Pom1 kinase links division plane position to cell polarity by regulating Mid1p cortical distribution

The correct orientation of the division axis, perpendicular to the mitotic spindle axis, is crucial for the successful outcome of mitosis. In animal cells, the division plane is dictated by signals emanating from the mitotic spindle ensuring a direct coupling of both orientations. This mechanism enables animal cells to adapt the orientation of the division plane to late changes in spindle orientation and provides a mean to control division plane orientation by spindle rotation events during the development of multicellular organisms (for reviews, see D'Avino et al., 2005; Glotzer, 2005; Piekny et al., 2005). In yeasts, which have a rigid cell wall and a stereotyped organization, orientation of these two axes is defined independently of each other, according to the cellular pattern: in budding yeast the division plane coincides with the budding site and the spindle is aligned along the mother-bud axis; in fission yeast, the spindle orientates along the long axis of the cell and division takes place perpendicular to this axis, in the middle of the cell, at the position of the nucleus (for reviews, see Chang and Nurse, 1996; Balasubramanian et al., 2004; Wolfe and Gould, 2005). In this organism, the displacement of nuclei by using cell centrifugation or optical tweezers recently revealed that the nucleus can actively influence the position of the division plane (Daga and Chang, 2005; Tolic-Norrelykke et al., 2005).

A major factor for division plane definition in fission yeast is the nuclear shuttling protein Mid1p (Chang et al., 1996; Sohrmann et al., 1996; Paoletti and Chang, 2005), which forms a band at the medial cortex during interphase (Paoletti and Chang, 2000). Mid1p concentrates in this band at the onset of mitosis, upon nuclear export (Bahler et al., 1998; Paoletti and Chang, 2000), and is thought to specify the division plane position by recruiting myosin II to the medial cortex, thereby initiating the assembly of the contractile ring in the middle of the cell (Wu et al., 2003; Motegi et al., 2004). The recruitment of myosin II is mediated by an interaction between Mid1p and the C-terminus of the tail of myosin II heavy chain Myo2p, and is temporally regulated by dephosphorylation of Myo2p (Motegi et al., 2004). Myosin II recruitment is followed by the recruitment of other ring components including the formin Cdc12p and the PCH protein Cdc15p which trigger the nucleation of an F-actin meshwork (Fankhauser et al., 1995; Chang et al., 1997; Carnahan and Gould, 2003; Wu et al., 2003). Cdc12p and Cdc15p recruitment also depends on Mid1p (Wu et al., 2006). All ring components then compact into a tight ring precisely located in the middle of the cell, that starts contracting in late anaphase (Wu et al., 2003).

How Mid1p binds preferentially to the medial region of the cortex, which is crucial for its function, is not understood. We have recently shown that Mid1p anchorage to the cell cortex depends on a dual binding mechanism (Celton-Morizur et al., 2004). First, the Mid1p N-terminal region is able to form faint cortical spots in the medial region of the cortex, in the vicinity of the nucleus. Second, the Mid1p C-terminal domain interacts with membranes through an amphipathic helix. The interaction is reinforced by a polybasic cluster, which also functions as a NLS when Mid1p is not membrane-bound. Strikingly, like the full-length protein both Mid1p N-terminal and C-terminal domains show preferential anchoring to the medial cortex compared with the cell tips (Celton-Morizur et al., 2004). This suggests that the spatial control of Mid1p distribution is highly regulated and may involve several overlapping mechanisms.

Another factor involved in the definition of the division plane is the DYRK kinase Pom1p (Bahler and Pringle, 1998), which displays genetic interactions with Mid1p (Bahler and Pringle, 1998). pom1 mutant cells often form misplaced contractile rings and septa. In addition, pom1 mutant cells have a monopolar pattern of growth, from the new or the old end, indicating that Pom1p also acts as a regulator of cell polarity. However, how Pom1 kinase regulates the position of the division plane remains unknown. Immunofluorescence experiments have shown that Mid1p is normally associated with the misplaced contractile rings in pom1 mutant cells (Bahler et al., 1998), but these experiments did not reveal the distribution of Mid1p on the cortex before mitosis, which is crucial to understand how the contractile ring is positioned in these cells.

Since Pom1p localization at the cell tips depends on microtubules and on the polarity factors Tea1p and Tea4p (Bahler and Pringle, 1998; Tatebe et al., 2005), we speculated that Pom1p negatively regulates the association of Mid1p with the cell cortex at cell tips and links Mid1p cortical distribution to the establishment of cell polarity. We therefore determined the localization of Mid1p in pom1 mutant cells. We indeed found that in these cells, Mid1p distribution on the cortex expands from the medial cortex towards the non-growing tip of the cell and is uncoupled from the position of the nucleus. We propose a model in which Pom1 kinase acts as a negative regulator of Mid1p distribution, excluding it from non-growing cell tips.

To efficiently detect Mid1p on the cortex during interphase, we fused Mid1p to a quadruple green fluorescent protein (GFP) tag in C-terminus (mid1-4GFP) (Maekawa et al., 2003). In a control mid1Δ strain where this construct was integrated at the leu1 locus and expressed under the control of the mid1 promoter (strain AP1442, Table 1), mid1-4GFP was expressed like endogenous Mid1p and septa were properly positioned (Fig. 1, see also Fig. 5D,E). The fluorescent signal of mid1-4GFP was brighter but very similar to mid1-GFP signal (Fig. 1A, left) (Paoletti and Chang, 2000). mid1-4GFP also complemented the synthetic lethality of mid1Δpom1Δ double deletion. This indicates that the 4GFP tag does not alter Mid1p function or distribution.

In pom1Δ cells (Strain AP1502, Table 1), mid1-4GFP cortical distribution was different from in wild-type cells (Fig. 1A, right). First, during interphase, the nuclear staining was reduced as reported previously (Bahler et al., 1998). Second, instead of being restricted to the medial cortex, mid1-4GFP cortical staining expanded towards one cell tip. As shown on fluorescence intensity graphs along the cortex (Fig. 1D), mid1-4GFP spots were often unevenly distributed along the cortex. To analyze the distribution of mid1-4GFP in detail, we counted the number of cells containing mid1-4GFP spots on the medial, lateral and tip cortex on one side of the cell (see scheme in Fig. 1E; a, b and c, respectively). 72% of pom1Δ cells displayed Mid1p spots in all three regions (abc cells) compared with 3% in wild-type cells. The distribution of mid1-4GFP spots did not vary during cell-cycle progression, monitored by measuring cell length (Fig. 1E). We also performed an automated detection of mid1-4GFP cortical spots and measured their distance to the medial axis of the cell (Fig. 6A). mid1-4GFP spots were more numerous in pom1Δ cells (12 spots per single focal plane in AP1502, n=77 and 7.5 spots per single focal plane in AP1442, n=107 in single focal planes) and very frequently located far from the medial cell axis on one side of the cell: 45% of spots were located more than 2 μm apart from the cell medial axis compared with 17% in wild-type cells; 68% of spot centroids, which represent the mean distribution point of spots in individual cells and whose position is indicative of the asymmetry of spots distribution, were located more than 0.75 μm apart compared with 12% in wild-type cells (Fig. 6B,C).

Similar redistributions of Mid1p were observed in pom1-1 mutant cells and in kinase-dead pom1-2 mutant cells (our unpublished observations) (Bahler and Nurse, 2001). Therefore, Pom1p activity is required to prevent Mid1p association with one of the two cell tips. These experiments also indicated that in the pom1 mutant, the coupling between Mid1p cortical distribution and nucleus position (Paoletti and Chang, 2000; Daga and Chang, 2005) is abolished. To further test this hypothesis, we ultracentrifuged cells in presence of the microtubule-depolymerizing drug MBC to move the nucleus away from the cell center (Khar and Mitchison, 1989; Daga and Chang, 2005) (see supplementary material Fig. S1a,c). Wild-type cells that were kept in MBC-containing medium slowly relocated Mid1p spots to the new cortical region facing the nucleus (see supplementary material Fig. S1b, arrowheads). By contrast, in pom1Δ cells, Mid1p failed to relocalize to the cortex facing the nucleus. Accordingly, the percentage of pom1Δ cells with cortical spots on one side and the nucleus on the other did not change over time after centrifugation (about 40% of cells with displaced nuclei, see supplementary material Fig. S1d, arrows). We conclude that, in absence of Pom1p activity, the coupling between Mid1p cortical distribution and nucleus position no longer operates.

One consequence of mid1-4GFP cortical redistribution in pom1Δ cells is that some mid1-4GFP spots were localized at one cell tip, in the same area as the cell polarity regulator Tea1p (Mata and Nurse, 1997), which is normally localized at the two cell tips in pom1 mutants (Bahler and Pringle, 1998). As pom1 cells exhibit a monopolar pattern of growth (Bahler and Pringle, 1998), we tested whether the tip containing mid1-4GFP corresponded to a specific tip by using fluostain I, which brightly stains growing tips. mid1-4GFP was consistently found at the non-growing tip (Fig. 1B). This result is of particular interest because it indicates that, in absence of Pom1p a second mechanism may prevent Mid1p association with the growing tip.

In wild-type cells, ring components are first recruited to the medial cortical compartment defined by Mid1p before compacting into a tight ring (Wu et al., 2003; Wu et al., 2006). This mechanism is thought to allow a precise positioning of the contractile ring in the middle of the cortical region defined by Mid1p. Cortical redistribution and uncoupling of Mid1p from the position of the nucleus in pom1 mutants may therefore directly affect the position of the division plane during mitosis.

To directly test this hypothesis, we made time-lapse movies of wild-type and pom1Δ cells expressing mid1-4GFP. All wild-type cells expressing mid1-4GFP that entered mitosis formed well assembled, centrally placed contractile rings (Fig. 2Aa). By contrast, in the pom1Δ mutant, six out of nine cells entering mitosis during the movies formed offset rings that were displaced towards the non-growing cell tip where mid1-4GFP cortical spots were located before mitosis (Fig. 2Ab,c). In these cells, the division plane position did not correspond to the position of the premitotic nucleus (see stars in Fig. 2Ab,c). In the three other cells, contractile rings assembled centrally but they either exhibited shape defects or appeared not to compact fully, as judged from our analysis in a single focal plane (our unpublished observations).

Moreover, we found that in pom1Δ cells, myosin II heavy chain Myo2p, whose recruitment to the medial cortex depends on Mid1p (Wu et al., 2003; Motegi et al., 2004), first accumulated asymmetrically on the cortex before compacting into offset rings on the same side of the cell (Fig. 2B). We therefore conclude that the redistribution of Mid1p towards the non-growing cell tip is responsible for the mispositioning of the contractile ring.

Finally, it has been reported that highly elongated cdc25 pom1Δ cells that are released from the cell-cycle block form contractile rings and septa at or near the cell middle (Bahler et al., 1998). To understand why this is the case, we analyzed the distribution of Mid1p in elongated pom1Δ cells that were obtained by a 5-hour treatment with hydroxyurea (HU) at 25°C (Fig. 1C). As in control cells, mid1-4GFP spots were asymmetrically redistributed towards the non-growing cell tip. The distribution of spots varied from one cell to the other but, generally, the number of spots was reduced at the cell tips compared with the lateral or medial cortex. This reduction in spot number at the cell tip in very long cells might explain why ring-positioning defects are not more dramatic in highly elongated pom1Δ cells.

As in wild-type cells Pom1p relocalizes to the newly forming end during septation (Bahler and Pringle, 1998) but its activity only increases when the new end starts growing at new-end take off (NETO) (Bahler and Nurse, 2001), we wondered whether Pom1p increased activity is necessary to remove Mid1p from the cortex of the new cell tip at NETO. However, in wild-type cells, most pre-NETO cells, ranging from 7-9 μm in length (Mitchison and Nurse, 1985), displayed a medial distribution of mid1-4GFP spots on the cortex (56%; Fig. 1E). Similarly, in wild-type cells that expressed mid1-4GFP and were blocked in S phase–before NETO (Mitchison and Nurse, 1985)–by treatment with hydroxyurea (HU), mid1-4GFP cortical staining was still restricted to the central cortex overlying the nucleus, like in non-treated control cells (Fig. 3). We conclude that Mid1p distribution is restricted to the central cortex before NETO in wild-type cells. Pom1p might, therefore, start to act on Mid1p distribution earlier in the cell cycle, before Mid1p activity increases.

To determine more precisely how early in the cell cycle Pom1p starts regulating Mid1p distribution, we recorded time-lapse movies during mitosis exit and septation. These movies show that in pom1Δ cells–like in wild type cells–mid1-4GFP spread laterally on the cell cortex surrounding the contractile ring in late anaphase, concomitantly with a reduction in the cortical staining and the reappearance of a nuclear pool of the protein (Fig. 4A).

We also analyzed mid1-4GFP distribution on single images to distinguish Mid1p cortical spots better during late stages of mitosis (Fig. 4B,C). Interestingly, cortical spots of mid1-4GFP were frequently detected at the old end, before or during sister-cell separation in pom1Δ cells, whereas they were already centrally located in wild-type cells. In conclusion, Pom1 kinase is not required for the release of Mid1p from the ring or septum region and may act on Mid1p cortical distribution immediately after the exit of mitosis.

Since Mid1p cortical distribution in pom1Δ cells correlated negatively with the monopolar pattern of growth, we analyzed whether this defect was specific to pom1 mutants or whether it could be observed in other strains deficient for bipolar growth.

In the bud6Δ strain (Glynn et al., 2001), in which 50% of the cells display a monopolar pattern of growth (Glynn et al., 2001), we observed a completely normal distribution of mid1-4GFP at the medial cortex (Fig. 5A,B; 10% of spots and 10% centroids located >2 μm and >0.75 μm apart from the medial axis, respectively, Fig. 6B,C). Accordingly, septa were normally positioned in this strain (Fig. 5D). This indicates that the defective distribution of Mid1p in pom1Δ cells is not a general consequence of their deficiency in establishing bipolar growth.

In tea3Δ cells, which grow monopolarly and have septum-position defects (Arellano et al., 2002), mid1-4GFP spots were most often restricted to the medial cortex (48.5±15% a cells instead of 66±11% in wild-type cells, Fig. 5A,B). The number of spots on the lateral cortex was slightly increased compared with wild-type cells (26% of spots and 22% centroids located >2 μm and >0.75 μm apart from medial axis, respectively, Fig. 6B,C). We also observed mild displacements of nuclei away from the cell center which could participate in the weak mispositioning of septa in this strain (25± 8.5%).

In pom1Δtea3Δ double mutants, which form strongly displaced septa at 36°C (Arellano et al., 2002) (our unpublished observations), mid1-4GFP displacement towards the non-growing tip was largely similar to what was observed in pom1Δ single mutants when cells were grown at 25°C (77.5±11.5% abc cells; Fig. 5A,B; 49% of spots and 67% centroids located more than 2 μm and 0.75 μm apart from medial axis, respectively; Fig. 6B,C). Some cells displayed strong defects in shape were not considered because their global polarity may be abnormal. Importantly, mid1-4GFP was still absent from the growing tip in pom1Δtea3Δ cells (Fig. 5A), indicating that Tea3p is not required to prevent Mid1p association with the growing tip in absence of Pom1p. We were unable to determine mid1-4GFP distribution at 36°C because the mid1-4GFP signal was lost in these conditions of growth (our unpublished observations).

Finally, the tea1Δ mutant provides a situation to test whether Pom1p localization to the cell ends is required to regulate Mid1p cortical distribution, because Pom1p localization to the cell ends was reported to be abolished in this background, whereas its kinase activity remained high (Bahler and Pringle, 1998; Bahler and Nurse, 2001). In tea1Δ cells, mid1-4GFP spots were frequently present on the lateral cortex of one cell end (50±1.5% ab cells compared with 20±9% in wild-type cells; Fig. 5A,B; 38% of spots and 46% centroids located more than 2 μm and 0.75 μm apart from medial axis, respectively; Fig. 6B,C). However, mid1-4GFP spots were infrequently found in the tip region (25±11% abc cells in tea1Δ cells compared with 64±5.5% in pom1Δ cells; Fig. 5A,B).

We did not expect a mild redistribution phenotype if Pom1p needs to concentrate at the cell ends to negatively regulate the distribution of Mid1p on the cortex. However, when we reexamined the pom1-GFP distribution in tea1Δ cells (Fig. 5C), we consistently observed some staining at the cell tips and in the septum region and, occasionally, on the lateral cortex. The fact that Pom1p does not dissociate fully from the cell tips in tea1Δ cells may therefore account for mid1-4GFP partial redistribution and septum mild mispositioning defects observed in this strain (40±10% compared with 5.5± 3% in wild-type cells and 90.5±0.3% in pom1Δ cells, Fig. 5D).

Altogether, these results strengthen our hypothesis that Mid1p distribution on the cortex is specifically controlled by Pom1p and suggest that Pom1 kinase activity may locally inhibit Mid1p association with the cortex at the cell tips.

We have found that in pom1-deficient cells, Mid1p distribution on the cortex expands from the medial cortex towards the non-growing tip of the cell and is uncoupled from the position of the nucleus. This defect is specific for pom1-deficient cells and not a general consequence of monopolar growth because we did not observe it in bud6Δ cells. Our analysis of the kinase-dead pom1-2 mutant (our unpublished results) and of the tea1Δ mutant, in which Pom1p is partially delocalized, also reveals that the regulation of Mid1p distribution requires active Pom1 kinase, properly localized at the cell tips. This indicates that Pom1p functions as a negative regulator in the distribution of Mid1p, preventing its association with the cortex at non-growing cell tips (see Fig. 7). Accordingly, we found that Mid1p cortical spots were more numerous in pom1Δ cells compared with wild-type cells.

Our observations of Mid1p cortical distribution in wild-type and pom1Δ cells during mitosis exit indicate that Pom1p is not necessary for Mid1p dissociation from the contractile ring or from the region of septum formation where Pom1 kinase relocates in late mitosis. Since in pom1Δ cells growth resumes at the new end in about half of the cases (Bahler and Pringle, 1998), a defect in the dissociation of Mid1p from the septum region would not have been sufficient to explain a consistent Mid1p redistribution to non-growing tips anyway. In wild-type cells, Pom1p might rather prevent Mid1p distribution to non-growing tips in very early stages of the cell cycle, at the old end, before growth resumes in this region and at the new end until NETO.

Interestingly, we found that in absence of Pom1p, Mid1p is still excluded from the growing tip of the cell. Mid1p distribution is therefore directly linked to the pattern of growth in these cells. This indicates that, in addition to Pom1p, other factors localized at the growing tip and absent from the non-growing tip can negatively regulate Mid1p cortical distribution. In wild-type cells under bipolar growth, Pom1p-dependent and independent mechanisms might constitute overlapping mechanisms preventing Mid1p association with growing tips, whereas Pom1 kinase might act alone at non-growing tips in early phases of the cell cycle (see Fig. 7).

The fact that contractile rings assemble in the region of the cell where Mid1p cortical spots are redistributed before mitosis indicates that, Mid1p misdistribution is responsible for the displacement of the contractile ring towards the non-growing cell tip that has previously been reported in pom1-deficient cells (Bahler and Pringle, 1998). These data are consistent with the fact that septum-position defects are generally weaker in pom1Δ cells than in mid1Δ cells, most septa forming perpendicularly to the long axis of the cell, generally not more than a few micrometers away from the medial axis of the cell (Bahler and Pringle, 1998; Bahler et al., 1998) (our unpublished observations). Indeed, this localization of septa correlates well with the distribution of centroids of Mid1p spots, spread between 0-2 μm apart from the cell medial axis in pom1Δ cells (see Fig. 6B). Moreover, we found that in highly elongated pom1Δ cells, Mid1p cortical spots are generally less numerous in the tip region compared with the lateral cortex. This might explain why ring-positioning defects are not aggravated upon artificial elongation of pom1Δ cells (Bahler et al., 1998).

The reasons for decreased Mid1p staining at the tips in long pom1Δ cells need to be clarified. One hypothesis is that Mid1p anchoring to the cortex requires positive signals generated in the medial region of the cell. These positive signals would be sufficient to allow anchoring all along the cortex of non-growing ends of short pom1Δ cells but would not reach cell tips in highly elongated cells. The fact that after ultracentrifugation of wild-type cells Mid1p cortical spots relocate in the vicinity of the displaced nucleus in presence of the microtubule depolymerizing drug MBC suggests that such positive signals could be generated by the nucleus or by nucleus-associated structures. However, other parameters, such as the growth activity of the tip in the previous cell-cycle might also affect the distribution of Mid1p at the cell tips because we observed important variations in the distribution of spots along the cortex from one cell to another in highly elongated as well as in short pom1Δ cells (see Fig. 1C,D).

It will be important to determine how Pom1 kinase acts on Mid1p to regulate its cortical distribution. One obvious hypothesis is that Pom1 kinase directly phosphorylates Mid1p, thereby regulating its association with the cortex. To test this hypothesis, we performed in vitro Pom1 kinase assays on Mid1p purified from cell extracts and obtained negative results (our unpublished observations). We also produced a triple mutant of Mid1p on the three potential Pom1 phosphorylation sites that we detected on the sequence according to the DYRK2-DYRK3 kinase consensus site (Campbell and Proud, 2002) (T34A, S218A, T517A). This mutant was correctly localized at the medial cortex and fully functional (our unpublished observations). Therefore, Pom1p may act indirectly on Mid1p.

In any case, in contrast to what was proposed from the observation that pom1 and mid1 have synthetic lethal interactions (Bahler and Pringle, 1998), our data indicate that Pom1p acts in the same pathway as Mid1p, controlling the position of the division plane. The synthetic lethality between the two deletions could either result from combined defects in cell polarity and morphogenesis, and in division plane positioning or from other functions of Pom1p in division plane placement, independent from Mid1p.

A search for additional factors having dual functions in the establishment of cell polarity and division plane placement may reveal intermediary factors through which Pom1p affects Mid1p distribution. It may as well identify the factors necessary to exclude Mid1p from the growing end of the cell independently of Pom1p. One candidate factor was Tea3p (Arellano et al., 2002). However, our analysis showed that Tea3p might only play a minor role in the regulation of Mid1p distribution on the cell cortex and, particularly, that Tea3p is not required to prevent Mid1p association with the growing tip in the absence of Pom1p.

Finally, our work provides the first evidence that the definition of the division plane by Mid1p is influenced by cell-polarity regulators and that Mid1p cortical distribution can be uncoupled from the position of the nucleus in their absence. Since Pom1 kinase is delivered to the cell tips in a microtubule-dependent manner (Bahler and Pringle, 1998), our work indicates that the nucleus controls the definition of the division plane indirectly through its ability to anchor antiparallel microtubule bundles at interphase microtubule organizing centers (iMTOCs) (Tran et al., 2001): microtubules bundles with their minus ends anchored on the nucleus and their plus ends at the cell tips deliver polarity regulators that in turn control Mid1p distribution on the cortex. It thus suggests that in fission yeast, like in animal cells, the microtubule network influences the definition of the division plane. One apparent objection to this model is that the depolymerization of microtubules does not affect Mid1p cortical distribution (Paoletti and Chang, 2000). This objection fails if Mid1p distribution is also regulated by microtubule-independent mechanisms or if the distribution of polarity regulators is multimodal, as recently described for Tea1p (Snaith et al., 2005).

All strains used in the present study are isogenic to the wild-type 972 strain and listed in Table 1. Standard S. pombe genetic approaches were used (Moreno et al., 1991).

pAP221 plasmid (pmid mid1-4GFP leu1+) was obtained from pAP146 (Celton-Morizur et al., 2004) by replacing the GFP cassette (NotI-SmaI fragment) with the 4GFP cassette from pSM1023 (fragment NotI-BglII blunted) (Maekawa et al., 2003) a generous gift from E. Schiebel (ZMBH, Heidelberg, Germany). pAP221 was digested with NruI and transformed into the SP1601 strain (mid1Δ::ura4+) (Sohrmann et al., 1996) to obtain strain AP1442. AP1442 was crossed with strain mid1Δ::KanMX6 (our collection) to obtain strain AP1487 (mid1Δ::KanMX6 + pAP221 integrated).

Strains AP1502, AP1717, AP1727, AP1737 and AP1758 were obtained by crossing strains AP1442 with AP1487, and strains derived by genetic crosses with JB110 (pom1Δ::ura4+) a generous gift from J. Bähler (Sanger Centre, Hinxton, UK) (Bahler and Pringle, 1998), PN4370 (tea3Δ::kanMX6) a generous gift from P. Nurse, (Arellano et al., 2002), PN1690 [tea1Δ::ura4+, generous gifts from P. Nurse (Cancer Research UK, London, UK) (Mata and Nurse, 1997)], and AP37 (bud6Δ::kanMX6 mid1Δ::ura4+) (Glynn et al., 2001) (our collection). Strain AP1329 was derived from strain JB115 (pom1-GFP::KanMX6, a generous gift from J. Bähler) (Bahler and Pringle, 1998), and strain AP1760 was obtained by crossing AP1329 with a strain derived from strain PN1690. Strain AP1823 was obtained by crossing JB110 (pom1Δ::ura4+, a generous gift from J. Bähler) (Bahler and Pringle, 1998) with a myo2-GFP strain derived from MBY498 (myo2GFP::ura4+, a generous gift from M. Balasubramanian, Temasek Life Science Laboratory, Singapore).

All strains were exponentially at 25°C in YE5S under shaking and imaged between slides and coverslips after a short centrifugation (10 seconds at 10,000 g) to concentrate the cells. Imaging was performed as described in Celton-Morizur et al. (Celton-Morizur et al., 2004), with acquisition times of 500 mseconds for time-lapse movies and of 2 seconds for mid1-4GFP or pom1-GFP and tea1-CFP single images. Cell walls were visualized by staining with fluostain I (2 μg/ml, Sigma, St Louis).

To move the nucleus away from the cell center, 1 ml of wild-type (AP1442) or pom1Δ cells (AP 1502) grown exponentially at 25°C in YE5S were treated for 5 minutes with MBC (25 μg/ml, Aldrich, Milwaukee), deposited in a ultra-clear centrifuge tube (Beckman Instruments, Palo Alto) on top of a 1% SeaKem LE agarose cushion (Cambrex, Rockland), and ultracentrifuged for 8 minutes at 100,000 g in a SW41 swinging rotor (Beckman Instruments, Palo Alto, USA) as described by Daga and Chang (Daga and Chang, 2005). After centrifugation cells were either imaged shortly after a brief centrifugation to concentrate the cells (i.e. 10-20 minutes after the ultracentrifuge stopped running), or shaken at 25°C in medium containing MBC and imaged 2-2.25 hours later.

Total cells extracts and western blot analysis were performed as described in Celton-Morizur et al. (Celton-Morizur et al., 2004), using anti-Mid1p antibody (1:150) (Paoletti et al., 2003) and TAT1 monoclonal antibody against tubulin (1:2000, generous gift from Keith Gull (University of Oxford, UK).

For each image, cells were located by user region drawing (see in Fig. 6Aa). Cells were precisely segmented using an automated local threshold inside the regions. The center of the cell was set to the segmented region. The cell orientation α_c was computed from an ellipse fitting on the region (see in Fig. 6Ab). (X_c, Y_c) and α_c define the major and minor axes of the cell, named long and medial axes respectively in the rest of the manuscript. Nuclei were detected using the MIA software developed on site, which allows the precise detection of objects with various sizes using `a-trous' wavelet-segmentation algorithms (Starck et al., 1998). Spots were detected outside from the nuclear region with the same software (see Fig. 6Ac,d). Coordinates of spots on the axes of the cell were automatically extracted and analysed for all strains. Coordinates on the long axis directly correspond to the distance between the spots and the medial axis of the cell. The position of spots centroids (or center of mass) was calculated with a similar weight for each spot. Coordinates of spots centroids on the long axis directly correspond to their distance from the medial axis.

We thank V. Doye, I. Loïodice, A-M Tassin and P. T. Tran for critical reading of the manuscript, J. Bähler, M. Balasubramanian, P. Nurse and E. Schiebel for strains and plasmids. This work was supported by ARC (grant no. 3623 to A.P.). S.C.M. received a doctoral fellowship from ARC and a short-term traveling fellowship from Journal of Cell Science.