PIASγ Is Required for Faithful Chromosome Segregation in Human Cells

PIASγ is required for timely anaphase onset in human cells

Based on the hypothesis that the E3 sumo ligase PIASγ contributes to the regulation of sister separation in human cells, we sought to test if depletion of PIASγ perturbs mitotic progression. We devised a protocol that allowed depletion of PIASγ from HeLa cells simultaneously with cell cycle synchrony in early S-phase (Fig. 1A,B ). Briefly, control or PIASγ-specific siRNA was added to asynchronously growing HeLa cells at 30% confluence, the siRNA was removed after 6 hours, then after 18 hours a double-thymidine synchrony procedure was employed. Samples were taken every 2 hours after release and processed for cytological and biochemical analysis. Upon release from early S-phase, control-treated and PIASγ-specific siRNA-treated cells first entered mitosis after ∼10 hours (Fig. 1C,I ). While most control cells had reached the subsequent G1 phase by 14 hours, PIASγ siRNA-treated cells accumulated in mitosis (∼35%) and the mitotic index remained high up to 24 hours after the release of the thymidine block (Fig. 1I,J and data not shown). This mitotic delay was also evident in the high levels of cyclin B and phosphorylated histone H3 (PH3) during this time frame (Fig. 1C ). The cells in mitosis lacked PIASγ protein as judged by Western blot, whereas PIASγ protein was detectable in the population of interphase cells (Fig. 1B ), indicating that those cells which accumulated in mitosis were the same cells in which PIASγ protein had been efficiently depleted. Analysis of the cytology of the mitotic cells revealed that most had metaphase chromosome arrangements indistinguishable from control metaphases (Fig. 1D,E ). That is, the chromosomes were organized into a metaphase plate arrangement and centromeric regions of the chromosomes possessed distinct primary constrictions. Many of these cells must have endured a prolonged delay in mitosis because at the later time points (18–24 hrs) they displayed over-condensed chromosomes typical of cells in which anaphase onset had been blocked (Fig. 1F,G ). Also at later time points, similar cells with overcondensed chromosomes had several chromosomes positioned off the metaphase plate (Fig. 1H ). Very few cells lacking PIASγ were observed with separated sister chromatids (Fig. 1J ). (For a more detailed description of these phenotypes of PIASγ-depleted cells see Fig. S1 and S2.) These data indicate that PIASγ is needed for the metaphase-to-anaphase transition in human cells and therefore suggest that sumoylation of mitotic substrates is a key component of anaphase initiation.

Metaphase plates form normally without PIASγ but chromosomes can then leave the plate

That a maximum of 35% of the cells accumulated in mitosis after PIASγ depletion (Fig. 1I and data not shown) suggested either that the cells entered mitosis more slowly than in controls, or that the cells eventually exited mitosis after delaying in metaphase (or a combination of both of these factors). Slower entry into mitosis was consistent with the result that, when the above time course experiment was repeated in the presence of nocodazole, PIASγ-depleted cells accumulated in mitosis to a lesser extent than controls (Fig. 1K ). However, to determine if some cells exited mitosis after PIASγ-depletion, and to gain a more detailed understanding of the mitotic defect in live cells, we performed 16–20 hour time-lapse analyses of HeLa cells possessing a fluorescent tagged histone (H2B-GFP) [23] following early S-phase synchrony (Fig. 2, Fig. S3, Table S1, and Movies S1–S4).

In controls, mitosis proceeded normally (Fig. 2A,B , Fig. S3 and Table S1), and cell synchrony was evidently sharp since 33/51 (64%) of the cells within a particular field progressed into mitosis during a short 4 hour time period (10–14 hours after release from early S-phase). PIASγ-depleted cells also reached mitosis quite synchronously with over half of the cells in a single field initiating mitosis within a 5 hour time frame after release from early S-phase (Fig. S3). Consistent with the time course experiment described in Fig. 1, most PIASγ-depleted cells became strongly delayed in mitosis. The average time spent in mitosis was 6 hours 35 minutes (s.d. = 3 hours 56 minutes; n = 26), compared with 69 minutes in control cells (s.d. = 24 minutes; n = 36).

PIASγ-depleted cells usually performed prometaphase with normal timing forming bona fide metaphase plates (Fig. 2C,D , Fig. S3 and Table S1). In these cells, metaphase lasted 1–6× longer than in control cells (on average 110 minutes with s.d. = 78 minutes, compared with 44 minutes with s.d. = 25 minutes in controls). The maximum length of metaphase recorded was 4 hours 10 minutes. Following the prolonged metaphase period, two alternative outcomes were observed, either (i) the cell initiated anaphase (Fig. 2C ) and occasionally, lagging chromosomes were observed (see bottom cell in Movie S3 that corresponds to Fig. 2C ; see arrows) or anaphase initiation was asynchronous (see Fig. S1P ), or (ii) several chromosomes departed from the metaphase plate and migrated toward a spindle pole (Fig. 2D, top cell). In this latter category, the cells appeared to have reverted to a prometaphase-like state; in some cases these cells recovered complete metaphase alignment of all chromosomes and at a later stage, after further delay in metaphase, were able to perform anaphase (Fig. 2D, bottom cell). Other cells in this category remained in this “de-congressed” metaphase state until the end of the time-lapse movie (Fig. 2D, top cell).

Together with the cytological analysis in Figure 1 and Figure S1, these live cell imaging studies provide a detailed description of the mitotic defects caused by PIASγ-depletion. Most cells reach metaphase in a timely manner and form apparently normal metaphase plates. After a prolonged metaphase delay cells either performed anaphase or a small number of chromosomes left the plate and migrated to the spindle poles. This de-congressed metaphase state was maintained in most cells for many hours.

The spindle checkpoint ought to be satisfied without PIASγ but centromeres remain cohered

That metaphase plates formed normally without PIASγ but anaphase failed to initiate on time seemed counterintuitive. The time-lapse studies indicated that when initiated, anaphase proceeded normally in some PIASγ-depleted cells, indicating a functional mitotic spindle and motors. In agreement, immuno-staining of alpha-tubulin revealed mitotic spindles that were indistinguishable from control metaphases in cells lacking PIASγ (Fig. 3A,G ). Moreover, the kinetochores of chromosomes at the metaphase plate in both control cells and cells that lacked PIASγ clearly possessed far less CENP-E than kinetochores of chromosomes away from the plate (Fig. 3B–F,H–K ). This staining pattern is typical of spindle checkpoint inactivation after congression of each chromosome to the metaphase plate [24], [25]. That metaphase plates formed in PIASγ-depleted cells which were indistinguishable from control metaphase plates, also argues that the spindle checkpoint ought to have been silenced because biorientation of chromosomes at the metaphase plate can only occur if proper kinetochore-microtubule attachments are made, resulting in a balance of opposite forces exerted by the spindle on each cohered sister and producing the proper spindle tension on each chromosome. Therefore, the prolonged metaphase delays in PIASγ-depleted cells indicate that a mitotic checkpoint is able to act in metaphase: this has only been previously reported to occur in response to inhibition of Topoisomerase II [26]–[28].

An Aurora B-dependent and Mad2-dependent checkpoint is activated following PIASγ depletion

The above findings indicated that a lack of PIASγ might have triggered a mitotic checkpoint. Although the spindle checkpoint appeared to have been silenced in metaphases lacking PIASγ, based on the reduction of CENP-E staining at kinetochores, we found that co-depletion, using RNAi, of Mad2 and PIASγ abrogated the accumulation of cells in mitosis following release from a double thymidine synchrony (data not shown). We also found that the small molecule inhibitor of Aurora B, ZM447439 [29], was able to relieve the metaphase arrest (Fig. 4). In control cells that were depleted of Hec1, a treatment that induces a persistent spindle checkpoint arrest [30], addition of ZM447439 resulted in rapid sister separation followed by mitotic exit in almost all cells (Fig. 4A–E ). Similarly, after we collected PIASγ-depleted mitotic cells by mitotic shake-off then added ZM447439, most cells rapidly performed sister chromatid separation and segregation, followed by mitotic exit (Fig. 4F–J ). Anaphase and mitotic exit could have been a result of cyclin B degradation upon ZM447439 treatment. Consistent with this, addition of the Cdk inhibitor roscovitine had similar effects on PIASγ-depleted cells (Fig. 4K–O ), inducing chromatid separation, segregation to the spindle poles and mitotic exit. These experiments therefore resolved several important issues. First, PIASγ-depleted cells are largely capable of removing the cohesion from between the sisters as long as Cdk-cyclin B is inactivated. Second, as seen in the time-lapse studies, PIASγ-depleted cells treated with ZM447439 or roscovitine must have possessed functional mitotic spindles because chromosomes segregated to the spindle poles before decondensing in telophase. Third, the metaphase delay imposed by PIASγ depletion is a checkpoint response.

Checkpoints can monitor defects in biological processes and respond by delaying cell cycle progression [8]. The metaphase delay in PIASγ-depleted cells is typical of a checkpoint response because anaphase onset is blocked despite being mechanically able to proceed, the delay can be bypassed by chemical inhibition of one of the checkpoint components (in this case Aurora B), and the delay is spontaneously overcome in some cells even though a successful anaphase could not always be completed. When overridden by ZM447439 or roscovitine, chromatid disjunction was rarely efficient (Fig. 4F–O ). Typically, bypass of a checkpoint delay reveals the cellular aberration that activated the checkpoint. In the case of the delay in PIASγ-depleted cells, checkpoint bypass revealed multiple chromatin bridges between the dividing nuclei in almost every cell. Laggard chromosomes were also seen in some of the anaphases in the time-lapse movies of PIASγ-depleted cells that had previously delayed in metaphase for a prolonged period. These data indicate that the defect in PIASγ-depleted cells might be an inability to efficiently separate the sister chromatids.

PIASγ is required for sister separation in hSgo1-depleted cells

The lack of PIASγ resulted in checkpoint activation at the metaphase-to-anaphase transition and the cells failed to remove cohesion from the centromeres. However, the above experiments did not address directly if the lack of PIASγ perturbed the removal of cohesin. A number of important studies have shown that sister chromatid cohesion cannot be maintained in cells depleted of the cohesin protector hSgo1, even under conditions in which the spindle checkpoint is active [13], [31]–[35]. These results demonstrated that without hSgo1, cells are oblivious to spindle checkpoint controls and that cohesin is rapidly removed from chromosomes upon nuclear envelope breakdown. If PIASγ depletion arrested cells in metaphase by activating the spindle checkpoint, but has no direct role related to chromosome cohesion, then depletion of hSgo1 should relieve sister chromatid cohesion after PIASγ knock-down. As previously described, we could efficiently deplete HeLa cells of hSgo1 by RNA interference (data not shown) and found that, similar to previous reports, the cells accumulated in mitosis with all of their sister chromatids fully separated, even in the presence of nocodazole [13]. In the absence of hSgo1, sister chromatids were similarly able to separate in cells in which APC/C activity had been eliminated by depletion of the Apc2 component of the catalytic site and by the simultaneous addition of nocodazole (Fig. 5A,B ). Furthermore, if the spindle checkpoint was persistently activated by depletion of Hec1 (Fig. 5C ), simultaneous depletion of hSgo1 resulted in complete loss of sister cohesion (Fig. 5D ). Each of these experiments confirms that the known spindle checkpoint pathways are unable to preserve cohesion in the absence of cohesin guardian hSgo1.

We depleted both PIASγ and hSgo1 in synchronized HeLa cells using RNAi (data not shown) and a double thymidine block protocol, then released into the cell cycle either with or without nocodazole. As previously reported, hSgo1-depleted cells accumulated in mitosis with separated sister chromatids, whether or not nocodazole was present in the medium, and cell cycle progression was blocked in a telophase-like state (Fig. 5G,J,L,N ). Unexpectedly, however, the doubly depleted HeLa cells lacking PIASγ and hSgo1 accumulated in metaphase/de-congressed metaphase with cohered sister chromatids (Fig. 5H–J,M,N ), just like the cells depleted of PIASγ only (Fig. 5E,F,K ). Thus, remarkably, even in the absence of the cohesin protector hSgo1, PIASγ is required for sister chromatid separation. The same result was observed when PIASγ was depleted simultaneously with Sororin, a protein that interacts physically with the cohesin complex and is required for sister chromatid cohesion in mitosis (data not shown) [36], [37]. Together these experiments indicate that PIASγ might be directly involved in the removal of cohesion.

PIASγ is not required for removal of cohesin from centromeres

The lack of sister separation in PIASγ/hSgo1 doubly depleted cells could be explained in one of two ways: either, (1) PIASγ is required for cohesin removal even in the absence of the cohesin guardian, or (2) sister chromatids remain cohered at the centromeres in the absence of cohesin. To test this we immuno-localized Rad21 in cells after PIASγ-depletion, hSgo1-depletion, or in doubly depleted cells. As expected, mitotic chromosomes in hSgo1-depleted cells lacked any detectable cohesin except before breakdown of the nuclear envelope, in which case cohesin was strongly detected throughout the nucleus (Fig. 5Q,Q′). PIASγ-depleted mitotic cells, however, like control cells, possessed clearly defined regions of centromeric Rad21 between the paired kinetochores of each cohered chromosome (Fig. 5P,P′ ). Some Rad21 was also seen between the chromosome arms (Fig. 5O,O′,P,P′ ). Strikingly, Rad21 could not be observed between the paired kinetochores or the arms of the cohered sisters in hSgo1/PIASγ doubly depleted cells (Fig. 5R,R′ ). Therefore, PIASγ is not required for removal of cohesin from chromosomes that occurs in the absence of hSgo1, but PIASγ is required for sister chromatid separation under the same experimental conditions. Thus, cohesion between sister kinetochores was maintained in the absence of detectable Rad21.

DNA catenations might maintain the centromeric primary constriction and cohesion at the centromere in the absence of cohesin

Since PIASγ was required for sister separation under two different conditions (absence of Sororin or hSgo1) in which cohesin-based cohesion cannot hold sisters together, and because we were unable to detect cohesin Rad21 at centromeres in PIASγ/hSgo1 depleted cells, we speculated that cohesin was not the sole component providing sister cohesion after PIASγ depletion. In yeast, components and regulators of the cohesin complex are modified by sumo ligases and, in addition, yeast Topoisomerase II is sumoylated. A known mechanism that joins sister chromatids, though not known to be strictly regulated, is DNA catenation, that arises as sister DNA molecules are synthesized during S-phase. In budding yeast and Xenopus, PIASγ-mediated sumoylation of DNA Topoisomerase II, the only enzyme capable of removing catenations from between sister chromatids, is thought to target Topoisomerase II to centromeres or pericentric regions of chromosomes during mitosis [16], [21]. It was therefore plausible that catenations, in addition to cohesin, linked the sister chromatids in PIASγ-depleted cells. This could explain why PIASγ and hSgo1 doubly depleted cells retained sister chromatid cohesion in the absence of cohesin and would be indicative of a need for PIASγ for catenation removal. To test this hypothesis we employed a specific inhibitor of Topoisomerase II, ICRF-193, that locks the enzyme in the so-called “closed-clamp” form, preventing concatenated sister duplexes from being resolved. We depleted PIASγ from HeLa cells before a double thymidine synchrony and then collected the cells that became arrested in mitosis after release from the S-phase block. As described in Figure 4, the Cdk inhibitor roscovitine or the Aurora B inhibitor ZM447439 caused such mitotic cells to separate most of their sister chromatids and then segregate them to the spindle poles, demonstrating that sister chromatid cohesion was largely removed. If PIASγ-depleted mitotic cells possess catenations that hold the sister DNA molecules together, then inhibition of Topoisomerase II ought to block the sister separation that is forced upon roscovitine or ZM447439 treatment. We added roscovitine (data not shown) or ZM447439 to the PIASγ-depleted mitotic cells simultaneously with ICRF-193 and prepared samples for cytology. Strikingly, inhibition of Topoisomerase II completely blocked sister chromatid separation in every cell observed. That Topoisomerase II was required for sister separation under these conditions, indicates that catenations were indeed present in the PIASγ-depleted metaphase-arrested cells (Fig. 6A–E ).

PIASγ is required for efficient localization of Topoisomerase II to centromere regions and chromosome cores in mitotic human cells

If indeed catenations could not be removed efficiently from the centromeres of PIASγ-depleted cells, then these catenations must have persisted despite the fact that Topoisomerase II is active in mitotic cells. One mechanism that could account for this apparent paradox would be if PIASγ helps to direct the decatenatory activity of Topoisomerase II to centromeric catenations. To test this hypothesis, we immuno-localized Topoisomerase IIα in control mitotic cells and in cells depleted of PIASγ (Fig. 6F–J ). During mitosis, Topoisomerase II is associated with the axial cores that run the length of condensed chromosome arms, but is also specifically concentrated at the centromere regions [38]–[43]. Using polyclonal antisera directed at Topoisomerase IIα, we reproducibly observed this staining pattern (core localization and intense staining at the centromere region) in almost 90% of the control cells (Fig. 6F,G,J ). Strikingly, however, fewer than 5% of PIASγ-depleted mitotic cells had this staining pattern. Instead, almost 40% of PIASγ-depleted mitotic cells had prominent staining of the chromosome cores along the chromosome arms, but lacked the intense staining at the centromere regions (Fig. 6I,J ). A further 48% of the PIASγ-depleted cells had a pattern of diffuse staining coincident with the chromatin, but not well localized to the cores or centromere regions (Fig. 6H,J ). Other proteins that specifically localize to centromere regions during mitosis, such as INCENP and CENP-F, localized to centromeres equally well in control and PIASγ-depleted mitotic cells (Fig. 6J and data not shown). These data are consistent with a need for PIASγ for proper localization of Topoisomerase II to centromere regions of chromosomes in mitosis and further suggest that localization to chromosome cores is less efficient in the absence of PIASγ.

Two different mechanisms regulate sister chromatid cohesion

Separation of sister chromatids at the metaphase-anaphase transition is the key moment of the mitotic cell cycle and its accuracy allows faithful partitioning of the duplicated genome. Groundbreaking studies have described a cohesin-based system that physically holds sister chromatids together and the mechanisms that regulate dissolution of this glue in preparation for anaphase [44]. In yeasts, firm genetic evidence has established that cohesin is the predominant, if not the sole, factor that accounts for sister cohesion and DNA catenations are removed from yeast chromosomes well before anaphase onset [45]. But in vertebrates, unlike in yeast, DNA catenations as well as cohesin complexes are present at centromeres until anaphase [46], [47]. Whether centromeric DNA catenations play an important functional role, rather than persisting as a mere coincidence, has not been understood because evidence that the catenation state of DNA is regulated through the cell cycle has been lacking. It is clear, however, that mechanisms must exist to ensure the efficient removal of centromeric catenations once the commitment to separate sister chromatids has been made. Here we provide the first hint that DNA catenations between human sister chromatids are specifically targeted for removal at or just prior to anaphase onset and that centromeric decatenation is regulated by distinct mechanisms from those which orchestrate removal of cohesin from the centromere. Our data suggest that a sumo ligase, PIASγ, promotes sister decatenation in preparation for or during sister separation, by specifically targeting Topoisomerase II to centromeric regions where catenations remain until anaphase.

In metaphase mammalian cells cohesin complexes persist at the centromeres [48]. DNA catenation must also exist at the centromeres in metaphase because Topoisomerase II inhibitors block chromatid disjunction when added to cells just prior to anaphase onset [46], [47]. We have shown that HeLa cells depleted of PIASγ suffer a prolonged metaphase block in which cohesin and catenations are maintained at the centromere, just like in normal metaphase cells. However, upon removal of cohesin from these centromeres by co-depletion of hSgo1, DNA catenations remain between the sister chromatids. Thus, although the regulation of cohesin binding at centromeres clearly depends on the cohesin guardian hSgo1, the persistence of DNA catenations at centromeres does not depend on hSgo1 (in the absence of PIASγ) nor does it depend on the presence of cohesin. In the apparent absence of cohesin at the centromere, PIASγ/hSgo1 doubly depleted cells block in metaphase: chromosomes with cohered sisters remain at the metaphase plate, and must therefore be under tension from the mitotic spindle, but remain physically tethered by catenations. These data provide two novel conclusions about the mechanism of sister cohesion. Firstly, the persistence of catenations and cohesion are regulated independently, cohesin by factors such as separase, hSgo1 and Plk1, and catenations by PIASγ and Topoisomerase II (Fig. 7). Secondly, catenations do not persist in metaphase simply because the sisters are held in close proximity by cohesin complexes.

Does the spindle assembly checkpoint monitor the ability to decatenate centromeric DNA?

The kinetics of progression through mitosis after PIASγ depletion revealed typical features of checkpoint activation: first, cells delayed progression to the next cell cycle stage (metaphase was prolonged compared to controls); second, the delay could be overridden spontaneously (as seen in the time-lapse movies) and by chemical treatment (Aurora B inhibition, Cdk inhibition); third, overriding the checkpoint revealed defects that the checkpoint was presumably aiming to avoid (anaphase chromatin bridges seen after chemical override, and laggard chromosomes and asynchronous anaphases after spontaneous override).

Our data indicate that without PIASγ, cells become blocked in mitosis due to activation of an Aurora B- and Mad2-dependent checkpoint. Previous studies of the role of Aurora B in spindle checkpoint control indicated that Aurora B helps to sense and/or correct mono-polarly orientated chromosomes and merotelically attached kinetochores. We have not determined the precise molecular event that triggers activation of the checkpoint after PIASγ depletion, but at a cytological level based on chromosome spreads, immuno-localization of CENP-E and live cell imaging, we could not detect a mitotic defect prior to metaphase chromosome alignment. That all of the chromosomes biorient at the metaphase plate in PIASγ-depleted cells, discounts the possibility that monopolar orientation of chromosomes and, possibly, merotelic attachments are the cause of checkpoint activation. The majority of HeLa cells depleted of PIASγ reached metaphase with kinetics indistinguishable from controls cells, then blocked in metaphase for a prolonged period (on average 2.5× the normal length of metaphase). In these metaphase cells filmed by time-lapse analysis, the typical movements of the metaphase plate occurred as in control cells, presumably indicating normal spindle and motor function. Indeed, proper alignment of all the chromosomes at the metaphase plate implies correct kinetochore, spindle and motor behavior as well as proper tension on each chromosome. The only molecular defects that we could provide evidence for in PIASγ-depleted cells were the inability to remove centromeric catenations and a lack of Topoisomerase II localization to centromere regions (and mitotic chromosome cores in some cells). It is therefore possible that the anaphase checkpoint is induced in response to persistent centromeric catenations or the inability to remove these catenations efficiently. Other studies lend support to this model in that inhibition of Topoisomerase II, the only enzyme that can decatenate DNA, prior to anaphase onset induces a metaphase checkpoint delay [26]–[28].

The separation of sister chromatids that occurs during mitosis in the Xenopus egg extract system also depends on PIASγ. But in this in vitro system, the lack of PIASγ did not result in a prolonged metaphase block, as we observed in HeLa cells depleted of PIASγ [15]. This difference may simply reflect a limitation of the egg extract system or could, more interestingly, mean that Xenopus embryonic-like cell cycles lack the ability to activate a checkpoint in response to PIASγ-depletion.

PIASγ might promote efficient decatenation at the centromere

Similar to PIASγ-depleted cells, Topo II inhibitor-treated cells block in metaphase with high levels of APC/C substrates such as securin and cyclin B [27]. An important difference between these two metaphase checkpoint arrested states, however, is that the Topo II inhibited cells are unable to separate their sister chromatids upon checkpoint bypass (due to the persistent catenations) whereas PIASγ-depleted cells can separate most of their sister chromatids upon inactivation of Aurora B or Cdk-cyclin B. It must therefore be the case that PIASγ function is not absolutely required for Topo II-mediated decatenation. We therefore suggest that PIASγ functions to promote efficient Topo II-mediated decatenation at the centromere, but that without PIASγ, decatenation can still occur but perhaps without the precision necessary for faithful chromosome segregation at the moment of anaphase onset. In agreement with this hypothesis, Topoisomerase II localized diffusely on the chromosomes of PIASγ-depleted cells rather than being concentrated at centromere regions and chromosome cores. It is noteworthy that although some of the PIASγ-depleted cells that we filmed by time lapse-microscopy could perform anaphase, we observed defects in the synchrony of sister separation; accordingly, PIASγ-depleted cells treated with Aurora B inhibitors could segregate their chromosomes, but not efficiently and chromosome bridges were always seen between the dividing cells. Together, these data indicate that PIASγ functions to promote faithful chromosome segregation, presumably by increasing the efficiency of decatenation, particularly at centromeres, but is dispensable for sister separation. It seems reasonable that human cells have evolved a mechanism to determine if decatenation cannot proceed efficiently, when Topo II activity is limited or in the absence of PIASγ, and to respond by delaying anaphase onset by activation of a checkpoint. Such a defect could be detected by monitoring the presence of catenations, the proper localization of Topoisomerase II or by assessing PIASγ activity.

In the experiments where we bypassed the prolonged metaphase delay seen in the absence of PIASγ, many of the PIASγ-depleted cells had been in metaphase for a considerable period of time before we added roscovitine or ZM447439. We can infer, then, that the catenations must have persisted despite the fact that Topoisomerase II is active in control metaphase cells. It is therefore plausible that sumoylation of Topoisomerase II helps to direct its decatenatory activity to centromeric catenations. This idea is consistent with the finding that PIASγ sumoylates Topoisomerase II in Xenopus and yeast. However, we have not ruled out the possibility that PIASγ promotes proper Topoisomerase II localization indirectly in human cells.

In summary, we have identified a human E3 sumo ligase required for efficient and timely anaphase separation of sister chromatids. Cells that lack PIASγ arrest in metaphase with cohered sisters even in the absence of the cohesin guardian hSgo1, and without cohesin Rad21 being detectable at centromeres. Topoisomerase II does not concentrate at centromere regions in such cells and catenations remain between the sister chromatids. Therefore, we provide the first evidence that the mechanism of loss of cohesion in human cells involves removal of DNA catenations stimulated by PIASγ. Moreover, our data indicate that catenations can act redundantly with cohesin complexes and that the failure to efficiently remove catenations triggers the spindle checkpoint.

Cell culture and siRNA methods

Cells were grown at 37°C with 5% CO₂ in DMEM containing high glucose, L-glutamine, sodium pyruvate and pyridosine hydrochloride (Gibco) plus Penicillin-Streptomycin (100 U/mL–100 µg/mL, Gibco). Cell synchronies were performed by double thymidine (2 mM) arrest and release into complete medium. The siRNA and synchrony protocol used for PIASγ depletion are depicted in Figure 1. PIASγ depleted using two different siRNA sequences gave similar results. Data presented here used PIASγ-B: 5′-GCUCUACGGAAAGUACUUA (dTdT)-3′. siRNAs were transfected using Lipofectamine2000 (Invitrogen) using the manufacturer instructions. 6 hours after siRNA transfection, medium was changed to normal medium containing half the normal concentration of Penicillin-Streptomycin. Nocodazole was used at 0.5 µM in DMSO.

Cytology

For cytological analysis cells were fixed with Carnoy's and chromosome spreads prepared or with paraformaldehyde for immuno-staining, as previously described [13]. Antibodies: anti-CENP-E, 1∶250 dilution (Immuquest); anti-myc, 1∶1000 dilution (Gramsch Laboratories); anti-Topoisomerase IIα, 1∶1000 dilution (CalBiochem). Rad21 was visualized in cells expressing myc-tagged Rad21 [49]. When quantifying cellular phenotypes a minimum of 1000 cells were counted per sample. Photomicrographs were acquired with a Zeiss Axioplan2 microscope, an alpha-Plan Fluar 100×/1.45 n.a. objective, and an AxioCam MRC5 camera with Axiovision software (Zeiss).

Biochemistry

Whole cell extracts were obtained and Western blots performed as previously described [13] using the following antibodies: anti-hRad21 (Abcam, 1∶1500), anti-hShugoshin (Adrian Salic and Tim Mitchison, Harvard Medical School, 1∶1000), anti-CyclinB1 (Abcam, 1∶1500), anti-phospho-H3 (Upstate, 1∶4000), anti-alpha-tubulin (Covance, 1∶1500), anti-Apc2 (Hongtao Yu, 1∶1000) and anti-Smc3 (Bethyl Laboratories, 1∶5000).

Live cell imaging

HeLa cells were plated onto poly-d-lysine coated 35 mm tissue culture dishes fitted with glass cover-slips (MatTek Cultureware). siRNA transfection and thymidine synchrony was performed as described in Fig. 1 except that upon release from the second thymidine arrest, the standard medium containing the thymidine was exchanged for DMEM without phenol red, supplemented with 10% FBS, penicillin/streptomycin and 200 µM Trolox (Calbiochem). The dishes were transferred to a microscope humidified stage incubator containing 5% CO₂ at 37°C, 4 hours after release from the early S-phase block. Cells were filmed at 120 second intervals with three z-sections for 16–20 hours, using a Zeiss Axiovert 200M microscope fitted with a 40×/1.3 n.a. Plan-Neofluar objective, an Axiocam HRm camera and using Openlab software.