e. the PP2A forms with Cdc55 or Rts1, respectively) have distinct functions and targets (Zhao et al., 1997). PP2ACdc55 dephosphorylates Cdk1 substrates during mitosis to ensure the correct order of cell cycle events (Godfrey et al., 2017). PP2ACdc55 is present in both the nucleus and the cytoplasm, and its function is dependent on its localization (Gentry and Hallberg, 2002; Rossio and Yoshida, 2011). Zds1 and Zds2, two yeast-specific proteins, physically interact with Cdc55 to trap it in the cytoplasm (Rossio and Yoshida, 2011; Yasutis et al., 2010). In the cytoplasm, PP2ACdc55 downregulates the mitotic Cdk1 inhibitor Swe1, resulting in mitotic entry (Lianga et al., 2013; Sorger and Murray, 1992; Yang et al., 2000). In the absence of Zds1 and Zds2, Cdc55 accumulates in the nucleus (Yasutis et al., 2010; Rossio and Yoshida, 2011), which inhibits mitotic progression by inhibiting the ubiquitin ligase anaphase-promoting complex (APC) and preventing Cdc14 release (Tang and Wang, 2006; Wang and Ng, 2006). During spindle damage, nuclear PP2ACdc55 dephosphorylates the APC components Cdc16 and Cdc27 to deactivate the APC and prevent anaphase onset (Rossio et al., 2013; Lianga et al., 2013). Nuclear PP2ACdc55 also prevents the release of Cdc14, a Cdk1-counteracting phosphatase necessary for mitotic exit (Wang and Ng, 2006). Thus, the two phosphatases Cdc14 and PP2ACdc55 are important for regulating cell cycle progression through mitosis.
Anaphase onset is a tightly regulated cell cycle event that results in sister chromatid segregation. Sister chromatid segregation is triggered when active separase (Esp1), cleaves the cohesin subunit Scc1 and releases poleward forces from the mitotic spindle in anaphase (Uhlmann et al., 1999; Ciosk et al., 1998). Securin (Pds1) is an anaphase inhibitor that binds to and inhibits Esp1 protease activity (Ciosk et al., 1998; Agarwal and Cohen-Fix, 2002). The Pds1–Esp1 interaction requires Pds1 phosphorylation by Cdk1 at three C-terminal sites (S277, S292, T304) (Agarwal and Cohen-Fix, 2002). The pds1-38 mutant, which contains S/T→A mutations at the C-terminal Cdk1 sites, does not bind to Esp1 (Cohen-Fix et al., 1996). In addition to Esp1 inhibition, the Pds1–Esp1 physical interaction promotes Esp1 nuclear localization (Agarwal and Cohen-Fix, 2002). Esp1 is localized to the spindle pole bodies (SPBs) and the spindle midzone during anaphase, and this localization requires Pds1 (Jensen et al., 2001). In pds1Δ cells, Esp1 does not associate with the spindle (Jensen et al., 2001). A temperature-sensitive esp1-C113 mutant, which contains a mutation in the C-terminus in the catalytic domain, fails to mediate elongation of spindles in mitosis (Baskerville et al., 2008). During the unperturbed cell cycle, Pds1 is degraded in an APC-dependent manner at the metaphase-to-anaphase transition (Cohen-Fix et al., 1996; Visintin et al., 1997). During anaphase, fully active Esp1 is released, resulting in Scc1 cleavage and spindle elongation (Uhlmann et al., 1999; Jensen et al., 2001). While the role of Esp1 in Scc1 cleavage has been extensively studied, there is less known about its function in spindle elongation.
Since Pds1 prevents sister chromatid segregation, it is a target of several checkpoint pathways that ensure accurate replication and partitioning of genetic information (Cohen-Fix et al., 1996; Tinker-Kulberg and Morgan, 1999). DNA replication stress compromises genome integrity due to replication fork stalling and collapse, which leads to chromosome breakage and DNA damage (Weinert et al., 1994). The intra-S checkpoint is a well-studied pathway that is activated upon replication stress which mainly targets and stabilizes Pds1 to arrest the cell cycle (Weinert et al., 1994; Yamamoto et al., 1996). The intra-S checkpoint is activated by the sensor kinase Mec1, which interacts with Ddc2 to form foci at replication stress sites (Paciotti et al., 2000; Weinert et al., 1994). Mec1 phosphorylates Rad9, which in turn recruits Rad53 kinase to disrupt the interaction between Pds1 and APCCdc20 (APC with its activator Cdc20) leading to Pds1 stabilization (Sanchez et al., 1999; Wang et al., 2001; Agarwal et al., 2003; Sanchez et al., 1996; Sun et al., 1996; Sweeney et al., 2005). Rad53 also inhibits mitotic Cdk1 activity to arrest the cell cycle (Palou et al., 2015). In addition to replication stress, Mec1 is also activated by DNA damage induced by methyl methanesulfonate (MMS), and activates a second effector kinase, Chk1, which phosphorylates and stabilizes Pds1 (Sanchez et al., 1999; Blankley and Lydall, 2004; Wang et al., 2001). Chk1 phosphorylation sites are distinct from the Cdk1 consensus sites, and Chk1 is only active during MMS-induced DNA damage but not during replication stress (Sanchez et al., 1999). The Cdk1 inhibitor Swe1 is also activated during replication stress independently from the intra-S checkpoint (Palou et al., 2015). Swe1 is a kinase that phosphorylates Cdk1 at Y19 to inhibit it (Sorger and Murray, 1992). Under normal conditions, Cdk1-Y19 phosphorylation is removed by the phosphatase Mih1 at mitotic entry (Russell et al., 1989). During replication stress, Cdk1 inhibition by Swe1 is redundant with Rad53 (Palou et al., 2015), that is, in the absence of the intra-S checkpoint, Swe1 prevents sister chromatid segregation during replication stress (Palou et al., 2015). In addition to the intra-S checkpoint and Swe1, the spindle assembly checkpoint (SAC) also prevents sister chromatid segregation during replication stress (Palou et al., 2016; Palou et al., 2015). The central SAC protein Mad2 physically disrupts binding between the APC and Cdc20 leading to Pds1 stabilization and metaphase arrest (Hwang et al., 1998; Cohen-Fix et al., 1996; Lim et al., 1998).
The function of PP2ACdc55 in reversing Cdk1 phosphorylation led us to hypothesize that PP2ACdc55 targets Pds1 and impacts the Pds1–Esp1 interaction. We focused on the nuclear PP2ACdc55 function using a localization mutant, cdc55-101, which excludes Cdc55 from the nucleus (Sasaki et al., 2000). In this study, we show that PP2ACdc55 dephosphorylates Pds1 both in vitro and in vivo. The Pds1–Esp1 protein interaction was enhanced in the cdc55-101 mutant, with Pds1 accumulation in the nucleus leading to accelerated spindle elongation. A Pds1 phosphorylation mutant showed aberrant spindle morphology with low abundance of Pds1 in the nucleus. Nuclear PP2ACdc55 also played a role under replication stress, which was independent from known replication stress response pathways.
Excluding PP2ACdc55 from the nucleus results in Pds1 hyperphosphorylation
Nuclear PP2ACdc55 stabilizes Pds1 during SAC activation by dephosphorylating APC subunits, which prevents Pds1 ubiquitylation (Lianga et al., 2013; Rossio et al., 2013). Since Pds1 is a Cdk1 substrate, and PP2ACdc55 counteracts Cdk1 phosphorylation events, it is possible that PP2ACdc55 regulates Pds1 phosphorylation status (Agarwal and Cohen-Fix, 2002; Godfrey et al., 2017). To examine how nuclear PP2ACdc55 affects Pds1 stability and phosphorylation, we monitored Pds1 protein in cdc55-101 cells, which exhibit Cdc55 nuclear exclusion during cell cycle (Rossio et al., 2013). We performed a time course in the presence of 100 mM hydroxyurea (HU) in order to slow cell cycle progression and observe transient Pds1 phosphorylation (Fig. 1A). Pds1 protein migrated as two distinct bands, with the upper band representing the hyperphosphorylated form, as previously reported (Agarwal and Cohen-Fix, 2002). cdc55-101 cells showed a higher ratio of hyper- to hypo-phosphorylated Pds1 protein compared to that seen in wild-type (WT) cells (Fig. 1A, right). Since HU depletes dNTP levels and causes replication stress, we next tested whether Pds1 hyperphosphorylation was due to intra-S checkpoint activity. We repeated the time course experiment using 50 mM HU to limit checkpoint activation (Fig. 1B). At a lower concentration of HU, Pds1 was hyperphosphorylated in cdc55-101 cells, similar to our findings at 100 mM HU (Fig. 1B, right). To confirm that Pds1 hyperphosphorylation was specifically due to a loss of nuclear Cdc55, we deleted the only other PP2A B-regulatory subunit, Rts1, and performed a time course in 50 mM HU (Zhao et al., 1997). In rts1Δ cells, there were no changes in Pds1 phosphorylation patterns between WT and rts1Δ cells (Fig. S1). In the absence of HU, Pds1 was transiently hyperphosphorylated in cdc55-101 cells, consistent with the trend observed at 50 mM and 100 mM HU (Fig. 1C). Pds1 hyperphosphorylation in cdc55-101 cells was most apparent at 45 and 60 min after release from G1 (Fig. 1C).
To confirm that the Pds1 upper band represents the phosphoprotein, Pds1 protein was immunoprecipitated from cdc55Δ cell lysate and treated with calf intestinal phosphatase (CIP) in the presence or absence of a phosphatase inhibitor (Fig. 1D). cdc55Δ cells were used instead of cdc55-101 for this immunoprecipitation to avoid unintended Pds1 dephosphorylation by cytoplasmic PP2ACdc55 in the cell lysate. The top Pds1 band (marked with **) was absent in CIP-treated cells, but present when the phosphatase inhibitor was added (Fig. 1D). We concluded that the top band represents hyperphosphorylated Pds1. Taken together, these results demonstrate a role for nuclear PP2ACdc55 in regulating Pds1 phosphorylation.
Sister chromatids remained cohesed in cdc55-101 cells
Phosphorylated Pds1 binds Esp1 and inhibits its protease activity (Ciosk et al., 1998; Agarwal and Cohen-Fix, 2002). Esp1 inhibition protects the cohesin complex and prevents sister chromatid segregation (Ciosk et al., 1998; Uhlmann et al., 1999). Since Pds1 was hyperphosphorylated in cdc55-101 cells, we considered whether nuclear PP2ACdc55 has a role in maintaining sister chromatid cohesion. We monitored cell cycle progression and chromatid segregation in HU-treated WT and cdc55-101 cells. Cells were synchronized in G1 by treatment with α-factor and released into medium containing 100 mM HU (Fig. 2A,B). Cell cycle progression was similar between WT and cdc55-101 cells (Fig. 2A). We examined sister chromatid segregation status in propidium iodide-stained cells (Fig. 2B). While neither WT nor cdc55-101 cells showed significant chromosome segregation during the course of HU treatment, a ‘stretched’ chromatid pattern was present in cdc55-101 cells (Fig. 2B, red bar). The stretched staining pattern was found in large-budded cells where connected staining extended to both cell bodies (Fig. 2B, right). At 180 min, 39% of cdc55-101 cells exhibited the stretched phenotype, compared to 18% in WT (Fig. 2B). Next, we tested whether the stretched chromosome pattern was due to partial chromosome segregation. A TetO-TetR system was used to visualize the region 2 kb away from centromere IV (Cen IV) (Li et al., 2002; Alexandru et al., 2001; Tang and Wang, 2006). GFP dots on Cen IV were counted in HU-treated WT and cdc55-101 cells at the indicated times (Fig. 2C). Neither WT nor cdc55-101 cells showed significant centromere separation, suggesting that chromosomes are not prematurely segregated in the cdc55-101 mutant (Fig. 2C).
PP2ACdc55 directly dephosphorylates Pds1 in vitro
Next, we tested whether PP2ACdc55 dephosphorylates Pds1 in vitro. First, Pds1–3HA was immunoprecipitated from a WT strain and incubated with purified Clb2–Cdk1 complex in vitro, resulting in Pds1 phosphorylation (Fig. 3A, lane 3). This phosphorylation is reduced in the Pds1-5A variant protein, which contains S/T→A mutations at all five Cdk1 consensus sites (T27A, S71A, S277A, S292A and T304A) (Fig. 3A, lane 2) (Holt et al., 2008). The phosphorylated WT Pds1–3HA from lane 3 was then incubated with purified PP2ACdc55, and Pds1 was shown to be fully dephosphorylated between 30–60 min (Fig. 3B). Without PP2ACdc55, the phosphorylated Pds1 remained hyperphosphorylated after 60 min incubation. This result confirms that PP2ACdc55 directly dephosphorylates Pds1 in vitro.
PP2ACdc55 nuclear exclusion is associated with accelerated spindle elongation
Esp1 plays a role in spindle elongation (Uhlmann et al., 1999; Jensen et al., 2001). We next analyzed spindle formation in cdc55-101 mutant (Agarwal and Cohen-Fix, 2002). TUB1-GFP was used to visualize tubulin in WT and cdc55-101 cells in the presence of 50 mM HU. Tub1–GFP signals that were longer than a dot were designated as spindles. Spindles that were >2 µm were categorized as long spindles. cdc55-101 cells showed long spindle formation at 90 min, while WT cells did not show long spindles until 120 min (Fig. S2). The time points where spindle elongation observed correlated with when Pds1 was most hyperphosphorylated in cdc55-101 cells (Fig. 1B; Fig. S2).
To test whether Cdc55 nuclear exclusion affected spindle elongation in unperturbed conditions, spindles were visualized in WT and cdc55-101 cells in a time course without HU. After 60 min from G1 release, 31±7.3% of cdc55-101 cells exhibited long spindles, compared to 13.3±1.9% of WT cells (mean±s.e.m.; Fig. 4A). Spindles that were <2 µm were categorized as short spindles. The number of cdc55-101 cells with a short spindle decreased at 60 min, due to elongation into longer spindles (Fig. 4A). These results demonstrate a correlation between Pds1 hyperphosphorylation and onset of long spindle formation in cdc55-101 cells (Fig. 1C and 4A, t=45 and t=60). WT and cdc55-101 cells had a similar cell cycle profile at 45 and 60 min, confirming that long spindle formation is not due to altered cell cycle stage (Fig. 4C). At 60 min, 30% of WT cells segregated sister chromatids, compared to 2% of cdc55-101 cells (Fig. 4D), which was opposite of the result that we predicted. Therefore, the long spindles that appeared early in cdc55-101 cells were not due to premature sister chromatid segregation.
A Pds1 C-terminal phosphorylation mutant shows fragile mitotic spindles
We next sought to determine which Pds1 phosphorylation sites affect spindle dynamics. Pds1 C-terminal phosphorylation facilitates a physical interaction with Esp1, which promotes spindle elongation in anaphase (Agarwal and Cohen-Fix, 2002). Therefore, we hypothesized that Pds1 C-terminal phosphorylation affects spindle elongation rate. We then used pds1-38 cells, which contain three S/T→A mutations at the C-terminal Cdk1 phosphorylation sites (Agarwal and Cohen-Fix, 2002). We first examined Pds1 phosphorylation status in WT and pds1-38 cells in a time course, and found that Pds1 phosphorylation was fully abolished in the pds1-38 mutant (Fig. 4B, top). To test whether Pds1 hypophosphorylation affects spindle elongation, we performed a time course using TUB1-GFP. Both WT and pds1-38 cells had a peak in short spindle accumulation at 60 min, but the number of cells with short spindles in pds1-38 cells was reduced compared to WT (Fig. 4B). At 60 min, 23±0.3% of pds1-38 cells had short spindles, compared to 41±2.8% in WT (mean±s.e.m.; Fig. 4B, left). At 75 min, both strains formed long spindles; 13±1.5% of pds1-38 versus 26.6±1.3% in WT (Fig. 4B, right). Therefore, the decay in short spindles was not coupled with accumulation of long spindles in pds1-38 cells, suggesting that the spindle is unstable. Long spindles in pds1-38 cells were structurally aberrant and asymmetrical compared to spindles in WT cells (Fig. 4B, red arrow). Frequently, after completing elongation, spindles in pds1-38 cells separated unequally between the mother and daughter cells (Fig. 4B, white arrows).
Similar to cdc55-101 cells, pds1-38 cells did not exhibit altered cell cycle progression compared to WT (Fig. 4C). 2% of pds1-38 cells had fully segregated sister chromatids compared to 30% of WT cells at 60 min (Fig. 4D). WT and pds1-38 cells both showed 58% of cells with fully segregated chromatids at 75 min (Fig. 4D). These findings indicate that the decrease in short spindles in pds1-38 cells at 60 min was not due to altered cell cycle stage, and did not coincide with sister chromatid segregation. In asynchronous conditions, the spindle length distribution between WT, cdc55-101 and pds1-38 cells are the same, which shows that the differences in spindle elongation are only apparent during synchronized cell cycle progression (Fig. S3). From these experiments, we conclude that Pds1 C-terminal phosphorylation status regulates spindle elongation and morphology.
Altered spindle elongation rates in cdc55-101 and pds1-38 are observed on a single-cell level
In order to examine spindle formation dynamics in live cells at the single-cell level, we observed TUB1-GFP in WT and cdc55-101 cells using time-lapse microscopy. We identified the time point where bud emergence occurred (red arrow), where a Tub1–GFP structure that had a greater length than width was first formed (short spindle), and where the spindle was first measured to be greater than 2 µm (long spindle) (Fig. 5A,B). WT and cdc55-101 cells spent an average of 39.17±0.98 and 40.44±1.25 min (mean±s.e.m.), respectively, between the time of bud emergence and short spindle formation (Fig. 5D). A drastic difference was observed in time from short spindle to long spindle formation, between WT (38.13±1.51 min) and cdc55-101 (34.32±2.3 min) cells (Fig. 5E). This result further indicates that spindle elongation is accelerated in cdc55-101 cells. Next, we monitored spindles in pds1-38 cells using time-lapse microscopy and showed that similar to WT cells, pds1-38 cells took an average of 42.3±1.85 min to progress from bud emergence to short spindle formation, and 40.0±2.3 min from short to long spindle formation (Fig. 5D,E). Consistent with results in Fig. 4B, pds1-38 cells displayed long spindles that were irregular in shape (Fig. 5C, blue arrow). In pds1-38 cells, we also found that short spindles shuttled between the mother and daughter cell bodies before elongating (Fig. 5C, yellow arrows). This shuttling movement was seen in 41% of pds1-38 cells compared to 23% in WT cells. Cells with spindles that did not fully elongate during the time-lapse experiment were not included in the analysis shown in Fig. 5D,E.
Nuclear PP2ACdc55 targets Pds1 C-terminal phosphorylation
To confirm whether Pds1 C-terminal phosphorylation sites are targeted by PP2ACdc55, we used a pds1-38 cdc55-101 double mutant strain to examine Pds1 phosphorylation status and spindle morphology. We found that spindle elongation rates in pds1-38 cdc55-101 double mutant cells were similar to those in the pds1-38 single mutant (Fig. 6A). The cdc55-101 pds1-38 double mutant cells also exhibited the irregular spindle structure that was present in the pds1-38 single mutant (Fig. 6A, right, white arrow). Pds1 was not phosphorylated in either the pds1-38 single mutant or the pds1-38 cdc55-101 double mutant, suggesting a role for Pds1 in regulating spindle dynamics downstream of PP2ACdc55 (Fig. 6B).
We treated cells with HU to slow down the cell cycle to test whether Pds1 was transiently phosphorylated in pds1-38 cdc55-101 cells. Under HU treatment, there was a faint Pds1 hyperphosphorylation band in the cdc55-101 pds1-38 double mutant that was not seen in pds1-38 (Fig. S4A). The presence of this band suggests that there are additional PP2ACdc55 dephosphorylation sites on Pds1 that are not the C-terminal Cdk1 sites. To test whether PP2ACdc55 regulates additional Pds1 phosphorylation sites, we used pds1-5A cdc55-101 cells, in which both N- and C-terminal Cdk1 phosphorylation sites are mutated. All of the pds1-38 mutation sites are included in pds1-5A, and, in addition to two N-terminal S/T→A mutations at T27 and S71 are present (Holt et al., 2008). The Pds1 phosphoprotein band was abolished in pds1-5A cells, suggesting that Pds1 N-terminal phosphorylation is too transient to detect during normal cell cycle conditions (Fig. S4B). Taken together, and consistent with prior findings, these results support a model where PP2ACdc55 targets Cdk1 consensus sites (Godfrey et al., 2017).
Since phosphorylated Pds1 binds to Esp1, we predicted that the Pds1 and Esp1 interaction is enhanced in the cdc55Δ mutant (Agarwal and Cohen-Fix, 2002). To test whether the physical Pds1–Esp1 interaction is affected by PP2ACdc55, a co-immunoprecipitation (co-IP) experiment was performed using asynchronous WT and cdc55Δ cells (Fig. 6C). Esp1–9MYC was immunoprecipitated and the amount of bound Pds1 protein was quantified to determine the relative ratio of Pds1 to Esp1 (Fig. 6C). cdc55Δ cells were used instead of cdc55-101 mutant to avoid unintentional cytoplasmic PP2ACdc55 activity in the cell lysate. The physical Pds1–Esp1 interaction was enhanced in cdc55Δ compared to WT, implicating a role for PP2ACdc55 in disrupting Pds1–Esp1 binding (Fig. 6C).
Pds1 phosphorylation status affects its localization
A previous study showed that Pds1 phosphorylation promotes Esp1 localization to the nucleus, but it is unknown whether this phosphorylation also affects Pds1 localization (Agarwal and Cohen-Fix, 2002). To test this possibility, we examined PDS1-GFP localization in WT and cdc55-101 cells. Pds1 was localized to the nucleus earlier in cdc55-101 cells compared to WT cells (Fig. 7A). At 45 min after G1 release, 62±4.8% of cdc55-101 cells showed Pds1 nuclear localization, compared to 31.3±6.2% of WT cells (mean±s.e.m.; Fig. 7A). At the 45-min time point, Pds1 was hyperphosphorylated in cdc55-101 cells compared to WT, suggesting that there is a correlation between Pds1 hyperphosphorylation and nuclear localization (Figs 4B and 7A). To confirm that Pds1 phosphorylation status affects the localization of Pds1, we examined localization of pds1-38-GFP during the cell cycle (Fig. 7B) (Agarwal and Cohen-Fix, 2002). At 60 min after G1 release, only 21.7±0.7% of pds1-38 cells showed nuclear Pds1 localization, compared to 41.3±0.6% of WT cells (Fig. 7B). Taken together, these findings demonstrate a positive relationship between Pds1 hyperphosphorylation and its nuclear localization.
Nuclear PP2ACdc55 is involved in the replication stress response
Since Pds1 is a target of replication stress response pathways and cdc55-101 cells showed enhanced Pds1 hyperphosphorylation during HU treatment, we considered whether PP2ACdc55-dependent Pds1 dephosphorylation is important for the response to replication stress. A role for PP2ACdc55 in the cell cycle response to replication stress is supported by evidence showing that pph21Δ, pph22Δ and cdc55Δ cells are sensitive to HU (Tang and Wang, 2006). However, the precise mechanism for PP2ACdc55 activity during replication stress is unknown. Therefore, we sought to determine whether PP2ACdc55 acts in one of the known replication stress response pathways.
There are four possible scenarios for nuclear PP2ACdc55 function during replication stress (Fig. S5A). In scenario 1, PP2ACdc55 is a factor in the intra-S checkpoint and acts downstream of Mec1 to stabilize Pds1. In scenario 2, PP2ACdc55 acts through the Swe1 pathway to inhibit mitotic Cdk1 activity. In scenario 3, PP2ACdc55 acts in the SAC pathway to dephosphorylate the APC components Cdc16 and Cdc27. Dephosphorylation of Cdc16 and Cdc27, in turn, would inhibit APCCdc20 activity and stabilize Pds1. In scenario 4, PP2ACdc55 is activated in replication stress conditions through a novel signaling pathway and acts on Pds1 directly. Then, replication stress separately activates Swe1 and inhibits mitotic Cdk1 activity. PP2ACdc55 could also act in the Swe1 dependent pathway.
We performed serial dilution assays using cdc55 mutants on plates containing HU to test whether PP2ACdc55 is involved in the replication stress response in vivo. We confirmed a growth defect in the cdc55Δ strain on HU plates (Fig. S5B, lane 2) (Liu and Wang, 2006). Since PP2ACdc55 activity is dependent on Cdc55 localization, we tested HU sensitivity of the cdc55-101 mutant, which showed the same degree of HU sensitivity as cdc55Δ cells (Fig. S5B, lanes 2 and 4). The most well-studied replication stress response pathway is the intra-S checkpoint (Fig. S5A, scenario 1). To determine whether PP2ACdc55 acts in the intra-S checkpoint, we tested for synthetic growth defects in cdc55-101 cells with mec1Δ sml1Δ and rad53Δ sml1Δ deletions on HU-containing plates (Fig. S5C). MEC1 and RAD53 are essential genes, but mec1Δ and rad53Δ lethality can be rescued by sml1Δ (Zhao et al., 1998). The mec1Δ sml1Δ double mutant is highly sensitive to a low concentration of HU at 2.5 mM (Fig. S5C, lane 1) (Weinert et al., 1994). mec1Δ sml1Δ showed a more-severe growth defect than cdc55-101 cells, making it unlikely that PP2ACdc55 acts upstream of Mec1 (Fig. S5C, lanes 1 and 4). The triple mutant cdc55-101 mec1Δ sml1Δ showed synthetic lethality compared to the double mutant mec1Δ sml1Δ (Fig. S5C, lanes 3 and 5). The enhanced growth defect of cdc55-101 mec1Δ sml1Δ supports a role for PP2ACdc55 that is independent of the Mec1 pathway. This was further supported by evidence that cdc55-101 rad53Δ sml1Δ mutant also showed synthetic growth defect on HU compared to the rad53Δ sml1Δ mutant (Fig. S5C, lanes 7 and 9). The Δsml1 single mutant did not show HU sensitivity and the cdc55-101 sml1Δ double mutant did not show a genetic interaction confirming that sml1Δ alone does not show HU sensitivity (Fig. S5C, lanes 10 and 14). During DNA damage, Mec1 activates a second effector kinase, Chk1, to phosphorylate and stabilize Pds1 (Sanchez et al., 1999). We tested whether CHK1 deletion shows genetic interaction with cdc55-101. chk1Δ cells did not show sensitivity to HU and there was no genetic interaction between chk1Δ and cdc55-101 (Fig. S5C, lanes 15 and 19). This result is consistent with previous results showing that Chk1 is not involved in the replication stress response (Sanchez et al., 1999). Taken together, these findings show that nuclear PP2ACdc55 acts independently of the Mec1-mediated signaling pathway during replication stress.
Recent studies have proposed that Swe1 and Mad2 are part of the replication stress response (Palou et al., 2015; Palou et al., 2016). Swe1 inhibits M-Cdk1 (mitotic Cdk1) to arrest the cell cycle during replication stress (Fig. S5A, scenario 2) (Palou et al., 2015). To test whether nuclear PP2ACdc55 acts in the Swe1 pathway, we performed a serial dilution using swe1Δ cdc55-101 cells on HU-containing plates. swe1Δ cells did not show HU sensitivity and swe1Δ cdc55-101 showed a similar growth defect to cdc55-101 cells (Fig. S6A, lanes 1, 4 and 5). This result indicates that the growth defect in swe1Δ cdc55-101 cells was solely due to the cdc55-101 mutation. Since Mih1 counteracts M-Cdk1 inhibitory phosphorylation by Swe1, we tested whether there was a genetic interaction between CDC55 and MIH1 (Russell et al., 1989). mih1 deletion showed a synthetic growth defect when combined with cdc55-101 in normal conditions, further suggesting that nuclear PP2ACdc55 is independent from Swe1-dependent M-Cdk1 regulation (Fig. S6B). Next, we tested whether PP2ACdc55 plays a role in the SAC pathway during replication stress (Fig. S5A, scenario 3) (Lianga et al., 2013). The SAC-deficient mad2Δ strain did not show HU sensitivity, indicating that the SAC is dispensable for the replication stress response in vivo (Fig. S6C, lane 1). Moreover, the growth defect in cdc55-101 was identical to the growth defect in cdc55-101 mad2Δ cells, indicating that the cdc55-101 mutation was solely responsible for HU sensitivity in the double mutant (Fig. S6C, lanes 4 and 5). We then examined the phosphorylation status of the APC component Cdc16 during HU treatment using a phos-tag reagent to better separate phosphoisoforms for immunoblotting analysis (Kinoshita et al., 2006). Cdc16 was hyperphosphorylated in cdc55-101 compared to WT when cells are untreated at time 0, as previously reported (Fig. S6D) (Lianga et al., 2013). Cdc16 phosphorylation status was unchanged after HU treatment, suggesting that phosphorylation of this APC component is not a target of the replication stress response (Fig. S6D). Taken together, our data indicate that PP2ACdc55 has a role that is independent of the SAC during replication stress. Finally, we confirmed that HU sensitivity in cdc55-101 is a result of Cdc55 nuclear exclusion by examining growth defects in CDC55-NES and CDC55-NLS cells (Fig. S6E). Cells expressing CDC55-NES, which contains a nuclear export signal, exhibited the same sensitivity to HU as cdc55Δ (Fig. S6E, lanes 1 and 4). CDC55-NLS cells, containing a localization signal, showed a severe growth defect on HU, indicating that both nuclear and cytoplasmic Cdc55 are involved in replication stress response, although they may act in separate pathways (Fig. S6E, lanes 1 and 3).
We attempted to generate a pds1Δ cdc55-101 strain, but double mutants were synthetic lethal, indicating that additional genetic interactions between PDS1 and CDC55 are present (Fig. S7). Synthetic lethality in pds1Δ cdc55-101 is consistent with an earlier finding showing that pds1Δ cdc55Δ double mutants are inviable (Tang and Wang, 2006).
PP2ACdc55 regulates Pds1 phosphorylation status
PP2ACdc55 regulates Pds1 phosphorylation status both in unperturbed and replication stress conditions, specifically at sites that are necessary for Esp1 interaction. A recent study showed that the Pds1 C-terminal segment (residues 258–373) is required for the Esp1 interaction, which is consistent with the idea that phosphorylation at the C-terminal sites (S277, S292 and T304) are required for Esp1 binding (Luo and Tong, 2017). Luo and Tong proposed a model showing that S277 should have favorable interactions with Esp1 due to a possible conformational change induced by S292 and T304 phosphorylation. When Pds1 and Esp1 are bound, the Pds1 C-terminal region is located in the Esp1 active site (Luo and Tong, 2017). While this explains how Pds1 could block Esp1 protease activity, it remains unclear whether the interaction site affects the function of Esp1 in spindle elongation. It was previously proposed that the Pds1–Esp1 interaction causes a conformational change in Esp1 that is a prerequisite for Esp1 proteolytic activity, although continued Pds1 binding would prevent Esp1 from interacting with a cleavage substrate (Hornig et al., 2002). The Esp1 C-terminus, which contains its catalytic domain, is necessary for spindle interaction and, thus, it is possible that an Esp1 conformational change is also a prerequisite for spindle elongation (Baskerville et al., 2008; Jensen et al., 2001). Pds1 dephosphorylation by PP2ACdc55 might prevent an Esp1 conformational change.
Pds1 phosphorylation status regulates spindle elongation and spindle structure
Anaphase spindle elongation requires Esp1 positioning at the SPBs by Pds1 (Jensen et al., 2001). Here, we show that Pds1 phosphorylation status affects Pds1 localization as well as spindle elongation. If hyperphosphorylated Pds1 is in a complex with Esp1, premature Pds1 localization to the nucleus may cause accelerated spindle elongation in cdc55-101. Thus, a likely scenario for spindle regulation is that Pds1 dephosphorylation prevents the Pds1–Esp1 complex from accumulating in the nucleus. Conversely, premature Pds1 translocation from the nucleus may cause the spindle instability observed in pds1-38 cells. In this scenario, the fragile spindles in pds1-38 may be a result of inefficient Pds1–Esp1 localization to the SPBs. A previous genetic screen using temperature-sensitive esp1-1 showed that there is a synthetic growth defect when microtubule minus-end motor Kar3 is overexpressed (Ho et al., 2015). Kar3 localizes to the SPB and generates inward force (Saunders et al., 1997). It is, therefore, possible that the presence of Pds1–Esp1 at the SPB inhibits Kar3 activity to prevent spindle collapse. Therefore, spindles in pds1-38 cells may undergo excessive inwardly directed force due to Kar3 activity.
An alternative explanation for accelerated spindle elongation in cdc55-101 is that Cdc14 release is altered in the absence of nuclear PP2ACdc55. This model would be consistent with previous studies showing that PP2ACdc55 prevents Cdc14 release from the nucleolus during SAC activation (Yellman and Burke, 2006; Wang and Ng, 2006). There are two separate pathways through which nuclear PP2ACdc55 may act in preventing release of Cdc14: inhibiting formation of the FEAR complex and reversing Cdc5 phosphorylation events. Based on our results, Pds1 in complex with Esp1 prematurely accumulated in the nucleus in cdc55-101 cells. The Esp1, Slk19 and Cdc5 proteins comprise the FEAR complex that releases Cdc14 from the nucleolus during anaphase (Stegmeier et al., 2002). Therefore, premature Esp1 nuclear recruitment may result in early FEAR complex assembly, resulting in precocious Cdc14 release. Untimely Cdc14 release cdc55-101 cells may also be due to increased Cdc5-dependent phosphorylation. Cdc5 phosphorylates Net1, a Cdc14 inhibitor, to stimulate Cdc14 release in early anaphase (Stegmeier et al., 2002; Shou et al., 2002). PP2ACdc55 reverses Net1 phosphorylation to maintain Cdc14 sequestration (Queralt et al., 2006). Thus, in cdc55-101 cells, Net1 should be hyperphosphorylated, resulting in unregulated Cdc14 release. After release, Cdc14 would dephosphorylate and activate the plus-end microtubule motor protein Cin8 and bundling protein Ase1 to promote spindle midzone assembly during anaphase (Khmelinskii et al., 2007; Roccuzzo et al., 2015). Therefore, accelerated spindle elongation in cdc55-101 cells may be a result of premature and enhanced Cdc14 release, which in turn results in elevated Cin8 and Ase1 activity.
Additional genetic interactions between CDC55 and PDS1 are revealed
It was recently shown that PP2ACdc55 dephosphorylates Esp1 at Cdk1 consensus sites (Lianga et al., 2018). These authors proposed a model in which Pds1 destruction and phospho-regulation of Esp1 act redundantly to ensure Esp1 activation in mitosis (Lianga et al., 2018). Combining the phospho-mimetic mutations esp1-3D with pds1Δ resulted in synthetic lethality (Lianga et al., 2018). Here, we found that pds1Δ cdc55-101 is synthetic lethal, which could be due to Esp1 hyperphosphorylation in cdc55-101 cells. Although Esp1 nuclear localization would be inefficient in pds1Δ cells, it would not be fully excluded (Agarwal and Cohen-Fix, 2002). Nuclear Esp1 would be fully active in the absence of both Pds1 and nuclear PP2ACdc55. Synthetic lethality in pds1Δ cdc55-101 double mutant is most likely due to uncontrolled nuclear Esp1 activity resulting in both premature Scc1 cleavage and spindle elongation.
Nuclear PP2ACdc55 acts independently of known replication stress response pathways
Previous reports have focused on the role of PP2ACdc55 in nocodazole-treated cells when the spindle is disrupted or when DNA is damaged (Minshull et al., 1996). During spindle disruption, Pds1 is prematurely degraded by APCCdc20 in cdc55Δ cells, resulting in precocious sister chromatid separation (Tang and Wang, 2006). In nocodazole-treated cells, PP2ACdc55 may target and inhibit APCCdc20 rather than Pds1. This pathway relies on the function of Esp1 as a protease to cleave Scc1. In the presence of DNA damage in cdc13-1 cells, the premature chromatid separation was independent from Pds1 degradation, although the molecular mechanism remained unclear (Tang and Wang, 2006). In this study, we used HU to study how PP2ACdc55 is involved in replication stress response. We found that Pds1 dephosphorylation inhibits mitotic spindle elongation but does not affect cohesin cleavage. PP2ACdc55-dependent Pds1 dephosphorylation releases Pds1 from the nucleus which might inhibit mitotic spindle formation during replication stress. We speculate that free Esp1 is not able to cleave Scc1 because it cannot be recruited to the nucleus by unphosphorylated Pds1.
Cdc55 nuclear exclusion resulted in Pds1 hyperphosphorylation and accelerated spindle elongation in both normal and replication stress conditions, suggesting that PP2ACdc55 may have additional functions during replication stress. A Pds1-independent function for PP2ACdc55 could include direct Esp1 inhibition. This scenario would be consistent with a prior report showing that PP2ACdc55 dephosphorylates Esp1 to inactivate it (Lianga et al., 2018).
PP2ACdc55 has distinct functions depending on its localization. Our results showed that both nuclear and cytoplasmic PP2ACdc55 activities are involved in the replication stress response. We showed that nuclear PP2ACdc55 prevents spindle elongation through the Pds1, and not through Mec1 or Mad2, pathways. Cytoplasmic PP2ACdc55 inhibits Swe1 by activating M-Cdk1 activity. It has been shown that Swe1 is stabilized in response to HU, but not to nocodazole (Liu and Wang, 2006). We propose a model whereby replication stress activates both nuclear and cytoplasmic PP2ACdc55, which dephosphorylate distinct targets to contribute to a mitotic block. We attempted to examine Cdc55 localization during replication stress, but there were no obvious changes detected (data not shown).
In summary, our findings show that nuclear PP2ACdc55 has a novel role in regulating spindle dynamics by dephosphorylating Pds1. We propose a model where PP2ACdc55 disrupts the Pds1–Esp1 interaction to inhibit spindle elongation. Nuclear PP2ACdc55 also has a role in the replication stress response that is independent of known checkpoint mechanisms. These findings may have implications in humans, as mammalian PP2A is a tumor suppressor, where mutations in human PP2A have been shown to be associated with solid tumors as well as various types of leukemia (Cristóbal et al., 2011; Lucas et al., 2011; Neviani et al., 2005; Yuan et al., 2009; Seshacharyulu et al., 2013). The PP2A regulatory subunit B55 (also known as PPP2R2; with PPP2R2A–PPP2R2D forms) is the mammalian homolog of Cdc55 and human PP2A-B55δ dephosphorylates and stabilizes human securin, the human homolog of Pds1 (Gil-Bernabé et al., 2006; Li and Virshup, 2002). Our work here suggests that it is of interest to study whether dephosphorylation of human securin by PP2A affects spindle behavior or has a role in the replication stress response.
MATERIALS AND METHODS
Plasmids and strains
Standard methods were used for mating, tetrad dissection, and transformation. HA and MYC tags were generated by a standard PCR method (Longtine et al., 1998). A full list of strains used is presented in Table S1. PDS1-6HA was generated using pYM3 plasmid (forward primer S3, 5′-CAGCGAAGAAGGCCTCGATCCTGAAGAACTAGAGGACTTAGTTACTCGTACGCTGCAGGTCGAC-3′; reverse primer S2, 5′-CTGTATATACGTGTATATATGTTGTGTGTATGTGAATGAGCAGTGGATATCGATGAATTCGAGCTCG-3′) (Janke et al., 2004). ESP1-9MYC was generated using pYM6 plasmid (S3, 5′-GGCGCAGCTCCTGTTATTTATGGGTTACCGATCAAGTTCGTATCACGTACGCTGCAGGTCGAC-3′; S2, 5′-CAATGCCTATATGAAATCTTTTCGAAACAGCCAGTACATGTAACAAATCGATGAATTCGAGCTC-3′) (Janke et al., 2004). CDC16-6HA was generated using pYM3 plasmid (S3, 5′-GCCTCGATCCTGAAGAACTAGAGGACTTAGTTACTCGTACGCTGCAGGTCGACACGCAGATATGGAACTGGAATTCCTCGCCCGCCTTCGTACT-3′; S2, 5′-CTGTATATACGTGTATATATGTTGTGTGTATGTGAATGAGCAGTGGATATCGATGAATTCGAGCTCGTTCCTCGCCCGCCTTCGTACT-3′). CDC55-GFP-NLS and CDC55-GFP-NES strains were constructed by isolating the C-terminal region of CDC55 from strains SY1808 and SY1811 (Satoshi Yoshida, School of International Liberal Studies, Waseda University, Japan) by PCR and transforming them into a W303 background using forward primer, 5′-GGGAACCGAAATGAATGAAATCG-3′; reverse primer, 5′-TCCTTTGATAGGAGTATTTGGGCGG-3′. chk1Δ was obtained from Euroscarf (Euroscarf, Oberursel, Germany), and the PCR product was transformed into the W303 strain. Standard cross methods were used to construct rad53Δ sml1Δ and swe1Δ strains using YGP24 and YGP98, respectively (David Quintana, Department of Biochemistry and Molecular Biology, Universitat Autónoma de Barcelona, Spain). pds1-38-3HA strains were constructed by crossing with RA2815 (Orna Cohen-Fix, NIDDK, National Institute of Health, USA), and pds1-5A-3HA strains were constructed using strain LH505 (Liam Holt, Department of Biochemistry and Molecular Pharmacology, NYU Langone Medical Center, USA). PDS1-GFP was constructed by crossing with BL123 (James Haber, Department of Biology, Brandeis University, USA). Pds1-38-GFP was constructed by isolating the C-terminal region of PDS1–GFP from BL123 (James Haber) strain by PCR and transforming it into the RA2815 strain background (Orna Cohen-Fix) (forward primer: 5′-GCATCACTCGGAATCAAG-3′; reverse primer, 5′-CCGCAGCACATTAGTAGAAAC-3′). The promURA::tetR::GFP::LEU cenIV::tetOx448::URA strain was obtained from a cross between 535-1-1 (Yanchang Wang, Dept of Biological Sciences, Florida State University, USA) and CDC55-MYC or cdc55-101-MYC strains.
Cell culture and media
Yeast extract peptone medium with glucose (YPD) was used for cell culture for western blot, flow cytometry and serial dilution experiments. Synthetic Complete (SC) medium with glucose was used for fluorescence microscopy experiments with TUB1-GFP strains. SC-low fluorescence medium with glucose was used for fluorescence microscopy experiments with CenIV–GFP dots (Sheff and Thorn, 2004). Cell culture was performed at 30°C. Where used, α-factor was added at a final concentration of 50 nM for 2 h at 30°C to arrest the cell cycle in G1, which was confirmed by counting cells showing >90% of 100 cells with shmoo morphology. Cells were washed three times to remove α-factor and resuspended in YPD or SD media to release the arrest.
cdc55Δ swe1Δ PDS1-6HA cells were grown in YPD, arrested in G1 phase by alpha-factor and released into YPD for 45 min. Cells were lysed and Pds1-HA was pulled down using anti-HA beads. IP samples were incubated at 37°C for 15 min in Cutsmart buffer (New England Biolabs) and protease inhibitor (Roche), with or without Calf Intestinal Phosphatase (CIP) (New England Biolabs). Control experiment was performed with 1× Phostop (Roche).
Cells were grown overnight in 3 ml YPD cultures at room temperature. Cells were diluted 10-fold and 5 μl was spotted on YPD plates containing the indicated concentration of HU. Plates were incubated for 3 days at room temperature.
Cell cycle profiles were monitored by flow cytometry using propidium iodide staining as previously described (Epstein and Cross, 1992). Flow cytometry analysis was performed using a BD Accuri C6 flow cytometer (BD Biosciences, San Jose, CA). 20,000 cells were analyzed per sample. Results were analyzed by FlowJo software (FlowJo LLC, Ashland, OR).
For all microscopy with the exception of time lapse microscopy, images were obtained with a Nikon Eclipse 90i fluorescence microscope using a 60×/1.45 NA Plan Apochromatic objective lens (Nikon, Tokyo, Japan) with an Intensilight Ultra High Pressure 130-W mercury lamp (Nikon, Tokyo, Japan). Images were taken with a Clara interline charge-coupled device camera (Andor, Belfast, UK). The images were captured with NIS-Elements software (Nikon, Tokyo, Japan). For TUB1-GFP and PDS1-GFP time course experiments, images were captured using the DIC filter at 80 ms exposure and the FITC filter for 200 ms exposure (Figs 4, 6 and 7; Fig. S2). For CenIV–GFP dots, Z-stacks were generated with seven 0.5 μm steps using the DIC filter at 80 ms exposure and FITC filter at 500 ms exposure (Fig. 2C). Images of flow cytometry samples using fixed cells were taken using the DIC filter at 80 ms exposure time and the TxRed filter at 50 ms exposure (Fig. 2B).
For time lapse microscopy, an Observer Z1 (Zeiss, Jena, Germany) microscope equipped with an automated stage and a plan-apo 63×/1.4 NA oil immersion objective was used (Fig. 5). Asynchronous cells were sonicated at room temperature for 5 s and transferred to 1.5% low-melting agarose pads made with SC medium containing glucose. Live-cell imaging was performed over 5 h with 3 min per frame. All images were prepared using FIJI software (NIH, Bethesda, MD) (Schneider et al., 2012). The cutoff point for long spindles was when spindle length was at its maximum length for each individual cell.
Western blotting, co-immunoprecipitation and Phos-tag analysis
Cells were lysed in TBT buffer containing inhibitors with glass bead agitation as previously described (Ikui et al., 2012). Proteins were separated using SDS-PAGE with Novex 4–20% Tris-glycine polyacrylamide gel (Invitrogen, Life Technologies, Carlsbad, CA). Western blot analysis was performed using peroxide-conjugated anti-hemagglutinin antibody at 1:250 dilution (12013819001, Sigma), anti-cMYC antibody 9E10 (M4439, Sigma-Aldrich, St Louis, MO) at 1:5000 dilution, and anti-Pgk1 (459250, Thermo Fisher) at 1:5000 as a loading control. Images were developed using a Fuji LAS 4000 Imager (GE Healthcare Life Sciences, Pittsburgh, PA). Co-immunoprecipitation was performed by incubating cell lysate with anti-MYC conjugated agarose beads (Sigma-Aldrich, St Louis, MO) for 1 h at 4°C. Phos-tag analysis was performed using phosphoproteins obtained by TCA precipitation and separated with phos-tag acrylamide gels as previously described (Fujifilm Wako Pure Chemical, Osaka, Japan) (Kinoshita et al., 2006; Kinoshita et al., 2009; Link and LaBaer, 2011).
In vitro phosphatase assay
Pds1–3HA was immunoprecipitated and phosphorylated by Clb2–Cdc28. The purification of the Clb2–Cdc28 complex was performed as previously described (Rudner et al., 2005; Lianga et al., 2013). Kinase reactions were performed with 1 μCi γ-[32P]ATP (Lianga et al., 2013). Phosphorylated Pds1–3HA was then treated with TAP-purified PP2ACdc55 complexes (Lianga et al., 2013; Lianga et al., 2018).
We gratefully acknowledge Orna Cohen-Fix for the pds1-38-HA strain, David Quintana for rad53Δ sml1Δ, swe1Δ, and pds1Δ strains, Satoshi Yoshida for CDC55-MYC, cdc55-101-MYC, CDC55-GFP-NES and CDC55-GFP-NLS strains, James Haber for the PDS1-GFP strain, Yanchang Wang for promURA::tetR::GFP::LEU cenIV::tetOx448::URA strain and Liam Holt for pds1-5A-3HA strain.
Conceptualization: S. Khondker, A.I.; Methodology: S. Khondker; Validation: S. Khondker, A.I.; Formal analysis: S. Khondker, A.I.; Investigation: S. Khondker, S. Kajjo, D.C.-B.; Resources: J.S., A.R., A.I.; Data curation: S. Khondker, A.I.; Writing - original draft: S. Khondker, A.I.; Writing - review & editing: J.S., A.R., A.I.; Visualization: S. Khondker; Supervision: J.S., A.R., A.I.; Project administration: A.I.; Funding acquisition: A.I.
This work was supported by the National Institutes of Health (NIH; 5SC1GM21242 to A.I.). Deposited in PMC for release after 12 months.
Peer review history
The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.243766.reviewer-comments.pdf
The authors declare no competing or financial interests.