Schizosaccharomyces pombe protein phosphatase 1 in mitosis, endocytosis and a partnership with Wsh3/Tea4 to control polarised growth

Reversible protein phosphorylation is one of the most widespread mechanisms by which eukaryotes modulate the activity of biological pathways. Although the fidelity of this signalling invariably relies upon the reversibility of the reaction, most genomes contain around 20 times fewer Serine/threonine (Ser/Thr) protein phosphatases than Ser/Thr protein kinases (Manning et al., 2002). The diversity amongst these kinases generally arises from the fusion of a particular type of catalytic domain with a distinct series of motifs that impart a unique set of properties to each member of a particular kinase family. By contrast, diversity amongst protein Ser/Thr phosphatases generally arises from the generation of holoenzymes with unique properties via the recruitment of generic catalytic subunits to a range of regulatory subunits at particular locations (Ceulemans et al., 2002). The physico-chemical properties of these catalytic subunits have been used to classify Ser/Thr phosphatases into four major classes; PP1, PP2A, PP2B, PP2C (Ingebritsen and Cohen, 1983). Type 1 protein phosphatases (PP1) have been linked, via the action of more than 50 regulatory subunits, to control events as diverse as cell cycle control and glycolysis (reviewed by Ceulemans and Bollen, 2004). Whereas mammalian cells contain four isoforms of the PP1 catalytic subunit (α, β, γ₁, γ₂ and δ), fission yeast contains two and the budding yeast just one (Ceulemans and Bollen, 2004; Ohkura et al., 1989). The unicellular lifestyle, sophisticated genetics and systematic approaches to study phosphorylation (Ptacek et al., 2005) makes yeasts particularly attractive models for the study of PP1 function (Stark, 1996).

Fission yeast cells grow by actin-mediated cell extension (Hayles and Nurse, 2001; Marks and Hyams, 1985; Mitchison and Nurse, 1985). The point at which this extension occurs is determined by a number of cell polarity determinants that are delivered to the cortex by interphase microtubules (Hagan, 1998; Hayles and Nurse, 2001). The founder member of this class of proteins is the kelch domain protein Tea1 that associates with Mal3, Tip1, Tea2 and Wsh3 (also known as Tea4; hereafter referred to as Wsh3/Tea4) (Browning et al., 2003; Browning et al., 2000; Busch and Brunner, 2004; Mata and Nurse, 1997). Mal3 and Tip1 are the fission yeast homologues of mammalian EB1 and CLIP170, respectively (Beinhauer et al., 1997; Brunner and Nurse, 2000). Tea2 is a kinesin that stabilises microtubules and is required for efficient delivery of Tea1, Tip1 and Mal3 to cell tips (Browning et al., 2003; Browning et al., 2000). Once delivered to cell tips, Tea1 relies upon a Tip-associated CAAX box protein, Mod5, to bind to the cell cortex at the tip (Snaith and Sawin, 2003). It is assumed that the subsequent association of Tea1 with Bud6, For3 and Wsh3/Tea4 promotes actin polymerisation to maintain polarised, linear growth of the rod shaped cells (Feierbach and Chang, 2001; Feierbach et al., 2004; Glynn et al., 2001; Martin et al., 2005; Tatebe et al., 2005). Wsh3/Tea4 contains a RVXF PP1 binding consensus motif (Meiselbach et al., 2006) and provides a physical and functional link with the stress response signalling cascade, so that cells that lack Wsh3/Tea4 are unable to control cell polarity following stress (Tatebe et al., 2005).

The polarity determinants operate at a second level of polarised growth. Following medial fission, daughter cells grow from the end that existed in the previous cell cycle, while the end that was generated by cytokinesis lies dormant until a discrete point known as new end take off (NETO) when this `new' end starts to grow (Mitchison and Nurse, 1985). NETO is trigged by the attainment of a critical cell volume and passage through S phase to an ill-defined point within G2 phase. This second level in the polarisation of cell growth also relies upon the cytoskeletal polarity determinants Tea1, Tea3, Bud6 and For3 alongside protein kinases such as Pom1 (Arellano et al., 2002; Bähler and Pringle, 1998; Feierbach and Chang, 2001; Glynn et al., 2001; Mata and Nurse, 1997).

Here we report the distribution of the two PP1 isoforms in fission yeast to new subcellular locations and demonstrate that the recruitment of Dis2 to cell tips by Wsh3/Tea4 is important for the regulation of the microtubule-mediated polarised tip growth.

In order to identify the location of Dis2 and Sds2 in the cell, sequences encoding EGFP were fused, in frame, to their genes at their native loci using the `marker switch approach' (MacIver et al., 2003a). Deletion of both PP1 encoding genes is lethal, whereas deletion of either alone is not (Ohkura et al., 1989). We therefore tested the functionality of each tagged allele by attempting to generate double mutants in which each PP1–tag gene fusion would be the sole source of PP1. An inability to generate such tag-deletion double mutants would indicate that fusion of the tag to the ORF had compromised gene function. sds21 and dis2 alleles in which EGFP was fused to the 3′ end of the gene could not act as the sole source of PP1 (data not shown). By contrast, the fusion of a single EGFP molecule and a spacer of three consecutive alanine residues to the N terminus of dis2⁺ and sds21⁺ did not affect the viability, morphology or generation time (data not shown). The actin and microtubule cytoskeletons of these strains were indistinguishable from the relevant dis2⁺ sds21.Δ or dis2.Δ sds21⁺ control strains (data not shown).

Imaging dis2.NEGFP sds21⁺ interphase cells established that Dis2.NEGFP fusion proteins were recruited to the nucleus, a bright spot on the nuclear periphery, numerous cytoplasmic dots at the cell periphery and zones at cell tips (Fig. 1A, supplementary material Movie 1). Double labelling with Hoechst 33342 to identify the hemispherical chromatin region (Toda et al., 1981) confirmed that Dis2.NEGFP gave a general nuclear stain rather than enrichment in a particular sub-domain of the nucleus (Fig. 1B). During mitosis the strong dot at the nuclear periphery was replaced, until anaphase, by up to six spots within the main body of the nucleus (Fig. 1C upper insets, supplementary material Movie 2). During anaphase, a single spot was seen at each end of the separating nuclei to persist as a nuclear peripheral dot in the next cell cycle (Fig. 1C lower insets). Dis2.NEGFP accumulated at the cell equator in early anaphase (Fig. 1A,C green arrow, supplementary material Movie 2).

Centromeres associate with the spindle pole body (SPB) throughout interphase. In mitosis they are transiently released before being pulled by microtubules to the two spindle poles. In anaphase they re-establish their affiliation with the interphase SPBs (Funabiki et al., 1993). As aspects of Dis2.NEGFP distribution mimic centromere behaviour and PP1 colocalises with kinetochores and centromeric proteins in other systems (Bloecher and Tatchell, 2000; Trinkle-Mulcahy et al., 2003), we imaged the green signal of Dis2.NEGFP alongside red signals arising from Pcp1.RFPC or Cnp1.Cherry to mark SPBs and centromeres, respectively (Fig. 2A-D). We were able to capture images in which the bright Dis2.NEGFP signals formed discrete dots adjacent to the SPB signal (Fig. 2A,B). Rapid movements of the SPB made it a challenge to capture the signal in two distinct fluorescence channels without any movement between the consecutive images. Depolymerisation of the microtubule cytoskeleton with carbendazim (CBZ) to block SPB-mediated nuclear migration (Hagan and Yanagida, 1997) suppressed this movement and enabled us to establish that, although the Pcp1.RFPC and Dis2.NEGFP signals were close, they very rarely overlapped, indicating that the two fluorochromes were associated with distinct structures (Fig. 2B).

The association of Dis2.NEGFP signals with SPBs was exceptionally rare from the commitment to mitosis until anaphase when they were seen just inside the nucleus adjacent to the SPB (data not shown). By contrast, the centromere marker overlapped with the Dis2 signal at all stages of the cell cycle (Fig. 2C,D). Chromatin immunoprecipitation (ChIP) established that Dis2.NPk associated with centromeric sequences. Sequences corresponding to the central core non-repetitive domains of the centromere that promote the formation of the kinetochore were enriched in immunoprecipitates whereas those in the flanking outer or inner repeats were not (Fig. 2E). Association with protein complexes in the central region is consistent with the inability of Dis2.NEGFP to colocalise with centromeres when an inner centromere component that is required for the loading of the centromere-specific histone H3 variant CenpA to the centromere, Mis6 is defective (Saitoh et al., 1997; Takahashi et al., 2000). mis6.302 dis2.NEGFP pcp1.RFPC cells were incubated at 36°C for 6 hours. Almost three-quarters of interphase cells (73%) did not have any punctate signals that would correspond to Dis2.NEGFP at centromeres near the SPB or dispersed throughout the nucleoplasm (data not shown). Thus, not only do the Dis2.NEGFP foci that associate with interphase SPBs and strong mitotic Dis2.NEGFP dots colocalise with centromeres, but they are disrupted upon disruption of centromere function. We conclude that these structures represent Dis2.NEGFP recruitment to centromeres.

Signals from the Sds21.NEGFP differed from those from Dis2.NEGFP in three respects (Fig. 3A). First, although Sds21.NEGFP was a nuclear protein it was enriched in the non-Hoechst 33342 staining nucleolus (Toda et al., 1981). Second, the Sds21.NEGFP nuclear signal was considerably fainter than that from Dis2.NEGFP (supplementary material Fig. S2B). Third, sds21.NEGFP expressing strains did not show any punctate signals in the cytoplasm or give the characteristic spot on the nuclear periphery that results from centromere association (Fig. 3A).

Although Sds21.NEGFP and Dis2.NEGFP have strikingly different distributions, cells that lack either molecule are viable and, yet, those that lack both are dead (Ohkura et al., 1989), suggesting either that the association of Dis2.NEGFP with some of the structures is not essential, or that the removal of one isoform enables the other to substitute and assume functions that were normally executed by the missing phosphatase. We therefore examined the distribution of each phosphatase in strains from which the second isoform had been deleted (Fig. 3B,C). Dis2.NEGFP distribution and signal intensity did not change when sds21⁺ was deleted (Fig. 3C,E). Significantly, there was no enhancement of the staining of the nucleolar region that might be expected if Dis2 were now to be recruited to sites formerly occupied by Sds21 (Fig. 3C,E). By contrast, deletion of dis2⁺ from sds21.NEGFP led to the incorporation of Sds21.NEGFP into all of the locations that would otherwise have been occupied by Dis2, with the exception of chromatin and the staining around the cell equator during division (Fig. 3B and supplementary material Fig. S1). Furthermore, when we marked sds21.NEGFP dis2⁺ cells by transient immersion in red fluorescent lectin before imaging them alongside sds21.NEGFP dis2.Δ we found that deletion of dis2⁺ prompted an increase in the intensity of the Sds21.NEGFP signal (Fig. 3D). Western blotting established that this increase in signal was due to an increase in protein level (Fig. 3F and supplementary material Fig. S2). Such an increase in Sds21 levels and recruitment to new sites suggested that Sds21 is competent to associate with targeting molecules that have a much greater affinity for Dis2 when Dis2 is present.

Treatment with the microtubule depolymerising drug carbendazim (CBZ) altered Dis2.NEGFP distribution (Fig. 4, supplementary material Movies 3, 4). The inclusion of nmt81.GFP.atb2 expressing cells (asterisks Fig. 4, supplementary material Movies 3, 4) (Garcia et al., 2001) alongside dis2.NEGFP cells enabled us to establish that microtubules had been destroyed by CBZ treatment because no discrete GFP signals were visible in the nmt81GFP.atb2 cells (Fig. 4C,D). Time-lapse series following microtubule depolymerisation revealed that cytoplasmic Dis2.NEGFP foci could be divided into two distinct subpopulations that differed in response to CBZ. Slightly larger foci that formed a cap at the cell tips (inset in Fig. 4B, blue brackets in Movies 1 and 3 in supplementary material) were lost following microtubule depolymerisation (inset Fig. 4D, supplementary material Movie 4), whereas the more numerous dots that were seen further back from the tips throughout the cytoplasm were unaffected by microtubule disruption (Fig. 4, compare insets of B and D, supplementary material Movies 3, 4).

Imaging single focal planes revealed that the small foci of Dis2.NEGFP appeared at the cortex and moved a short distance into the cytoplasm before disappearing again (supplementary material Movie 1, yellow brackets and arrows). Continual imaging of three consecutive z sections of 0.2 μm over short intervals established that the disappearance of dots reflected a genuine dispersal or disassociation of Dis2.NEGFP from these structures or dissolution of the structures rather than movement out of a particular focal plane (supplementary material Movie 5, data not shown). Generating kymographs by compressing the region delineated by green bars in Fig. 5A to give a single slice and then lining up consecutive slices one after another, established that dots persisted for an average of 7.9 seconds (a total of 61 dots monitored).

Inward migration of spots is reminiscent of that of endocytic proteins such as Sla2, Myo1 and Arp2/3 (Castagnetti et al., 2005; Sirotkin et al., 2005). We therefore fused sequences encoding the red fluorescent protein `Cherry' (Shaner et al., 2004) to the 3′ end of the end4⁺ (also known as sla2⁺) gene that is required for endocytosis (Iwaki et al., 2004) and recorded consecutive Sla2.Cherry and Dis2.NEGFP signals from the same cell (Fig. 5B). The strong accordance between the two signals is consistent with the two proteins associating with each other or with a common structure. Moreover, efficient movement of Dis2.NEGFP foci relied upon Sla2 function because 67% of Dis2.NEGFP foci (n=239) appeared at the cell cortex, but failed to migrate towards the cell centre, in the temperature sensitive sla2.Δ background at 33°C [Fig. 5C; 100% of foci (n=221) migrate towards the cell centre before disappearing in sla2⁺ cells; data not shown]. Furthermore, attempts to combine sla2.Δ with the dis2.11 mutant failed, indicating that the presence of the two mutations in the same cell conferred synthetic lethality (data not shown). We conclude that the small dots are likely to represent an association with endocytic vesicles to regulate some aspects of early stages of endocytosis. We refer to these dots as Dis2.NEGFP endocytic associated foci (DEAF).

As actin plays a key role in endocytosis, mixed cultures of dis2.NEGFP or dis2.NEGFP sla2.Cherry and crn1.GFP (Pelham, Jr and Chang, 2001) were treated with latrunculin A (Lat A) to depolymerise F-actin. This treatment abolished the crn1.GFP signal, indicating that it effectively removed the F-actin cytoskeleton (asterisk in Fig. 5E). In the neighbouring cells Lat A treatment froze the migration of the Dis2.NEGFP and Sla2.Cherry DEAFs at the cortex and they failed to migrate into the cytoplasm (Fig. 5E,F). The distribution of the tip-associated Dis2.NEGFP foci was unaffected by disruption of the actin cytoskeleton. However, fewer DEAFs than in unperturbed cells appeared throughout the cytoplasm and now localised in the cell tip region (insets in Fig. 5D,E).

The disappearance of foci from the cell tips upon the disruption of the microtubule cytoskeleton (Fig. 4D, supplementary material Movie 4) prompted us to investigate the relationship between this population of Dis2.NEGFP and polarised tip growth. Staining cells with fluorescent lectin to stain the cell wall can differentiate between growing and non-growing tips if cells are returned to the cell culture to resume growth and generate non-fluorescent, freshly grown zones (May and Mitchison, 1986). It can therefore be used to stage cells relative to the NETO transition in G2 phase (Mitchison and Nurse, 1985). Transient lectin staining and live cell imaging established that Dis2 associated with cell tips irrespective of whether they were growing or not (Fig. 6A,B). Furthermore, both ends of cells in which cell cycle progression had been arrested before NETO, with the cdc10.129 mutation, recruited Dis2.NEGFP to the same degree as those arrested at the G2-M boundary, after NETO, with the cdc25.22 mutation (Fig. 6C,D).

Because the dependence of the tip association of Dis2.NEGFP upon an intact microtubule cytoskeleton mirrors that of polarity determinants such as Tea1, we assessed the impact of abolition of the function of polarity determinants upon Dis2.NEGFP distribution. Dis2.NEGFP associated with the tips of tea3.Δ, pom1.Δ and bud6.Δ cells (Fig. 7A-C) but failed to form caps at the tips of tea1.Δ, mal3.Δ, tip1.Δ and tea2.1 cells (Fig. 7D-G). As microtubules do not reach the tips of mal3.Δ, tip1.Δ and tea2.1 cells (Beinhauer et al., 1997; Browning et al., 1998; Brunner and Nurse, 2000; Verde et al., 1995), the lack of a Dis2.NEGFP cap at the tips of these cells could simply be due to the inability of microtubules to reach these tips. By contrast, microtubules often fail to stop elongating when they reach the tips of tea1.Δ cells (Mata and Nurse, 1997) raising the possibility that it is simply the incorrect behaviour of microtubules at cell tips that affects Dis2.NEGFP association with cell tips. We therefore examined Dis2.NEGFP distribution in peg1.1 mutants in which the deficiency in CLASP function compromises the termination of microtubules at cell tips so that they also curl around cell tips without affecting Tea1 recruitment (Grallert et al., 2006). Dis2.NEGFP association with cell tips was normal (data not shown). We therefore conclude that Tea1 function is required for the association of Dis2.NEGFP foci with cell tips to form the cap-like structure that is typical of wild-type cells.

DEAFs were distributed between cell tips and either side of wild-type nuclei (Fig. 1A, Fig. 8B). Whereas there was a moderate bias towards one end of wild-type cells (presumably the old end), there was a striking polarisation of DEAFs in cdc10.129 cells that failed to enter the cell cycle and so arrested before NETO (Fig. 6C, Fig. 8B). Strikingly, DEAFs were greatly enriched on the old, growing end of tea1.Δ cells (Fig. 7E, Fig. 8). Such hyper-polarisation was not seen in tea3.Δ, bud6.Δ, mal3.Δ or tea2.1, although a slight enrichment on one side was observed in pom1.Δ and tip1.Δ.

Budding yeast PP1, Glc7, is recruited to bud tips by the conserved Bud14 protein (Knaus et al., 2005). The fission yeast homologue, Wsh3/Tea4, contains the PP1 binding consensus sequence RVXF (Knaus et al., 2005; Meiselbach et al., 2006), binds Tea1 and the formin For3 at cell tips, and is required to maintain cell polarity during unperturbed cell cycles and following osmotic stress (Martin et al., 2005; Tatebe et al., 2005). Imaging dis2.NEGFP wsh3⁺ cells that had been labelled with red lectin alongside unlabelled dis2.NEGFP wsh3.Δ cells revealed a cap of Dis2.NEGFP fluorescence in the wsh3⁺ but not wsh3.Δ background (Fig. 9A, supplementary material Movie 5). Depolymerisation of microtubules abolished the fluorescent cap in both backgrounds (Fig. 9A right panel). The colocalisation of Dis2 and Wsh3/Tea4 at cell tips and the ability to co-immunoprecipitate Dis2.NEGFP with epitope-tagged Wsh3 supported the prediction that Dis2 is recruited to cell tips by Wsh3/Tea4 (Fig. 9B, Fig. 10A, supplementary material Movie 6). Mutation of amino acids 223 and 225 of Wsh3/Tea4 (the V and F residues of the RVXF motif, respectively) to alanine blocked both the ability to co-immunoprecipitate Dis2 with Wsh3/Tea4 and the recruitment of Dis2.NEGFP to tips (Fig. 10A,B and supplementary material Fig. S3) indicating that Wsh3/Tea4 mirrors the relationship between Glc7 and Bud14 in recruiting Dis2.NEGFP to cell tips. Dis2.NEGFP generally associated with Wsh3.C2tdTom foci after they reached the cell tips, however, occasional foci did transiently associate with Wsh3/Tea4 foci along the cell cortex (see supplementary material Fig. S4). Dis2.NEGFP association with Wsh3/Tea4 foci at either the tip or the general cortex was dynamic as the signal intensity at individual foci often oscillated (see inset in Fig. 9B) or faded completely while the intensity of the Wsh3.C2tdTom signals maintained a steady signal (supplementary material Fig. S4).

To address the significance of the recruitment of PP1 to cell tips by Wsh3/Tea4, we investigated whether PP1 docking was critical for the recruitment of this polarity determinant itself to cell tips. We did this in a wsh3.Δ background by expressing wsh3⁺, wsh3.V223A or wsh3.F225A genes in which sequences encoding three Pk epitope tags had been inserted, in frame at the C terminus. Wsh3.PkC was recruited to cell tips irrespective of whether it bound Dis2 or not (Fig. 10C). We next addressed the critical role executed by Wsh3/Tea4 in maintaining polarised growth following osmotic stress (Tatebe et al., 2005). We induced varying levels of wsh3⁺, wsh3.V223A or wsh3.F225A expression from the thiamine repressible nmt41 promoter in wsh3.Δ cells by the inclusion of increasing levels of the repressor thiamine in the medium. 20 μM thiamine represses wsh3⁺ production to such a degree that no wsh3⁺ is detectable at the cell tips (Fig. 10B). Scoring cell morphogenesis 3 hours after induction of osmotic stress established that expression of the wild-type, but not either mutant gene suppressed the morphological defects arising from stressing these wsh3.Δ cells (Fig. 10D). The repeated initiation of growth of the old cell end following mitosis in the septation-defective mutant cdc11.132, accentuates defects in polarised tip growth that accompany cell cycle progression of wsh3.Δ cells (Fig. 10E) (Martin et al., 2005). We therefore compared the morphology of cdc11.132 wsh3.Δ cells expressing wsh3⁺, wsh3.V223A or wsh3.F225A and found that the wild-type but neither mutant gene suppressed the morphological defects of wsh3.Δ cdc11.132 cells (Fig. 10E).

Wsh3/Tea4 binds to the polarity determinant Tea1 and is required to ensure the recruitment of Tea1 to the non-growing tip (Fig. 10F) (Martin et al., 2005). Given the clear requirement for PP1 recruitment to cell tips by Wsh3/Tea4 to control polarity (Fig. 10D,E), we investigated whether it was required for Wsh3/Tea4 to promote Tea1 recruitment to both cell tips. To this end we induced wsh3⁺, wsh3.V223A or wsh3.F225A in a wsh3.Δ tea1.pkGFP background. Tea1.PkGFP was recruited to both cell tips irrespective of the ability of Wsh3 to recruit Dis2.NEGFP. Thus, the morphological defects arising from mutation of the PP1 binding sites in Wsh3/Tea4 are not a consequence of altered Tea1 recruitment.

In addition to binding Tea1, Wsh3/Tea4 binds to the formin For3 (Martin et al., 2005). Over production of Wsh3/Tea4 promotes the polymerisation of actin to form prominent F-actin cables and alters cell morphology [Fig. 10B, 20 hours, see supplementary material Fig. S2 for levels of expression, Fig. 10G (Martin et al., 2005)]. Significantly, whereas high levels of expression of wsh3⁺ induced morphological defects and induced excessive F-actin filaments, expression of wsh3.V223A or wsh3.F225A did not.

We conclude that the docking of PP1 to Wsh3/Tea4 is not required for the recruitment of either Wsh3/Tea4 or Tea1 to cell tips. It is, however, critically important for the role played by Wsh3/Tea4 in maintaining polarised tip growth as cells transit the cell cycle or recover from osmotic stress. This function is likely to involve modulation of the F-actin cytoskeleton as the ability of Wsh3/Tea4 to promote F-actin polymerisation is compromised when it cannot bind PP1.

Tracking Dis2-NEGFP and Sds21-NEGFP fusion proteins enabled us to identify processes that are likely to be regulated by these PP1 holoenzymes. We extend the investigation by Ohkura et al. in which immunofluorescence with antibodies that recognised both Dis2 and Sds21 highlighted an enrichment in the nucleus in general and, in particular, the nucleolar region (Ohkura et al., 1989). The lack of a fixation step and the enhanced sensitivity of modern live cell imaging technologies, identified new locations at the cell tip, centromeres and endocytic vesicles, and a distinction between the distribution of Sds21 and Dis2.

Whereas the predicted amino acid sequences of Dis2 and Sds21 are strikingly similar [79% identical (Ohkura et al., 1989)], their distributions are significantly different. Dis2 was found at significant levels at a variety of locations including the entire nucleus, whereas Sds21 was found to accumulate in the nucleolus. Deleting dis2⁺ enhanced Sds21 levels and Sds21 was then found at all the locations previously occupied by Dis2 with the exception of any enhancement of association with chromatin and the central dots seen prior to cytokinesis. However, it is important to note that these elevated levels of Sds21 were still below those of Dis2 (Ohkura et al., 1989) (see supplementary material Fig. S2). One interpretation is that free Sds21 has a limited half-life and yet is stabilised when bound to targeting sub-units, however, Dis2 has a greater affinity for many of these targeting sub-units so that the reduced affinity of Sds21 for these sub-units is reflected in reduced protein levels. The most important distinctions between the two molecules lie in their nuclear distribution as, at the crude level of detection signals, they do not appear to substitute for one another at these sites. This suggests that the actions at these sites might not be essential functions, but clearly demonstrates that the minor sequence divergence is functionally highly significant.

Such distinctions echo the fact that Dis2 contributes the majority of PP1 activity (Kinoshita et al., 1990) and a number of functional distinctions between Sds21 and Dis2. Enhanced Sds21 levels arrest cell cycle progression in G2 whereas elevating Dis2 protein do not (Yamano et al., 1994). This divergence in function was attributed to the ability of Dis2 to be downregulated by phosphorylation on a Cdc2 cyclin B phosphorylation site, T316, as overexpression of a dis2.T316A mutant provoked a G2 arrest (Ishii et al., 1996; Yamano et al., 1994). However, T316 does not appear to be essential for Dis2 function as a multi-copy plasmid expressing a version of dis2 that lacks the last 30 amino acids (297-327) can support growth of dis2.Δ sds21.Δ cells (Kinoshita et al., 1991). A further distinction between Dis2 and Sds21 is illustrated by the unique role played by Dis2 in promoting recovery from DNA damage checkpoint arrest (den Elzen and O'Connell, 2004).

PP1 distribution in S. pombe is reminiscent of PP1 distribution in other systems. GFP fusion protein versions of the highly related α β δ and γ PP1 catalytic sub-unit isoforms of human cells have distinct distributions and show preferential recruitment to specific nuclear compartments. Whereas PP1α is excluded from the nucleolus in a diffuse nuclear pool, PP1β and PP1δ are equally distributed between the chromatin and nucleolus, but do not accumulate within the nucleoli to the same degree as PP1γ (Trinkle-Mulcahy et al., 2001). In addition, immunolocalisation data suggest that PP1α associates with centrosomes, PP1γ with the mitotic spindle and PP1δ with mitotic chromosomes (Andreassen et al., 1998; Fernandez et al., 1992). The single PP1 isoforms of A. nidulans and S. cerevisiae, BimG and Glc7, respectively, also accumulate in nuclei. Glc7 is permanently enriched in the nucleolus whereas BimG is a general nuclear protein in rich conditions, but accumulates in nucleoli during growth on a poor carbon source (Bloecher and Tatchell, 2000; Fox et al., 2002). As in S. pombe, discrete Glc7 spots co-localise with proteins associated with centromeres at the apex of anaphase nuclei (Bloecher and Tatchell, 2000). A discrete BimG dot mirrors the behaviour of centromeric Dis2 in associating with the SPB region throughout the cell cycle with the exception of the period when centromeres leave the SPB from mitotic entry until anaphase (Fox et al., 2002; Funabiki et al., 1993). However, it remains to be established whether this distribution arises from the association of PP1 with centromeres. Association of Glc7 with a kinetochore component is consistent with its role in counteracting the aurora kinase Ipl1 in the generation of unattached chromosomes to initiate the spindle checkpoint response (Francisco and Chan, 1994; Francisco et al., 1994; Hsu et al., 2000; Pinsky et al., 2006; Sassoon et al., 1999; Zhang et al., 2005). This role is likely to be conserved as human PP1γ shows a highly dynamic association with kinetochores either side of a more centrally located aurora B (Trinkle-Mulcahy et al., 2003), and levels of ark1⁺ expression that do not kill wild-type cells (Petersen and Hagan, 2003) are fatal when dis2 function is compromised (our unpublished data).

The cortical associations of Dis2 also echo data from the other fungi. Both Glc7 and BimG accumulate around the cytokinetic ring, the tips of growing hyphae or buds and sites from which hyphae will branch, or buds will emerge, respectively (Bloecher and Tatchell, 2000; Fox et al., 2002). Cortical PP1 has been studied in greatest detail in budding yeast, but it is clear that BimG is required for polarised growth during sporulation and vegetative growth (Doonan and Morris, 1989b; Fox et al., 2002).

The extensive characterisation of endocytosis in S. cerevisiae provides a framework within which to view Dis2 association with endocytic vesicles (Kaksonen et al., 2003; Kaksonen et al., 2005; Kaksonen et al., 2006). The motility of fusions between endocytic proteins and fluorescent proteins such as GFP defines three distinct phases in the endocytic vesicle cycle (Kaksonen et al., 2003; Kaksonen et al., 2005; Kaksonen et al., 2006). The site of endocytosis is determined in a non-motile phase by the co-operative action of molecules such as clathrin, Sla1, Sla2 and Pan1. The association of these proteins with the emerging vesicle persists through the next, slow motile, phase during which the recruitment of myosin 5 and polymerisation of actin accompanies the formation of the clathrin pit. The vesicle then enters a rapid motile phase following the cessation of further actin polymerisation, the departure of myosin 5 and the scission of the pit to form the vesicle. Actin polymerisation is accompanied by the recruitment of homologues of the higher eukaryotic proteins N-WASP and the Arp2/3 complex, which harness actin polymerisation to promote scission and the eventual long range, fast motility of the free vesicle in the final phase (Kaksonen et al., 2003; Kaksonen et al., 2005; Kaksonen et al., 2006).

Glc7 regulates endocytosis via its association with the endocytic molecule Scd5 (Chang et al., 2002; Henry et al., 2002). Scd5 shuttles between the cytoplasm and the nucleus, associates with other endocytic proteins, Sla2 and Rvs167 and is required to maintain a wild-type actin cytoskeleton and for efficient endocytosis (Chang et al., 2006; Chang et al., 2002; Henry et al., 2002; Henry et al., 2003). Scd5 binds to Glc7 via a classic PP1 binding motif. Mutation of this motif compromise endocytosis, actin structure and confers temperature sensitive lethality (Chang et al., 2002). The suppression of scd5 mutant phenotypes following deletion of the gene encoding the endocytic kinase Prk1 combines with biochemical data to suggest that Prk1 negatively regulates Scd5 (Henry et al., 2003), leading, in turn, to the proposal that following recruitment by Scd5 the Glc7-Scd5 holoenzyme counteracts Prk1 phosphorylation (Chang et al., 2006; Henry et al., 2003). As Prk1 phosphorylation acts at multiple levels to regulate actin dynamics and the association of endocytic proteins (Henry et al., 2003; Huang et al., 2003; Kaksonen et al., 2006; Sekiya-Kawasaki et al., 2003; Zeng and Cai, 1999; Zeng et al., 2001) a dynamic interplay between Scd5-Glc7 and Prk1 is likely to be critical to control endocytosis.

Although endocytosis is less extensively characterised in fission yeast, where it has been studied it mirrors that of budding yeast (Castagnetti et al., 2005; Ge et al., 2005; Sirotkin et al., 2005). The fission yeast Sla2 (also known as End4) protein localises in an identical manner to its budding yeast orthologue and is similarly required for endocytosis (Iwaki et al., 2004). By analogy with the three phases in budding yeast (Kaksonen et al., 2006), we found that Dis2 associates with vesicles during the non-motile phase and leaves just prior to the highly motile one. It was recruited to the patches after Sla2/End4. S. cerevisiae Sla2 physically binds to Scd5 and its recruitment to patches requires Scd5 (Chang et al., 2002; Henry et al., 2002). If conserved in S. pombe this relationship suggests that Scd5 first recruits Sla2/End4 and then Dis2 to the forming patches. As has been previously reported for Sla2/End4, depolymerisation of actin with Lat A did not block recruitment of Dis2 to the cell cortex but did freeze Dis2.NEGFP foci at the cell cortex before internalisation (Castagnetti et al., 2005; Ge et al., 2005; Sirotkin et al., 2005). In contrast to Wsp1 and Vrp1 (Sirotkin et al., 2005), neither Dis2 nor Sla2 formed aggregates at cell tips following actin depolymerisation, indicating that recruitment of Wsp1 and Vrp1 but not Dis2 or Sla2 depends upon actin dynamics. It is currently unclear whether the polarisation of endocytosis between growing tips and the nucleus (Castagnetti et al., 2005; Ge et al., 2005) is a secondary consequence of underlying growth polarity or is actively controlled to confer some advantage upon the cell.

The recruitment of a second population of Dis2.NEGFP foci to cell tips by Wsh3/Tea4 in a manner that is dependent upon the Kelch domain protein Tea1 (Martin et al., 2005; Tatebe et al., 2005) is highly reminiscent of budding yeast where Bud14 protein relies upon the Kelch domain proteins Kel1 and Kel2 to recruit Glc7 to the bud tip to control dynein activity during spindle orientation (Knaus et al., 2005). The S. pombe Bud14 homologue, Wsh3/Tea4 (Martin et al., 2005; Tatebe et al., 2005), both co-localised with Dis2 and physically associated with it. This association relied upon the conserved PP1 binding consensus site in Wsh3/Tea4. Mutation of this sequence blocked the recruitment of Dis2, but neither Tea1 nor Wsh3/Tea4 itself to cell tips.

Mutation of the PP1 binding sites in Wsh3/Tea4 compromised the control over the establishment and choice of polarised tip growth that accompanies cell cycle progression of unperturbed cultures and led to a major deficiency in re-establishing polarised growth from existing tips following osmotic stress. This function may be due, in part, to the impact of Wsh3/Tea4 upon actin polymerisation as the ability of excess Wsh3/Tea4 to induce excessive F-actin cables was abolished by mutation of the PP1 binding sites.

Thus, like its A. nidulans homologue BimG, Dis2 function is required for tip growth (Doonan and Morris, 1989; Fox et al., 2002). The requirement for PP1 function during recovery from stress, echoes the requirement for phosphorylation of the polo kinase Plo1 on serine 402 to promote tip growth during recovery from centrifugation and heat stress (Petersen and Hagan, 2005). However, it remains to be established whether Dis2 counteracts the action of polo kinase or any of the other protein kinases, such as Pom1, Kin1, Pck1 and Pck2 that modulate cell morphogenesis in fission yeast (Bähler and Pringle, 1998; Drewes and Nurse, 2003; La Carbona et al., 2004; Toda et al., 1993). Alternatively, it could play a critical role in modulating the activity of the kinases of the stress response pathway that associate with Wsh3/Tea4 (Tatebe et al., 2005). In this scenario the impact on morphogenesis would be mediated by stress signalling to either halt growth in the initial response to the stress, or to re-initiate it once cell composition has been changed to deal with this altered state (Petersen and Hagan, 2005).

Although Bud14 or Glc7 is required for spindle orientation rather than cell morphogenesis (Knaus et al., 2005), the conservation of Bud14-related molecules across species indicates that the identification of the substrates of the Dis2-Wsh3/Tea4 holoenzyme and clarifying its relationship with the cell polarity and morphogenesis kinases will shed much light on the control of the microtubule cytoskeleton and cell morphogenesis by protein phosphorylation in eukaryotes in general.

Strains are listed in supplementary material Table S1. Standard fission yeast and molecular biology approaches were used throughout (Moreno et al., 1991). For imaging, cells were grown overnight in filter sterilised, appropriately supplemented, EMM2 (EMM2-FS) to mid log phase and mounted in a Bioptechs FC2 chamber with lectin (Sigma, L2380; 0.5 mg/ml). TRITC-labelled lectin (Sigma L5264) was used to preferentially label one strain in fields containing two strains (May and Mitchison, 1986). Latrunculin A (Lat A; Calbiochem 4280) and carbendazim (CBZ; Sigma PS1077) were dissolved in DMSO to generate stock solutions of 10 mM and 5 mg/ml, respectively.

Vectors to enable the C-terminal tagging of native loci with the fluorescent proteins tdTomato and Cherry integration of PCR fragments were gifts from Kayoko Tanaka (University of Leicester, UK) and Michael Knop (EMBL, Heidelberg, Germany). The marker switch approach of MacIver et al. (MacIver et al., 2003a) in which the ura4⁺ marker was inserted 266 and 640 bp upstream of the dis2⁺ or sds21⁺ ATG, respectively, was used to generate fusions in which the 3 Pk or EGFP tags from pGEM.Pk and pGEM.EGFP (Craven et al., 1998) were inserted at the initiator methionine codon after it had been mutated to introduce an NdeI site followed by sequences encoding three alanine residues immediately after the initiator methionine. Fragments that encompassed the sequences 435 and 70 bp upstream and 220 and 710 bp downstream of the Dis2 and Sds21 ORFs, respectively, were used for integration. After colony PCR confirmed that the tag had been successfully introduced, two rounds of backcrossing were undertaken. Deletion of wsh3⁺ with nat was carried out according to the method of Hentges (Hentges et al., 2005).

A Perkin Elmer Ultraview spinning disk confocal head on a Zeiss Axiovert 200 or Deltavision Spectris system were used. Spinning disk images for Fig. 1A,C, Fig. 2A,B, Fig. 4, Fig. 5D,E, Figs 6, 7 and 8, and Movies 2-4 were captured as sets of 10-12×0.4 μm slices with a 20-second delay and the z series was compressed to a maximal projection. Images for Fig. 1B, Fig. 2C,D, Fig. 3, Fig. 5B,C,F, Figs 9, 10, supplementary material Fig. S1, Fig. S2B, Fig. S3A, Fig. S4 were captured with the Deltavision Spectris system. One focal plane snapshots were captured for Fig. 1B, Fig. 3. Images for Fig. 2C,D, Fig. 10F and supplementary material Fig. S2B were captured as sets of 20×0.3 μm slices whereas images and the z series were compressed to give the images in the panels. Images for Fig. 9, Fig. 10B, and supplementary material Fig. S3A, Fig. S4 were captured as sets of 3×0.3 μm slices and the z series was compressed to a maximal projection. To study DEAF movement, one single focal plane was captured continuously, without delay between time points, for 100-300 seconds with the spinning disk for Fig. 5A and Movie 1 and with the Deltavision Spectris system for Fig. 5B,C,F. All image analysis used Imaris (Bitplane) software. The stage and phase turret of the axiovert 200 microscope were encapsulated in a heated Perspex environmental chamber (Solent scientific, Segensworth, UK) and the Zeiss 100× 1.45 NA α plan-FLUAR lens was fitted with a Bioptecs objective heater. To inactivate temperature-sensitive gene products, the chamber, objective heater and Bioptechs coverglass were all heated to 33°C or 36°C.

Standard western blotting approaches following TCA precipitation were as described previously (MacIver et al., 2003b). For western blotting, immunofluorescence and immunoprecipitation α-V5 (bethyl A190-120A), α-GFP polyclonal antibodies were used. Chromatin immunoprecipitation PCR utilised the procedures and primers outlined by Pidoux et al. (Pidoux et al., 2004).

We would like to thank Tony Carr, Roger Tsien, Kayoko Tanaka, Takashi Toda, Elmar Schiebel, Paul Nurse, Stephania Castagnetti, Fred Chang, Michael Knop and Kathryn Ayscough for strains and reagents, and Steve Bagley for support in the Advanced Imaging Facility in the Paterson Institute. We are particularly indebted to Alison Pidoux and Wolfgang Reiter for advice on ChIP procedures. This work was supported by Cancer Research UK [CRUK] and the Basque Government.