The abscission checkpoint is normally active in only a small fraction of cultured cells because at any given time, few cells are in the cell cycle phase of cytokinesis and, further, because checkpoint activity is normally transient. We have overcome these experimental hurdles by combining treatments that (1) prevent checkpoint satisfaction (Mackay et al., 2010), using siRNA-mediated depletion of the nuclear pore basket proteins Nup153 and Nup50 (referred to throughout as siNups) and (2) synchronize the cell cycle, using thymidine addition to arrest cells in G1/S phase, followed by release and synchronous progression through the remaining cell cycle (Figure 1B,C). These conditions elicited an active checkpoint in up to ~80% of HeLa cells in culture (Figure 1D,E). We did not observe increased chromatin bridges in these checkpoint-active samples (Figure 1—figure supplement 1A,B), indicating that lagging chromatin did not significantly contribute to sustained checkpoint activity. Importantly, cells enriched with an active abscission checkpoint by this method remained competent for abscission, as they all completed division within 60 min following deactivation of the checkpoint with an AurB inhibitor (AurBi, ZM 447439) (Figure 1D,E) or after >5 hr without treatment (not shown).

Using this new checkpoint enrichment protocol, we tested whether ESCRT factor recruitment to midbodies was altered by the abscission checkpoint. The ESCRT adaptor CEP55 and one of the two early-acting ESCRT factors, TSG101/ESCRT-I, localized normally to midbodies when the checkpoint was active (Figure 2A,B, Figure 2—figure supplement 1A, Figure 2—figure supplement 2A,B). In contrast, checkpoint activity significantly delayed midbody recruitment of the other early-acting ESCRT factor, ALIX. This delay was particularly pronounced in early-stage midbodies (Figure 2C,D) and was still striking when all midbodies were tracked over time (Figure 2—figure supplement 1B), despite a similar decrease in abundance of early-stage midbodies in both siCon and siNups samples (e.g., 15 hr post-thymidine, Figure 2—figure supplement 1C). To control for potential cell-to-cell variation, we also quantified CEP55 and ALIX intensity within the same cells to determine their relative ratios at individual midbodies. This experiment confirmed that an active checkpoint delays recruitment of ALIX to the midbody relative to CEP55, with an approximately threefold decrease in the relative ratio 11 hr post-thymidine release (Figure 2E,F, Figure 2—figure supplement 1D), and an approximately threefold decrease relative to TSG101/ESCRT-I in an analogous experiment (Figure 2—figure supplement 2C,D). By 16 hr after thymidine release, ALIX levels at the midbody had largely, but not completely recovered (Figure 2D–F, Figure 2—figure supplements 1B,D and 2D). This is consistent with the duration of abscission delay under these conditions, which starts to lift after 16 hr (not shown). AurBi treatment rapidly rescued ALIX recruitment at 11 hr (Figure 2G), confirming that delays in ALIX recruitment were checkpoint dependent. We also found that checkpoint activity delayed recruitment of IST1, an ESCRT-III subunit that functions downstream of ALIX and is required for abscission (Figure 2—figure supplement 3A,B; Agromayor et al., 2009; Bajorek et al., 2009). IST1 recruitment was likewise restored by brief AurBi treatment (Figure 2—figure supplement 3C). Taken together, these observations reveal a new dimension of regulation that helps explain why abscission is delayed when this checkpoint is not satisfied. Both ALIX and IST1 localize to midbodies and are required for cytokinesis in HeLa cells (Agromayor et al., 2009; Bajorek et al., 2009; Carlton and Martin-Serrano, 2007; Morita et al., 2007), and abscission timing is therefore expected to be delayed when their recruitment is delayed.

We next investigated what other cellular changes occur concomitantly with delayed ALIX recruitment. We have previously shown that cells expressing CHMP4C with mutations that inhibit ALIX binding bypass the abscission checkpoint (Sadler et al., 2018), indicating that interaction of CHMP4C and ALIX is key to checkpoint activity. We therefore first tested whether CHMP4C localization was also altered. Stably expressed HA-CHMP4C localized to midbodies regardless of checkpoint status, whereas ALIX recruitment was again delayed when the checkpoint was active (Figure 3—figure supplement 1A–D). Thus, consistent with previous reports (Sadler et al., 2018), bulk CHMP4C protein localization does not change with checkpoint activity. We next examined the localization of specific AurB-activated phospho-isoforms of endogenous CHMP4C using phospho-specific antibodies (Capalbo et al., 2016). As reported previously (Capalbo et al., 2016), singly phosphorylated CHMP4C (pCHMP4C) formed a single ring at the Flemming body and was recruited with the same timing whether or not the checkpoint was active (Figure 3—figure supplement 1E–G). Tri-phosphorylated CHMP4C (pppCHMP4C) also formed single Flemming body rings and, additionally, formed double rings, one on either side of the Flemming body (Capalbo et al., 2016), particularly when the checkpoint was active (Figure 3—figure supplement 1G). However, checkpoint-dependent changes in pppCHMP4C distribution were much more dramatic at cytoplasmic sites (Figure 3A,B). In control midbody-stage cells, pppCHMP4C exhibited a granular nuclear localization, whereas cells with sustained checkpoint activity accumulated pppCHMP4C in 0.5–2 µm diameter cytoplasmic foci that we have named Abscission Checkpoint Bodies (ACBs). Pan-CHMP4C antibodies decorated ACBs, whereas antibodies specific to singly phosphorylated pCHMP4C did not, consistent with the tri-phosphorylated subpopulation of CHMP4C preferentially localizing to this site (Figure 3—figure supplement 1E, Figure 3—figure supplement 2A). Note, however, that detection of endogenous CHMP4C in ACBs with the pan-antibody required removing soluble cytoplasmic proteins by pre-fixation treatment with buffer (PHEM) to preserve general cell morphology followed by detergent to pre-permeabilize cell membranes. This treatment leaves larger cellular structures, including ACBs, intact while removing cytoplasmic background to increase ACB staining clarity. ACBs were observed in the majority of checkpoint-active cells throughout the entire post-thymidine time course (Figure 3C). ACBs were abundant (26 ± 14 per midbody-stage cell), and the number of ACBs per midbody-stage cell remained constant throughout the time course (Figure 3—figure supplement 2B). In summary, previously uncharacterized bodies that contain the abscission checkpoint factor pppCHMP4C form in the cytoplasm of cells in which the abscission checkpoint is active due to Nup153/50 depletion.

Initial screens for additional ACB components revealed that a second CHMP4 isoform, CHMP4B, was also present at this site. These two CHMP4 isoforms have similar sequences, yet perform quite different roles in cell division: CHMP4C regulates abscission timing but is not required for abscission, whereas CHMP4B is required for abscission but does not have a regulatory role (Capalbo et al., 2012; Carlton et al., 2012; Carlton et al., 2008). CHMP4B is an abundant cytosolic ESCRT-III protein, and like bulk CHMP4C, its ACB localization was also best visualized after pre-permeabilizing cells to remove soluble cytoplasmic proteins (Figure 3D). An antibody specific for the T232 phosphorylated, active form of AurB (pAurB) (Yasui et al., 2004) also stained ACBs (Figure 3D), explaining our previous observation that abscission checkpoint activity caused pAurB to localize to cytoplasmic foci of unknown composition (Mackay et al., 2010). In the absence of checkpoint activation, CHMP4B exhibited a punctate cytoplasmic distribution in pre-permeabilized cells, but these puncta were significantly smaller than ACBs and did not colocalize with either pppCHMP4C or pAurB. Hence, in addition to pppCHMP4C, pAurB and CHMP4B are ACB components, and in each case, their ACB localization was strongly checkpoint dependent (Figure 3E).

We further characterized ACBs by testing whether they corresponded to a variety of known cellular assemblies and organelles (Figure 4—figure supplement 1A–C). ACBs did not colocalize with a host of organelles, including P-bodies, which appeared the most similar in character and appearance yet were clearly distinct (Figure 4—figure supplement 1D). Unexpectedly, our survey revealed that ACBs are specifically detected with an antibody that recognizes SR domain-containing splicing factors as well as an antibody (mAb SC35) reported to primarily recognize the splicing factor SRRM2 (Ilik et al., 2020; Figure 4A,B Figure 4—figure supplement 2A,B). These antibodies are known to detect cytoplasmic bodies termed mitotic interchromatin granules (MIGs) (Fu and Maniatis, 1990; Li and Bingham, 1991; Reuter et al., 1985). During interphase, MIG components reside in a nuclear compartment called nuclear speckles where these components are hypothesized to be concentrated to increase splicing efficiency (Beck, 1961; Chen and Belmont, 2019). Both MIGs and nuclear speckles are compartments with liquid–liquid phase separation characteristics (Rai et al., 2018; Strom and Brangwynne, 2019), and they primarily contain factors that function in mRNA biogenesis, particularly splicing (Mintz et al., 1999; Saitoh et al., 2004; Uversky, 2017). After being released from nuclei upon mitotic nuclear envelope disassembly, these factors associate into small cytoplasmic MIG foci during the metaphase to anaphase transition. MIGs then normally disappear in late telophase, concomitant with stepwise reassembly of nuclear speckles within newly formed nuclei (Prasanth et al., 2003; Reuter et al., 1985; Spector and Lamond, 2011; Spector and Smith, 1986; Tripathi and Parnaik, 2008). In contrast, extended abscission checkpoint activity promoted the appearance of ACBs, which are larger than MIGs and become apparent in late midbody-stage cells (Figure 4B). Nuclear speckle assembly in the nucleus was delayed concurrently with ACB appearance (Figure 4A,B). ACBs were not an artifact of fixation as we observed them in live cells expressing SRRM2-mCherry (Figure 4C). We therefore propose that cytokinetic ACBs are derived from MIGs and that abscission checkpoint signaling plays a pivotal role in inducing the transition.

When abscission checkpoint satisfaction occurred promptly, pAurB, pppCHMP4C, and CHMP4B, newly discovered here to be MIG constituents, relocalized individually to different sites in control cells at telophase, rather than transitioning to ACBs. Interestingly, the checkpoint maintenance factors pAurB and pppCHMP4C relocalized to nuclei where they partially colocalized with nuclear speckles (Figure 4A, Figure 4—figure supplement 2A,B). Nuclear speckle localization was not observed for CHMP4B, which remained cytoplasmic and also localized to midbodies (Figure 3D, Figure 4—figure supplement 1B; Capalbo et al., 2012; Carlton et al., 2012). Abscission checkpoint activity delayed timely nuclear relocalization of pAurB and pppCHMP4C, as well as splicing factors recognized by the canonical MIG marker mAb SC35, until abscission or just prior (Figure 4A, Figure 4—figure supplement 2A,B). These observations indicate that MIGs/ACBs play a role in constraining the migration of their constituents to various different cellular sites.

To probe for direct ACB functions in abscission delay, we tested whether abscission timing was altered when we artificially stimulated their formation/maintenance without otherwise preventing abscission checkpoint satisfaction. This was accomplished by inhibiting DYRK3, a kinase that prevents premature MIG formation in metaphase (Rai et al., 2018), or by inhibiting CLK1, a kinase that promotes component release from nuclear speckles and thus may similarly regulate MIG stability, allowing them to persist and at least partially mimic their maturation into ACBs (Araki et al., 2015; Colwill et al., 1996). CLK inhibition was previously reported to accelerate abscission (Petsalaki and Zachos, 2016), but our studies employed a CLK1/2-specific inhibitor (CLK1i) rather than a pan-CLK inhibitor, and our treatment windows were much shorter than those employed previously (30 min vs. 5 hr). For our studies, we treated with CLK1 inhibitor or DYRK3 inhibitor for 30 min to limit effects on preceding cell-cycle stages. Under these conditions, inhibition of either CLK1 or DYRK3 kinase increased the percentage of midbody-stage cells with SC35-positive cytoplasmic foci from 13% (control) to 23% (DYRK3i) or 21% (CLK1i) (Figure 4D–F) and increased median abscission times in the bulk population by 11 min (DYRK3i, 19% delay) or 6 min (CLK1i, 10% delay), as measured in live cell imaging experiments (Figure 4G). Similarly, the time course of midbody resolution in a synchronized cell population was delayed upon treatment with either inhibitor (Figure 4—figure supplement 3). Conversely, when we overexpressed CLK1 in a background of Nup153/50 depletion, midbody-stage cells decreased ~25% (Figure 4H). CLK1 overexpression did not completely dissolve ACBs, but did decrease the number of ACBs per midbody pair (Figure 4I, Figure 4—figure supplement 4A), and concomitantly increased the proportion of SRRM2 nuclear staining (Figure 4—figure supplement 4B,C), indicating that CLK1 overexpression induced SRRM2 relocalization from ACBs into nuclei. Thus, treatments that promote MIG/ACB formation and maintenance delay abscission in cells independently of perturbations that prevent the abscission checkpoint from being satisfied, and treatments that promote ACB dissolution accelerate abscission, even when the abscission checkpoint is active. These effects are modest, but the experiments target just one aspect of abscission regulation in isolation and therefore support the model that ACBs have a functional role in abscission delay.

To confirm that ACBs are a broadly relevant component of the abscission checkpoint, we investigated their formation under the three other conditions known to prevent abscission checkpoint satisfaction: replication stress, intercellular tension, and lagging chromatin bridges within the midbody (Lafaurie-Janvore et al., 2013; Mackay and Ullman, 2015; Steigemann et al., 2009). We observed that ACB levels significantly increased in midbody-stage cells under conditions of replication stress (Figure 5—figure supplement 1) or high intercellular tension (Figure 5—figure supplement 2), but not in the presence of chromatin bridges (Figure 5—figure supplement 3). Delayed ALIX midbody recruitment followed this same trend (Figure 5—figure supplements 1G, 2D, and 3D). Thus, ACB formation and delayed recruitment of ALIX to the midbody were correlated and both were seen under three of four different conditions known to prevent satisfaction of the abscission checkpoint.

To test whether this newly identified checkpoint mechanism is deployed in other cellular contexts, we probed the non-transformed epithelial cell-line RPE1 for key checkpoint hallmarks. RPE1 cells exhibited abscission checkpoint activity in response to Nup depletion, as measured by an increase in midbody-stage cells and a delay in ALIX recruitment to the midbody (Figure 5—figure supplement 4A–C). This response was less robust than observed in HeLa cells, suggesting that corrective or compensatory pathways may mitigate errors that keep the abscission checkpoint from being satisfied or that RPE1 cells may be less dependent on the ESCRT pathway for abscission. Nonetheless, ACBs were once again detected in nearly all midbody-stage RPE1 cells subjected to the checkpoint enrichment protocol and contained pppCHMP4C, pAurB, CHMP4B, and splicing factors recognized by mAb SC35 (Figure 5A–C, Figure 5—figure supplement 4D). These observations demonstrate that checkpoint-dependent delay of ALIX midbody recruitment and ACB formation occur in distinct cell types and are broadly relevant as abscission checkpoint mechanisms.

To probe the relationship between the abscission checkpoint and ACBs further, we depleted the key abscission checkpoint regulator CHMP4C and assayed ACBs. As expected, co-depletion of CHMP4C in Nup-depleted cells abrogated the checkpoint and reduced midbody-stage cells (Figure 5D, full bars). This treatment also significantly reduced the percentage of midbody-stage cells with ACBs (Figure 5D, shaded regions). The dependence of ACB formation on CHMP4C reinforces the connection between abscission checkpoint signaling and ACB appearance. The suppression of ACB formation in the absence of CHMP4C, despite Nup depletion, also demonstrates that ACBs do not form due to altered function of the nuclear pore per se (Figure 5D, Figure 5—figure supplement 5).

To investigate the connection between the presence of ACBs and the delay of ALIX recruitment to the midbody, we tested whether ALIX itself might target to ACBs. To avoid the strong signal for ALIX present in the cytoplasm (Figure 2C), we again used pre-permeabilization treatment to remove soluble cytoplasmic content. Using this method, we detected ALIX in ACBs in both HeLa and RPE1 cells (Figure 6A,B). To determine whether this localization was specific, we performed a co-depletion with siALIX and siNups, which was informative at several levels. First, as previously established, ALIX depletion alone resulted in cytokinesis failure (Carlton and Martin-Serrano, 2007; Morita et al., 2007; Figure 6—figure supplement 1A,B), and when abscission checkpoint satisfaction was prevented by concurrent treatment with siNups, enrichment of midbody-stage cells still took place (Figure 6C, full bars). This checkpoint activity corresponded to a significant reduction in cytokinesis failure (Figure 6—figure supplement 1A–B). These observations are consistent with the conclusion that reduced ALIX activity prevents abscission during checkpoint regulation. Second, depletion of ALIX confirmed its localization to ACBs as its detection at ACBs diminished concordantly (Figure 6—figure supplement 2A,B). Third, in the absence of ALIX, although ACB-like structures still persist in late midbody-stage cells in response to checkpoint activation (Figure 6C, shaded regions), the average cross-sectional area of these structures decreased by ~35% (corresponding to ~50% reduction in volume), as measured using two different ACB markers (Figure 6D, Figure 6—figure supplement 1C). Concomitantly, the number of the focal ACB-like structures increased in checkpoint-active midbody-stage cells depleted of ALIX (Figure 6—figure supplement 1D). Thus, ALIX is a component of ACBs and is required to create full-sized ACBs or to maintain their integrity. Interestingly, although pAurB and pppCHMP4C clearly localize to MIGs (Figure 6—figure supplement 2E), we could not detect ALIX in MIGs (Figure 6—figure supplement 2C–D). These observations indicate that ALIX is present in ACBs and, although we cannot rule out that our detection method is limiting for MIGs, it appears that ALIX recruitment contributes to ACB maturation from MIGs.

The presence of ALIX in ACBs suggested that the checkpoint-dependent delay in ALIX recruitment to the midbody may occur because a modified subpopulation of ALIX that would normally be targeted to the midbody is instead retained in ACBs. To determine whether ACB formation and ALIX midbody recruitment delay are connected, cells were scored for both the presence of ACBs and the intensity of ALIX at the midbody, under conditions that enrich for abscission checkpoint-active cells. Consistent with a direct ALIX sequestration model, we found that the presence of ACBs strongly correlated with decreased ALIX levels at the midbody in individual cells (Figure 6E, Figure 6—figure supplement 2F). This correlation was striking even late in the post-thymidine time course, when overall ALIX midbody recruitment has begun to recover.

Details regarding antibodies used in this study can be found in Supplementary file 1.

Details regarding plasmids used in this study can be found in Supplementary file 2. DNA was amplified using PCR and ligated into the pLVX-inducible vector using Gibson Assembly according to the manufacturer’s instructions (NEB, Rowley, MA).

HeLa cells were cultured and maintained at 37°C and 5% CO₂ in DMEM supplemented with 10% FBS. The Tet-On HeLa dox-inducible and HA-CHMP4C (Carlton et al., 2012) cell lines were supplemented with 100 µg/ml G418 to maintain Tet-On or HA-CHMP4C expression respectively (Invitrogen, Carlsbad, CA). RPE1 cells were supplemented with 10 µg/ml hygromycin (Invitrogen) to maintain hTERT expression. The SRRM2-mCherry cell line was maintained in 100 µg/ml G418 to maintain Tet-On expression and 10 µg/ml puromycin (InvivoGen, San Diego, CA) for SRRM2-mCherry expression. At the outset of these studies, HeLa, HeLa Tet-On, and RPE1 cells were tested and found negative for mycoplasma using a PCR mycoplasma detection kit (ABM, Bellingham, WA). Cell types were authenticated by sequencing 24 loci (University of Utah Sequencing Core). Unless otherwise labeled in individual panels or figure legends, all experiments used HeLa cells.

To generate a stable cell line, Day 1: Hela Tet-On cells were plated in a 24-well dish. Day 2: cells were transfected with 500 ng/well of the SRRM2-mCherry plasmid using Lipofectamine LTX with Plus Reagent according to the manufacturer’s instructions (Thermo Fisher Scientific, Waltham, MA). Day 3: cells were split into a 10 cm dish with 500 μg/ml G418 and 10 μg/ml Puromycin for selection. After drug selection for 14 days, colonies were harvested into individual wells of 12-well plates. Clones were validated for inducible SRRM2-mCherry expression using immunofluorescence imaging and an antibody specific to mCherry (Abcam, Cambridge, UK, ab167453).

Cells were singly transfected with siRNA for 42–72 hr, as indicated in figure legends, using Lipofectamine RNAiMax (ThermoFisher) according to the manufacturer’s instructions. Media was exchanged 24 hr after transfection, and cells were incubated between 18 and 48 hr before harvesting. For co-transfections, cells were seeded with combinations of (1) siALIX, siCon, or siNup153/50, or (2) siCHMP4C, siCon, or siNup153/50. Media was exchanged 24 hr after transfection. Cells were fixed 72 hr post-transfection to allow adequate time for CHMP4C or ALIX knockdown. Details regarding siRNAs used in this study are provided in Supplementary file 2.

Cells were seeded on acid-washed, 10 µg/ml fibronectin-treated, glass coverslips and then treated according to the individual experimental protocol. To fix, cells were removed from the incubator, washed once in PBS pH 7.4, and then fixed by one of the three methods as notated in Supplementary file 1:

After fixing, cells were rinsed twice with 2 ml PBS each time and then blocked (3% FBS and 0.1% Triton X-100 in PBS) for 30 min on the bench top. Primary antibodies were applied at the dilution noted in Supplementary file 1 for at least 1 hr, 23°C in blocking solution. After 1 wash with 2 ml PBS, secondary antibodies (Thermofisher) were applied for 45 min to 1 hr, and cells were washed in 2 ml PBS. For Figure 2E, Figure 2—figure supplement 2C, DNA was stained with Hoechst for 10 min at 23°C. For all other images, coverslips were mounted with ProLong Gold Antifade Reagent with DAPI (Thermofisher) on a microscope slide.

Images were acquired using four different microscopes:

The SRRM2-mCherry-inducible cell line was plated in four-well LabTek chamber slides. Day 1: cells were treated with 2 μg/ml Doxycycline to induce SRRM2-mCherry expression. Day 2: cells were treated with siCon or siNups as above. Day 4: cells were treated with one drop NucBlue reagent (ThermoFisher) per 2 ml of imaging media and 1 μM Tubulin Tracker Green (Thermofisher) 45 min prior to imaging in HEPES buffered, Phenol Red-free DMEM/F-12 (ThermoFisher) containing 10% FBS. Cells were imaged on a Leica SP8 White Light Nikon Ti-E widefield inverted microscope at 60× objective.

Midbody-stage cells were identified by α-tubulin staining and were always counted as one cell. Cell populations were counted and sorted into interphase, midbody-stage, mitotic, recently abscised pair (counted as one cell), multinucleate, and failed bridge (counted as one cell). Cells were counted unblinded while on the microscope.

Midbodies were designated as ‘early’ or ‘not early’ using α-tubulin staining before scoring for the examined phenotype. Midbodies were classified as early based on midbody width, level of midbody-pinching, nuclear area, and flatness of cells, see Figure 4B for illustration.

Midbodies were scored as having ACBs if they had at least one ACB. However, most midbody-stage cells with ACBs had 10–40 ACBs. For most figures, cells were counted as having or not having ACBs, unblinded while on the microscope. In Figure 3—figure supplement 2B and Figure 4I, ACBs were counted from images using ‘Find Maxima’ with Fiji software (NIH). To quantify ALIX signal in ACBs, deconvolved z-slices were stacked and then uniformly adjusted for brightness and contrast. The SC35 or pAurB channels were uniformly thresholded, and then cytoplasmic objects between 0.025 and 5 µm² were marked as regions of interest (ROIs). ROIs were used to measure individual ACB intensity and area in the non-thresholded ALIX channel. Values were background corrected with large ROIs in the cytoplasm that excluded ACBs.

Quantification of fluorescence staining intensity at the midbody was also done with Fiji software (NIH). The freehand selection tool was used to outline the region of interest at the midbody, and staining intensity within this area was measured. Signals were background corrected using measurements from adjacent regions. In cases where the protein was not visible at the midbody, these methods sometimes generated a small negative intensity value after background correction because the Flemming body is a natural dark zone. In these cases, intensity was valued at zero.

To quantify SRRM2 subcellular distribution, the DAPI channel was used to threshold the nuclei and create ROIs using Fiji software. These ROIs were then used to determine the mean SRRM2 nuclear intensity. For each midbody-stage cell, an N:C ratio was determined by dividing the average mean nuclear intensity by the baseline cytoplasmic signal (the mean cytoplasmic intensity measured in an area devoid of ACBs).

Cells were lysed in NP40 lysis buffer (50 mM Tris pH 7.4, 250 mM NaCl, 5 mM EDTA, 50 mM NaF, 1 mM Na₂VO₄, 1% Nonidet P40, 0.02% NaN₃, with freshly added PMSF [1 mM], aprotinin, leupeptin) for 30 min, 4°C, vortexing every 10 min. Lysates were clarified by spinning at 13,200 g for 10 min at 4°C, and protein contents in clarified lysates were quantified by Bradford assay before gel loading. Gel and western conditions are listed in Supplementary file 1. Twenty to 35 µg lysate per sample was prepared with SDS loading buffer, resolved by SDS–PAGE, and following wet transfer for 2 hr at 40 V, membranes were blocked with 5% milk in Tris-buffered saline (TBS) for at least 30 min, 23°C, and were incubated with primary antibodies in 5% milk in TBS-T (0.1% Tween 20/TBS) overnight at 4°C. Membranes were washed in TBS-T and incubated with the corresponding secondary antibodies conjugated with HRP (Thermofisher) or near-infrared fluorescent dyes (Abcam, Cambridge, UK) in TBS-T for 1 hr, 23°C, and washed again. Blots were detected with Western Lightning PLUS ECL (PerkinElmer, Waltham, MA) on Hyblot CL film (Thomas Scientific, Swedesboro, NJ). Proteins with infrared fluorescent dyes were detected using a Li-Cor Odyssey Infrared scanner and Image Studio version 5.2 software.

The abscission checkpoint was kept active using one of four different methods.

For fixed-imaging time courses, HeLa cells were seeded on glass coverslips in 24-well dishes, and 2 mM thymidine was added 8 hr post-seeding. After 24 hr, thymidine was washed out with PBS. Thirteen hours post-thymidine release, after most cells had completed metaphase, 1 µM DYRK3 inhibitor (GSK 626616, Tocris, Minneapolis, MN), 1 µM CLK1/2 inhibitor (534350, Millipore, Burlington, MA), or DMSO were each added individually to six separate wells. Coverslips from wells corresponding to each treatment were fixed directly after inhibitor/vehicle was added for the first timepoint at 13 hr post-thymidine release and then at hourly intervals until 18 hr post-thymidine release.

Live-imaging experiments to determine abscission timing used HeLa cells expressing H2B-mCherry and GFP-α-tubulin. Cells were seeded in a Lab-Tek II 8-chambered #1.5 German Coverglass System and incubated for 48 hr. Stage positions were first set on the microscope, and then 1 µM DYRK3i, 1 µM CLK1/2i, or DMSO were added just prior to initiation of imaging. Imaging was carried out for 3 hr on a Nikon Ti-E widefield inverted microscope (Nikon 20x N.A. dry objective lens) equipped with Perfect Focus system and housed in a 37°C chamber (OKOLAB, Ambridge, PA) with 5% CO₂. Multiple fields of view were selected at various x and y coordinates, and images were acquired using a high-sensitivity Andor Zyla CMOS camera (Andor, Manchester, CT) controlled by NIS-Elements software. Images were acquired every 5 min, and abscission time was measured as the time from midbody formation to disappearance. In parallel, cells were seeded on glass coverslips, incubated for 48 hr, and either DYRK3i, CLK1i, or DMSO were added for 60 min prior to fixing and detection of foci labeled with mAb SC35 or antibodies against pAurB.

On Day 1, cells were seeded into 24-well dishes both with and without glass coverslips and transfected with either siCon or siNups siRNA as described above. On Day 2, cells were transfected with 800 ng/well of empty vector DNA or Myc-CLK1-WT DNA using Lipofectamine LTX according to the manufacturer’s instructions (Thermofisher). At the time of transfection, 2 µg/ml doxycycline was added to the medium to induce expression. On Day 3, media was changed and fresh media added with 2 µg/ml doxycycline. On Day 4, cells were harvested for immunofluorescence and immunoblot as described above.

Five statistical tests were used to evaluate the significance of data. When comparing two samples, p-values were calculated using an unpaired t-test or Mann–Whitney test for non-normally distributed datasets (Figure 4I, Figure 1—figure supplement 1A, Figure 4—figure supplement 4B, Figure 5—figure supplement 2A,B, Figure 5—figure supplement 4A). When comparing two samples with multiple cells in each replicate, p-values were calculated using Stratified Analysis with Nonparametric Covariable Adjustment to adjust for replicate-level variation (Figure 4G, Figure 6D,E, Figure 5—figure supplement 1C,G, Figure 5—figure supplement 2D, Figure 5—figure supplement 3D, Figure 6—figure supplement 1D, Figure 6—figure supplement 2B,D). When comparing two complete time curves, p-values were calculated using Welch’s t-test (Figure 3C). When comparing datasets with three or more samples, p-values were calculated using one-way ANOVA with Sidak’s multiple comparisons test (Figure 2F,G, Figure 3E, Figure 4F,H Figure 5C,D, Figure 6C, Figure 2—figure supplement 1D, Figure 2—figure supplement 2D, Figure 2—figure supplement 3C, Figure 3—figure supplement 2B, Figure 4—figure supplement 3C,D, Figure 5—figure supplement 3B, Figure 6—figure supplement 1B). When comparing two or more datasets over a time course or multiple categories within a sample, p-values were calculated using two-way ANOVA with Sidak’s multiple comparisons test (Figure 1—figure supplement 1B, Figure 3—figure supplement 1G, Figure 4—figure supplement 3B, Figure 5—figure supplement 1B,E Figure 5—figure supplement 4C).