The roles of the cytoskeleton during cellulose deposition at the secondary cell wall


*(fax +44 161 275 3938; e-mail


During secondary cell wall formation, developing xylem vessels deposit cellulose at specific sites on the plasma membrane. Bands of cortical microtubules mark these sites and are believed to somehow orientate the cellulose synthase complexes. We have used live cell imaging on intact roots of Arabidopsis to explore the relationship between the microtubules, actin and the cellulose synthase complex during secondary cell wall formation. The cellulose synthase complexes are seen to form bands beneath sites of secondary wall synthesis. We find that their maintenance at these sites is dependent upon underlying bundles of microtubules which localize the cellulose synthase complex (CSC) to the edges of developing cell wall thickenings. Thick actin cables run along the long axis of the cells. These cables are essential for the rapid trafficking of complex-containing organelles around the cell. The CSCs appear to be delivered directly to sites of secondary cell wall synthesis and it is likely that transverse actin may mark these sites.


Cellulose is the major component of many plant cells walls and is considered to be the world’s most abundant biopolymer. In the primary cell walls of expanding cells, cellulose is essential in regulating both the extent and orientation of cell expansion and consequently is a major factor in determining the shape and morphology of plant cells (Cosgrove, 2005). As the largest component of woody secondary cell walls, cellulose is essential in providing mechanical stability to plants and serves as a major global sink for carbon. Consequently, cellulose has recently attracted much attention as a potential renewable source of biofuels with low net greenhouse gas emissions (Farrell et al., 2006).

Cellulose is synthesized by a large complex, the cellulose synthase complex (CSC) that moves through the plane of the plasma membrane. The complex has been visualized in freeze-fracture studies of the plasma membrane as a six-lobed structure known as a rosette. Measurements suggest that the complex has a diameter of 25–30 nm (Kimura et al., 1999; Mueller and Brown, 1980). Each complex is believed to simultaneously synthesize of the order of 36 linear chains of β-1,4-linked glucose that are extruded into the cell wall, where they pack together to form a rigid microfibril (Somerville, 2006). It is believed that the rigid cellulose microfibril is essentially stationary and that extension of the cellulose chains can only occur by movement of the CSC through the plasma membrane (Delmer and Amor, 1995). There are no analogous systems described in which such a large enzyme complex has to be moved through the plasma membrane in a specified direction.

Developing xylem vessels have been widely used for the study of secondary cell wall synthesis, since they rapidly deposit large amounts of cellulose, but only at specific sites on the plasma membrane. These sites contain a high density of CSCs relative to regions of the plasma membrane that lie between secondary wall thickenings (Herth, 1985). For more than 40 years it has been known that bands of cortical microtubules (MTs) mark these sites of secondary cell wall deposition (Hepler and Newcomb, 1964). These MTs are believed to direct cellulose synthesis to specific regions at the plasma membrane and control the orientation of cellulose deposition; however, the mechanism by which cortical MTs influence the orientation of cellulose deposition remains unclear. Several models, based upon observations for primary wall synthesis, suggest that the CSC is directly attached to cortical MTs or that the cortical MTs act as a physical barrier that constrains the complex to move between them (reviewed in Hepler and Palevitz, 1974). More recently, various studies with the temperature-sensitive mutant mor1 and with chemical inhibitors suggest that microfibril order can be maintained in the absence of MTs (Himmelspach et al., 2003). This is supported by recent experimental data showing that linear movement of the primary wall complex still occurs when the underlying MTs depolymerize (Paredez et al., 2006). Consequently, more recent models suggest that either MTs subtly influence nascent microfibril orientation or that they affect microfibril length, but are not essential to maintain the orientation of microfibrils (Baskin, 2001; Wasteneys, 2004).. For the secondary walls of developing xylem, immunolabelling studies suggest that MTs are continuously required to maintain proper localization of the CSC (Gardiner et al., 2003).

Very little is known about the role of actin in secondary wall deposition. Labelling of actin by fluorescein isothiocyanate (FITC)-conjugated phalloidin during xylem differentiation in Zinnia cell culture revealed the presence of thick filaments lying parallel to the long axis of the cell (Kobayashi et al., 1987). This longitudinal alignment of actin has also been seen in the xylem of fixed Arabidopsis seedlings (Gardiner et al., 2003). In developing Zinnia tracheary elements that will exhibit a spiral-like pattern of secondary wall deposition, actin filaments became localised transversely beneath the sites of future cell wall deposition, just as secondary cell wall deposition commences. In young differentiating cells exhibiting semi-reticulate patterned secondary walls, actin appears in patches in spaces between the MTs. As the cell walls thicken, the actin becomes more concentrated at the wall edges in a similar fashion to the MTs (Kobayashi et al., 1988). Treatment with the actin-depolymerizing drug cytochalasin B results in a failure in transverse alignment of MTs at the onset of differentiation, resulting in longitudinal secondary walls (Kobayashi et al., 1988). This suggests a role for actin in controlling correctly patterned deposition of the secondary cell wall.

This study uses live cell imaging to obtain a detailed picture of both actin and MT organization in developing xylem and their role in both intracellular trafficking and plasma membrane localization of the CSC.


Microtubules are arranged into thick bands beneath sites of secondary wall synthesis

In order to visualize the cortical MT array, the mouse MAP4 MT-binding domain fused to yellow fluorescent protein (YFP) (Marc et al., 1998), was used. Expression was driven using the promoter from the IRX3 gene, a marker for secondary cell wall formation that gives xylem-specific expression in roots (Figure 1). pIRX3::YFP-MBD (YFP-MBD) has recently been shown to encode an accurate reporter of MT behaviour that exhibits no effect on plant development and behaves in a manner comparable with YFP-tubulin reporters (Wightman and Turner, 2007). In developing roots, the youngest protoxylem vessels are always found closest to the apical meristem near the root tip, while more mature vessels are localized progressively further from the tip (Dolan et al., 1993). Consequently, the root may be viewed as a developmental series. At the earliest stages of vessel development MTs were visualized as narrow bands, orientated transversely relative to the long axis of the cell (Figure 1a). Later stages of vessel development were characterized by patterned secondary cell wall deposition that could be distinguished using brightfield optics (Figure 1d). These bands of secondary cell wall thickening became more obvious in older vessels (Figure 1f) and there appeared to be a good spatial correlation between bands of MTs and secondary cell wall deposition (Figures 1 and 2). In older vessels, bands of MTs appear to be composed of at least two distinct bundles with a narrow gap down the centre (Figures 1e and 2). This finding agrees with a previous study that examined MTs in Arabidopsis cell culture undergoing xylogenesis (Oda et al., 2005). In addition, these split bundles appear to flank the secondary cell wall thickening (Figure 2b).

Figure 1.

 Microtubule development in xylem.
A 5-day-old seedling of wild-type Arabidopsis expressing the pIRX3::YFP-MBD reporter. Images were taken at approximately equal distances from the root tip (a, b) to the termination of yellow fluorescent protein (YFP) fluorescence (e, f). Left panels show YFP and right panels show brightfield. Gap between split microtubules is labelled by the arrow. Bar = 10 μm.

Figure 2.

 Microtubules bundle and then appear to split.
(a) Yellow fluorescent protein (YFP)-microtubule (MT) reporter (false colour red) in a late-developing xylem where bundles appear to be divided. The filled arrow indicates a MT ‘bridge’ that exists between adjacent hoops. Barbed arrow shows a clear gap between the split bundles at the cell periphery where the fluorescence signal is much brighter. Bar = 10 μm.
(b) Microtubules with associated brightfield image. The right panel shows a merged image of divided MT regions flanking secondary wall thickenings. Bar = 1 μm.
(c) Three-dimensional restoration of deconvolved confocal images of the YFP-MT reporter (false colour yellow). The image is a snapshot from an interactive virtual reality movie where the developing xylem has been rotated and tilted (Movie S1). The arrowhead indicates a complete hoop at the termination of a spiral. Inset: close-up of two hoops showing MT bridges.

Using 3D restoration microscopy of confocal z-stacks, the bands of MTs could be seen to form discrete hoops and short spirals (Figure 2c and Movie S1). In addition, the ends of each spiral finished as a complete hoop (Figure 2c and Movie S1). Although, superficially, both spirals and hoops appear as discrete structures, the cortical MT network was found to be continuous, with bridges of either single MTs or narrow MT bundles linking the adjacent hoops and spirals (Figure 2a,c, inset).

Localization of the CSC to sites of secondary wall growth is dependent upon underlying MT bundles

To observe the CSC, the protein IRX3 was fused at its N-terminus to YFP and transformed into irx3-1 mutant plants, as described previously for GFP (Gardiner et al., 2003). To investigate the role of MTs, a MT-reporter fused to the cyan fluorescent protein (pIRX3::CFP-MBD) was introduced into the YFP-IRX3 line (pIRX3::YFP-IRX3). As shown in previous studies using immunofluorescence of fixed seedlings (Gardiner et al., 2003), at certain stages of vessel development YFP-IRX3 localized to transverse bands that appear to be part of discrete hoops or discrete spirals (Figure 3a). These bands represent peripheral localization around the circumference of the developing xylem vessel. In Figure S1, a developing xylem vessel is in longitudinal section that is slightly tilted relative to the focal plane. Consequently the position of the focal plane within the vessel differs along its length. As the plane of the section passes through the centre of a vessel, YFP-IRX3 is seen to be confined to a region directly beneath the cell indentations resulting from secondary wall formation (Figure S1, inset). The banded pattern of YFP-IRX3 expression is very similar to that observed for the MTs (Figure 3a,b). In addition, YFP-IRX3 was seen in intracellular organelles of the xylem (Figure 3a).

Figure 3.

 Cellulose synthase complexes (CSCs) at the cell surface are associated with thick microtubule (MT) bundles.
(a) Yellow fluorescent protein (YFP)-IRX3 labelling in xylem. The YFP can be seen to label bands and intracellular organelles. Image spans two vessels divided by a perforation plate (vertical dashed line), older developing xylem on the left and younger xylem on the right. Bands of CSCs appear to be split in older developing xylem. Contrast enhancement and a 2-pixel-radius Gaussian filter of three split bands are shown for clarity (inset).
(b) The same developing xylem showing MTs labelled with the cyan fluorescent protein (CFP) reporter. Bar = 10 μm.
(c) Correlation plot comparing YFP-IRX3 band widths with those of CFP-MTs in older and younger developing xylem. Measurements were taken along a line that runs along the long central axis of the cells. R2 = 0.86.

Two xylem vessels at different stages of development were observed. The younger vessel was characterized by relatively narrow bands of YFP-IRX3 and MTs compared with the older vessel (Figure 3a,b). In the older vessel, YFP-IRX3 bands can be resolved into at least two distinct regions (Figure 3a, inset), in an identical manner to that previously observed for using the YFP-MT reporter (Figure 2). However, for the CFP-MT reporter, bands cannot be resolved into distinct regions (compare Figures 2b and 3b). This may be due to a small amount of cell wall autofluorescence filling in the area between the two separated regions when using the CFP filter set during microscopy. There is an excellent correlation between the width of the MT bands and the width of the bands of YFP-IRX3 (Figure 3c). No YFP-IRX3 fluorescence was observed between the bands that might correspond to the narrow bridges that link MT bands (Figure 2). Together these data suggest that the bands representing thick MT bundles tightly constrain the CSC to regions of the plasma membrane underlying sites of secondary wall deposition.

To investigate further the extent of CSC dependence upon the MT bundles, a MT-depolymerizing herbicide, oryzalin, was added to seedlings containing YFP-IRX3 and the CFP-MT reporter. Loss of surface-localized banding of YFP-IRX3 occurred concurrently with the loss of CFP-labelled MT bundles within 45 min of the addition of oryzalin (Figure 4a and Movie S2). A small noticeable effect was first seen within 15 min after the addition of oryzalin (Movie S2). The time-course of oryzalin-induced loss of MT and CSC banding is identical to that previously reported for immunofluorescence (Gardiner et al., 2003). YFP-IRX3 was only seen at the plasma membrane wherever MTs persisted (Figure 4b). No banding of CSCs, as observed by the lack of fluorescence from YFP at these sites, was found in the absence of MTs. These data demonstrate that in developing xylem vessels the MTs are required for maintenance of the CSC at the sites of cell wall synthesis.

Figure 4.

 Localization of the cellulose synthase complex (CSC) is microtubule (MT) dependent.
(a) Bands of CSCs are observed to disappear after MT depolymerization using oryzalin (Orz).
(b) Two different regions of the root from a 5-day-old seedling, treated for 40 min with oryzalin, are shown (upper and lower panels). Merged images (yellow fluorescent protein, green; cyan fluorescent protein, red) are shown in the right panels. Arrows indicate CSCs associated with MT bundles. Arrowheads indicate transport organelles. Bars = 5 μm. For comparison developing xylem of untreated seedlings is shown in Figure S2.

Treatment of seedlings with the cellulose synthesis inhibitor isoxaben has been shown to result in the disappearance of plasma-membrane CSCs from sites of primary wall synthesis (Paredez et al., 2006). To determine the effect of isoxaben upon the secondary wall CSC, short image sequences were taken at 10-min intervals of isoxaben-treated seedlings containing the YFP-IRX3 fusion. In contrast to what is observed for untreated seedlings (Figure S2), bands of YFP-IRX3 in treated seedlings were seen to gradually disappear until only internal organelles were visible moving around the cell (Figure 5a and Movie S3). In Figure 5(b), three clearly visible bands of YFP-IRX3 were still visible shortly after the addition of isoxaben. Seven minutes later the bands are much less well defined and have almost disappeared by +8 min (Figure 5b, right panel).

Figure 5.

 Loss of YFP-IRX3 bands after isoxaben treatment.
(a) A developing xylem cell before (left panel) and 30 min after isoxaben treatment (right panel).
(b) Three bands (arrows) containing YFP-IRX3 in a seedling mounted in Isoxaben (left panel). Images of the bands taken after a further 7 min (middle panel) and 8 min (right panel) show the bands to be barely visible (arrows). Bars = 5 μm.

CSC-containing organelles move unidirectionally around the xylem and pause at sites of wall synthesis

Organelles labelled with YFP, and therefore containing CSCs, were observed to move around the cell in a single direction. Movie S4 is a time-lapse series where 7-ms exposures are constantly streamed from an electron-multiplying charge-coupled device (CCD) camera. This rapid acquisition allows individual organelles to be tracked during the time lapse. The upper exposure limit for useful discrimination and tracking of organelles was found to be 650 msec. For time-lapse movies where frame intervals could be determined, organelles were observed to move at maximum speeds of more than 2.6 μm sec−1. By projecting a stack of images from the time-lapse series, the organelles are observed to follow the same path around the cell (Figure 6a). Once an organelle reaches the end of the cell, it crosses over to the opposite side and moves away in the opposite direction (Figure 6b and Movie S4). The speed of organelle movement was not observed to be linear; instead, particle tracking showed organelles to pause. The path taken by a single organelle during particle tracking is shown in Figure 7(a) and Movie S5. This path crossed several bands of YFP-IRX3, representing regions of cellulose deposition at the cell wall. A plot of total distance travelled against time shows the organelle to frequently pause, and these pause sites correspond to the bands of secondary cell wall deposition that are marked by the transverse bands of YFP-IRX3 fluorescence (Figure 7b). Incidences of organelles pausing at these sites cannot be explained solely by an obstruction, generated by the secondary wall thickenings protruding into the cytoplasm. In Movie S6, several organelles are tracked along the same region of a developing xylem. The organelle labelled ‘blue’ pauses at a band for approximately 3 sec before proceeding along the length of the developing xylem cell. Subsequent organelles move along the same track but are not observed to pause at the same location, suggesting a pausing mechanism that does not rely on physical obstruction of the organelle.

Figure 6.

 Unidirectional behaviour of organelle transport.
(a) Standard deviation projection of a 1400-frame time lapse of YFP-IRX3 taken at 7-msec intervals. Arrows show the direction of organelle movement.
(b) Tracking of a single organelle at the end of the developing xylem cell. The cell terminus (dashed white line) and the path taken by the organelle (white line) are shown. Bar = 10 μm.

Figure 7.

 Organelles pause at sites of wall synthesis.
(a) Average projection of a time-lapse sequence (shown as Movie S5). The path of a tracked organelle is shown (dashed line). Bar = 10 μm.
(b) Distance travelled by the organelle during the time-lapse sequence. A density scan of grey levels is shown on the y-axis. Peaks (density scan) and arrows show the positions of bands of YFP-IRX3 and are seen to correspond to organelle pause events.

Statistical data relating to pausing of organelles are presented in Table 1. As a proportion of the total tracking period, the time spent during pause varies greatly between different organelles. However, the majority of pause events are located at bands of CSCs. The data also show that there appears to be a level of discrimination between pauses at individual bands, since some organelles stop at relatively few bands during their journey along the cell.

Table 1.   The frequency and location of organelle pause events in developing xylem. Tracking data were obtained for nine randomly chosen organelles
Organelle pause (% total time during tracking)
 Min. = 14
 Max. = 64
 Mean = 40
Pause event occurring at bands (% total pause)
 Min. = 72
 Max. = 100
 Mean = 90
Percentage of bands at which a pause event was observed
 Min. = 18
 Max. = 100
 Mean = 62

Figure 8(a) shows a region of a developing xylem where the bands of YFP-IRX3 can be seen in the average projection of Movie S7. Particle tracking shows the organelle to pass beneath two bands before pausing at the third band for 4 sec (Figure 8b and Movie S7). The pause event is accompanied by an increase in fluorescence of the overlying band of YFP-IRX3 from an average pixel intensity of 3.5 (Figure 8a, time point 1) to an average intensity of 19.5 (Figure 8a, time point 2 and Figure 8c).

Figure 8.

 Distribution of fluorescence around a band after organelle pause.
(a) A tracked organelle (white arrow) is seen to pause at a band (middle panel) where average fluorescence intensity, measured within the area outlined by the dashed box increases from a value of 3.5 intensity units (time point 1, left panel) to a value of 19.5 (time point 2, middle panel) intensity units. The maximum possible intensity for any pixel is 255. The tracked organelle reaches a maximum of 243 intensity units during the time-lapse sequence. Bar = 10 μm.
(b) Distance travelled by the organelle during the time-lapse sequence. The organelle is seen to pause at the band for 4 sec. The arrow indicates the position of the fluorescence peak shown in (C).
(c) Plot of average intensity of the dashed box during the time-lapse sequence. The peak of 19.5 intensity units is labelled by the arrow.

During cellulose deposition at the primary wall, a subset of organelles containing either the primary wall CSC or another enzyme involved in cellulose biosynthesis, called Korrigan, have been shown to represent the Golgi apparatus (Paredez et al., 2006; Robert et al., 2005). This indicates the rapidly moving YFP-IRX3-containing organelles within the xylem may also represent Golgi. To identify these organelles, a xylem-specific Golgi reporter, consisting of CFP fused to mannosidase I (pIRX3::ManI-CFP), was introduced into the YFP-IRX3 line. The co-localization data are shown in Movie S8 and Figure S1 and show a subset of YFP-labelled organelles to co-localize with the CFP marker, suggesting that these organelles are Golgi.

Organelle movement is actin dependent

A singular and unidirectional path for organelle transport suggests a role for some part of the cytoskeleton. Disruption of the MT array by oryzalin treatment does not stop organelle movement (data not shown). However, treatment with the actin-depolymerizing drug lantrunculin B was found to completely stop organelle movement (Movie S9). Additionally, the cessation of organelle movement was tightly linked to the disappearance of bands of YFP-IRX3 at the sites of wall synthesis (Figure S3a). For comparison, a developing xylem from an untreated seedling is shown in Movie S10.

The mechanism of CSC delivery was further investigated by constructing a reporter consisting of the mCherry red fluorescent protein fused to the actin-binding domain2 (fABD2) of the AtFIM1 protein. Fusion of fABD2 to fluorescent proteins has been shown to be a good reporter of actin behaviour (Sheahan et al., 2004; Wang et al., 2004). Using the pIRX3::mCherry-ABD2 reporter, thick actin cables were easily visualized; these ran in pairs, longitudinally along the length of the cell (Figure 9a). These data agree with the longitudinal actin observed by immunofluorescence of fixed Arabidopsis seedlings (Gardiner et al., 2003). The actin cables were seen to be arranged at opposite sides of the cell (Figure 9a). Time-lapse acquisitions showed YFP-IRX3-containing organelles to move along these cables (Movie S11). Taken together, these data show that organelle transport around the xylem is an actin-dependent process.

Figure 9.

 Distribution of actin in developing xylem.
(a) A pair of actin cables (arrowheads) running longitudinally along a developing xylem cell as visualized by a pIRX3::mCherry-fABD2 reporter. The focal plane is within the developing xylem. Bar = 10 μm.
(b) Developing xylem cell at an early stage of secondary wall synthesis exhibiting faint YFP-IRX3 banding. The focal plane is at the upper surface of the developing xylem.
(c) Faint transverse actin fibres are visualized by the mCherry-FABD reporter. The positions of five transverse actin fibres are shown by dashed lines in (b). Bar = 5 μm.
(d) Transverse actin (tA) appears to emanate from a thick actin cable (AC) running along the length of the cell. Actin bridging the cable and the narrow transverse bundle is shown by arrowheads. Bar = 5 μm.
(e) Oryzalin-treated cell showing persistence of transverse actin (examples marked by arrowheads). Bar = 5 μm.

Transverse actin is associated with sites of wall synthesis

One candidate for marking the pause sites for the CSC-containing organelles is the thick MT bundle found beneath the bands of YFP-IRX3. Removal of the bands of MT bundles at the sites of cell wall synthesis would be expected to result in a loss of pause events in addition to the loss of YFP-IRX3 from sites of wall synthesis. To test this hypothesis, the effect upon pause events of the MT-depolymerizing drug oryzalin was observed for seedlings containing the CFP-MT reporter in the YFP-IRX3 background (Figure 10). After MT depolymerization (Figure 10a), time-lapse microscopy coupled with particle tracking of four organelles showed that pause events continued (Figure 10b). A more global view of organelle movement is shown in the kymograph of Figure S4 where jagged lines indicate organelle pause. In addition, stationary organelles remained at regular intervals, as shown by vertical lines in the kymograph. The data suggest that MTs are not essential in marking the sites for CSC delivery.

Figure 10.

 The effect of microtubule depolymerization upon organelle pause.
(a) Cyan fluorescent protein (CFP)-microtubules (MTs) of a developing xylem cell taken before (left panel) and after addition of oryzalin (right panel). Bar = 10 μm.
(b) Four organelles tracked before (left plot) and after (right plot) oryzalin treatment. Examples of pause events are shown by arrows.

An alternative marker for CSC delivery may be polymerized actin lying beneath the sites of cell wall synthesis (Kobayashi et al., 1987). As only the thick actin cables described above have so far been observed in live imaging techniques, transverse actin may be arranged as either single fibres or narrow bundles and therefore may yield a weaker fluorescence signal from the mCherry reporter. Using a sensitive confocal microscope, accumulation of fluorescence signal from multiple confocal scans revealed transverse actin to be barely visible in addition to the strongly fluorescent longitudinal cables. The existence of transverse actin in the developing xylem was confirmed by immunolabelling using an anti-actin antibody (Figure S3b). In Figure 9(b) a young developing xylem at an early stage of secondary wall synthesis, as visualized by the faint bands of YFP-IRX3, is seen to possess transverse narrow actin bundles, as visualized by the mCherry reporter, that are located within or at the side of bands of YFP-IRX3 (Figure 9b,c). The locations of several actin bundles are drawn as dashed lines on the bands of YFP-IRX3 (Figure 9b). The narrow actin bundles do not appear to span the width of a band, which is in stark contrast to the thick MT bundles (Figure 3). Therefore, the narrow bundles are located within the vicinity of sites of secondary wall synthesis but do not always co-localize with the bands of YFP-IRX3. The fibres that make up these narrow bundles appear to originate from polymerized actin within the longitudinal cables (Figure 9d), indicating the existence of a continuous actin network between longitudinal and transverse actin. Transverse actin is a good candidate for marking the pause sites for moving organelles as it persists in the xylem of oryzalin-treated cells (Figure 9e).


Thick MT bundles maintain CSCs beneath sites of secondary wall synthesis

Using live cell imaging on roots of intact Arabidopsis seedlings, microtubules were observed to form thick bands beneath the sites of secondary wall synthesis in the developing xylem. These bands, representing thick bundled MTs, were linked by individual MTs or very narrow bundles. The CSC, fused to the YFP reporter, co-localized with the MT bands, suggesting that CSCs are either absent at the sites of narrow MT bridges or present at densities too low for detection by fluorescence microscopy. No secondary cell wall material has been reported between the characteristic rings and spirals of the protoxylem, despite the many studies using transmission electron microscopy (TEM), suggesting a clear distinction between the thick MT bundles and the narrow bridges (Figure S5).

In older developing xylem, the MT bands were observed to divide in two (Figures 2 and 3), as previously observed in cell culture (Oda et al., 2005). This suggests that thickening of MT bundles, followed by separation, will result in cellulose synthesis being directed to the edges of nascent secondary walls during the later stages of development. Consistent with this observation, Oda et al. (2005) found that differentiating cells of a taxol-treated culture possessed a narrower gap between the two MT bundles when compared with control cells. This reduced distance between the split bundles resulted in narrower secondary walls.

Previous work using immunofluorescence of fixed seedlings showed that MTs are required for localization of the CSC to sites of secondary cell wall synthesis (Gardiner et al., 2003). Here we used live-cell imaging to look at the effect of loss of MT bundles upon the CSCs. We saw no evidence of cell surface localization of CSCs without underlying MTs. Instead, in the absence of MTs the bands of CSCs disappeared, suggesting that for secondary wall synthesis MTs are continuously required to maintain CSCs to sites of cellulose synthesis at the plasma membrane.

Delivery of CSCs requires actin

Organelles containing YFP-IRX3 were observed to move rapidly in a single direction around the cells of developing xylem. The velocities of these organelles were in excess of 2.6 μm sec−1 and are similar to those velocities observed for Golgi stacks in tobacco BY2 cells (Nebenführ et al., 1999). Some of the CSC-containing organelles were identified as Golgi due to their co-localization with a xylem-specific Golgi marker and is consistent with observations made during live cell imaging of the primary wall complex and the Korrigan endoglucanase (Paredez et al., 2006; Robert et al., 2005). In addition to Golgi, Korrigan has been found to label late endosomes. Some of the CSC-containing organelles may also label this compartment as CSCs from cotton fibres have been shown to possess short half-lives measured in minutes (Jacob-Wilk et al., 2006). Additionally, CesA proteins that are not part of the CSC are believed to be targeted for degradation (Taylor, 2007). Bands of YFP-IRX3 in the developing xylem are fainter and harder to visualize than for YFP-IRX3-containing organelles, suggesting complexes do not stay at the plasma membrane indefinitely. By visualizing an actin reporter, consisting of the fluorescent mCherry protein fused to an actin-binding domain, thick cables running along the length of the cells were seen to ferry the organelles through the cytoplasm. The organelles pause at sites of secondary wall synthesis and can be sometimes observed to result in an increase in fluorescence of the banded YFP-IRX3. This increase in fluorescence intensity could represent an offloading event of CSCs either directly into the plasma membrane or through the rapid distribution of small transport vesicles. Organelle pause does not occur at every band, suggesting there is targeting of organelles to particular bands along the developing xylem and may reflect the requirement for replenishment of the CSC at the plasma membrane. The incidences of pause occur in the presence of oryzalin, indicating that MTs may not play an essential role in marking sites for CSC delivery. Instead faintly labelled actin fibres or narrow actin bundles orientated transversely and positioned close to the sites of secondary wall synthesis were observed and were seen to persist in the absence of MTs after treatment with oryzalin.

This transverse actin seems to originate from the longitudinal cables, and organelle pause may depend on the fate of the attached fibre within the cable; organelles may only pause if the actin fibre exits the cable to form a narrow transverse bundle. The lack of correlation between the widths of the bands of YFP-IRX3 and the location of transverse actin (Figure 9c) suggests that actin does not have a role once CSCs are at the plasma membrane, unlike the MTs which span the entire width of the bands (Figure 3c). However, disruption of actin by lantrunculin B results in loss of the bands of CSCs in a similar fashion as seen for MT depolymerization in the presence of oryzalin. The similar loss of bands could be due to different effects of the inhibitors. The tight correlation between MTs and the membrane-bound CSCs suggest that oryzalin treatment results in CSCs no longer being maintained at the plasma membrane. In contrast, the actin does not exhibit the same tight correlation with CSC localization, and depolymerization of transverse actin may result in loss of bands through lack of replenishment of CSCs from the organelles.

A model for cellulose synthesis at the secondary wall

A model illustrating the organization of the cytoskeleton and how it affects the delivery and maintenance of CSCs to the sites of secondary wall synthesis within the developing xylem is shown in Figure 11. Actin cables transport the CSC-containing organelles around the cell which pause at sites marked by transverse actin. The CSCs incorporated into the plasma membrane are maintained beneath sites of wall synthesis by MT bundles. Hogetsu (1991) proposed that the MTs maintain a partitioned plasma membrane beneath the site of secondary wall synthesis. Such partitions may contain highly fluid membrane, facilitating the movement of the CSC. In our model, the MTs also act as barriers preventing CSCs leaving the partitions. As the secondary wall thickens, the MT bundles split and thus maintain cellulose synthesis towards the edges of the walls. Finally, endocytosis of spent or non-functional CSCs is balanced by the entry of new or recycled complexes.

Figure 11.

 Model showing the role of the cytoskeleton in cellulose synthesis in developing xylem.
A schematic of a portion of a developing xylem cell is shown. A longitudinal cross section of an area surrounding a secondary cell wall thickening is shown within the dashed box. AC, actin cable; tAF, transverse actin fibre; G, yellow fluorescent protein-containing organelle, arrows next to organelles indicate the direction of movement; SCW, secondary cell wall; PM, plasma membrane; V, transport vesicle; MTB, microtubule bundle which is made up of individual microtubules, MT; CSC, cellulose synthase complex; CM, cellulose microfibril which is made up of multiple glucan chains.

Experimental procedures

Plant growth conditions for microscopy

Arabidopsis thaliana ecotype Landsberg was used throughout this study. Seedlings were surface sterilized in 30% (v/v) sodium hypochlorite solution containing 1% (v/v) Triton X-100 and washed five times in distilled water. Seeds were resuspended in 0.1% (w/v) Agar (Sigma, and spotted in lines on growth medium containing 2.2 g l−1 MS supplemented with Gamborg B5 vitamins (Duchefa,, 1.5% (w/v) Agar, pH 6.8. Seeds were vernalized for 3 days at 4°C and then placed upright in a Gallenkamp environmental test chamber and incubated under constant light (∼100 μmol m−2 sec−1) at 22°C. Whole seedlings were viewed under the microscope after 5 days’ growth.

Fluorescent protein fusions

The CSC was labelled with the yellow fluorescent protein EYFP (Clontech, The EYFP coding sequence minus the STOP codon was amplified by PCR using primers that appended NheI sites at either terminus. NheI-digested EYFP was then cloned into NheI-cut pCS12, an IRX3 genomic fragment which includes 1.75 kb of promoter sequence in pBluescript KS (Taylor et al., 1999). The resultant N-terminal fusion was subcloned, on a XhoI/SacI-digested fragment, into SalI/SacI-digested pCB2300 (Cambia, and transformed into irx3 mutant plants by vacuum infiltration (Clough and Bent, 1998). A transformed line yielding medium levels of YFP fluorescence emission (RWY6a) was found to fully rescue the irx3 mutant phenotype. To label microtubules, EYFP or ECFP (Cyan, Clontech) were fused N-terminally to the 1.25 kb MT-binding domain (MBD) of the mouse MAP4 gene as previously described (Wightman and Turner, 2007). The EYFP-MBD fusion was transformed into wild-type plants and a YFP-expressing line, called RWYM4 was identified. The ECFP-MBD fusion was transformed into line RWY6a to give RWYC2.

To label actin, a region of FIM1 encoding actin-binding domain 2 (fABD2; Sheahan et al., 2004) was amplified from a pSPORT Arabidopsis cDNA library (LeClere and Bartel, 2001) using primers FABDmCher2fwd (5’-CGGCGGCATGGACGAGCTGTACAAGGATCCTCTTGAAAGAGCTGAATTGG-3’) and FABD2-2rev (5’- GCCTCTAGATCATGACTCGATGGATGCTTCC-3’). The gene encoding the mCherry fluorescent protein (Shaner et al., 2004) minus the STOP codon was amplified using primers CherryCLA (5’-GCCATCGATATGGTGAGCAAGGGCGAGGAGGATAAC-3’) and mCherFABDrev (5’-CCAATTCAGCTCTTTCAAGAGGATCCTTGTACAGCTCGTCCATGCCGCCG-3’). The fABD2 and mCherry DNA were then purified and used as templates in a single PCR with 0.4 μm of primers CherryCLA and FABD2-2rev and a 1:9 mixture of Pfu polymerase and Bio-X-Act polymerase (Bioline, Reaction conditions were the same as for the FP-MBD construct described in Wightman and Turner (2007). A 1.8 kb fragment was gel purified, digested with ClaI/XbaI and cloned into ClaI/XbaI-cut pBluescript-EXFP-MBD-NosT (Wightman and Turner, 2007). The construct was subcloned as a SacI/KpnI DNA fragment into pCB1300 and transformed into line RWY6a to give RWYmCA.

To label Golgi, a construct, pAN57, containing a mannosidase-CFP fusion was obtained as a gift from Andreas Nebenführ, University of Tennessee, USA (Nebenführ et al., 1999). The 0.8 kb SacI/NheI DNA fragment of pAN57 containing the 35S promoter was swapped for a SacI/NheI fragment containing the IRX3 promoter for xylem-specific expression. The construct was subcloned as a SacI/KpnI DNA fragment into pCB1300 and transformed into line RWY6a to give RWYGolgi1.


Microscopy was carried out using the following imaging systems: A Leica DMR microscope fitted with a SPOT Xplorer 4MP camera (Diagnostic Instruments, and equipped with either a HCX PL APO ×100 oil numerical aperture (NA) 1.4–0.7 objective (part no. 506220, Leica Microsystems,; Figures 1, 2a,b, 3, 4a, 9a, Movie S11 and Figure S1) or a HCX PL APO CS ×63 water NA 1.2 objective (part no. 11506212; Figures 4b, 7, 8 and Movies S5 and S7). The gain on the camera was set to either 8 ×  or 16 × . For the image sequence in Figure 8, the gain was set to maximum (64 × ) and intensity values were background subtracted. A Leica DM5500 fitted with a Photometrics Cascade II 512B EMCCD camera (Photometrics UK) and a 63 ×  water objective (Figures 5a, 6 and 10, Figure S3a and Movies S3, S4, S6, S9 and S10). A Leica ASMDW live-cell imaging system equipped with a xenon bulb for fluorescence, 63 ×  water objective and Cascade II camera (Movie S3). For the epifluorescence microscopes the filter cubes were all supplied by Chroma Technology Corp. ( YFP (part no. 41028), CFP (31044v2) mCherry (41043) and YFP/mCherry dual (51019). Confocal microscopy was carried out on a Leica SP2 TCS confocal equipped with a 63× water objective (Figure 2c and Movie S1) and a Leica SP5 TCS upright confocal equipped with both oil and water 63× objectives (Figures 5b, 9b–e, Figures S1B and S3B and Movie S8). Oryzalin (Riedel-de-Haen; was used at a working concentration of 10 μm in 0.1% DMSO; isoxaben (Riedel-de-Haen) at 10 μm in 0.1% DMSO; lantrunculin B (Calbiochem, at 1 μm in 0.1% DMSO. All drugs were flushed under the coverslip of mounted seedlings. Deconvolution was carried out using Volocity software (Improvision, Images have been contrast enhanced to aid visualization using ImageJ to yield no more than 0.5% saturated pixels.

Observation of narrow transverse actin

Actin labelled by the mCherry-fABD2 reporter was visualized using the SP5 confocal and water objective. The power on the HeNe 543 and 594 laser lines was set to maximum and between four and six scans were accumulated for each section of a software optimized z-series. The z-series was processed in ImageJ (W. Rasband, National Institutes of Health) using the walking average plugin (J. Rietdorf and A. Seitz, EMBL, Heidelberg) set to a 3-z average. Scans that encompassed the upper surface of the xylem cell were then averaged using the Z-project command of ImageJ. Immunolabelling of actin was carried out according to the protocol described by Gardiner et al. (2003).

Tracking organelle movement

Organelles were tracked using the Manual Tracking plug-in of ImageJ (F. Cordelieres, Institut Curie, Orsay, France). For Table 1, three tracks from each of three seedlings (nine tracks in total) were determined. Each time point was scored for location, labelled as ‘band’ or ‘no band’. To take into account errors in locating the same point of the organelle in each frame, a pause event was defined as movement of less than 1 pixel (255 nm) per frame.


The authors wish to thank Thomas Nuhse for proofreading the manuscript. This work was supported by the Biotechnology and Biological Sciences Research Council, grant no. 34/C19282 and a grant from The Gatsby Foundation.