Rho-dependent transfer of Citron-kinase to the cleavage furrow of dividing cells

Cytokinesis is the final step in cell division in which a parent cell is divided into two daughter cells. After segregation of chromosomes to the opposite poles in anaphase, a cleavage furrow is formed around the equator of a dividing cell, which deepens in telophase finally to separate two daughter cells. In classical experiments using fertilized eggs of sea urchin or newt eggs, a ring composed of actomyosin is observed beneath the cleavage furrow and it is suggested that the constriction of this ring leads to the cleavage of the cell. Indeed, disruption of this ring with F-actin-depolymerizing compounds results in failure of cytokinesis. However, how cytokinesis is temporally linked with nuclear division and how the cytokinetic apparatus is constructed spatially in a dividing cell remain largely unknown (Satterwhite and Pollard, 1992; Fishkind and Wang, 1995; Glotzer, 1997; Hales et al., 1999; Robinson and Spudich, 2000). The small GTPase Rho is suggested as a crucial regulator in these processes of cytokinesis. For example, inactivation of Rho with botulinum C3 exoenzyme prevents fertilized eggs of sea urchin or Xenopus embryos from entering into cytokinesis after nuclear division and produces multinucleate cells (Kishi et al., 1993; Mabuchi et al., 1993; Drechsel et al., 1997). Furthermore, injection of C3 exoenzyme into cells undergoing cytokinesis causes dissolution of the contractile ring and regression of the cleavage furrow, and the cells cannot continue cytokinesis (Mabuchi et al., 1993). These results strongly suggest that Rho is activated during cell division and works as a switch to induce and maintain the cytokinetic apparatus. Recently, a putative activator of Rho in this process has been identified. This GDP-GTP exchanger for Rho, Pebble in Drosophila and ECT-2 in mammalian cells, is found to be activated after the nuclear division (Prokopenko et al., 1999; Tatsumoto et al., 1999). Indeed, the GTP-bound active form of Rho accumulates during division of HeLa cells and this accumulation was abolished by expression of a dominant negative form of ECT-2, resulting in formation of multinucleate cells (Kimura et al., 2000). Thus, Rho appears to induce cytokinesis by organizing the cytokinetic apparatus. However, how Rho exerts this action has not been fully elucidated. Because of the crucial role of the actin cytoskeleton in cytokinesis, many studies have been carried out to examine the distribution and behavior during cell division of actin itself and several actin-binding proteins such as myosin, profilin and cofilin (Robinson and Spudich, 2000). However, their relation to the Rho signaling has not been clarified. Rho acts on downstream effectors to elicit its actions. They include the ROCK/ROK/Rho-kinase family of protein kinases, protein kinase PKN, Citron and Citron-kinase (Citron-K) and adapter proteins such as mDia, rhophilin and rhotekin (Narumiya, 1996). Among these molecules, ROCK, mDia and Citron-K are found to localize to the cytokinetic apparatus (Madaule et al., 1998; Kosako et al., 1999; Kato et al., 2001). ROCK is found to accumulate in the cleavage furrow and is proposed to be involved in elicitation of the myosin-based contractility and in disassembly of intermediate filaments during division (Kosako et al., 2000). Involvement of mDia in cytokinesis has been suggested by cytokinesis defect in Drosophila diaphanous mutants as well as induction of cytokinesis failure by microinjecting anti-mDia antibody to cultured mammalian cells (Castrillon et al., 1994; Tominaga et al., 2000). mDia contains the polyproline-rich FH1 region that binds profilin, and is suggested to induce actin polymerization through this interaction (Watanabe et al., 1997). Citron is present both as N-terminally truncated nonkinase isoforms and as an N-terminally extended kinase isoform; the former is expressed in a rather limited way in the neuronal tissues but the latter is ubiquitously expressed in various tissues and cells (Madaule et al., 1995; Madaule et al., 1998). We previously found that Citron-K accumulates in the cleavage furrow in dividing cells and persists in the midbody between divided cells. It was also demonstrated that overexpression of Citron-K deletion mutants causes cytokinesis defect in cultured mammalian cells, indicating that Citron-K plays also an important role in cytokinesis (Madaule et al., 1998). These results strongly suggest that Rho mobilizes several downstream effectors to execute its function in cytokinesis. However, the molecular mechanism through which Rho mobilizes these effectors has not yet been clarified. In the present study, we have taken Citron-K as an example and analyzed how activated Rho mobilizes this effector to accumulate in the cleavage furrow. Citron-K is particularly interesting in this respect, because previous studies suggest that Citron-K acts only in cytokinesis (Madaule et al., 1998; Kosako et al., 2000) and a recent study showed that disruption of its gene results in cytokinesis defect in vivo (Di Cunto et al., 2000).

HeLa cells were grown in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal calf serum (FCS) at 37°C with an atmosphere containing 5% CO₂. We enriched mitotic cells by synchronizing the growth of cells with the thymidine block method. Briefly, cells were incubated with 10 mM thymidine in DMEM containing 10% FCS for 16 hours at 37°C. The cells were then washed with Dulbecco’s phosphate-buffered saline (PBS[−]) three times and cultured in DMEM with 10% FCS at 37°C for 11 hours. For disruption of microtubules, cells were incubated with 500 ng/ml nocodazole (Wako Pure Chemicals, Osaka, Japan) at 37°C for 45 minutes. For disruption of F-actin, cells were incubated either with 2 μM cytochalasin D (Wako Pure Chemicals) at 37°C for 30 minutes, or with 2 μg/ml latrunculin A (Wako Pure Chemicals) at 37°C for 1 hour. For inhibition of ROCK kinase, cells were incubated with 10 μM Y-27632, a specific ROCK inhibitor (Uehata et al., 1997), for 30 minutes at 37°C. Treatment with C3 exoenzyme was performed as described previously (Kato et al., 2001). Briefly, C3 exoenzyme was prepared as described previously (Morii and Narumiya, 1995) and was electroporated into HeLa cells 6 hours after the release from the thymidine block. The cells were plated on glass coverslips and cultured in DMEM supplemented with 10% FCS. After 6-16 hours culture, cells were subjected to fixation and stained.

pCAG-myc-Citron-K has been previously described (Madaule et al., 1998). Green Fluoresecent Protein (GFP)-tagged Citron-K was produced by subcloning the inserts in pEGFP-C1 (Clontech) using BsiWI and NotI restriction sites newly created in this plasmid, and the construct was confirmed by nucleotide sequencing. Sources of pEXV-myc-Val14RhoA, pCMV-myc-Asn19RhoA, pCMV-myc-Asn17Cdc42, pEXV-myc-Val12Rac1, pCMV5-FLAG-Asn17Rac1 were described previously (Ishizaki et al., 1997; Hirose et al., 1998; Kimura et al., 2000). pCMV-myc-Val12Cdc42 was kindly provided by M. Symons (Picower Institute for Medical Research, Manhasset, NY). For construction of pEGFP-RhoA, pEGFP-Asn19RhoA, and pEGFP-Val14RhoA, the respective constructs in pBTM (Watanabe et al., 1997) were digested with BamHI and EcoRI and the resulting fragments were inserted into the BglII and EcoRI sites of pEGFP-C1. Transfection of these plasmids to HeLa cells were performed using Lipofectamine Plus (Gibco/BRL) in OPTI-MEM (Gibco/BRL) as described by Fujita et al. (Fujita et al., 2000).

HeLa cells grown on 20×20 mm glass coverslips were fixed with 4% formaldehyde in PBS[−] for 15 minutes at 4°C except in experiments shown in Fig. 4 and Fig. 8C, where the TCA fixation method by Hayashi et al. (Hayashi et al., 1999) was used. Fixed cells were washed with PBS-Tx (0.1% Triton X-100 in PBS[−]) several times. After blocking in PBS-Tx containing 1% bovine serum albumin (PBS-Tx-BSA) for 1 hour at room temperature, immunocytochemistry was performed with following antibodies and fluorescence reagents. The primary antibodies used were rabbit polyclonal anti-Citron antibody (Madaule et al., 1998), rabbit polyclonal anti-Nedd 5 antibody (Kinoshita et al., 1997), mouse monoclonal anti-β-tubulin antibody (clone TUB 2.1, Sigma), mouse monoclonal anti-c-Myc antibody (9E10, Santa Cruz), rabbit polyclonal anti-c-Myc antibody (A-14, Santa Cruz), rabbit polyclonal anti-FLAG antibody (D8, Santa Cruz), mouse monoclonal anti-RhoA antibody (26C4, Santa Cruz) and rabbit polyclonal anti-RhoA antibody (119, Santa Cruz). The primary antibodies were added at 1:200 dilution in PBS-Tx-BSA and incubation was carried out at room temperature for 45 minutes. The cells were then washed with PBS-Tx several times, and incubated with Texas Red-X phalloidin (Molecular Probes), TOPRO3 (Molecular Probes) or 4,6-diamidino-2-phenylindole (DAPI) (Molecular Probes), and/or with following secondary antibodies; FITC-conjugated donkey anti-mouse IgG (Jackson Immuno Research), Texas Red-conjugated donkey anti-mouse IgG (Jackson Immuno Research), Cy5-conjugated donkey anti-mouse IgG (Jackson Immuno Research), FITC-conjugated donkey anti-rabbit IgG (Jackson Immuno Research), Texas Red-conjugated donkey anti-rabbit IgG (Jackson Immuno Research) and Cy5-conjugated donkey anti-rabbit IgG (Jackson Immuno Research). For blocking anti-Citron antibody, His-tagged antigenic Citron peptide (amino acid residues 674-870 of Citron-N) was prepared as described previously (Madaule et al., 1995) and added at 1 μg/ml to the incubation with anti-Citron antibody. Fluorescence images were acquired by an MRC 1024 laser-scanning confocal microscope imaging system (Bio-Rad) equipped with a Zeiss Axiovert 100TV microscope.

HeLa cells expressing GFP-Citron-K were identified on CELLocate coverslips (Eppendorf) using an Olympus IX70 fluorescence microscope equipped with a cooled CCD camera (SenSys 0400, 768X512 pixels; Photometrics) and their phase-contrast and fluorescence images were recorded together with their location information. Cells were then fixed with 2.5% glutaraldehyde, 0.2% tannic acid and 0.05% saponin in 0.1M cacodylate buffer, pH 7.4, for 1 hour at room temperature. After washing with 0.1 M cacodylate buffer (pH 7.4) three times (5 minutes each), cells were postfixed with ice-cold 1% OsO₄ in the same buffer for 45 minutes. The samples were rinsed with distilled water, stained with 0.5% aqueous uranyl acetate for 2 hours at room temperature, dehydrated with ethanol and embedded in Epon 812. Prior to ultra-thin sectioning, the CELLocate coverslip was detached from the Epon block. The location information of the coverslip was transferred onto the surface of the block, which enabled the identification of the GFP-Citron-K-expressing cells based on the recorded images. Ultra-thin sections of the GFP-Citron-K-expressing cells were cut, doubly stained with uranyl acetate and lead citrate and viewed with a JEM 1010 transmission electron microscope (JEOL).

We previously demonstrated that Citron-K is enriched in the cleavage furrow of mitotic cells during cytokinesis (Madaule et al., 1998). Because it is an effector of the small GTPase Rho, the enrichment of Citron-K in the cleavage furrow is presumed to be carried out by the action of Rho. To understand the mechanism of Citron-K mobilization, we first used anti-Citron antibody and monitored the intracellular localization of endogenous Citron-K during the cell cycle. We also added recombinant Citron fragment containing the antigenic epitope to the incubation to identify specific signals. As shown in Fig. 1, although the Citron-K immunostaining is relatively weak, we could successfully identify specific signals by comparing immunofluorescence images in the absence and presence of the competing peptide. First, punctate signals were detected by the anti-Citron antibody and were abolished by the addition of the epitope peptide in the cytoplasm of interphase cells. In prometaphase, these specific Citron-K signals disintegrate and disperse diffusely in the cytoplasm. From anaphase to telophase, Citron-K accumulates in the cleavage furrow and finally to the midbody in the post-mitotic stage. These results demonstrate that Citron-K changes its localization during cell division from dot-like structures in interphase, to the cytoplasm in prometa- and metaphases and finally to the cortex of cleavage furrow in telophase.

To examine whether Rho is involved in this change of Citron-K localization, and if so, to identify a step regulated by Rho in this localization change, we electroporated C3 exoenzyme into HeLa cells enriched in S phase. We first examined whether the C3 exoenzyme treatment interfered with cytokinesis of HeLa cells, because a previous study by O’Connell et al. (O’Connell et al., 1999) showed that C3 exoenzyme injection into cultured NRK epithelial cells induced abnormal cortical activity and resulted in ectopic division. The C3 exoenzyme treatment of HeLa cells resulted in almost 100% production of binucleate cells 16 hours after the treatment (Fig. 2A), suggesting that Rho also plays a crucial role in regulation of cytokinesis of mammalian cells. Localization of Citron-K in these C3 exoenzyme-treated cells in various mitotic phases were then examined (Fig. 2B). C3 exoenzyme treatment, and consequently inactivation of Rho, did not affect dispersion of Citron-K into the cytoplasm in prometa- and metaphase. However, transfer of Citron-K to cleavage furrow in ana- to telophase was completely prevented by this treatment. Instead, Citron-K in the treated cells stayed in the spindle midzone. Co-staining with microtubules demonstrated that Citron-K associates with the central spindles in these cells. All of these signals appear to reflect the behavior of endogenous Citron-K, because they were abolished by the addition of the antigenic peptide.

To confirm these findings, we constructed GFP-tagged Citron-K and expressed it in HeLa cells. The GFP fusion protein showed the same pattern of phase-dependent change in the intracellular localization as endogenous Citron-K (Fig. 3A). GFP-Citron-K again shows punctate signals in interphase cells, becomes dispersed in the cytoplasm in prometaphase and concentrates in the cleavage furrow after anaphase. To clarify the identity of the punctate signals seen in interphase cells, GFP signals were examined with electron microscopy. As shown in Fig. 3B, they appeared as amorphous materials not enclosed with lipid bilayer, suggesting that they are protein aggregates and not vesicles. The GFP signals in the cleavage furrow in dividing cells often appear punctate, suggesting that aggregates of the overexpressed protein are not completely disassembled and migrate. When HeLa cells expressing GFP-Citron-K were treated with C3 exoenzyme, GFP signals were again seen in association with the midzone spindles in ana- and telophase cells (Fig. 3C). These results corroborate the above findings with endogenous Citron-K and suggest that the GFP-fusion of Citron-K can be used as a probe to monitor the behavior of the endogenous protein. Essentially the same localization was observed when Citron-K was expressed as a Myc-tagged protein. Expression of these recombinant Citron-K proteins did not interfere with cytokinesis.

The above results indicate that Rho catalyzes the transfer of Citron-K to the cell cortex in cleavage furrow, possibly from the midzone spindles. Because Citron binds to the GTP-bound, active form of Rho (Madaule et al., 1995), we wondered whether Rho and Citron-K colocalize in cleavage furrow. We therefore expressed GFP-Citron-K in HeLa cells and examined the colocalization of GFP-Citron-K and endogenous Rho in mitotic cells (Fig. 4). In interphase cells, Rho is present diffusely in the cytoplasm, whereas Citron-K are in dot-like structures as described. When cells undergo cytokinesis, Rho accumulates in cleavage furrow and stays to the midbody. This is consistent with a previous finding on the Rho localization in fertilized eggs of sea urchin (Nishimura et al., 1998). When both Rho and Citron-K were visualized in these cells, the Rho signal colocalizes with the Citron-K signal from the beginning of cytokinesis in the cleavage furrow to the midbody of postmitotic cells (Fig. 4B-E) suggesting, although not proving, that Citron-K makes a complex with Rho in this cytokinetic apparatus. Given that Citron binds only to active Rho (Madaule et al., 1995), these results suggest that Rho present in this complex is the GTP-bound active form. In addition to these structures in the cell cortex, a portion of overexpressed GFP-Citron-K remains as aggregates in the cytoplasm, where no colocalization with Rho was observed (Fig. 4B,C, arrowheads).

The above findings that inactivation of Rho interfered with the transfer of Citron-K to the cortex and that transferred Citron-K appeared to make a complex with Rho in cleavage furrow strongly suggest that active Rho binds Citron-K and they move together to the cortex. To test this hypothesis, we examined the effect of expression of Val14RhoA (a dominant active RhoA mutant) on the localization of overexpressed Citron-K in interphase cells. Myc-tagged Citron-K and GFP-tagged Val14RhoA were co-expressed in HeLa cells. Cells expressing both constructs were identified by GFP fluorescence and the Myc-tag staining, and the localization was examined by confocal microscopy. As shown in Fig. 5A, Citron-K forms band-like structures at the bottom of the cells co-expressing Val14RhoA. This band-like structure consists of numerous small patches in which the Myc-Citron-K signal and the GFP-Val14RhoA signal overlap completely. By contrast, no overlap of the Citron-K signals and GFP-Rho was found in remaining aggregates in the middle of the cells (Fig. 5A, middle slice). This was supported by examination in a vertical view of cells co-expressing Citron-K and Val14RhoA, which shows that the band-like structures were seen only on the cell cortex (Fig. 5B). Given that Citron binds directly GTP-Rho (Madaule et al., 1995), this colocalization of Citron-K and Val14RhoA in small patches in band-like structures suggests that activated Rho is present in these structures in a complex with Citron-K. Indeed, co-expression of either wild-type RhoA or dominant negative Asn19RhoA with Citron-K failed to induce the formation of the band-like structures, and Citron-K remains as aggregates in the cytoplasm (Fig. 5C). We also co-expressed Citron-K with other Rho-family small GTPases, Rac1 and Cdc42. Although Citron is able to bind to Rac1 in a yeast two hybrid system and in an in vitro overlay assay (Madaule et al., 1995), expression of neither dominant active nor dominant negative Rac1 (Val12Rac1 and Asn17Rac1, respectively) affected the localization of Citron-K in interphase cells (Fig. 5D). The localization of Citron-K was not affected either by expression of a dominant active or a dominant negative Cdc42 (Val12Cdc42 and Asn17Cdc42, respectively) (Fig. 5E), suggesting that Citron-K translocation to cell cortex to form band-like structures is specific to Rho activation. These results taken together suggest that the binding of Citron-K and Val14RhoA induces the transfers of Citron-K from the cytoplasm to the cell cortex. In these experiments, expression of Val14RhoA appeared to disintegrate large Citron-K aggregates in the cytoplasm. However, detailed inspection revealed that Citron-K is still present in small aggregates after transfer to the cortex, which are seen as small patches. Thus, disintegration observed in this overexpression system is probably not the same as dispersion of endogenous Citron-K seen in prometaphase of mitotic cells (see Discussion).

It is well known that Rho regulates actin cytoskeleton and that active Rho induces stress fibers (Hall, 1998). Given that Citron-K forms the band-like structure on Rho activation at the bottom of the cells, we wondered whether it has some connection with actin stress fibers. To this end, we cotransfected pEGFP-Citron-K and pEXV-myc-Val14RhoA into HeLa cells and subjected the cells for phalloidin staining. Covisualization of Citron-K and F-actin revealed that the band-like structure of Citron-K is present at the same plane as stress fibers, but that these two structures exclude each other as suggested by no signal overlap of GFP-Citron-K and F-actin (Fig. 6A-C). We were also interested in the relationship between Citron-K and a septin, because the latter molecule also accumulates in the cleavage furrow and is involved in cytokinesis (Field and Kellogg, 1999). However, when endogenous Nedd5, one of the septins, was stained in HeLa cells co-expressing GFP-Citron-K and Val14RhoA, it was associated with actin stress fibers, as previously observed (Kinoshita et al., 1997), and its signals and the GFP-Citron-K again excluded each other (data not shown).

We next examined the spatial relation between the Citron-containing structures and F-actin in mitotic cells. We studied this issue on both endogenous Citron-K (Fig. 7A,B) in HeLa cells and cells expressing GFP-Citron-K (Fig. 7C,D). Both studies demonstrated a clear separation of the structures containing Citron-K and the F-actin in the cleavage furrow. The Citron-containing structure appears to be concentrated beneath F-actin structure in the cleavage furrow by being encircled in all directions by F-actin during cytokinesis except in the earliest stage, where it was difficult to separate the two signals (Fig. 7A).

We next addressed the interaction between the Citron-enriched cortical structures and the cytoskeletons by disrupting either microtubules or F-actin. Nocodazole was used to depolymerize microtubules in cells co-expressing Citron-K and Val14RhoA. Depolymerization of microtubules did not affect the colocalization of Citron-K and active Rho, and the band-like structures of the Citron-K-Val14RhoA complex was maintained as those found in nontreated cells (data not shown), suggesting that the formation of the band-like structures does not depend on the integrity of microtubules. By contrast, when F-actin was disrupted either with cytochalasin D or latrunculin A treatment, the band-like structures were disintegrated, and Citron-K-containing small patches spread all over the ventral surface of the cell cortex (Fig. 8A). Because stress fibers are formed by virtue of actomyosin-based contractility that is exerted by the action of a Rho effector ROCK, we examined the effect of the ROCK inhibitor Y-27632 on this accumulation. Disruption of stress fibers by inactivation of ROCK resulted again in dispersion of the accumulation of Citron-containing patches (Fig. 8B). In some cells which have remaining stress fibers, some of the Citron-containing band-like structures were also conserved (Fig. 8B, right-hand cell). These results suggest that band-like arrangement of Citron-K-active Rho patches depends on orderly organization of F-actin by the action of the actomyosin system. We then examined the effect of F-actin depolymerization on the concentration of the Citron-containing structures in cleavage furrow (Fig. 8C). Mitotic cells were enriched by the use of the thymidine block and treated with either cytochalasin D or latrunculin A. In the dividing cells treated with either reagent, Citron-K was not seen as a ring-like structure in the cleavage furrow but dispersed as patches all around the cell cortex. However, the colocalization of Citron-K and endogenous Rho was spared in the dispersed patches. However, unlike Citron-K, a portion of Rho still remained at the original site of the cleavage furrow. Y-27632 was without effect on mitotic cells, which is consistent with the dispensable action of ROCK in cytokinesis (Madaule et al., 1998; Ishizaki et al., 2000).

In this study we monitored the localization of Citron-K during mitosis both by immunofluorescence study of endogenous protein with anti-Citron antibody and by expression of GFP-Citron-K fusion in HeLa cells. Both studies have revealed the multi-step change of Citron-K localization during mitosis. Citron-K is present as aggregates in interphase cells, disperses into the cytoplasm in prometaphase, translocates to the cell cortex in anaphase and accumulates in cleavage furrow in telophase. Although not all interphase cells contain visible aggregates in immunofluorescence, we think that Citron-K is present as aggregates also in these cells in a form not detected by this method. Oligomer formation has been reported for MRCK, a kinase homologous to Citron-K (Tan et al., 2001). Using C3 exoenzyme to inactivate Rho, we have found that the above sequential change is catalyzed by consecutive activation of a Rho-independent and a Rho-dependent mechanism. Thus, dispersion of Citron-K occurs normally in C3 exoenzyme-treated cells. However, it does not move to the cortex but stays in association with the midzone spindles in anaphase cells. These results indicate that Citron-K moves to the cortex via the midzone spindles in a Rho-dependent manner. This is an intriguing finding because, in mammalian cells, the cleavage signal is suggested to come from the central interdigitating spindle microtubules (Cao and Wang, 1996). Previously, several proteins have been reported to associate with the midzone spindles. They include TD-60, INCENP, aurora kinase and survivin and are collectively termed chromosomal passenger proteins (Andreassen et al., 1991; Adams et al., 2001; Skoufias et al., 2000). However, these proteins first locate at centromeres of chromosomes in metaphase, then associate with the midzone spindle extending to the cortex in anaphase, stay there in telophase and concentrate in the intracellular bridge after mitosis. However, we did not see any attachment of Citron-K to centromeres or chromosomes. It appears to bind to the midzone spindles from the cytoplasm and transfers to the cortex upon Rho activation. Thus, Citron-K is likely to be a novel type of passenger protein.

The above observation also indicates the presence of a Rho-independent, yet cell-cycle-dependent signal for the initial mobilization of Citron-K. At present, we do not know the identity of this signal. Cell-cycle-dependent mobilization was also reported for other proteins working in mitosis. For example, expression study showed that survivin is present as dots in the cytoplasm in interphase cells but concentrates in distinct spots on chromosomes in prophase (Skoufias et al., 2000). ECT-2, the Rho exchange protein involved in cytokinesis, is present in the interphase nucleus, disperses in the cytoplasm in prometaphase, concentrates in the spindle in metaphase and transfers to the cortex in anaphase (Tatsumoto et al., 1999). In the latter case, phosphorylation-dependent activation was suggested to occur.

As discussed, active Rho appears to transfer Citron-K to cell cortex and to concentrate it in cleavage furrow in telophase. We mimicked the transfer of Citron-K to the cortex by co-expression of Val14RhoA and Citron-K in interphase cells. Although the experiments in interphase cells naturally do not exactly simulate the process in dividing cells, the results obtained have provided many implications. In the latter experiment, the transferred proteins form band-like structures in the ventral cortex of cells. It should be mentioned that Citron-K is present still as small aggregates in these band-like structures, partly because this procedure skipped the natural dispersion process seen during mitosis and partly because of the high amount of overexpressed Citron-K. Because Citron-K binds to the active form of Rho selectively, this result suggests that the binding of Citron-K to active Rho takes Citron to cell cortex and cleavage furrow. A number of other Rho effector proteins including ROCK, mDia, Rhophillin, Rhotekin and PKN have been identified (Ishizaki et al., 1997; Watanabe et al., 1997; Watanabe et al., 1996; Reid et al., 1996). Although membrane translocation associated with activation of Rho has been reported on some of these effectors, such a marked translocation as that seen in Citron-K has never been observed. This is probably because other effectors are transiently translocated and used only in a small amount in response to local activation of Rho. However, transfer of Citron-K requires extensive and widely spread activation of Rho as induced by Val14RhoA expression in interphase cells. High accumulation of GTP-Rho during mitosis was already reported (Kimura et al., 2000). It should also be mentioned that almost all of endogenous Citron-K is transferred by Rho activation and accumulates in the cell cortex. At its accumulation site, Citron-K appears to make a complex with active Rho because colocalization of Citron-K and Rho is persistently observed in the band-like structures formed by co-expression with Val14Rho and in cleavage furrow. These results indicate that Rho in the active GTP-bound form serves as a structural component in the Citron-K-containing cytokinetic apparatus. We previously found that accumulation of GTP-Rho continues to be present during cytokinesis after the decline of the Rho exchange activity, and suggested the presence of a stabilization mechanism for GTP-Rho (Kimura et al., 2000). Our present finding is consistent with this suggestion. However, this is in contrast to the presumed activation mechanisms for other Rho effectors, in which active Rho transiently interacts with effectors. As shown by the previous study (Nishimura et al., 1998) and also shown here, most of Rho present in the cell accumulates in the cleavage furrow during cytokinesis. One mechanism of this Rho accumulation is the accumulation of Citron-K in this region. The translocation and accumulation of Citron-K appears to be linked to its function in cytokinesis. Di Cunto et al. (Di Cunto et al., 2000) disrupted selectively the gene for Citron-K in the kinase domain and showed that the kinase domain of Citron-K is crucial in cytokinesis. Thus, this study presents a new mode of stimulus-activated construction of a functional cytokinetic apparatus.

During cleavage of eggs of the echinoderms and Xenopus embryos, F-actin forms a distinct structure known as the contractile ring, which cooperates with myosin and constricts to divide the cell. The actin cytoskeleton also plays a crucial role in cytokinesis of mammalian cells, although the actin contractile ring is not so discernible in these cells. Because Rho is involved in reorganization of several types of the actin cytoskeleton such as stress fibers, we were interested in the relationship between the Citron-containing structures and the actin cytoskeleton. Unexpectedly, we have found that the Citron-K-containing structures and the actin cytoskeleton are present by excluding each other. In interphase cells co-expressing Citron-K and Val14Rho, the band structure that contains Citron-K is present at the same plane of the ventral cell cortex as stress fibers and the two exclude each other. In the cleavage furrow of mitotic cells, the Citron-K-containing structure is encircled by the F-actin. At the very bottom of the cleavage furrow, strong accumulation of Citron-K was observed but not of F-actin. Oegema et al. (Oegema et al., 2000) have reported the similar absence of F-actin in the center of the cleavage furrow in BHK-21 cells. Furthermore, we have found that disruption of F-actin with either cytochalasin D or latrunculin A abolished the Citron-K-containing structures and dispersed small patches containing Citron-K throughout the cell cortex. These results suggest that Citron-K molecules are put together in the band-like structures in interphase cells and in the cleavage furrow of mitotic cells by the force of the actin cytoskeleton. Because Citron-K plays an essential role in cytokinesis at least in some populations of neuronal cells (Di Cunto et al., 2000), one function of the actin cytoskeleton in cytokinesis of mammalian cells is to make Citron-K accumulation in the cleavage furrow. It should be emphasized that our present data do not exclude the role of the actomyosin cytoskeleton in elicitation of the contractile force in cytokinesis. There has been no report showing the contractile force generation by Citron-K. However, a number of reports have shown the involvement of actin binding proteins including myosin in generation and processing of the contractile ring (Robinson and Spudich, 2000).

The data reported above have clarified how activated Rho mobilizes Citron-K during cell division. Citron-K exists in the cytoplasm in interphase, moves to midzone spindles, binds then to activated GTP-Rho, transfers to cell cortex and accumulates in the cleavage furrow. How do Citron molecules accumulated in the cleavage furrow exert a crucial function in cytokinesis? One plausible possibility is that Citron-K accumulated there by the force of F-actin in turn regulates functions of actin and other cytoskeletons, although there has been no direct evidence to support this hypothesis. Citron-K most probably exerts this action by phosphorylating some substrate(s) in this process. Identification of these substrates will clarify this issue. The above mobilization pathway has also raised several other important questions such as the identity of the initial Rho-independent, cell-cycle-dependent signal, how Citron-K moves to the midzone spindles, how Rho mobilizes Citron-K there to cell cortex and the function of the midzone spindles in this process. These questions may be solved by identification of domains of Citron-K responsible for each mobilization step and their binding partner there. This approach may also give us a new insight into the interaction among Rho, Citron-K, microtubules and actin cytoskeleton during mitosis.

We thank S. Tsukita, H. Bito, T. Tsuji, M. Okamoto, Y. Takada, F. Oceguera, K. Kimura, M. Maekawa and T. Furuyashiki for helpful advice and discussion, M. Kinoshita and M. Noda for generous supply of anti-Nedd 5 antibody, K. Nonomura for technical assistance, and T. Arai and H. Nose for secretarial help.