The rotation of cellulose synthase trajectories is microtubule dependent and influences the texture of epidermal cell walls in Arabidopsis hypocotyls

In 1974, Heath hypothesised that cortical microtubules in higher plants provide a spatial template for cellulose-synthesising particles that glide along the plasma membrane extruding cellulose microfibrils into the wall (Heath, 1974). This was recently substantiated by Paredez and colleagues (Paredez et al., 2006), who made 10- to 20-minute movies showing fluorescently tagged CESA6 particles moving along underlying microtubules. To explain the higher-order question of how successive layers of cellulose microfibrils build up the texture of the cell wall, it is necessary to see how this short-term relationship develops over time.

Early ideas, derived from filamentous algae, suggested that transverse cellulose microfibrils in the innermost part of the plant cell wall ‘hoop-reinforce’ the cell, encouraging anisotropic expansion at right angles to the hoops (Green, 1962). Later, however, Green and colleagues (Lang et al., 1982) reported that pea epidermal cells lay down about a half of their microfibrils longitudinally. The texture was not, therefore, composed exclusively of transverse hoops and many other studies have shown that the situation in multicellular higher plant organs is more complex than in freely growing algal filaments one cell thick. For example, in his freeze-fracture studies of elongating parenchyma cells, Itoh (Itoh, 1975; Itoh, 1976) found that layers of longitudinal microfibrils occurred at the cytoplasmic face of the cell wall – a finding at odds with the idea that the innermost microfibrils must be transverse to provide hoop-reinforcement against lateral expansion. These longitudinal cellulose microfibrils alternated with transverse microfibrils throughout the thickness of the wall. In epidermal cells, microfibrils are also deposited in alternating transverse and longitudinal alignments (Chafe and Wardrop, 1972). This structure was observed in xylem parenchyma and the term ‘crossed-polylamellate’ was proposed for the plant cell wall in general (Chafe and Chauret, 1974). In epidermal cells of mung bean hypocotyls, cellulose microfibrils have been reported to be laid down in strata of alternating ‘criss-cross’ alignment separated by fibrils intermediate between these orthogonal directions (Roland et al., 1977) – a pattern also reported in epidermal cells of oat (Sargent, 1978) and sunflower (Hodick and Kutschera, 1992). It has been reported to take about 3 hours for microfibrils to rotate through 180° (Vian and Roland, 1987). The progressive change in alignment between wall layers observed in chemically extracted and silver-stained mung bean stems by transmission electron microscopy has also been seen in freeze-fracture studies of the same tissue (Satiat-Jeunemaitre et al., 1992). Treatment of mung bean (Vian et al., 1982; Satiat-Jeunemaitre, 1984) and azuki bean (Takeda and Shibaoka, 1981) shoots with the anti-microtubule agent colchicine suppressed the normal shifting pattern of the polylamellate outer epidermal wall in favour of a thick invariant layer adjacent to the cytoplasm, which suggests that there is a connection between cortical microtubules and wall lamellation.

When they were first discovered, cortical microtubules were described as occurring in transverse, hoop-like arrays in elongating cells (Ledbetter and Porter, 1963). However, numerous subsequent studies have demonstrated that, in stem cells, microtubules adopt longitudinal and oblique as well as transverse orientations (Iwata and Hogetsu, 1988; Flanders et al., 1989; Ishida and Katsumi, 1992; Zandomeni and Schopfer, 1993; Duckett and Lloyd, 1994; Nick et al., 1994; Yuan et al., 1994; Mayumi and Shibaoka, 1996; Takesue and Shibaoka, 1999; Hejnowicz et al., 2000; Paolillo, 2000; Sawano et al., 2000; Furutani et al., 2000; Buschmann et al., 2004; Le et al., 2005; Ishida et al., 2007; Chan et al., 2007). Many of these investigations (e.g. Paolillo, 2000; Sawano et al., 2000; Buschmann et al., 2004; Le et al., 2005) emphasise the point that all three microtubule orientations can be found while the tissue is still actively elongating. As an explanation for this diversity, studies have shown that microtubules actively cycle between these different orientations in stem epidermal cells. For example, in pea epidermal cells labelled with microinjected fluorescent tubulin, microtubules could be seen to reorient from transverse to longitudinal, via an intermediate discordant stage (Yuan et al., 1994); when gibberellic acid was added, longitudinal arrays reoriented back to transverse (Lloyd et al., 1996). This ability to reorient was considered to be sensitive to a range of factors, including light and hormones (Lloyd, 1994). Shibaoka and colleagues extended this notion by proposing that microtubules not only reorient in response to experimental intervention, but also undergo endogenous oscillatory cycles (Mayumi and Shibaoka, 1996; Takesue and Shibaoka, 1998). According to this idea, microtubules reorient from transverse to longitudinal then cycle back again in about 6 hours, in agreement with the time it took for cellulose microfibrils to go through a similar cycle in the polylamellate outer epidermal wall of azuki bean epicotyls (Takeda and Shibaoka, 1981). This kind of repetitive or cyclic behaviour appears to be confined to shoots, because it has not been reported in roots, which seem to display only a one-way reorientation from transverse to longitudinal as the growth rate declines (Baskin et al., 1999; Sugimoto et al., 2000; Granger and Cyr, 2001).

Based on his work with aldehyde-fixed sunflower epidermal cells, in which he also saw variably aligned microtubules during cell elongation, Hejnowicz (Hejnowicz, 2005) then proposed that microtubules underwent not just oscillatory cycles, flipping between transverse and longitudinal, but full rotary cycles in defined clockwise or anticlockwise directions. Arabidopsis seedlings expressing AtEB1a-GFP revealed microtubules rotating in clockwise or anticlockwise directions in living epidermal cells (Chan et al., 2007). Although individual microtubules tended to treadmill in linear or curvilinear paths, the more durable bundles or tracks along which microtubules move were themselves seen to move more slowly in the plus-end direction of their component microtubules and over the longer term, to veer sideways, resulting in rotary movements. It was hypothesised that this could help explain the crossed-polylamellate texture of stem epidermal cell walls. Here, we tested that hypothesis. We show that cellulose synthase particles also undergo full rotations over the several hours required for any rotary or oscillatory model for cell wall alignment (Takeda and Shibaoka, 1981; Vian and Roland, 1987). Over this time period, both polymer-promoting taxol and microtubule-destabilising oryzalin stopped the rotation of the cellulose synthase trajectories without stalling their rate of movement. This cessation of rotation resulted in the build-up of an abnormally thick, invariant inner-wall layer. We conclude that microtubule rotation is the basis of cellulose synthase track rotation and that these movements contribute to the variable texture of the polylamellate cell wall.

To study the movement of the cellulose synthase GFP-CESA3, over the several hours required to deposit several wall layers (Takeda and Shibaoka, 1981; Vian and Roland, 1987), it was first necessary to calibrate the long-term behaviour of microtubules in cells grown in illuminated microscopy chambers (Chan et al., 2007). Microtubule rotation was studied by making z-stacks every 15 minutes for up to 24 hours in seedlings that expressed the tubulin marker TUA6-GFP.

Rotation of microtubules was seen at days 2 and 3, when cells were elongating at an average rate of 1.2 μm/hour (n=23, 9 hypocotyls, s.d.=0.6). In stills taken from movies and arranged across the page, these rotations could be seen to trace arcing patterns. Fig. 1 illustrates one kind of rotation where the tracks continuously reorient first in a clockwise direction then anticlockwise. The time taken to rotate through 180 degrees was variable (average=332 minutes, n=23, s.d.=133). In addition to bow-shaped arcs produced by continuous rotation, and reversed arcs produced by changing the direction of rotation, semi-arcs were produced by discontinuous reorientation, i.e. when microtubules ‘jump’ between different alignments as predicted by Hejnowicz (Hejnowicz, 2005) and demonstrated by Chan and co-workers (Chan et al., 2007). Rotations are therefore typically variable within a cell and need not be synchronised between neighbouring cells (Chan et al., 2007).

Next, the behaviour of cellulose synthase tracks was investigated at the stage when microtubules were rotating. This was performed using spinning-disk confocal microscopy on seedlings expressing GFP-CESA3 under the control of its own promoter in the mutant background cesa3^je5 (Desprez et al., 2007). GFP-CESA3 labelled similar compartments in light-grown hypocotyl cells, as reported for dark-grown cells: i.e. cellulose-synthesising particles at the cortex and relatively fast-moving Golgi bodies in the cytoplasm (Desprez et al., 2007). At the outer epidermal wall, particles were aligned in rows to form tracks, as previously described for CESA6 (Paredez et al., 2006) and CESA3 (Desprez et al., 2007). Here, we investigated how the CESA tracks behaved over the longer time established for microtubules to undergo rotation. By limiting the z-sectioning range, it was possible to collect movies of CESA tracks over several hours (see supplementary material Movies 1 and 2, which are best viewed looped). In supplementary material Movie 1, the CESA tracks in the upper cell can be seen to rotate anticlockwise (asterisk 1). The stills from this movie are shown in Fig. 2. Rotations could be clockwise or anticlockwise. supplementary material Movie 2 shows CESA trajectories rotating fully 360° anticlockwise in just over 4 hours.

By cutting strips from cells at successive time intervals the resulting kymographs (Fig. 3A,B) illustrate the arcing patterns followed by these CESA trajectories. It took 185 minutes (mean of 10 cells, s.d.=77) for these trajectories to describe 180° rotations. Rotation of CESA tracks did not occur synchronously across the tissue and the time taken to rotate clockwise or anticlockwise was variable, as reported for rotation of microtubule tracks (Chan et al., 2007).

To functionally test the role of microtubules in the rotation of CESA3-GFP tracks, we examined the consequence of adding the microtubule-stabilising drug, taxol. Notably, taxol did not inhibit the linear velocity of GFP-CESA3 particles [taxol, 259±56.4 nm/minute (mean ± s.d.); control, 285±34.1 nm/minute; n=500; five hypocotyls per treatment, 25 cells]. After several hours, taxol caused the cells to swell radially, but although the cells remained elongate, the drug could be seen to suppress the longer-term rotary movement of CESA tracks. Fig. 4 illustrates CESA tracks maintaining a right-hand oblique alignment over a 6 hour period, without undergoing rotation.

Microtubules in Arabidopsis hypocotyl cells are stabilised by addition of taxol (Chan et al., 2007), so to test the effect of microtubule destabilisation, seedlings expressing GFP-CESA3 were treated with 200 nM oryzalin. Fig. 5 shows that, as for taxol (Fig. 4), the CESA trajectories did not undergo rotation and remained in a net oblique alignment over a 6 hour period, before the cells began to swell significantly. Similarly to taxol, oryzalin had no significant effect on the linear velocity of GFP-CESA3 particles [oryzalin, 278±37.5 nm/minute (mean±s.d.); control, 285±34.1 nm/minute; n=500]. Kymographs of the CESA trajectories in taxol, oryzalin and control cells are illustrated in supplementary material Fig. S1.

Next, electron microscopic sections of the outer epidermal wall were examined to determine whether the taxol-induced suppression of CESA rotation affected the wall texture (Fig. 6). When stained by the silver periodate method (Thiéry, 1967), the polysaccharides in the outer epidermal walls of Arabidopsis hypocotyls appeared as fibrils of varying length. In controls (Fig. 6A), the inner portion of the wall showed darker bands of transverse fibrils parallel to the plasma membrane interspersed with smaller lines and dots representing fragments of longitudinal or oblique fibrils running in other directions. Refrégier and colleagues (Refrégier et al., 2004) saw fibrils parallel to the membrane in both transverse and longitudinal sections, establishing that the outer epidermal wall of dark-grown Arabidopsis hypocotyls is polylamellate. We also sectioned hypocotyls transversely and longitudinally, and in both cases saw bands of fibrils parallel to the plasma membrane (therefore not shown), confirming the coexistence of longitudinal and transverse fibrils in the light-grown outer epidermal wall. In the presence of taxol (Fig. 6B), or the anti-microtubule herbicide oryzalin (Fig. 6C), the fibrils formed a characteristically thick, monotonous inner layer that was about half the width of the wall. In this monotonous layer, fibrils appeared to run more or less perpendicularly away from the plasma membrane. However, these sections were not cut at right angles to the hypocotyls, but at 45°, to increase the travel through the thickness of the wall and see its structure more clearly (Chafe and Doohan, 1972; Hodick and Kutschera, 1992). As explained previously (Preston, 1982), tilting of sections (or, as here, cutting at an angle) causes neighbouring fibrils to appear to overlap. In this case, layers of longitudinal fibrils parallel to the plasma membrane, and enhanced by silver proteinate, appeared to tilt away from the plasma membrane. Such thick monotonous layers were not seen in similar sections from wild-type controls.

In roots, microtubules tend to be transverse to the growth axis in the zone of most rapid elongation, reorienting through 90° to longitudinal as the growth rate declines (Liang et al., 1996; Baskin et al., 1999; Granger and Cyr, 2001; Sugimoto, 2000). However, in shoots, the picture is more complex. Although the elongation zone is reasonably well defined in roots, rapid elongation growth in light-grown shoots can be distributed ‘almost homogeneously over the length of the organ’ (maize coleoptile) (Zandomeni and Schopfer, 1993) and this dispersion of the growth zone has been confirmed for the light-grown Arabidopsis hypocotyl (Gendreau et al., 1997). In contrast to the zonal alignment of microtubules in the Arabidopsis root (Baskin et al., 1999), microtubule alignment in light-grown Arabidopsis hypocotyls becomes highly variable, with transverse, oblique and longitudinal arrays occurring simultaneously (Le et al., 2005). This variability has been particularly noted by the Shibaoka laboratory (Sakiyama and Shibaoka, 1990; Sakiyama-Sogo and Shibaoka, 1993; Kaneta et al., 1993; Mayumi and Shibaoka, 1995; Mayumi and Shibaoka, 1996), who hypothesised that microtubules in stem epidermal cells undergo endogenous cycles of reorientation (Mayumi and Shibaoka, 1996; Takesue and Shibaoka, 1998). Hejnowicz (Hejnowicz, 2005) suggested that such cycles would be rotary rather than oscillatory, and rotary cycles have now been seen in Arabidopsis hypocotyl epidermal cells (Chan et al., 2007). In this paper, we have investigated the effect of microtubule rotation on the rotation of cellulose synthase tracks and the resulting effect on cell wall texture.

Cortical microtubules translocate by treadmilling (Shaw et al., 2003) and over longer time periods, the tracks along which they move undergo slow rotary movements that are blocked by adding taxol (Chan et al., 2007). Here, we found that the tracks formed by GFP-CESA3 particles also describe clockwise or anticlockwise cycles that are blocked by addition of taxol. Microtubules were still present under this treatment, except neither their tracks nor the CESA trajectories rotated over the longer term. Lack of rotation did not affect the rate at which the synthases moved along the plasma membrane, illustrating that microtubule turnover is not required for the linear movement of CESA3. However, the failure of CESA3 tracks to rotate in taxol-treated cells indicates that microtubule dynamics are normally required for the tracks to move out of their fixed alignment and to undergo rotary movement. The same effect was obtained with 200 nM oryzalin: CESA particles continued to move unabated in the presence of the drug, but the longer-term rotation of their trajectories was prevented. This blocking of CESA track rotation by oryzalin or taxol resulted in the production of a characteristically monotonous, thick band of fibrils at the cytoplasmic face of the outer epidermal cell wall, showing that perturbation of microtubule dynamics affects wall texture. Similar effects of anti-microtubule drugs on wall texture have been reported previously. In mung bean hypocotyls (Vian et al., 1982) and maize coleoptiles (Satiat-Jeunemaitre, 1984), the addition of colchicine inhibits the regular layering normally seen in the inner portion of the outer epidermal cell wall. Instead, a new, thick, monotonous layer is deposited at the cytoplasmic face that was not seen in controls. Colchicine has also been reported to perturb the crossed-polylamellate structure of the outer epidermal wall in azuki beans, resulting in a thick inner band in which microfibrils all run in the same direction (Takeda and Shibaoka, 1981). The production of a thick, monotonous inner layer in the presence of anti-microtubule drugs recalls early studies (Quader et al., 1978) on the alga Oocystis. Its wall normally consists of alternate layers of left-handed and right-handed oblique cellulose microfibrils. In the presence of anti-microtubule drugs, this regular alternation is blocked, building up a thick invariant ‘sandwich layer’ – the switching of alignment only resuming upon removal of the drug.

These experiments demonstrate that GFP-CESA3 was evidently still delivered to the plasma membrane in the presence of taxol or oryzalin, the linear velocity of CESA particle movement along the plasma membrane was unaffected and wall continued to be made. The main effect of these drugs is not, therefore, to stop wall formation but to affect its texture. According to self-assembly mechanisms for wall formation (Neville et al., 1976; Vian and Roland, 1987), each layer of microfibrils is aligned relative to its predecessor, without the need for a cytoplasmic template, as occurs in cholesteric liquid crystals. However, our studies demonstrate that the linear translocation of synthase particles is unable to be converted into rotary movement in the presence of microtubule-directed drugs, arguing against the idea that the drive for rotation originates from the wall. Instead, this shows that the normal polylamellate texture of the outer epidermal cell wall not only requires the presence of microtubules but for those microtubules also to remain dynamic and capable of reorientation. Although these findings do not rule out the possibility that biophysically driven realignment of microfibrils might subsequently change the lamellation pattern during the course of normal cell expansion, they do indicate that the initial wall texture is heavily influenced by the self-organising properties of microtubules.

Arabidopsis thaliana plants expressing ³⁵S::GFP:TUA6 and CESA3::CESA3:GFP are described (Chan et al., 2007). Cold-treated seeds were sown on filter paper soaked with distilled water, placed in microscopy chambers and, following a 2 day cold treatment at 4°C, the chambers were transferred to the growth room at 22°C under continuous lighting. Chambers were then imaged on day 2 and day 3, after transfer to 22°C. 10 mM stock solutions of taxol and of oryzalin were prepared in DMSO (Sigma) and working-strength solutions were prepared by dilution in distilled water. The chamber was opened, the 2-day-old seedlings washed three times with 4 μM final concentration taxol (Sigma) or 200 nM oryzalin. After re-sealing the chamber, the hypocotyls were imaged.

The middle regions of hypocotyls were imaged using either a VisiTech (UK) spinning disc confocal microscope or a Leica SP2 confocal laser scanning microscope using a ×40/1.3 NA oil objective lens. GFP was excited using the 488 nm line of an argon ion laser and emitted light filtered through a 500–550 nm band-pass filter. For spinning-disk microscopy, fluorescence was detected using a Hamamatsu Orca ER cooled CCD camera with 1× binning, set at 0.7–1.0 seconds (to visualise GFP-tubulin) and 2–3 second (GFP-CESA3) exposure time. Time-lapse images were acquired with a time delay of 15–30 minutes. Cell measurements and kymographs were constructed using the re-slice tool of ImageJ (http://reb.info.nih.gov/ij/). Long-term movies of growing hypocotyls were aligned using Stackreg (http://bigwww.epfl.ch/thevenaz/stackreg/) plug-ins of ImageJ. Movies were edited using the brightness and contrast tool of ImageJ and the Time stamper plug-in (http://rsb.info.nih.gov/ij/plugins/stamper.html).

Seedlings were fixed in 2.5 % (v/v) glutaraldehyde in 0.05 M sodium cacodylate, pH 7.3, for 16 hours followed by three washes in 0.05 M cacodylate buffer and post-fixation in 1% (v/v) OsO₄ in 0.05 M sodium cacodylate for 1 hour at room temperature. After three washes in distilled water, samples were extracted for 18 hours in 40% methylamine (Sigma) under continuous gentle agitation, and finally rinsed in water. Samples were then dehydrated through a graded ethanol series and embedded in LR White resin according to the manufacturer's instructions (London Resin Company, Reading, UK).

The material was sectioned with a diamond knife using a Leica UC6 ultramicrotome (Leica, Milton Keynes, UK) at a 45° angle to the direction of growth, within the elongation zone of the hypocotyl. Ultrathin sections of ~90 nm were picked up on 200 mesh gold grids that had been coated with pyroxylin and carbon. Sections were stained with periodic acid, thiocarbohydrazide, silver proteinate (PATAg) for visualisation of the wall polysaccharides according to Thiéry (Thiéry, 1967): 1% (v/v) periodic acid, 20 minutes; distilled water, 10 seconds (three times), then 10 minutes (three times); 0.2% (w/v) thiocarbohydrazide in 20% (v/v) acetic acid overnight; 10% acetic acid, 10 seconds (three times), then 10 minutes (three times); 5% (v/v) acetic acid, 10 seconds; 2.5% acetic acid, 10 seconds; distilled water 10 seconds (three times), then 10 min (three times); 1% (w/v) silver proteinate (TAAB Laboratories, Aldermaston, UK) in distilled water, 30 minutes in the dark; distilled water 10 seconds (three times), then 10 minutes (three times). The grids were examined in a FEI Tecnai 20 transmission electron microscope (FEI, Eindhoven, The Netherlands) at 200 kV and imaged using an AMT digital camera (Deben, Bury St Edmunds, UK).

This work was supported by a grant-in-aid by the BBSRC to the John Innes Centre. We thank the Gatsby Foundation for financing M.E. E.C. was supported by the National Agency for Research Project ‘IMACEL’ ANR-06-BLAN-0262. We also thank the European Union Framework Program 6 (FP6) ‘CASPIC’ NSET-CT-2004-028974, the National Agency for Research Project ‘Wall Integrity’ ANR-08-BLAN-0292 and the FP6 program 037704 ‘AGRONOMICS’ for their support. Deposited in PMC for release after 12 months.