A mutation in the Arabidopsis γ-tubulin-containing complex causes helical growth and abnormal microtubule branching

Although microtubule polymers can assemble in vitro from purified tubulins under appropriate conditions, eukaryotic cells tightly control the formation of new microtubules (termed `nucleation') temporally and spatially. γ-Tubulin, a member of the tubulin family, and its associated proteins have dominant roles in microtubule nucleation and spindle assembly in evolutionally distant eukaryotic species (Wiese and Zheng, 2006). In fungal and animal cells, γ-tubulin is found in complexes with additional proteins, termed γ-tubulin complex proteins (GCPs). The γ-tubulin small complex (γ-TuSC) is a core nucleation unit composed of two molecules of γ-tubulin associated with one molecule each of GCP2 and GCP3. Animal cells also contain a larger complex called the γ-tubulin ring complex, which contains additional proteins, including GCP4, GCP5 and GCP6, and shows high nucleation activity in vitro, whereas different sets of proteins are associated with larger γ-tubulin complexes in yeasts (Wiese and Zheng, 2006).

The organization of non-centrosomal microtubule arrays in higher plants involves numerous nucleation sites dispersed on the cell cortex (Chan et al., 2003), on the nuclear surface (Erhardt et al., 2002), and possibly on spindles and other endomembranes (Shimamura et al., 2004). Although γ-tubulin is essential for the nucleation of plant microtubules (Murata et al., 2005; Binarová et al., 2006; Pastuglita et al., 2006), the nature of γ-tubulin-containing complexes and how they are recruited to individual nucleation sites are not known. In Arabidopsis, tandem affinity purification of γ-tubulin-associated proteins recovered GCP2 and GCP3, indicating that they are in the same complexes in vivo (Seltzer et al., 2007). Sequenced genomes of higher plants contain homologs of other accessory GCP proteins found in animal γ-tubulin ring complexes (e.g. Pastuglia and Bouchez, 2007), implying that the fundamental organization of plant nucleation complexes is similar to that in animal cells.

Cortical microtubules in interphase plant cells are predominantly nucleated from the γ-tubulin-containing sites on the lattices of previously established microtubules (Murata et al., 2005). A notable feature of the microtubule-dependent nucleation is that the branching angle between the mother microtubule and the newly formed daughter microtubule is well defined with an average of around 40°. Assembly of nascent microtubules at acute angles to the pre-existing microtubules was also observed in algal cells after depolymerization of cortical microtubules by oryzalin (Wasteneys and Williamson, 1989). The minus-end of daughter microtubules is eventually released from the nucleation site (Shaw et al., 2003; Murata et al., 2005), perhaps by the severing activities of katanin (Burk and Ye, 2002). Free microtubules then migrate on the cell cortex by a hybrid treadmilling mechanism (Shaw et al., 2003), interact with each other with outcomes of selective stabilization or depolymerization, and finally generate a particular array pattern (Dixit and Cyr, 2004). In rapidly elongating cells, such as the epidermal cells in the growth zones of the root or the etiolated hypocotyl, for example, single and bundled microtubules align approximately transversely to the cell's main axis of growth. Cortical microtubules, particularly the bundled microtubules, might guide the movement of cellulose synthase complexes through the plasma membrane, thereby playing crucial roles in determining the ordered deposition of load-bearing cellulose microfibrils and anisotropic cell expansion (Paradez et al., 2006; Lucas and Shaw, 2008).

It is largely unknown how the microtubules are organized into specific array patterns (Ehrhardt and Shaw, 2006). Genetic screening recovered dozens of Arabidopsis twisting mutants that possess either right- or left-handed helical arrays of cortical microtubules in rapidly elongating cells (Ishida et al., 2007b; Perrin et al., 2007; Korolev et al., 2007). These twisting mutants are caused by mutations or overexpression of α- or β-tubulin, microtubule-associated proteins, and a mitogen-activated protein kinase phosphatase-like protein. Although exactly how the otherwise transverse microtubule array is transformed into helical microtubule arrays in mutant cells has yet to be clarified, microtubule stability and microtubule dynamics at either plus- or minus-ends are proposed to be important for the proper organization of cortical arrays (Ishida et al., 2007a; Yao et al., 2008).

In this paper, we report that a partial loss-of-function mutation of Arabidopsis GCP2 generates a left-handed helical array of cortical microtubules and right-handed helical growth. Interestingly, the branching angle of microtubule nucleation is less tightly controlled in the mutant, providing evidence of a role for GCP2 in the positioning of γ-tubulin-containing nucleating complexes on the wall of pre-existing microtubules.

In our continued screening for twisting mutants of Arabidopsis thaliana, we isolated a novel right-handed helical growth mutant, designated spiral3 (spr3). This mutant was generated in the Wassilewskija ecotype, and is caused by a recessive single-locus mutation (data not shown). The spr3 plants grew normally and were fertile, but several microtubule-dependent morphological characteristics were altered compared with wild-type plants (Fig. 1). When Arabidopsis seedlings were grown on a hard agar plate, wild-type roots grew downward in the direction of the gravity vector, whereas spr3 roots sharply skewed to the right side of the plate when viewed in front of the plate. A pair of wild-type cotyledons extended in a straight line in opposite directions, whereas spr3 cotyledons twisted in a counterclockwise direction when viewed from above. Similar counterclockwise twisting was observed in spr3 petals but not in wild-type petals. At the cell level, epidermal cell files of etiolated hypocotyls and roots were straight along the main axis of the organs in the wild type, but formed right-handed helices in the spr3 mutant.

In addition to having twisting elongating axial organs, spr3 plants were impaired in cell morphogenesis of trichomes and pavement cells. Trichomes on the 14-day-old wild-type true leaves (n=389) had mainly three branches (63.5%) or two branches (35.5%). By contrast, trichomes on spr3 leaves (n=419) mostly contained two branches (95.9%); three branched trichomes were scarce (0.4%) and needle-like trichomes with no branching point were also observed (3.6%). Reduced trichome branching has been reported in tubulin mutants (Abe et al., 2004) and in wild-type seedlings treated with microtubule-disrupting drugs (Mathur and Chua, 2000).

Arabidopsis leaf pavement cells develop from slightly elongated polygons (stage I) to stage II cells with multiple shallow lobes alternating with indentations or necks (Fu et al., 2002). Wild-type leaf pavement cells of late stage II show a characteristic jigsaw puzzle appearance with interdigitating lobes and indentations, whereas lobe outgrowth was reduced in the spr3 mutant (see Fig. 1G). The complexity of cell shapes was estimated by using circularity values (Le et al., 2006), where 1.0 (4π×area/perimeter²) indicates a perfect circle and an increasingly elongated polygon provides a smaller value. The circularity values for the wild-type pavement cells (n=35) and the spr3 cells (n=41) were 0.16±0.03 and 0.38±0.07, respectively, indicating that the spr3 pavement cells are less interdigitated than the wild-type cells (P<0.01; Student's t-test).

Since defects in mutant cell morphogenesis indicate a link between microtubule organization and SPR3 function, we analyzed the organization of cortical microtubule array. When root epidermal cells were analyzed by immunohistochemistry with an anti-tubulin antibody, it was found that the cortical microtubule arrays in the wild-type roots were mostly aligned transverse to the long axis of the cells, whereas those in the spr3 roots were arranged in left-handed helices (Fig. 2A). Quantitative analysis of individual microtubule orientations showed that the spr3 microtubules were skewed by ∼11° on average in the direction of the left-handed helix relative to the wild-type distribution (Fig. 2B).

The microtubule organization in pavement cells was visualized in a microtubule marker line expressing GFP-β-tubulin-6 [GFP-TUB6 (Nakamura et al., 2004)]. In the late stage II control cells in which a jigsaw puzzle appearance was established, transversely ordered cortical microtubules were restricted to the neck region (Fu et al., 2002). By contrast, abundant and thick transverse cortical microtubules were found throughout the spr3 cells (Fig. 2C).

We cloned the SPR3 gene by a map-based cloning approach (supplementary material Fig. S1). The SPR3 locus was mapped to a 30-kb interval on chromosome 5, in which seven genes were predicted to exist. After sequencing these candidate genes, we found a Gly-to-Arg substitution in the tenth exon of At5g17410, which encodes a subunit of the γ-tubulin-complex, GCP2 (Fig. 3A). The mutation changes the invariant Gly305 residue to Arg in the conserved Grip motif 1 (Fig. 3B). When an 8.7-kb region of the wild-type GCP2 gene was introduced into the spr3 mutant, all the anisotropic growth phenotypes, including the twisting growth, were complemented (Fig. 3C). Therefore, we conclude that the Gly305-to-Arg mutation in GCP2 causes the spr3 phenotypes.

To examine the biochemical consequences of the spr3 mutation, we assayed the interaction of GCP2 with GCP3, another subunit of the γ-tubulin-containing core complex, with a yeast two-hybrid system (Fig. 3D). When wild-type GCP2 protein fused to the Gal4 DNA-binding domain and a GCP3 protein fused to the Gal4 transcription-activating domain were expressed in the nucleus, they enabled growth of yeast cells on the selective medium and reconstituted the β-galactosidase activity. By contrast, a combination of the spr3-type G305R GCP2 mutant and wild-type GCP3 gave poor growth on the selective medium and highly reduced activity of β-galactosidase. Thus, the G305R mutation in GCP2 considerably compromises its interaction with GCP3 in yeast.

Since abnormalities in microtubule nucleation were suspected in plants with the spr3 allele of the GCP2 mutant, we compared the nucleation frequency of cortical microtubules in wild-type and spr3 plants, in which microtubules were visualized using the GFP-TUB6 marker (Fig. 4A). The GFP-TUB6-expressing Arabidopsis plants do not show any morphological abnormalities (Abe and Hashimoto, 2005). The frequency of nucleation was evaluated in two independent experiments (Fig. 4B). In the first experiment, microtubule nucleation events were monitored in arbitrary regions (400 μm²) of hypocotyl epidermal cells for 4 minutes. The microtubule nucleation frequency (approximately 1.3 events per minute per 400 μm²) did not differ significantly between the control cells and the spr3 cells. In the second experiment, nucleation events were counted along the length of preexisting cortical microtubules in hypocotyl epidermal cells, because cortical microtubule density might influence nucleation frequency. The calculated frequency values (approximately 0.002 events per minute per μm of microtubule) were indistinguishable between the control and spr3. Therefore, we conclude that microtubule nucleation efficiency was not measurably affected in the hypocotyl epidermal cells of spr3 plants under standard growth conditions.

Next, we extended our analysis to microtubule dynamics by using the GFP-TUB6 marker. Parameters of dynamic instability for cortical microtubules were obtained in hypocotyl epidermal cells of a control line (the marker line), spr3, and a complemented transgenic line (GCP2/spr3) in which the wild-type GCP2 gene was introduced into spr3. Table 1 shows rates of microtubule growth and shrinkage, frequencies of microtubule catastrophe and rescue, and percentages of time microtubules spent in periods of growth, pause or shrinkage. At the leading (plus) end of microtubules, these dynamic parameters were not significantly different between the control and spr3 mutant and between GCP2/spr3 and spr3, except that the growth rate in spr3 plants was reduced by 19-20% compared with that in the control or GCP2/spr3. The biological consequences of this small but statistically significant difference are not clear. At the lagging (minus) end, the rescue frequency of spr3 microtubules (0.119±0.090 events/second) was moderately but significantly (P<0.01; t-test) increased compared with the control value (0.058±0.050 events/second) and the value for GCP2/spr3 (0.077±0.054 events/second), whereas microtubule minus-ends of spr3 tended to spend more time in pause and less time in shrinkage, compared with minus-ends of control and GCP2/spr3. Other parameters at the minus-end were indistinguishable among the three types of cells. In summary, microtubule dynamics was not markedly affected by the spr3 mutation, although depolymerizing minus ends of spr3 microtubules stopped shrinking somewhat more frequently than those of control and reference microtubules.

After careful observation of microtubule nucleation events, we noticed that the tight control of microtubule nucleating angles was compromised in the spr3 cells. In GFP-TUB6-expressing control epidermal cells of hypocotyls and cotyledons, most cortical microtubules grew from the sides of other microtubules at a narrow range of angles, averaging approximately 40° to the wall of the existing microtubule (Fig. 4C). In the spr3 cells, microtubule-nucleating angles distributed more widely, as evidenced by the larger s.d. values, and the average nucleating angles increased by 7-10°, compared with those in the control. When the spr3 phenotypes of twisting morphology and a skewed microtubule organization were rescued by introducing the wild-type GCP2 gene, the distribution of microtubule-nucleating angles returned to the patterns observed for control cells, thereby confirming that the spr3 mutation is responsible for the relaxed and altered distribution of the microtubule-nucleating angles.

The above observations identified two abnormalities in microtubule behavior of the spr3 cells; slightly less dynamic minus-ends and less tight regulation of microtubule-nucleating angles. To assess the importance of dynamic minus-ends, we examined the morphological and cellular phenotypes of the katanin spr3 double mutant. The katanin allele used had a mutation in the 60 kDa ATPase catalytic subunit of the heterodimeric katanin (McNally and Vale, 1993), which severs microtubules and might be responsible for the release of nascent microtubules from the cortical nucleation sites in plant interphase cells (Burk et al., 2001; Wasteneys, 2002). In the katanin single mutant, all types of interphase cells, except for tip-growing cells, exhibited a dramatic reduction in length and an increase in width, as shown for the epidermal cells of etiolated hypocotyls and roots in Fig. 5A. The epidermal cells of katanin spr3 were short and swollen, and formed right-handed helical cell files; the additive cell morphology of katanin and spr3.

Immunostaining with an anti-tubulin antibody showed that, in katanin and the double mutant, cortical microtubule arrays of elongating root epidermal cells were organized aberrantly, and most microtubules were oriented in various angles with a wide deviation from the transverse direction (Fig. 5B,C). Many cortical microtubules converged at common sites, in an aster-like organization. A similar organization of cortical microtubules has been reported for various cell types with other katanin alleles (Bichet et al., 2001; Burk et al., 2001; Burk and Ye, 2002; Bouquin et al., 2003). It was noted that a minority of the epidermal cells with the katanin mutation had mostly aligned arrays, which were oriented in the transverse direction in katanin [as reported by Burk and Ye (Burk and Ye, 2002)] and shifted slightly in the left-handed helical direction in katanin spr3.

Time-lapse analysis of the microtubule nucleation events revealed that the minus-ends of cortical microtubules were not released from the nucleation sites on the wall of preexisting microtubules in the GFP-TUB6-expressing cells of katanin (Fig. 5D) and katanin spr3 (not shown), indicating that the microtubule-severing activity of katanin is necessary for the microtubule release. Distribution patterns of microtubule nucleating angles in the epidermal cells of hypocotyls and cotyledons were indistinguishable between the control (Fig. 4C) and katanin (Fig. 5E). By contrast, nucleating angles in the double mutant cells were distributed more broadly and shifted to elevated values (Fig. 5E), as observed in the spr3 single mutant cells (Fig. 4C). These results show that katanin is not involved in specifying the nucleating angles of nascent microtubules.

In conclusion, the observed mild defect in the minus-end microtubule dynamics is not essential to generate the helical growth in the spr3 plants.

Since spr3 does not appear to be a fully non-functional allele, we searched the publicly available T-DNA or transposon insertion databases for insertion alleles of GCP2. In the gcp2-1 allele (Wassilewskija ecotype), a single T-DNA was inserted after the first nucleotide G of the 17th exon and an accompanying 33-bp deletion removed after the insertion site (Fig. 3A). The heterozygous gcp2-1 plants were indistinguishable from wild-type plants, but we could not recover homozygous plants from the selfed progeny of the heterozygous plants. The siliques of self-pollinated gcp2-1 heterozygous plants contained desiccated ovules and aborted seeds (Fig. 6A), thereby reducing the number of fully developed seeds in a silique to approximately 40% of that in the self-pollinated wild-type plants (Fig. 6B). When an 8.7-kb genomic region of the wild-type GCP2 gene was introduced, we could obtain transgenic plant lines that were homozygous for the gcp2-1 allele, which grew normally and set seeds at the wild-type level (Fig. 6B). Therefore, we conclude that the null allele of GCP2 does not produce viable seeds, and that spr3 is a partial loss-of-function allele.

A high percentage of undeveloped ovules indicates that gametophyte functions might be impaired in gcp2-1. To analyze the transmission efficiency of female and male gcp2-1 gametes, we used heterozygous plants in reciprocal crosses with wild-type plants. When wild-type pollen was pollinated on heterozygous gcp2-1 pistils, only 3.4% of the progeny plants (n=143) carried the gcp2-1 mutant allele, instead of the 50% expected for full transmission. When pollen of gcp2-1 heterozygous plants was used for the cross to wild-type plants, the gcp2-1 mutant allele was recovered in 17% of the progeny (n=205). Therefore, the transmission of mutant gametes was reduced by 93% on the female side and by 66% on the male side, showing that the gcp2-1 mutation drastically affects the development and function of both male and female haploid gametophytes.

We thus examined mature ovules by confocal microscopy according to the procedure used to visualize the embryo sac (Fig. 6C) (Christensen et al., 1997). Less than 1% of ovules had abnormal gametophytes (n=127) in wild-type plants. By contrast, abnormal gametophytes with reduced numbers of nuclei were found in 38% of mature ovules (n=426) in pistils of gcp2-1/+ plants. Synergids nucleoli were frequently absent and a large vacant space was generally found in the micropylar region of the abnormal female gametophytes.

To directly compare the development of gcp2-1 and wild-type male gametophytes, we used the quartet (qrt) mutant, which produces four microspores that remain associated as a tetrad (Preuss et al., 1994). At the mature stage, more than 99% of the qrt tetrads contained four pollen grains, which possess a tricelled structure containing two sperm cells enclosed in a vegetative cell (Fig. 6D). In the qrt/qrt; gcp2-1/+ plants, however, 69% of the tetrads (n=146) contained one or two abnormal pollen grains, in which 22% of the tetrads had a bicellular pollen grain with one large sperm cell nucleus, 40% had one aborted microspore, and 7% had two aborted microspores (Fig. 6D,E). These results show that GCP2 is required for normal development of both male and female gametophytes.

As expected from the essential role of the γ-tubulin core complex in microtubule nucleation, Drosophila GCP2 mutants die early, during the first and second instars (Colombié et al., 2006). The deletion of GCP2 orthologs in budding and fission yeasts is also lethal (Knop et al., 1997; Vardy and Toda, 2000). In this study, we showed that a T-DNA insertion null allele of GCP2, gcp2-1, does not produce viable homozygous mutant seeds, and the formation and function of gcp2-1 gametophytes, especially embryo sac development, were drastically affected. These mutational studies in various organisms thus suggest that the conserved GCP2 subunit of the γ-tubulin core complex is essential for the survival of various organisms, including plants.

In Arabidopsis, the simultaneous disruption of two γ-tubulin genes generates aberrant microtubule structures and severely impairs cell division in gametophytes; >90% of the female gametophytes and >60% of the male gametophytes are non-functional (Pastuglia et al., 2006). These effects of γ-tubulin mutations on gametes are strikingly similar to those of gcp2-1, which reflects that these mutations target different components of the same complex. As discussed previously (Pastuglia et al., 2006), a pool of γ-tubulin (and GCP2) might be carried over from parental sporocytes into the gametophytes, and might sustain cell division until the cellular concentration reaches a critical level. This might explain the phenotypic variation seen in the developmental defects in mutant gametophytes (e.g. Fig. 6). The female gametophyte is more severely affected than the male gametophyte, possibly because development of the embryo sac involves three successive mitoses and a larger cellular volume.

The structure of the Saccharomyces cerevisiae γ-TuSC was recently analyzed by electron microscopy (resolution, 25 Å), γ-tubulin gold labeling and in vivo FRET (Kollman et al., 2008). In the proposed model, the complex is Y-shaped, with an elongated body (corresponding to the N-termini of Spc97p and Spc98p, the yeast homologs of GCP2 and GCP3) connected to two arms (corresponding to the C-termini of Spc97p and Spc98p, each bound to one molecule of γ-tubulin). The G305R spr3 mutation in the grip1 motif of GCP2 appears to be located near the neck region at the end of the dimerized body part, although precise placement is not possible at this level of resolution. Our yeast interaction assay also suggests that the G305R mutation compromises interaction between GCP2 and GCP3. That MT nucleation frequency is not significantly affected in the spr3 cells implies that the mutant GCP2 is incorporated into γ-TuSC and its interaction with GCP3 is perhaps partly stabilized in vivo by other interacting subunits in γ-TuSC or a larger complex.

We observed mild defects in the microtubule dynamics of cortical microtubules; a slight decrease in the growth rate at the plus-end, and a slight increase in rescue frequency at the minus-end. In a partial loss-of-function mutant of the fission yeast GCP2 homolog, the plus ends of interphase and spindle microtubules exhibited abnormal behavior, including cycles of growth and breakage, and hyperstability (Zimmerman and Chang, 2005). Abnormally long microtubules or alterations in microtubule dynamics have also been reported when γ-tubulin, or the homologs of GCP2 and other γ-tubulin complex subunits, are mutated or depleted in yeasts and Drosophila (Colombié et al., 2006) (reviewed by Raynaud-Messina and Merdes, 2007). It is quite possible that defective microtubule nucleation increases the cytoplasmic pool of free tubulins, which consequently causes altered microtubule dynamics and an abnormal microtubule structure. Alternatively, the function of the microtubule ends might be directly affected by the microtubule-nucleation complex. In vitro reconstitution experiments suggest that the γ-tubulin ring complex can act as a minus-end-capping protein, by binding directly to the minus-end and preventing microtubule depolymerization (Wiese and Zheng, 2000). A hypothetical model has been postulated, in which a small number of γ-tubulin complexes incorporate into the plus-end microtubule sheets during their closure into polymerizing tubes (Raynaud-Messina and Merdes, 2007).

A striking feature of the spr3 mutation is its effect on the distribution of microtubule-nucleating angles. In wild-type cells of Arabidopsis and tobacco, the angle between the mother microtubule and the newly formed daughter microtubule is well defined with an average value of approximately 40° (Murata et al., 2005) (and this study). The spr3 mutation of GCP2 relaxes the narrow distribution of the branching angle, and shifts the average to a slightly increased value. The branching points have been shown to contain γ-tubulin (Murata et al., 2005), and might also contain GCP2 in the form of γ-TuSC or a larger complex. The compromised distribution pattern of the branching angle in spr3 suggests that GCP2 has an important role in defining the precise physical interaction between the postulated γ-tubulin-containing nucleation complex and a pre-existing mother microtubule. Recent structural analysis revealed how the actin-related protein 2/3 (Arp2/3) complex docks onto a pre-existing mother actin filament to form a new filament at a characteristic ∼70° angle (Rouiller et al., 2008). In the actin branch, the Arp2 and Arp3 subunits form a short pitch helix dimer and together contribute the first two subunits of the daughter actin filament. Moreover, all seven subunits of the Arp2/3 complex make some contact with the mother filament. It will be a future challenge to elucidate how an activated γ-tubulin-containing nucleating complex recognizes the side of a mother microtubule in higher plant cells.

Previous analyses of Arabidopsis twisting mutants suggested a correlation between the generation of helical microtubule arrays and the dynamics of individual microtubules (Ishida et al., 2007a). The alteration of microtubule dynamics in the spr3 cells was not great but could be observed at both ends. However, the helical growth phenotype of the katanin spr3 double mutant, in which the minus-end of daughter microtubules remained attached to the side of mother microtubules, indicates that the observed minus-end dynamics is not essential for generating twisted growth. We here propose three models that might underlie the mutant array organization.

When a growing plus-end of cortical microtubule encounters a pre-existing microtubule, possible outcomes are influenced by the collision angle. Encounters at angles over ∼40° mostly result in crossover or depolymerization, whereas at angles below ∼40°, the formation of bundles, where the incoming microtubule is captured and grows in parallel to the encountered microtubule, is favored (Dixit and Cyr, 2004; Ehrhardt, 2008). The critical collision angle of ∼40° is intriguingly similar to the nucleation angle of the wild-type cells. The divergent nucleation and convergent capture of nascent microtubules are proposed to be important to maintain the transverse array of cortical microtubules (Murata et al., 2005). When the nucleating angles become more divergent and steeper, as in the spr3 cells, the intricate balance of collision consequences might be affected. We indeed observed that the microtubule-microtubule encounters in the spr3 cells result in an increased frequency of depolymerization at the plus-end of incoming microtubules (supplementary material Table S1). Therefore, abnormal microtubule nucleating angles may affect the establishment and maintenance of the transverse array organization.

Alternatively, the altered encounter outcomes as discussed above might result from the slight decrease in the growth rate at the plus end of spr3 microtubules, and might be independent on the encounter angles. For example, it was recently shown that Arabidopsis mutants lacking the CLASP microtubule-associated protein have increased encounter frequencies and are more prone to form bundles at greater encounter angles even though plus-end dynamics do not appear to be distinct from that in the wild type (Ambrose and Wasteneys, 2008).

Finally, γ-tubulin-containing nucleation complexes might be important for generating unskewed cortical microtubules. Microtubules assembled in vitro from pure tubulin dimers are composed of varying numbers of protofilaments (from 12 to 17), in which 14 protofilament microtubules dominate (Wade and Chrétien, 1990; Chrétien and Wade, 1991), whereas in vivo in most cell types, including those of plants, microtubules mostly consist of 13 protofilaments (e.g. Ishida et al., 2007a). Only in the 13-protofilament microtubules, do the protofilaments align strictly parallel to the axis of the microtubule. The protofilaments in the 14 protofilament microtubules, for example, are skewed and form left-handed helices along the microtubule axis. One important function of the γ-tubulin-containing complexes might be to provide a nucleation template to ensure the assembly of 13-protofilament microtubules. In the spr3 cells with impaired γ-TuSC, the otherwise tight control of the protofilament number might be compromised, resulting in a microtubule population containing some skewed microtubules. Such skewed microtubules might serve as a chirality determinant, which should persist even after individual microtubules are assembled into higher-order bundles (Ishida et al., 2007b). The development of efficient techniques to visualize the protofilament ultrastructure of plant microtubules in vivo (Bouchet-Marquis et al., 2007) is highly anticipated.

All the Arabidopsis thaliana plants used in this study were of the Wassilewskija ecotype. gcp2-1 was obtained from INRA (Versailles, France). spr3 and katanin mutants were isolated from seeds mutagenized with ethylmethane sulfonate (Lehle Seeds, Round Rock, TX). The katanin allele had a one-base insertion after the 54th base T of the fourth exon, which resulted in the creation of a stop codon at amino acid position 295. A GFP-TUB6-expressing line was generated in the Wassilewskija background, by using the same plasmid reported previously (Nakamura et al., 2004). Plants were grown as described (Furutani et al., 2000).

F2 progenies from a cross between spr3 (Wassilewskija ecotype) and the wild type (Columbia ecotype) were used as a mapping population. The SPR3 locus was mapped by using DNA markers that were available from The Arabidopsis Information Resource database (http://www.arabidopsis.org) or were newly developed (supplementary material Fig. S1). The two markers closest to the SPR3 locus (marker names based on the resident BAC clone, primer sequences, and restriction enzymes used) are T10B6 (5′-TGAACCGAGTCCATTTACAGA-3′, 5′-TGAGACCTACGCAACCTTGT-3′ and AluI) and K3M16 (5′-GCGAGTAGAAAGAAGATTTTGTTGATC-3′, 5′-CAACTAGGAAACTAAATGACATAACCCA-3′ and HinfI). Use of these markers confined the SPR3 locus to a 30-kb segment on chromosome 5. Seven open reading frames predicted in this region were amplified by PCR from the spr3 genome and sequenced.

The isolation and manipulation of DNA were performed by using standard molecular techniques. A SPR3 genomic region, which contained a 1058-bp region 5′ upstream of the initiation ATG, all exons and introns, and a 1908-bp region 3′ downstream of the stop codon, was subcloned from a BAC clone T10B6 into pDONR221 by PCR and BP reactions, and then transferred to a Gateway binary vector pGWB1 (Nakagawa et al., 2007) by an LR reaction. The resulting binary vector was introduced into Agrobacterium tumefaciens strain MP90 and used to transform spr3 and atgcp2-1/+ plants by the floral dip method (Clough and Bent, 1998). To confirm complement tests, glufosinate (BASTA) resistance in the atgcp2-1 mutant plants were used.

Microtubules of 4- or 5-day-old seedlings were immunostained, essentially as described by Abe and Hashimoto (Abe and Hashimoto, 2005). Anti-α-tubulin antibody YL1/2 (Abcam, Cambridge, UK) and Alexa Fluor 568-conjugated anti-rat IgG (Molecular Probes, Eugene, OR) were used as primary and secondary antibodies, respectively. Root epidermal cells were observed with a C1 ECLIPSE E600 confocal laser-scanning microscope (Nikon, Tokyo, Japan).

Female gametophyte development was analyzed according to the method of Christensen et al. (Christensen et al., 1997). The 488-nm argon laser of a Nikon C1 confocal microscope was used to illuminate female gametophytes. To analyze male gametophytes, anthers containing mature pollen were brushed on a microscope slide. After staining with DAPI (0.1% Nonidet P-40, 10% DMSO, 50 mM PIPES, pH 7.2, 5 mM EGTA, pH 7.5, and 5 μg/ml DAPI) (Schnurr et al., 2006), pollen grains were covered with a coverslip and observed by using a mercury lamp and a Nikon ECLIPSE E1000 microscope.

Four-day-old seedlings expressing GFP-TUB6 were used for the analysis of microtubule dynamics and nucleating angles in the epidermal cells of upper hypocotyls and in pavement cells of cotyledons. We used a DMRE microscope (Leica, Allendale, NJ) equipped with a CSU10 scanning head (Yokogawa, Tokyo, Japan), a 488-nm argon ion laser (Omnicrome, CA), and an ORCA-ERCCD camera (Hamamatsu Photonics, Shizuoka, Japan). Images were taken every 4 seconds during the course of 4 or 6 minutes at 23°C. Only microtubules for which both leading and lagging ends were visible were analyzed. Dynamic instability parameters were obtained as we described previously (Nakamura et al., 2004). Acquired images were processed and analyzed by using Scion Image (http://www.scioncorp.com/) and ImageJ version 1.40 (http://rsb.info.nih.gov/ij).