ISWI is a RanGTP-dependent MAP required for chromosome segregation

In addition to their function as data storage devices, chromosomes have more recently been shown to play an important organizational role in the cell. In interphase they organize the nucleo-cytoplasmic transport, and during mitosis they drive spindle assembly in metaphase and nuclear envelope and nuclear pore complex assembly in telophase (Clarke and Zhang, 2008). In each case, the guanine nucleotide exchange factor for Ran (RCC1) is bound to chromosomes and the Ran-GTPase activating protein (RanGAP) is distributed throughout the cytoplasm. As a consequence, the GTP bound form of Ran (RanGTP) is locally produced inside the nucleus in interphase and in a gradient surrounding the chromosomes during mitosis (Hetzer et al., 2002). The RanGTP gradient induces spindle assembly through the local activation of several NLS-containing molecules (Kalab and Heald, 2008), examples being the microtubule (MT) nucleator TPX2 and the MT stabilization factor CDK11 (Gruss et al., 2001; Yokoyama et al., 2008). Both TPX2 and CDK11 are inhibited by the importin-α/β heterodimer in the mitotic cytoplasm and released from it when RanGTP binds to importin-β around chromosomes. Because all NLS proteins are released from importins locally around chromosomes in mitosis, they are potentially involved in spindle assembly or other chromosome-dependent processes. We have recently established an affinity purification method to isolate pure NLS proteins from Xenopus egg extracts, by optimizing the elution of NLS proteins from importin-β affinity column (Yokoyama et al., 2008). We are in the process of characterizing the functions of components in the NLS fraction.

Here, we report on one of these NLS proteins, ISWI, initially characterized as a chromatin-remodeling ATPase (Brown et al., 2007). We find that ISWI is a RanGTP-dependent microtubule-associated protein (MAP) in vitro. This protein is not required for spindle assembly, but is essential for chromosome segregation. This anaphase function involves a RanGTP-dependent stabilization of spindle MTs.

Previously, NLS-containing MAPs had been prepared by sequential purification of MAPs and importin-β–binding proteins, resulting in the identification of an NLS-MAP, Xnf7 (Maresca et al., 2005). Following this approach, the NLS fraction prepared from egg extracts (Yokoyama et al., 2008) was incubated with taxol-stabilized MTs and the bound proteins were eluted with high salt (Fig. 1 A). Only a very small proportion of the total NLS proteins bound to the MTs. Nevertheless, the bound fraction that was eluted by high salt still contained a considerable number of proteins (Fig. 1 A). Because ISWI is a nuclear protein that had been reported to bind MTs (Trachtulcová et al., 2000; Liska et al., 2004), we blotted the NLS-MAP fraction with a commercially available anti–human SNF2H antibody (a human orthologue of Xenopus ISWI), and indeed we found ISWI in the MAP fraction (Fig. 1 A). We further checked the behavior of ISWI throughout the fractionation of NLS proteins and MAPs (Fig. 1 B). ISWI was depleted from extracts with importin-β beads and efficiently eluted from MTs with high salt, indicating that ISWI is an NLS protein and a MAP.

We then prepared recombinant Xenopus ISWI (Fig. 1 C) and examined its ability to bind MTs in a sedimentation assay. Purified ISWI clearly bound to taxol-stabilized MTs (Fig. 1 D). To determine the binding affinity of ISWI for MTs, ISWI (0.50 µM) was sedimented with increasing concentrations of MTs (Fig. S1 A). All of the ISWI was bound to MTs when the MT/ISWI molar ratio was higher than 4. The dissociation constant (Kd) of ISWI at 0.50 µM was calculated to be ∼0.28 µM (the tubulin concentration required to bind 50% of ISWI; Fig. S1 B), indicating that the affinity is relatively high. In addition, ISWI bound to MTs at 1:1 molar ratio under ISWI saturation (Fig. S1 A). This binding was analyzed by electron microscopy showing that ISWI decorates the outer surface of MTs apparently in an organized way (Fig. S1 C).

We then examined whether the MT binding of ISWI was regulated by RanGTP and importins. Addition of importin-β weakly inhibited the binding of ISWI to MTs, whereas importin-α did not (Fig. 1 E). The importin-α/β heterodimer, however, had a significant inhibitory effect on the binding of ISWI to MTs that was reversed by RanQ69L (a mutant of Ran that cannot hydrolyze GTP; Fig. 1 E). Moreover, when we used an ED mutant of importin-α that cannot bind NLS sequences (Gruss et al., 2001), the importin-β/ED heterodimer did not inhibit the MT binding (Fig. 1 E). This demonstrates that importin-α/β binds to ISWI via an NLS and that this binding inhibits ISWI–MT interaction.

The identification of ISWI as a MAP suggested that it might bundle MTs. Indeed, recombinant ISWI strongly bundled taxol-stabilized MTs (Fig. 2 A).

Moreover, recombinant ISWI triggered microtubule assembly when incubated with pure tubulin. At 2.5 µM tubulin, ISWI induced the formation of MT aggregates containing many short MTs (Fig. 2 B). Electron microscopy confirmed that those structures are composed of a number of aster-like structures of MTs (Fig. 2 C). At higher tubulin concentrations (20 µM), ISWI induced the assembly of longer MTs that also frequently formed aster-like structures (Fig. 2 B).

We therefore examined whether ISWI could nucleate MTs. Incubation of GFP-ISWI alone produced some ISWI aggregates (dots; Fig. S2, A–C). In the presence of 2.5 µM tubulin, however, the number of dots increased ∼10 times and they formed MT asters (Fig. S2, B and C). GFP-ISWI and Cy3-tubulin showed the same localization on the dots and MTs. Furthermore, recombinant ISWI-coated beads induced MT growth in pure tubulin solutions, proving that ISWI can nucleate MTs in vitro (Fig. S2 D).

Because ISWI is a RanGTP-dependent MAP (Fig. 1 E), it was possible that RanGTP could also regulate the MT assembly by ISWI. To examine this, we immunoprecipitated ISWI from a Xenopus cytostatic factor–arrested metaphase egg extract (CSF extract) using antibodies against full-length ISWI (Fig. 3 A) and tested the MT promoting activity of the immunoprecipitates in the presence and absence of RanQ69L (Fig. 2 D). ISWI immunoprecipitates did not induce MT assembly in 20-µM tubulin solutions, but they did upon addition of RanQ69L (Fig. 2 D). This suggested that the importin-α/β heterodimer might coprecipitate with ISWI. It had been reported that ISWI is coimmunoprecipitated with several polypeptides from Xenopus CSF extracts (MacCallum et al., 2002), and our immunoprecipitates showed similar band patterns (Fig. 2 E). Indeed, we identified importin-α and -β in the immunoprecipitates by immunoblot (Fig. 2 E). The interactions are specific because importin-α and -β did not coprecipitate with ISWI from the extracts supplemented with RanQ69L (unpublished data). These results indicated that ISWI nucleates, stabilizes, and bundles MTs in vitro, and the activities are up-regulated by RanGTP.

ISWI is an ATPase, the activity of which is essential for chromatin remodeling (Corona et al., 1999). To examine whether the ATPase activity of ISWI was important for its effect on MTs, we prepared a Xenopus ISWI mutant in which lysine 204 was replaced with arginine (K204R; Fig. S3 A). This residue was located in a conserved ATP-binding site (GXGKT) and the corresponding mutation in Drosophila ISWI abolished its ATPase activity as well as nucleosome remodeling (Corona et al., 1999). In fact, the K204R mutant did not show ATPase activity (Fig. S3 B). Nevertheless, the K204R mutant bound and assembled MTs with the same efficiency as the wild-type protein (Fig. S3, C and D). Moreover, we examined whether the addition of ATP or AMPPNP affected the MT binding or polymerization activities of wild-type ISWI, but it did not affect either activity. These results indicated that the MT-regulating activity of ISWI does not require its ATPase activity in contrast to the chromatin-remodeling activity.

To examine the subcellular localization of endogenous ISWI, we prepared three independent ISWI antibodies that specifically recognize ISWI in Xenopus XL177 cell lysates (Fig. 3 A) and performed immunofluorescence staining of the cells using a mixture of all three antibodies (Fig. 3 B). In interphase, ISWI was localized in nuclei (Fig. 3 B). In prometaphse, after nuclear envelope break down, ISWI signal on the centrosomes increased, and some ISWI stayed on chromosomes while the rest was distributed throughout the mitotic cytoplasm. In metaphase the centrosome staining further increased and ISWI also became visible on spindle MTs (Fig. 3 B). In anaphase, ISWI remained localized on spindle poles and MTs, and both signals started to decrease at late anaphase (Fig. 4 B; unpublished data). The centrosomal signal decreased further in telophase while ISWI reentered the newly formed nuclei.

We also added GFP-ISWI to Xenopus egg extracts to examine its localization during the cell cycle (Fig. 3 C). In interphase extracts, GFP-ISWI was localized inside the nucleus (Fig. 3 C), as expected. The localization of ISWI drastically changed during progression toward mitosis. During the early stages of spindle assembly, GFP-ISWI was detected on chromosomes and along the length of spindle MTs, while none was visualized at the formed spindle pole (Fig. 3 C, 10 min; Yokoyama et al., 2008). The majority of GFP-ISWI then became localized at spindle poles in metaphase, while some were still detected on spindle MTs and chromosomes (Fig. 3 C). In anaphase, GFP-ISWI remained bound to them during chromosome segregation, although ISWI localization on spindle MTs apparently increased (Fig. 3 C). Immunofluorescence staining of egg extract spindles showed the same localization (Fig. 3 D).

Our results in culture cells and egg extracts are consistent with previous reports showing that the majority of ISWI dissociates from chromatin during the cell cycle progression from interphase to mitosis (Demeret et al., 2002; MacCallum et al., 2002). Thus, ISWI binds spindle MTs in mitosis, whereas it stays inside the nucleus during interphase and therefore does not bind cytoplasmic MTs.

We then tested whether RanGTP could regulate the interaction of ISWI with spindle poles and MTs. To address this possibility, we examined the localization of ISWI protein on RanQ69L-induced pseudo spindles and asters (Carazo-Salas et al., 1999). Both GFP-ISWI addition and immunofluorescence staining showed that ISWI localizes on those structures (Fig. 4, A and B). Moreover, centrosomal asters in CSF extracts did not incorporate GFP-ISWI, but they did when stabilized to form bigger asters in the presence of RanQ69L (Fig. 4 C; Carazo-Salas et al., 2001; Yokoyama et al., 2008). Thus, ISWI binds to spindle poles and MTs in the absence of chromatin, most likely in a RanGTP-dependent manner.

We then examined the role of ISWI in cell cycle progression using Xenopus egg extracts. Previous work had shown that the depletion of ISWI from egg extracts did not affect DNA replication, chromosome condensation, sister chromatid cohesion, or chromatin decondensation (Demeret et al., 2002; MacCallum et al., 2002), but the role of ISWI in MT organization had not been examined. The addition of the full-length ISWI antibody to interphasic extracts did not inhibit the decondensation of sperm chromatin during nuclear assembly (Fig. 5 A). In metaphase extracts, spindles also assembled normally in the presence of the antibody with condensed chromosomes properly aligned on the metaphase plate (Fig. 5 A). However, subsequent chromosome segregation was strongly inhibited. 10 min after calcium addition, spindle shapes still remained, but the amount of MTs was clearly decreased compared with control extracts (Fig. 5, A and B). At 20 min, spindles started to disappear and chromosomes remained clustered in one single structure, whereas they were already segregated in control spindles (Fig. 5, A– C). After 30–40 min, MTs disappeared completely and chromosomes did not segregate, whereas in the control extracts MTs were still present around the segregating chromosomes (Fig. 5, A and B). In the presence of ISWI antibodies, the nonsegregated chromosomes still decondensed after 1 h, as in control extracts (Fig. 5 A). In both extracts, inactivation of histone H1 kinase activity proceeded with identical timing after calcium addition (Fig. 5 D), indicating that cell cycle progression was not affected by the ISWI antibody addition. Similar results were obtained when antibodies were added just before inducing metaphase or anaphase and in extracts from which ISWI was immunodepleted (Fig. S4, A and B).

In those experiments, calcium was used to induce the transition from metaphase to anaphase. Because calcium is also known to depolymerize MTs, we wondered whether the observed phenotype reflected anaphase onset or a general MT destabilization caused by calcium. To address this, we induced anaphase by adding calcium/calmodulin-dependent kinase II (CaMKII) to the metaphase-arrested extracts (Hannak and Heald, 2006; Reber et al., 2008). In the presence of ISWI antibodies, spindle MTs still disappeared at anaphase onset and chromosomes did not segregate (Fig. S4, C and D). Thus, ISWI is required for the stability of anaphase spindle MTs and chromosome segregation, but not for chromosome condensation, spindle assembly, chromosome alignment on the metaphase plate, or chromosome decondensation.

Because ISWI is a chromatin-remodeling complex protein, it might have been involved in some aspect of chromosome separation in anaphase. To address this possibility, we examined the effect of ISWI inactivation on spindles assembled around plasmid DNA–coated beads (Heald et al., 1996). The DNA bead spindles do not have kinetochores and do not segregate in anaphase. When anaphase was induced by calcium addition to control extracts, spindles remained intact over 20 min before dissociating without performing anaphase elongation movements (Fig. 6 C). In ISWI-depleted extracts (Fig. 6 A), spindles assembled properly in metaphase, but they disappeared rapidly 5 min after calcium addition (Fig. 6, C and D). Spindle MTs seemed to depolymerize evenly rather than at any specific site (Fig. 6 C). Addition of recombinant ISWI to the depleted extracts rescued the phenotype (Fig. 6, C and D). In all three extracts, histone H1 kinase activity decreased with exactly the same timing (Fig. 6 E). Thus, in ISWI-depleted extracts, MTs depolymerize at anaphase onset independently of kinetochore assembly and chromosome segregation.

These results indicated that in contrast to other known RanGTP-dependent MAPs (Kalab and Heald, 2008), ISWI is not essential for spindle assembly, but it is required to maintain MT stability during anaphase, and this is essential for chromosome segregation.

Because spindles assembled around sperm nuclei or DNA beads containing chromatin, we could not formally exclude the possibility that ISWI requirement in anaphase was mediated through its potential chromatin-remodeling activity. Therefore, we analyzed the stability of RanGTP spindles in anaphase, which do not contain chromatin. In control extracts, the spindle-like structures disappeared quickly, 5–10 min after calcium addition (Fig. S5), consistent with the previous report (Reber et al., 2008). The lifetime of RanGTP spindles (∼10 min) in anaphase was much shorter than that of sperm and DNA bead spindles (∼40 min; Fig. 5; Fig. 6; Fig. S5). The addition of excess recombinant ISWI (4 µM) did not affect the number and the shape of the RanGTP spindles in CSF extracts, but it maintained the spindle-like structures during anaphase (∼40 min; Fig. S5). The spindles were then disassembled as the extract state changed to interphase (Fig. S5 B). Smaller concentrations of ISWI (<1.4 µM) did not prolong the lifetime of Ran spindles. These results indicated that in anaphase ISWI stabilizes spindle MTs as a MAP, but this process does not require the chromatin-remodeling activity of ISWI.

The results reported above suggested that ISWI might regulate MT stability in anaphase in a RanGTP-dependent manner. To address this, we examined the stability of MTs nucleated by centrosomes during anaphase in the presence or absence of RanQ69L (Fig. 7). In control extracts in the absence of RanQ69L, astral MTs remained short and dynamic at 20 min after calcium addition, and then they changed into long MTs at 60 min as the interphase state was established (Fig. 7). RanQ69L clearly stabilized MTs in the anaphase extracts (20 min), although the stabilization effect was weaker than that in CSF extracts (Fig. 7). This result indicated that RanGTP stabilizes MTs in anaphase, whereas it does not nucleate MTs, as shown in Fig. S5. The addition of the ISWI antibody did not affect the RanGTP-dependent MT stailization in CSF extract, but specifically inhibited MT stabilization in anaphase (Fig. 7). Thus, RanGTP stabilizes MTs in anaphase by activating ISWI.

To examine the role of ISWI in vivo and in a different organism, we performed RNAi against ISWI in Drosophila S2 cells. We used these cells because Drosophila has one ISWI gene, whereas mammals have two. Three independent dsRNAs for ISWI were prepared according to the sequence information on the Genome RNAi database (Horn et al., 2007). Treatment of S2 cells with the dsRNAs efficiently down-regulated the protein level of ISWI after 2 d, in comparison to cells treated with control dsRNA for GFP (Fig. 8 A, day 4).

ISWI-depleted cells, treated with all three dsRNAs, grew slowly and eventually stopped growing (see Fig. 10 B). FACS analysis at 4 d did not show a difference between control and ISWI-depleted cells. After 12 d, however, ISWI-depleted cell populations contained more G2/M phase cells and a decreased level of S phase cells (Fig. 8 C; Table I). Cells with less than 2N chromosomes increased also, indicating the occurrence of apoptotic and/or aneuploid cells (Fig. 8 C; Table I).

To examine spindle morphology and chromosome segregation, the cells were immunostained for tubulin and DNA after 4 d (Fig. 8, D and E). In control cells, spindles elongated when chromosomes were segregated at anaphase (Fig. 8 D). In ISWI-depleted cells (MRC015_H12), we found mitotic cells that did not exist in the control. They were classified into four phenotypes and the proportion of each phenotype in the total mitotic cells was quantified (Fig. 8 E; Table II). Longer spindles were seen with chromosomes staying in the central region (phenotype 1; 6.7 ± 1.8%). Most frequently, we detected longer spindles with one or more chromosomes separated and located close to the spindle poles (phenotype 2; 17.9 ± 5.6%). In these two phenotypes, the morphology of the spindle MTs seemed to be normal. We also observed longer defective spindles or apparently shrinking spindles having chromosomes scattered through the spindle (phenotype 3; 2.9% ± 0.1%). Furthermore, anaphase or telophase spindles, judged by chromosome separation, were found with defective MTs and chromosomes lagging or missegregating (phenotype 4; 7.4 ± 6.5%). In phenotypes 3 and 4, MTs were often detached from the spindles together with chromosomes, but still connected to spindle poles (Fig. 8 E, arrows). Mitotic indices were not affected in ISWI-depleted cells for at least 4 d (Table II). The proportions of prophase, prometaphase, and metaphase in ISWI-depleted cells were similar to those in control cells, but the proportions of proper anaphase and telophase decreased in ISWI-depleted cells (Table II). The other two dsRNAs for ISWI (HFA07446 and OB2651) showed the same phenotypes with similar frequencies, indicating that the observed defects were caused by the lack of ISWI.

To address directly whether ISWI is required in anaphase, we conducted live-cell imaging of S2 cells stably expressing GFP–α-tubulin and mCherry-centromere identifier (CID, a kinetochore marker; Coelho et al., 2008) at 4 d after dsRNA treatment (Fig. 9; Videos 1–3). We found normal metaphase spindles in ISWI RNAi cells with similar frequency to control cells, consistent with our quantitation of fixed cells (Table II). When we filmed the metaphase cells to follow their cell cycle progression, however, the spindles in ISWI-depleted cells frequently disintegrated and chromosomes did not segregate during an imaging period of 1 h (7 out of 12 cells filmed; Fig. 9 B; Videos 2 and 3). Before the MTs disintegrated, sister chromatids separated and moved slightly poleward (Fig. 9 B; Videos 2 and 3), suggesting that the cells entered anaphase (Matos et al., 2009). Spindles then became longer until they reached the plasma membranes, and MTs were detached from spindles with chromosomes or started to shrink (Fig. 9 B;,Videos 2 and 3).

This demonstrates that ISWI is essential for Drosophila cells to grow and progress from G2/M phases, and is required for anaphase MT stability and chromosome segregation. Thus, the significant role of ISWI in anaphase is conserved in Drosophila.

We identified a known chromatin remodeling factor, ISWI, as a potential NLS-MAP by sequential affinity purification of NLS proteins and MAPs. In vitro, ISWI binds to pure MTs with high affinity and decorates the lattice of MTs with a likely stoichiometry of one ISWI per one tubulin dimer. The MT binding of ISWI is specifically inhibited by the importin-α/β heterodimer and the inhibition is reversed by RanGTP. ISWI bundles preexisting MTs and induces MT assembly from pure tubulin, often forming MT asters. The ISWI-dependent MT assembly is also positively regulated by RanGTP in vitro. Consistently, ISWI accumulates on spindle MTs during mitosis in Xenopus culture cells and egg extracts, and is required for RanGTP-dependent MT stabilization during anaphase in extracts. These results demonstrate that ISWI is a bona-fide nuclear MAP regulated by RanGTP and suggest that analysis of all the protein bands found in the NLS-MAP fraction will lead to the identification of additional MAPs that function locally around chromosomes during mitosis.

ISWI is one of the ATPases included in chromatin-remodeling complexes (Brown et al., 2007). The fundamental function of the complexes is to slide nucleosomes along the DNA in order to make specific DNA sequences accessible or inaccessible to regulators. Each complex contains one ATPase, and the ATPase activities are essential for the complexes to mobilize nucleosomes (Corona et al., 1999). Currently, the catalytic ATPase subunits fall into three families, the SWI/SNF family, the ISWI family, and the CHD family (Brown et al., 2007). The three families of ATPase subunits have distinct roles and exhibit homology restricted to the ATPase domain (Brown et al., 2007). Among them, only ISWI has been reported to bind MTs (Trachtulcová et al., 2000; Liska et al., 2004). In this study, we demonstrated that the ATPase activity of ISWI is not necessary for its MT binding and assembly activities in vitro, indicating that the chromatin-remodeling and MT-regulating activities are distinct. SANT and SLIDE domains specifically contained in ISWI could be important for the MT-regulating activity (Grüne et al., 2003).

Although the role of ISWI in chromatin remodeling has been established in vitro, the function of ISWI in vivo remained unclear. In Drosophila, null ISWI mutants die in late larval stages when the maternal contribution of ISWI is depleted (Deuring et al., 2000). In mammals, there are two ISWI genes (SNF2H and SNF2L). SNF2H is mainly expressed in proliferating cells, whereas SNF2L is expressed in differentiated cells (Lazzaro and Picketts, 2001). SNF2H-null mice are embryonic lethal (Stopka and Skoultchi, 2003). In Xenopus laevis, only one ISWI is known, which is expressed primarily during development and poorly detectable in adult differentiated tissues (Demeret et al., 2002). Down-regulation of the ISWI protein in Xenopus embryos inhibited their early development (Dirscherl et al., 2005). These results suggest that ISWI is essential for cell growth, differentiation, or viability, but do not clarify which cellular process ISWI is required for.

Previously in Xenopus egg extracts, immunodepletion of ISWI was shown not to affect nucleosome assembly, DNA replication, chromosome condensation, sister chromatid cohesion, or chromatin decondensation (Demeret et al., 2002; MacCallum et al., 2002). Consistently, we see no effect on these processes in our ISWI immunodepletion or antibody addition. However, in both types of experiment we observe a significant reduction of MT stabilization in anaphase associated with chromosome segregation defects. In addition, ISWI depletion in Drosophila cells results in MT and chromosome segregation defects in anaphase or telophase. And the depleted cells eventually stop growing, with accumulation in G2/M phases. Thus, the critical role of ISWI in anaphase is conserved in different organisms. The lethality observed in null ISWI mutants in Drosophila and mouse may be due to the lack of proper chromosome segregation, although this remains to be investigated further.

It has been reported in Xenopus egg extracts that the majority of ISWI dissociates from chromosomes when the cell cycle state of the extract is converted from interphase to mitosis (Demeret et al., 2002; MacCallum et al., 2002). In agreement with this observation, we show that ISWI binds predominantly to chromosomes and partially to MTs at the beginning of spindle assembly, whereas it mainly accumulates on spindle poles during metaphase. Thus, in the process of spindle assembly, ISWI seems to move from chromosomes to spindle microtubules and poles. These localization results suggest that ISWI may function as a chromatin-remodeling ATPase in interphase and as a MAP in mitosis (Fig. 10 A).

ISWI is a RanGTP-dependent MAP that localizes to spindle poles and MTs throughout mitosis. Depletion of other known RanGTP-dependent MAPs in extracts or cells results in defects in spindle assembly in metaphase (Kalab and Heald, 2008). However, in ISWI-depleted extracts and ISWI antibody-treated extracts, we only detect defects in MT stability and chromosome segregation during anaphase. In metaphase, spindles assemble with normal kinetics and morphology, and chromosomes are properly aligned on the metaphase plate. The down-regulation of ISWI in Drosophila cells also shows clear defects in anaphase but apparently not in prophase, prometaphase, and metaphase. These results therefore indicate that differently from other Ran-regulated MAPs, ISWI is specifically required during anaphase.

On the other hand, we find that RanGTP stabilizes centrosomal MTs in anaphase extracts while it does not nucleate MTs. This is interesting because in metaphase RanGTP induces both stabilization and nucleation of MTs that are essential for spindle assembly (Gruss et al., 2001; Yokoyama et al., 2008). ISWI is required for the RanGTP-dependent MT stabilization in anaphase but not in metaphase. In addition, recombinant ISWI stabilizes and maintains RanGTP spindles during anaphase. Importantly, both assays do not contain chromatin. These results demonstrated that ISWI stabilizes and maintains spindle MTs during anaphase in a RanGTP-dependent manner, and that this process does not require chromatin. The maintenance of spindle MTs in anaphase must be essential for chromosomes to be segregated along the MTs.

The above finding raises the question of why there is a specific requirement for ISWI in anaphase. Although this remains to be investigated, we propose two possibilities. First, ISWI may be activated throughout mitosis by RanGTP, but redundant with other MT stabilization proteins in early steps of mitosis. Indeed, some MAPs required for spindle assembly, like TPX2 and HURP, undergo degradation in anaphase (Hsu et al., 2004; Stewart and Fang, 2005). Because of this, ISWI function in anaphase may be no longer redundant, but essential, for the maintenance of spindle MT stability. The second possibility is that ISWI may not have any function in metaphase regardless of its binding to spindles and may only be activated in anaphase. In that case, ISWI could be regulated by another mechanism, such as an interacting protein or a modification of ISWI protein, in addition to RanGTP.

In summary, we have shown here that, in addition to its role as a chromatin-remodeling factor, ISWI functions as a RanGTP-dependent MAP required for MT stability in anaphase (Fig. 10 A). Moreover, RanGTP induces spindle assembly at the beginning of mitosis as well as nuclear envelope and pore complex formation at the end of mitosis (Fig. 10 B; Hetzer et al., 2002; Clarke and Zhang, 2008). The present results demonstrate a novel distinct role for RanGTP in chromosome segregation during anaphase, extending the function of chromosomes in mitosis to controlling their own segregation (Fig. 10 B).

Xenopus cytostatic factor–arrested metaphase egg extracts (CSF extracts) were prepared as described previously (Murray, 1991). In brief, Xenopus eggs were dejellied by cysteine treatment, washed with XB buffer (10 mM K-Hepes, 100 mM KCl, 1 mM MgCl₂, 0.1 mM CaCl₂, and 50 mM sucrose, pH 7.7) and subsequently CSF-XB buffer (10 mM K-Hepes, 100 mM KCl, 3 mM MgCl₂, 0.1 mM CaCl₂, 50 mM sucrose, and 5 mM EGTA, pH 7.7), and crushed by centrifugation at 20,000 g for 20 min at 16°C. The straw-colored middle layer was recovered as a CSF extract. DNA bead preparation, cycled spindle assembly, anaphase progression, immunodepletion, and antibody addition were performed as described previously (Hannak and Heald, 2006). In brief, endogenous ISWI was depleted from CSF extracts by 3- or 4-time incubations of 60% (vol/vol) Dynabeads Protein A coupled with anti–full-length or N-terminal ISWI antibodies. Recombinant ISWI was added back to the depleted CSF extracts before sending to interphase. In antibody addition experiments, rabbit IgG (Sigma-Aldrich) or anti–full-length ISWI antibodies were added to CSF extracts at 0.37 mg/ml before sending to interphase. The reactions were squashed or spun down on coverslips, as described previously (Hannak and Heald, 2006). MT intensity around sperm or DNA beads was quantified using a macro written in MatLab (The MathWorks). Histone H1 kinase activity was assayed as described previously (Félix et al., 1990). Constitutively active CaMKII was added as described previously (Reber et al., 2008).

NLS proteins were isolated from Xenopus CSF extracts as described previously (Yokoyama et al., 2008). In brief, CSF extracts (13.4 ml) were incubated with a 40% wet bead volume (5.36 ml) of z-RanQ69L beads at 4°C for 1 h. The beads were removed by centrifugation, and the extract (activated extract) was incubated with a 120% wet bead volume (16 ml) of GST–importin-β beads at 4°C for 1 h. The beads were separated from the extract (depleted extract) by centrifugation, washed five times with a wash buffer (CSF-XB, 100 mM NaF, 80 mM β-glycerophosphate, 0.1 mM sodium vanadate, 0.03% digitonin, and 1 mM DTT), and incubated with a 75% volume (10 ml) of an elution buffer (20 µM RanQ69L-GTP, 500 mM NaCl, 1 mM GTP, 1 mM ATP, and 10% glycerol in the wash buffer) at 4°C overnight. The eluate (the NLS protein fraction) was dialyzed to CSF-XB supplemented with 10% glycerol and 1 mM DTT.

The NLS protein fraction (10 ml) was incubated with 2 µM taxol-stabilized MTs in the presence of 1 mM GTP and 10 µM taxol at 20°C for 15 min, and the MTs were spun down at 100,000 g for 10 min at 20°C through a 40% glycerol cushion in CSF-XB containing 20 µM taxol. The MT pellet was resuspended with a wash buffer (CSF-XB containing 1 mM DTT, 1 mM GTP, and 20 µM taxol) and spun down again through the cushion. The washing was repeated one more time. The MT pellet was incubated with 500 µl of 500 mM NaCl in the wash buffer for 10 min at room temperature, and the MAPs were recovered in the supernatant after centrifugation at 100,000 g for 10 min at 20°C.

A cDNA clone (IMAGE:6317687, RZPD) covering the complete Xenopus ISWI cDNA was amplified by PCR and cloned into EheI–HindIII sites of pFastBac HTa (Invitrogen). Baculoviruses were prepared following the manufacturer’s instructions. Recombinant ISWI was expressed in Sf21 insect cells, and purified on Talon beads (BD) and a Mono S column (GE Healthcare). GFP-ISWI was prepared as described for ISWI. The K204R mutant of ISWI, in which lysine 204 was replaced with arginine, was constructed by site-directed mutagenesis using PfuTurbo DNA polymerase (Agilent Technologies), and prepared as described for ISWI. N-terminal ISWI (1–400 aa) and C-terminal ISWI (746–1046 aa) were amplified by PCR and cloned into BamHI–XhoI or SmaI–XhoI sites of pGEX-6P-1 (GE Healthcare), respectively. The GST fusion proteins were expressed in bacteria and purified with Glutathione Sepharose (GE Healthcare).

Histidine-tagged RanQ69L-GTP, importin-β, importin-α, and the ED mutant of importin-α were prepared as described previously (Yokoyama et al., 2008).

Rabbit polyclonal antibodies against Xenopus full-length ISWI (1–1046 aa), N-terminal ISWI (1–400 aa), and C-terminal ISWI (746–1046 aa) antibodies were prepared and purified against recombinant proteins according to a standard protocol. A rabbit polyclonal antibody against hSNF2H was purchased from Santa Cruz Biotechnology, Inc.

MT sedimentation assay: recombinant ISWI was incubated with various concentrations of taxol-stabilized MTs at room temperature for 15 min in BRB80 buffer (80 mM K-Pipes, 1 mM MgCl₂, and 1 mM EGTA, pH 6.8) or CSF-XB buffer containing 1 mM GTP, 10 µM Taxol, and 0.05 mg/ml BSA. The samples were centrifuged at 100,000 g for 10 min at 20°C, and supernatant and pellet were analyzed by SDS-PAGE for Coomassie staining or immunoblot.

MT bundling assay: taxol-stabilized MTs (0.3 µM) incorporating Cy3-labeled tubulin (Hyman et al., 1991) were incubated with recombinant ISWI in BRB80 containing 10 µM taxol at room temperature for 20 min, squashed between slides and coverslips without fixation, and imaged immediately.

MT polymerization assay: recombinant ISWI was incubated at 37°C for 30 min with various concentrations of tubulin purified from pig brains (Castoldi and Popov, 2003) in BRB80 containing 1 mM GTP. The tubulin contained 10% Cy3-labeled tubulin. The samples were fixed with 0.25% glutaraldehyde, 10% glycerol, and 0.1% Triton X-100 in BRB80, and then spun down at 20,000 g for 12 min at 20°C onto 12-mm round coverslips through a 25% glycerol cushion in BRB80. The coverslips were post-fixed with cold methanol at −20°C for 10 min, washed with PBS, and mounted on slides.

MT nucleation assay: Talon Dynabeads (10 µl slurry; Invitrogen) were incubated with or without recombinant ISWI (8 µg) at 4°C for 30 min, and washed with PBS containing 0.01% Tween 20. The beads (0.1 µl slurry) were then incubated at 37°C for 30 min in 20 µl of BRB80 containing 20 µM tubulin (2 µM Cy3-tubulin) and 1 mM GTP. The samples were fixed and spun down as described in the MT polymerization assay.

After the assays, images were acquired using a Cellobserver microscope, a Plan-Apochromat 63x NA 1.4 oil objective lens, a Cy3 emission filter, an AxioCam MRm camera, and AxioVision software (all from Carl Zeiss, Inc.).

In vitro reactions as described above were spotted onto copper mesh grids, washed with water, and stained with 1.5% uranyl acetate. Images were acquired on an electron microscope (model CM-120; Biotwin).

Protein A Dynabeads (Invitrogen) were coupled with rabbit IgG or anti–full-length ISWI antibodies following the manufacturer’s instructions. Each bead sample (60 µl slurry) was incubated with CSF extracts (100 µl) at 4°C for 90 min, and washed two times with CSF-XB and two times with CSF-XB containing 0.1% Triton X-100. The immunoprecipitates were resolved by SDS-PAGE for Coomassie staining and Western blotting. To examine the MT polymerization activity of the immunoprecipitates, they were incubated at 37°C for 30 min with 20 µM pure tubulin (2 µM Cy3-tubulin) in BRB80 containing 1 mM GTP in the presence or absence of 15 µM RanQ69L, then fixed, and spun onto coverslips.

The assay was conducted as described previously (Corona et al., 1999).

Xenopus XL177 tissue culture cells were fixed in cold methanol at −20°C for 10 min, and stained with a mixture of full-length ISWI (1–1046 aa), N-terminal ISWI (1–400 aa), and C-terminal ISWI (746–1046 aa) antibodies (total 3 µg/ml of each 1 µg/ml) followed by Alexa 488–labeled anti–rabbit IgG (Invitrogen). Tubulin was stained with anti–α-tubulin DM 1A (Sigma-Aldrich) followed by Alexa 568–labeled anti–mouse IgG (Invitrogen). DNA was stained with Hoechst 33342. Images were acquired using a Cellobserver microscope, a Plan-Apochromat 63x NA 1.4 oil objective lens, an AxioCam MRm camera, and AxioVision software (all from Carl Zeiss, Inc.).

RanGTP-dependent MT nucleation assay: CSF extracts were incubated with 15 µM RanQ69L-GTP and 1 µM Cy3-tubulin at 20°C for 30 min. The samples (Ran spindles and asters) were fixed with 4% formaldehyde and 30% glycerol in BRB80, and spun down at 20,000 g for 12 min at 20°C onto 12-mm round coverslips through a 40% glycerol cushion in BRB80. The coverslips were post-fixed with cold methanol at −20°C for 10 min, and mounted on slides directly or immunostained before the mount.

RanGTP-dependent MT stabilization assay: performed as described previously (Yokoyama et al., 2008). In brief, CSF extracts were incubated with 1 µM Cy3-tubulin, 0.15 mg/ml anti-TPX2 antibody, and 2,000 centrosomes/µl in the presence or absence of 15 µM RanQ69L-GTP at 20°C for 30 min. The samples (centrosomal asters) were fixed, spun down, and mounted on slides. The MT length and intensity of centrosomal asters were quantified using a macro written in MatLab.

Drosophila Schneider S2 cells were cultured and RNAi was performed as described previously (Bettencourt-Dias and Goshima, 2009). In brief, dsRNAs against Drosophila ISWI (MRC015_H12, HFA07446, and OB2651) were prepared based on the sequence information from the GenomeRNAi database (Horn et al., 2007). Cells (10⁶) were incubated with 30 µg dsRNA at 25°C, and analyzed by cell counting and FACS. For immunofluorescence, the part of the cells was attached on coverslips, fixed, and stained with anti–α-tubulin DM 1A and anti–phospho-histone H3 (Millipore) followed by Alexa 568–labeled anti–mouse IgG and Alexa 488–labeled anti–rabbit IgG. DNA was stained with Hoechst 33342. Images were acquired at 0.25-µm z steps using a confocal microscope (model SP5; Leica) equipped with a 63x Plan-Apochromat NA 1.4 oil-immersion objective lens. They were subsequently processed using ImageJ (National Institutes of Health, Bethesda, MD).

Live-cell imaging was performed using S2 cells stably expressing GFP–α-tubulin and CID-mCherry (Coelho et al., 2008) after dsRNA treatment. 4D datasets covering the entire cell volume were collected every 1.5 min with 1-µm z steps at 25°C using a spinning disc confocal microscope (Ultraview Vox; PerkinElmer) equipped with an EC Plan-Neofluar 100x/1.3 oil-immersion objective lens. They were processed using ImageJ.

Fig. S1 shows that ISWI binds to MTs with high affinity. Fig. S2 shows that ISWI nucleates MTs in vitro. Fig. S3 shows that the ATPase activity of ISWI is not necessary for MT regulation. Fig. S4 shows that in ISWI-depleted extracts, spindle MTs disappear at anaphase onset and chromosomes do not segregate. Fig. S5 shows that excess ISWI stabilizes RanGTP spindles during anaphase. Video 1 shows a control S2 cell expressing GFP-tubulin and CID-mCherry. Videos 2 and 3 show S2 cells expressing GFP-tubulin and CID-mCherry after dsRNA treatment for ISWI.

We thank H. Maiato for providing an S2 cell line, A. Perez-Gonzalez for helping FACS analysis, O. Gruss for recombinant active CaMKII, F. Nedelec for writing macros, P. Bieling for the histidine-tagged GFP protein plasmid, K. Crnokic for maintaining frogs, and M. Bettencourt-Dias for advice regarding RNAi experiments. We also thank the Advanced Light Microscopy Facility in EMBL.

H. Yokoyama was supported by the EMBL postdoctoral fellowship.