Helical Growth of the Arabidopsis Mutant tortifolia2 Does Not Depend on Cell Division Patterns but Involves Handed Twisting of Isolated Cells W

Several factors regulate plant organ growth polarity. tortifolia2 ( tor2) , a right-handed helical growth mutant, has a conservative replacement of Arg-2 with Lys in the a -tubulin 4 protein. Based on a published high-resolution (2.89 A ˚ ) tubulin structure, we predict that Arg-2 of a -tubulin forms hydrogen bonds with the GTPase domain of b -tubulin, and structural modeling suggests that these contacts are interrupted in tor2 . Consistent with this, we found that microtubule dynamicity is reduced in the tor2 background. We investigated the developmental origin of the helical growth phenotype using tor2 . One hypothesis predicts that cell division patterns cause helical organ growth in Arabidopsis thaliana mutants. However, cell division patterns of tor2 root tips appear normal. Experimental uncoupling of cell division and expansion suggests that helical organ growth is based on cell elongation defects only. Another hypothesis is that twisting is due to inequalities in expansion of epidermal and cortical tissues. However, freely growing leaf trichomes of tor2 mutants show right-handed twisting and cortical microtubules form left-handed helices as early as the unbranched stage of trichome development. Trichome twisting is inverted in double mutants with tor3 , a left-handed mutant. Single tor2 suspension cells also exhibit handed twisting. Thus, twisting of tor2 mutant organs appears to be a higher-order expression of the helical expansion of individual cells.


INTRODUCTION
As a result of cell division and cell elongation, plant organs expand into space to generate the growth axis. That axis may be straight but in some cases can be composed of cell files that are not aligned in parallel but are twisted around the axis. Expression of this feature during organ expansion is called helical growth. In Arabidopsis thaliana, helical growth can be induced by mutation. Indeed, the analysis of helical growth mutants has proved to be especially informative concerning the relationship between microtubule behavior and the direction of organ growth (for review, see Ishida et al., 2007a). The molecular identification of mutant loci has revealed mutations in aand b-tubulin (Thitamadee et al., 2002;Ishida et al., 2007b) as well as in several microtubule-associated proteins from plants (Whittington et al., 2001;Buschmann et al., 2004;Nakajima et al., 2004;Sedbrook et al., 2004;Shoji et al., 2004;Kaloriti et al., 2007;Perrin et al., 2007;Nakamura and Hashimoto, 2009). When applied in moderate concentrations, microtubule drugs also result in helical organ torsions (Furutani et al., 2000). Depending on the allele, helical growth mutants show either left-handed or right-handed organ torsions. Interestingly, although exceptions have been reported (Furutani et al., 2000;Thitamadee et al., 2002;Yuen et al., 2003), left-handed growth is usually epistatic over right-handed growth.
There is a long-standing paradigm that, during rapid cell growth, microtubules direct the alignment of cellulose microfibrils (for a discussion of models, see Baskin, 2001Baskin, , 2005. In many cases, microtubules and microfibrils encircle the cell perpendicular to the direction of cell expansion, and it is thought that microfibril orientation determines the polarity of cell expansion. Recent analysis of a fluorescence-tagged cellulose synthase has confirmed that this enzyme travels along the tracks provided by microtubules, supporting the idea that microtubules serve as a cytoplasmic template for microfibril orientation (Paredez et al., 2006). Apart from this interphase function in cell wall assembly, microtubules play important roles during mitosis and cytokinesis where they are involved in the regulation of tissue patterning and organ shape (Jü rgens, 2005;Lloyd and Buschmann, 2007;Mü ller et al., 2009).
Two opposing models may explain the development of a helical growth phenotype in Arabidopsis organs. According to the model by Hashimoto and coworkers (Furutani et al., 2000;Hashimoto, 2002), organ twisting is based on differences in cell elongation between epidermal cells and the underlying cortical cells. The epidermal cells of both Arabidopsis mutants or wildtype plants treated with microtubule-targeting drugs grew in helical files, and the cortical cells were often found to be swollen and impaired in longitudinal expansion. The cortical microtubules of epidermal cells were arranged in an oblique (rather than a strictly transverse) fashion, therefore shifting the axis of growth anisotropy to an oblique angle. As a result of the discrepancy in expansion between the two tissues, epidermal cells are assumed to tilt over to produce a rotation of the distal part of the organ.
Another model for helical organ growth can be derived from the cell division patterns in meristems. Helical patterns of division in a ring of initial cells give rise to the root epidermis and lateral root cap (Baum and Rost, 1996), and it is conceivable that these helical division patterns could impact on organ expansion and therefore organ shape. Interestingly, the twisting mutants lopped1/tornado1 and tornado2 show a number of defects in root meristem organization (Cnops et al., 2000). Based on this observation, Wasteneys and Collings (2004) speculated that helical growth could be based on an interference with the stereotypical division patterns in meristems.
Microtubules are deeply involved in the helical growth phenomenon (Lloyd and Chan, 2002). Microtubules are polymers consisting of 13 files of tubulin arranged to form a hollow cylinder (Lö we et al., 2001). Since tubulin is mainly added at the microtubule plus end, but is lost at the minus end, microtubules exhibit treadmilling (Shaw et al., 2003). Tubulin dimers are added head to tail onto microtubules so that the exposed fast-growing plus end is crowned by b-tubulin subunits (Nogales et al., 1999). Tubulin itself is a heterodimer made from aand b-tubulin subunits and has GTPase function. The docking of an arriving tubulin dimer is assumed to trigger GTPase function in the b-tubulin subunit of the previous dimer. This induces a conformational switch in tubulin, which appears to alter its capacity to support the tubulin lattice (Buey et al., 2006). Microtubule plus ends exhibiting a GTP cap (i.e., b-tubulin subunits carrying mainly unhydrolyzed GTP) are stable and continue polymerization. However, when the GTP cap is lost, microtubules are likely to undergo depolymerization (catastrophe). The aand b-tubulin proteins are an outstanding example of evolutionary protein optimization among the eukaryotic kingdoms (McKean et al., 2001;Buschmann and Lloyd, 2008). They are important drug and herbicide targets, and mutations leading to hyper-or hyposensitivity have been reported (Anthony et al., 1998;Richards et al., 2000;Morrissette et al., 2004). In yeast, tubulins impaired in GTPase function were reported to result in hyperstabilized microtubules (Anders and Botstein, 2001). A collection of plant tubulin mutations assumed to impact on microtubule stability was reported for plants. However, high-resolution structural insight allowing a mechanistic interpretation was not employed (Ishida et al., 2007b).
In this article, we present the helical growth mutant tortifolia2 (tor2), named after the helical growth mutant prototype tortifolia, described by Reinholz (Reinholz, 1947;Buschmann et al., 2004). The phenotype of this mutant is based on a conservative amino acid exchange of the Arg-2 residue of a-tubulin 4. We significantly improved earlier modeling of mutations in a-tubulin (Ishida et al., 2007b) by employing a higher-resolution crystallographic structure (at 2.89 Å ; Nettles et al., 2004) and inspecting the original electron density map. This allowed us to predict hydrogen bonds between residues and to assess the potential effect of the tor2 mutation.
Our analysis suggests that the Arg-2 residue of a-tubulin interacts via hydrogen bonds with the GTPase region of b-tubulin and that the Lys residue found in the tor2 mutant does not form these bonds. In accordance with this prediction, we find that several microtubule dynamic instability parameters are significantly altered in the tor2 background. The tor2 mutation produces right-handed helical growth in various organs and has a major impact on epidermal cell form. We rule out the possibility that division patterns are involved in helical growth by manipulating the direction of organ twisting in hypocotyls that only undergo interphase cell elongation. Furthermore, by presenting examples of helical growth for Arabidopsis cells growing in isolation, we show that torsion is inherent within the single cell and is not necessarily an expression of tissue interactions. We discuss these findings in relation to models for helical cell and organ growth.

RESULTS
Arabidopsis plants with helically twisted leaf petioles were isolated in a screen for mutants impaired in leaf development. In addition to the formerly identified tor1 mutant (Fabri and Schä ffner, 1994;Buschmann et al., 2004), two further complementation groups were found and termed tor2 and tor3. tor2 and tor3 mutants show right-handed and left-handed petiole torsions, respectively. This article focuses on the analysis of the tor2 mutant but employs the tor3 mutant to resolve a specific question.
The tor2 Phenotype Is Caused by a Missense Mutation in a-Tubulin 4 Fine mapping of the semidominant tor2 locus identified its position in the top region of chromosome 1. Sequencing of the a-tubulin 4 candidate gene (At1g04820) in this region revealed a G-to-A transition for the fifth nucleotide of its open reading frame. This mutation results in a conservative exchange of the second amino acid Arg (R) with Lys (K). The R2 residue of a-tubulin is highly conserved and can be found in several hundred GenBank entries for a-tubulin. We next confirmed that the R2K mutation of a-tubulin 4 is responsible for the tor2 phenotype ( Figures 1A and  1B). A T-DNA insertion mutant of the a-tubulin 4 gene was identified in the SALK collection. RT-PCR analysis of a homozygous line confirmed the knockout of a-tubulin 4. However, respective plants did not show any obvious phenotype ( Figure  1C; see Supplemental Figure 1 online). When wild-type or a-tubulin 4 knockout plants were transformed with a construct carrying the entire a-tubulin 4 R2K gene, a typical tor2 phenotype was generated ( Figure 1D). The same plant lines transformed with the wild-type a-tubulin 4 gene did not show any obvious phenotype. These experiments confirm that the tor2 phenotype results from the expression of the dominant-negative R2K mutation of a-tubulin 4.

The R2K Mutation Is Predicted to Interrupt Interactions with Residues of the GTPase Switch of b-Tubulin
To establish the spatial placement and potential interactions of the a-tubulin R2 residue, we employed the published bovine tubulin dimer structure (Lö we et al., 2001). Importantly, all interacting residues are fully conserved in Arabidopsis a-tubulin and b-tubulin proteins when compared with this heterologous model. We were able to localize the a-tubulin R2 residue with high confidence and found that it is likely to form interdimer interactions through hydrogen bonds with Q94 and D74 of b-tubulin (of the following tubulin dimer) ( Figure 1E). To further confirm this interpretation, we then used the corresponding structure factors of the model with the highest available resolution (2.89 Å , PDB-ID: 1TVK) (Nettles et al., 2004). Electron density showed that the positions of the amino acids and their side chains in the structural model are justified and that a-tubulin R2 and b-tubulin Q94, in particular, are localized with high confidence. The region around b-tubulin D74 was less well defined. Both intermolecular contacts of a-tubulin R2 with the b-tubulin subunit could affect the GTP/GDP binding site of b-tubulin, since D74 and Q94 locate to a region in b-tubulin that corresponds to (E) Interactions of wild-type TUA-R2 with TUB-D74 (distance 5.6 Å ), TUB-Q94 (distance 2.6 Å ), and TUA-T130 (distance 4.9 Å ). (F) The mutant isoform TUA-K2 is modeled to interact with TUA-T130 (distance 2.9 Å ) only. The predicted distance of TUA-K2 to TUB-D74 is larger than 10 Å . (G) Overview of this region for wild-type a-tubulin shows the proximity to the GTP nucleotide in TUB. b-Tubulin (TUB), blue backbone; a-tubulin (TUA), green backbone. Small yellow numbers indicate distances along the yellow lines in Å . Images were prepared with coordinates from published structures (PDB-ID: 1TVK) (Nettles et al., 2004). Bars = 1 mm in (A) to (D).
the Switch I and Switch II region of classical GTPases (Lö we et al., 2001). Next, we used the structural model to evaluate the potential effect of the a-tubulin R2K mutation. Based on energetically favorable confirmations from rotamer libraries, we found that K2 can be aligned sufficiently close to a-tubulin T130 (2.9 Å distance) and to a-tubulin E3 (2.5 Å ; see Supplemental Figure 2 online) to allow stable intramolecular interactions ( Figure 1F). By contrast, the intermolecular interactions with Q94 and D74 of b-tubulin appear to be less favored in the a-tubulin mutant R2K. Thus, intramolecular interactions are likely to be strengthened by the R2K mutation, whereas the interdimer contacts are less likely to form and are probably weakened ( Figure 1F). Our results are consistent with the possibility that the a-tubulin R2K mutation alters the GTPase activating function of b-tubulin, thereby affecting microtubule assembly (see overview of this region for the wild type in Figure 1G).

Microtubules in tor2 Show Altered Dynamics
To test the prediction that the a-tubulin 4 R2K mutation affects microtubule dynamic instability, we created 35S pro :EB1-green fluorescent protein (GFP) and 35S pro :GFP-TUB6 microtubule marker lines in tor2 and Landsberg erecta (Ler) backgrounds. Microtubule dynamics were analyzed in 3-d-old plants germinated on Murashige and Skoog (MS)-based agar plates. The rate of microtubule polymerization was reduced in tor2 compared with the Ler reference line. This was true for both EB1-GFP measured in hypocotyls and GFP-TUB6 measured in cotyledon epidermal cells (Figure 2A). We found that the rate of microtubule depolymerization was also reduced in tor2. Plus end catastrophe and lagging end shrinkage were significantly slower in tor2 than in the Ler background ( Figures 2B and 2C). Based on the Student's t test, the difference between tor2 and Ler polymer growth and shrinkage rates were statistically significant for all measurements (Figures 2A to 2C). When compared with Ler, leading ends of tor2 microtubules spent less time polymerizing but spent an increased amount of time in a paused state and in shrinkage ( Figure 2D). Consistently, microtubule dynamicity was significantly reduced in tor2 ( Figure 2E). Catastrophe and rescue frequencies, calculated over the entire lifetime of randomly selected microtubules, were significantly increased in tor2 compared with Ler ( Figure 2E). This was also evident from kymographs, where Ler often showed leading end polymerization throughout the entire period of observation (4 min). By contrast, tor2 microtubules showed a hesitant forward and backward movement ( Figure 2F). Indeed, a difference in Ler and tor2 microtubule behavior can be readily observed in accelerated movies (see Supplemental Movie 1 online). We next asked whether these defects in tor2 microtubule dynamics would impact the number of microtubules or microtubule bundles present in the cortical array. Microtubule density does not, however, seem to be significantly altered in tor2 when compared with Ler (see Supplemental Figure 3 online). Instead, the tor2 defect appears to affect the length of the EB1 comets at the microtubule plus end. Visual comparison suggests only a slight difference between EB1 comet size in tor2 and Ler (see Supplemental Figure 4 online); however, when we quantified the fluorescence distribution along normalized EB1-GFP comets, we found that the tail of tor2 EB1 comets is significantly more fluorescent ( Figure 2G). This suggests that tor2 plus end comets are longer.

The tor2 Mutation Renders Root Elongation Hypersensitive to Microtubule Drugs
We next asked whether tor2 responds differentially to drugs affecting microtubule stability. Root length of mutant and wildtype plants germinated on agar plates containing microtubule drugs was measured ( Figure 3). We found that tor2 root elongation was hypersensitive to the microtubule destabilizing drugs oryzalin and propyzamide. However, elongation of tor2 roots was also slightly hypersensitive to the microtubule stabilizing drug taxol. The more sensitive response to the drug application was statistically significant for at least two concentrations in all cases ( Figure 3). All experiments were repeated and yielded the same trend.

Right-Handed Helical Growth of tor2 May Be Based on a Genetic Pathway Independent from tor1
Young tor2 leaves show straight leaf petioles similar to those of the wild type. However, during expansion growth, petioles start to twist, leading to a right-handed rotation of the lamina with an average speed of 248 per day (calculated for days 13 to 21) ( Figure 4A). To investigate whether tor2 was affected in the same genetic pathways as the right-handed tor1 (spr2) mutant, homozygous double mutants were bred. TOR1 encodes a plantspecific microtubule-associated protein that functions in orienting cortical microtubules (Buschmann et al., 2004;Shoji et al., 2004). The tor1 tor2 double mutant clearly shows an additive leaf twisting phenotype. The twisting of tor1 tor2 double mutant leaves was faster (498 per day) than that of the tor1 and tor2 single mutants, and the final extent of leaf rotation was more pronounced ( Figure 4B). In tor1 tor2 double mutant plants, we frequently observed leaf rotations larger than a full turn, demonstrating the developmental plasticity of the process ( Figure 4C).
Because these results suggested that tor2 constitutes a righthanded helical growth pathway independent from tor1, we analyzed its phenotype in more detail. Apart from leaf petioles, right-handed growth was seen in roots, hypocotyls, stems, and flowering organs of tor2 ( The tor2 mutation has a particularly strong impact on epidermal cell growth. In roots, soon after cell file twisting becomes apparent, epidermal cells start to swell and the final cell shape is often spherical (Figures 4E to 4H). In tor2 hypocotyls, epidermal cells occasionally resume a new axis of growth leading to cone-like protrusions (see Supplemental Figure 6 online). In plants, microtubules have several functions during cell division. To detect a possible defect in chromosome segregation and/or cytokinesis in tor2, we analyzed the DNA content of tor2 mutant cells by flow cytometry. However, flow cytometry of nuclei from rosette leaves and excised shoot meristems did not show a significantly altered pattern of DNA content compared with the wild type (see Supplemental Figure 7 online). Thus, the a-tubulin 4 mutation observed in tor2 has no observable effect on the segregation of chromatin. (A) Microtubule polymerization was assessed by measuring the speed of leading (plus) end growth using GFP markers EB1 and TUB6. Irrespective of the marker or the tissue type, tor2 plus end growth was significantly slower (t test, P < 0.05). (B) Microtubule depolymerization at the plus end was assessed by measuring shortening during catastrophe in cotyledon cells. tor2 microtubules show significantly slower catastrophes (t test, P < 0.05). (C) Additionally, microtubule depolymerization was assessed by measuring the speed of lagging end (minus end) shortening in cotyledon cells. The t test suggested a significantly slower rate of depolymerization (P < 0.05). Each value (6SD) presented in (A) to (C) is based on n $ 40 microtubules. (D) Based on randomly selected microtubules, we measured the life history (percentage of time spent) of the leading end of wild-type and tor2 microtubules in cotyledon cells. tor2 spends more time in pause and shrinkage when compared with the wild type. (E) Based on randomly selected microtubules, we measured microtubule dynamicity and the frequency of catastrophe and rescue. These analyses were performed in cotyledon cells, since they contain less crowded and less bundled microtubule arrays than hypocotyls or roots, allowing far more accurate measurements. Leading end dynamicity (cumulative forward and backward movement of the plus end, unit = mm/min) is dramatically altered in tor2. Catastrophe and rescue frequencies were calculated in relation to total time spent. In tor2, both catastrophe and rescue were significantly more frequent. Ler = 5256 s observation time; tor2 = 6728 s observation time.
(F) Kymographs recording the leading end using GFP-TUB6 in cotyledon cells for 4 min each. Wild-type microtubules (Ler) often showed prolonged periods of polymerization, whereas microtubules in tor2 behaved erratically and frequently switched between pause, polymerization, and depolymerization.
(G) Plus end comets of tor2 dim less quickly than those of the wild type. Fluorescence intensity of comets was measured using images of EB1-GFP in hypocotyl cells. The maximum signal was set to one and given the position zero. n = 25 comets for both Ler and tor2. The fluorescence values at the points between the two arrows are statistically different for Ler and tor2 (P < 0.05).

Helical Organ Growth Induced by Drugs or the tor2 Mutation Are Not Caused by Effects on Cell Division
The Arabidopsis root meristem exhibits a ring of initial cells that generates the epidermis and peripheral root cap (Baum and Rost, 1996). These T-junctions, where the founder cell splits longitudinally to generate two cell files, arise on a rising spiral around the apical meristem. In longitudinal sections of root tips, the staggered position of these T-junctions is indicative of the spiral division pattern by which they were generated. We therefore asked whether this pattern of T-junctions can be identified in root tips of tor2 and whether it is altered in this mutant. Staining of roots with propidium iodide and imaging with a confocal microscope readily revealed the T-junctions (Figures 5A and 5B). In wild-type and tor2 roots, T-junctions were present in a staggered manner consistent with a spiral pattern of cell division in the ring of initials. To assess a possible effect on organ growth, we counted the number of cells between the first and second T-junction for the wild type and tor2, corresponding to the steepness of the helix produced. However, no significant difference between tor2 and the wild type was found (tor2 = 5.3 and wild type = 5.2; n = 13). Although in tor2 columella cells and the quiescent center occasionally appeared slightly misshapen, no obvious effect on tissue patterning was detected.
To conclusively rule out the possibility that aberrant cell division patterns produce helical organ growth, we used Arabidopsis hypocotyls as a model. Postgermination hypocotyl elongation does not involve significant epidermal or cortical cell divisions (Gendreau et al., 1997;Collett et al., 2000). When grown on MS-based media without drugs, light-grown tor2 hypocotyls show obvious right-handed growth and wild-type hypocotyls were straight ( Figure 5C). Propyzamide (3 mM) produced lefthanded helical growth in wild-type hypocotyls, indicating that drug-induced helical growth in wild-type hypocotyls is based on an effect on cell elongation and not cell division. tor2 hypocotyl growth remained right-handed at this propyzamide concentration. At the highest concentration of 9 mM propyzamide, wildtype hypocotyl growth became somewhat erratic, showing straight, left-handed, and right-handed hypocotyls. However, tor2 plants growing on 9 mM propyzamide showed an inversion of hypocotyl growth from a right-handed helix to a left-handed helix ( Figure 5C). This shows that the direction of tor2 hypocotyl growth can be overridden from right-handed to left-handed at a late developmental stage, when organ growth solely depends on cell elongation.

Leaf Trichomes of tor2 Show Right-Handed Twisting and Increased Branching
Since our analyses indicated that helical growth is indeed based on cell elongation, we asked whether the tor2 mutation has an effect on freely growing cells outside of an organ context. Whereas root hairs of tor2 appeared largely normal, leaf trichomes had a striking phenotype: tor2 trichomes were twisted and bent, trichome branches were usually longer than in the wild type, and the average number of branch points was increased ( Figures 6A and 6B, Table 1). When viewed from above, the majority of tor2 trichomes showed branches that bend to the right ( Figure 6B). Interestingly, we found that the direction of trichome twisting was inverted in tor2 tor3 double mutants ( Figures 6B and 6D). tor3 is a novel mutation that produces left-handed helical growth in petioles (see Supplemental Figure 8 online); however, the tor3 single mutant has no obvious trichome Seedlings were grown for 8 d on agar plates supplemented with 0 to 10 mM propyzamide, 0 to 0.50 mM oryzalin, and 0 to 10 mM taxol as indicated (see Methods). Relative lengths of primary roots of the wild type (triangles, dotted lines) and tor2 (rhombs) measured using Image J software (n = 13 to 35, 6SD) are displayed; wild-type control roots were approximately twice as long as tor2 control roots. Statistical tests based on the normally distributed values indicate that hypersensitivity of tor2 is significant in the critical regions of concentration in all three cases as indicated by asterisks (P values < 0.005 and <0.02 for asterisks in brackets). phenotype ( Figure 6C). The above observations were confirmed by quantification using a binocular microscope (Table 1). Under the scanning electron microscope, using the papillae (cuticular wax droplets) as landmarks, branches and stalks of tor2 trichomes showed torsions of consistently right-handed direction, whereas wild-type trichomes were straight only ( Figures 6E to  6H). These results show that the right-handed twisting phenotype of tor2 is expressed in a freely growing cell and does not require tissue-based interactions for it to develop. zwichel mutants have fewer trichome branches than wild-type plants, and this is combined with a moderate branch elongation defect (Hü lskamp et al., 1994;Luo and Oppenheimer, 1999). ZWICHEL encodes a calmodulin-regulated kinesin (Oppenheimer et al., 1997;Reddy and Day, 2000). Transient microtubule stabilization using taxol was reported to increase branch initiation in the zwichel background (Mathur and Chua, 2000). Overexpression of an N-terminally tagged a-tubulin, which produces right-handed organ torsions in Arabidopsis, was reported to mimic the taxol effect on zwichel trichomes (Abe and Hashimoto, 2005). We therefore crossed tor2 into the zwichel background. For comparison, we additionally quantified the effects of taxol on zwichel single mutants. The tor2 zwichel double mutant shows significantly more branch points when compared with the zwichel single mutant (Table 1, Figures 6J and 6K). We further confirmed that taxol is capable of producing a second branch point in zwichel single mutant trichomes, although with a comparatively low efficiency of 5.5%. In the tor2 zwichel double mutant, only a subset of the initiated branches elongated and those that did tended to stay short ( Figures 6J and 6K). This suggests that the branch elongation defect seen in zwichel mutants was enhanced in the tor2 zwichel double mutant background. Indeed, tor2 zwichel double mutant trichomes often showed an enhancement of the right-handed twisting seen in the tor2 background. This was most obvious in unbranched trichomes occasionally seen in tor2 zwichel double mutants ( Figure 6L).

Left-Handed Cortical Microtubule Arrays in Roots and
Trichomes of tor2 EB1-GFP lines backcrossed to tor2 were used to analyze microtubule orientation relative to the cell's axis (Figure 7). In tor2 roots, epidermal cells of the early elongation zone showed left-handed microtubule arrays, and this was seen before the onset of cell file twisting. Wild-type microtubule arrays of a comparable region were transverse ( Figures 7A and 7B). We therefore asked whether freely growing cells like trichomes would also show helical microtubule arrays, focusing on an early stage of trichome development, after the epidermal outgrowth has acquired a cylindrical shape but before branching. Wild-type trichome cells of this stage show cortical microtubule arrays that are denser near the cell's tip compared with the base. However, we routinely observed a spot devoid of microtubules at the apex of the cell. In wild-type trichomes, EB1 comets appeared to encircle the apex, creating a halo of microtubules (see Supplemental Movie 2 online). Interestingly, we found that this cortical array of microtubules was lefthanded helical in tor2 mutants but was mainly transverse in the wild type (Figures 7C and 7D; see Supplemental Movies 2 and 3 online). A quantitative analysis of array orientations based on EB1 movement in movies showed that the majority of tor2 microtubule arrays in unbranched trichomes was shifted to lefthanded positions but was transverse in the wild type ( Figure  7E). This demonstrates that the right-handed twisting seen in adult trichomes of tor2 is preceded by the formation of lefthanded cortical microtubule arrays.

Cell Suspensions Created from tor2 Show the Twisting Phenotype
To validate the finding that the tor2 mutation is capable of producing helical growth in isolated cells, we generated cell suspension cultures from wild-type (Ler) and tor2 roots (three lines each). This allowed us to study cell elongation outside the context of a plant organ. Microscopy observation revealed that wild-type lines predominantly formed elongated, cylindrical cells. By contrast, the elongation growth of tor2 lines was obviously compromised ( Figure 8A). In tor2 lines, we were able to distinguish several characteristic cell-shape classes: elongated cells (data not shown), multipolar cells (top left), rounded cells (bottom left), and helical cells of varying degrees (middle and right). We next quantified the cell shape classes in our suspension lines ( Figure 8B). Whereas elongated shapes were predominant in wild-type cells (>70%), this class was diminished in tor2 at the expense of (in order of frequency) multipolar, round, and helical cells. Helical cells in tor2 accounted for >10% and these were predominantly right handed. Thus, helical growth of individual cells was a general capacity of the tor2 mutant and was not restricted to specialized cell types like trichomes.

Helical Organ Growth Is Based on a Cell Elongation Defect
Two opposing models may explain helical organ growth observed in Arabidopsis mutants and drug-treated plants. One model predicts that twisting is based on division patterns in meristematic tissues. Cell divisions in the ring of initials that creates the lateral root cap and epidermis occur nonsimultaneously. The pattern of divisions follows a spiral sequence that produces a helical pattern of cell junctions in the apical meristem (Baum and Rost, 1996). It is conceivable that this could impact upon the direction of root growth. Indeed, wild-type root growth shows a subtle bias toward the left-hand direction (Simmons et al., 1995). That helical division patterns could be the basis for helical growth in Arabidopsis mutants was also speculated by Wasteneys and Collings (2004). We therefore asked whether the striking twisted growth of Arabidopsis helical growth mutants could be based on helical patterns of division in meristems. We were able to identify the staggered T-junctions in the epidermis of both tor2 and the wild type but could not detect a significant change in the spacing of the T-junctions. Furthermore, we did not observe the patterning defects seen in the tornado mutants (Cnops et al., 2000;Figures 5A and 5B).
To gain positive evidence that helical organ growth is based on cell elongation and not cell division, we used postgermination hypocotyl growth as a model. Because hypocotyl growth is largely independent of cell division, it has been used for investigations on cell elongation (Gendreau et al., 1997;Collett et al., 2000). Using hydroxyurea (Pfosser et al., 2007), we confirmed that suppressing cell division did not abolish hypocotyl growth (data not shown). In this study, we further showed that propyzamide is capable of altering the direction of hypocotyl growth in wild-type and tor2 plants. Taken together, these results suggest that helical growth of wild-type and tor2 plants is not based on helical division patterns.
Hashimoto and colleagues (Furutani et al., 2000;Hashimoto, 2002) presented an alternative model of helical organ growth according to which, "the driving force for helical growth should come from radial swelling of inner cells" and the "elongating epidermis may function passively just to determine the direction of skewing." Furutani et al. (2000) speculate that cortical microtubules of inner cells have a stronger response to mutation or drug treatment, leading to a severe loss of growth anisotropy, whereas outer cells respond by producing helical microtubule arrays. In this model, epidermal cells are assumed to contain oblique microtubule arrays underneath the load-bearing outer cell wall, and this is thought to result in the deposition of oblique cellulose microfibrils. This is assumed to produce polar cell growth in the direction perpendicular to microtubule and microfibril alignment and, therefore, to helical organ growth (Furutani et al., 2000). The hypothesis of Hashimoto and coworkers is attractive because the majority of helical growth mutants show oblique microtubule arrays in tissues prone to undergo twisting (a few noteworthy exceptions may be found in Sugimoto et al., 2003;Sedbrook, 2004;Ishida et al., 2007a). Our results show that this general tendency holds true for right-handed tor2. However, we note that tor2 shows a dramatic loss of anisotropy in epidermal tissues of roots, and the impact on epidermal cell form may be even more pronounced than the effect on cortical cells ( Figure  4G). This contrasts with results obtained for spiral1 and could mean that the model by Hashimoto and coworkers, where epidermal cells are less affected than cortical cells, may not apply to tor2 (Furutani et al., 2000). Since tor2 shows righthanded twisting of cells growing in isolation (Figures 6 and 8), it is tempting to speculate that a somewhat simpler mechanism is responsible for organ twisting in tor2: that is, the simultaneous occurrence of cell twisting in different tissues, in the same direction, may lead to a rotation of the entire organ. Evidently, growing organs are capable of integrating the behavior of independent cells into one coherent growth direction.

Handed Cells Elongate in a Helical Fashion
Whereas twisted trichomes have been observed among the distorted class of Arabidopsis mutants (Feenstra, 1978;Hü lskamp et al., 1994;Szymanski et al., 1999), tor2 is the only mutant reported to show trichome torsions of defined handedness. Although exceptions may eventually be revealed, the identification of genes affected in distorted-type trichome mutants has pointed to actin-dependent processes involved in cell elongation. However, some distorted mutants additionally show defects in microtubular organization (e.g., Saedler et al., 2004;Zhang et al., 2005). Our results on the tor2 mutation of a-tubulin indicate that right-handed trichome twisting is preceded by lefthanded microtubule arrays in early trichomes. In tor2, helical growth is not limited to trichomes because suspension cells generated from tor2 mutants also show right-handed cell torsions. It is possible that a closer inspection of other helical growth mutants will show that trichome twisting is a more general feature of this mutant class. The observation that the tor3 mutation alters  the direction of trichome twisting in tor2 tor3 double mutants (Table 1, Figure 6) seems to support this notion. The fact that Arabidopsis tor2 shows helical growth at the single-cell level suggests a mechanistic similarity to other cell types with helical growth (e.g., the Phycomyces sporangiophore, the charophyte Nitella, and the filamentous algae Chaetomorpha and Cladophora). These organisms have served as early model organisms to test principles of growth anisotropy in freely growing cells (for review, see Roelofsen, 1965;Preston, 1974). Based on a wide range of experimental approaches, including birefringence analyses and electron microscopy of cell walls, it was concluded that helical growth in these freely elongating cells is based on the handedness exhibited by the helical cell wall texture (Roelofsen, 1951;Frei and Preston, 1961;Probine, 1963).

The Relationship between the Helical Sign of Cell Wall Organization and the Direction of Growth
In all cases, helical growth is thought to occur in a direction perpendicular to the mean alignment of wall microfibrils (Roelofsen, 1965;Hashimoto, 2002;Wasteneys and Collings, 2004;Baskin, 2005). Therefore, the angle of helical growth is a direct translation of the angle of microfibrils in the wall, and Roelofsen (1965) has discussed this in terms of actual values. This means that cells with left-handed wrappings of wall microfibrils should show right-handed growth and vice versa (see Figure 9 for explanation). Some support for this is provided by our observation that the predominantly left-handed cortical microtubules of tor2 trichomes are accompanied by the opposite right-handed pattern of cuticular wax papillae (Figures 6 and 7). For modeling purposes, the elongation of cells with helically arranged microfibrils has been compared with pulling out a spring (Roelofsen, 1950;Lloyd and Chan, 2002). This showed that certain springs meet the criteria explained above: that is, a line drawn vertically down the spring rotates with the opposite helical sign when the spring is stretched. Helical lamellae of actual plant cell walls are likely to behave differently in detail (shorter microfibrils, loosening and breakage of bonds, etc.; Figure 9), but such models do explain how the expansion or stretching of helically based material can produce twisting with the opposite helical sign.
Modeling Shows That Interactions of a-Tubulin with the GTPase Region of b-Tubulin Are Interrupted by the R2K Mutation If the helicity of the microtubule array and cell twisting can be considered as expressions of microtubule organization, then what impact does the tor2 amino acid substitution have on this?
We showed that the tor2 phenotype is based on the R2K mutation of a-tubulin 4. The R2 residue of a-tubulin is highly conserved in all eukaryotes. Only a single sequence showing an exchange toward K was found among several hundred GenBank entries. However, the cognate a-tubulin from the planarian Schmidtea polychroa is also rather divergent from a-tubulins in other regions of the polypeptide (Simoncelli et al., 2003). The functional importance of the N terminus of a-tubulin was revealed by Ala scanning of the yeast protein (Richards et al., 2000). In Saccharomyces cerevisiae, the R2A mutation of a-tubulin (together with an E3A mutation in the same molecule) yielded benomyl supersensitivity and cold sensitivity as well as heat sensitivity. The R2K mutation of a-tubulin described in this article leads to right-handed helical growth in Arabidopsis organs and single cells. In a recent publication, Ishida et al. (2007b) presented a set of tubulin mutations leading to helical growth, among those an allele showing the same mutation as tor2. They located the mutations to the tubulin heterodimer of a structural model with a maximum resolution of 3.7 Å (Nogales et al., 1999). At 3.7 Å , for structures derived from two-dimensional crystals, the resolution varies considerably in different orientations so that atomic models leave considerable room for interpretation, particularly for the amino acid side chains. Consistently, Ishida et al. did not provide information on hydrogen bonds formed and gave no mechanistic interpretation concerning the effect of the R2K mutation (Ishida et al., 2007b). By contrast, by choosing a structure with 2.89-Å resolution (Nettles et al., 2004), and by confirming the observed interactions through inspection of the respective electron density region, we are able to predict that the wild-type residue R2 of a-tubulin is engaged in an interdimer interaction with Q94 and D74 of b-tubulin via hydrogen bonds. We next modeled the submolecular consequences of the tor2 amino acid exchange (Figure 1). Results suggest that in the R2K a-tubulin mutant, interdimer interactions are weakened and that intramolecular hydrogen bonds (to T130 and E3 of the same a-tubulin subunit) are favored instead. The predicted a-tubulin R2 binding partners Q94 and D74 of b-tubulin are located in a region that corresponds to the Switch I and II regions of classical GTPases (Lö we et al., 2001). Switch I and II regions of classical GTPases are involved in GTP binding and hydrolysis (Nogales et al., 1998). It is assumed that the Switch I and II regions of b-tubulin are responsible for the rate of GTP hydrolysis and that this impacts on microtubule dynamic instability (Nogales et al., 1999). Interestingly, in the collection of The models link the alignment of structural features with the direction of cell twisting. Arrow on the left indicates the initial direction of expansion. The spring model (center) emphasizes that stretching a left-handed spring can lead to a right-handed displacement of surface landmarks. The model on the right stresses that the load-bearing microfibrils are not interconnected and may reorient during growth. The direction of growth is perpendicular to the mean orientation of microfibrils. This leads to a rotation of the cell's tip relative to its base. tubulin mutants presented by Ishida et al. (2007b), two b-tubulin mutants, S95F and G96D, are reported to show right-handed growth. Indeed, both residues are part of the Switch II region and are located next to Q94, which our models predict to interact with the R2 residue of a-tubulin (Figure 1). This suggests that the region Q94 to G96 of b-tubulin is a hotspot for mutations leading to right-handed growth and supports the idea that this region is important for GTP hydrolysis and normal dynamic instability. Taken together, the results suggest that the R2K mutation exerts its pivotal effect by changing the rate of GTP hydrolysis and thereby microtubule dynamic instability.
It is interesting that the R2K mutation has only been described for the a-tubulin 4 isotype but not for the other a-tubulin genes present in the Arabidopsis genome. Although this might indicate a specific function for a-tubulin 4, it seems likely that our screen and the screens by others (Thitamadee et al., 2002;Ishida et al., 2007b) were not at a saturating level to reveal this specific mutation in other a-tubulin isotypes.

The R2K Mutation and Its Effect on Microtubule Dynamics
Because the tor2 mutation was predicted to impact on microtubule dynamic instability, we analyzed the effect of microtubule drugs on tor2 root elongation. We found that tor2 root elongation was hypersensitive to depolymerizing drugs oryzalin and propyzamide. This suggests that microtubules in tor2 are destabilized. However, we found that tor2 root elongation was also hypersensitive to the microtubule-stabilizing drug taxol (Figure 3). This argues that the tor2 defect is not caused by microtubule stabilization or destabilization per se but possibly through a more subtle effect on microtubule behavior. We next asked whether tor2 is hypersensitive to cold. The effect of cold was studied during hypocotyl etiolation growth, as the phenotypically similar spr1/sku6 mutants respond strong to this procedure (Furutani et al., 2000;Sedbrook et al., 2004). However, we could not observe an obvious enhancement of the tor2 phenotype in the cold (data not shown). Right-handed growth similar to tor2 has been observed in plants expressing N-terminally tagged versions of a-tubulin and in an a-tubulin 5 mutant carrying a point mutation and was speculated to be the result of microtubule stabilization. Measuring several parameters of microtubule dynamics prompted those authors to speculate that microtubules in these right-handed mutants are stabilized. It appears that microtubule stability based on drug application was not quantitatively assessed in the case of these mutants (Abe and Hashimoto, 2005;Ishida et al., 2007b). In this regard, it is interesting that the application of the microtubule-stabilizing drug taxol does not produce right-handed, but left-handed growth (Furutani et al., 2000). Taken together, the results for tor2 do not strictly support a role for microtubule stabilization in contrast with destabilization; rather, they indicate that microtubule stabilization itself is not the driving force behind righthanded helical growth of tor2.
To gain better insight into tor2 microtubule dynamics, we used GFP-tagged EB1 and TUB6 markers and investigated microtubule behavior in detail. This showed that not only was the rate of microtubule polymerization reduced in tor2, but the rate of microtubule depolymerization also. As a result, microtubule dynamicity is diminished in tor2. Microtubule dynamicity is a measure for the total amount of polymer gained or lost per time. The fact that tor2 microtubules spent more time in pause would decrease dynamicity further (Figure 2). It is interesting that we find both catastrophe and rescue frequencies increased in tor2, and it is possible that this functionally mirrors the increased sensitivity to stabilizing and destabilizing microtubule drugs ( Figure 3). Next, we measured the density of microtubules at the cell's cortex but could not find significantly more microtubules or microtubule bundles in cells of tor2 (see Supplemental Figure 3 online). This differs from what was obtained for an a-tubulin 5 point mutation (Ishida et al., 2007b) and adds support to the idea that microtubules are not simply stabilized in tor2. Since EB1 has been suggested to have an affinity for the GTP cap of a microtubule (Tirnauer et al., 2002), which is a measure of continuing microtubule growth, we examined the length of the EB1 comets. Results confirm that EB1 plus end comets in tor2 plants are larger than those in the wild type ( Figure 2G). In conclusion, and as suggested from our structural analysis, the tor2 mutation of a-tubulin has a strong impact on microtubule dynamic instability, and this is likely the result of interfering with the GTPase function of b-tubulin. The observed reduction in dynamicity may reduce the capacity of the array to reorient in response to external or internal cues.

Putative Mechanisms of Helical Growth
There is a long history of helically based cell walls producing the twisting of filamentous algae (Frei and Preston, 1961;Probine, 1963;Roelofsen, 1965), and this study suggests that the growth pattern of higher plants may be traced back to the behavior of their isolated cells (as seen in trichome cells of tor2). As left-handed microtubule arrays are seen very early during trichome development, we suggest that this is the reason for right-handed cell features of mature tor2 trichomes (Figures 6 and 7). The question then is how are these helical arrays established in the first place?
Plant microtubule arrays are thought to be constructed from the way that treadmilling microtubules (Shaw et al., 2003) interact. Collisions at shallow angles are thought to lead to parallelization of growth, while steeper collisions lead either to catastrophic depolymerization (Dixit and Cyr, 2004) or crossover (Shaw et al., 2003;Wightman and Turner, 2007). Microtubules are born either at random sites at the cortex (Shaw et al., 2003) or branch from foci upon mother microtubules (Murata et al., 2005). Although no obvious asymmetries have so far been described at this level of plant microtubule behavior, at a larger scale microtubules have been reported to produce uniform tissue fields with a common helical sign in wild-type Arabidopsis and maize (Zea mays; Liang et al., 1996). It seems possible that helical growth as described for tor2 has its origin in such naturally occurring microtubule array asymmetries but under circumstances where growth and microtubule reorientation become desynchronized. In roots, for example, helical arrays develop only at the end of the cell elongation phase (Liang et al., 1996;Sugimoto et al., 2000). These handed arrays may have little effect on wild-type organ shape (however, see Simmons et al., 1995), but it is conceivable that in, for example, left-handed helical growth mutants, this reorientation to right-handed helical arrays occurs prematurely, thereby causing helical organ growth. Likewise, it seems possible that a change in microtubule dynamicity observed in tor2 impacts on microtubule reorientations, desynchronizing growth and microtubule positioning.
In conclusion, the results presented here show that the R2K mutation of a-tubulin leads to right-handed helical growth of organs and isolated cells. Our high-resolution analysis of the tubulin structure shows that the R2 residue of a-tubulin interacts through hydrogen bonds with the GTPase Switch regions of b-tubulin and that these interactions are not formed by the mutant tubulin. These results suggest that the tor2 phenotype is likely to result from an effect on microtubule dynamic instability. Indeed, testing several parameters of microtubule dynamics in living Arabidopsis cells clearly reveals an effect on microtubule behavior. We next used the tor2 mutant to investigate the developmental origin of helical growth. Because the application of the antimicrotubule drug propyzamide allows the manipulation of the twisting direction in nondividing hypocotyls, we reject the hypothesis that organ twisting is based on cell division patterns. Moreover, tor2 is the first Arabidopsis mutant reported to show helical growth in isolated cells. We show that mature trichomes of tor2 have a consistent right-handed twist, and this is preceded by the formation of left-handed cortical microtubule arrays in developing trichomes. According to these observations, we hypothesize that in tor2, the twisting of individual cells directly translates into the higher-order twisting of organs.

Mutant Strains and Mutation Mapping
Mutants displaying the tor phenotype were identified in an ethane methylsulfonate-mutagenized Arabidopsis thaliana Ler population. Allelism tests revealed several complementation groups. The tor2 mutation is inherited as a semidominant trait. Additional genetic markers for chromosome 1 facilitated the identification of tor2. New markers were developed using the method of DNA duplex analysis (Hauser et al., 1998).
Double mutants with tor2 were identified in F2 populations generated by crossings to the respective mutant lines. The double homozygous situation was verified by PCR (flanking markers, deletions) or by outcrossing into the wild type (zwichel 9311-11 mutant). The zwichel 9311-11 mutant was kindly provided by David Oppenheimer. A homozygous knockout a-tubulin 4 mutant (SALK_080530) was identified in the SALK collection (Alonso et al., 2003), and the absence of the corresponding transcript in the mutant was confirmed by RT-PCR in comparison to PCR from wild-type cDNA (see Supplemental Figure 1 online).
Transgenic Ler lines expressing EB1-GFP were generated by transforming Ler wild-type plants with a 35S pro :EB1-GFP construct (Chan et al., 2003). Obtained lines were crossed to the homozygous tor2 mutant and control, and tor2 lines were bred from progeny of such crosses. Transgenic lines expressing a GFP-TUB6 fusion were obtained by transforming tor2 and Ler plants with pBGW-35S pro :sGFP-TUB6. For this construct, three fragments encompassing the cauliflower mosaic virus 35S pro :synthetic GFP sequence (Chiu et al., 1996), the Arabidopsis TUB6 open reading frame, and the nopaline synthase terminator were fused by a PCR strategy and subsequently recombined into pBGW (Karimi et al., 2002). The construct allows the constitutive expression of sGFP with C-terminally linked full-length TUB6. The linker between sGFP and TUB6 is ala-gly-ala-gly. All plant transformations were performed by floral dip (Clough and Bent, 1998).

Structural Analyses
Structural models and structure factors were downloaded from Protein Data Bank (www.rcsb.org). All detailed analyses were based on structural information with the highest available resolution (Nettles et al., 2004; PDB-ID: 1TVK; resolution of 2.89 Å ). We used the program Coot (Emsley and Cowtan, 2004) to perform symmetry operations for the generation of complete structural models. Structure factors were converted into MTZ files using the CCP4 program suite (Collaborative Computational Project, 1994) and analyzed with Coot. The R2K mutation was generated in Coot and optimized using standard rotamer library. Images were generated with Pymol (DeLano Scientific).

Drug Treatments
For the analysis of root elongation and hypocotyl twisting, plants were germinated on MS-based media containing the indicated drug concentration. Stocks of microtubule drugs (Sigma-Aldrich) were prepared in DMSO. Experiments to generate taxol-induced trichome branching were based on a procedure described elsewhere (Mathur and Chua, 2000).

Cell Suspensions
tor2 and Ler callus cultures were generated from root cuttings by placing them on agar medium as described earlier (May and Leaver, 1993). After sufficient callus growth (2 to 3 weeks) cells were transferred to fresh plates and eventually into liquid MS medium supplemented with 2,4-D. Obtained suspension lines were subbed weekly.

Confocal Microscopy, Analysis of Image Data, and Microtubule Dynamics
Propidium iodide-based microscopy of roots was performed on living specimens using the Leica SP2 confocal microscope. Sufficient staining was achieved using 100 mg/mL propidium iodide (Sigma-Aldrich) in diluted MS medium for 30 min. GFP-based microscopy of trichomes and roots and for the analysis of microtubule dynamics was performed using the Visitech spinning disc confocal microscope using a 60-fold immersion objective. Data were collected using Metamorph software. Movies were assembled and analyzed with the help of the ImageJ software (W. Rasband, National Institutes of Health).
Movies for the analysis of microtubule dynamics were based on 61 projections from four z-sections acquired every 4 s (4 min total length; measured at 248C). As a rule of thumb, microtubules were randomly selected in stills, and their behavior was then followed in the respective movies. Rate of polymerization and depolymerization was measured from kymographs. This involved the plugin Multiple Kymograph and the macro Read Velocities from Tsp (J. Rietdorf, EMBL, Heidelberg, Germany). EB1 comet profiles were measured using the Plot Profile function of ImageJ. Catastrophe and rescue frequencies were calculated in relation to total time spent because catastrophe and rescue were often observed to occur out of the paused state (especially in tor2). This was based on randomly selected microtubules regardless of their previous life history. This seemed appropriate since in plant cortical arrays the origin of microtubules is often beyond the area of observation and polymerization can occur for longer than the time frame of observation (see Figure 2F). These microtubules would otherwise escape the measurement. For a somewhat different approximation, see Cassimeris et al. (1988). For statistical analyses, data were checked for normality and F-tests and t tests were performed using Microsoft Excel.

Scanning Electron Microscopy
Leaf samples (fourth or fifth true leaf) were mounted on an aluminum stub using O.C.T. compound (BDH Laboratory Supplies). The stub was then immediately plunged into liquid nitrogen slush to cryopreserve the material. The sample was transferred onto the cryostage of an ALTO 2500 cryo-transfer system (Gatan) attached to a Zeiss Supra 55 VP FEG scanning electron microscope. Sublimation of surface frost was performed at 2958C for 3 min before sputter coating the sample with platinum for 2 min at 10 mA, at temperatures below 21108C. After sputter coating, the sample was moved onto the cryostage in the main chamber of the microscope and held at approximately 21308C. The sample was imaged at 3 kV, and digital TIFF files were stored.

Accession Numbers
Sequence data from this article can be found in the Arabidopsis Genome Initiative or GenBank/EMBL databases under the following accession numbers: TUB9 (At4g20890), EB1 (At3g47690), and sGFP (EF090408). tor2 mutants carry a point mutation in the a-tubulin 4 gene of Arabidopsis (Arabidopsis Genome Initiative number At1g04820).

Supplemental Data
The following materials are available in the online version of this article.