From an insertional mutagenesis screen designed to identify non-essential genes regulating cell polarity in fission yeast (H.A.S., unpublished observations), we identified four genes whose loss-of-function phenotype resembles that of a tea1Δ strain, which forms bent or branched cells after a variety of stresses4,5 (see Supplementary Information). One of these genes was tea1+ itself (Fig. 1a), and two of the others were previously identified genes that are known to affect microtubule organization and consequently the normal delivery of tea1p to cell tips (tea2+, tip1+ (refs 9, 10)). The fourth gene, which we termed mod5+ (for morphology defective 5), has not previously been characterized and is identified as open reading frame SPBC530.04 in the S. pombe genome ( Deletion of the complete open reading frame of mod5+ (mod5Δ) did not affect cell viability and yielded the same mutant phenotype as the original insertion mutant (Fig. 1b).

Figure 1: mod5Δ cells fail to localize tea1p at cell ends.
figure 1

a, b, Phenotype of tea1Δ cells (a) and mod5Δ cells (b) on solid medium 4 h after refeeding. Wild-type cells are uniformly cylindrical (see, for example, c) c, d, Anti-tea1p staining in wild-type cells (c) and mod5Δ cells (d). e, f, Anti-tubulin staining in wild-type cells (e) and mod5Δ cells (f). g, h, Merged images of c and e, and d and f, respectively. tea1p accumulates at cell tips in wild-type cells but is restricted to microtubule ends in mod5Δ cells. i, mod5Δ cell stained for tubulin and tea1p, showing tea1p on microtubule ends. j, Enlargement of the cell tip shown in i. Scale bar, 5 µm (ai) and 1 µm (j).

In immunofluorescence experiments with wild-type cells, tea1p was present both at cell tips and at the ends of microtubules (Fig. 1c, e, g)4. Strikingly, in mod5Δ cells, although tea1p remained associated with microtubule ends, it no longer accumulated to high levels at cell tips (Fig. 1d, f, h–j; see also Supplementary Information), and 91% of tea1p spots observed at cell tips in mod5Δ cells were associated with a microtubule end (n = 613 spots). Immunoblotting experiments confirmed that tea1p levels are not altered in mod5Δ cells (data not shown). Microtubule organization was also generally similar between wild-type and mod5Δ cells (Fig. 1e, f), although a small percentage of mod5Δ cells contained a microtubule curling around the cell tip4,8 (see Supplementary Information).

Because tea1p is normally targeted to cell ends by microtubules4,8, we further investigated the role of mod5p in the cortical localization of tea1p by disrupting microtubules during polarity re-establishment experiments, in which cells are first grown to stationary phase and then diluted into fresh medium. This procedure increases the penetrance of polarity mutant phenotypes during the first cell cycle after dilution into fresh medium (ref. 9 and our unpublished observations) and was performed both in the presence and in the absence of the fungal microtubule inhibitor 2-methyl benzimidazolylcarbamate (MBC). Within 3 hours of being returned to fresh medium, wild-type cells not treated with the drug re-established a growth polarity axis, with tea1p uniformly decorating cell tips (Fig. 2a). By contrast, in mod5Δ cells not treated with drug, tea1p was found only at microtubule ends and did not accumulate at cell tips (Fig. 2b). It has been reported that the localization of tea1p at cell tips is strongly dependent on microtubules4. However, in wild-type cells treated with MBC during polarity re-establishment, at least one cell tip was able to accumulate significant levels of tea1p, despite the presence of only very short microtubule remnants (Fig. 2c). In contrast, mod5Δ cells exhibited nearly no detectable tea1p at cell tips after treatment with MBC (Fig. 2d; see Supplementary Information). We conclude that a major defect associated with the loss of mod5p is a failure to retain tea1p at cell tips, and that this is likely to be independent of the microtubule-based targeting of tea1p.

Figure 2: Microtubule-independent targeting of tea1p to the cortex depends on mod5p.
figure 2

Cells are stained for tea1p (green) and tubulin (red). a, b, Wild-type cells (a) and mod5Δ (b) cells 3 h after release to growth in the presence of dimethyl sulphoxide (control). c, d, Wild-type cells (c) and mod5Δ cells (d) 3 h after release to growth in the presence of 50 µg ml-1 MBC. Scale bar, 5 µm. e, Percentage of cells forming branches in wild-type, mod5Δ and tea1Δ cells 3 h after release to growth, in the absence and in the presence of MBC. The percentages shown in the first three columns are 0%, 3% and 1%, respectively.

These results also indicate that the targeting of tea1p to cell tips might be a two-step process in which microtubule-dependent delivery is followed by a microtubule-independent, but mod5p-dependent, cortical anchoring mechanism. We speculate that, in the absence of microtubules, diffusion of tea1p might allow its eventual cortical association, provided that mod5p is present. This interpretation is supported by the frequency of abnormal, branched cells observed in polarity re-establishment experiments (Fig. 2e). In tea1Δ cells a high frequency of cell branching was seen both in the absence and in the presence of MBC (88% and 92% of cells, respectively; n = 200 cells). In comparison, very few wild-type cells formed branches under either condition (0% and 3%, respectively). However, in mod5Δ cells the frequency of branched cells increased markedly in the presence of MBC, from 1% to 84%. We interpret these data to indicate, first, that the absence of tea1p from cell tips during polarity re-establishment leads to a high frequency of branching; second, that in mod5Δ cells not treated with the drug, the small amount of tea1p present at cell tips in association with microtubule ends might often be sufficient for normal polarized growth; and last, that in MBC-treated mod5Δ cells, the further loss of this small amount of microtubule-associated tea1p produces a phenocopy of tea1Δ.

To determine whether mod5p might have a role not only in the anchoring of tea1p at the cortex but also in the microtubule-based transport of tea1p, we examined live cells expressing green fluorescent protein (GFP)-tagged derivatives of tea1p and α-tubulin (atb2p, the non-essential α-tubulin in S. pombe11,12,13,14 (K.E.S., unpublished observations)). In cells expressing GFP–tubulin at near-endogenous levels, no significant differences in microtubule dynamics were apparent between wild-type and mod5Δ cells (data not shown). The localization of tea1p–GFP in live wild-type and mod5Δ cells was similar to that of endogenous tea1p in fixed cells (Fig. 3a, b; see also Supplementary Information). Time-lapse videomicroscopy of wild-type cells revealed small particles of tea1p–GFP originating in the medial perinuclear region of the cell and translocating towards the cell tips at 2.43 ± 0.40 µm min-1 (mean ± s.d.; n = 27 particles; Fig. 3a), where they were then retained, as described recently8. In mod5Δ cells, tea1p–GFP particles moved from the nuclear periphery towards the cell tips at similar rates (2.69 ± 0.55 µm min-1; n = 27 particles; Fig. 3b) but upon reaching cell tips either moved away from the cortex and/or disassembled, which is consistent with observations of fixed time-points (see Supplementary Information).

Figure 3: tea1p is transported on microtubule ends to cell tips in mod5Δ cells.
figure 3

a, b, Time lapse series of tea1p–GFP in wild-type cells (a) and mod5Δ cells (b). Images in a and b were taken at 10-s intervals. Yellow arrowheads indicate tea1p–GFP particles moving towards cell tips. In b, green arrows indicate tea1p–GFP spots moving in a stream. c, d, Double-labelled time-lapse series of tea1p-YFP (green) and CFP-atb2p (red) in wild-type cells (c) and mod5Δ cells (d). Images in c and d were taken at 15-s intervals. QuickTime movie files of this figure are provided in Supplementary Information. Scale bar, 5 µm.

Three other differences in the properties of tea1p–GFP particles were also apparent in mod5Δ cells. First, mod5Δ cells contained 6.52 ± 0.98 tea1p–GFP particles per cell (mean ± s.d.; n = 116 cells), compared with 2.73 ± 1.24 in wild-type cells (n = 131 cells). In addition, tea1p–GFP particles in mod5Δ cells were 2–3-fold brighter than in wild-type cells (see Supplementary Information). Moreover, in many instances several tea1p–GFP particles seemed to move in a single stream towards the cell tips (Fig. 3b; see also Supplementary Information). The increased number and intensity of particles can be explained in part by the fact that mod5Δ mutants might contain a larger free cytoplasmic pool of tea1p–GFP because it is no longer sequestered at cell tips. In wild-type cells, tea1p–GFP particles translocating towards cell tips have been shown to be specifically associated with the plus ends of growing microtubule bundles8. The streaming behaviour of tea1p–GFP particles observed in mod5Δ cells indicated that several particles might be tracking along the length of a single microtubule bundle rather than being uniquely associated with microtubule ends. However, by imaging tubulin tagged with cyan fluorescent protein (CFP) together with tea1p tagged with yellow fluorescent protein (YFP) we found tea1p–YFP particles present only at the ends of microtubules in both wild-type and mod5Δ cells (Fig. 3c, d; see also Supplementary Information). These results indicate that long microtubule bundles in mod5Δ cells might contain some shorter individual microtubules whose ends might not be detectable with current imaging methods. Alternatively, these microtubule ends might exist in wild-type cells as well, but might not normally be occupied by tea1p unless the free cytoplasmic pool is increased, as in mod5Δ cells. In any case, we do not find evidence for a fundamentally different mode of tea1p transport in mod5Δ cells.

To determine whether mod5p acts locally to retain tea1p at cell tips, we investigated its intracellular distribution. Immunostaining of wild-type cells with anti-mod5p antibodies showed mod5p at cell tips (Fig. 4a; see also Supplementary Information), and this was confirmed in live cells expressing amino-terminally tagged GFP–mod5p (Fig. 4b). GFP–mod5p localization to cell tips was partly affected by acute disruption of microtubules with MBC (see Supplementary Information) but was unaffected by disruption of actin filaments with latrunculin B (data not shown). The last four amino acids of the predicted mod5p-coding sequence are a consensus signal for C-terminal prenylation (CaaX, where C is cysteine, a is an aliphatic amino acid and X is any amino acid); mod5p is therefore predicted to be membrane associated15. We tested the functional significance of this motif both by deleting the last four amino acids of mod5p (mod5ΔCaaXp) and by mutating the putative prenylated cysteine residue to serine (mod5C519Sp); in both cases the normal localization of mod5p was disrupted (Fig. 4c; data not shown). Mod5ΔCaaXp appeared in a small number of spots throughout the cell, which were typically found near cell tips (Fig. 4c), and in the mod5ΔCaaX mutant the localization of tea1p was defective (Fig. 4d). These results indicate that the proper targeting of mod5p to the plasma membrane at cell tips might be essential for its function and that a normal microtubule distribution might contribute to its localization.

Figure 4: mod5p localization to cell tips depends on a functional CaaX sequence and on tea1p.
figure 4

a, Anti-mod5p immunofluorescence in wild-type cells. b, Localization of GFP–mod5p in wild-type cells (expressed from the nmt81 promoter, replacing endogenous expression, about 3-fold higher than in the wild type). c, Localization of GFP–mod5ΔCaaXp (expressed from the nmt41 promoter, replacing endogenous expression). d, Expression of mod5ΔCaaXp mimics the phenotype of mod5Δ. mod5ΔCaaX cells stained with antibodies against tea1p (green) and tubulin (red). e, Localization of GFP–mod5p in tea1Δ cells. f, Localization of GFP–mod5p in tea3Δ cells. Scale bar, 5 µm.

What trans-acting factors might be necessary for localizing mod5p at cell tips? We found that in tea1Δ cells, GFP–mod5p is no longer restricted to the cell tips but instead spreads out across the entire plasma membrane (Fig. 4e). tea1p is thought to be important for the cell-tip localization of several other effector proteins involved in polarized growth, including tip1p, tea2p, tea3p, pom1p and bud6p9,10,16,17,18. In tip1Δ cells, in which short microtubules lead to defects in tea1p targeting to cell tips10, cortical GFP–mod5p was similarly mislocalized, but in tea2-1 mutants, which have a similar microtubule defect, GFP–mod5p localization was nearly normal (see Supplementary Information), indicating that the role of tip1p in mod5p localization might extend beyond the regulation of microtubule dynamics. tea3Δ cells showed a slight spreading of GFP–mod5p away from cell tips (Fig. 4f; see also Supplementary Information), indicating that tea3p might also contribute to restricting mod5p to cell tips (see below). In pom1Δ and bud6Δ strains we observed normal GFP–mod5p localization (see Supplementary Information), indicating that they might not be directly involved in the tea1p-mediated regulation of mod5p localization.

Collectively, these results suggest a model of fission yeast cell polarity regulation involving a positive-feedback loop in which cortically localized mod5p at cell tips promotes the anchoring of microtubule-delivered tea1p. Correctly anchored tea1p then acts reciprocally to prevent mod5p from spreading out across the plasma membrane, and this spatial restriction of mod5p ultimately leads to the proper subsequent anchoring of additional tea1p (see Supplementary Information for model). So far we have not been able to demonstrate a biochemical association between mod5p and tea1p, but mod5p is extracted from the cell cortex only under partly denaturing conditions (H.A.S., unpublished observations). In a two-hybrid screen we have identified an interaction between mod5p and tea3p (H.A.S., unpublished observations), which is related in structure to tea1p and interacts with tea1p by two-hybrid analysis, although this has not yet been confirmed biochemically16. This indicates that mod5p and tea1p might physically interact only indirectly, possibly linked by mutual interactions with tea3p. Recently it has been shown that the deletion of a coiled-coil region at the C terminus of tea1p prevents it from anchoring at the cell cortex8; thus, an interaction between mod5p and tea1p might occur through this region of tea1p.

We suggest that the tea1p–mod5p system provides spatially selective anchoring of tea1p at cell tips in the context of dynamic targeting. If tea1p were simply deposited at the cortex after associating with the plus ends of dynamic microtubules, the partly stochastic nature of the orientation of microtubule growth would leave open the possibility of mistargeting, and thus only low-fidelity positioning of tea1p. However, we would argue that this is held in check by the localization interdependence between tea1p and its cortical anchoring factor, mod5p. Interactions between plus-end-associated microtubule-binding proteins and cortical or plasma-membrane proteins, such as the association of EB1 family proteins with adenomatous polyposis coli (APC) protein (or Kar9p in budding yeast)19,20,21,22, or of CLIP-170 family proteins with CLIP-associated proteins (CLASPs) and Ras GTPase-activating-protein-like protein (IQGAP1) (refs 23, 24), are thought to be important in a variety of microtubule-based eukaryotic cell behaviours, including mitotic spindle and/or nuclear positioning, and directed cell migration (see ref. 3 for references). It is thought that in most instances the purpose of these interactions is to regulate microtubule dynamics and/or the attachment of microtubules to subcellular structures2,3,25,26. In contrast, the primary role of cortical mod5p might be to increase the fidelity of the positioning of microtubule-targeted tea1p in the cell, thus ensuring that the correct identity of the polarized cell tip is continuously asserted. In conjunction with the selective stabilization of microtubule dynamic instability27, spatially selective anchoring of microtubule-targeted proteins might be a more general principle regulating cell polarity.


Yeast methods

S. pombe methods were as described28. Mutant cells were identified by starving cells on rich medium plates (YE5S) for 2 days, replica-plating them to fresh plates, and examining their cell shape after 4 h at 32 °C. Deletion of mod5+ and GFP-tagging of mod5p and tea1p were performed by polymerase chain reaction (PCR)-based gene targeting with the kanMX selectable marker29; strains and primers used are listed in Supplementary Information. The C terminus of tea1p was tagged with GFP or YFP. Because of the CaaX box at the C terminus of mod5p, the N terminus of mod5p was tagged with GFP, and GFP–mod5p was expressed under the control of the weak nmt81 promoter. Before analysis, expression of GFP–mod5p was induced for 2 days at 32 °C in medium lacking thiamine. The nmt81 promoter drives the expression of GFP–mod5p to a concentration about threefold that of the endogenous protein (see Supplementary Information; additional data not shown). Deletion of the C-terminal CaaX box and mutation of C519S in mod5 were achieved by PCR-based integration with a ura4+ selectable marker29, and confirmed by sequencing. Because steady-state expression concentrations of these mutant mod5p proteins from the endogenous mod5+ promoter are lower than in wild-type cells, expression was restored to wild-type levels by replacing the endogenous promoter with the nmt41 promoter (data not shown).

In polarity re-establishment experiments, strains were grown to stationary phase in YE5S liquid medium for 3 days at 25 °C before dilution 1:20 into fresh medium at 32 °C, in the presence or absence of 50 µg ml-1 MBC9. Cells were fixed with formaldehyde and branching was scored by light microscopy.


Anti-tea1p antibodies were raised in sheep against His6-tea1p (ref. 4). Anti-tea1p antiserum gave only background staining in tea1Δ cells (see Supplementary Information), and was used at a dilution of 1:1,000. TAT1 anti-tubulin hybridoma supernatant30 was a gift from K. Gull, University of Manchester, UK, and used at a dilution of 1:30. Anti-mod5p antibodies were raised in sheep against a GST fusion protein containing amino acids 28–426 of mod5p. The IgG fraction from anti-mod5p antiserum was purified on Protein G–agarose, and used at a dilution of 1:100. Alexa-conjugated fluorescent secondary antibodies (Molecular Probes) were used for immunofluorescence.


For immunostaining with anti-tea1p and anti-tubulin antibodies, cells were harvested by filtration and fixed in methanol at -80 °C for more than 10 min before being processed essentially as described5. Twelve Z-sections at 0.3-µm spacing were captured with a Leica TCS-SP confocal microscope and projected onto a single plane. Because neither mod5p nor GFP–mod5p could be preserved by methanol or aldehyde fixation methods (data not shown), for immunostaining with anti-mod5p antibodies, cells were harvested by filtration and fixed overnight in 10% trichloroacetic acid at 4 °C (ref. 16), then processed as above. Anti-mod5p immunofluorescence images were collected with a Nikon TE300 microscope equipped with Chroma filters, a PiFoc piezofocus device (Physik Instrumente), and a CoolSnapHQ camera (Roper Scientific), controlled by MetaMorph software (Universal Imaging). Seven Z-sections at 0.2-µm spacing were projected onto a single plane. Live-cell imaging of GFP, YFP or CFP was performed on the Nikon TE300 microscope. For time-lapse acquisition of tea1p–GFP, three Z-sections at 0.5-µm spacing were collected with 600-ms exposure time, at 10-s intervals, with appropriate neutral density filters. For time lapse acquisition of cells doubly labelled with tea1p–YFP and CFP–tubulin, two Z-sections at 0.6-µm spacing were collected with exposure times of 1,250 ms for YFP and 1,000 ms for CFP, at 15-s intervals, with neutral density filters. CFP–atb2p was expressed from a plasmid (pRL72 (ref. 18), a gift from F. Chang, Columbia University, USA) under control of the nmt1 promoter induced for 20–24 h at 25 °C in 50 nM thiamine.

For measurements of the number of tea1p–GFP particles per cell, particles were counted from Z-stacks containing the entire cell volume; cell tips (terminal 10% of cell length) were not included.