Cortical microtubule arrays are initiated from a non-random pre-pattern driven by atypical microtubule initiation

The ordered arrangement of cortical microtubules in growing plant cells is essential for anisotropic cell expansion and hence for plant morphogenesis. These arrays are dismantled when the microtubule cytoskeleton is rearranged during mitosis and reassembled following completion of cytokinesis. The reassembly of the cortical array has often been considered as initiating from a state of randomness, from which order arises at least partly through self-organizing mechanisms. However, some studies have shown evidence for ordering at early stages of array assembly. To investigate how cortical arrays are initiated in higher plant cells, we performed live cell imaging studies of cortical array assembly in tobacco BY-2 cells after cytokinesis and drug-induced disassembly. We found that cortical arrays in both cases did not initiate randomly, but with a significant over-representation of microtubules at diagonal angles with respect to the cell axis, which coincides with the predominant orientation of the microtubules before their disappearance from the cell cortex in preprophase. In Arabidopsis root cells, recovery from drug-induced disassembly was also non-random, and correlated with the organization of the previous array, although no diagonal bias was observed in these cells. Surprisingly, during initiation only about half of the new microtubules were nucleated from locations marked by GFP-GCP2 tagged gamma-nucleation complexes ( γ -TuRC), therefore indicating that a large proportion of early polymers was initiated by a non-canonical mechanism not involving γ -TuRC. Simulation studies indicate that the high rate of non-canonical initiation of new microtubules has the potential to accelerate the rate of array re-population. performed high time resolution observations to quantify nucleation complex recruitment, nucleation rates and microtubule nucleation angles. We found evidence for a highly non-random initial ordering state that features diagonal microtubule orientation and an atypical microtubule initiation mechanism. Simulation analysis indicates that these atypical nucleations have the potential to accelerate the recovery of cortical array density.


Introduction (5813 characters including spaces)
Higher plant cells feature ordered arrays of microtubules at the cell cortex (Ledbetter and Porter, 1963) that are essential for cell and tissue morphogenesis, as revealed by disruption of cortical arrays by drugs that cause microtubule depolymerization (Green, 1962) or stabilization (Weerdenburg and Seagull, 1988), and by loss of function mutations in a wide variety of microtubule associated proteins (MAPs) (Baskin, 2001;Whittington et al., 2001;Buschmann and Lloyd, 2008;Lucas et al., 2011). The structure of these arrays is thought to control the pattern of cell growth primarily by its role in the deposition of cellulose microfibrils, the loadbearing component of the cell wall (Somerville, 2006). Functional relations between cortical microtubules and cellulose microfibrils have been proposed since the early sixties, even before cortical microtubules had been visualized (Green, 1962). Recent live cell imaging studies have confirmed that cortical microtubules indeed guide the movement of cellulose synthase complexes that produce cellulose microfibrils (Paredez et al., 2006) and have shown further that microtubules position the insertion of most cellulose synthase complexes into the plasma membrane (Gutierrez et al., 2009). These activities of ordered cortical microtubules are proposed to facilitate the organization of cell wall structure, creating material properties that underlie cell growth anisotropy.
While organization of the interphase cortical array appears to be essential for cell morphogenesis, this organization is disrupted during the cell cycle as microtubules are rearranged to create the preprophase band, spindle and phragmoplast during mitosis and cytokinesis (reviewed by Wasteneys, 2002). Upon completion of cytokinesis, an organized interphase cortical array is regenerated, but the pathway for this reassembly is not well understood.
The plant interphase microtubule array is organized and maintained without centrosomes as organizing centers (reviewed in Wasteneys, 2002;Bartolini and Gundersen, 2006;Ehrhardt and Shaw, 2006) and microtubule self-organization is proposed to play an important role in cortical microtubule array ordering (Dixit and Cyr, 2004). In electron micrographs microtubules have been observed to be closely associated with the plasma membrane (Hardham and Gunning, 1978) and live cell imaging provides evidence for attachment of microtubules to the cell cortex (Shaw et al., 2003;Vos et al., 2004). The close association to the plasma membrane restricts the cortical microtubules to a quasi two-dimensional plane where they interact through polymerization-driven "collisions" (Shaw et al., 2003;Dixit and Cyr, 2004).
Microtubule encounters at shallow angles (< 40 degrees) have a high probability of leading to bundling, while microtubule encounters at steeper angles most likely result in induced catastrophes or microtubule crossovers (Dixit and Cyr, 2004). Several computational modeling studies have since shown that these types of interactions between surface-bound dynamical microtubules can indeed explain spontaneous coalignment of microtubules (Allard et al., 2010;Eren et al., 2010;Hawkins et al., 2010;Tindemans et al., 2010).
The question of how the orientation of the cortical array is established with respect to the cell axis is less well understood. One possibility is that microtubules are selectively destabilized with respect to cellular coordinates (Ehrhardt and Shaw, 2006). Indeed, recent results from biological observations and modeling suggest that catastrophic collisions induced at the edges between cell faces, or heighted catastrophe rates in cell caps could be sufficient to selectively favor microtubules in certain orientation and hence determine the final orientation of the array (Allard et al., 2010;Eren et al., 2010;Ambrose et al., 2011;Dhonukshe et al., 2012).
To date, all models of cortical array assembly assume random initial conditions. However, experimental work by Wasteneys and Williamson (1989a, b) in Nitella tasmanica showed that, during array reassembly after drug-induced disruption, microtubules were initially transverse. This was followed by a less ordered phase and later by the acquisition of the final transverse order. A non-random initial ordering was also observed in tobacco BY-2 cells by Kumagai et al. (2001), who concluded that the process of transverse array establishment starts with longitudinal order, but did not provide quantitative data for the process of array assembly. The initial conditions for the cortical microtubule array formation are important to consider, as they may strongly influence the speed at which order is established, and could even prevent it from being established over a biologically relevant time scale.
In the present study, we used live cell imaging to follow and record the whole transition from the cortical microtubule-free state to the final transverse array and used digital tracking algorithms to quantify the microtubule order. Nucleation stands out as a central process to characterize during array initiation. Lacking a central body to organize microtubule nucleations, the higher plant cell has dispersed nucleation complexes (Wasteneys andWilliamson, 1989, 1989;Chan et al., 2003;Shaw et al., 2003;Murata et al., 2005;Pastuglia et al., 2006;Nakamura et al., 2010). Therefore we

After cytokinesis, microtubules in BY2 cells reappear with a transient diagonal order
To investigate array initiation, we used tobacco Bright Yellow 2 (BY-2) suspension cells expressing GFP fused to tobacco α -Tubulin (GFP-TUA). These cells feature highly ordered arrays oriented transversely to the axis of growth, have a relatively high mitotic index, and are ideal for drug treatment in flow cell experiments. Furthermore, the potential crosstalk with neighbors is limited because BY-2 cells generally grow in cell files that break up into individual cells (Chan et al., 2011;Crowell et al., 2011;Fujita et al., 2011).
Using point-scanning confocal microscopy, we acquired images from the plane of the cell cortex every 3-5 minutes and measured microtubule length density and ordering after cytokinesis. The first visible microtubules appeared in the cortex after the phragmoplast reached the optical plane of the cell cortex (Figure 1a, Figure S1 and Movie S1) and within ~45 minutes the length density, defined as microtubule length per square micrometer, leveled at around 0.5 µm/µm 2 (= µm -1 , mean of 6 cells; Figure 1b). With the increase in length density, the microtubules also became increasingly bundled, as indicated by increases in the fluorescence intensities of individual microtubule structures. As our focus was on microtubule orientation, we gave bundles the same weight as individual microtubules.
The angles of microtubules with respect to the cell elongation axis were measured and visualized in a contour plot ( Figure 1c). Time is presented along the xaxis and the angular distribution over the interval from 0° to 180° along the y-axis (20 bins). The color range represents the fraction of the total microtubule length, so that orientation patterns at both low and high microtubule densities can be compared.
Surprisingly, the majority of the microtubule length was diagonally oriented at 45° and 135° angles to the elongation axis in the early stages of array reformation, forming two clear peaks in the angular frequency histogram.
To quantify the transition from the diagonal to the transverse cortical microtubule order, the angular distribution data were filtered to produce the weighted diagonal order parameter D and the weighted transverse order parameter T (see Supplementary Information). From the means of the D and T order parameters over time, we infer that the diagonal ordering was dominant for the first ~25 minutes after which it was replaced by transverse ordering (Figure 1d).

Transient diagonal ordering during recovery from oryzalin treatment in BY2 cells
To establish if the mechanism of transverse microtubule ordering via a transient diagonal phase is generic or cell cycle dependent, we immobilized BY-2 cells expressing GFP-TUA in flow cells and treated them for 1 hour with 20 µM oryzalin to reversibly depolymerize the cortical microtubule array (Morejohn et al., 1987) (Figure 2a and Movie S2). This concentration and duration of oryzalin treatment was sufficient to eliminate all detectible GFP-TUA labeled microtubules. Both the microtubule length density increase and the development of ordering after oryzalin wash out were similar as observed after cell division ( Figure 2b). The average plateau density was reached ~25 minutes after appearance of the first cortical microtubules, which is ~45 minutes after the start of the oryzalin wash out (mean of 8 cells). The first microtubules reappeared at diagonal angles to the elongation axis (45° and 135°; Figure 2c). On average, the transient diagonal ordering was replaced by the final transverse ordering after ~40 minutes ( Figure 2d). Thus, it appears that both the pattern and kinetics of assembly and ordering are similar whether the array is disassembled by native mechanisms during the cell cycle, or by drug treatment.

Diagonal ordering also occurs during array disassembly in BY2 cells
Interestingly, a diagonal bias for microtubule orientation was also observed during late stages of array disassembly as cells exit interphase and form preprophase bands (observations from 5 cells, Figure S2). Likewise, the same bias was observed in late stages of microtubule depolymerization caused by oryzalin application (Figure 2c and d). The microtubule length density started to decrease less than a minute after drug application and reached zero microtubules after ~16 minutes. Within 2 minutes after oryzalin addition, a diagonal microtubule order took over the dominant transverse order and lasted until the last microtubules were depolymerized ( Figure   2d). Thus diagonal biasing of microtubule orientation appears to be a feature both of the last stages of array disassembly and the first stages of array re-assembly, irrespective of whether arrays are taken apart by cellular mechanisms or by drug treatment.

Microtubule nucleation has a diagonal bias during array initiation in BY2 cells
A bias in microtubule orientation might occur because microtubules are preferentially created in specific orientations, or because they are selectively destabilized, or if they are reoriented once initiated. To assess the origin of the diagonal microtubule ordering, we made movies at high time resolution (2s intervals) of BY2 cells expressing GFP-TUA cytokinesis and oryzalin wash out (Movie S3). We observed that in the first 30 minutes the majority of new microtubules were nucleated at the cell cortex at locations free of other detectable microtubules. The majority of nucleations during this period were free nucleations (274 out of 352, 77%, in 6 cells after cytokinesis, and 73 out of 117, 62%, in 5 cells after oryzalin wash out). These observations are in contrast to those of interphase nucleation, where microtubuleassociated microtubule nucleations have been observed to comprise greater than 99% of nucleations in wild type Arabidopsis cells (Murata et al., 2005;Nakamura et al., 2010;Kirik et al., 2012). We measured the angles of these free nucleations with respect to the cell axis after both cytokinesis ( Figure 3a) and oryzalin wash out ( Figure 3b). We did not analyze microtubule nucleations in the same orientation as the microtubule they nucleated on, as they do not give rise to new microtubule orientations. A Bayesian statistical analysis of these data (see Materials and Methods) revealed a significant bias for nucleations to occur along the diagonal directions both after cytokinesis and oryzalin wash out.

In Arabidopsis root cells, a large fraction of nucleations during array initiation
are free of labeled γ -tubulin complexes We found it remarkable that the nucleation bias had the same orientation as the cortical microtubule order just before disappearance. This suggested that a 'memory' of the previous array organization might be maintained at the cell cortex.
We could imagine three alternative models. First, nucleation complexes recruited to the previous array might persist at the cell cortex, retaining orientational information.
Second, there might be other orientational information at the cell cortex that acts to orient newly recruited nucleation complexes as they initiate the next array. Finally, a subset of the previous array might be resistant to disassembly either by native mechanisms or by drugs, and they may be either small enough to evade detection by GFP-TUA5 labeling (or that the specific isoform of alpha tubulin we labeled is not or 1 1 less incorporated into these structures). These disassembly-resistant remnants might act as orientated seeds for initiating new polymerization during array reassembly.
To distinguish among these hypotheses, we assayed the localization and dynamic behavior of After drug washout, we acquired images of the cell cortex at high time resolution (2s intervals). We observed no evidence for persistent GFP-labeled nucleation complexes at the cell cortex, thus refuting the first hypothesis; that nucleation complexes recruited to the previous array might persist at the cell cortex to initiate the new array.
We then scored all observed nucleation events in the field of view, asking if GCP2-3xGFP was present at the position of microtubule nucleation. As labeled complexes are present and motile in the streaming cytosol (Nakamura et al., 2010), we required that punctae GFP signal be present at the position of nucleation for at least two consecutive image frames to be scored positively. In control cells that were not pretreated with oryzalin, we found that 68 out of 70 nucleations (97%) colocalized with the GCP2-3xGFP label ( Figure 4a and Movie S5, data acquired from 6 cells on 6 plants), a frequency in good agreement with the ~98% found by Nakamura et al. label in the first 20 minutes after the start of oryzalin wash out, a dramatically lower proportion (p << 0.0001, one-tailed binomial test, 8 cells). Thus, while only ~3% of nucleations was not observed to be accompanied by GCP2-3xGCP in mature arrays, this frequency raised to ~44% during early stages of array assembly (Figure 4b). The lack of detectable γ -TuRC label at nearly half of the early nucleations argues strongly against the second hypothesis for diagonal nucleation orientation; that orientational information at the cell cortex directs the orientation of new nucleation complexes recruited to the cell cortex during early array assembly. We also found no evidence for involvement of two candidates for such orientational information, the cortical actin cytoskeleton and cellulose microfibrils, by disruption experiments with latrunculin B ( Figure S3) or isoxaben ( Figure S4), respectively.

2
On the other hand, the marked reduction in GCP2-3xGFP co-localization was consistent with the third hypothesis; that a large and significant proportion of nucleations during early array recovery arise from seeds not associated with γ -tubulin complexes. We term these nucleation events non-canonical nucleations because they lack association with detectible γ -TuRCs as determined by GCP2-3xGFP labeling, an essential subunit of the core γ -TuRC.
In the Arabidopsis root epidermal cells, we also determined the orientation the orientation of the cortical microtubule array during oryzalin treatment and the recovery from oryzalin wash out. We found that in these cells, the orientational bias of the initial array after oryzalin wash out was the same as the bias that we found during the depolymerization (Figure 4c and d, Figure S5, Movie S6). This was the case not only for transversely oriented cortical microtubule arrays, but also for oblique and longitudinal arrays (Figure 4c and d, Figure S5, Movie S6).

Simulations
We performed mechanistic simulations to ask if the observed prevalence of diagonal microtubule nucleation was sufficient to explain the degree of observed diagonal ordering during the initial stages of array assembly and to ask what affect these non-canonical nucleations might have on the evolution of array density and ordering. In the simulations, cortical microtubules interact on a cylindrical cell-shaped surface of dimensions similar to that of the tobacco BY-2 cells used in our in vivo experiments ( Figure 5a) (Tindemans et al., 2010;Deinum et al., 2011). To test for the influence of the non-canonical nucleation events, these nucleations were treated as a separate class, their density and orientations was chosen to match the distributions determined from the live cell experiments as described above in BY2 cells.

Discussion
The transverse arrangement of the cortical microtubule array is essential for anisotropic growth, yet little was known about how it arises from a disassembled state, the situation that recurs after each cell division during the life of the cell. The currently accepted self-organization models for transverse cortical microtubule array establishment, based on microtubule interactions (Allard et al., 2010;Eren et al., 2010;Hawkins et al., 2010;Tindemans et al., 2010), assume that new microtubules appear with random initial orientations. We found that the first microtubules in new arrays of tobacco BY-2 cells were in fact not randomly oriented, but showed significant ordering at orientations of both 45° and 135°. This was true both for array assembly during the cell cycle as well as reassembly of arrays after oryzalin washout.
Organization in these arrays did not evolve by gradual ordering from a disorganized, random state, but by a transition from one ordered state to another.
Exploration of the cause for the non-random initiation of array establishment revealed that there was a significant bias in the orientation of early microtubule nucleations, sharing the same 45° and 135° bias relative to the cell axis that was observed for array ordering. Results from simulation studies incorporating these oriented nucleations matched experimental observations very closely, indicating that this population of directionally biased nucleations is sufficient to explain the fastforming initial diagonal ordering state of new arrays, and together with self-ordering based on microtubule interactions, is sufficient to explain the transition from this transiently ordered array into the final transverse array.
Analysis of oryzalin treatment in Arabidopsis root epidermal cells showed that also in these cells the initial bias of microtubule orientation correlates with the last orientation before disassembly. This was true for not only the transverse cortical microtubule array, but also for oblique and longitudinal orientation, which strongly suggests that the bias depends on elements of the microtubule cytoskeleton before depolymerization.
We considered several alternative ideas for the mechanism of nucleation orientation. In interphase cells, the majority of nucleations at the cell cortex occurs the disassembly phase, suggesting that another factor may serve as the nucleator or that initiations may arise from stabilized short fragments of microtubules.
We therefore speculate that short fragments of microtubules may act as nucleators during early cortical array recovery. This would provide a mechanism by which the orientation of new initiations is tied to that of the previous array, in the absence of a microtubule scaffold that positions and orients nucleation complexes.
Previous studies by Wasteneys and Williamson (1989b) in Nitella are also consistent with this possibility. These investigators observed that while Nitella microtubules returned in their original transverse orientation during recovery from oryzalin, orientation was random after longer, and presumably more complete, oryzalin treatment (Wasteneys and Williamson, 1989Williamson, , 1989. In our studies, incomplete drug action cannot explain observations of array re-assembly following cytokinesis, since there was no drug treatment in these cells and the extremely similar mode and kinetics of array reassembly we observed between these cells and those recovering from oryzalin treatment suggest that a similar mechanism is responsible in both situations.
It is as yet unclear how such sort pieces of microtubules persist at the cell cortex during drug treatments for over an hour and even longer during cytokinesis.
We did observe that the rate of disassembly slows down over time. It is well possible that microtubule disassembly is in part dependent on how much the microtubule is physically loaded with microtubule associated proteins. Indeed, disassembly of microtubules could elevate the cytosolic concentration of microtubule associated proteins by decreasing the microtubule binding surface. An increased concentration of free MAPs in turn could drive association with any polymer remaining, potentially acting to stabilize short microtubule fragments. A second puzzle is why the presumed source of the oriented seeds -the last cohort of microtubules at the end of array disassembly -has a diagonal bias to the cell axis in the BY2 cells. One possibility is that the bias arises from the normal formation of the newest microtubules by branching nucleation at about 40 degrees to their mother polymers (Wasteneys andWilliamson, 1989, 1989;Murata et al., 2005;Chan et al., 2009;Nakamura et al., 2010). In a transversely oriented array, these nucleations would lie approximately at 45° and 135° to the cell axis, and would have a high likelihood of interacting with the dominant population of transverse microtubules.
These interactions can lead to incorporation into bundles by treadmilling motility (Shaw et al., 2003;Dixit andCyr, 2004), or catastrophe (Dixit andCyr, 2004), both of which would tend to diminish the population of diagonally oriented polymers.
However, as the microtubule array is broken down and microtubule density drops, encounters would be predicted to be less frequent and therefore the likelihood of aligning or eliminating branching microtubules will be reduced. Irrespective of whether the source of oriented nucleation in early array assembly is due to seeds from the previous array or another mechanism, our observations reveal the existence of a substantial class of non-canonical nucleations not associated with γ -TuRCs that contribute to the initiation of the cortical array. In simulation studies we explored how these non-canonical nucleations may affect array reassembly and found this class of oriented nucleations to have the potential to significantly accelerate recovery of array density without significantly impeding the acquisition of the ultimate ordering which is driven by microtubule interactions. The existence of this mechanism may address a fundamental dilemma the plant cell faces in rebuilding an array from scratch. In interphase cells, nucleation from

Specimen mounting
Transformed cells were imaged in thin ~10 to 20 µL gas permeable microchambers lined on one side with Biofoil (VivaScience, Hannover, Germany) and a 24 x 24 mm coverslip on the other side as described earlier (Vos et al., 2004). Slides For the nucleation analysis we used a confocal spinning disk microscope described earlier (Gutierrez et al., 2009), except that a Nikon Eclipse Ti microscope with the perfect focusing system and a 100x 1.45 NA oil objective replaced the Zeiss Axiovert 200. Alternatively, we used a total internal reflection fluorescence (TIRF) microscopy on a Nikon Eclipse Ti microscope with the perfect focusing system. We used a 100x 1.49 NA TIRF oil objective and excited with a solid-state 478nm laser (Cobolt AB) and using a Semrock 535/39 emission filter. The microscope was equipped with a manual Nikon TIRF arm and a QuantEM EM-CCD camera (Photometrics). We used 800 ms exposure time and a 2 or 2.14 s time interval for the spinning disk and TIRF microscope respectively.  Figure S7). Both functions have the property that a randomized system yields a value of zero. A system that is perfectly ordered (in the transverse direction for T and the diagonal direction for D) produces a value of 1.

ordering (see Supplementary Information and
For the nucleation analysis we determined the position in the cell, the time point, the angle with respect to the cell axis, whether the nucleation was free or microtubule bound and the angle of the seed microtubule in case of branching nucleation. For further analysis we only used the microtubule nucleations that were unbound. To assess whether a bias exists for nucleation along diagonal directions, we defined 15° degree bins around the 45°, 135°, 225° and 315° degree directions, and scored microtubules in these bins as being diagonal. We introduced a diagonal biasing     M  i  c  r  o  t  u  b  u  l  e  s  i  n  t  o  O  b  l  i  q  u  e  A  r  r  a  y  s  .  M  o  l  e  c  u  l  a  r  B  i  o  l  o  g  y  o  f  t  h  e  C  e  l  l  2  1  :   2  6  7  4  -2  6  8  4  F  u  j  i  t  a  M  ,  H  i  m  m  e  l  s  p  a  c  h  R  ,  H  o  c  a  r  t  C  H  ,  W  i  l  l  i  a  m  s  o  n  R  E  ,  M  a  n  s  f  i  e  l  d  S  D  ,  W  a  s  t  e  n  e  y  s  G  O   (  2  0  1  1  )  C  o  r  t  i  c  a  l  m  i  c  r  o  t  u  b  u  l  e  s  o  p  t  i  m  i  z  e  c  e  l  l  -w  a  l  l  c  r  y  s  t  a  l  l  i  n  i  t  y  t  o  d  r  i  v  e  u  n  i  d  i  r  e  c  t  i  o  n  a  l  g  r  o  w  t  h  i  n  A  r  a  b  i  d  o  p  s  i  s  .  P  l  a  n  t  J  o  u  r  n  a  l  6 6 : 9 1 5 -9 2 8 G r e e n P B (   C  a  n  a  d  i  a  n  J  o  u  r  n  a  l  o  f  B  o  t  a  n  y  -R  e  v  u  e  C  a  n  a  d  i  e  n  n  e  D  e  B  o  t  a  n  i  q  u  e  6  6  :   1  7  0  7  -1  7  1  6  W  h  i  t  t  i  n  g  t  o  n  A  T  ,  V  u  g  r  e  k  O  ,  W  e  i  K  J  ,  H  a  s  e  n  b  e  i  n  N  G  ,  S  u  g  i  m  o  t  o  K  ,  R  a  s  h  b  r  o  o  k  e  M  C  ,  W  a  s  t  e  n  e  y  s  G  O   (  2  0  0  1  )  M  O  R  1  i  s  e  s  s  e  n  t  i  a  l  f  o  r  o  r  g  a  n  i  z  i  n  g  c  o  r  t  i  c  a  l  m  i  c  r  o  t  u  b  u  l  e  s  i  n  p  l  a  n  t  s  .  N  a  t  u  r  e  4  1  1  : 6 1 0 -6 1 3