The shape of plants depends on cellulose, a biopolymer that self-assembles into crystalline, inextensible microfibrils (CMFs) upon synthesis at the plasma membrane by multi-enzyme cellulose synthase complexes (CSCs). CSCs are displaced in directions predicted by underlying parallel arrays of cortical microtubules, but CMFs remain transverse in cells that have lost the ability to expand unidirectionally as a result of disrupted microtubules. These conflicting findings suggest that microtubules are important for some physico-chemical property of cellulose that maintains wall integrity. Using X-ray diffraction, we demonstrate that abundant microtubules enable a decrease in the degree of wall crystallinity during rapid growth at high temperatures. Reduced microtubule polymer mass in the mor1-1 mutant at high temperatures is associated with failure of crystallinity to decrease and a loss of unidirectional expansion. Promotion of microtubule bundling by over-expressing the RIC1 microtubule-associated protein reduced the degree of crystallinity. Using live-cell imaging, we detected an increase in the proportion of CSCs that track in microtubule-free domains in mor1-1, and an increase in the CSC velocity. These results suggest that microtubule domains affect glucan chain crystallization during unidirectional cell expansion. Microtubule disruption had no obvious effect on the orientation of CMFs in dark-grown hypocotyl cells. CMFs at the outer face of the hypocotyl epidermal cells had highly variable orientation, in contrast to the transverse CMFs on the radial and inner periclinal walls. This suggests that the outer epidermal mechanical properties are relatively isotropic, and that axial expansion is largely dependent on the inner tissue layers.
Spatial organization of microtubules at the cell cortex has a critical but as yet unclear function in determining the mechanical properties of plant cell walls (Wasteneys, 2004; Wasteneys and Fujita, 2006; Geitmann and Ortega, 2009). This function is central to axis formation, which is a defining feature of plant development and essential for many processes, including organ specification during embryogenesis and root and stem elongation. Cell expansion is generally at right angles to the predominant orientation of cellulose microfibrils, which, together with cross-linking hemicelluloses, are the major stress-resisting component of cell walls, due to tight packing of glucan chains into a rigid crystalline superstructure. Ultrastructural evidence (Ledbetter and Porter, 1963; Bowling and Brown, 2008), together with tracking of fluorescently tagged cellulose synthase enzymes in living hypocotyl cells (Paredez et al., 2006), supports the idea that cellulose microfibrils are produced at the plasma membrane from multi-enzyme cellulose synthase complexes (CSCs), and that CSC displacement trajectories are along pathways subtended by cortical microtubules. Recent studies suggest that CSCs are also inserted into and retracted from the plasma membrane near cortical microtubules (Crowell et al., 2009; Gutierrez et al., 2009), but it is unclear how the spatial correlation of microtubules in the cortex and CSCs at the plasma membrane affects the physical properties of cellulose, and how this controls directional cell expansion.
The simplest model for cell axis maintenance states that cortical microtubules act by specifically controlling the orientation of cellulose microfibrils. Treating cells with microtubule-depolymerizing drugs has been reported to alter the orientation patterns of cellulose microfibrils (Takeda and Shibaoka, 1981; Mueller and Brown, 1982), but this has not been demonstrated in other studies (Emons et al., 1992; Baskin, 2001; Sugimoto et al., 2003). Indeed, cellulose microfibril orientation is largely unaffected in cells undergoing radial swelling as a consequence of microtubule disruption in the temperature-sensitive mor1-1 mutant (Himmelspach et al., 2003; Sugimoto et al., 2003) of Arabidopsis thaliana, or after drug-induced disassembly within time frames that do not affect cell division (Sugimoto et al., 2003; Baskin et al., 2004). In contrast, specifically inhibiting cellulose synthesis results in random cellulose microfibril orientation (Sugimoto et al., 2001; Himmelspach et al., 2003), and can also lead to a significant increase in dispersion of cortical microtubules about the transverse axis (Himmelspach et al., 2003; Chu et al., 2007; Paredez et al., 2008). These results, together with the observation that CSCs continue to be inserted at the plasma membrane and can move along uniformly oriented trajectories in the absence of microtubules (Paredez et al., 2006), suggest that cellulose microfibril orientation may be a self-organized, microtubule-independent process that is in itself insufficient to maintain unidirectional cell expansion.
How then do cortical microtubules control expansion direction? It has been proposed that, by closely associating with CSCs, cortical microtubules are able to modulate the mechanical properties of cellulose microfibrils, such as their length and relative strength (Wasteneys, 2004). The degree of crystallinity, a measure of the ratio of crystalline to amorphous cellulose in the whole cell wall, is one of the main characteristics determining the overall mechanical properties of cellulose microfibrils and the cell wall itself. High cellulose crystallinity makes microfibrils highly inextensible, but also limits cross-linking by hemicelluloses (Chambat et al., 2005), which is crucial for resisting mechanical stress during rapid cell expansion. In this study, we show that the proportion of crystalline cellulose decreases when cell elongation rates increase, but that, if the microtubule polymer mass is reduced, as occurs in the mor1-1 temperature-sensitive mutant, the proportion of crystalline cellulose remains high. We determined that the reduced microtubule polymer mass in the mor1-1 mutant reduces the proportion of CSCs that track in microtubule-defined domains, and that this is correlated with an increased velocity of CSCs. Under these perturbed growth conditions, we found no correlation between the altered CSC trajectories and the cellulose microfibril deposition patterns, further highlighting the significance of variable crystallinity in modulating the mechanical integrity of plant cell walls.
Cell-wall crystallinity is high when microtubules are destabilized
To determine how microtubule organization affects the chemical properties of the primary cell wall, we adapted an X-ray diffraction method previously used for wood samples (Coleman et al., 2009) to measure the proportion of crystalline cellulose in the cell wall, referred to here as the degree of crystallinity, in Arabidopsis mutants that have altered microtubule spatial organization and polymer mass, together with associated changes in directional cell expansion. Phenotypes of these mutants were previously characterized in roots, hypocotyls and leaves, but these organs have a mixture of growing and non-growing cells, and also have too little biomass to measure cell-wall crystallinity by wide-angle X-ray diffraction of intact organs. We therefore measured the crystallinity from the mid-point of the rapidly expanding upper 3 cm of inflorescence stems, organs that provide sufficient biomass to generate diffraction patterns from which the crystalline content of cellulose I may be calculated (Figure S1). To avoid potential effects of chemical extraction of the wall components, we used dried but otherwise intact inflorescence stems. Importantly, the microtubule and morphological phenotypes for all mutants examined were also seen in the inflorescence stems, as shown in Figure 1 and Figure S2.
We first compared the degree of crystallinity in wild type and the temperature-sensitive mutant mor1-1 (AT2G35630). At 21°C, mor1-1 mutants are identical in stature and morphology to wild type, but at 29°C, unidirectional expansion is impaired after cortical microtubules become short and lose parallel order as a result of reduced dynamics (Kawamura and Wasteneys, 2008; Allard et al., 2010). Stems that had reached a height of 5–8 cm were marked in 1 cm increments, and then grown for one more day at permissive (21°C) or restrictive (29°C) temperatures. Consistent with previous studies on leaves, hypocotyls and roots (Whittington et al., 2001; Sugimoto et al., 2003), cells from the upper growing regions of inflorescence stems of mor1-1 plants grown at 29°C contained microtubules that were shorter and had reduced parallel order relative to wild type (Figure 1a,b). As previously shown in mor1-1 roots (Sugimoto et al., 2003), cellulose microfibrils observed by field emission scanning electron microscopy retained parallel transverse orientation at the inner periclinal and radial walls (Figure 1c–e). We determined that neither the α-cellulose content (Figure 1f) nor crystallite width (Table S1) was altered in mor1-1 plants relative to wild type at 29°C. At 21°C, there was no significant difference in the degree of crystallinity between wild type (23.6 ± 2.5%) and mor1-1 (21.6 ± 2.2%) (Figure 1g). At 29°C, however, which stimulated increased stem elongation rates in wild type but more isotropic expansion in mor1-1 plants (Figure 1h), cell-wall crystallinity declined (17.8 ± 1.8%) in wild type but remained high (21.4 ± 1.4%) in mor1-1 (Figure 1g), suggesting that cortical microtubules influence the extent of crystallization during cellulose synthesis during rapid anisotropic growth. Another temperature-sensitive allele, mor1-2 (Whittington et al., 2001), also retained a high degree of crystallinity at restrictive temperature (Figure 2a), and had a reduced elongation rate relative to wild type (Figure 2b).
As our measurements were made on unmodified cell walls, the degree of crystallinity could reflect the ratio of crystalline cellulose to the other cell-wall components, including non-crystalline cellulose and matrix polysaccharides. We were interested to determine whether the decline in crystallinity during rapid growth in wild type could be explained by changes in this ratio. We therefore used GC/MS to measure the sugar content in the growing regions of inflorescence stems. The data shown in Table S2 indicate that, in wild type, the ratio of trifluoroacetic acid-insoluble glucose (cellulose content) to other sugars (matrix polysaccharide content) remained unchanged at the higher temperatures that stimulate rapid growth. In mor1-1, the ratio was the same as that of wild type at permissive temperature but increased at the restrictive temperature. Based on this analysis, we conclude that the changes in crystallinity detected by X-ray diffraction are independent of the ratio of cellulose to other polysaccharides.
To determine whether the high degree of crystallinity upon microtubule disorganization is specific to the mor1 temperature-sensitive mutants, we measured cell-wall crystallinity in two other genetically modified Arabidopsis lines that have altered microtubule organization at 21°C (Figure 2a).
The 60 kDa ATPase katanin (AT1G80350) is a microtubule-severing protein (Stoppin-Mellet et al., 2006). The bot1 loss-of-function allele, like the mor1-1 mutant, has disordered microtubule arrays and radially swollen cells (Bichet et al., 2001). Unlike mor1 mutants, microtubule lengths in bot1 mutants are not reduced (Bichet et al., 2001). We found that cell-wall crystallinity in bot1 (22.1 ± 3.1%) was not significantly different from that in wild type (23.6 ± 2.5%) (Figure 2a). We then confirmed by immunofluorescence that the cells of bot1 inflorescence stems have relatively long microtubules with reduced parallel order (Figure S2a,c), and that stem elongation rates are greatly reduced (Figure 2b). This result indicates that loss of parallel transverse microtubule orientation alone has no effect on wall crystallinity, and suggests that the reduced microtubule polymer mass in the mor1-1 mutant is more likely to account for the failure of wall crystallinity to decrease during rapid growth.
In order to determine whether wall crystallinity is correlated with microtubule polymer mass, we performed X-ray diffraction analysis on the inflorescence stems of a transgenic line in which the microtubule-associated RIC1 protein (AT2G33460) is over-expressed (RIC1-OX3). A previous study on leaf pavement cells in the RIC1-OX3 line demonstrated that cortical microtubules were well-ordered, dense and bundled (Fu et al., 2005). In RIC1-OX3 inflorescence stems, we observed many narrow and highly elongated cells in which well-aligned and apparently bundled cortical microtubules were present (Figure S2b,d). The degree of crystallinity in RIC1-OX3 inflorescence stems (19.6 ± 1.9%) was significantly lower than that in wild type (23.6 ± 2.5%) (Figure 2a). This result provides additional evidence that the abundance and stability of microtubules in the plant cell cortex are inversely correlated with the degree of wall crystallinity.
Temperature and microtubule organization affect YFP–CesA6 particle velocity
To obtain a better understanding of how the abundance of microtubule polymers in the cell cortex is related to wall crystallinity, we analyzed the movement of CSCs in the plasma membrane under conditions in which wall crystallinity is altered. We tracked the movement of CSCs using YFP–CesA6, a functional yellow fluorescent protein-tagged reporter of CesA6 (AT5G64740), which is a primary cell-wall CesA that is normally expressed in rapidly elongating hypocotyls (Paredez et al., 2006). After determining that the original YFP–CesA6 reporter described in Paredez et al. (2006) was homozygous lethal, presumably because of its insertion in a vital gene, and that the transgene insertion was linked to the MOR1 locus, we re-screened primary transformants to obtain a line in which the transgene could be maintained homozygously and introgressed into the mor1-1 background (see Experimental procedures). YFP–CesA6 particles, presumed to be subunits of active CSCs, were imaged in dark-grown hypocotyl epidermal cells using a spinning disc confocal microscope equipped with a temperature-controlled stage and an objective lens heater set at either 21 or 29°C. As cellulose biosynthesis is a biochemical reaction and is expected to be influenced by temperature, temperature control during imaging was critical for measuring CSC kinetics. The tendency for the lens and stage to warm during extended observation made it imperative to monitor and maintain the temperature carefully for all analyses.
We first compared CSC velocity and trajectories in wild type cells at 21 and 29°C, temperatures that stimulate moderate and rapid hypocotyl growth, respectively. Cellulose polymer formation is thought to provide the force to displace CSCs, so, assuming greater cellulose production at 29°C, we predicted an increase in CSC velocity. Projections of 5 min time-lapse recordings of CSC trajectories (Figure 3a–d) demonstrated that YFP–CesA6 particles move with very similar patterns at both temperatures, producing locally parallel trajectories that were transverse or oblique relative to each cell’s longitudinal axis (Figure 3a,b, Figure S3a and Movies S1 and S2). According to these time-lapse movies and the criss-cross patterns of moving particles displayed in kymographs (Figure 3e–h and Movies S1–S4), YFP–CesA6 particles moved in opposite directions along the same pathways (Figure 3e,f), as previously described (Paredez et al., 2006). The velocity of YFP–CesA6 particles increased dramatically from 84 ± 47 nm min−1 at 21°C (Figure 3i) to 311 ± 89 nm min−1 after 3 h at 29°C (Figure S3c), and then to 354 ± 73 nm min−1 after 24 h at 29°C (Figure 3j). This 4.2-fold increase in CSC velocity over an 8°C temperature shift suggests that temperatures stimulating increased cell growth also increase the rate of cellulose production by each CSC.
Next we compared YFP–CesA6 velocities in the mor1-1 mutant at 21 and 29°C. At 21°C, the trajectories of YFP–CesA6 particles, as with the wild type, were locally parallel and either transverse or oblique relative to the cell’s long axis (Figure 3c and Movie S3). At 29°C, particle trajectories were sometimes curved, showed reduced parallel order, and ranged in orientation from transverse to longitudinal (Figure 3d, Figure S3b and Movie S4). As with wild type, YFP–CesA6 particles in mor1-1 at 29°C moved bi-directionally along linear pathways (Figure 3g,h). The YFP–CesA6 particle mean velocity increased in mor1-1 from 71 ± 42 nm min−1 at 21°C (Figure 3k) to 339 ± 84 nm min−1 after 3 h (Figure S3d), and then to 410 ± 79 nm min−1 after 24 h at 29°C, representing a 5.8-fold increase (Figure 3l). The mean CSC velocity in mor1-1 is significantly greater than the mean velocity recorded for wild type at 29°C, indicating that the temperature-dependent increase in CSC velocity is enhanced under conditions of reduced microtubule abundance.
Microtubule reduction in mor1-1 increases the proportion of CSCs that track in plasma membrane domains devoid of microtubules
To directly assess the distribution and motility of CSCs in relation to cortical microtubule polymer abundance, we introduced a red fluorescent protein-tagged β-tubulin 6 (AT5G12250) (mRFP–TUB6) microtubule reporter into the YFP–CesA6-expressing lines, and recorded 5 min time-lapse series in the red and yellow channels. Instead of using conventional merge or overlap line scan analysis tools (Paredez et al., 2006), we used a different approach to quantify areas occupied by microtubules and CSC trajectories in two dimensions to determine whether CSCs have a preference to coincide with microtubules when microtubules are perturbed. We used a rolling ball algorithm to eliminate background fluorescence before overlap quantification (Figure 4a–d and Movies S5–S8). At 21°C, wild type and mor1-1 showed no significant differences in the areas occupied by either microtubules or CSCs, or in the proportion of the microtubule-occupied areas that overlapped with CSC trajectories (Figure 4e). At 29°C, there was also no difference in the overall area occupied by CSCs over 5 min periods or in the proportion of the microtubule domains coincident with CSC trajectories. However, the total area occupied by microtubules was significantly reduced in mor1-1 at 29°C (41.4 ± 10.1%) compared to wild type (55.8 ± 11.6%). Consequently, the proportion of CSCs that track in microtubule-free domains increased in mor1-1 (51.8 ± 11.9%) compared with wild type (37.2 ± 9.9%) at the restrictive temperature.
The lower degree of coincidence between CSCs and cortical microtubules in mor1-1 at 29°C (77% relative to wild type) may be attributed in part to CSCs continuing to track within the same domains after the disappearance of microtubules, as previously reported for wild type cells (Paredez et al., 2006). By breaking down the 29°C time-lapse analysis into sub-series (Figure S4), we found a lag between formation of a microtubule domain and the appearance of CSCs at that site, and then continued tracking of CSCs for a short while after the disappearance of microtubules from the domain. However, this behaviour was common to both the mor1-1 mutant and the wild type. This further supports the conclusion that the reduced coincidence of CSCs and cortical microtubule domains in mor1-1 is mainly a function of the reduced area of the cortex occupied by microtubules.
The lower wall crystallinity measured in rapidly expanding wild type cells at 29°C could be the result of increased wall loosening, which would dilute the cellulose content in a given area. However, this possibility is not supported by measurements of CSC density. We found no difference between wild type and mor1-1 in terms of the area occupied by CSCs at either 21 or 29°C (Figure 4e), despite differences in the degree of crystallinity.
Microfibril orientation patterns are complex at the outer periclinal face of hypocotyl epidermal cells and do not appear to be affected by microtubule disruption
So far none of the studies in which CSC trajectories have been identified by tracking fluorescent CesA subunits has examined the final cellular alignment of cellulose microfibrils. Previous studies showed that cellulose microfibrils in mor1-1 mutant root epidermal cells maintain a parallel transverse orientation when microtubule arrays are disrupted (Himmelspach et al., 2003; Sugimoto et al., 2003). However, in these studies, the images of microtubules and cellulose microfibrils were taken from the radial and inner periclinal walls, whereas tracking YFP–CesA6 particles is so far only possible at the outer periclinal face of epidermal cells. The cryoplaning technique employed usually removes this outer periclinal surface, but occasionally some folded-back outer periclinal walls are retained so that the inner surface of this wall face can be observed. Microfibrils produced at 29°C as well as at 21°C at the outer periclinal wall of dark-grown hypocotyls of both wild type and mor1-1 were found to be locally parallel but variably oriented (Figures 5 and 6; data not shown for 21°C). Importantly, the microfibril patterns in the mor1-1 mutant did not show the disorder predicted by the CSC trajectories. Cellulose microfibrils in radial and inner periclinal walls had a transverse parallel orientation in both wild type and mor1-1 (Figure S5), consistent with previous results for mor1-1 roots (Sugimoto et al., 2003) and the present results for inflorescence stems (Figure 1c,d). Our analysis of cellulose microfibril deposition patterns indicates that the outer periclinal face of epidermal cells in dark-grown hypocotyls is unusual, with a highly variable orientation of cellulose microfibrils that, unlike microfibrils at most other cell surfaces examined to date, are not oriented perpendicular to the major axis of expansion.
Our study has uncovered a link between cortical microtubules and the physical properties of expanding cell walls. Specifically, we found that the proportion of crystalline cellulose declined at temperatures promoting rapid cell elongation, but that there was no such reduction when microtubule assembly was compromised in the mor1-1 mutant at the same high temperature. Several results indicate that modification of cell-wall crystallinity depends on maintaining a critical microtubule polymer mass rather than parallel transverse microtubule orientation. First, in the mor1-1 mutant, a reduced microtubule polymer mass resulted in a greater proportion of CSCs that track in non-microtubule domains, and this was correlated with a relatively high degree of crystallinity. Second, in the bot1 mutant, in which microtubule arrays are disordered but microtubule length and abundance are not reduced (Bichet et al., 2001), there was no significant change in the wall crystallinity. Third, over-expressing the Rho of plants (ROP) effector protein RIC1, which promotes microtubule accumulation in bundles (Fu et al., 2005), reduced wall crystallinity.
The X-ray diffraction technique that we used on inflorescence stems to compare wall crystallinity was designed to restrict our measurements to primary walls of the expansion zone. Recent studies using X-ray diffraction have used Arabidopsis biomass prepared from oven-dried organs, which are then ground, treated with boric acid, and centrifuged (Harris and DeBolt, 2008; Harris et al., 2009). These whole-organ extracts contain a composite of cellulose from primary and secondary cell walls.
Despite the different methods of sample preparation for X-ray diffraction, the values calculated for relative crystallinity in the growing regions of the inflorescence stem are in close agreement with the values obtained by Harris and DeBolt (2008) for hypocotyls. It is also interesting to note that, in addition to our finding a reduction in the degree of crystallinity in infloresence stems from 24% at 21°C to 18% at 29°C with increased elongation rates, Harris and DeBolt (2008) reported a similar reduction in the crystallinity index from 27% in light-grown hypocotyls to 23% in rapidly growing dark-grown hypocotyls. Reduced crystallinity could therefore be an important adaptation for maintaining anisotropic expansion during rapid growth. In vitro adsorption analysis showed that the binding capacity of the hemicellulose xyloglucan is higher when cellulose has a low degree of crystallinity (Chambat et al., 2005). Thus, a higher proportion of amorphous cellulose during rapid wall expansion could promote a greater degree of tethering by xyloglucan. This in turn may limit lateral slippage of cellulose microfibrils, so that they maintain controlled separation to prevent isotropic expansion during rapid growth.
Figure 7 summarizes schematically the mor1-1 and wild type microtubule, CSC and cellulose microfibril phenotype analysis at 21 and 29°C. In the remainder of this discussion, we first consider how microtubule polymer status and temperature can affect the velocity and activity of CSCs, and then speculate on how the activity of CSCs within plasma membrane domains subtended by microtubules affects the quality of the cellulose produced. Finally, we consider the lack of correlation between altered microtubule orientation and the most recently deposited cellulose microfibrils, as determined by high-resolution SEM analysis.
The dramatic increases in the velocity of CSC displacement at higher temperature observed in this study are consistent with the need to couple cellulose synthesis with cellular growth rates. The 4.2-fold increase in CSC velocity in wild type following a temperature increase of 8°C closely corresponds to the 4-fold increase in cellulose production over a 10°C temperature range in secondary walls of cotton (Pillonel and Meier, 1985; Roberts et al., 1992). However, we measured changes in primary walls, so any increase in rate of cellulose synthesis per CSC would be reflected in an increased growth rate rather than net cellulose content. It is not feasible to compare our CSC velocity values with those of previous studies (Paredez et al., 2006; DeBolt et al., 2007) because the previous studies did not specifically control or monitor temperature.
The reduction in microtubule polymer mass in the mor1-1 mutant at the restrictive temperature affected CSC activity in two ways. First, a higher proportion of CSCs tracked in microtubule-free domains. Second, the mean CSC displacement velocity was increased relative to wild type. This suggests but does not prove that CSC displacement is faster in microtubule-free domains.
Microtubules might control CSC velocity by locally modulating the composition and fluidity of the plasma membrane through which the active CSCs move as they produce cellulose. The observed loss of distinct transverse banding of the GPI-anchored, putative cellulose-interacting protein COBRA upon microtubule depolymerization (Roudier et al., 2005) provides evidence for microtubule-dependent plasma membrane heterogeneity. Thus, microtubule-defined plasma membrane microdomains might control membrane fluidity, which has been suggested to influence CSC movement (Emons et al., 1992). In the mor1-1 mutant, the relatively high CSC displacement velocity suggests that there is a higher membrane fluidity than in wild type.
Although the relationship between CSC enzyme activity (and therefore displacement velocity) and the crystalline content of the cell wall is likely to be complex, it is interesting to note that, in the mor1-1 mutant, there is a moderate increase in both the CSC velocity and the degree of crystallinity at restrictive temperature. It is unclear whether there is a causal relationship between CSC velocity and the degree of crystallinity as a result of altered microtubule polymer mass, and this question will need to be addressed in future research.
Do plasma membrane domains populated by cortical microtubules modulate the proportion of crystalline and amorphous cellulose that is synthesized by membrane-bound CSCs? One way this could occur is by microtubules controlling the secretion of non-cellulosic polysaccharides or proteins that modify polysaccharides in the vicinity of the CSCs. Indeed, as shown here (Table S2), there was a decrease in other wall polysaccharides relative to crystalline cellulose when microtubule polymer mass was reduced in the mor1-1 mutant at restrictive temperature. Cellulose microfibrils synthesized in the absence of matrix polysaccharides in vitro have a higher degree of crystallinity compared with endogenous cellulose from plant cells (Lai-Kee-Him et al., 2002). Thus, if matrix polysaccharide secretion or cross-linking is enhanced in microtubule microdomains, the proportion of crystalline cellulose is predicted to be reduced. In contrast, in the mor1-1 mutant, in which reductions in both microtubule polymer levels and the proportion of matrix polysaccharides occur, the predicted reduction in matrix–cellulose interaction should result in a higher proportion of crystalline cellulose, as was indeed observed.
Despite the general correspondence between cortical microtubule domains and CSC trajectories, disrupting cortical microtubule orientation in mor1-1 had no obvious effect on the orientation patterns of the most recently deposited cellulose microfibrils at the outer periclinal face of hypocotyl epidermal cells, which were locally parallel but variably oriented in both mor1-1 and wild type. In agreement with previous studies on roots (Sugimoto et al., 2000; Sugimoto et al., 2003), microfibrils were parallel and transverse at the inner periclinal and radial walls. Unfortunately, it remains technically challenging to detect YFP–CesA6 movement at the inner periclinal and radial walls where microtubules in the mor1-1 mutant have been shown to be disorganized (Sugimoto et al., 2003). The apparent lack of correlation between CSC trajectories and microfibril orientation suggests that cellulose orientation is consolidated after synthesis. It seems unlikely, but we cannot rule out the possibility that other active CSCs not detected by the YFP–CesA6 reporter, i.e. CesA2, 5 and 9, may be partially redundant with CesA6 (Robert et al., 2004; Desprez et al., 2007; Persson et al., 2007), and may replace CesA6 in some of the CSCs.
The highly variable orientation patterns of both microtubules and cellulose microfibrils at the outer periclinal face of hypocotyl epidermal cells during elongation is in sharp contrast to the consistently transverse arrangement found along radial and inner periclinal faces. The outer epidermal wall is relatively thick, but its bulging appearance suggests relatively isotropic mechanical properties, consistent with variably arranged cellulose. This finding suggests that the inner periclinal and radial walls of the epidermis, together with the inner tissues, contribute greatly to the architecture that shapes the organ.
When first described nearly 50 years ago, the close correspondence between cortical microtubule orientation and the deposition patterns of cellulose microfibrils (Ledbetter and Porter, 1963) appeared to confirm a previous speculation that spindle fibre-like proteins were responsible for aligning cellulose (Green, 1962). Subsequent studies on the microtubule–cellulose relationship have paid little attention to exploring the physico-chemical features of the cellulose. The present study identifies a potential role for cortical microtubules in controlling unidirectional cell expansion through modulation of cellulose crystallinity.
Plant materials and growth conditions
The Arabidopsis thaliana homozygous mor1-1 mutant (Whittington et al., 2001), which had been back-crossed eight times, and wild type segregants from this back-cross, were used throughout this study. Other microtubule-related mutant lines included mor1-2 (Whittington et al., 2001), botero1 (bot1) (Bichet et al., 2001), provided by Dr Herman Höfte (Centre de Versailles-Grignon, INRA, France), and RIC1-OX3 (Fu et al., 2005), provided by Dr Zhen Biao Yang (Center for Plant Cell Biology, University of California at Riverside, CA). Plants were grown as described previously (Kawamura and Wasteneys, 2008). Ten-day-old seedlings were transferred to soil and grown at 21°C for harvesting inflorescence stems.
A T1 generation of YFP–CesA6 transgenic lines described by Paredez et al. (2006) was obtained from the laboratory of Dr Chris Somerville (Department of Plant Biology, Carnegie Institution for Science, Stanford, CA; currently Energy Bioscience Institute, University of California at Berkeley, CA).
Introgressing YFP–CesA6 into the mor1-1 mutant
A homozygous YFP–CesA6 segregant in the prc1-1 mutant (i.e. CesA6 null) background was obtained in the T3 generation and crossed with a homozygous mor1-1 mutant. F2 seedlings were screened for YFP–CesA6 expression in mor1-1/prc1-1 double homozygotes. To identify prc1-1 homozygotes expressing YFP–CesA6, 3-day-old dark-grown seedlings were observed by spinning disc confocal microscopy to screen for lines expressing YFP particles at the cell periphery. To verify that the selected lines carried the prc1-1 mutation, genomic DNA was extracted from leaf tissue, and the region of DNA containing the prc1-1 mutation (a 3.5 kb fragment) was amplified using primers prcF (5′-AGTGGCTGCGGATAAGAA-3′) and prcGR (5′- CCTTCACAGAAGCACCGAA-3′).
Generation of mRFP–TUB6 in YFP–CESA6, prc1-1 and mor1-1/prc1-1 transgenic lines
The binary vector pCAMBIA1300 containing CaMV 35S promoter-driven mRFP fused with TUB6 was obtained from Dr Richard Cyr (Biology Department, Penn State University, University Park, PA), and transformed into Agrobacterium tumefaciens strain GV3101 by electroporation. After transformation of pCesA6::YFP-CesA6/mor1-1/prc1-1 triple homozygotes by floral dipping (Weigel and Glazebrook, 2002), T2 and T3 plants were used for experiments.
Immunofluorescence labelling of microtubules in the epidermal cells of inflorescence stems
Inflorescence stem preparation for immunofluorescence labelling was performed as described previously (Kazama and Mineyuki, 1997; Sugimoto et al., 2000). The elongation zones of wild type and mor1-1 inflorescence stems grown at 29°C for 24 h were used. Alexa 488 fluorescent images of immunolabelled samples were obtained using an upright AxioImager M1 microscope (Carl Zeiss, http://www.zeiss.ca/) equipped with a Zeiss PASCAL Excite two-channel scan head, using the 488 nm line from an argon laser, with a 488 nm dichroic filter and a 505 nm emission filter, a 63 × NA 1.4 oil-immersion lens, using Kalman averaging (n =2). Images were recorded using lsm software (Carl Zeiss) and processed using ImageJ (http://rsb.info.nih.gov).
Cell-wall preparation for field emission scanning electron microscopy (FESEM)
Cell walls from inflorescence stems and hypocotyls were prepared for analysis of cellulose microfibril orientation patterns as described previously for Arabidopsis roots (Himmelspach et al., 2003). Hypocotyl images were taken from dark-grown 3-day-old YFP–CesA6/prc1-1 and YFP–CesA6/mor1-1/prc1-1 seedlings. More than three samples were observed for each treatment. Cellulose microfibrils were detected using a Hitachi S4700 scanning electron microscope (http://www.hitachi.com) set at 3 kV and 10 μA. High-magnification images were obtained mid-way along the cell.
Measurement of cellulose microfibril angle
A 5 × 7 array of reference points was set on each image taken at 35 K magnification, and the orientation of cellulose microfibrils at each reference point was measured relative to the long axis of the cell using ImageJ. A total of 420 cellulose microfibrils from 12 epidermal cells (15–35 μm long) from four stems in wild type and the mor1-1 mutant were examined.
Determination of α-cellulose content
After 1 day at 21 or 29°C, the growing regions of inflorescence stems were excised and oven-dried at 90°C for 1 day. Dried stems were ground and extracted with HPLC-grade acetone for 24 h. After evaporation, sequential lignin and hemicellulose extractions were performed using a microanalytical method (Yokoyama et al., 2002). α-cellulose content (%) was determined as the proportion of dried weight of α-cellulose per total weight of dried inflorescence stems.
Measurement of the degree of crystallinity in the wall
The growing regions of inflorescence stems grown at 21 and 29°C for 1 day were excised with scissors, flattened and dried between heavy books for 1 month. X-ray diffraction patterns were recorded using a Bruker AXS Advance D8 X-ray diffractometer equipped with an area array detector (http://www.bruker.com). CuKα radiation was generated at an accelerating voltage of 40 kV with a current of 30 mA. A flattened dried stem was placed vertically in the sample holder. Diffraction intensities were collected using the diffractometer at Bragg angles between 4 and 40°, and diffraction data from 4 to 40° in the 2θ angle range were integrated using gadds software (Bruker). Crystallite width was calculated from the breadth of the X-ray diffraction peak representing the cellulose 002 plane using Scherrer’s formula (Scherrer, 1918).
The collected data were normalized and resolved using a crystallinity calculation program based on the Vonk method (1973). Briefly, the background diffraction signal was subtracted, then the amorphous curve was fitted to the diffraction pattern, and finally a linear regression analysis was performed to obtain cell-wall crystallinity.
Measurement of inflorescence stem elongation rate
Once the length of inflorescence stems of wild type and mor1-1 grown at 21°C reached 5–8 cm, they were marked 2 cm from the top. Plants grown at 21 or 29°C for 1 day were used for measuring the total length of the region. Mean lengths of the region were then calculated from data from 20 inflorescence stems of wild type, mor1-1 and other mutant lines.
Carbohydrate analysis of the cell wall
Plants were grown either at 21°C for 29 days followed by 30°C for 7 days, or at 21°C for 42 days, the top growing regions of the stems were harvested, and carbohydrate analysis was performed as described previously (Lane et al., 2001).
YFP–CesA6 image acquisition
Plants were cultured on Hoagland’s medium in 1.2% agar in vertically held Petri plates covered with aluminium foil. Seeds were incubated at 4°C for 4 days, then germinated at 21°C. Hypocotyls were obtained from dark-grown 3-day-old YFP–CesA6 wild type and YFP–CesA6 mor1-1 seedlings that were either (i) grown for the full 3 days at 21°C, (ii) grown for 3 days at 21°C and for 3 h at 29°C, or (iii) grown for 2 days at 21°C and for 1 day at 29°C. All tools such as glass slides, cover slips, mounting media and forceps were either pre-cooled or pre-warmed to the desired temperature of 21 or 29°C.
To keep the temperature stable around the mounted samples, a temperature-controlled stage, bionomic controller BC-110 equipped with heat exchanger HEC-400 (20/20 Technology Inc., http://www.20-20tech.com), was set up to maintain temperature at either 21 or 29°C. An objective lens heater (Bioptechs, http://www.bioptechs.com) was used for the 29°C setting. The sample temperature was monitored immediately after imaging by measuring the temperature of glycerol on the cover slip.
Images were obtained with 600 msec exposure times for both YFP–CesA6 and mRFP–TUB6. Two frames were taken at each time point, and kalman averaged to reduce background fluorescence from rapidly moving Golgi bodies. Live-cell imaging was performed at 10 sec intervals over a 5 min period.
YFP–CesA6 velocity measurement
To adjust tissue translocation during imaging, the StackReg (Thévenaz et al., 1998), MultiStackReg or Align Slice plugins for ImageJ were used to align two-frame averaged images. The images were smoothed by averaging three continuous frames using the Walking Average plugin (ImageJ), and kymograph analysis was performed using Multiple Kymograph plugins (ImageJ).
Coincidence analysis of microtubules and CSCs
Background fluorescence was subtracted from the projections of 5 min time-lapse images of mRFP–TUB6 and YFP–CesA6 using a rolling ball algorithm (ImageJ) and setting thresholds (Figure S6). The area for the analysis was selected based on two criteria: (i) both microtubules and CSCs are in sharp focus, and (ii) no Golgi background fluorescence is present. Total areas occupied by either microtubules or CSCs, and by both microtubules and CSCs, were calculated. The degree of coincidence between microtubule domains and CSCs was determined by dividing the area occupied by both microtubules and CSCs by the area occupied by microtubules. The proportion of microtubule-independent CSCs was determined by subtracting the area occupied by both microtubules and CSCs from the total area occupied by CSCs, then dividing it by the total CSC area.
A Shapiro–Wilk normality test was used to determine whether the data were normally distributed. For the data showing abnormal distributions, a Wilcoxon rank sum test was used to determine whether the means under different conditions were significantly different. For small sample numbers, standard deviations were calculated, and means were compared by the independent Student’s t test for samples with unequal variance at a significance level of 0.05.
We thank C.R. Somerville (Energy Bioscience Institute, University of California at Berkeley, CA), R.J. Cyr (Biology Department, Penn State University, PA), J.C. Ambrose, Z.-B. Yang (Center for Plant Cell Biology, UC Riverside, CA) and H. Höfte (Centre de Versailles-Grignon, INRA, France) for provision of experimental materials, K. Hodgson and D. Horne from the University of British Columbia Bioimaging Facility for imaging assistance, and J.C. Ambrose, C.J. Douglas and A.L. Samuels for helpful discussions. This work was supported by grants to G.O.W. from the Australian Research Council, the Natural Sciences and Engineering Research Council of Canada (298264-2009), The Canadian Institutes of Health Research (178544), and the Canadian Foundation for Innovation.