Two distinct regions of Mto1 are required for normal microtubule nucleation and efficient association with the γ-tubulin complex in vivo

Microtubule (MT) nucleation in eukaryotic cells involves the γ-tubulin complex (γ-TuC), a multi-protein complex containing γ-tubulin and several additional proteins (for reviews, see Luders and Stearns, 2007; Raynaud-Messina and Merdes, 2007; Wiese and Zheng, 2006). In higher eukaryotes, the γ-TuC exists primarily as a large (∼2 MDa) complex, the γ-tubulin ring complex (γ-TuRC). The γ-TuRC consists of several copies of a smaller complex (γ-tubulin small complex; γ-TuSC) containing γ-tubulin and the proteins GCP2 and GCP3, plus other proteins that are γ-TuRC specific. Obvious homologues of γ-TuSC proteins are found in all eukaryotes, whereas homologues of γ-TuRC-specific proteins may not be ubiquitous – for example, they have not been found in budding yeast, which has a relatively simple mode of microtubule organization. However, they have been identified and characterized in the fission yeast Schizosaccharomyces pombe, although they are significantly diverged from their higher eukaryotic counterparts (Anders et al., 2006; Fujita et al., 2002; Venkatram et al., 2004).

Fission yeast serves as a useful model organism for understanding microtubule nucleation by the γ-TuC, not only because of its conservation of γ-TuC components but also because of its diversity of cytoplasmic microtubule organizing centres (MTOCs) (for a review, see Sawin and Tran, 2006). Only a small number of cytoplasmic MTs are present at any given time in fission yeast (Hoog et al., 2007), and these can be imaged by expressing GFP-tubulin at physiological concentrations. During interphase, MTs are nucleated not only from the spindle pole body (SPB; the yeast centrosome equivalent) but also from several interphase microtubule organizing centre `satellite' particles (iMTOCs) present on the nuclear envelope, in the cytoplasm and on MTs themselves. During late mitosis, cytoplasmic astral MTs analogous to those seen in budding yeast are nucleated from the duplicated SPBs. At the end of cell division, a post-anaphase array (PAA) of MTs is nucleated from the equatorial MTOC (eMTOC) at the site of septation. Because fission yeast MTOCs change throughout the cell cycle, they may provide a useful model for understanding the changes in MT organization that accompany differentiation of many higher eukaryotic cell types, for which relatively little is known, especially in relation to non-centrosomal MT nucleation (reviewed by Luders and Stearns, 2007; Raynaud-Messina and Merdes, 2007; Wiese and Zheng, 2006).

Two key regulators of cytoplasmic MT nucleation in fission yeast are the non-essential proteins Mto1 (initially called Mbo1 or Mod20) and Mto2 (Janson et al., 2005; Samejima et al., 2005; Sawin et al., 2004; Venkatram et al., 2004; Zimmerman and Chang, 2005). In a variety of assays mto1Δ mutants are completely defective in nucleation of MTs from all types of cytoplasmic MTOCs (Sawin et al., 2004; Venkatram et al., 2004). However, mto1Δ mutants are viable, because intranuclear mitotic spindle assembly and function is independent of Mto1; intranuclear MT nucleation is most probably driven by an Mto1 paralogue, Pcp1, which is present on the nucleoplasmic face of the SPB (Flory et al., 2002). Mto1 localizes to cytoplasmic MTOCs and co-immunoprecipitates the γ-TuC, and collectively these results have led to a model in which Mto1 functions by recruiting the γ-TuC to prospective cytoplasmic MTOCs (Sawin et al., 2004; Venkatram et al., 2004; Zimmerman and Chang, 2005).

Mto2 colocalizes with and physically interacts with Mto1, almost certainly directly (Samejima et al., 2005; Venkatram et al., 2005). In addition, Mto2 can co-immunoprecipitate the γ-TuC (Janson et al., 2005), and in mto2Δ mutants, Mto1 fails to co-immunoprecipitate the γ-TuC (Samejima et al., 2005). This could suggest a simple model in which Mto2 provides a `bridge' between Mto1 and the γ-TuC. However, unlike mto1Δ mutants, mto2Δ mutants are defective in nucleation from only a subset of cytoplasmic MTOCs; they can nucleate MTs from the SPB in both interphase and mitosis, and they can also nucleate a small number of MTs from the eMTOC at the end of mitosis. Thus, an alternative interpretation, based primarily on mutant phenotypes, might be that Mto1 retains some ability to recruit and/or activate the γ-TuC in the absence of Mto2 (e.g. at the SPB), even if to date this has not been detected biochemically.

Mto1 is a large coiled-coil protein with an ∼60-amino acid amino-terminal region (centrosomin motif 1; CM1) that is conserved in fungal and animal cells (Sawin et al., 2004; Zhang and Megraw, 2007). Proteins containing CM1 are generally similar in structure: they are all large proteins (100-300 kDa); they contain several predicted coiled-coil regions; and CM1 is typically found towards the amino terminus of the coding region. Most of these proteins have not been extensively studied, but those that have been characterized, including the Drosophila protein Centrosomin (Cnn) and the human proteins myomegalin and CDK5RAP2, have been implicated in microtubule organization, centrosome function and/or brain development (Sawin et al., 2004; Bond et al., 2005; Ching et al., 2000; Fong et al., 2008; Megraw et al., 1999; Terada et al., 2003; Veith et al., 2005; Verde et al., 2001; Zhang and Megraw, 2007). To address the question of how Mto1 and Mto2 may function in MT nucleation in more detail, we constructed site-directed mutations in Mto1 and examined the effects of the mutations on protein localization, interaction and function using a range of cytological and biochemical assays. Our data suggest not only that the conserved CM1 region is critical for Mto1 interaction with the γ-TuC, but also that binding of Mto2 to Mto1 is equally important for interaction with the γ-TuC. Further experiments suggest that Mto1 and Mto2 act cooperatively to promote interaction of an Mto1/2 complex with the γ-TuC, and thereby efficient MT nucleation.

We targeted the CM1 region of the single chromosomal copy of fission yeast mto1⁺ to construct three different site-directed mutants: mto1-9A1, mto1-9A2 and mto1-9A3 (Fig. 1A,B). Each mutation converts nine consecutive amino acids of CM1 to alanine. In a morphology assay, mto1-9A3 cells were slightly curved relative to wild-type cells, whereas mto1-9A1 and mto1-9A2 cells were strongly curved, like mto1Δ cells (Fig. 1C) (Sawin et al., 2004). This suggests that mto1-9A1 and mto1-9A2 mutations strongly impair Mto1 function, whereas the mto1-9A3 mutation has only a mild effect.

Upon cold-induced MT depolymerization in wild-type cells, Mto1 and associated γ-TuC proteins redistribute to sites on the surface of the nuclear envelope, and when cells are re-warmed, MTs are rapidly nucleated from these sites (Fig. 1D; supplementary material Fig. S1) (Sawin et al., 2004). By contrast, in the same MT-regrowth assay mto1Δ mutants are completely defective in cytoplasmic MT nucleation and instead slowly nucleate intranuclear MTs (most probably via Pcp1). These intranuclear MTs eventually escape into the cytoplasm, acting as progenitors of essentially all cytoplasmic MTs in the mutant cells and forming abnormal MT bundles that often curve around the cell tips (Fig. 1D; during steadystate growth of mto1Δ mutants, intranuclear spindle MTs escaping into the cytoplasm at the end of mitosis perform an equivalent function) (Sawin et al., 2004; Zimmerman and Chang, 2005). In the MT-regrowth assay we found that mto1-9A1 and mto1-9A2 mutants were indistinguishable from mto1Δ cells, both before cold-induced MT depolymerization and during recovery (Fig. 1D). mto1-9A3 mutants behaved like wild-type cells, except that both before depolymerization and after repolymerization, some cells in mto1-9A3 mutants had curved microtubule bundles. However, this was not as extreme as in mto1Δ, mto1-9A1 or mto1-9A2 mutants (Fig. 1D).

Analysis of movies of live cells co-expressing GFP-tubulin (GFP-Atb2) (Sawin et al., 2004) and the SPB marker Sad1-dsRed (Chikashige et al., 2004) also showed mto1-9A1 and mto1-9A2 mutants to be indistinguishable from mto1Δ cells. All three of these mutants showed essentially no new nucleation of interphase cytoplasmic MTs at steady state, while mto1-9A3 mutants showed approximately half the nucleation frequency of wild-type cells (Fig. 1E; supplementary material Movies 1-5; see Materials and Methods). Consistent with this, mto1-9A1 and mto1-9A2 mutants also showed the unusual interphase MT behaviours characteristic of mto1Δ mutants, such as bend-breakage when MT bundles curve around cell tips (Sawin et al., 2004), and a strong reduction in SPB oscillations, which probably reflects reduced interactions between cytoplasmic MTs and the SPB (Fig. 1E; supplementary material Movies 2-4). In dividing cells, mto1-9A1 and mto1-9A2 mutants, like mto1Δ cells, did not nucleate either cytoplasmic astral MTs or PAA MTs (Sawin and Tran, 2006), whereas mto1-9A3 mutants behaved like wild-type cells (Fig. 1E; supplementary material Movies 1-5). Taken together, these results show that mto1-9A1 and mto1-9A2 mutants are null for Mto1-dependent MT nucleation, whereas mto1-9A3 mutants are partially functional.

We investigated the ability of the Mto1-mutant proteins to physically interact with Mto2 and the γ-TuC (Samejima et al., 2005; Vardy and Toda, 2000; Venkatram et al., 2005). In immunoprecipitation experiments with Myc-tagged Mto1 proteins expressed from the endogenous mto1⁺ promoter, we found that all of the mutant proteins were present at normal wild-type levels, and all co-immunoprecipitated Mto2 to the same extent as wild-type cells (Fig. 2A). However, although both wild-type Mto1-Myc and Mto1-9A3-Myc co-immunoprecipitated γ-tubulin (Gtb1) and HA-tagged Alp4 (GCP2 homologue), a well-characterized component of the fission yeast γ-TuSC (Vardy and Toda, 2000), Mto1-9A1-Myc and Mto1-9A2-Myc did not co-immunoprecipitate these γ-TuC components (Fig. 2A).

We next examined the localization of the Mto1-mutant proteins. In the MT regrowth assay described above, all Mto1-mutant proteins redistributed to the nuclear envelope upon cold-induced MT depolymerization, as in wild-type cells (supplementary material Fig. S1) (Sawin et al., 2004). To investigate localization more closely in live cells, we tagged the Mto1-mutant proteins with GFP and also tagged Alp4 with the RFP tandem-dimer Tomato (tdT) (Shaner et al., 2004; Snaith et al., 2005). In wild-type cells we observed colocalization of Mto1-GFP with Alp4-tdT at SPBs, eMTOCs, and at iMTOC satellites (Fig. 2B,C) (Zimmerman and Chang, 2005). Interestingly, although most Mto1-GFP interphase satellites contained Alp4-tdT, some satellites did not, suggesting that the γ-TuC is not always stably associated with Mto1 at prospective MTOCs. This result differs slightly from a previous study using CFP- and YFP-tagged proteins (Zimmerman and Chang, 2005) and may be attributable to our improved imaging conditions.

In mto1 mutants, we found that Mto1-9A1-GFP, Mto1-9A2-GFP and Mto1-9A3-GFP localized to SPBs, to the eMTOC and to interphase satellites that appeared to track the cytoplasmic MTs formed in mto1 mutants by the nuclear escape mechanism (Fig. 2D-F). However, we did not observe any Alp4-tdT associated with either equatorial or interphase-satellite Mto1-9A1-GFP or Mto1-9A2-GFP. We did observe Alp4-tdT at SPBs, consistent with previous observations of Alp4-GFP and other γ-TuC proteins at SPBs in mto1Δ mutants; this most probably represents γ-TuC interacting with Pcp1 on the nucleoplasmic face of the SPB (Flory et al., 2002; Sawin et al., 2004; Venkatram et al., 2004; Zimmerman and Chang, 2005).

Collectively, these results indicate that the failure of Mto1-9A1 and Mto1-9A2 to associate with γ-TuC is not due to a mislocalization of the Mto1-mutant proteins, and thus that CM1 is important for γ-TuC binding per se. In addition, the demonstration that Mto1 can localize to prospective MTOCs independent of its ability to bind the γ-TuC supports our previous hypothesis that Mto1 is `upstream' of the γ-TuC in determining sites of cytoplasmic MT nucleation (Samejima et al., 2005; Sawin et al., 2004).

In order to test whether the CM1 region of Mto1 is sufficient for interaction with the γ-TuC, we created several different Myc-tagged carboxyl-terminal truncations of Mto1, expressed from the endogenous mto1 promoter, and performed immunoprecipitation experiments using strains also expressing HA-tagged Alp4 (Fig. 3; see also Fig. 1A). All of the truncated Mto1 proteins were efficiently immunoprecipitated, some much better than full-length Mto1. Interestingly, although all but the shortest truncation contained the entire CM1 region, we found that truncations up to and including Mto1-(1-500)-Myc did not co-immunoprecipitate Alp4-HA, whereas Mto1-(1-800)-Myc and larger proteins did (Fig. 3). Thus, in spite of the clear importance of CM1 for the interaction of Mto1 with the γ-TuC, additional sequences beyond CM1 are also important for this interaction.

Previously we showed that amino acids 476-549 of Mto1 are important for binding to Mto2 in vitro (Samejima et al., 2005). This suggested the possibility that in addition to the CM1 region, the Mto2-binding region of Mto1 might also be required for the interaction of Mto1 with the γ-TuC. We therefore wanted a means to test directly whether binding of Mto2 to Mto1 is required for Mto1 association with the γ-TuC, and/or whether Mto2 has additional roles in cytoplasmic microtubule nucleation independent of its association with Mto1.

We constructed several small deletions within the putative Mto2-binding region of Mto1 and targeted these to the chromosomal copy of mto1⁺ (Fig. 1A; additional data not shown). An allele deleting amino acids 523-537, mto1-334, produced a protein that failed to interact with Mto2 in immunoprecipitation experiments (Fig. 4A) as well as in vitro (data not shown). The mto1-334 mutant also showed cell-shape defects like other mto1 and mto2 mutants (Fig. 4B; Fig. 1C). In addition, we found that Myc-tagged Mto1-334 did not co-immunoprecipitate the γ-TuC (Fig. 4C).

Thus far, these results would be consistent with mto1-334 being a null allele of mto1. However, in MT regrowth assays we found that mto1-334 mutants, like mto2Δ mutants, were able to nucleate cytoplasmic MTs from the SPB, but not from non-SPB sites (Fig. 4D) (Samejima et al., 2005). By contrast, mto1Δ mutants cannot nucleate MTs from any cytoplasmic sites (Fig. 1D) (Sawin et al., 2004). In addition, live mto1-334 mutant cells expressing GFP-Atb2 and Sad1-dsRed displayed a profile of cytoplasmic MT-nucleation behaviours identical to mto2Δ mutants but distinct from mto1Δ mutants. This included: nucleation of cytoplasmic MTs during interphase, but only from the SPB (often accompanied by strong SPB oscillations); nucleation of cytoplasmic astral MTs in mitosis; and very limited MT nucleation from the eMTOC at the end of mitosis (Fig. 4E; supplementary material Movies 6 and 7).

Upon cold-induced depolymerization of MTs in MT-regrowth assays, Mto1 in mto2Δ mutants does not redistribute to the surface of the nuclear envelope, although the biological significance of this is not known (Samejima et al., 2005) (see Discussion). In this assay, Mto1-334 localization in mto2⁺ (wild-type) cells was identical to Mto1 (wild-type) localization in mto2Δ cells (supplementary material Fig. S1). In addition, when we tagged Mto1-334 with GFP, we found that the localization of Mto1-334-GFP in mto2⁺ (wild-type) cells was identical to Mto1-GFP (wild-type) in mto2Δ cells, as was the localization of Alp4-tdT in the two strains (Fig. 4F). Collectively these results suggest that mto1-334 is not a null allele but rather an interaction-specific allele of mto1 that disrupts binding of Mto1 to Mto2. Because the mto1-334 phenotypes (i.e. in a mto2⁺ background) are indistinguishable from those of mto2Δ, we conclude that the function of Mto2 in MT nucleation is mediated primarily, if not exclusively, via its binding to Mto1. Moreover, the failure of Mto1-334 to immunoprecipitate the γ-TuC indicates that the Mto1-Mto2 interaction is required for an efficient association of Mto1 with the γ-TuC, as detected in cytoplasmic extracts.

In experiments in which anti-Mto2 immunoprecipitates from cell extracts were washed with different salt concentrations, we found that Mto1 is much more stably associated with Mto2 than are γ-TuC proteins, suggesting that Mto1 and Mto2 may be part of a distinct protein complex or subcomplex that can interact with the γ-TuC (Fig. 5A). To investigate this further we constructed transheterozygous diploid strains, in order to determine whether both the CM1 region and the Mto2-binding region of Mto1 need to be present on the same polypeptide for proper Mto1 function. This was the case, as all mto1 mutants were complemented by wild-type mto1⁺, whereas we did not observe any complementation in the trans-heterozygous mutant strains (Fig. 5B,C). This suggests that CM1 and the Mto2-binding region work closely together in a common function to promote interaction of Mto1 with the γ-TuC.

We therefore considered the possibility that Mto1 and Mto2 might each bind the γ-TuC independently, albeit very weakly, achieving stronger, cooperative binding when both species are present in an Mto1/2 complex. We TAP-tagged Mto1 and Mto2 (Gould et al., 2004; Venkatram et al., 2005) and partially purified TAP-tagged Mto1 from wild-type and mto2Δ cells, as well as TAP-tagged Mto2 from wild-type and mto1Δ cells. The TAP-tagged proteins were expressed at physiological levels (Fig. 6A), and these partial purifications gave higher yield and purity than conventional immunoprecipitation methods used in this and previous work (data not shown). We probed western blots of the purifications with antibodies to γ-tubulin as well as with newly raised antibodies to Alp4 and Alp6 (GCP2 and GCP3 homologues; Fig. 6B,C; see Materials and Methods for technical details). As expected from previous studies (Janson et al., 2005; Sawin et al., 2004; Venkatram et al., 2004), the γ-TuC proteins co-purified both with Mto1-TAP and with Mto2-TAP (Fig. 6B, lanes 1, 2). Under normal exposure conditions, we did not detect an association between Mto1 and γ-TuC in mto2Δ strains (Fig. 6B, lane 3), consistent with previous immunoprecipitation experiments (Samejima et al., 2005). Moreover, we did not observe an association between Mto2 and the γ-TuC in mto1Δ strains (Fig. 6B, lane 4). Thus not only does Mto1 association with γ-TuC depend on Mto2, but also Mto2 association with γ-TuC depends on Mto1. Interestingly, however, with more sensitive detection and longer exposures (see Materials and Methods) we found evidence for a physical interaction between Mto1 and the γ-TuC in the absence of Mto2, and also a possible weaker interaction between Mto2 and the γ-TuC in the absence of Mto1 (Fig. 6C, lanes 3, 4). We did not detect the γ-TuC in purifications from several negative-control strains, including both untagged strains and unrelated Mod5-TAP strains (Snaith and Sawin, 2003) (also Hilary Snaith, University of Edinburgh, UK, personal communication) used to assess non-specific binding of γ-TuC to TAP-tagged proteins as well as possible unforeseen aberrant properties (e.g. `stickiness') of γ-TuC subunits in mto1Δ and/or mto2Δ backgrounds (Fig. 6B,C, lanes 5-7). These results suggest that Mto1 and Mto2 may bind cooperatively to the γ-TuC.

The work presented here provides novel insights into how fission yeast proteins Mto1 and Mto2 act to promote cytoplasmic MT nucleation via interactions with the γ-TuC. Our cytological and biochemical analysis of mto1-9A1, mto1-9A2 and mto1-334 mutants demonstrates that two distinct regions of Mto1 – the CM1 region and the Mto2-binding region – are required for an efficient association of Mto1 with the γ-TuC. mto1-9A1 and mto1-9A2 mutants completely fail to nucleate cytoplasmic MTs (Table 1), and Mto1-9A1 and Mto1-9A2 mutant proteins do not interact with the γ-TuC, although the mutant proteins still interact with Mto2 and localize to all prospective MTOC sites in vivo. Together with the conservation of CM1 and work in other organisms (see below) these results implicate the CM1 region of Mto1 as a likely γ-TuC-binding region. At the same time, however, our characterization of mto1-334 mutants, which phenocopy mto2Δ mutants (Table 1), demonstrates that binding of Mto1 to Mto2 is of comparable importance in promoting the interaction of Mto1 with the γ-TuC.

The stability of the Mto1-Mto2 interaction, together with our observation that the CM1 region and the Mto2-binding region of Mto1 must be present on the same polypeptide in order to achieve proper Mto1 function, suggested that Mto1 and Mto2 might form a complex in which Mto1 and Mto2 bind independently but cooperatively to the γ-TuC. Accordingly, using TAP purifications and sensitive detection methods we obtained evidence for an association of Mto1 with the γ-TuC in mto2Δ mutants, and an association of Mto2 with the γ-TuC in mto1Δ mutants. In both cases, however, these associations were significantly weaker than those observed in wild-type cells. A model of cooperative binding of an Mto1/2 complex to the γ-TuC explains these results, as well as several aspects of both mto1Δ and mto2Δ phenotypes and the differences between them (Janson et al., 2005; Samejima et al., 2005; Sawin et al., 2004; Venkatram et al., 2004; Venkatram et al., 2005; Zimmerman and Chang, 2005). In mto2Δ mutants, the limited amount of cytoplasmic MT nucleation observed, primarily at the SPB, can be attributed to a high local concentration of Mto1 weakly interacting with the γ-TuC, probably via CM1, and possibly supplemented by additional mechanisms that may be specific to the SPB. By contrast, the absence of cytoplasmic MT nucleation in mto1Δ mutants can be understood if Mto2 were able to bind the γ-TuC weakly but not `activate' it for nucleation; in any case, because the localization of the Mto1/2 complex is governed primarily by Mto1 and not by Mto2 (Samejima et al., 2005) (see also below), the γ-TuC would not be recruited and/or concentrated to specific subcellular sites in mto1Δ cells. In the future it will be of interest to investigate the nature of the interactions between Mto1, Mto2 and the γ-TuC at a deeper molecular level. Thus far, we have not been able to detect interaction of either Mto1 or Mto2 with individual components of the γ-TuC in directed yeast-two hybrid assays or in vitro binding experiments (our unpublished data). Given the weakness of the interactions detected in the Mto1- and Mto2-TAP purifications from mto2Δ and mto1Δ strains, addressing this question may require additional and/or novel methods.

One aspect of our work suggests that Mto2 could have additional roles in promoting Mto1 function beyond that suggested here, although such potential roles remain ill-defined. Specifically, compared with wild-type cells, Mto1 (or Mto1-334) in mto2Δ and mto1-334 mutants does not accumulate on the nuclear envelope upon cold-induced MT depolymerization (supplementary material Fig. S1) (Samejima et al., 2005). Although this is admittedly an `artifactual' situation, and nearly all other aspects of Mto1 localization remain unchanged in mto2Δ or mto1-334 mutants, in these mutants there also appear to be fewer Mto1-GFP (or Mto1-334-GFP) particles associated with interphase MTs during steadystate growth (Fig. 2C, Fig. 4F) (see also Samejima et al., 2005). The molecular basis for these differences is unknown. We note, however, that comparable differences in Mto1 localization are not observed with Mto1-9A1 and Mto1-9A2 proteins (Fig. 2D,E; supplementary material Fig. S1); it is therefore unlikely that the effects of disrupting Mto1-Mto2 interaction on these more subtle aspects of Mto1 localization are directly linked to aberrant MT distribution and/or reduced interaction of Mto1 with the γ-TuC. Further work on the biochemistry of the Mto1/2 complex may help to illuminate whether and how Mto2 may further contribute to Mto1 function.

While this paper was in preparation, two studies were published investigating the role of the CM1 region in higher eukaryotic proteins. Although there are some general similarities between these studies and our own, there are also some important differences. A cytological study in Drosophila showed that a deletion within the CM1 region of the Cnn protein leads to reduced and/or altered recruitment of γ-tubulin and two other proteins, D-TACC and Msps, to mitotic spindle poles, with a concomitant impairment of some microtubule-dependent processes (Zhang and Megraw, 2007). Allowing for the different nature of the fly and yeast systems, these results are consistent with our own. However, in the Drosophila work, there was no biochemical investigation of protein-protein interactions. In addition, the single deletion mutation within CM1 did not indicate whether all of CM1 or only a particular sub-region was important for function. By contrast, we have shown that mto1-9A1 and mto1-9A2 mutants are functionally null, whereas mto1-9A3 mutants, which are altered in an immediately adjacent part of CM1, have only slightly impaired function. Moreover, in immunoprecipitation experiments we have shown a specific defect in the physical interaction of Mto1-9A1 and Mto1-9A2 with the γ-TuC, with no effects on the interaction with Mto2. Thus our findings significantly extend the mutational analysis of CM1 in Drosophila.

Another recent study, in human cells, has provided evidence that the CM1 region of one of the two human Mto1/Cnn-related proteins, CDK5RAP2, is itself sufficient to interact with the γ-TuRC (Fong et al., 2008), at least when overexpressed well above probable physiological levels. These findings are also consistent with our own, although to date we have not been able to demonstrate convincingly an interaction between the fission yeast γ-TuC and bacterially expressed Mto1 fragments containing CM1 (our unpublished data). Beyond this, however, we have further shown that in fission yeast, an additional region of Mto1, which binds the interacting partner Mto2, is also required for efficient association of Mto1 with the γ-TuC, at least when Mto1 is present at physiological levels. It will be interesting to determine whether the proposed mechanism of cooperative interaction of the Mto1/2 complex with the γ-TuC in fission yeast also applies in some form to higher eukaryotic proteins. Presently there are no recognizable homologues of Mto2 in higher eukaryotes, or even in fungi (our unpublished data). However, it is possible that these organisms possess functional equivalents of Mto2 that are highly diverged in sequence, as several fission yeast proteins involved in MT nucleation are only poorly related in primary sequence to their higher-eukaryotic counterparts. Examples of this include the non-essential fission yeast proteins Gfh1, Mod21 and Alp16, which have been shown by combined biochemical, cytological and genetic analyses to be orthologues of mammalian γ-TuC proteins GCP4, GCP5 and GCP6, respectively, even though the actual amino acid sequence identity between the fission yeast and mammalian proteins is too low to be convincing in its own right (Anders et al., 2006; Fujita et al., 2002; Venkatram et al., 2004). Another example is Mto1 itself, in which only the CM1 region is conserved from yeast to higher eukaryotes, even though all of the CM1-containing proteins are generally similar in their overall predicted structural features. Indeed, CDK5RAP2 has recently been identified as a particularly rapidly evolving protein in primates, which may relate to its role in regulating brain size (Bond et al., 2005; Evans et al., 2006). Future biochemical analyses of the higher eukaryotic homologues of Mto1 should shed light on this question.

Standard yeast genetic and physiological methods were used throughout (Moreno et al., 1991). Strains and genotypes are listed in supplementary material Table S1. The mto1-9A1, -9A2 and -9A3 mutants were constructed by a two-step gene replacement at the mto1 locus. In the first step, the CM1 region was deleted and replaced with the ura4⁺ gene, using PCR-based tagging. The ura4⁺ gene was then replaced by mutated CM1 sequences (Fig. 1B) that had been generated by mega-primer PCR. The resulting strains were confirmed by genomic sequencing. The mto1-334 mutant was constructed in a similar manner. In the first step, amino acids 494-600 of Mto1 were deleted and replaced with the ura4⁺ gene. In the second step, the ura4⁺ gene was replaced with a linearised plasmid fragment from which the coding sequence for amino acids 523-537 (LMRMEQQWREDVDQL) had been deleted. The resulting strains were confirmed by colony PCR and western blotting. Diploid strains were made by ade6-210-ade6-216 transcomplementation.

PCR-based tagging and truncations of genes were performed using methods and template plasmids as described previously (Bahler et al., 1998; Brune et al., 2005; Snaith et al., 2005). Apart from the nmt81:GFP-tubulin (Sawin et al., 2004), all tagged proteins were tagged at their C termini and expressed from endogenous promoters. Tagging of the endogenous (essential) Alp4 with the tandem-dimer-Tomato (tdT) RFP tag (Shaner et al., 2004) did not lead to significant deleterious effects on MT behaviour. However, when crossing the Alp4-tdT strain with various mutants we sometimes noticed subtle effects on mutant cell morphology. This suggests that Alp4-tdT may have minor effects on MT function, as has been observed with other tagged versions of Alp4 (Anders et al., 2006).

Cell morphology assays and DIC imaging in return-to-growth experiments were performed as described previously (Anders et al., 2006; Snaith and Sawin, 2003); cells were grown to stationary phase in rich medium for 2–3 days and then refed with fresh medium, and fixed and imaged 3 hours after refeeding.

For Myc- and HA-tagged proteins, monoclonal antibodies 9E10 and 12CA5 were used. For Mto1 and Mto2, serum or affinity-purified polyclonal antibodies were used as described previously (Samejima et al., 2005; Sawin et al., 2004). For γ-tubulin (Gtb1), monoclonal antibody GTU-88 (Sigma) was used. Antibodies to Alp4 and Alp6 were raised by immunization of sheep with SDS-PAGE-purified GST-Alp4 and GST-Alp6 fusion proteins (from purified inclusion bodies), and either serum (anti-Alp4) or blot-affinity-purified antibody (anti-Alp6) was used for western blots.

For immunoprecipitations, extracts were made from liquid-nitrogen frozen cells as described previously (Anders et al., 2006; Samejima et al., 2005; Sawin et al., 2004). Protein A Sepharose, protein G Sepharose, or protein G Dynabeads were used for immunoprecipitations, depending on the primary antibody. For salt-washes of immunoprecipitations, two intermediate wash steps were supplemented with KCl to the final salt concentration indicated.

A single-step purification of TAP-tagged proteins, up to and including release of proteins bound to IgG beads by cleavage with TEV protease, was performed following the steps described by Gould et al. (Gould et al., 2004), with only minor modifications. In the experiments shown in Fig. 6 we note that different amounts of Mto1 and Mto2 are seen in lanes 1-4 under `TEV elution, short exposures'. These differences are likely to have a simple technical explanation. Comparing lanes 2 and 4, it can be seen that more Mto2(–TAP) is purified from mto1Δ cells than from wild-type cells, even though equal levels of Mto2(–TAP) are present in the cell extracts. This suggests that the TAP tag on Mto2 may not be easily accessible when Mto2-TAP is part of the Mto1/2 complex, consistent with previously published work that did not successfully detect γ-TuC proteins by mass spectrometry after Mto2-TAP purification (Venkatram et al., 2005). At the same time, comparing lanes 1 and 2, it can be seen that comparable amounts of Mto2 are purified from Mto1-TAP and Mto2-TAP purifications in wild-type cells, while in the same cells, significantly less Mto1 is purified from Mto2-TAP purification as compared with Mto1-TAP purification. We interpret this as follows. In the Mto1-TAP purification from wild-type cells, all of the Mto2 purified is derived from the Mto1/2 complex. In the Mto2-TAP purification from wild-type cells, most of the Mto2(–TAP) purified may be derived from `free', uncomplexed Mto2, because much of the Mto2(–TAP) that is present as Mto1/2 complex in these cell extracts is poorly accessible and is therefore not purified. Therefore, only a small proportion of Mto2(–TAP) that is purified here is derived from the Mto1/2 complex, and that is why there is less Mto1 in lane 2 than in lane 1. The increased Mto2 in lane 4 compared to lane 2 is, accordingly, the result of a higher proportion of total Mto2 in cell extracts being accessible to purification. According to this view, the different amounts of γ-TuC purified in lanes 1 and 2 would be correlated with the different amounts of total Mto1/2 complex that were purified. We also note that exposures were chosen to illustrate these intensity differences, and therefore intensities can be compared within panels but not between panels.

Microtubule cold-shock and regrowth experiments were performed exactly as described previously, using methanol fixation (Samejima et al., 2005; Sawin et al., 2004). For anti-tubulin immunofluorescence, monoclonal antibody TAT-1 was used (Woods et al., 1989), For anti-Mto1 immunofluorescence, affinity-purified polyclonal antibody to Mto1 was used (Sawin et al., 2004). TAT1 anti-tubulin-stained cells were further stained with Alexa-Fluor-488-labelled secondary antibody (Invitrogen) and observed by spinning-disk confocal microscopy, using a Nikon 100×/1.45 NA PlanApo objective on a Nikon TE2000 inverted microscope, connected to a Yokogawa CSU-10 spinning disk confocal head (Visitech) and a Coolsnap HQ CCD camera (Photometrics). Images were acquired using Metamorph software (Molecular Devices). Anti-Mto1-stained cells were further stained with Alexa-Fluor-568-labelled secondary antibody, counterstained with DAPI and observed by wide-field microscopy, using a Nikon 100×/1.40 NA Plan Apo objective on a Nikon TE300 inverted microscope, connected to a Coolsnap HQ CCD camera. Images were acquired using Metamorph software. Note that methanol fixation leads to `shrunken' cells relative to live-cell imaging; this can be seen comparing cell lengths to scale bars.

Live-cell imaging used a system and methods described previously (Anders et al., 2006). Cells were grown in minimal medium at room temperature (23–25°C). For videomicroscopy of cells expressing GFP-labelled tubulin (nmt81:GFP-Atb2, with the promoter induced) and RFP-labelled spindle pole bodies (Sad1-dsRed), 8 z-sections, 0.6 μm apart, were acquired every 20 seconds, using Metamorph software (Molecular Devices). Video sequences were deconvolved using Softworx (Applied Precision), and maximum projections were used for movies at 10 frames per second (i.e. 200× time compression relative to real time), using Quicktime 7.4.1 (Apple) and H.264 video compression. Maximum projections are also shown in the image stills of Fig. 1E and Fig. 4E. Still images of cells expressing Mto1-GFP or mutant Mto1-GFP with Alp4-tdT were acquired as above but not deconvolved. Maximum projections are shown, except for Fig. 2B. Apart from deconvolution for movies, all microscope images were adjusted only by linear contrast enhancement using Photoshop (Adobe), and all relevant panels within a given figure were adjusted identically, allowing direct comparison of intensities.

For measurements of new MT nucleation frequency in wild-type cells and mto1 mutants expressing GFP-Atb2 and Sad1-dsRed, 20 cells of each genotype were analysed from movies (Fig. 1E; supplementary material Movies 1-5). The appearance of new MT bundles was measured in each cell for 26 minutes (80 time points); values shown are the mean nucleation frequency per minute per cell (± s.d.) Obvious intranuclear MT nucleation, which is sometimes apparent in mto1Δ, mto1-9A1 and mto1-9A2 mutants, was ignored. The low levels of apparent cytoplasmic nucleation measured in mto1Δ, mto1-9A1 and mto1-9A2 mutants may reflect occasional mis-scored intranuclear MT nucleation (note that the standard deviation is greater than the mean for these mutants).

We thank Keith Gull, Tom Pollard, Hilary Snaith, Takashi Toda and Karsten Weis for useful reagents, strains and methods, Joanne Konkel for early contributions to this project, members of our laboratory for useful comments, and H. Ohkura and K. Hardwick for comments on the manuscript. V.J.M. and L.M.G. are supported by a Doctoral Training Grant from the BBSRC. K.E.S. is a Wellcome Trust Senior Research Fellow in Basic Biomedical Sciences. This work was supported by the Wellcome Trust.