Report

Polo-Like Kinase Cdc5 Controls the Local Activation of Rho1 to Promote Cytokinesis

See allHide authors and affiliations

Science  07 Jul 2006:
Vol. 313, Issue 5783, pp. 108-111
DOI: 10.1126/science.1126747

Abstract

The links between the cell cycle machinery and the cytoskeletal proteins controlling cytokinesis are poorly understood. The small guanine nucleotide triphosphate (GTP)–binding protein RhoA stimulates type II myosin contractility and formin-dependent assembly of the cytokinetic actin contractile ring. We found that budding yeast Polo-like kinase Cdc5 controls the targeting and activation of Rho1 (RhoA) at the division site via Rho1 guanine nucleotide exchange factors. This role of Cdc5 (Polo-like kinase) in regulating Rho1 is likely to be relevant to cytokinesis and asymmetric cell division in other organisms.

Cytokinesis, the physical separation of daughter cells after mitosis, involves the dynamic reorganization of the cortical cytoskeleton and is precisely regulated both in time and space (1, 2). Successful cytokinesis is critical to the maintenance of genome stability: Cytokinesis failure can lead to chromosome aberrations and cancer (3). The timing and mechanism of contractile actin ring (CAR) assembly is an important but poorly understood aspect of the eukaryotic cell cycle.

The Polo-like kinase Cdc5 controls many aspects of cell division, including cytokinesis (4, 5). Polo-like kinase is involved in cytokinesis in animal cells, but its precise role has been difficult to address because manipulations that inhibit Polo-like kinase produce spindle assembly defects, which affect CAR organization (2, 6, 7). RhoA is an attractive candidate to be under Polo-like kinase regulation because of its essential role in CAR assembly and contraction (1, 2, 8, 9). In yeast, cytokinesis is largely independent of microtubules (8), making yeast an advantageous system to study the role of Polo-like kinase. Additionally, the contractile ring is not essential in most budding yeast strains; another mechanism involving cell wall deposition can substitute for CAR function (10, 11). This facilitates genetic analysis because null alleles affecting CAR function can be studied in viable cells. Despite these differences, the core machinery for CAR assembly is conserved in budding yeast: CAR assembly requires Rho1, the functional homolog of RhoA in animal cells, which activates actin filament assembly via formins (8, 12).

The timing of CAR assembly was characterized by actin staining and by labeling of the CAR with a fusion between tropomyosin and green fluorescent protein (Tpm2-GFP) (Fig. 1A and fig. S1A). To identify cell cycle regulators required for CAR assembly, we reexamined CAR assembly in mutants that block mitotic exit (11, 13). The CAR assembled in cdc15-2 or cdc14-1 strains, mutations that compromise the canonical mitotic exit network (MEN) (Fig. 1A and fig. S1B). By contrast, CAR assembly was impaired in mutants lacking Cdc5 (Fig. 1A and fig. S1B).

Fig. 1.

Cdc5 is required for CAR assembly. (A) CAR assembly is defective in cdc5- but not cdc15- or cdc14-arrested cells. Cells were released from a G1 block (α factor) and examined at intervals after release at the nonpermissive temperature 34.5°C for budding index (diamonds), percentage of anaphase cells by labeling nuclear DNA (squares), and bud neck localization of Tpm2-GFP (triangles). The α factor was re-added at 60 min to prevent progression into a second cell cycle. (Bottom) Representative images of cells from the 120-min time point. Scale bar indicates 5 μm. (B) Cdc5 is required to maintain the CAR. Log-phase cultures of cdc5-2 and control cells were arrested in telophase by overexpression of BFA1 from the GAL1 promotor for 3 hours. The telophase-arrested cells were shifted to a restrictive temperature (34.5°C) to inactivate cdc5-2 and cdc15-2 and fixed, and the CAR formation was visualized with the use of Tpm2-GFP. (C) Cdc5 is required for the subcellular targeting of Bni1 and Rho1. Bni1-3xGFP expressed at an endogenous level was visualized after 2-hour incubation of the cells at 34.5°C. Rho1 was visualized by immunofluorescence labeling with a rabbit antibody against Rho1 130 min after release to 34.5°C from a G1 block. Arrows point to the bud neck signals in cdc15 cells and the bud tip signals in cdc5 cells.

The Cdc5 requirement for CAR formation was observed either by using a very severe conditional allele, cdc5-2, or by examining the null phenotype after Cdc5 depletion (fig. S2A) (14). The defect in CAR assembly in cells lacking Cdc5 was not a secondary consequence of the anaphase defect in these cells: Cdc5 was not only required to establish CAR assembly but also to maintain it. When cells were arrested before mitotic exit by overexpression of a MEN inhibitor, Bfa1, CAR maintenance still required Cdc5 [Fig. 1B and Supporting Online Material (SOM) Text]. Furthermore, early anaphase pathways regulated by Cdc5 were not required for CAR assembly (SOM Text and fig. S2B). In the absence of Cdc5, a residual fraction of cells were observed that had thin CARs (Fig. 1A and fig. S2B). Thus, Cdc5, although not absolutely essential for CAR assembly, is an important regulator.

Next we characterized the nature of the CAR assembly defect in cdc5-2 cells at the restrictive temperature, with the use of the MEN mutant cdc15-2 as a control. Cdc5 is required for septin disassembly after mitotic exit but is not required for septin assembly (15). The septin Shs1, the type II myosin Myo1, and the IQGAP homolog Iqg1 (Cyk1) were properly recruited to the bud neck both in the cdc5-2 and in the cdc15-2 control cells (fig. S2C). By contrast, although the formin Bni1 and Rho1 localized normally to the neck in cdc15-2–arrested cells, neither Bni1 nor Rho1 was detectable at the bud neck in cdc5-2–arrested cells (Fig. 1C). Furthermore, Bni1 was often mistargeted to the bud tip in cdc5-2–arrested cells (Fig. 1C). Thus, Cdc5 is required for the neck recruitment of Rho1 and Bni1, a key downstream factor required for CAR assembly.

To identify potential Cdc5 substrates relevant to Rho1 regulation, we screened all known budding yeast Rho guanosine triphosphatase (GTPase) regulators for interaction with the Polo-box domain (PBD) of Cdc5 (7, 16, 17). PBD binding partners include the Rho1 guanine nucleotide exchange factors (GEFs) Rom2 and Tus1, the Rho1 GTPase activating protein (GAP) Sac7, and a putative Rho-GAP, Ecm25 (Fig. 2, A and B). The PBD also binds Bem3, a GAP for Cdc42, the major small GTPase that controls polarized morphogenesis in budding yeast. These interactions are highly specific to the priming phosphorylation of the candidates, because the interaction was lost with a Polobox pincer mutation, PBD* (Fig. 2, A and B), that abolishes phosphospecific recognition of the substrates by PBD (17). All the candidates had multiple potential Polo-box binding sequences, binding motifs that are often primed by cyclin-dependent kinase (CDK)–dependent phosphorylation (16). Tus1, Bem3, Sac7, and Ecm25 are also good CDK substrates (18).

Fig. 2.

A screen for Polo-box binding partners suggests that Cdc5 controls multiple Rho-type GTPase regulators. All potential Rho GTPase regulators in the yeast genome were screened for physical interaction with the Cdc5 PBD and a control pincer mutation (PBD*). (A) In cases where epitope-tagged constructs were not available from the yeast tandem affinity purification (TAP)–tagged library (24), epitope-tagged constructs were generated. Tus1 and Rom1 were overexpressed from the GAL1 promoter for this initial screen. After precipitation assay, Western blotting was performed to detect the indicated epitope tags of the bound proteins. + indicates the wild-type GST-PBD; +* indicates the pincer mutant. (B) PBD or PBD* precipitation assays with lysates from strains from the yeast TAP-tag fusion library. (C) Endogenous-level Tus1 binds to the PBD. This binding is abolished by the ST mutation and thus requires the Polo-box binding motif. WT, wild type. (D) Hyperphosphorylation of Tus11-300-myc requires Cdc5 and the Polo-box binding motifs. Shown are Western blots to detect endogenous-level Tus11-300-myc or its ST derivative in the indicated strains after 2.5 hours at the restrictive temperature of 34.5°C. (E) Phosphomimetic mutations at Polo-consensus phosphorylation sites partially bypass the requirement for Cdc5 for CAR assembly and Rho1 recruitment. The indicated strains were arrested with α factor and released at the nonpermissive temperature (33.5°C) for 140 min. The CAR was visualized with the use of Tpm2-GFP, and Rho1 was visualized by immunofluorescence. More than 400 cells were counted for each sample. The experiment was repeated three times with near-identical results. By a Student's t test, the difference in Tpm2-GFP labeling in cells expressing Tus1-6DE versus Tus1 was significant (P = 0.011). Error bars indicate SEM. Scale bar, 5 μm; arrows point to the bud neck signal.

Because of the requirement of Rho1 for CAR assembly, we focused our subsequent analysis on the Rho1 GEFs, Tus1 and Rom2 (19). Different patterns of synthetic lethal interactions distinguish genes required for CAR function from genes required for the CAR-independent pathway for cytokinesis (8, 11). Analysis of CAR assembly and synthetic lethal interactions strongly implicated Tus1 and Rom2 in cytokinesis, with Tus1 playing the dominant role in CAR assembly (SOM Text, table S1, and fig. S3). These findings are consistent with the requirement for the fission yeast Tus1 homolog in cytokinesis (2022). The pattern of synthetic lethality observed with cdc5-2 also provided genetic evidence for a role for Cdc5 in the CAR-dependent mechanism (SOM Text and fig. S4).

We next evaluated the possibility that Tus1 is a direct Polo-like kinase substrate. The N terminus of Tus1 (Tus11-300), containing two Polo-box binding motifs (Ser-Ser/Thr-Pro), is sufficient for Polo-box binding (fig. S5, A and B). The serine residues immediately preceding the CDK consensus site in the Polo-box binding motifs of Tus1 (Ser7 and Ser92) were mutated to threonine (hereafter referred to as ST mutations), a well-characterized change that disrupts binding to the PBD (16, 17) but that should not affect CDK-dependent phosphorylation. Glutathione S-transferase (GST)–PBD precipitation experiments revealed that neither Tus1-ST1–300 nor full-length Tus1-ST protein bound to the PBD (Fig. 2C and fig. S5B). Similar ST substitutions were introduced into the coding sequence for Rom2; the Rom2-ST protein also had a reduced affinity for the Cdc5 PBD (fig. S6). Thus, Tus1 and Rom2 interact with Cdc5 by a mechanism similar to that of other known Polo substrates.

Biochemical experiments suggested that Tus1 is a bona fide in vivo Cdc5 substrate. Although we were not able to detect a clear mobility shift of the full-length Tus1 (1307 amino acids), the Tus1 Polo-box interacting domain (PID), Tus11-300, was phosphorylated in vivo as evidenced by a motility shift that was abolished by phosphatase treatment in cdc15-2–arrested cells (figs. S2D and S5C). Most notably, the slowest migrating form of Tus11-300 (hereafter referred to as hyperphosphorylated) was absent in cdc5-2–arrested cells (Fig. 2D). This hyperphosphorylated form could also be induced in cycling cells by overexpression of Cdc5 (fig. S5D). Furthermore, Tus1-ST1-300 also lost this hyperphosphorylated band (Fig. 2D). Thus, Tus11-300 is phosphorylated in late mitosis in a manner that depends on Cdc5. The fact that this Cdc5-dependent phosphorylation also required physical interaction between Tus11-300 and the Polo box strongly suggests that the Cdc5-dependent phosphorylation of Tus11-300 is direct. Indeed, mass spectrometry confirmed that the Tus11-300 was readily phosphorylated by a human Polo-like kinase, Plk1, in vitro (table S2), identifying at least 12 phosphorylated residues. Although we were not able to map phosphorylation sites directly on full-length Tus1 because of its low expression (23), together with our genetic data these findings strongly suggest that Tus1 is a Cdc5 substrate.

We attempted similar experiments to characterize the potential Polo-like kinase dependent phosphorylation of Rom2. We identified several in vitro Plk1 phosphorylation sites by mass spectrometry (table S3). However, we were not able to identify a mobility shift for full-length Rom2, nor were we able to detect a mobility shift for several N-terminal Rom2 fragments. These fragments were not well expressed in Escherichia coli and thus may not fold properly. Nevertheless, the similar phenotypes of rom2-ST and tus1-ST strains (SOM Text, table S4, and fig. S7) suggest that in vivo Cdc5 regulates Rom2 similarly to Tus1.

A Tus1 protein containing several phosphomimetic mutations at putative Cdc5 phosphorylation sites provided further evidence for the importance of Cdc5-dependent phosphorylation of Tus1 for CAR assembly. Phosphomimetic mutations (Ser to Asp or Thr to Glu) were introduced at Ser or Thr residues within Plk1 phosphorylation consensus sites (Asp/Glu-X-Ser/Thr, where X is any amino acid). This consensus is optimal for Plk1-dependent phosphorylation of vertebrate Cdc25 (23) and perfectly matches the in vivo Cdc5 target phosphorylation sites on Mcd1 (Scc1) (14). Three out of four Plk1 phosphorylation consensus sites within Tus11-300 were phosphorylated by Plk1 in vitro (table S2). A Tus1 mutant containing six putative phosphomimetic changes (Tus1-6DE) partially bypassed the requirement for Cdc5, both for CAR formation and for Rho1 localization to the division site (Fig. 2E). A control Tus1 mutant where the putative phosphorylation sites were mutated to Ala failed to restore CAR assembly and Rho1 localization in cdc5-2 cells, demonstrating that only the negatively charged residues caused a gain-of-function effect at these sites. The magnitude of the gain-of-function phenotype induced by TUS1-6DE is likely to be an underestimate, because the introduction of negatively charged residues appears to reduce the already low steady state expression of Tus1 (fig. S5E) (24).

The synthetic lethal interactions observed with tus1-ST and rom2-ST suggested that these mutations were specifically defective in the CAR-dependent pathway for cytokinesis but not other Rho1-dependent cellular functions (SOM Text, table S4, and fig. S7). Thus, the ST mutants are separation-of-function alleles that are specifically defective in contractile ring function. Supporting this hypothesis, tus1-ST strains arrested in telophase were defective in CAR assembly, and this defect was enhanced in tus1-ST rom2-ST double mutant strains (Fig. 3A).

Fig. 3.

Polo regulation of Tus1 and Rom2 is critical for CAR assembly and local Rho1 activation. (A) Polo regulation of Tus1 and Rom2 is required for CAR assembly and for recruitment of active Rho1 to the bud neck. CAR assembly was assayed with use of Tpm2-GFP. Active Rho1 was detected by immunofluorescence by using an antibody that specifically recognizes Rho1-GTP but not Rho1-GDP (25). Error bars indicate SEM; scale bar, 5 μm; arrows point to the bud neck signals. (B) The total cellular level of active Rho1 requires Cdc5- but not Polo-dependent phosphorylation of Tus1 and/or Rom2. The precipitation assay was performed with the use of Pkc1-RBD.

Our experiments suggest that Cdc5 controls Rho1 through Tus1 and Rom2. This was directly tested by monitoring the amount of active Rho1 both globally and locally at the bud neck. The total amount of active GTP-bound Rho1 was detected by a precipitation assay with the Rho1-binding domain (RBD) of Pkc1 (Fig. 3B). The amount of active Rho1 was diminished in cdc5-2–arrested cells in comparison with cdc15-2–arrested cells (Fig. 3B). We also visualized active Rho1 by immunofluorescence using an antibody that specifically detects GTP-bound Rho1 (25). Consistent with the precipitation experiment of active Rho1, activated Rho1 was barely detectable in cdc5-2–arrested cells but was abundant and recruited to the bud neck region in cdc15-2–arrested cells (Fig. 3A). By contrast, tus1-ST rom2-ST cells arrested at the cdc15-2 block had active Rho1 at the bud cortex but lacked active Rho1 at the bud neck (Fig. 3A). Consistent with this observation, the precipitation assay indicated that Rho1 activity was not diminished in tus1-ST rom2-ST strains and in fact was slightly elevated (Fig. 3B). The fact that total Rho1 activity was high in tus1-ST rom2-ST strains but low in cdc5-2 strains suggests that Cdc5 also controls global Rho1 activity through another mechanism, perhaps through Rho1 GAPs such as Sac7 (Fig. 2B). Thus, Cdc5-dependent regulation of Tus1 and Rom2 controls the local activation of Rho1 at the division site but not the global levels of active Rho1 in the cell.

Next, we determined whether Cdc5-dependent phosphorylation is required to recruit Tus1 and Rom2 to the bud neck. Consistent with its important role in CAR assembly, Tus1-GFP was recruited to the bud neck just before CAR constriction and formed a single ring that colocalized and contracted with Myo1-CFP (fig. S8). The bud neck localization of Tus1 was dependent on Cdc5 activity but not on other MEN components (Fig. 4A). Furthermore, Tus1-ST-GFP failed to localize to the division site in cdc15-2–arrested cells that had high Cdc5 activity (Fig. 4A).

Fig. 4.

Cdc5 targets Tus1 to the division site. (A) Tus1-GFP requires Cdc5 and its Polo-box binding motifs to localize to the bud neck. Tus1-GFP or the ST derivatives were imaged in cdc5-2 or in cdc15-2 2 hours after incubation at 34.5°C (n > 100), along with the percentages of the large budded cells with GFP signal at the bud neck. Scale bar, 5 μm. (B) Tethering Tus1-ST to the bud neck by fusion to Mlc2 corrects the CAR assembly defect of tus1-ST rom2-ST strains and partially bypasses the requirement for Cdc5 in CAR assembly. (Left) Numerical summary of the data. The bars indicate the percentage of cells with detectable contractile rings. Error bars indicate SEM. The dark bars indicate the percentage of cells with robust contractile rings. (Right) Examples of cells. (Top) GFP imaging of the indicated fusions. (Bottom) Actin was visualized with alexa568-phalloidin. Arrows indicate cells with robust CAR labeling.

We next tested whether the requirement for Cdc5 regulation of Tus1 could be bypassed by tethering the ST variant to the bud neck. The coding sequence for Tus1-ST-GFP was fused to the coding sequence for Mlc2, a nonessential type II myosin light chain (26), and expressed from the endogenous TUS1 promoter. Indeed, in cdc15-2–arrested cells, Tus1-ST-Mlc2-GFP protein localized to the bud neck and activated CAR assembly, unlike Tus1-ST-GFP (Fig. 4B). Furthermore, Tus1-Mlc2-GFP could partially restore CAR assembly in cdc5-2–arrested cells (Fig. 4B). Thus, a critical function of the Cdc5-dependent regulation of cytokinesis is to recruit Tus1, and thus Rho1, to the site of CAR assembly. These results do not exclude the additional possibility that Cdc5 also affects Tus1 GEF activity.

Our study reveals a molecular pathway by which Cdc5 controls contractile ring assembly in budding yeast. Cdc5 is required for the recruitment of Rho GEFs to the division site that in turn is necessary for recruitment and activation of Rho1. The failure of this mechanism results in the mistargeting of the formin Bni1 from the bud neck to the bud tip. This causes a defect in CAR assembly, because the actin filaments in the contractile ring are assembled by formins (1, 2, 8). Because of the conservation of many components in the pathway and because the RhoA GEF Ect2 is a substrate for Plk1 in vitro (27), the recruitment of Rho1 activity to the division site by Polo-like kinases might be conserved in animal cells.

It is possible that Cdc5 has more broad roles in Rho-type GTPase regulation than the local activation of Rho1 at the division site. Cdc5 is required for the global activation of Rho1 in late mitosis, whereas the Cdc5 regulation of Tus1 and Rom2 is required for local Rho1 recruitment and activation. Cdc5 could exert additional control of Rho1 via Rho GAPs. It is also possible that Cdc5 could exert Polo-box independent regulation of Rho1 GEFs. Lastly, given that fission yeast Polo kinase was recently shown to be required for stress-induced polarized growth (28), it is intriguing that our biochemical screen identified Bem3, a Cdc42 GAP, as a Polo-box binding partner. Our findings thus elucidate a mechanism by which Polo-like kinase governs Rho1 during cytokinesis and raise the possibility that Polo-like kinase controls other aspects of the cortical cytoskeleton through other Rho GTPases.

Supporting Online Material

www.sciencemag.org/cgi/content/full/1126747/DC1

Materials and Methods

SOM Text

Figs. S1 to S8

Tables S1 to S6

References

References and Notes

View Abstract

Navigate This Article