Insights into cortical microtubule nucleation and dynamics in Arabidopsis leaf cells

Immediately after the completion of cytokinesis, newly divided plant cells form cytoplasmic microtubules (MTs) that radiate from the nuclear envelope toward the plasma membrane, and generate dynamic MT arrays at the cell cortex, which persist during interphase (Hashimoto, 2015). The MTs in these cortical arrays guide the movement of cellulose synthase complexes along the plasma membrane, thus regulating the orientation of the deposited cellulose microfibrils and ultimately dictating the shape of plant cells (Paredez et al., 2006). Interphase MTs in plants are not anchored at MT-organizing centers such as centrosomes in animal cells, but are continuously reorganized through polymerization and depolymerization processes at both ends of the MTs, through interactions between neighboring MTs and through MT severing at crossover sites (Wasteneys and Ambrose, 2009; Hashimoto, 2015). Cortical MTs migrate within the cell cortex because of different rates of polymerization at their ends, with their dynamic plus-ends biased toward net growth, and their minus-ends toward slow depolymerization (Shaw et al., 2003). The dynamic behavior of individual MT ends, termed dynamic instability (Mitchison and Kirschner, 1984), is characterized by MT polymerization that is interrupted by a sudden and rapid depolymerization phase, followed by a ‘rescue’ process, after which polymerization resumes. The molecular mechanism underlying the rescue process is not fully understood (Gardner et al., 2012). Although various MT-associated proteins (MAPs) are known to regulate dynamic instability (Akhmanova and Steinmetz, 2015), the mechanisms by which plant MAPs remodel cortical MT arrays are unclear (Hamada, 2010).

The organization of MT arrays is also regulated by the location, timing and the manner of de novo MT formation (Hashimoto, 2013). Most plant cortical MTs are generated at the sides of preexisting MTs, either at a branching angle of ∼40° (Murata et al., 2005; Chan et al., 2009) or in a parallel fashion, with the daughter MT immediately bundling with the mother MT (Chan et al., 2009). In bundle-forming nucleation, the parallel configuration, in which the plus-ends of both the daughter and the mother MTs grow in the same direction, is favored over the antiparallel configuration (Chan et al., 2009). Because MT bundles in the Arabidopsis cortical array are predominantly antiparallel (Shaw and Lucas, 2011), it remains unclear how parallel MT bundles are transformed into antiparallel ones. MT-free nucleation, in which a solitary MT is polymerized in the cell cortex in the absence of existing MTs, occurs for a small percentage of nucleation events (Shaw et al., 2003; Nakamura et al., 2010).

Evolutionarily conserved MT nucleation complexes, which contain γ-tubulin ring complexes (γTuRCs), have been shown to be essential for MT nucleation in plants (Hashimoto, 2013). γTuRC is composed of 13 γ-tubulin molecules and five types of related γ-tubulin complex proteins (GCP2 to GCP6; Kollman et al., 2011). A mitotic-spindle-organizing protein associated with a ring of γ-tubulin 1 (MOZART1, or MZT1; Hutchins et al., 2010) binds to a preassembled γTuRC and facilitates its activation (Cota et al., 2017). Targeting factors subsequently direct the primed γTuRC to specific subcellular locations to initiate local MT nucleation. Eight-subunit augmin complexes are required for branch-forming MT nucleation in mitotic spindles (Kamasaki et al., 2013; Hotta et al., 2012) and in plant cortical arrays (Liu et al., 2014). Augmin is thought to bind to the sides of MTs and recruit γTuRC through the targeting adaptor NEDD1 (Zeng et al., 2009; Petry and Vale, 2015). Plant cells contain functional γTuRC, MZT1, augmin and NEDD1, but it is not known how these conserved nucleation factors regulate the free, branched or parallel forms of nucleation observed in plant cells (Sánchez-Huertas and Lüders, 2015).

In this work, we used a live-cell MT imaging system in which MTs and γTuRC were fluorescently labeled to investigate several aspects of MT dynamics in cortical MT arrays in Arabidopsis cells. Our results indicate that bundle-forming MTs are nucleated in an antiparallel orientation, non-nucleating γTuRCs contribute to MT stability, and cortical MTs are anchored to the plasma membrane at their plus-ends.

To visualize γTuRC and MTs simultaneously, we co-expressed MZT1a–GFP (hereafter MZT1–GFP) (Nakamura et al., 2012) and mCherry–β-tubulin (TUB) in stably transformed Arabidopsis transgenic lines. The MZT1–GFP construct has been shown to fully complement the embryonic lethal phenotypes of mzt1 double mutants (Nakamura et al., 2012), and an mCherry fusion of GCP2, an integral subunit marker of γTuRC (Nakamura et al., 2010), mostly colocalized with MZT1–GFP (93.1%, n=159 in 4 cells; Fig. S1). In cotyledon pavement cells, most of the MZT1–GFP-labeled γTuRCs appeared on the lattice of cortical MTs, and roughly half of the γTuRCs nucleated nascent MTs, with branch-forming nucleation occurring more frequently than bundle-forming nucleation (Movie 1). MT-free nucleation occurred in less than 2% of total nucleation events. These observations are consistent with previous reports (e.g. Shaw et al., 2003; Dixit and Cyr, 2004; Wightman and Turner, 2007; Chan et al., 2009; Nakamura et al., 2010).

We reinvestigated the polarity of daughter MTs in relation to that of the preexisting mother MTs during MT-dependent MT nucleation. The plus-ends of mother MTs were identified by their high dynamicity profiles. In the branch-forming nucleation (n=391 in 27 cells), the angle between newly formed daughter MTs and the mother MTs was ∼40° towards the plus-end of the mother MT (Fig. S2). In agreement with previous results (Chan et al., 2009), no bias for nucleation on the right or left side was observed (Fig. 1A; Movie 2). In 20% of bundle-forming nucleation events (n=258 in 32 cells), daughter MTs grew toward the plus-ends of mother MTs, forming parallel bundles (Fig. 1B; Movie 3), whereas the remaining 80% of bundle-forming nucleation events resulted in antiparallel bundles (Fig. 1C; Movie 4). Thus, our results indicate that antiparallel nucleation is highly favored over parallel nucleation.

However, Chan et al. (2009) reported that parallel nucleation is favored over antiparallel nucleation, with a factor of 4.4, based on an analysis in which the plus-ends of microtubules were labeled with GFP-tagged end-binding 1 protein. We believe that our analysis is more accurate because we unambiguously identified MT nucleation events by the simultaneous observation of γTuRC and MT dynamics, whereas Chan et al. (2009) relied on data interpretation to distinguish between de novo nucleation and rescue events. Our current finding is consistent with a report that found that the polarity of MT bundles in the hypocotyl epidermal cells of light-grown Arabidopsis seedlings is predominantly (71%) antiparallel (Shaw and Lucas, 2011).

The finding that bundle-forming MT nucleation is predominantly, but not only, antiparallel indicates that a mode of γTuRC recruitment exists that is distinct from the recruitment in branch-forming nucleation. Bundling nucleation may not require the same targeting factors as those used for branching nucleation. Interestingly, when NEDD1 was knocked down in Arabidopsis leaf pavement cells, branching nucleation events were significantly reduced, whereas bundle-forming nucleation events were minimally affected (Walia et al., 2014). Similarly, knockdown of the AUG6 augmin subunit gene in Arabidopsis leaf pavement cells affected branching nucleation more severely than bundle-forming nucleation (Liu et al., 2014). Thus, to promote bundle-forming MT nucleation, some γTuRC may be recruited to existing MTs independently of known targeting factors, such as augmin and NEDD1.

Time-lapse imaging of γTuRC particles in the Arabidopsis pavement cells revealed that a small fraction of the particles (∼5%) moved in one direction for several micrometers (105 motile particles observed out of 1960 γTuRC particles in 13 cells). An analysis of these motile particles (236 events in 24 cells) revealed that 28% were located on MT plus-ends, 37% on MT minus-ends and 35% on the sides of existing MTs (Table S1).

Of all the γTuRC particles at the MT plus-ends, 92.5% (62 events) localized to depolymerizing plus-ends, while the remaining 7.5% (5 events) localized to growing plus-ends. In one observation, a γTuRC particle localized very close to the growing plus-end of a cortical MT and was carried along by the polymerizing end for 4.6 µm until this MT end collided with another MT, after which the γTuRC particle dissociated from the MT (Fig. S3A; Movie 5). In another observation, a γTuRC particle that localized to the plus-end of a growing MT in a MT bundle was carried by the growing end for 2.4 µm, and then switched its localization to the MT lattice of a different companion MT in the bundle (Fig. S3B; Movie 6). These MT plus-end-localized particles did not affect the polymerization rate of the MT ends.

Recruitment of γTuRC particles to the depolymerizing plus-end was often associated with a rescue event. When a depolymerizing plus-end passed the lattice-associated γTuRC particle, the depolymerizing end dragged the particle for 1.2 µm, and then switched to a polymerization phase, leaving the particle at the rescue site (Fig. 2A; Movie 7). In another observation (Fig. 2B; Movie 8), a depolymerizing plus-end was transiently associated with one γTuRC particle, and later associated again with another brighter particle (possibly an oligomerized particle). The depolymerization rate continued to decrease while γTuRC particles accumulated, until the plus-end switched to the polymerization phase. We analyzed all 50 γTuRC-associated depolymerizing plus-ends to determine whether the depolymerization rate changed after particle association (Fig. 2C). Although depolymerization rates of individual MTs were variable, binding of γTuRC to the depolymerizing plus-end tended to reduce the rates. The differences between the depolymerization rates before (v0) and after (v1) MZT1–GFP recruitment exhibited a Gaussian distribution, with an average of −0.15 µm s⁻¹ (Fig. 2D). The null hypotheses of zero median (P=3.06×10⁻⁶<0.01) and zero mean (P=1.96×10⁻⁶<0.01) were rejected based on a Wilcoxon signed-rank test and a Student's t-test, respectively, thereby demonstrating the statistical significance of the velocity reduction. In astral MT arrays in Drosophila mitotic cells, γTuRC depletion affects MT stability, indicating a nucleation-independent role for γTuRC (Bouissou et al., 2014). Other examples in several phyla show that mutations or deficiencies in γ-tubulin or GCPs alter plus-end MT dynamics (Oakley et al., 2015). These results suggest that, in addition to its essential role in nucleation, γTuRC directly regulates MT dynamics.

When γTuRC is recruited to the lattice of preexisting MTs, most of these lattice-bound γTuRC particles dissociate from the MTs within 20 s, without nucleating daughter MTs (Nakamura et al., 2010). To examine whether these transiently stabilized and lattice-bound γTuRC particles modulate MT dynamics, we monitored MT dynamics near the γTuRC-bound sites. We found that depolymerization of a plus-end was in some cases inhibited, and regrowth was initiated precisely at the location of γTuRC (Fig. 3A; Movie 9). Next, we compared the locations of rescue sites (n=91) with those of the lattice-bound γTuRC particles, and found that the rescue frequency was ∼3- to 4-fold higher at γTuRC-bound sites (n=29) than at sites that lacked γTuRC (Fig. 3B). These results suggest that non-nucleating γTuRC particles that are transiently localized to the MT lattice contribute to MT rescue events.

Cytoplasmic linker-associated protein (CLASP) is a MAP that promotes MT rescue and suppresses MT catastrophe events (Al-Bassam and Chang, 2011). The molecular mechanisms regulating CLASP-mediated MT rescue are not fully understood, but the lattice-bound non-nucleating γTuRCs may function in a similar manner to modulate MT dynamics. Consistent with our observations, immunolocalization of γTuRC in combination with live imaging analysis of MT dynamics in Drosophila interphase cells showed that pause and rescue events frequently occur at the γTuRC attachment sites along interphase MTs (Bouissou et al., 2009).

γTuRCs are preferentially recruited to cortical MTs, but are sometimes localized to the MT-free cell cortex area (Nakamura et al., 2010). These singular γTuRCs are also capable of nucleating MTs. We found that some of these unbound γTuRC particles (seven observations) moved from their initial sites in the cell cortex after MT nucleation. In the sequence shown in Fig. 4 and Movie 10, a singular γTuRC particle nucleated a MT immediately after the γTuRC particle was recruited to the MT-free cell cortex, and remained at the initial location until the new MT reached 1.2 µm in length. The γTuRC particle was subsequently pushed backward at a velocity of 8.6×10⁻² µm s⁻¹ for 24 s, while plus-end growth apparently ceased. The γTuRC particle stopped moving when it collided with the lattice of another MT, and plus-end polymerization resumed at the same time. At 68 s after its initial transient stabilization at the cortex, this γTuRC particle dissociated from the minus-end of the new MT, which subsequently slowly depolymerized. When we re-plotted the kymograph of this nucleation event in such a manner that the γTuRC particle stayed at a fixed location, the polymerization at the MT plus-end was at a constant velocity (Fig. 4B). We interpret this dynamic behavior of γTuRC as shown in the model in Fig. 4C. The γTuRC particle that was not associated with a cortical MT was weakly associated with the plasma membrane. After a new MT was nucleated from such a γTuRC particle, the polymerizing plus-end was captured by a protein complex, which was tethered to the plasma membrane. If the membrane-binding affinity of the plus-end-binding complex was stronger than the membrane-binding affinity of the γTuRC particle, γTuRC was pushed in the direction of the minus-end, while the plus-end continued to polymerize, but was anchored at a fixed location. This means that the membrane-associated plus-end complex allows tubulin incorporation at the growing plus-end, while anchoring the plus-end to the membrane. When MTs become long enough to accumulate sufficient numbers of lattice-anchoring proteins, a newly formed MT may become stably associated with the plasma membrane, and the growing plus-end may be allowed to extend forward while attaching to the end-anchoring complex (Fig. 4C).

A cortical MT has occasionally been reported to ‘wag its plus-end tail’ when the plus-end is dissociated from the cortex (Shaw et al., 2003; Ambrose and Wasteneys, 2008; Hashimoto, 2015), indicating that the plus-end provides an important cortical attachment site. MT plus-end-tracking proteins (+TIPs; Kumar and Wittmann, 2012; Akhmanova and Steinmetz, 2015) are obvious candidates for mediating the plus-end association of MTs with the cell cortex, and may associate with the plasma membrane directly, possibly by interacting with phospholipids or membrane-anchored proteins, or indirectly through unknown adaptor proteins. An Arabidopsis homolog of the animal canonical +TIP, CLASP, associates with the plus-ends and sides of cortical MTs in Arabidopsis cells, and the loss of CLASP results in partial detachment of MTs from the cortex (Ambrose and Wasteneys, 2008). Other plant +TIPs may also be involved in tethering MTs to the plasma membrane at their plus-ends.

To obtain a transgenic plant for the simultaneous visualization of γTuRC and MT, the mzt1 double mutant line complemented with the MZT1a–GFP construct (Nakamura et al., 2012) was transformed with the UBI10 promoter-driven mCherry–TUB6 construct (Fujita et al., 2013) using the Agrobacterium-mediated floral dip method (Clough and Bent, 1998). A transgenic line was selected that did not show any apparent morphological abnormalities, such as radial swelling or twisting, and expressed both fluorescent markers at appreciable levels. Seedlings were grown on 1.5% agar medium containing half-strength Murashige and Skoog salt mixture (pH 5.7) and 1% sucrose at 22°C under continuous light conditions.

Pavement cells in the cotyledons of 13- to 16-day-old seedlings were analyzed. Time-lapse imaging of fluorescent protein localization was performed according to Wong and Hashimoto (2017). For dual-color visualization, mCherry was excited and its emission light collected with an exposure time of 0.1 s, and GFP was subsequently excited and its emission signal collected with an exposure time of 0.2 s. Image processing and analyses were performed using Fiji/ImageJ software.

We thank Yuichi Sakumura (NAIST) for statistical analysis, Takehide Kato (NAIST) for sample preparation, and Masayoshi Nakamura (Nagoya University) for providing us with Arabidopsis plants expressing both GFP–MZT1a and GCP2–mCherry.