Schizosaccharomyces pombe Pak-related protein, Pak1p/Orb2p, phosphorylates myosin regulatory light chain to inhibit cytokinesis

The family of conserved p21-activated kinase (Pak) serine/threonine kinases are key downstream effectors of the Cdc42 family of GTPases (Manser et al., 1994). Pak kinases, like cdc42 GTPase, have been extensively studied and implicated in regulating filamentous actin (F-actin) cytoskeleton and cellular morphogenesis in several organisms. For example, Pak1 kinase activity is essential for cylindrical morphology of Schizosaccharomyces pombe (Ottilie et al., 1995), bud morphology in Saccharomyces cerevisiae (Weiss et al., 2000), and hyphal morphology in Candida albicans (Leberer et al., 1997). Pak kinases are also important for F-actin organization and turnover in mammalian cells (Manser et al., 1997) as well as axonal guidance (Hing et al., 1999) and epithelial polarity (Conder et al., 2007) in Drosophila melanogaster. Cytokinesis in S. cerevisiae requires Pak kinase for septin collar formation (Cvrckova et al., 1995; Kadota et al., 2004; Versele and Thorner, 2004). Pak kinase has also been implicated in cytokinesis in Dictyostelium discoideum, although the mechanism of regulation of cytokinesis by Pak kinase has not been clarified (de la Roche et al., 2005).

In recent years, the fission yeast S. pombe has become an attractive model organism for the study of the F-actin cytoskeleton, cellular morphogenesis, and cytokinesis. Two Pak kinase–related proteins have been identified in S. pombe, of which Pak1p/Orb2p/Shk1p is essential for cell viability (Ottilie et al., 1995) and localizes to the cell ends and to the cell division site (Qyang et al., 2002), whereas the second protein, Pak2p/Shk2p, has been shown to be dispensable for growth and colony formation (Sells et al., 1998; Yang et al., 1998).

In this study, we uncover a novel role for Pak1p kinase in the regulation of cytokinesis in fission yeast. We provide evidence that, in addition to localizing to the cell ends, Pak1p localizes to the actomyosin ring at the cell division site. Pak1p appears to phosphorylate the regulatory light chain of myosin II (Rlc1p) and thereby inhibits cytokinesis until complete segregation of the genetic material.

To better understand the role of fission yeast Pak1p, we first characterized the intracellular distribution of this protein. A strain in which the pak1 coding sequences were fused in frame with the gene encoding GFP was generated. As described previously in a fission yeast strain overexpressing pak1 (Qyang et al., 2002), wild-type Pak1p was detected at the cell ends of interphase cells (Fig. 1 A, i). In cells undergoing mitosis and cytokinesis, Pak1p was detected at the cell division site. To precisely determine the temporal regulation of Pak1p localization to the cell division site, we characterized its localization in cells expressing Sid4p-GFP, a marker of the spindle pole body (SPB) that also served to monitor cell cycle progression (Chang and Gould, 2000). In cells with two closely spaced SPBs, which is indicative of prometaphase, Pak1p signal was still detected at the cell ends (Fig. 1 A, ii). Interestingly, after prometaphase and before anaphase A, Pak1p relocated from the cell ends to the cell division site (Fig. 1 A, iii). The intensity of the medial signal increased as the cell proceeded through anaphase B (Fig. 1 A, iv). In septating cells, Pak1p signal was present at the septum region (Fig. 1 A, v).

To investigate whether Pak1p localization to the cell division site coincides with the presence of the actomyosin ring in mitosis, a strain coexpressing Pak1p-GFP and Rlc1p-Cherry was imaged (Fig. 1 B). In fission yeast, the actomyosin ring is assembled before anaphase A, whereas its constriction occurs after completion of anaphase B (Wu et al., 2003). This medial Pak1p-GFP signal colocalized with Rlc1p-Cherry (Fig. 1 B, i). During actomyosin ring constriction, Pak1p-GFP was detected in the vicinity of the constricting actomyosin ring as well as with the developing septum (Fig. 1 B, ii). In some tilted view, Pak1p-GFP was found to clearly organize into a ring structure (Fig. 1 B, iii).

To further determine whether Pak1p is associated with the actomyosin ring, we examined its localization in β-1,3 glucan synthase cps1-191 mutant that is defective in septum assembly at the nonpermissive temperature (Liu et al., 1999). At 36°C, most cells arrested with an actomyosin ring (Rlc1p-Cherry signal) and Pak1p-GFP were observed to localize at the division site (Fig. 1 C). Similarly, in β-tubulin nda3-KM311 mutant at the nonpermissive temperature, Pak1p-GFP was localized at the division site in these prometaphase-arrested cells (42% of the cell population arrested in prometaphase, n = 200; Fig. 1 D). These experiments established that Pak1p was localized in the vicinity of the actomyosin ring as well as the division septum, suggesting a potential role in the regulation of cytokinesis.

The available mutant alleles of pak1 were substantially compromised for cell shape and were unsuitable for the study of cytokinesis. Therefore, we created a mutant allele of pak1 (pak1-M460G) in which the ATP-binding site was mutated. The kinase activity of Pak1p-M460G relative to that of wild-type Pak1p was measured with myelin basic protein (MBP) phosphorylation (Fig. 1 E). This analysis established that the kinase activity of Pak1p-M460G was significantly reduced as compared with that of Pak1p.

The pak1-M460G strain was viable at 18–36°C laboratory conditions (Fig. S1, available) but displayed defects in growth patterning (Fig. 1 F). Pak1p-M460G–expressing cells exhibited a monopolar growth pattern. After cell separation, one daughter cell grew from the new end (Fig. 1 F, white arrows), whereas the other daughter cell grew from the old end (Fig. 1 F, black arrows). Interestingly, Pak1p-M460G was capable of localizing to the growing zones and to the cell division site (Fig. 1 F, 180'), suggesting that maximal kinase activity of Pak1p is not essential for its cellular localizations. Pak1p-M460G was present at the division site in 37% of cell populations (n = 200) that represented the prometaphase-arrested cells under these conditions (Fig. 1 G).

To understand the role of Pak1p function in cytokinesis, we studied the dynamics of actomyosin ring assembly and constriction in wild-type and pak1-M460G cells expressing Rlc1p-GFP (actomyosin ring marker) and Sid4p-GFP (SPB marker) by time-lapse microscopy. Kymographs of the Rlc1p-GFP actomyosin ring beginning at early anaphase (defined by two SPBs at 5–7 μm apart) are shown in Fig. 2 A. Assembly of the actomyosin ring was not perturbed in pak1-M460G cells in that early anaphase cells contained normal actomyosin rings. Interestingly, whereas the process of cytokinesis lasted a mean value of 36.2 min in wild-type cells, the process was considerably faster with a mean value of 31.6 min in pak1-M460G–expressing cells (Fig. 2, A and B). Mitotic spindle elongation rates in wild-type– and pak1-M460G–expressing cells were comparable (unpublished data), suggesting that the accelerated cytokinesis in pak1-M460G cells was not caused by accelerated mitotic progression.

We considered two possible explanations for the observed accelerated cytokinesis in pak1-M460G–expressing cells. First, it was possible that actomyosin ring constriction was accelerated in pak1-M460G cells. Second, it was possible that actomyosin ring constriction was initiated earlier than normal in pak1-M460G cells. Kymographic analyses revealed that the mean time of ring constriction was similar in wild-type and pak1-M460G cells, but the initiation of ring constriction seemed to be earlier in pak1-M460G cells (Fig. 2 A, arrows).

To confirm our analysis that the actomyosin ring constriction was initiated earlier in pak1-M460G cells, we synchronized cells in prometaphase using the cold-sensitive nda3-KM311 mutant. Under these conditions, nda3-KM311 and the nda3-KM311 pak1-M460G mutants arrested in prometaphase with stable actomyosin rings. Return of the arrested cells to the permissive temperature led to spindle assembly and a synchronous resumption of downstream events of mitosis, such as chromosome segregation and cytokinesis (Fig. 2 C). In nda3-KM311 cells (Fig. 2 D), as in wild-type cells, actomyosin ring constriction was largely initiated upon dissolution of the mitotic spindle. In contrast, actomyosin ring constriction in the majority of nda3-KM311 pak1-M460G cells was initiated when the mitotic spindle was still present (Fig. 2, D and E). Collectively, these experiments established that compromise of Pak1p kinase function led to premature constriction of the actomyosin ring in anaphase cells.

To understand the molecular mechanism underlying pak1 kinase regulation of cytokinesis, we took a candidate approach to identify potential substrates whose phosphorylation might regulate actomyosin ring dynamics. Mammalian pak1 inhibits myosin II activity by inhibiting myosin light chain kinase and thereby reducing the activating phosphorylation of its myosin regulatory light chain (MRLC; Sanders et al., 1999). Conversely, a previous study has also shown that mammalian Pak can directly phosphorylate MRLC and activates myosin II (Chew et al., 1998). The amino acids in S. pombe Rlc1p that are equivalent to the phosphorylation sites in mammalian MRLC are shown in Fig. S2 (available). The codons for the serine residues 35 and 36 in S. pombe rlc1 were substituted with those for alanine (rlc1-S35AS356A [SA]). S. pombe Pak1p kinase, Rlc1p, and Rlc1p-SA were expressed in bacteria and purified. In vitro kinase assays revealed that Pak1p kinase phosphorylated wild-type Rlc1p but not Rlc1p-SA (Fig. 3 A). We further analyzed Rlc1p phosphorylation in vivo in synchronized early anaphase cells. By isoelectric focusing and SDS-PAGE analyses, Rlc1p from wild-type cells was resolved into two spots (Fig. 3 B, arrowheads) with the smaller spot (nearer to pH 3), possibly a phosphorylated form (Fig. 3 B, top). Interestingly, Rlc1p from pak1-M460G cells was resolved as one spot indicating lack of protein modification (Fig. 3 B, bottom). Thus, Rlc1p is directly phosphorylated either by Pak1 kinase or by another kinase downstream of Pak1 in vivo.

Protein–protein interaction of Pak and its substrate has been shown to be important for its localization in mammalian cells (Manser et al., 1998). Because Rlc1p is a putative substrate of Pak1p, we tested whether Pak1p-GFP localization to the actomyosin ring depended on Rlc1p function. To this end, nda3-KM311 rlc1Δ cells were arrested at prometaphase, and Pak1p-GFP localization was monitored. In these rlc1Δ cells, the actomyosin ring was less compacted (Fig. 3 C, actin). In nda3-KM311 and nda3-KM311 rlc1Δ cells, F-actin was observed in medial rings and bands in 65–75% of cells. However, Pak1p-GFP was observed in these medial rings and band structures in <5% of cells and was largely detected as aggregates in 63% of the nda3-KM311 rlc1Δ cells (n = 200; Fig. 3 C, arrows). The fact that Pak1p localization to cell ends (Fig. 3 C, asterisks) was not affected established that loss of Rlc1p function was not generally destabilizing the localization of Pak1p at all cellular locations but was specific for its effects on the actomyosin ring. Collectively, these experiments established that Pak1p likely phosphorylates Rlc1p at serine 35 and 36 and that the putative substrate played an important role in the localization of the kinase to the actomyosin ring.

We have shown that cells compromised for Pak1p function undergo premature constriction of the actomyosin ring and that Pak1p phosphorylates Rlc1p on serine 35 and 36 both in vitro and in vivo. We investigated the physiological consequences resulting from compromised Pak1p function. In fission yeast cells, the actomyosin ring is assembled upon entry into mitosis and constricts upon completion of anaphase B and spindle dissolution. Prometaphase-arrested nda3-KM311 pak1-M460G cells did not undergo actomyosin ring constriction and septation (unpublished data). Thus, Pak1p function was not required to restrain cytokinesis in metaphase-blocked cells.

We tested whether cells that undergo slower anaphase progression would depend on Pak1p function to restrain actomyosin ring constriction and septation. To this end, we created double mutants defective in Ase1p, a protein important for spindle integrity and nuclear segregation (Loiodice et al., 2005), and pak1-M460G. This strain exhibited a deleterious genetic interaction and was not viable (unpublished data). As a result, nmt1-pak1 (pak1 expression regulated by nmt1 promoter at its genomic locus) strain was used in place of pak1-M460G for this analysis. nmt1-pak1 cells were similar to pak1-M460G cells (Fig. 4 A). We tested genetic interactions and cytokinesis timing in nmt1-pak1 ase1Δ double mutants. Interestingly, nmt1-pak1 ase1Δ double mutants displayed defective cytokinesis characterized by the presence of fully formed septa before full segregation of the chromosomes. A similar effect was observed in double mutants defective in ase1 and rlc1-SA (in which serine 35 and 36 were replaced with alanine; Fig. 4 B). Quantification of these experiments is shown in Fig. 4 C. Live cell imaging of ase1Δ and rlc1-SA ase1Δ double mutants was also performed to determine the temporal regulation of ring constriction (Fig. 4 D). In ase1Δ cells, actomyosin ring constriction (followed using Rlc1p-GFP as a marker) was initiated at least 20 min after ring assembly and nuclear segregation. However, in the double mutant, ring constriction was initiated earlier and before nuclear segregation, leading to a variety of aneuploidy events, including the formation of anucleate cells in several instances. The phenotype is classified into two major types: (a) septated cell with the nucleus displaced to one compartment and (b) septation through the segregating nuclei. Collectively, these experiments suggested that Pak1p by phosphorylating Rlc1p might function to delay actomyosin ring constriction and septum assembly until completion of chromosome segregation and spindle disassembly.

In summary, we have shown that fission yeast Pak1p localizes to the actomyosin ring during early mitosis. Pak1p localization is similar to that of Rga8p, a Rho–GTPase-activating protein that interacts with Pak1p (Yang et al., 2003). It is presently unclear whether the Rho–GTPase-activating protein Rga8p might participate in the regulation of cytokinesis as well. Although compromising Pak1p function did not affect the timely assembly of actomyosin rings, actomyosin ring constriction was prematurely initiated in the pak1-M460G mutant. We have shown that Pak1p phosphorylates serine 35 and 36 of Rlc1p, the regulatory light chain of fission yeast myosin II. Compromising Pak1p function or Rlc1p function (by replacement of serine 35 and 36 with the nonphosphorylatable amino acid alanine) in cells that are slowed down in anaphase leads to aneuploidy caused by premature ring constriction. However, premature ring constriction was not observed upon compromise of Pak1p or Rlc1p function in prometaphase- and/or metaphase-arrested cells. These observations suggest that at least two mechanisms might exist to prevent actomyosin ring constriction until completion of chromosome segregation. The first mechanism might involve prevention of activation of the septation initiation network in metaphase by high levels of Cdk activity (Guertin et al., 2000). Our current experiments reveal that Pak1p, via Rlc1p phosphorylation, might prevent actomyosin ring constriction in anaphase A and B cells, thereby contributing to genomic stability (Fig. 4 E).

In metazoans, phosphorylation of the major isoform MRLC at amino acids (equivalent to serine 35 and 36 of the fission yeast Rlc1p) leads to activation of myosin II function and appears to regulate cytokinesis (Yamakita et al., 1994; Matsumura et al., 1998). In contrast, our experiments suggest the possibility that phosphorylation of Rlc1p at serine 35 and 36 leads to down-regulation of myosin II function. Although these observations appear to contradict each other, it is possible that other MRLC isoforms might exist in animal cells whose phosphorylation might prevent myosin II activation. Future studies should test this possibility.

Pak1p-GFP has its 5′ start codon replaced with a stop codon, and its 3′ stop codon is replaced with a six-glycine linker and two copies of Myc epitopes and is cloned in frame to GFP. Mutations were generated by site-directed mutagenesis (Agilent Technologies). Plasmids were linearized with XhoI and integrated into the pak1 locus by homologous recombination; its expression was under the control of the native promoter. The nmt1-pak1 strain was constructed by PCR-based targeting at pak1 genomic locus (Bahler et al., 1998).

S. pombe strains used in this study are listed in Table I. The ase1-null strain was provided by S. Oliferenko (Temasek LifeSciences Laboratory, Singapore). All Pak1p-GFP–expressing strains were grown in rich yeast extract plus supplements (YES) media, wild-type and Pak1p-M460G–expressing cells were grown at 24°C to mid-log phase, and strains in cps1-191 and nda3-KM311 genetic background were grown at temperatures specified in the figure legends. Cells were spotted on YES 2% agarose pad, and imaging was performed with a microscope with a Plan Apo 100×/1.45 NA oil lens (IX71; Olympus) equipped with a camera (CoolSnap ES; Photometrics) and MetaMorph software (version 6.2r6; MDS Analytical Technologies) at 22°C. cdc25-22 cells were grown at 24°C to mid-log phase, were shifted to 36°C for 4 h to arrest cells in G2, and were shifted down to 24°C to allow synchronous entry into mitosis. Conditions for time-lapse imaging were performed according to the protocol described previously (Tran et al., 2004) with a confocal microscope (LSM510; Carl Zeiss, Inc.) with a 24°C heated Plan-Neofluor 100×/1.30 NA oil lens. Images were converted into a 3D stack with LSM Image Browser (version 3,5,0,376; Carl Zeiss, Inc.), and a kymograph was generated with MetaMorph (version 6.1). nda3-KM311 strains were methanol fixed for indirect immunofluorescence staining with mouse anti-TAT1 (provided by K. Gull, University of Manchester, Manchester, England, UK) and rabbit anti-cdc4 antibodies at 1:200 dilutions and secondary antibodies Alexa Fluor 488 and 594 conjugate. Images were captured with a microscope (DM IRE2; Leica) with a 100×/1.4 NA oil lens equipped with a CoolSnap ES camera and MetaVue software (version 6.2r4; MDS Analytical Technologies). nmt1-pak1 and nmt1-pak1 ase1–null strains were grown in minimum medium liquid media at 24°C, and subsequently these cultures were split into two halves: one half remaining in minimum medium liquid and the other half grown in YES (supplemented with 15 μM thiamine) to repress pak1 expression. rlc1 ase1–null stains were grown in YES liquid media at 24°C.

Recombinant GST-Pak1p, GST-Rlc1p, and GST-Rlc1p-S35AS36A were expressed in a BL21 DE3 bacteria strain and purified. 1 μg Pak1p kinase was incubated in kinase buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 10 mM MgCl₂, and 1 mM MnCl₂) containing 10 μCi of freshly added γ-[³²P]ATP, 10 μM cold ATP, and substrate (20 μg Rlc1p, 20 μg Rlc1p-S35AS36A, or 2.5 μg MBP) at 22°C for 30 min. The reaction was terminated with SDS sample buffer and resolved with SDS-PAGE for autoradiography.

cdc25-22 cells were harvested 30 min after synchronous entry into mitosis. Cells were lysed by glass bead disruption in buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 2 mM EDTA, 50 mM NaF, 20 mM β-glycerophosphate, 1% Triton X-100, 100 nm okadaic acid, 1 mM DTT, protease inhibitor cocktail, and PMSF). An immobilized pH gradient strip (Bio-Rad Laboratories), pH 3–10, was rehydrated with 160 μg of clarified protein lysate overnight before focusing. The immobilized pH gradient strip was transferred to 10% SDS-PAGE. Western blotting was performed with anti-Myc antibodies and protein A–HRP conjugate.

Fig. S1 shows that kinase-compromised pak1 alleles were viable. Fig. S2 shows the sequence alignment of mammalian MRLC and S. pombe Rlc1p from the eukaryotic orthology (YOGY) database.

We thank Drs. K. Gull for antitubulin antibodies, S. Oliferenko for ase1-null strain, T.G. Chew for experimental help, and T.G. Chew and others of the Cell Division Laboratory for discussions and/or critical reading of the manuscript.

T.-H. Loo is a recipient of a fellowship from the Singapore Millennium Foundation, and this work was funded by research funds from the Temasek Life Sciences Laboratory.