Kinesin-14 and kinesin-5 antagonistically regulate microtubule nucleation by γ-TuRC in yeast and human cells

Bipolar spindle assembly is a critical control point for initiation of mitosis through nucleation and organization of spindle microtubules and is regulated by kinesin-like proteins. In fission yeast, the kinesin-14 Pkl1 binds the γ-tubulin ring complex (γ-TuRC) microtubule-organizing centre at spindle poles and can alter its structure and function. Here we show that kinesin-14 blocks microtubule nucleation in yeast and reveal that this inhibition is countered by the kinesin-5 protein, Cut7. Furthermore, we demonstrate that Cut7 binding to γ-TuRC and the Cut7 BimC domain are both required for inhibition of Pkl1. We also demonstrate that a yeast kinesin-14 peptide blocks microtubule nucleation in two human breast cancer cell lines, suggesting that this mechanism is evolutionarily conserved. In conclusion, using genetic, biochemical and cell biology approaches we uncover antagonistic control of microtubule nucleation at γ-TuRC by two kinesin-like proteins, which may represent an attractive anti-mitotic target for cancer therapies.

T he microtubule cytoskeleton is a self-assembling network that underlies specialized, often polarized, cellular functions in eukaryotes. Knowledge of its mechanisms is fundamental to understanding normal development and disease and is expected to assist new technologies through biomimicry. The microtubulebased mitotic spindle apparatus is perhaps the best studied selfassembly platform 1,2 and a primary target for cancer therapeutics 3 . Spindle pole microtubule-organizing centres (MTOCs) utilize a g-tubulin template within a ring complex (g-tubulin ring complex, g-TuRC) to orchestrate addition of a-/b-tubulin heterodimeric microtubule building blocks into 25 nm polarized microtubules [4][5][6][7][8][9] . Conserved protein structural features of the g-TuRC MTOC have been identified through crystallography studies from multiple model organisms and include a-/b-tubulin 10 , g-tubulin 11 , GCP4 (ref. 12) and the g-tubulin small complex (g-TuSC) cryo-EM structure 13 . Conserved structural features are additionally supported by cross-species analysis 14,15 . Still unknown is how dynamic control over MTOC functions for microtubule nucleation and organization is achieved. The fission yeast Schizosaccharomyces pombe provides an ideal eukaryotic platform to address conserved MTOC mechanisms [14][15][16][17] .
The coordination of spindle microtubules into a bipolar array requires kinesin-like proteins (Klps), though Klp mitotic functions are not limited to interactions solely on microtubules. Studies of the functionally diverse kinesin-14 Klp family across eukaryotes have indicated an ability by some members to affect microtubule number and organization at spindle poles [18][19][20][21] . In fission yeast, kinesin-14 Pkl1 interacts directly with the g-TuRC MTOC to alter its composition and function 17,22,23 . Conservation of the kinesin-14 g-TuRC regulatory mechanism is expected from yeast to human, as human kinesin-14 HSET replaces fission yeast kinesin-14 Pkl1 (ref. 23) and all human g-TuSC protein components are also compatible 14,15 . Nearly as ubiquitous and complex in eukaryotes as kinesin-14 Klps are members of the kinesin-5 family that oppose kinesin-14 function. In fission yeast, kinesin-5 Cut7 opposes the action of kinesin-14 Pkl1 in mitosis, but the detailed mechanism is not yet characterized. Elucidating this mechanism could be informative for understanding g-TuRC regulation and spindle bipolarity.
In this study, we expand the mechanism for kinesin-14 regulation of g-TuRC. Studies from our lab and others describe genetic interactions of Pkl1 with g-TuRC proteins 22,[24][25][26] , checkpoint pathways 20,26 and spindle pole organization 20 . More recently we identified key Tail elements in Pkl1 that function along with Motor binding to g-tubulin to regulate g-TuRC 17,22,23 .
Here we demonstrate that kinesin-14 Pkl1 asymmetrically blocks microtubule nucleation in vivo in fission yeast and that a kinesin-14 Pkl1 Tail peptide can similarly prevent nucleation and generate mitotic arrest in two human breast cancer cell lines. We reveal that, in fission yeast, kinesin-5 Cut7 counters Pkl1 ability to block nucleation by also associating with g-TuRC and binding similarly to g-tubulin. This counteraction requires the additional conserved kinesin-5 BimC domain. Balanced regulation by kinesin-14 Pkl1 and kinesin-5 Cut7 generates optimal mitotic fidelity, although both proteins are codispensable as determined by genetic analysis of single and double mutants, biochemical approaches and timelapse fluorescence microscopy. Analysis of pkl1D single and pkl1D cut7D double mutants also reveals separate mitotic roles for both kinesin-14 Pkl1 and kinesin-5 Cut7. Our findings identify kinesin-14 Pkl1 as a Klp-negative regulator of microtubule nucleation at g-TuRC and demonstrate conservation of this mechanism in human breast cancer cells, resulting in mitotic arrest. We expect these discoveries to be broadly relevant to the microtubule cytoskeleton field with potential as a novel strategy and target for future development in cancer therapeutics.

Results
Kinesin-5 is dispensable in the absence of kinesin-14 Pkl1. Spindle bipolarity in fission yeast requires kinesin-5 Cut7 (ref. 27). The mechanism underlying its essential nature remains unclear as another Klp, kinesin-6 Klp9, is capable of crosslinking antiparallel microtubules and is required for spindle elongation 28 . In eukaryotes, an opposing relationship between kinesin-5 and kinesin-14 Klps in microtubule regulation is highly conserved. We previously demonstrated that kinesin-14 Pkl1 directly binds and downregulates g-TuRC function 17,23 . We tested the hypothesis that a required role of kinesin-5 Cut7 (cut7 gene), which localizes at spindle poles, is to oppose kinesin-14 Pkl1 (pkl1 gene). By homologous recombination (Fig. 1a), we simultaneously deleted the cut7 gene while marking the locus with ura4 (cut7D::ura4 þ ) in a strain deleted for Pkl1 (pkl1D::his3 þ ). The pkl1D cut7D double mutant strain exhibits robust viability by serial growth assays, similar to wildtype cells (Fig. 1b). Spindle pole body (SPB) separation is not affected in the double mutant pkl1D cut7D versus wild type (Fig. 1c,d) nor is mitotic progression through anaphase affected as compared with wild type or pkl1D strains (Fig. 1c,e-g). However, spindle breakdown is delayed as indicated by a persistent spindle following anaphase B elongation (Fig. 1c,g). We demonstrate that kinesin-5 Cut7 is dispensable in the absence of Pkl1, indicating that a kinesin-5-independent mechanism for spindle assembly can exist in fission yeast. This also supports a required role for Cut7 that is to counter Pkl1, which is a direct negative regulator of the g-TuRC MTOC.
Kinesin-5 Cut7 binds c-TuRC through Motor and Tail domains. Kinesin-14 Pkl1 negatively regulates g-TuRC through two internal protein domains that include elements of its Motor and Tail regions 17,22 . To determine whether Cut7 binds the g-TuRC MTOC by a similar mechanism, we performed Fast Protein Liquid Chromatography (FPLC) 15,17 using V5-tagged deletion derivatives of Cut7 previously generated 29 , immunocytology using newly generated V5-tagged deletion and site-directed mutagenesis derivatives, and Pkl1 peptide co-immunoprecipitation assays (Fig. 2). Fractionation of whole-cell extracts carrying V5-tagged full-length Cut7 and two Cut7 truncation constructs (Cut7-Head-Stalk or Cut7HS, aa 1-88; Cut7-Stalk-Tail or Cut7-ST, aa 443-1,085) were examined in the pkl1D cut7D double mutant and in the gtb1-K5A strain that inhibits Pkl1 regulation of g-TuRC by blocking its Motor domain binding to g-tubulin (Figs 2a,b and 3a-c) 22 . A FLAG-tagged truncated Pkl1 construct that retains full function and localizes to g-TuRC (pkl1D95) 23 was used as a positive control along with Alp4 (gamma complex protein GCP2 yeast orthologue) and g-tubulin proteins that are core subunits of the g-TuSC and the 42,000 kDa high-molecular weight g-TuRC MTOC. The g-TuRC peaks in FPLC fractions 15 and 16 (Fig. 2b). Profiles of the three cut7 constructs in pkl1D cut7D double mutant cells that are pREP81V5/cut7, pREP81V5/cut7HS and pREP81V5/cut7ST are shown (Fig. 2a,b). Western blot analysis following FPLC fractionation reveals that Cut7, Cut7HS and Cut7ST exhibit similar high-molecular weight profiles as Pkl1D95 and all peak in identical fractions to core g-TuRC proteins g-tubulin and Alp4. To further confirm Cut7ST binding to g-TuRC, we performed coimmunoprecipitations from high-molecular weight FPLC fraction 15 in strain alp4-HA pREP81V5/cut7ST (Fig. 2d,e). Magnetic beads with histidine affinity were conjugated to small His-tagged Pkl1 Tail peptides (PgT or PgM) 17 that bind g-TuRC (PgT, targeting) or cannot interact with fission yeast g-TuRC (PgM, mutated). Using this approach, we detect g-TuRC core subunit proteins g-tubulin and Alp4-HA in addition to Cut7ST (V5 tag) by western blot analysis after elution off of beads (PgT). These proteins were not recovered by the mutated peptide, as expected. We were unable to detect a-/b-tubulin in high-molecular weight fractions following FPLC, which further suggests that the Klps detected were directly bound to g-TuRC.
We previously demonstrated that mutation of a conserved lysine residue to alanine in g-tubulin helix 11 (gtb1-K5A) abolishes Pkl1 Motor domain binding to g-TuRC and blocks its full function in vivo 22 . To determine whether Cut7 similarly binds to g-tubulin through helix 11, we examined its FPLC profile in the gtb1-K5A strain (Fig. 3a). The mutant g-tubulin-K5A fractionates similarly to wild-type g-tubulin by FPLC. The V5-Cut7 signal in high-molecular weight g-TuRC fractions is significantly reduced in the gtb1-K5A strain, whereas steady-state expression levels of V5-Cut7 in gtb1 wild type and gtb1-K5A mutant backgrounds by western blot analysis of whole-cell extracts is similar (Fig. 3c). Fluorescence microscopy of V5-Cut7 in the gtb1-K5A mutant reveals that, while pole localization is reduced, Cut7 retains localization to spindle microtubules, which is enhanced. Truncated Cut7HS that contains the Cut7 Motor and Stalk domains and lacks the Tail is completely absent from high-molecular weight g-TuRC fractions in the gtb1-K5A background (Fig. 3a). Similarly to V5-Cut7, it retains the ability to bind to spindle microtubules as observed by immunocytology (Fig. 3d) Figure 1 | Spindle assembly and cell viability remain high in the pkl1D cut7D strain. (a) cut7 knockout and integration of ura4 þ at this locus in the pkl1D::his3 þ single mutant. The PCR-based approach used long (500 nts) flanking tracts of homology and the genomic deletion/integration event was confirmed by colony PCR (lane 3: 5 0 , lane 5: 3 0 oligonucleotide pairs b and c, d and e, respectively). Lane 1 is a positive control using oligonucleotide pair a and b. Lanes 2 and 4 are negative controls using oligonucleotide pair b and c on pkl1D single-mutant cells. (b) Serial dilution growth assays at 25°C permissive and 36°C restrictive temperatures (top). Cells were plated on rich YES plates at increasing dilution. Cell viability was analysed by Phloxine B stain (bottom). (c) Average spindle pole body separation versus time (Pcp1-GFP) for wild type (green curve; n ¼ 5 time series), pkl1D single mutant (red curve; n ¼ 7 time series) and pkl1D cut7D double mutant cells (purple curve; n ¼ 9 time series) following hydroxyurea synchronization. (d) Timelapse fluorescence microscopy of prophase spindle pole body separation in wild type and pkl1D cut7D cells using spindle pole marker Pcp1-GFP. Timelapse fluorescence microscopy of mitosis from metaphase is shown for wild type in (e), pkl1D in (f) and pkl1D cut7D in (g). GFP-Atb2 marks microtubules (green) and DNA is stained with Hoechst (blue). Scale bar, 5 mm.    association with g-TuRC (Fig. 3a). The V5-Cut7ST with a fused nuclear localization single (NLS) to allow nuclear localization 29 is sufficient to target spindle poles in mitosis (Fig. 3d). This indicates that, similar to Pkl1, Cut7 contains distinct Motor and Tail domain elements that offer independent binding sites to g-TuRC.
The eukaryotic BimC domain is highly conserved across kinesin-5 members 30,31 . The domain was first identified in the Aspergillus nidulans kinesin-5 BimC protein 32 , but its precise role in mitosis has remained unknown for two decades. To examine the consequences of a mutation to the BimC domain of Cut7 on spindle pole localization, we used the loss of function temperature sensitive allele cut7-22 that contains a single-point mutation within this region (1,021 Pro to Ser) 22 . To determine whether the BimC box is the primary spindle pole targeting site in the Cut7 Tail domain, we utilized V5-tagged Cut7 deletion and/or sitedirected mutagenesis constructs with a fused NLS (Fig. 3e) 23 . These constructs encode the fusion proteins V5-NLS-Cut7ST, V5-NLS-Cut7ST 22 (point mutation at 1,021 Pro to Ser resulting in lost bipolarity) 22 , V5-NLS-Cut7T (Cut7-Tail, aa 888-1,085) and V5-NLS-Cut7T 22 and were analysed by immunocytochemistry of pkl1D cut7D cells that were fixed after shifting to 36°C. V5-NLS-Cut7ST and V5-NLS-Cut7T lack the Motor domain and retain spindle pole localization. In contrast, mutated V5-NLS-Cut7ST 22 and V5-NLS-Cut7T 22 constructs are unable to localize to poles but are expressed and retained within the nucleus, suggesting that the BimC sequence constitutes a key domain in the Cut7/g-TuRC interaction. Together, our data support the model in which Cut7 interacts physically with the g-TuRC MTOC in a manner poised to allow opposition to Pkl1 activity.
Pkl1 regulates spindle morphology from c-TuRC. Mitotic phenotypes in strains carrying single pkl1D or pkl1D cut7D double mutants versus wild type were evaluated by live cell and timelapse imaging of microtubules (a-tubulin as mCherry-Atb2 or GFP-Atb2), spindle poles (g-TuRC pericentrin protein 1, Pcp1-GFP), antiparallel microtubules and chromatin (anaphase B and chromatin binding kinesin-6 member Klp9-GFP) as well as Hoechst staining of DNA. An increase in spindle thickness is observed in the pkl1D cut7D double mutant strain versus wild type as seen in single-strain imaging ( Fig. 4a) or by live cell imaging of a mixed culture with wild-type cells (Fig. 4b). This increased thickness is also present in the pkl1D strain (Fig. 4c). In Fig. 4b, the ratio of double mutant to wild-type cells at 2:1 resulted in an accordant increase in the ratio of thick:thin spindles. Thick spindles (40.5 mm in a spindle of length 4 to 6 mm) are observed in 4±2% of wild-type cells, 63±8% pkl1D single mutant and 68 ± 8% of pkl1D cut7D double mutant cells (mean±s.e.m., n ¼ 90 cells for each strain). Although the pkl1D cut7D double mutant background generates a slight increase in the thick spindle phenotype (Po0.05 by Student's t-test), the change versus pkl1D single mutant alone is small. This indicates that it is primarily the loss of Pkl1 that induces this phenotype, which is exacerbated by additional loss of Cut7. The nature of the morphological change to spindle thickness could be the result of an increase in the number of antiparallel microtubules from both poles, parallel microtubules that emanate from a single pole, or due to unattached microtubules or disorganized arrays 20 at a single pole. To distinguish among these possibilities, we used multiple approaches. kinesin-6 Klp9-GFP crosslinks antiparallel microtubule arrays 28 and can be used to preferentially mark the extent of microtubule overlap (generally midzone), and is used with a-tubulin (mCherry-Atb2) to visualize spindle microtubules and length of the mitotic spindle. In wild-type cells, the Klp9-GFP signal spans the entire spindle midzone width (Fig. 4d, left images), whereas in the double mutant ( Fig. 4d, right images) we observe microtubule staining adjacent to the zone of antiparallel overlap that appears to emanate primarily from a single pole. We do not detect (d) Live cell fluorescence microscopy of wild type and pkl1D cut7D cells with mCherry-Atb2 and Klp9-GFP suggests that the increased spindle thickness we observe is due to parallel microtubules that emanate from a single pole (highlighted by cartoon schematics). Klp9-GFP marks antiparallel microtubules at the spindle midzone. In the schematic, yellow marks Klp9-GFP/antiparallel microtubule overlap, red marks parallel microtubules that extend from either pole (white circle), and green is Klp9-GFP signal on chromatin. Images were oriented similarly for convenience. Similar results to pkl1D cut7D cells were observed for the pkl1D single mutant. (e) Plating of wild type, pkl1D and pkl1D cut7D (top to bottom) cells on medium containing three concentrations of the microtubule-depolymerizing drug TBZ (0 mg ml À 1 left, 10 mg ml À 1 middle, 20 mg ml À 1 right). Scale bar, 5 mm.
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms6339 ARTICLE increased resistance to the microtubule-depolymerizing drug Thiabendazole (TBZ) in pkl1D or pkl1D cut7D strains versus wild type ( Fig. 4e) consistent with no or limited changes to microtubule number. We favour the interpretation that increased spindle width is likely due to asymmetric spindle pole effects (lost organization) in the absence of Pkl1 as seen 20 and to an increased number of spindle microtubules. Morphological changes in spindle thickness do not affect mitotic progression in the pkl1D cut7D double mutant cells through anaphase as seen by timelapse imaging and kymographic analysis versus wild type (Fig. 5). However, following anaphase B, timely spindle breakdown is delayed in a significant percentage of pkl1D cut7D cells. In 78% of double mutant cells (n ¼ 15 timelapse series) spindles remain intact significantly longer than wild type (n ¼ 7 timelapse series) for an additional 24 ± 7 min, and can persist beyond formation of equatorial MTOC arrays (Fig. 5a-c). Three patterns of spindle microtubule density in pkl1D cut7D anaphase cells were observed (Fig. 5a-c,e) and are referred to as Type 1, Type 2 and Type 3. In Type 1, central microtubule antiparallel overlap is identical to wild type. In Type 2, microtubule density is highly biased to one pole, and, in Type 3, central microtubules are diminished compared with thicker pole-biased microtubules. The relative frequencies of these patterns averaged over three time points after hydroxyurea synchronization (120, 140 and 160 min) are indicated in a stacked histogram (Fig. 5f). Our findings indicate that changes occur to spindle width and organization in both the pkl1D cut7D double mutant and pkl1D single mutant strains. Thickness along the spindle length is distributed in three patterns, two that are distinct from wild type. Spindle morphology phenotypes in the pkl1D cut7D double mutant are only modestly altered versus the single pkl1D mutant strain. These findings indicate that kinesin-14 Pkl1 is the primary kinesin regulating microtubule organization at g-TuRC, a loss that results in broader spindles with asymmetric microtubule density along the spindle length.
Daughter pole disorganization persists in the double mutant. An asymmetric effect on SPB organization with loss of the typical plaque-like appearance at one pole has been observed by TEM analysis of the pkl1D strain 20 . Similarly in the pkl1D cut7D double mutant we observe an asymmetric effect on spindle poles, including altered astral microtubule arrays as previously shown 33 .
Here we additionally identify the daughter pole as being primarily affected (Fig. 6). Asymmetry in astral microtubule lengths from opposing poles is observed and orientation is parallel to the spindle axis beginning in early mitosis. Asymmetric astral arrays are observed in 35±6% of pkl1D cells and 33±6% of pkl1D cut7D cells ( Fig. 6b; mean ± s.e.m., n ¼ 200 cells per strain). A small percentage of cells in both strains (8±3% and 6±2%, respectively) exhibits parallel arrays that are symmetric, and the remainder of cells examined have normal appearing astral arrays.  Cells without astral microtubules were also observed, but excluded from this analysis. We do not observe protrusions in the nuclear envelope as observed by co-imaging with nuclear envelope and SPB markers shown in Fig. 6c (n ¼ 0/57 cells), suggesting that these arrays are cytoplasmic. The septationinitiation network (SIN) protein Cdc7 loads primarily to the daughter pole in mitosis 34,35 . We found that the longer abnormal astral microtubules extend from poles that harbour Cdc7-GFP (68 ± 8% of cells with a signal, n ¼ 29 cells) and that this is also the pole with increased thickness (Fig. 6a,d). These data extend the asymmetric SPB disorganization phenotype observed in pkl1D cells 20 , identifying a role for Pkl1 in maintaining daughter pole organization from g-TuRC that indirectly alters astral microtubule arrays in the cytoplasm. Further, the data are consistent with daughter pole disorganization contributing to the thicker spindle phenotypes we observe in both pkl1D and pkl1D cut7D mutant cells.
We monitored chromosome segregation in the double mutant by live cell timelapse fluorescence microscopy and immunocytology of microtubules and DNA, as well as mini chromosome loss (Fig. 7). Three types of chromosome segregation defects are present in the double mutant that are unequal segregation, lagging chromosomes and lost chromosomes (Fig. 7a). Compared with wild-type cells, the pkl1D cut7D double mutant strain has increased lagging chromosomes and unequally segregated chromosomes, as well as a minor increase in lost chromosomes 36 . However, we along with others 20 observe that the efficiency of chromosome segregation is already markedly reduced in the absence of kinesin-14 Pkl1. Compared with the pkl1D cut7D double mutant strain, no lost chromosomes are observed in pkl1D and unequal segregation is reduced, though lagging chromosomes are prominent. Figure 7b is a histogram representation of the relative frequencies of these phenotypes in wild type, pkl1D single mutant and pkl1D cut7D double mutant strains. Shown are frequencies (n ¼ 500 cells) for unequal segregation (Type 1; wild type: 0%, pkl1D: 7 ± 3% and pkl1D cut7D: 9 ± 3%, mean ± s.e.m.), lagging chromosomes (Type 2; 3 ± 2, 23 ± 5 and 30 ± 5%) and chromosome loss from the spindle (Type 3; 0, 0 and 1%). We further quantified chromosome missegregation by monitoring the loss rate of a mini chromosome (cen1-7L) sup3E in the three strains (Fig. 7c). Mini chromosome loss in the pkl1D cut7D double mutant is greater than that in wild type (27% increase) but reduced by 5% with respect to pkl1D (32%). Our findings on chromosome segregation with pkl1D are consistent with previous studies 20 and reveal both rescue (lagging chromosomes) and exacerbation (unequal segregation) of the pkl1D phenotype in the double mutant along with the additional lost chromosome phenotype.
In wild-type cells, Mad2 monitors proper bipolar attachment of spindle microtubules to the kinetochore (KT), then transitions from KTs to both SPBs coincident with anaphase A onset. In anaphase B, Mad2 becomes asymmetric and makes a subsequent transition from the daughter pole to the equatorial MTOC, but remains asymmetrically localized at the mother pole 26 . We observe Mad2-GFP associated with the lost chromosome that is attached by an intranuclear microtubule (Fig. 7d), similar to what was observed previously with the lost chromosome phenotype 36 . This lost chromosome event occurs in 1% of pkl1D cut7D cells (n ¼ 3/300 cells) and is not observed in pkl1D cells. Mad2-GFP was also useful in confirming that the increased spindle microtubule thickness we observe in Type 2 morphologies (Fig. 5b,e) is associated with the daughter pole. By imaging cells with Mad2-GFP and mCherry-Atb2, the increased spindle ARTICLE microtubule thickness extends primarily from the daughter spindle pole that is unmarked or dimly marked by Mad2-GFP in early anaphase B (Fig. 7e), consistent with our Cdc7-GFP data (Fig. 6). Our findings indicate that the chromosome segregation defects in the double mutant strain generally resemble those of pkl1D except for the additional presence of the rare lost chromosome phenotype.
An asymmetric block on c-TuRC nucleation competency by Pkl1. We previously demonstrated that kinesin-14 Pkl1 physically interacts with g-TuRC in prophase to downregulate its function and oppose bipolar spindle assembly 17,22 . We established an in vivo assay system to examine the impact of elevated levels of Pkl1 in double mutant cells where kinesin-5 Cut7 is absent (Fig. 8a). Multi copy pkl1 was expressed using the low strength thiamine repressible nmt promoter (pREP81/pkl1). The pkl1D cut7D cells containing pREP81/pkl1 were grown with or without 5 mg ml À 1 thiamine and were fixed after 17 h. Cells were stained for a-tubulin and DNA using TAT1 antibody and Hoechst, respectively (Fig. 8a). After thiamine release we observed that 89 ± 9% of cells were arrested with unformed spindles. This frequency was only 6±2% when cells were grown in the presence of thiamine (Fig. 8c, left plot; mean ± s.e.m., n ¼ 200 cells for both conditions). In cells with unformed spindles, tubulin signal was concentrated near the cell centre. We were unable to obtain pkl1D cut7D cells transformed with pkl1 expressed under a higher strength promoter (pREP90x/pkl1), even under promoter repressing conditions in the presence of thiamine. We additionally performed live cell fluorescence microscopy with GFP-Atb2 (a-tubulin) in the cut7-22 temperature sensitive strain with or without the native pkl1 gene (Fig. 8b). Cells were synchronized by hydroxyurea for 4 h, released at the 36°C restrictive temperature and imaged 4 h after release. We observed similar unformed spindles in 82±9% of cells containing Pkl1. The pkl1D cut7-22 cells grown similarly exhibit robust spindle assembly (Fig. 8c, right plot; n ¼ 200 cells for both conditions), and these data are consistent with the serial growth assays performed with these two strains at 36°C (Fig. 1b). We envision three possible models in which spindle assembly would fail via a spindle pole based mechanism that are based on tubulin seeding at g-TuRC (Fig. 8d). This includes no addition of tubulin to g-TuRC, symmetrical addition but without microtubule elongation, or asymmetric addition of tubulin at a single spindle pole. Our data indicate the presence of short microtubules, ruling out model 1. To distinguish between models 2 and 3, we generated the strain [pkl1 þ cut7-22 pcp1-GFP mCherry-atb2] to simultaneously visualize poles and microtubules (compare Fig. 8b with Fig. 8e). Our data support model 3. We observe that in 100% of arrested cells where two distinct poles could be identified, the tubulin signal was associated with a single pole (n ¼ 25 cells). In vivo nucleation shown by cold depolymerization and reformation of microtubules in Fig. 8f demonstrates that, in pkl1D cut7D cells, g-TuRC can nucleate both interphase and spindle microtubules, consistent with our other data. Together, the in vivo data here are consistent with our model in which a required role for Cut7 in spindle assembly is to bind g-TuRC to oppose Pkl1 activity. Therefore, of these Klps, kinesin-14 Pkl1 in fission yeast is the primary negative regulator of g-TuRC microtubule nucleation.
Kinesin-14 PcTR peptide arrests human breast cancer cells. All human g-TuSC proteins are functional in fission yeast 14,15 . We previously developed biochemical tools in the form of kinesin-14 Tail peptides (Fig. 2e) 17 that regulate g-TuRC in vitro and here tested the conserved capability of these peptides to block microtubule nucleation in human MCF-7 breast cancer cells (Fig. 9). By immunocytology in fixed MCF-7 cells we demonstrate that 6-His tagged PgT targets the centrosome (Fig. 9a) and colocalizes with g-tubulin (Fig. 9b). By microtubule nucleation assays with or without g-TuRC, we also investigated the effect of kinesin-14 peptides on human g-TuRC nucleation competency in vitro (Fig. 9c). The control targeting peptide PgT is insufficient to block g-TuRC microtubule nucleation alone. However, PgTR that contains an additional g-tubulin binding and regulatory sequence 17 is a potent inhibitor of nucleation. Peptides PgT and PgR (separated TR elements) isolate human g-TuRC components in vitro similar to their action in fission yeast (Figs 2e and 9d). That is, PgT co-immunoprecipitates human GCP2, the yeast Alp4 counterpart, in addition to g-tubulin, while PgR binds and removes g-tubulin from the complex 17 . By live cell transfection of human MCF-7 breast cancer cells that exhibit low aggressiveness or MDA-MB-231 cells that are highly aggressive, we show that PgTR is a potent mitotic spindle protein (MSP) class regulator of mitotic arrest (Fig. 9e-g). Cell lines were transfected with 1 mg of 6-His tagged PgTR in 2 ml (108 mM) using  Fig. 9f). In arrested cells, residual clumped microtubules are present but do not appear to extend from centrosomes, suggesting that PgTR has an inhibitory effect on g-TuRC microtubule nucleation in vivo in human cells. These findings suggest that kinesin-14 action at g-TuRC is conserved from the yeast SPB to the human centrosome. We expect that PgTR will be a powerful mechanistic tool to elucidate g-TuRC function for microtubule nucleation in multiple model organisms and a potential therapeutic tool for preventing cell proliferation in disease.

Discussion
Microtubule organizing centres play major roles in specialized eukaryotic processes of broad interest such as spindle assembly, neuronal function and immunological synapse formation that involves cell polarization. Understanding the detailed mechanisms for microtubule nucleation requires combined knowledge of the underlying structure along with regulatory insights. In this work we demonstrate that the ability of fission yeast kinesin-14 Pkl1 to bind and alter g-TuRC structure and function 17 results in blocked microtubule nucleation in vivo generating failed spindle bipolarity and mitotic arrest. Conservation of this mechanism is revealed through use of a kinesin-14 Pkl1 peptide PgTR in human breast cancer cells that localizes to centrosomes and is sufficient to arrest nucleation in the two breast cancer cell lines investigated by preventing bipolar spindle formation. In fission yeast, kinesin-5 Cut7 but not kinesin-14 Pkl1, is an essential mitotic protein 18,27 . To better understand kinesin-14 Pkl1 function at g-TuRC, and kinesin-5 counteraction of this Klp, we applied genetic analysis, biochemistry and timelapse imaging. Here we show that kinesin-5 Cut7 is dispensable in the absence of kinesin-14 Pkl1 and that counteraction of Pkl1 by Cut7 requires Cut7 binding to g-TuRC through its Motor and BimC domains. These Klps are the first identified to directly bind and regulate g-TuRC, actions that are sufficient to impact microtubule nucleation capability. These findings are expected to have significant impact in the cytoskeleton field, particularly in understanding MTOC function as well as in potential therapeutic anti-cancer applications that utilize mitotic spindle protein agonists. Distinct mitotic phenotypes are present with loss of either Pkl1 or Cut7. The loss of pkl1 in the presence of cut7, although viable, results in an asymmetric effect on daughter spindle pole organization that influences spindle width and impairs chromosome segregation. Compared with the double mutant pkl1D cut7D, no amelioration or exacerbation of the pkl1D phenotype is observed, indicating that these phenotypes are likely due to loss of Pkl1. The additional loss of Cut7, however, does result in delayed spindle breakdown for mitotic exit. Cut7 localizes to the spindle midzone in anaphase 37 , and, although not required for anaphase B spindle elongation 29 , our data indicate it may contribute to normal progression through this stage. A primary role for kinesin-5 Cut7 is therefore to counteract kinesin-14 Pkl1 at g-TuRC. Only in the presence of Pkl1 does removal of Cut7 or inactivation of the Cut7 BimC domain (cut7-22) allow an asymmetric block on g-TuRC microtubule nucleation to be imposed that results in failed spindle bipolarity (Fig. 8).
Extensive studies demonstrate both the importance of kinesin-5 motors in spindle assembly along with kinesin-5-independent mechanisms. In the latter, force generation by other microtubule motors such as nuclear envelope-associated dynein and kinesin-12 operate and include microtubule pushing forces on the opposing pole and kinetochore-mediated microtubule interactions in prophase 38 . Our ability to remove kinesin-5 Cut7 in the absence of kinesin-14 Pkl1 reveals that in fission yeast kinesin-5independent mechanisms exist to establish spindle bipolarity.
Spindle phenotypes in the double mutant and single pkl1D strains do not include changes to timing for prophase SPB separation or mitotic progression through anaphase B versus wild type. The increase in spindle thickness on loss of Pkl1 that we observe is reminiscent of phenotypes induced by loss of kinesin-14 Kar3 in budding yeast 19 . The thick spindle morphology did not result in increased resistance to the microtubuledepolymerizing drug TBZ at 10 or 20 mg ml À 1 concentrations in single or double mutant backgrounds versus wild type, consistent with no change in spindle microtubule number. We favour the model that altered microtubule organization of parallel microtubules emanating from the daughter pole results in increased spindle width at this pole as opposed to an increase in spindle microtubule number. This is consistent with studies by ref. 20 in which TEM analysis of pkl1D cells revealed a decrease in pole organization characterized by loss of the typical plaque-like structure with apparent normal microtubule number. The replication of the SPB and centrosome is semi-conservative with the mother pole used as a template. To identify whether the mother or daughter pole is affected by loss of pkl1, we applied live cell fluorescence microscopy with asymmetric pole markers Cdc7-GFP and Mad2-GFP along with mCherry-Atb2 to mark microtubules. Our studies reveal that the daughter pole is affected in both pkl1D single mutant and pkl1D cut7D double mutant cells. Daughter pole disorganization additionally affects cytoplasmic astral microtubule arrays. Mitotic events can influence licensing and semi-conservative centrosome replication in the succeeding G1/S. In human cells, separase and polo kinase license centrosomes for duplication in the next cell cycle during mitosis 39 . Whether loss of kinesin-14 Pkl1 impacts subsequent cell cycle events outside of mitosis is not known. However, the changes to daughter spindle pole integrity without Pkl1 indicate a broader role beyond microtubule nucleation for spindle assembly, such as maturation or integrity of the daughter MTOC. We did not detect a similar role with kinesin-5 at the daughter MTOC and additional loss of Cut7 does not exacerbate these phenotypes. The concept of asymmetric events at spindle poles is well known. In budding yeast, g-tubulin mutants have been isolated that block robust microtubule nucleation from a single pole as seen by transmission electron microscopy 40 . In human cells, mother centriole stability is asymmetrically affected by kinesin-13 Kif24 (ref. 41). As well, regulation of poles can be asymmetric and is observed in mitotic checkpoint pathways 26,34,35,[42][43][44][45][46][47] that monitor spindle assembly, positioning and timing to help ensure the accurate segregation of chromosomes in cell division.
In this study, we demonstrate that the role of kinesin-14 at g-TuRC is to block microtubule nucleation and that key domains are required for this mechanism. This ability to localize to g-TuRC at spindle pole bodies is conserved with g-TuRC in the mammalian centrosome. We additionally identified the ability of kinesin-5 to bind g-TuRC as a key component in the Klp/g-TuRC regulatory mechanism in fission yeast. Pkl1 interacts with g-TuRC through two domains, a Motor domain that binds to g-tubulin helix 11 and distinct Tail domain binding to the complex. The combined domains provide strongest interactions with g-TuRC 17 . Our data indicate that similar to Pkl1, the Tail domain of Cut7 is the primary g-TuRC targeting element. We hypothesize that the Motor domain plays a role in assisted targeting to the g-TuRC site at spindle poles and in competition with Pkl1 binding to this site. Consistent with our previously published findings on Pkl1, we observe that the combined domains of Cut7 provide the strongest interactions with g-TuRC.
However, unlike Pkl1 that has low affinity to microtubules 48 , Cut7 retains the ability to bind strongly to spindle tubulins when g-tubulin specific binding is prevented. This alternative site of interaction may lower the pool of Cut7 at g-TuRC. Thus, as a consequence of blocked loading due to the helix 11 mutation and retained high microtubule binding affinity, we would expect reduced binding of Cut7 to g-TuRC complexes with g-tubulin-K5A present as compared with Cut7ST, as observed. Interestingly, the cut7-22 mutation lies within a MAP kinase phosphorylation consensus sequence in the conserved BimC sequence of the Cut7 Tail domain 29 , indicating that phosphorylation at this or other kinase sites within this domain may be important in the g-TuRC mechanism. Finally the dual regulatory relationship of kinesin-14 and kinesin-5 at the g-TuRC in fission yeast, along with the ability of PgTR peptide to block nucleation and spindle bipolarity in breast cancer cells, is expected to be impactful in regard to cancer therapy 49 . We are interested in exploring development of this new class of mitotic spindle protein (MSP) reagents as an addition to combined cancer therapies, in particular with Ispinisib/Monastrol 50,51 that targets human kinesin-5 or in response to taxol-resistant cancers.
Our findings are of particular interest in regard to multiple clinically oriented studies 52-60 that demonstrate overexpression of g-tubulin and other centrosomal proteins is characteristic of tumorigenesis and human malignancies in multiple tissues. In this case supernumerary microtubule-nucleating centrosomes are often observed and result in abnormal multipolar mitoses, aneuploidy and ultimately cell death 58,59,61 . Similarly, overexpression of g-tubulin in malignant cells can also produce ectopic microtubule nucleation in the cytoplasm. This is thought to result from g-tubulin-centrosome decoupling as well as subcellular sorting changes to soluble cytoplasmic pools or insoluble centrosomal complexes 52,57,58,60 as well as insoluble cytoplasmic aggregates 53 . Interestingly, an increase in the percentage of soluble cytoplasmic g-tubulin is associated with cell lines of higher aggressiveness and poorer prognosis versus those of low or moderate aggressiveness 58 . Further, g-tubulin can be incorporated within the a-/b-tubulin lattice of cytoplasmic microtubules that may impact drug resistance 58 . The ability of the PgTR peptide to target complexed g-tubulin could allow a means to prevent ectopic microtubule nucleation, although currently this is untested. Regardless, the peptide can mitotically arrest both the MCF-7 low and MDA-MB-231 high aggressive cell lines with similar efficiency. The benefit of PgTR versus other anti-mitotic agents that solely target microtubules is yet to be determined but is of interest in particular for malignant cell lines that are difficult to arrest and that often develop resistance.

Methods
General yeast strains and growth conditions. Standard procedures for genetic manipulation of fission yeast are as described 62 (S. pombe strains in Table 1). Cultures grown in fully supplemented YES-rich medium or minimally supplemented medium are also as described 62 . For yeast transformations, we used the EZ-YEAST Transformation Kit (MP Biochemicals). In growth assays, cells were grown to logarithmic phase in 10 ml rich YES media at 27°C. Cells were counted by haemocytometer and equalized and spotted at an initial concentration of 2 Â 10 7 cells ml À 1 with followed by 1/10 serial dilutions. Plates at 30°C and 36°C were grown for 4-5 days (n ¼ 3 experiments). Plates at 25°C were grown for 7 days. For promoter induction using the pREP81 low strength 63 or pREP90x high strength 64 nmt plasmids, cells were maintained on plates containing 5 mg ml À 1 thiamine before inoculation in 10 ml selective media with (control) or without (test) 5 mg ml À 1 thiamine for 17 h. Plates used to assess viability contained 5 mg l À 1 Phloxine B (Sigma-Aldrich). Mini chromosome loss was measured as described 65 . Growth curves were obtained using haemocytometer.
Yeast strain constructions. Integration of the ura4 gene at the cut7 locus was done using a PCR-based gene-targeting approach with long tracts of flanking homology as previously described 66 (Epicentre MasterAmp Extra-Long PCR Kit). We used 500 bp homology upstream and downstream of the cut7 open reading frame and verified stable integrants by colony PCR (Fig. 1a). Plasmid Integration was done with pREP vectors using homologous recombination at the autonomous replication site (Mlu1, New England Biolabs). All genetic crosses were done on minimal sporulation media, followed by marker selection and colony PCR.
Synchronous yeast culture. Cultures were grown overnight in YES-rich or selective minimal media at 27°C using baffled flasks. Hydroxyurea (11 mM) was added to cultures in logarithmic phase and incubated for 4 h. Cells were then washed twice with 10 ml sterile water before release in fresh media. Depending on the experiment, cells were released at 27°C (permissive temperature), 32°C (microtubule repolymerization) or 36°C (restrictive temperature).
For live cell peptide transfection we used the Chariot system (Active Motif) according to the manufacturer's instructions. Cells were seeded into 35 mm tissue culture-treated dishes containing coverslips and grown in complete medium to B60% confluency. One microgram of kinesin-14 Tail peptide PgTR (GenScript) was diluted in 100 ml 1 Â PBS on ice. Six microlitres of 1/10 PBS-diluted Chariot reagent was further diluted into 100 ml sterile water on ice in a separate tube. Diluted peptide and diluted Chariot were combined and incubated at room temperature for 30 min to allow the Chariot-peptide complex to form. Following incubation, media was aspirated from cells, which were then washed once in 1 Â PBS. The entire 200 ml volume was added to cells in 400 ml serum-free media and gently rocked to ensure even delivery. Plates were returned to 37°C for 1 h. Next, DMEM complete growth medium was added to 2 ml without removing the peptide delivery solution (108 mM peptide in 2 ml). Cells were further incubated at 37°C overnight and fixed after 24 h.
Fluorescence microscopy and immunocytochemistry. Fluorescence microscopy was performed using Zeiss Observer.Z1 inverted microscope with 63X Plan-Apochromat 1.4 NA oil and Â 100 oil 1.45 PlanFLUAR DIC objectives. Data were obtained using a Hamamatsu ORCA ER CCD camera with Zeiss Axiovision Rel 4.8 acquisition software. We acquired 20-image 0.1 mm Z-stacks. Timelapse series were acquired every 30 s to 6 min, with a median interval of 2 min. The 10-image 0.1-micron Z-stacks were superimposed on each timelapse image in a series. With live cells, GFP-Atb2 was imaged at 50-60 ms exposure and mCherry-Atb2 was imaged at 500 ms exposure. Only GFP-Atb2 was used in timelapse. Average temperature in the imaging room was 23°C. Using À 20°C methanol fixation, we were able to preserve GFP signal for quantifying phenotypes. In immunocytochemistry, microtubules were stained with a primary TAT1 antibody (1:25) 67 , followed by secondary goat antimouse Alexa Fluor 488 IgG (1:50; Life Technologies cat. A-11001) and DNA was stained in 1 mg ml À 1 Hoechst. A monoclonal anti-V5 primary antibody conjugated to FITC was used for viewing V5-tagged Cut7 constructs in fixed cells (1:500; Life Technologies cat. R963-25). Cells were imaged immediately using the Zeiss Observer.ZI system. In Fig. 8e, zoomed images were made high contrast. Z-stacks were made into 2D projections using ImageJ. Cold depolymerization and repolymerization of in vivo microtubules was performed as previously described 68 . Wild type and pkl1D cut7D cells containing the integrated GFP-atb2 (a-tubulin) plasmid were fixed to preserve GFP signal and analysed using the Zeiss system.
Human cells were fixed on glass coverslips in À 20°C methanol for 10 min, washed with 1 Â PBS and permeabilized in 0.5% Triton X-100 for 20 min. Following further washes, cells were blocked for 30 min in 1% bovine serum albumin/knock out serum replacement (BSA/KOSR). For peptide localization to centrosomes, 1 mg ml À 1 of peptide was applied after blocking and before primary antibody application. Antibodies used were primary mouse anti-g-tubulin monoclonal (1:5,000; Sigma-Aldrich cat. T5326) or primary mouse anti-a-tubulin monoclonal DM1A (1:1,000; Santa Cruz Biotech cat. sc-32293), primary His-tag polyclonal antibody (1:1,000; Cell Signaling cat. 2365), secondary goat anti-mouse Alexa Fluor 488 IgG (1:1,000; Life Technologies cat. A-11001) and secondary goat anti-rabbit Texas Red (1:1,000; Life Technologies cat. T-6391). Antibodies were diluted in 1% BSA/KOSR antibody dilution buffer. After secondary antibody application and washes, DNA was stained with 1 mg ml À 1 Hoechst solution in 1 Â PBS for 10 min followed by three times final PBS washes and mounted on slides with ProLong Gold anti-fade (Life Technologies). The 40-image 0.1 mm Z-stacks were made into maximum intensity 2D projections using ImageJ.
In vitro microtubule nucleation assays. In vitro microtubule nucleation assays were performed in a total volume of 5 ml. That is, 3 ml for the sample and 2 ml of tubulin at a 1:5 ratio of Rhodamine:unlabelled tubulin (Cytoskeleton). Total tubulin concentration was 3.75 mg ml À 1 in 2.5 Â tubulin working buffer (2.5 Â BRB80: 200 mM PIPES, 2.5 mM MgCl 2 , 2.5 mM EGTA at pH 6.8 and 2.5 mM GTP). For whole-cell extract nucleation analysis with peptide, 2 ml of whole-cell extract was added with 1 ml of peptide at 300 nM for a final peptide concentration of 60 nM. This 3 ml combination was added first followed by tubulin working buffer. For samples with no peptide, the 5 ml final volume comprised 3 ml RIPA buffer and 2 ml of whole-cell extract. The 5 ml reaction was combined on ice, quickly spun and returned to ice before incubating in a 37°C water bath for 4 min. Sample incubation was staggered at 20 s intervals to allow for pipetting. After 4 min, 50 ml of 1% glutaraldehyde fixing solution was added and tubes were incubated at room temperature for 3 min. Samples were completed by addition of 1 ml 1 Â BRB80, inverting multiple times to mix. For analysis, 50 ml of this mixture per sample was sedimented by ultracentrifugation at 173,000 Â g through a 15% glycerol cushion onto glass coverslips and imaged by Rhodamine epifluorescence using the Zeiss system at Â 630. Images of multiple fields were collected and the average microtubule number per field was determined.
Statistical analysis. For statistical analysis of phenotypes, n values were chosen as number of cells per strain needed to ensure adequate power to detect significant outcomes. P-values were generated using Student's t-test and statistical significance was considered for Po0.05 as appropriate. All statistical data in this study are reported as mean ± s.d. or ± s.e.m., as indicated. For cell cycle arrest by 1 mg transfection of PgTR, 12 fields at Â 200 were counted. Arrested cells with positive peptide signal were taken as a percentage of the entire population. Cells that were negative for peptide signal did not arrest.