A fundamental aspect of microtubule (MT) array organization in all eukaryotes is the spatial organization of MT polarity. The best characterized mechanism driving MT organization requires a MT organizing center, such as a centrosome, which initiates MT polymerization and anchors these MTs with their (+)-ends radiating outward (Ou and Rattner, 2004 ; Varmark, 2004 ). However, this mechanism does not seem to operate to organize the plant cortical MT array because higher plants lack traditional centrosomes (Schmit, 2002 ). The acentrosomal plant cortical MT array is nonetheless capable of adopting several organizational states (Dixit and Cyr, 2004a ; Lloyd and Chan, 2004 ), which play pivotal roles in plant morphogenesis and regulate both cell division and cell elongation events (Mineyuki, 1999 ; Wasteneys, 2004 ).
Plant cortical MTs are plasma membrane associated dynamic structures (Dixit and Cyr, 2004a ) that are nucleated from dispersed sites throughout the cortex (Wasteneys and Williamson, 1989 ; Yuan et al., 1994 ; Granger and Cyr, 2001 ; Shaw et al., 2003 ). Significantly, the (–)-ends of these MTs are unanchored and consequently both MT ends are dynamic with the (+)-ends showing dynamic instability and the (–)-ends undergoing slow depolymerization, collectively known as hybrid treadmilling (Shaw et al., 2003 ). The polymer dynamics of plant cortical MTs can be described by stochastic parameters, which by themselves do not explain how cortical MT organization can occur. Recently, it was shown that the stochastic cortical MT dynamics are deterministically modified by the angle at which they encounter one another (Dixit and Cyr, 2004b ). These deterministic events provide a mechanistic route that results in the parallel arrangement of cortical MTs; however, the relationship between MT polarity and organization in cortical MT arrays is not well understood.
To study the relationship between MT polarity and organization, we utilized the Arabidopsis End Binding 1 (EB1) family of MT-binding proteins because EB1 proteins are best known for their ability to bind MT (+)-ends (Bisgrove et al., 2004 ; Akhmanova and Hoogenraad, 2005 ). The Arabidopsis EB1 protein family consists of three members, EB1a, EB1b, and EB1c (Chan et al., 2003 ; Mathur et al., 2003 ), which show between 82 and 47% overall amino acid sequence similarity to one another. The N-termini (1–133 amino acids) of EB1a and EB1b are ~93% similar and they are ~73% similar to the N-terminus of EB1c. In comparison, the C-termini of EB1a and EB1b are ~71% similar and they are ~32% similar to the C-terminus of EB1c. EB1a and EB1b are also more similar to human and yeast EB1 orthologues (30–35%), compared with EB1c (20–30%). Predictably, EB1a and b proteins would make ideal MT (+)-end markers; however, earlier studies using overexpressed proteins surprisingly showed that they extensively bind MTs and result in (–)-end (Chan et al., 2003 ) and endomembrane (Mathur et al., 2003 ) labeling.
In this study, we engineered both Arabidopsis and tobacco plant cells to stably express GFP (green fluorescent protein)-tagged EB1a and b at the proper levels of expression in order to better analyze cortical MT organization. We show that under low expression conditions, GFP-tagged EB1a and b, as expected, exclusively bind the (+)-end of MTs. With this tool, we demonstrated that the majority of MTs in organized cortical arrays, in a given cell, have the same polarity and that the polar arrangement of cortical MTs occurs simultaneously with parallel MT arrangement. Finally, we present evidence that the polar MT alignment involves a selective stabilization of MTs of the same polarity and that MT polymerization rates do not substantially influence this process.
Heat-shock induction was achieved by exposing BY-2 cells (25 ml of cells in a 125-ml flask at 100 rpm rotation) and Arabidopsis plants (in a humid chamber) to 34°C for 2 h. Preliminary experiments showed that this treatment regime did not perturb cell growth, MT dynamics, or mitotic index. Although this treatment was mild, we routinely followed heat shock with recovery for at least 3 h at room temperature before conducting observations.
Wide-field microscopy was conducted with a Zeiss Axiovert S100 TV microscope. Images were captured with a CoolSNAP HQ camera (Roper Scientific, Tucson, AZ) using 20–30% light intensity from a 100 W mercury arc-lamp and GFP (460–500-nm excitation, 510–560-nm emission) filter set at 5–10-s intervals using 1-s exposure times over a 5–10-min period. MT orientation was quantified by determining the angle of MT growth relative to the long axis of the cell. For these measurements, up to 20° difference in MT angles was considered to represent the same direction because MT encounters at this angle resulted in MT coalignment 90% of the time (Dixit and Cyr, 2004b ). MTs that started out at divergent angles at the beginning of the movie but which subsequently became coaligned with other MTs after MT encounters were considered to be oriented in the same direction. However, if a rare MT changed direction more than 20° at the end of the movie, it was scored as a divergent MT.
Confocal microscopy was conducted with a Zeiss 510 Meta laser scanning microscope. The 488-nm laser line, at 3–5% power output, was selected for GFP excitation, and images were collected using 1-s scan times. For the fluorescence recovery after photobleaching (FRAP) experiments, photobleaching was achieved using three scans at 100% laser power, followed by image acquisition using 3% laser power at 3-s intervals.
To determine whether the EB1-GFP localization patterns are conserved across plant families, EB1-GFP proteins were similarly expressed in stably transformed tobacco BY-2 cells. Our data showed that, as in Arabidopsis, low expressing EB1a-GFP and EB1b-GFP appeared in a cometlike pattern (1.01 ± 0.11 μm for EB1a; 1.11 ± 0.18 μm for EB1b; n = 50; Figure 1, A and B, and Supplementary Movie 2), whereas EB1c-GFP is nuclear (Figure 1C). The cometlike character, with the bright side leading has been shown to be diagnostic of (+)-end growth (Tirnauer et al., 2002 ) and in all instances this behavior was observed in plant cells that expressed EB1 at low levels. In addition, during mitosis, EB1a-GFP and EB1b-GFP comets are seen to migrate toward and accumulate at the spindle midzone (i.e., away from the spindle poles), which is consistent with MT (+)-end localization because (+)-ends extensively interdigitate in this area of the spindle. Accumulation was not observed at the spindle poles (Supplementary Movie 3). EB1a-GFP and EB1b-GFP are indistinguishable in terms of comet growth velocity (5.10 ± 0.7 μm/min vs. 4.98 ± 0.78 μm/min; n = 100 MTs, respectively). Notably, in contrast to the low expression conditions, high expression of EB1b-GFP in BY-2 cells, using the 35S promoter, results in loss of (+)-end-labeling specificity and more extensive labeling of MTs, which indicates the lower affinity sidewall binding site is present in plant EB1 (Supplementary Figure 2).
To further investigate whether functions of EB1 are evolutionarily conserved, we examined whether EB1 can rescue MT defects in fission yeast that lack the EB1 ortholog Mal3. As shown in Figure 1D, mal3 null cells are hypersensitive to drugs that destabilize MTs, such as TBZ (Chen et al., 2000 ), and this defect is more efficiently rescued by EB1b and only weakly rescued by EB1c.
Collectively, we concluded from these results that EB1a and EB1b, like the mammalian and yeast counterparts, display a strong MT (+)-end binding activity under low expression conditions, whereas EB1c is a more divergent form.
In isodiametric BY-2 cells, the EB1a-GFP and EB1b-GFP comets are oriented in random directions, suggesting that these MTs are not organized and thus displaying random MT (+)-end polarity (Supplementary Figure 3 and Supplementary Movie 4). However, in elongated BY-2 cells, the EB1a-GFP and EB1b-GFP markers revealed that the majority of MT (+)-ends (76 ± 12% with EB1a; >600 microtubules; and 79 ± 13% with EB1b; >800 microtubules; 10 cells each) are oriented transversely with the same polarity (Figure 2A and Supplementary Movie 5). Moreover, optical sectioning of these cells showed that the cortical MT (+)-end polarity is continuous along all faces of the cell (i.e., the polarity is observed circumferentially about the cell; unpublished data).
To better understand this phenomenon in organized tissues, we also analyzed Arabidopsis plants that were stably transformed with a construct expressing EB1b-GFP, under the control of the native EB1b promoter, and found that epidermal cells containing transverse (indicative of a rapid elongative state) or longitudinal (indicative of tissues that have ceased elongating; Dixit and Cyr, 2004a ) cortical arrays also display polar (76 ± 9%; >500 microtubules; 12 cells; 5 independent lines) MT (+)-end organization (Figure 2B and Supplementary Movie 6).
Using the same analytical approach as with the cultured BY-2 cells, we closely examined cells in these organized tissues (i.e., all cells had ordered cortical MTs) and observed an interesting pattern in which ~70% of cells with organized cortical MT arrays exhibit global cortical array polarity; that is, the same net MT polarity was observed in all regions of the cortex in these cells. Furthermore, the net cortical MT array polarity appeared to be cell autonomous, i.e., one cell can contain MTs of one polarity, while its neighbor can contain MTs of the opposite polarity. In the remaining 30% of cells with organized cortical MT arrays, regional variation in net MT polarity was observed. In such cells, distinguishable belts or hoops of opposing MT polarity occur in the cell cortex (Figure 3, A and B). In these cells, we noted a boundary area where MT dynamics appeared somewhat depressed as judged by a lower frequency of comets in this area. These boundary regions were not included in our quantitative analyses. Similar to cells with global MT array polarity, the net polarity within individual belts of MT polarity is ~80% on average. In all cases, we typically observed the same net organization in neighboring cells (i.e., similar net transverse, oblique, or longitudinal arrangement) but the net direction of microtubule polarity could differ.
Our data obtained from EB1b-GFP-expressing BY-2 cells (Supplementary Figure 4) show that the average “lifespan” of the comets is significantly greater among MTs oriented in the majority direction (85 ± 15 s vs. 67 ± 11 s; p-value of 0.002 using t test; >500 MTs). MT comets oriented in the majority direction frequently followed the tracks of previous comets (Supplementary Figure 4), consistent with MT bundling following shallow-angle MT encounters (Dixit and Cyr, 2004b ). On the other hand, MT comets oriented in the minority direction frequently disappeared after encountering MT comets growing in the majority direction. These observations are consistent with the hypothesis that cortical MT array polarity is established through selective stabilization (and therefore enrichment) of copolar MTs compared with antiparallel MTs. We found no significant difference in the comet velocities between MTs of opposing directions, suggesting that MT polymerization rate is not the mechanism leading to net cortical array polarity.
Although the organization of MTs by the centrosome elegantly explains how many MT structures are assembled, it is challenging to envision how organized MT arrays may be assembled without a centrosome. The plant cortical MT arrays thus represent an ideal model system to study such an event. Our data show that MTs in these arrays, in a given cell, are organized such that they predominantly project with the same polarity. Our data further illustrated that the ordering of cortical MTs occurs soon after cortical MT nucleation at the end of cytokinesis, during which MTs with the same polarity gradually enrich and become parallel to one another. Finally, we revealed that cortical MT array polarity is associated with selective stabilization of MTs oriented in the majority direction.
The study of densely packed cortical MTs has been hindered by the lack of good MT end markers. Our study suggests that EB1a or EB1b is a powerful tool for this type of analysis, provided that they are properly expressed. At low expression levels, we observed that EB1a and EB1b primarily label growing MT (+)-ends throughout the cell cycle, whereas at high levels they extensively decorate MT sidewalls. This difference in how EB1 interacts with MTs, dependent on the level of expression, is consistent with the idea that EB1 has two binding sites for MTs: one with high-affinity that is specific for (+)-end labeling and one with lower affinity specific for binding to the sidewall. Data in support of this comes from work with budding yeast and Xenopus (Tirnauer and Bierer, 2000 ; Schuyler and Pellman, 2001 ; Tirnauer et al., 2002 ) and here we show a similar phenomenon, which indicates the duality in MT interactions is likely an evolutionary conserved feature of EB1 proteins. We did not detect any (–)-end labeling by EB1a and EB1b as was reported by Chan et al. (2003 ), who worked with cells that were expressing high levels of EB1. This difference in how plant EB1 interacts with MTs most likely is due to the difference in protein expression levels because native promoter-driven, low-level EB1 expression by the same group also did not result in MT (–)-end labeling (Chan et al., 2005 ).
Our studies with plant EB1 also reveal that EB1c may in fact represent a new member of the EB1 family, with a distinct function, because it does not associate with the (+)-ends of cortical MTs but rather shows nuclear localization during interphase and spindle and phragmoplast localization during mitosis (Van Damme et al., 2004a and our unpublished results). Furthermore, it does not replace yeast EB1 as efficiently as EB1b. Because EB1c diverges most extensively from other EB1 proteins in the C-terminus, we speculate that the C-terminus of EB1 proteins can impart specific functions shared by a given EB1 subfamily.
The observed polarity of cortical MTs was first deduced from studying EB1 and later it was corroborated by FRAP experiments on GFP-TUB6. Both techniques effectively sample the entire cortical MT array based on the well-established fact that plant cortical MTs are highly dynamic. Although the EB1 markers localize only to growing (+)-ends, they are expected to label all cortical MTs at some point during our observation period because of the high dynamicity and rescue frequency of cortical MTs (Chan et al., 2003 ; Dhonukshe and Gadella, 2003 ; Mathur et al., 2003 ; Shaw et al., 2003 ; Dixit and Cyr, 2004b ; Nakamura et al., 2004 ; Van Damme et al., 2004a ; Vos et al., 2004 ; Abe and Hashimoto, 2005 ). We specifically confirmed the high dynamicity of cortical MTs by performing FRAP analyses on GFP-TUB6 that showed complete recovery of photobleached MTs within 2–3 min. Therefore, we are confident that the described organization of MTs is a general feature of the cortical array and is not influenced by which marker was used for the analysis.
Although we have demonstrated that organized cortical MTs, in a given cell, have a predominant polarity, some previous reports suggested that cortical MT arrays do not have a well-defined polarity after examining MTs data sets pooled from many cells (Shaw et al., 2003 ; Tian et al., 2004 ; Vos et al., 2004 ). Our study provides some important clues to allow for a better understanding of these conflicting conclusions. First, tracking MT dynamics using GFP-tubulin, rather than using an end marker, can underestimate the extent of MT coalignment because of limitations in sampling MTs in a crowded environment. It is difficult to resolve growing (+)-ends of bundled MTs using GFP-tubulin, whereas, the (+)-ends of bundled MTs are easily detectable by EB1-GFP as discrete comets. Furthermore, MTs oriented in the same direction frequently track along existing MTs (Dhonukshe et al., 2005 ) and therefore, predictably, remain unresolved when observed using GFP-tubulin. In addition, our study found that MT coalignment is readily detectable in elongated cells, whereas cells that are nearly spherical do not exhibit such coalignment. Finally, in ~30% of cells with array polarity, their MTs can be divided into two or more regions of opposing polarity. Hence, in about one third of cells there is no “net” coalignment of MTs along the entire cell. Although hook decoration of cortical MTs suggested mixed MT polarity (Tian et al., 2004 ), it is important to highlight that the authors also noted that only 30% of cortical MTs were hook decorated (Tian et al., 2004 ). Because of this technical limitation, this study relied on quantification of MT polarity by pooling data obtained from small sample sets (3–6 MTs) from numerous different cells. Because we have reported here that the net MT polarity can vary from cell to cell, pooling data from large numbers of cells will predictably give the impression of mixed polarity. In support of our observations, examination of the literature shows undiscussed cases of net MT polarity within plant cortical MT arrays. For example, Figure 2D in Chan et al. (2003 ) shows ~74% net MT polarity; and, Figure 1B in Dhonukshe et al. (2005 ) shows ~75% net MT polarity. Furthermore, analyses of serial electron micrograph sections from individual cells also show that adjacent cortical MT ends share a common directionality (Hardham and Gunning, 1978 ).
In terms of the mechanism of polar coalignment of cortical MTs, our data show that cortical MT array polarity is established progressively after MT nucleation at the end of cytokinesis and is concomitant with the parallel ordering of cortical MTs. Cortical MT nucleation predominantly occurs from existing cortical MTs in a γ-tubulin-dependent manner (Murata et al., 2005 ). This MT-dependent nucleation results in branching patterns of cortical MTs (Falconer et al., 1988 ; Wasteneys and Williamson, 1989 ; Wasteneys, 2002 ; Murata et al., 2005 ) that are subsequently resolved into coaligned cortical MTs, probably due to intermicrotubule interactions (Dixit and Cyr, 2004b ). Specifically, branching MTs colliding with neighboring MTs at shallow angles would become bundled and stabilized, whereas, those MTs colliding with neighboring MTs are steep angles would depolymerize and be lost from the population. Over time, these intermicrotubule interactions foster the generation of coaligned cortical MTs (Dixit and Cyr, 2004b ). In this scheme, selective stabilization of cortical MTs in a particular direction would mean that subsequent MTs nucleated from these “pioneer” MTs would possess roughly the same polarity and eventually resolve into a polar coaligned array. In support of this hypothesis, we show that as MTs achieve coalignment, there is an enrichment of MTs with the same polarity relative to those with the opposite polarity, and our data show that this process is associated with a selective stabilization of copolar MTs, and not by any change in MT polymerization rate. From a molecular standpoint, MT crossing-linking proteins, such as MAP65 isoforms, may facilitate the process of aligning MTs that are close to one another. Indeed, AtMAP65-1 and AtMAP65-5 predominantly decorate coaligned cortical MTs (Van Damme et al., 2004b ; Mao et al., 2005 ) and it will be interesting to determine whether these MTs have uniform polarity.
From the perspective of the cortical MT array function, we speculate that the cortical MT array polarity influences the polar deposition of cellulose wall microfibrils, which themselves are polar polymers (i.e., the reducing ends of the β-1,4 glucan chains are oriented distal to the cellulose synthase complex). The inherent polarity of these two interacting systems (i.e., MTs and cellulose) may play an integral role in the cross-talk between these two systems (Fisher and Cyr, 1998 ). One possibility is that cellulose synthase complexes interact, in some manner, with the cortical MT array in a chiral manner, as suggested by the unidirectional microfibril deposition by groups of cellulose synthase complexes (Brown and Montezinos, 1976 ; Herth, 1984 ; Kudlicka et al., 1987 ; Delmer and Amor, 1995 ; Tsekos et al., 1999 ). The predilection for unidirectional higher order cellulose microfibril bundles also suggests polar microfibril deposition (Brett, 2000 ). In addition, the Arabidopsis FRA1 (Fragile fiber 1) gene encodes a kinesinlike protein with predicted (+)-end motor activity that is required for the oriented deposition of cellulose microfibrils (Zhong et al., 2002 ), consistent with the idea that polarity within the cortical MT array is conveyed to the cellulose microfibril array. This notion suggests we rethink the paradigm of simple coalignment between MTs and microfibrils and extend it to encompass their mutual, polar coalignment.
We thank Imelda Mercado for her expert help with the yeast rescue experiments and Takashi Hashimoto for GFP-TUB6-expressing Arabidopsis plants. Work in the Chang lab is supported by a National Institutes of Health grant (CA90464) and work in the Cyr lab is supported by a DOE grant (DE-FG02-91ER20050).