Polo boxes form a single functional domain that mediates interactions with multiple proteins in fission yeast polo kinase

The cell division cycle in eukaryotic cells is a highly regulated process requiring a range of events to be carried out sequentially. The polo-like kinases (or polo kinases) are a family of protein kinases that are implicated in numerous events throughout mitosis (reviewed by Glover et al., 1996; Lane and Nigg, 1997; Glover et al., 1998; Donaldson et al., 2001). Although mammalian cells contain more than one polo-like kinase, yeasts and Drosophila have only one. An understanding of how polo kinases regulate mitotic events will be essential to understanding molecular mechanisms of the mitotic progression as a whole.

One of the most intriguing aspects of polo kinase function is the variety of tasks they execute throughout the cell cycle. Polo kinases are required at several key points through mitosis, starting from control of the G2/M transition through phosphorylation of Cdc25C and mitotic cyclins(Abrieu et al., 1998; Karaiskou et al., 1999; Kumagai and Dunphy, 1996; Ouyang et al., 1997; Qian et al., 1998; Toyoshima-Morimoto et al.,2001) and a role in the DNA damage checkpoint to prevent entry into mitosis (Sanchez et al.,1999; Smits et al.,2000; Toczyski et al.,1997). At the beginning of mitosis, various proteins are recruited to the centrosomes, a maturation process which requires polo kinases(Sunkel and Glover, 1988; Lane and Nigg, 1996). Polo kinases are also required for the establishment of a bipolar spindle(Ohkura et al., 1995; Lane and Nigg, 1996; Qian et al., 1998), a conserved function which is evident from the phenotype of the original Drosophila polo¹ mutant(Llamazares et al., 1991; Sunkel and Glover, 1988).

Equally, polo kinases are important for exit from mitosis. A role in the metaphase to anaphase transition via an interaction with the anaphase promoting complex/cyclosome (APC/C) has been demonstrated(May et al., 2002; Descombes and Nigg, 1998; Charles et al., 1998; Shirayama et al., 1998). In addition, budding yeast polo kinase phosphorylates cohesin to allow proteolysis by separase in order to initiate anaphase(Alexandru et al., 2001).

Fission yeast polo kinase, Plo1, is required for formation and correct positioning of the septum and overexpression induces septation even in interphase cells (Ohkura et al.,1995; Bahler et al.,1998). Overproduction of murine or budding yeast polo kinases in budding yeast cells also induces septation in a non-catalytic domain dependent manner (Lee and Erikson, 1997; Song et al., 2000), and a physical interaction has been demonstrated between the budding yeast polo kinase, Cdc5p, and septins (Song and Lee,2001). This function in cytokinesis has also been shown to be conserved in Drosophila (Carmena et al., 1998; Herrmann et al.,1998).

Dynamic, cell cycle regulated localisation and kinase activation during mitosis is a feature common to all members of the family. Localisation to the SPB/centrosome occurs early in mitosis and persists until late anaphase, when the protein has been seen to relocalise to the midbody or to the future site of cell cleavage (Bahler et al.,1998; Golsteyn et al.,1995; Logarinho and Sunkel,1998; Moutinho-Santos et al.,1999; Mulvihill et al.,1999; Shirayama et al.,1998; Song et al.,2000). Detailed study in fission yeast has indicated that localisation to the SPB takes place very early in mitosis, prior to full activation of catalytic activity (Tanaka et al., 2001). This cell cycle regulated catalytic activation and localisation is likely to play an important role in the execution of polo kinase function.

Polo-like kinases are characterised by an amino terminal catalytic domain,and a carboxy terminal non-catalytic domain consisting of three blocks of conserved sequences known as polo boxes(Glover et al., 1996)(Fig. 1). Studies have been carried out to identify the role of this non-catalytic domain in budding yeast and mammalian cultured cells. In mammalian cells, the C-terminus of mammalian Plk1 alone directs localisation to centrosomes(Jang et al., 2002; Seong et al., 2002). A mutation in polo box 1 abolished the ability of mammalian or budding yeast polo kinase to localise to the mitotic apparatus in budding yeast(Lee et al., 1998; Song et al., 2000) or to complement a budding yeast mutant (Jang et al., 2002; Lee et al.,1998; Song et al.,2000). Also, the effects of overexpression of the kinase in budding yeast or mammalian cells were shown to be dependent on polo boxes(Lee et al., 1998; Seong et al., 2002). However,whether the non-catalytic region consists of functionally separable sub-domains, in particular in the light of multi-functional nature of polo kinase, has not been addressed.

Fission yeast is an excellent model organism for the dissection of the multi-functional nature of polo kinase due to the ease with which it is genetically manipulated and the clarity with which multiple functions are observed. In fission yeast, deletion of the polo kinase gene, plo1⁺, results in three major defects — failure to establish spindle bipolarity, failure to form septa or misplacement of septa(Ohkura et al., 1995; Bahler et al., 1998). In addition, overexpression of the polo kinase results in two clear phenotypes,cells displaying multiple septa and mitotically arrested cells with monopolar spindles (Ohkura et al.,1995). Involvement in entry into mitosis has been suggested through the interaction with the SPB protein Cut12 and the NimA-like kinase Fin1 (Mulvihill et al., 1999; Grallert and Hagan, 2002) and a role in metaphase to anaphase transition via an interaction with the APC/C subunit Cut23 has also been demonstrated for fission yeast Plo1(May et al., 2002). Furthermore, it has been shown that Plo1 is at the top of the septum initiating network (SIN) (Tanaka et al.,2001) demonstrating that many of the conserved functions of polo kinases may be studied using fission yeast as a model organism.

In this study we examined the roles of conserved amino acid sequences in the fission yeast polo-like kinase (Plo1) and in particular their relationship with multiple functions. Functions were explored by overexpression experiments and by stable expression of mutant genes at wild-type levels in a plo1 deletion background. We have shown, using a combination of site directed and a novel random mutagenesis method, that the polo boxes form a single functional unit required for in vivo function, SPB localisation and multiple protein-protein interactions.

General molecular techniques were carried out according to Sambrook et al.(Sambrook et al., 1989). Site directed mutagenesis of plo1 was carried out in pTZ19U (Biorad) by using the Muta-Gene Phagemid in vitro mutagenesis kit (Biorad) according to the manufacturer's instructions. Mutations were confirmed by sequencing of the plo1 gene.

Fission yeast analyses were carried out as described(Moreno et al., 1991; Ohkura et al., 1995). Strains used in this study are shown in Table 1.

In overexpression experiments, plo1 mutants were expressed from the nmt1 promoter in pREP1(Maundrell, 1993). Cells were grown in the presence of thiamine and then washed in sterile water before growth in two parallel cultures, either in the presence or absence of 4 μM thiamine, for 15-16 hours.

For plasmid shuffling experiments, spores were prepared from a diploid strain heterozygous for deletion of plo1 (Sp269) carrying wild-type plo1 on a ura4⁺ marked plasmid[pURplo1⁺ (May et al.,2002)] and mutant plo1 in pREP1. Germination was carried out in the presence of thiamine and plo1Δ haploids selected. The presence of pURplo1⁺ maintains viability of the deletion while thiamine represses expression of the plo1 mutants. The strains were replica plated three times to media containing uracil (both in the presence and absence of thiamine) to allow loss of the pURplo1⁺ plasmid. Cells that had lost the plasmid were selected by using 5-FOA, which kills Ura⁺ cells. Cells were spotted on selective media in the presence or absence of both 5-FOA and thiamine. Abilities of each plo1 mutant to complement a plo1deletion were tested either at approximately wild-type levels of expression (+thiamine, nmt1 repressed) or when overexpressed (- thiamine, ∼100 times that of wild-type expression levels, nmt1 de-repressed).

Spore germination analysis was carried out to determine whether expression of the mutant genes at wild-type level from an integrated copy were able to support growth in the absence of endogenous plo1⁺. HA-tagged plo1 mutants under the control of the attenuated nmt1 promoter (derived from pREP41) were integrated at the leu1 locus of a plo1::his3⁺/plo1⁺ diploid strain(Sp269). Sp269 was created by replacing the entire coding region of one plo1 gene with his3⁺ using a PCR-mediated method(Bahler et al., 1998). Integration was confirmed by PCR using primers 5′ to leu1 and within the polo box domain of plo1. Comparison of the number of His⁺ (plo1Δ) and His^-(plo1⁺) haploids obtained following spore germination revealed whether a particular mutant complemented the plo1disruptant. Full rescue of the deletion phenotype resulted in the ratio of His^- to His⁺ haploids of 1:1. Where no rescue occurred. no His⁺ haploids were found.

Samples were prepared and fixed as described previously(May et al., 2002; Ohkura et al., 1995). Localisation of the Plo1 mutant proteins in vivo was observed in strains expressing GFP-Plo1 under the control of the attenuated nmt1 promoter by autofluorescence while Tat1 antibody(Woods et al., 1989) and Sad1 antibody (Hagan and Yanagida,1995) were used for visualising microtubules and SPB,respectively. Cells were observed with an Axioplan 2 microscope (Zeiss). Images were captured using a CCD camera (Hamamatsu) and processed using Openlab2 (Improvision) and Photoshop (Adobe).

Two-hybrid screening using plo1⁺ as a bait was carried out as described (May et al.,2002) and positive interactors confirmed by isolation and reintroduction of the prey plasmids along with either pBTM116plo1⁺ or pBTM116.

Isolation of plo1 mutants, which retain only a subset of interactions was carried out as follows. For random mutagenesis of plo1, pBTM116plo1⁺ was used as template in a PCR reaction using Taq polymerase without proofreading activity (Roche). Primers used in the reaction were complementary to the 5′ end of plo1⁺ and to the ADH1 terminator sequence, which is 3′ to the plo1⁺ gene in pBTM116plo1⁺. Strain L40 was transformed with gapped pBTM116 plasmid and PCR product. Gap repair in vivo resulted in the recreation of pBTM116plo1 with mutations in plo1(Muhlrad et al., 1992). Each L40 stain carrying mutagenised plo1 bait constructs was mated with Y187 strains carrying prey plasmids. Mating efficiency was determined by growth on selective media and the two-hybrid interaction assessed by expression of the HIS3 reporter gene (by growth on selective media lacking histidine and containing 10 mM 3-AT after 2 days at 32°C). Plasmids were isolated from strains with pBTM116plo1 constructs of interest, re-transformed to confirm the interaction pattern, and sequenced to determine the mutation site.

Polo kinases have two conserved regions. One coincides with an amino-terminal catalytic (kinase) domain and the other is found in the carboxy-terminal non-catalytic region, mainly as three blocks called the polo boxes (Glover et al., 1996)(Fig. 1). In an attempt to define functional domains, we targeted highly conserved residues in both the catalytic and non-catalytic regions of the fission yeast polo kinase, Plo1,for site-directed mutagenesis (Fig. 1) and assayed functions of the mutant genes.

First the mutant genes were overexpressed in wild-type cells under the control of the nmt1 promoter on a multicopy plasmid, pREP1. Overexpression of the wild-type plo1 gene resulted in two major cell defects (Ohkura et al., 1995)(Fig. 2B). One is the induction of untimely septation, which can be seen in cells with multiple septa or mononucleated cells with septa. The other is a failure to establish the bipolarity of the mitotic spindle resulting in mitotic arrest, which can be seen as cells with overcondensed chromosomes and monopolar spindles.

To see whether the non-catalytic domain consists of functionally separable units, we tested various mutants in the polo boxes. These include various point mutations at highly conserved amino acids in polo box 1 (W497F, G505A,YQL508AAA and FN519AA), polo box 2 (L577A) or polo box 3 (DHK625AAA), and series of C-terminal deletions (50 amino acids each; 1-633, 1-583, 1-533, and 1-483). We found that all of the mutations created in the non-catalytic domain show identical effects upon overexpression. They all abolished the mitotic arrest phenotype but had little effect on induction of untimely septation(Fig. 2C,D,K as examples, and data not shown for other mutations). This demonstrates that the overexpression effects of plo1⁺ on spindle formation and septation are separable and that the former is dependent on the integrity of all of the polo boxes but the latter is not. Mutations in any one of the three polo boxes led to an identical effect suggesting that the polo boxes may act as a single functional unit.

On the other hand, mutations in the kinase domain of Plo1 result in the opposite effect. We tested various mutations (K69R, K69Q, D181R, D181N, E193V,and T197V) in conserved residues that are known or thought to be important for the activity of polo kinases. All of the mutants in the catalytic domain resulted in mitotic arrest and, in some cases, lack of septum upon overexpression (Fig. 2E,L). Immunostaining indicated that the mitotic arrest was associated with monopolar spindles in all kinase domain mutants (Fig. 2G-I). As one of the mutants (Plo1K69R) has been shown to have no or little kinase activity (Tanaka et al.,2001), this is likely to be due to dominant negative effects caused by overexpression of inactive (or less active) kinase. To test whether the mitotic arrest was due to activation of the spindle checkpoint, Plo1K69R was overproduced in the checkpoint-defective mad2 deletion mutant. Inactivation of the spindle checkpoint completely abolished the mitotic arrest upon overexpression of kinase domain mutants, and instead gave rise to cells with large nuclei probably resulting from a continuation of the cell cycle without nuclear division (Fig. 2L).

We also examined whether this dominant negative effect is dependent on intact polo boxes. Mutations in either polo box 1 (YQL508AAA) or 3 (DHK625AAA)completely abolish the effects of overexpression of an inactive kinase Plo1K69R (Fig. 2F), confirming the importance of polo boxes for the plo1 function.

To determine whether the plo1 mutants we created retain function in vivo, we first tested complementation of a plo1 disruptant using plasmid shuffling. In these experiments, we first constructed a plo1disruptant (plo1::his3⁺) which contains a wild-type copy of plo1⁺ (pURAplo1⁺) on a ura4⁺ marked plasmid and mutant plo1(pREP1plo1*) on a LEU2 marked plasmid. We then assayed the ability of pREP1plo1* to support the growth of the plo1disruptant when pURAplo1⁺ is lost (selectable by 5-FOA which kills Ura⁺ cells).

The wild-type plo1 gene was able to fully support the growth of a plo1 disruptant. Various mutants in the kinase domain which we tested(Plo1K69R, K69Q, D181R, T197V, and E193V) supported little or no growth of the disruptant. Mutants in the non-catalytic domain also failed to support the growth of the disruptant (W497F, G505A, YQL508AAA, FN519AA, YM572AA,DHK625AAA, 1-633 and 1-583).

In the above experiments it was not possible to control expression levels accurately or to observe cytological phenotypes. Therefore we examined whether expression of these mutants at wild-type levels from an integrated copy can support growth of a plo1 disruptant. The viability of plo1disruptants carrying integrated HA-tagged mutant plo1 under the control of nmt41, which results in expression at a level comparable to the native promotor (Mulvihill et al.,1999) (Fig. 4E),were assayed through spore germination (see Materials and Methods). Wild-type plo1 expressed in this way was able to fully complement the lethality of the disruptant. On the other hand, complementation was abolished by mutations in the kinase domain (K69R, D181R, D181N, T197V and E193V) or the non-catalytic domain (YQL508AAA, DHK625AAA, 1-633, 1-583 and 1-533). Cytological examination of germinating spores indicated that 50%-90% of disruptant cells expressing these mutant genes exhibited similar defects to those seen in a plo1 disruptant(Fig. 3A,B). These include septation defects (either a lack of septation or the formation of defective septa) and mitotic arrest. Multiple septation or untimely septation, which is typical upon overexpression of plo1⁺, was not observed. As a control, less than 5% of disruptant cells expressing the wild-type gene show abnormalities upon spore germination. It is clear then that all three polo boxes of Plo1, in addition to the catalytic domain, are required for cellular function of the protein. The fact that mutations in different polo boxes resulted in the same consequences is consistent with the idea that polo boxes together form one functional unit.

The effect of the mutations on the cellular localisation of Plo1 protein was tested in vivo by GFP tagging of the mutant proteins at the N-terminus. Mutated plo1 genes fused in frame with the GFP gene were expressed under the control of an attenuated nmt1 promoter from a copy integrated at the leu1 locus of wild-type cells. Wild-type Plo1 tagged in this way is expressed at a level comparable to the endogenous protein and displays cell cycle localisation identical to the endogenous protein (Mulvihill et al.,1999). In the case of catalytically inactive Plo1 (Plo1K69R),localisation has been shown to be unaffected by the mutation(Tanaka et al., 2001).

There were no differences observed between localisation of kinase domain mutants (GFP-Plo1K69R and T197V) and wild-type protein. Like wild-type(Fig. 4A), GFP-Plo1K69R and GFP-Plo1T197V localised initially to unseparated SPBs and remained there before becoming weaker during anaphase(Fig. 4B). This prompted us to examine whether the entire catalytic domain is dispensable for cell-cycle regulated localisation. We constructed a strain expressing GFP fused to the non-catalytic domain of Plo1 protein (GFP-Plo1.313-683). GFP-Plo1.313-683 localises to the SPBs in a similar manner to GFP-Plo1. To follow this localisation pattern as cells progress through the cell cycle, newly-divided short cells (early G2 cells) were collected by centrifugal elutriation. These synchronised cells were cultured and samples taken at regular time intervals. Like wild-type Plo1, GFP-Plo1.313-683 accumulates at SPBs in late G2 or very early mitosis before separation of SPBs (where only one SPB signal is visible)and remains on SPBs as they separate (two SPB signals) until late anaphase(Fig. 4D). Unfortunately, as a protein consisting of just the polo box domain of Plo1 (GFP-Plo1.472-683) was not detected by immunoblotting (Fig. 4E), it could not be determined if this region of the protein alone is sufficient for cell-cycle localisation to the SPBs. However, these results indicate that the non-catalytic domain of Plo1 is sufficient for cell-cycle-regulated localisation to the SPBs.

In contrast, localisation to the SPBs was abolished in all of the non-catalytic domain mutants tested (W497F, YQL508AAA, DHK625AAA, 1-633,1-583, 1-533). This was not due to differences in the amount of GFP-Plo1 proteins, as all except Plo1.1-483 and Plo1.472-683 (not included in our cytological analysis) were detected at comparable levels to that of untagged Plo1 by immunoblotting using an antibody against the non-catalytic domain of Plo1 (Fig. 4E).

In summary, SPB localisation is dependent on the polo boxes and the non-catalytic domain of Plo1 is sufficient for its cell-cycle-regulated localisation.

Protein protein interactions may play an important role in Plo1 function. To identify proteins potentially interacting with Plo1, we carried out a two-hybrid screen of a mitoitc S. pombe cDNA library using full length Plo1 as a bait. Positive two-hybrid interactors included known interactors (cut23 and dmf1/mid1), genes previously described in another context [sum2, sck1 and abp2(Forbes et al., 1998; Jin et al., 1995; Sanchez et al., 1998)] and previously uncharacterised genes (Table 2). cut23 and dmf1/mid1 encode a subunit of the anaphase promoting complex, and a protein which localises to the pre-division site and is required for correct septum positioning, respectively. Both gene products have been shown to functionally and physically interact with Plo1(Bahler et al., 1998; May et al., 2002). sum2,sck1 and abp2 encode a protein which may have a role in G2/M transition, a non-essential protein kinase and a putative ARS-binding protein,respectively (Forbes et al.,1998; Jin et al.,1995; Sanchez et al.,1998). The following three genes are previously uncharacterised but encode proteins which have structural motifs or limited homologies to other proteins — SPAC1006.03c (containing predicted coiled-coil regions), SPAC6B12.08 (with a Dna-J domain) and SPAC26H5.05(containing ankyrin repeats).

Cut23 and Dmf1/Mid1 have been shown to interact with Plo1 through the non-catalytic domain (Bahler et al.,1998; May et al.,2002). To identify the region of Plo1 that mediates the interaction with each interactor we have isolated, we tested each of them against various Plo1 mutants in a directed two-hybrid assay. Without exception, these interactors were all able to interact with the dead kinase mutant (Plo1K69R) and Plo1 lacking the entire kinase domain (Plo1.313-683). On the other hand, any of the mutations in the non-catalytic domain that we tested (YQL508AAA, DHK625AAA, 1-633, 1-583, 1-533) abolished the interaction with all of the two-hybrid interactors(Table 2). This indicates that the non-catalytic domain is sufficient for the two-hybrid interactions, and that the integrity of the polo boxes is essential for interaction with all of the proteins identified in our screen. This suggests that the polo boxes together form a domain which interacts with multiple proteins.

Our site-directed mutagenesis of conserved amino acids indicated that the polo boxes are essential for the interaction with all of the proteins we examined. It is not clear, however, whether the polo boxes determine which proteins interact with Plo1. To gain an insight into this issue, we attempted to isolate mutations in plo1 which specifically disrupt interactions with only a subset of proteins by random mutagenesis.

The entire plo1 gene was randomly mutagenised by PCR and cloned into a bait plasmid in yeast by gap repair. In this experiment, the mutation rate assayed by sequencing was roughly 1 point mutation in every 1 kb. These mutant genes were simultaneously tested for interaction against four of the two-hybrid interactors (cut23, dmf1/mid1, SPAC1006.03c, and SPAC6B12.08) by mating individual yeast strains containing mutant plo1 with other yeast strains containing different prey plasmids (see details in Materials and Methods).

Of 1035 potential mutants tested, 60% were positive for interaction with all (except the empty activation domain vector control), 29% did not interact with any of the prey constructs and 11% displayed differential interactions(i.e., interact with some but not others). Among mutants which displayed differential interactions, the protein interaction profiles did not show a tendency for any two of the interactors behave similarly.

To determine which residues are responsible for the specificity of the protein interactions, we sequenced some of the mutants which display differential interactions. All of these mutations mapped within or close to the polo boxes, except those disrupting the interaction with SPAC6B12.08(Fig. 5B), indicating that the polo boxes play a crucial role in determining protein interactions. At least three mutations which disrupt interaction with Cut23 mapped in three different polo boxes, confirming the view that the polo boxes together form one domain. Those mutations which disrupt the interaction with SPAC6B12.08 mapped in a cluster of residues in subdomain X of the catalytic domain (K251E, I252T and S256P). As the entire catalytic domain including subdomain X is dispensable for the interaction with SPAC6B12.08, this may be due to stereo-hindrance caused by a structural change.

We hoped that mutations which differentially affect protein interactions might disrupt a subset of plo1 functions in vivo. We tested complementation of a plo1 disruptant by expression of these mutant genes from an integrated copy in the genome. None of these mutants were able to fully support the growth of a plo1 disruptant. Further cytological analysis did not reveal defects specific to each of the mutants, perhaps reflecting the fact that Plo1 is likely to interact with a number of other proteins in vivo and that in no case merely a single interaction was compromised by the mutations. Nevertheless, our screen for differential two-hybrid interactions highlights the importance of the polo boxes for determining protein interactions.

The role of the polo boxes has been studied previously by site-directed mutagenesis in budding yeast and mammalian cultured cells(Lee et al., 1998; Lee et al., 1999; Seong et al., 2002; Song et al., 2000). In these studies, only one or a few mutations were made and a single function was assayed in vivo. Therefore it was unclear whether the non-catalytic domain containing the polo boxes consisted of functionally separable domains. In our mutational study we address this question by making a series of point mutations and truncations in the non-catalytic region, in particular at conserved residues within the polo boxes. The functions or activity of these mutants were examined by four assays — overexpression effects (two functions could be determined), complementation of a disruptant (three functions), localisation and protein interaction (with multiple interactors). All of these assays show that various mutations in the polo boxes and serial truncations from the carboxy terminus produce indistinguishable results,indicating that the polo boxes form one single functional domain.

Our functional study is consistent with a very recent structural study of the Sak polo-box domain (Leung et al.,2002). It suggests that this part of the polo-box domain(equivalent to polo box 1 and 2) autonomously folds and can interact with the second polo box domain (equivalent to polo box 3 and the C-terminal tail) to form a putative ligand-binding domain.

Upon overexpression of plo1, mutations in the polo boxes do not affect induction of septation but abolish the ability to interfere with spindle formation. This is the first example that polo box mutations exhibit distinct effects on two functions of polo-like kinase. The opposite effects of mutations in either the kinase domain or in the polo boxes upon overexpression of plo1 indicate that the polo boxes are not simply required for kinase activity. In mammalian cells, carboxy terminal truncations of polo-like kinase resulted in an increase in overall kinase activity(Jang et al., 2002; Mundt et al., 1997). Our preliminary results confirmed that polo box mutants retain cell cycle regulated kinase activity in fission yeast (N.R. and H.O., unpublished).

In contrast to the overexpression assay, replacement of wild-type plo1 gene by mutants indicated that the polo boxes are essential for at least three detectable functions of Plo1 kinase in vivo. These functional studies suggest that the catalytic and non-catalytic domains work in concert but that the requirement for the polo boxes is not simply a requirement for catalytic activity.

So what is the role of the polo boxes? It has been shown in S. cerevisiae, S. pombe, X. laevis, D. melanogaster and mammalian cultured cells that polo kinases localise to the centrosomes/SPBs in a cell cycle regulated manner and that mutations in the polo boxes abolish this localisation (Bahler et al.,1998; Golsteyn et al.,1995; Lee et al.,1998; Logarinho and Sunkel,1998; Moutinho-Santos et al.,1999; Mulvihill et al.,1999; Shirayama et al.,1998; Song et al.,2000). In mammalian cultured cells, the non-catalytic domain alone has been shown to be sufficient for the localisation(Jang et al., 2002).

Consistently, in fission yeast it has been shown that a kinase inactive mutant can localise to the SPBs (Tanaka et al., 2001). Here we show that the polo boxes are essential, and the non-catalytic domain is sufficient for cell-cycle regulated localisation of Plo1 to the SPBs. Therefore at least one molecular role of the polo boxes is to form an autonomous domain which directs cell cycle regulation of SPB localisation.

Although this and previous studies have identified a role for the polo box domain in localisation to the SPB/centrosome, our study suggests that the polo box domain is likely to play a more general role, which is to mediate interaction with multiple proteins. It has been suggested that localisation is required for polo kinase function based on the observation that mutations in the polo boxes disrupt both localisation and in vivo function(Lee et al., 1998; Song et al., 2000). In the light of our findings, more caution is necessary to interpret these results as polo box mutations simultaneously disrupt interaction with many proteins.

Site-directed mutagenesis in any individual polo box disrupts all of the functions and protein interactions of Plo1 kinase that we have examined. Therefore it is unlikely that each polo box forms a distinct subdomain which interacts with a different set of proteins. Most likely, the polo boxes together form one protein interaction domain.

Then how do the polo boxes participate in protein-protein interactions? It is possible that the polo boxes directly recognise interacting proteins and thereby determine specificity. Alternatively they may simply facilitate folding of the domain to allow intervening sequences to recognise target proteins. We have isolated mutations which disrupt only a subset of protein interactions. These mutations mapped mostly within or close to the polo boxes,despite the fact that the polo boxes occupy less than 20% of the non-catalytic domain. Therefore it is possible that the polo boxes play a crucial role in determining the specificity of protein interaction.

The two-hybrid interactors that we isolated include two previously identified interactors, Dmf1/Mid1 and Cut23. Dmf1/Mid1 is a medial ring protein required for positioning of the division site and Plo1 has been shown to interact with it and to be required for its localisation(Bahler et al., 1998). Fission yeast Cut23 is a subunit of the APC/C which interacts with Plo1, and a mutation in Plo1 which compromises that interaction fails to activate APC mediated proteolysis (May et al.,2002). Therefore at least some of the two-hybrid interactors we isolated have strong functional connections with Plo1 kinase.

Although the other two-hybrid interactors we have isolated have not yet been shown to have a clear functional relationship with Plo1 kinase, our preliminary results indicate that at least one of them (SPAC1006.3c) are indeed co-immunoprecipitated with Plo1 (N.R. and H.O., unpublished). Moreover,some studies, although limited, may suggest possible connections between some of the interactors and Plo1 function. For example, sum2⁺(suppressor of uncontrolled mitosis) is implicated in the G2/M transition, as it was originally isolated as a suppressor of cdc25⁺overproduction (Forbes et al.,1998). Abp2 was originally identified as a putative ars binding protein, but the deletion mutant shows aberrant chromatin and septal structures and fails to arrest cell cycle when replication is inhibited(Sanchez et al., 1998). Further detailed study of these two-hybrid interactors will reveal the significance of these interactions.

In other organisms, several mitotic proteins, such as human TCTP, Drosophila Asp, Xenopus Cdc25C, budding yeast septins and tubulins, have been shown to physically interact with polo kinase. TCTP is a microtubule associated protein which is phosphorylated by Plk1 kinase. Overexpression of a non-phosphorylatable form disrupts nuclear division(Yarm, 2002). Asp is another microtubule associated protein which is implicated in microtubule assembly from centrosomes (Gonzalez et al.,1990; Wakefield et al.,2001). Asp interacts physically with polo kinase and phosphorylation by polo kinase is required for its activity(do Carmo Avides et al., 2001; Gonzalez et al., 1998). Septins are required for cytokinesis in budding yeast and have been shown to interact with the budding yeast polo kinase Cdc5p both physically and functionally (Song and Lee,2001). In most cases, these interactions were mediated entirely through the non-catalytic domain with the exceptions of the tubulins(Feng et al., 1999) and GRASP65(Lin et al., 2000). Therefore protein interactions through the non-catalytic domain are likely to play a crucial role for polo kinase function in general.

Then what are the roles of these protein-protein interactions? It is possible that some of the interactors are substrates of polo kinase. As the region of the protein required for protein-protein interactions is separate from the catalytic domain, this is unlikely to be a simple substrate/kinase interaction. Rather, it is likely that physical interactions via the polo box domain act as a `docking' mechanism to enhance the efficiency of substrate recognition. If the role of an interaction is in docking, the interactor does not have to be a direct substrate of polo kinase. The interactors can act as`adaptors' which bring substrate and kinase together by interacting with both polo kinase and particular substrates.

It is also possible that these interactors may act as regulators to influence kinase activity directly, either positively or negatively. Indeed polo kinase is catalytically activated in a cell cycle regulated manner. However, our preliminary results suggest that Plo1 kinase which has mutated polo boxes still exhibits cell cycle regulation, suggesting its catalytic activity is regulated in other ways.

Polo kinases exhibit multiple functions at different stages of mitosis. Cell cycle regulation of kinase activity alone may not be sufficient to achieve this complex task. Interaction with multiple mitotic regulators may provide means for complex temporal and spatial regulation of polo kinase,perhaps via independent control of interaction with individual proteins. Therefore the characterisation of these interactors and an analysis of their mode of interaction will be crucial to understanding the function and regulation of polo kinase in vivo.

We thank Stuart MacNeill, Alison Crawford and Vladimir Nekrasov for help with elutriation, cell cycle analysis of Plo1 localisation and two-hybrid screening, and Kevin Hardwick and Ken Sawin for strains. We also thank Stuart MacNeill for critical reading of the manuscript and other members of Ohkura laboratory for discussion and encouragement. The work was funded by The Wellcome Trust. N.R. held a Wellcome Trust Prize PhD Studentship, and H.O. holds a Wellcome Trust Senior Research Fellowship.