Research Article

Visualization of Cellulose Synthase Demonstrates Functional Association with Microtubules

See allHide authors and affiliations

Science  09 Jun 2006:
Vol. 312, Issue 5779, pp. 1491-1495
DOI: 10.1126/science.1126551


Expression of a functional yellow fluorescent protein fusion to cellulose synthase (CESA) in transgenic Arabidopsis plants allowed the process of cellulose deposition to be visualized in living cells. Spinning disk confocal microscopy revealed that CESA complexes in the plasma membrane moved at constant rates in linear tracks that were aligned and were coincident with cortical microtubules. Within each observed linear track, complex movement was bidirectional. Inhibition of microtubule polymerization changed the fine-scale distribution and pattern of moving CESA complexes in the membrane, indicating a relatively direct mechanism for guidance of cellulose deposition by the cytoskeleton.

Cellulose is synthesized in vascular plants by a plasma membrane–localized enzyme, cellulose synthase, that has been visualized by freeze fracture of plasma membranes as 25- to 30-nm-diameter symmetrical rosettes with six resolved subunits (1). From measurements of the dimensions of cellulose microfibrils, it has been inferred that each of the six subunits of a rosette synthesizes approximately six β-1,4-glucan chains, which hydrogen bond with each other to form a microfibril of about 36 chains that is extruded into the extracellular space and can reach more than 7 μm in length (2). The only known components of cellulose synthase in higher plants are a family of 10 CESA proteins. Genetic studies of cellulose synthesis during secondary cell wall formation have shown that at least three different CESA proteins must be simultaneously present to support cellulose synthesis (3).

The deposition of cellulose fibrils is generally oriented perpendicular to the axis of cellular expansion in growing tissue, a feature that has been postulated to facilitate directional, or anisotropic, cell growth. Disruption of microfibril organization by the anti–spindle fiber drug colchicine, and accompanying isodiametric cell expansion, led Green to propose that spindle fibers might play a role in orienting the deposition of cellulose microfibrils and constraining the pattern of cell expansion (4). Soon thereafter, cortical microtubules were discovered (5) and were frequently observed to lie parallel to the cellulose fibrils [reviewed in (6)]. The alignment hypothesis for cellulose deposition has since evolved into two major forms: Microtubules have been proposed to act as molecular rails, directly guiding cellulose synthase rosettes as they synthesize microfibrils (7); alternatively, microtubules have been proposed to serve as passive constraints, forming channels that confine the lateral movement of synthesizing complexes such that a net coorientation of microfibrils and microtubules results (8), a model that has been popularized in current textbooks (9). The alignment hypothesis, under either model, makes two predictions: that microtubules and microfibrils should be coaligned, and that changes in microtubule organization should cause changes in microfibril arrangement (10). However, many inconsistencies between the orientation of microtubules and cellulose alignment have been observed, and a role for microtubules in cellulose alignment remains controversial (1014). Here we test the alignment hypothesis by simultaneous and dynamic visualization of cellulose synthase and microtubules in living plant cells.

Visualization of dynamic CESA6 complexes in the cell membrane. Until recently, the only way to simultaneously visualize microtubules and cellulose synthase was by electron microscopy of fixed tissue, where the dynamic relationship between molecules cannot be observed and only a very limited region of the cell can be viewed at a time. Thus, transient states of molecular association could not be examined, and global patterns of organization and association were difficult to observe. Other methods to visualize the outcome of cellulose synthase activity, such as polarization microscopy or dye labeling of material in the cell wall (11), do provide global views of accumulated cellulose organization, but they do not report on the activity and distribution of cellulose synthase during the process of cell wall creation. To visualize cellulose synthase, we produced transgenic Arabidopsis plants in which an N-terminal fusion of citrine yellow fluorescent protein (YFP) (15) to the CESA6 protein was expressed under control of the native CESA6 promoter in a cesa6-null mutant background [prc1-1(16)]. This construct complemented the mutant phenotypes of reduced hypocotyl elongation and radialized cell expansion, indicating that the fusion protein was functional (table S1). A similar construct for the CESA7 gene was previously shown to complement a mutation affecting secondary cell wall synthesis in the vasculature (17).

Observation of expanding hypocotyl cells by spinning disk confocal microscopy revealed YFP:CESA6 fluorescence in three subcellular locations (Fig. 1). First, the label was detected in the focal plane of the plasma membrane as discrete particles at or below the resolution limit of the microscope (Fig. 1, A and B). These particles were not randomly dispersed, but tended to be organized into linear arrays (Fig. 1, A and B). Imaged over time, these particles were also observed to be motile (Movie S1), tracing roughly linear trajectories along the axes of the particle arrays. These trajectories were visualized in static images by averaging the frames of the time series (Fig. 1C). In kymograph analysis, particle motions described straight traces that were parallel to each other, indicating that velocities were steady and highly similar from particle to particle (Fig. 1D). The cross-hatching of traces in the kymographs indicates that particle movement was bidirectional within the tracks defined by particle translocation. Thus, particle motility does not show polarity relative to individual tracks nor to the cell axis. The slopes of these traces revealed an average velocity of 330 nm/min, with a range of 150 to 500 nm/min and a standard deviation of 65 nm/min (Fig. 1E). These rates of movement correspond to the addition of ∼300 to 1000 glucose residues per glucan chain per minute, values roughly one-third of those previously predicted for in vivo synthetic rates in algal cells (18). Labeled particles were seen to appear de novo in optical sections and to begin translation immediately (Movie S6), indicating a very short lag time (<10 s) for particle motility to commence after appearance at the cell membrane. A short lag time for the initiation of particle motility is also supported by frame averaging analysis, where evidence for stationary particles was seldom observed (Figs. 1 and 2). Particle lifetimes have not yet been accurately determined because the high density of the particles, the limited field of view, and loss of signal over extended observation due to photo-bleaching make it difficult to obtain compelling numbers. However, we have observed individual particles moving at a constant rate for at least 15 min.

Fig. 1.

YFP:CESA6 localization and motility in etiolated hypocotyl cells. Optical sections of plasma membrane (A, B, C, and F) and adjacent cytoplasm (G) in upper hypocotyl cells of etiolated prc1-1 seedlings expressing YFP:CESA6 (3 days old). Images in all figures were acquired by spinning disk confocal microscopy. (A) Average of five frames (10-s intervals) acquired at the plane of the cell membrane. (B) Enlarged region of (A) corresponding to the yellow frame showing arrays of CESA particles marked with green arrowheads. (C) Average of 61 frames showing movement of labeled particles along linear tracks during 10 min. (D) Kymograph of region highlighted in (C), displaying steady, consistent, and bidirectional particle translocation. (E) Histogram of particle velocities calculated from kymograph analysis of 303 particles in 32 tracks measured in seven cells from six seedlings. (F) and (G) Adjacent focal planes of the same cell showing YFP:CESA6 particles in the cell membrane, Golgi [red circles in (G)], and a particulate cytosolic compartment [yellow arrowheads in (G)]. Scale bars: 10 μm (A and G), 5 μm (B).

Fig. 2.

Colocalization of microtubules and YFP:CESA6 in etiolated hypocotyl cells. YFP:CESA6 labeling is in green, microtubules labeled with CFP:TUA1 are in red, and combined images are on the right. Image acquisition interval was 10 s. (A to C) Average of five image frames showing colocalization of YFP:CESA6 particles and microtubules. (D to F) Average of 30 frames reveals YFP:CESA6 particle paths along microtubules and microtubule bundles. Arrowheads mark prominent areas of colocalization in (A) to (F). Brackets indicate an area sparsely populated by either microtubules or YFP:CESA6. (G) Plot of a line scan through a region of (F) (dashed line) showing a strong correlation between CFP:TUA1 and YFP:CESA6. (H to M) Correlated shift in orientation of both YFP:CESA6 and microtubules. (H to J) Average of first five frames of a 10-min image sequence taken adjacent to a region in the same cell that had already been imaged for 10 min. (K to M) Average of the last five frames of the same 10-min sequence. Reorientation of the microtubules and the YFP:CESA6 particle tracks do not occur in unimaged controls mounted for 20 min. Scale bar, 10 μm.

Citrine-YFP fluorescence was also seen to accumulate in cytoplasmic compartments (Fig. 1G). Organelles labeled by a prominent ring of fluorescence were confirmed to be Golgi by colocalization with a cyan fluorescent protein (CFP):MANNOSIDASE marker (fig. S1) (19). Labeling of the Golgi by YFP:CESA6 was nonuniform and characterized by formation of distinct puncta (Fig. 1G). The relative brightness of these puncta suggests a higher concentration of labeled protein than is observed in puncta at the plasma membrane, perhaps representing the package of CesA6 protein into secretory vesicles (20). It is also possible that assembly of CESA rosettes in the Golgi could contribute to the punctate appearance of label in the Golgi, consistent with previous observations by electron microscopy (20). In addition to particulate localization at the cell membrane and to the Golgi, YFP:CESA6 also labeled a population of small organelles, at or near the resolution limit of the imaging system (Fig. 1G). These organelles were distinguished from the particles in the membrane by their focal plane (Fig. 1, F and G), higher fluorescent intensities, and faster and less steady patterns of movement (Movie S3).

The herbicide isoxaben specifically inhibits cellulose synthesis in plants. Missense mutations in CESA3 and CESA6 confer resistance to isoxaben, suggesting that these subunits are direct targets (21, 22). However, the residues affected are remote from the presumed active site of the enzyme, suggesting that the mode of action is not direct inhibition of catalysis. Treatment of seedlings with 100 nM isoxaben resulted in the rapid loss of YFP:CESA6 particles from the plasma membrane (fig. S2, Movie S2). Within 5 min, a decrease in particle density was observed, and within 20 min most of the plasma membrane YFP:CESA6 was lost (fig. S2). The ability of the YFP:CESA6 construct to rescue the mutant, together with its localization pattern, dynamic behavior, velocity of movement, and sensitivity to drug treatment, suggest that the observed YFP-labeled particles at the cell cortex were individual cellulose synthase biosynthetic complexes. The fluorescence intensities of the individual YFP:CESA6 particles within a given cell were not uniform (Fig. 1, Movie S1), suggesting either that the stoichiometry of CESA subunits within a complex is not fixed, or that individual particles may be composed of a variable number of rosette complexes.

Dynamic colocalization with cortical microtubules. To observe the spatial relationship between microtubules and membrane-localized cellulose synthase, we crossed a CFP:TUA1 α-tubulin marker into the YFP:CESA6 line. Two-channel confocal imaging of expanding hypocotyl cells revealed extensive overlap between the two patterns of label distribution (Fig. 2), and labeled CESA particles were observed to move along tracks defined by microtubules (Fig. 2, D to F, and Movie S3; ∼60% of pixels labeled with YFP:CESA6 above the local background were also labeled with CFP:TUA1 above the local background). Although most microtubules in these growing cells had an orientation that was roughly transverse to the cell axis, there were regional differences within cells for net array orientation, and many microtubules exhibited discordant angles and curved configurations. The high coincidence of YFP:CESA6 label with regional microtubule organization and with discordant and curved microtubules (Fig. 3) further indicate that the observed degree of colocalization was not due to coincidental overlap of two linear arrays with similar orientation. Furthermore, bidirectional tracking of labeled rosettes along curved and discordant microtubules is not expected if rosettes are simply passively channeled between the visually resolved spaces between elements of the cortical microtubule array.

Fig. 3.

Dynamic relation between YFP:CESA6 and CFP:TUA1. In all images YFP:CESA6 labeling is in green, CFP:TUA1 labeling is in red, and combined images are on the right. Each image is the average of three frames. Image acquisition rates are as in Fig. 1. Carets mark two regions where distinct microtubule bundles are decorated with CESA6 at the first time point (A to C). At subsequent time points (D to L), microtubule bundles marked by carets have depolymerized and YFP:CESA6 remains. Yellow arrowheads mark a newly assembled microtubule bundle (E and F) that persists over the remaining course of the experiment (H and I, K and L). Corresponding CESA label is not detected in this position (D and F, G and I) until after the new bundle is created (J and L). Scale bar, 10 μm.

Despite the widespread colocalization of YFP:CESA6 and cortical microtubules, there was not complete overlap between the two distributions. Time-lapse observation revealed that the dynamics of these two interacting molecular systems contributes to the observed degree of colocalization. For example, cortical microtubules migrate by polymer treadmilling at rates that are roughly four times the mean velocity of membrane-localized particles (23). Microtubule catastrophe can occur at up to 75 times this velocity (23). Thus, microtubules are considerably more dynamic than is cellulose synthase. Frequently, microtubules with associated YFP:CESA6 were observed to depolymerize rapidly while the YFP:CESA6 persisted and continued to translocate along their original trajectories, producing a local YFP:CESA6 linear array without the presence of a colinear microtubule (Fig. 3, Movie S5). Once complexes are set in motion, the rigidity of crystalline cellulose may be enough to maintain an initial trajectory (12). Linear arrays of YFP:CESA6 were observed to persist for as long as 4 min after their associated microtubules had depolymerized; these arrays eventually dissipated (Fig. 3, Movie S5), suggesting that the microtubules not only predict the localization of YFP:CESA6, but also stabilize the linear arrays of dynamic CESA6 complexes. Events in which microtubule polymerization outpaced the establishment of related YFP:CESA6 linear arrays were also observed (Fig. 3, Movie S5). Single treadmilling microtubules are the most dynamic elements of the cortical cytoskeleton (23) and were frequently observed in time-lapse movies (Movie S3), but YFP:CESA6 seemed to be primarily associated with the brighter and more stable elements of the microtubule cytoskeleton that correspond to microtubule bundles.

Microtubules guide cellulose synthase distribution and behavior. We investigated the causal relation between microtubules and organization of cellulose deposition by two methods: observation of dynamic patterns of colocalization during organizational transitions and disruption of the cytoskeleton with the drug oryzalin. In hypocotyl cells just below the hook, microtubules are organized into dense arrays of transverse bands, and the YFP:CESA6 complexes show the same orientation (Fig. 2, A to F, H to J; fig. S3, A to C). About 20 min after stimulation of these cells with blue light (488-nm excitation laser), a marked reorganization of CFP-labeled microtubules was observed, from a net transverse to a net longitudinal orientation (∼30 observations in 30 plants) (Fig. 2, I and L); this response is consistent with previous observations of blue light–stimulated reorientation of the cortical microtubule array in pea hypocotyls (24). Reorientation was not observed in control specimens mounted for observation but kept in the dark. This stimulated rearrangement of microtubules allowed us to examine the coupling between the cytoskeletal and cellulose synthase arrays. Dual label imaging revealed that the arrays of microtubules and YFP:CESA6 changed orientation concurrently (recordings of eight cells in eight plants) (Fig. 2, H to M) (Movie S4). Assembly of new microtubule tracks was observed to precede the appearance of linear arrays of YFP:CESA6 protein at the same position (Fig. 3). Correlated changes in the arrangement of microtubules and YFP:CESA6 localization were also observed along the axis of the hypocotyl in the course of normal development (fig. S3).

If cortical microtubule organization plays a role in establishing YFP:CESA6 organization, then global disruption of microtubules by destabilizing drugs should change the patterns of YFP:CESA6 distribution and movement. Treatment of intact seedlings with 10 μM oryzalin for 3 hours abolished the microtubule arrays in hypocotyl cells and caused marked changes in YFP:CESA6 organization and behavior, but did not deplete the CESA particles from the membrane (∼100 cells observed in 20 plants) (Fig. 4). YFP:CESA6 particles continued to translocate, but particles aggregated and moved in swarms that were not seen in untreated cells. Most of the “resistant” YFP:CESA6 tracks overlapped with oryzalin-resistant microtubules. In some cases, YFP:CESA6 tracks could be seen continuing past the end of microtubules (Fig. 4H). We infer that microtubules confer orientation on the movement of the CESA complexes but are not required for CESA motility per se. When microtubules were nearly completely depopulated from the cortex of etiolated hypocotyl cells (20 μM oryzalin for 7 hours; ∼30 cells observed in 16 plants), the pattern of CESA distribution changed. Rather than accumulating in dense swarms after partial loss of the cortical array, YFP:CESA was much more uniformly dispersed (compare Fig. 4, A and I). Notably, rosettes in these cells moved in roughly linear and parallel tracks set at oblique angles to the cell axis (Fig. 4I). Thus, in the course of disrupting the cortical cytoskeleton with oryzalin, the pattern of YFP:CESA6 distribution and trajectories in the membrane makes a transition through three distinct states. In the absence of oryzalin, YFP:CESA6 is distributed in a highly organized state defined by the organization of the intact cortical microtubule array. As the cytoskeleton becomes partially disrupted, YFP:CESA6 distribution and trajectories change markedly, but continue to be dominated by the influence of a small number of microtubules. When the cortical array is nearly completely disassembled, YFP:CESA6 acquires a second state of high organization in which particles are again more uniformly dispersed and trajectories appear to be more uniformly aligned.

Fig. 4.

Microtubule depolymerization changes YFP:CESA6 organization. Gray images: YFP:CES6, average of 5 frames; green images: YFP:CES6, average of 30 frames; and red images: CFP:TUA1, average of 30 frames. Image acquisition rates are as in Fig. 1. (A to D) Control treatment with 0.02% methanol for 3.5 hours. (E to H) Treatment with 10 μ Moryzalin (in 0.02% methanol) for 3.6 hours causes loss of microtubule organization with correlated changes in YFP:CESA6 localization (E) and trajectories (F). (I to K) Near-complete loss of cortical microtubules after treatment with 20 μM oryzalin for 7.6 hours. YFP:CESA6 is more uniformly dispersed (I) than in controls or cells with partial cortical arrays, and particle trajectories are highly cooriented (J). (D) Merge of (B) and (C). (H) Merge of (F) and (G), where (G) has been background subtracted (31). White arrowhead in (H) indicates CESA label extending beyond a defined microtubule track. Colored arrowhead in (I) and (J) highlight organized tracks of YFP:CESA6 in the absence of microtubules. Scale bar, 10 μm.

Conclusions and discussion. Taken together, these observations demonstrate that cellulose synthase complexes containing CESA6 are organized in the cell membrane by a functional association with cortical microtubules. The distribution and movement of YFP:CESA6 along trajectories defined by discordant and curved microtubules, and the high level of coordination between YFP:CESA6 and microtubules observed during reorganization events, show that CESA localization and guidance display tight spatial and temporal coupling to microtubules, and therefore, that the two arrays of molecules are likely to be in intimate contact with each other. These observations effectively rule out a model in which CESA complexes are guided solely by passive channeling between the optically resolved microtubules of the cortical array. One possibility is that each cortical microtubule, or microtubule bundle, allows for lateral interaction with the cytosolic domains of CESA complexes, leading to organization of two linear arrays of CESA, one on either side of the microtubule. This model is consistent with freeze-etching experiments that revealed rosettes to lie alongside, but not directly on top of, cortical microtubules (8). The arrangement of rosettes into two linear arrays along each microtubule also suggests possible mechanisms to account for bidirectional movement along individual CESA tracks. Our observations of normal rates of rosette movement following microtubule depolymerization demonstrate that any interactions with microtubules, whether by direct contact of CESA protein or through linker proteins, are not required for cellulose synthase motility and further support the idea that the motive force for complex motility is provided primarily by cellulose polymerization (25, 26). Our results also show that, although cortical microtubules have a defining influence on the distribution and guidance of CESA rosettes if they are present, in the absence of cortical microtubules, the movement of CESA complexes does not appear to be random, suggesting an intrinsic capacity to self-organize (12) or the action of a second extrinsic organizational mechanism. This surprising observation may explain previous experiments that did not support a causal relation between microtubules and the orientation of cellulose deposition (14, 27).

Previous experiments that challenged the microtubule alignment model for CESA guidance should be reexamined with live-cell confocal microscopy and with careful attention paid to the physiological and developmental stage of the tissue. In this respect, we emphasize that we have examined the localization of only a single type of CESA protein in one tissue type. It could be that other CESA proteins do not have localization patterns that are so tightly coupled to the cytoskeleton, that the behavior of CESA proteins varies within different organs, or that only a subset of the microtubules is involved in guidance in some cell types and that these were overlooked in some previous studies.

In vitro measurements of cellulose synthesis have been fraught with technical difficulties, and substantial effort is required to demonstrate enzyme activity in tissue extracts (28). The methods and results described here create new opportunities to assay the effects of genetic, developmental, and environmental variation on cellulose synthesis. In principle, improved single-molecule optical methods may be used to measure not only the rate of synthesis but also the duration and orientation of deposition—factors that have profound effects on the physical properties of cellulose. Such assays may facilitate an understanding of the roles of genes such as COBRA and KORRIGAN, whose effects on cellulose synthesis and cell-grown anisotropy are poorly understood (29, 30).

Supporting Online Material

Materials and Methods

Figs. S1 to S3

Table S1

Movies S1 to S8


References and Notes

View Abstract

Navigate This Article