The Arabidopsis COBRA protein facilitates cellulose crystallization at the plasma membrane

Background: The COBRA gene is highly coexpressed with cellulose synthase genes, but its function remains unclear. Results: COBRAlocalizesattheplasmamembraneandbindsglucanchains.NMRstudiesindicatestructuraldefectsincellulose in the mutant despite normal polymerization rate. Conclusion: COBRA functions downstream of cellulose biosynthesis. Significance: This work suggests that alignment of glucan chains into cellulose fibrils is facilitated by one or more proteins. Mutations in the Arabidopsis COBRA gene lead to defects in we present that COBRA localizes to discrete particles in the plasma membrane and is sensitive to inhibitors of cellulose synthesis, suggesting that COBRA and the cellulose synthase in close proximity on the plasma membrane.

Mutations in the Arabidopsis COBRA gene lead to defects in cellulose synthesis but the function of COBRA is unknown. Here we present evidence that COBRA localizes to discrete particles in the plasma membrane and is sensitive to inhibitors of cellulose synthesis, suggesting that COBRA and the cellulose synthase complex reside in close proximity on the plasma membrane. Live-cell imaging of cellulose synthesis indicated that, once initiated, cellulose synthesis appeared to proceed normally in the cobra mutant. Using isothermal calorimetry, COBRA was found to bind individual ␤1-4-linked glucan chains with a K D of 3.2 M. Competition assays suggests that COBRA binds individual ␤1-4-linked glucan chains with higher affinity than crystalline cellulose. Solid-state nuclear magnetic resonance studies of the cell wall of the cobra mutant also indicated that, in addition to decreases in cellulose amount, the properties of the cellulose fibrils and other cell wall polymers differed from wild type by being less crystalline and having an increased number of reducing ends. We interpret the available evidence as suggesting that COBRA facilitates cellulose crystallization from the emerging ␤1-4-glucan chains by acting as a "polysaccharide chaperone." Cellulose microfibrils comprise the core component of the cell walls that surround each plant cell, allowing a mode of growth based on turgor pressure (1). In higher plants, cellulose is synthesized at the plasma membrane by the cellulose synthase complex (2,3), which is comprised of multiple cellulose synthase proteins (CESA), 2 although their exact stoichiometry is unknown. Upon activation, the CESA proteins synthesize ␤1-4-glucan chains (4), which align with each other, and by hydrogen bonding, form the cellulose fibril (3). Cellulose synthase complex movement at the plasma membrane has been visualized by live imaging (5), and a connection to cortical microtubules via the CSI1 protein has been demonstrated (2,6).
The mechanism by which individual glucan chains form cellulose microfibrils remains elusive. Because it is challenging to reconstruct active plant cellulose synthesis in vitro, genetic methods have been used to identify and characterize the genes for proteins that participate in cellulose synthesis and deposition (7). KORRIGAN was one of the earliest genes that was implicated in cellulose synthesis (8). KORRIGAN was shown to act as an endo-1,4-␤-glucanase (9), and it was also shown to have multiple effects on the composition of the cell wall (10,11). Although the function of KORRIGAN is known and an interaction with the cellulose synthase complex has been shown (12,13), the role of KORRIGAN in cellulose formation is still unclear.
Co-expression analysis helped to identify several other genes required for different aspects of cellulose formation (14). Like CSI1 and KORRIGAN, chitinase-like genes are co-expressed with the CESAs and were shown to be involved in cellulose formation (15). Chitinase-like proteins were suggested to play a key role in establishing interactions between cellulose microfibrils and hemicelluloses, which are thought to coat the cellulose microfibrils.
COBRA is one of the genes that is most highly co-expressed with the primary CESA family (14). The Arabidopsis COBRA gene was originally identified in a screen for root defects (16). In Arabidopsis, the COBRA family contains 11 members with different expression patterns. Of these genes, COBRA has the highest and most widely distributed expression (17).
Mutations in COBRA cause dramatic reduction in cellulose levels (18). Inspection of the orientation of cellulose fibrils in cobra mutants by electron microscopy suggested that the orientation of the cellulose fibrils was affected (19). COBRA was shown to be glycophosphatidylinositol (GPI) anchored, as well as N-glycosylated (19). GPI-anchored proteins are typically targeted to the outer leaflet of the plasma membrane (20). Based on phase separation, COBRA was detected in both endomembranes as well as in the plasma membrane fraction, whereas immunolabeling demonstrated localization at the Golgi and the cell wall, but not at the plasma membrane (19). However, immunolabeling of the rice ortholog showed that COBRA localized mainly at the plasma membrane in young, developing tissue, and primarily in the cell wall in mature tissue (21). Therefore, it was suggested that its localization is dependent on tissue type and developmental stage.
The COBRA sequence harbors a putative cellulose binding motif (CBM) (22), and the rice COBRA CBM was shown to bind cellulose (21). COBRA was tested against a variety of cell wall polymers, and was found to bind specifically to cellulose. However, the binding was significantly higher when tested against a crude cellulose fraction compared with Avicel, a purified cellulose powder (21).
In this work, we succeeded in using live imaging to visualize COBRA and establish its presence in distinct particles at the plasma membrane. By using isothermal titration calorimetry we showed that COBRA can bind individual ␤1-4-glucan chains with an dissociation constant (K D ) of 3.2 Ϯ 0.1 M. Furthermore, COBRA binding to individual ␤1-4-glucan chains competes with its binding to pure cellulose powder. These observations suggest that COBRA acts at the initial step of cellulose biosynthesis, when emerging individual chains align to form cellulose fibrils. Using solid-state nuclear magnetic resonance (ssNMR) we characterize the cell wall of the cobra mutant in detail, demonstrating the effect of disrupting cellulose crystallization.

Molecular Cloning
All plasmids are listed in Table 1. Oligonucleotides are listed in Table 2. Gateway cloning was performed according to the manufacturer's instructions (Invitrogen). COBRA was subcloned into pGEX4T-2 with oligonucleotide primers NS173 and NS174 to create GST-His 6 -COBRA (pNS87). The COBRA CDS (lacking the signal peptide and GPI motif sequence) with the sequence CATCATCCAGGCCATCAT (encodes the FlAsH tag, -Cys-Cys-Pro-Gly-Cys-Cys) inserted after tyrosine 37 (FlAsH-COBRA) was synthesized by DNA2.0. FlAsH-COBRA was amplified using oligonucleotides NS29 and NS30 and cloned into pDONOR207 to create plasmid NS38. 2.5 kb of the COBRA promoter was amplified using oligonucleotides NS31 and NS33 and cloned into pDONOR P4-P1R to create plasmid NS29. Plasmids NS29, NS38, and NS60 were used to create the expression vector pCOBRA::FlAsH-COBRA:NOS in pH7 mu34GW (plasmid NS79) by multigateway reaction.

Plant Growth Conditions
Arabidopsis ecotype Columbia seeds and various mutant lines (Table 3) were sterilized and germinated on Murashige and Skoog (MS) plates (one-half-strength MS salts, 0.8% agar, and 0.05% MES, pH 5.7). Seedlings were grown vertically on the agar at 22°C under continuous light for 5 days before being transferred to pots in a growth chamber at 22°C under a 16-h light/8-h dark cycle.

Confocal Microscopy
YFP-CESA6 imaging-For analyses of YFP-CESA6 proteins, seeds were germinated on MS agar plates and grown vertically in darkness for 5 days at 22°C. Seedlings were mounted between two coverslips in water. Imaging was performed on a Yokogawa CSUX1 spinning-disk system featuring the DMI6000 Leica motorized microscope and a Leica ϫ100/ 1.4 numerical aperture oil objective. YFP was excited at 488 nm, and a band-pass filter (510/20 nm) was used for emission filter-   ing. Image analysis was performed using Metamorph (Molecular Devices) and Imaris (Bitplane) software. Movies were collected at 5 days from randomly selected seedlings of the various genotypes. Movies were taken at ambient temperatures. An approximately equal number of acquisitions were attempted for each, with poor focal quality movies discarded during post-processing (data not shown).
Image analysis was performed with ImageJ and Imaris software. Movies were first contrast enhanced in ImageJ, and a walking average of four frames was taken using the kymograph plug-in for ImageJ. These steps were performed to improve the accuracy of automated particle recognition performed in subsequent steps. These images were opened in Imaris 6.2.1 and switched from Z-series to time series. The pixel size was set to 135 nm/pixel based on measurements from the microscope, and the time interval was set to 5 s. The particle-recognition algorithm in Imaris was performed with a spot size of 250 nm. High-intensity signal was filtered to eliminate Golgi signal. Following this, the connected components program was run, which determines particle identity over several frames and converts the movement of the particle into tracks. All tracks present for less than 60 s (12 frames) were discarded. The displacement and duration of the remaining tracks were exported to a spreadsheet, and their average velocity was calculated.
FlAsH-COBRA Imaging-FlAsH labeling was done as reported previously (23). Seeds of FlAsH-COBRA were surface sterilized in 2.4% sodium hypochlorite with 0.1% SDS, washed four times in water, and suspended in 0.15% agar at 4°C in the dark for at least 2 days before planting. Seeds were germinated on vertical 0.8% agar plates containing 2.2 g/liter of Murashige and Skoog salts (Caisson Labs) and 0.6 g/liters of MES (0.5ϫ Murashige and Skoog medium), pH 5.6, in 24 h with 120 mol m Ϫ2 s Ϫ1 light at 21°C for 5 days. Seedlings were removed from the plate and placed in 1.5-ml microcentrifuge tubes containing 2 M Lumio green (Invitrogen, number 12589-057) in liquid 0.5ϫ Murashige and Skoog medium lacking agar for 60 min in the dark with gentle shaking. The seedlings were washed twice with Disperse Blue 3 (Invitrogen), and mounted on slides for observation as described in the YFP-CESA6 imaging section.
Isoxaben Treatments-Isoxaben treatments were carried out as previously reported (24). GFP-CESA3 and FlAsH-COBRA seeds were surface sterilized and plated as described in the GFP-CESA3 imaging section. Seedlings were treated for 30 min with 100 nM Isoxaben.
Plasmolysis-FlAsH-COBRA seedlings were plasmolyzed by incubation in 0.8 M NaCl for 10 min. Seedlings were mounted on microscope slides in the plasmolysis solution.

Protein Expression and Purification
Escherichia coli DH5␣ was used for DNA propagation. For protein expression, pNS87 was transformed into E. coli BL21 CodonPlus DE3 RIL (Stratagene) cells. Cells were grown to A 600 of 0.6. Protein expression was induced by adding 2 mM isopropyl ␤-D-thiogalactopyranoside. Protein extraction and purification was carried out with AKTA-prime (GE Healthcare, UK) using HisTRAP FF and GSTrap HP columns according to the manufacturer's instructions (GE Healthcare, UK).

Isothermal Titration Calorimetry
Isothermal titration calorimetry experiments were performed in NanoITC 2G (TA Instruments) (25). Measurements were made at 21°C. The proteins were dialyzed extensively against 50 mM Tris-Cl buffer, pH 7.5, and the cellohexaose was dissolved in the same buffer to minimize heats of dilution. Titrations were carried out in triplicate, and the errors are the mean Ϯ S.D. of these replicates. During a titration experiment the 50 M protein sample stirred at 250 rpm in a 1.43-ml reaction cell was injected with 15 successive 15-l aliquots of 750 M cellohexaose at 300-s intervals. Integrated heat effects, after correction for heats of dilution, were analyzed by Sigmaploid dose-response ϩ Hill slope (SigmaPlot). The fitted data yielded the dissociation constant (K d ), number of binding sites on the protein (n), and the enthalpy of binding (⌬H).

COBRA, Avicel, and Cellohexaose Competition Assay
Experiments were done based on Ref. 26. Proteins were incubated with Avicel PH-101 (Sigma, 11365) (1 mg/ml) in 20 mM Tris-Cl buffer, pH 8, and 150 mM NaCl for 120 min. Avicel recovered by centrifugation was washed 3 times with 20 mM Tris-Cl buffer, pH 8, and 150 mM NaCl. Avicel-bound proteins were released by heating in SDS sample buffer for 5 min at 100°C and subjected to SDS-PAGE.

Cell Wall Analysis
Seedlings grown vertically on plates were placed in the dark for 24 h to reduce starch levels before harvesting. The alcoholinsoluble residues fraction was prepared by grinding 150 mg of seedling tissue in liquid nitrogen, followed by one extraction with 1 ml of 70% ethanol and three extractions with 1 ml of chloroform:methanol (1:1, v/v). The pellet was washed in 1 ml of acetone and dried overnight under vacuum. Cellulose analysis was based on a modified version of Saeman hydrolysis according to Bauer and Ibáñez (27). In brief, cellulose content was determined by subtracting hemicellulosic glucose released by hydrolysis with 4% sulfuric acid from total glucose released by a two-stage acid hydrolysis. From each sample, two 1-mg aliquots of alcohol-insoluble residue were prepared. The first was treated with 50 l of 72% sulfuric acid (room temperature, 1 h) to disrupt the cellulose crystalline structure; sulfuric acid was then diluted to 4% by adding 1.4 ml of H 2 O and the mixture was heated (121°C, 60 min). The second aliquot was treated with 1.45 ml of 4% sulfuric acid shortly before heating (121°C, 60 min). After cooling at room temperature, the reaction mixture was vortexed, and a 1:50 dilution was used for HPLC analysis.
Released monosaccharides were analyzed using an ICS-3000 HPLC system (Thermo Fisher Scientific) equipped with a pulse-amperometric detector. Samples were injected onto a 150 ϫ 3-mm CarboPac PA20 column (Thermo Fisher Scientific) with a 50 ϫ 3-mm guard column of the same material and eluted at 30°C with 2 mM potassium hydroxide at a flow rate of 0.4 ml/min. Cellulose content (mg of cellulose/mg of dry tissue) was calculated by subtracting the amount of glucose released by 4% sulfuric acid hydrolysis from the amount of glucose released by 72% sulfuric acid treatment. For neutral sugar and uronic acid analysis, a 1-mg aliquot of alcohol-insoluble residue was hydrolyzed with 1.45 ml of 4% sulfuric acid and the mixture was heated (121°C, 60 min). Samples were cooled and vortexed, and a 1:50 dilution was used for HPLC analysis as above. For uronic acids, samples were injected onto a 250 ϫ 3-mm CarboPac PA200 column with a 30 ϫ 3-mm guard column and eluted at 30°C using a 50 to 200 mM sodium acetate gradient (in 0.1 M NaOH) at 0.4 ml/min.

Cell Wall Structural Analysis by 13 C-MAS ssNMR
All 13 C-ssNMR spectra were acquired on a 17.6 T (700 MHz 1 H frequency) Bruker Avance spectrometer equipped with a 3.2-mm E free triple resonance magic angle spinning (MAS) probe from Bruker. The spectra were recorded at a 14 kHz magic angle spinning (MAS) rate and a temperature of 280 K. 13 C-Labeled chemical shifts were referenced externally to adamantane signal at 38.48 ppm (28). 13 C direct polarization experiments with a long recycle delay, 20 s, were used for quantitative measurements. Spectra were acquired with a 48-ms acquisition time, 83 kHz 1 H SPINAL-64 decoupling (29), and a 5-s pulse width on 13 C. 13 C-Labeled chemical shifts were assigned based on reported literature assignments (30).
Two-dimensional 13 C-13 C chemical shift homonuclear correlation spectra were acquired with a 50-ms dipolar-assisted rotational resonance mixing time (31). The data were acquired at 273 K, 14 kHZ spinning frequency for ϳ60 h. Other acquisition parameters were 0.5 ms 1 H-13 C CP contact time, 83 kHz 1 H SPINAL-64 decoupling, 10-ms t 1 acquisition time (1024 Rows X 9.5 us) with States-TPPI detection, 48-ms t 2 acquisition time (5120 ϫ 9.4 s), and a 2-s pulse delay. The data were processed with 0.28 ppm (50 Hz) net line broadening (Lorentzian-to-Gaussian apodization) and zero filled to 8192 (F2) ϫ 2048 (F1) complex points before Fourier transformation. These experiments were performed to obtain through-space correlations and to observe changes in the cell wall structure.

RESULTS
To visualize COBRA in live tissue, we created two transgenic lines expressing YFP-COBRA under the native promoter. In the first line, YFP was inserted after Pro 42 , five amino acids after the predicted secretion peptide cleavage site. In the second line, the YFP tag was inserted before the GPI site, at Pro 427 . In both lines we used 2.5 kb upstream of the ATG as a promoter sequence. However, these constructs did not complement the phenotype of the mutant in the 10 independent lines examined, implying the fusion protein was non-functional.
Thus, to visualize COBRA in live tissue we used FlAsH labeling, which involves the modification of the COBRA protein by the addition of a small peptide motif that binds derivatives of fluorescein (23). The tetracysteine motif (Cys-Cys-Pro-Gly-Cys-Cys) was placed immediately after the predicted secretion peptide cleavage motif at Tyr 37 . The FlAsH-COBRA construct expressed under the control of the native COBRA promoter complemented the cob-6 phenotype, implying functionality of FlAsH-COBRA. Due to low penetration of the FlAsH dyes into hypocotyl cells, FlAsH-COBRA imaging was done in elongating root cells in 5-day-old light grown seedlings (Fig. 1). FlAsH-COBRA was found to localize mainly at the plasma membrane in a punctate pattern (Fig. 1A). Plasma membrane localization of FlAsH-COBRA was further confirmed by plasmolysis (Fig.  1B). Unfortunately, the FlAsH tag bleached rapidly during laser illumination so we were not able to obtain useful measurements of FlAsH-COBRA movement. However, we observed movement of the particles, although they appeared to move more slowly than the GFP-CESA particles and were significantly less abundant (Fig. 1A).
To further investigate the association between COBRA and the cellulose synthase complex, seedlings expressing FlAsH-COBRA were treated with the cellulose synthase complex inhibitor, isoxaben (32). Isoxaben is thought to specifically act on the cellulose synthase complex because point mutations in CESA3 confer resistance to the drug (33). Treatment of seedlings with 100 nM isoxaben for 45 min caused almost complete depletion of the cellulose synthase complex from the plasma membrane (5) (Fig. 1C). To detect the different focal planes, we used the outer epidermal cell wall autofluorescence as a reference point, the focal plane below (200 nm) was referred to as plasma membrane, and 3 m below this was referred to as the focal plane of the cytoplasm (Fig. 1C, lower scheme). Isoxaben treatment of seedlings expressing FlAsH-COBRA showed a similar phenomenon, as the signal from the plasma membrane depleted, and FlAsH-COBRA could be detected only within the cells (Fig. 1C). The similar effect of isoxaben on both GFP-CESA and FlAsH-COBRA puncta suggests a relationship between the two proteins. One possibility is that the presence of COBRA on the plasma membrane is dependent upon the presence of the CESA complex.
The effect of the cob-6 mutation on cellulose synthesis was visualized by live-cell imaging of cellulose synthase complexes labeled with YFP (Fig. 1D). The average velocity of the cellulose synthase complex in 5-day-old seedlings was 225 nm/min in WT and 289 nm/min in cob-6 (Fig. 1E). The average apparent lifetime of the CESA6-YFP-labeled complex in the cob-6 background, measured as "tracks" in time-averaged images, was shorter, but the directionality of the tracks was unaffected and linear movement was maintained (Fig. 1D). Thus, the basis for the cellulose deficiency was not apparent from direct observation of cellulose synthesis. These sets of data suggest that COBRA acts at the plasma membrane, adjacent to the cellulose synthase complex. Therefore, we tested whether COBRA could interact with the product of the cellulose synthase complex, ␤1-4-linked glucan (Fig. 2).
COBRA contains a putative CBM (19), and it was recently shown that the purified CBM region from Brittle Culm1, a COBRA-like protein from rice, can bind cellulose (21). The fact that COBRA contains a GPI anchor (19) implies that this association occurs on the plasma membrane outer leaflet. Therefore, we tested if COBRA could interact with individual ␤1-4glucan chains. We expressed COBRA from Arabidopsis in E. coli without the secretion peptide and the GPI motifs that are cleaved during protein maturation. Binding, affinity, and stoichi- ometry of COBRA for cellohexaose was determined by isothermal titration calorimetry (Fig. 2). From these results, the number of ligand molecules bound to COBRA (n), the dissociation constant (K D ), and the binding enthalpy (⌬H) were evaluated. Monomeric carbohydrate-binding proteins generally have a 1:1 ratio with their oligosaccharide ligands (34). Similarly, COBRA was also found to bind one molecule of cellohexaose per protein molecule (n ϭ 0.94 Ϯ 0.02). The enthalpy of binding, ⌬H, is negative with a value of Ϫ8.2 Ϯ 0.4 kcal mol Ϫ1 (Fig. 2) implying that the reaction is exothermic. The binding association constant of COBRA to cello-hexaose was found to be 3.2 Ϯ 0.1 M, which is similar to other proteins containing CBMs (35).
Because COBRA was shown to bind cellulose (21), we tested the affinity to crystalline cellulose versus the affinity to individual glucan chains (Fig. 3). Avicel is a 95% pure crystalline cellulose with an average particle size of 50 m. Because Avicel is not soluble, competition assays with cellohexaose were carried out using pulldown assays (26). COBRA was incubated for 2 h at 21°C with Avicel in the presence of increasing concentrations of cellohexaose. The samples were analyzed using an anti-His 6 FIGURE 1. COBRA is localized at the plasma membrane, is sensitive to isoxaben, and when mutated disrupts cellulose synthesis. A, FlAsH-COBRA and GFP-CESA3 were visualized in the hypocotyl elongation zone of 5-day-old dark grown Arabidopsis seedlings. FlAsH-COBRA showed a punctate pattern of labeling in the plasma membrane focal plane. B, plasma membrane localization was verified by plasmolysis with 0.8 M NaCl, which causes shrinkage and condensation of the cytoplasm and vacuole, and detachment of the plasma membrane from the cell wall (arrowhead). C, treatment of seedlings with 100 nM isoxaben caused clearance of GFP-CESA3 particles from the plasma membrane resulting in intracellular accumulation. The same observation was made for FlAsH-COBRA treated with isoxaben. D, time-lapse confocal images of YFP-CESA6 in hypocotyl cells of 5-day-old etiolated Col-0 (left) and cob-6 (right) plants were compared. Each image is an average of 36 frames taken at 5-s intervals (n ϭ 9). CESA complex velocities were measured using Imaris. Data passed normality and equal distribution tests after natural log transformation. The complex velocity in cob-6 (289 Ϯ 162 nm/min) was found to be significantly different from that of wild type (225 Ϯ 147 nm/min) (t test, p Ͻ 0.01). Aside from the differences in the velocity rate, in cob-6, cellulose synthesis occurs in a relatively linear direction. Scale bars are 5 m in A, B, and D and 10 M in C.
antibody, the washed pellet representing COBRA was bound to Avicel, and the soluble fraction COBRA was not bound to Avicel (Fig. 3). It is important to bear in mind that in this type of experiment, the individual glucan chains are soluble, whereas in planta they would be constrained by attachment to the cellulose synthase complex and/or hydrogen bonding to cellulose microfibrils. In the absence of cellohexaose, there is no detectable COBRA in the unbound portion (Fig. 3, 0 M C6). As mentioned above, the COBRA K D for cellohexaose was found to be 3.2 M (Fig. 2), so we tested a range of cellohexaose concentrations below and above the K D . When 1 M cellohexaose was added to the reaction, a considerable fraction of COBRA remained in the soluble fraction, and this fraction increased when 2 M cellohexaose was added to the reaction. With addition of 3 M cellohexaose, more protein was detected in the soluble fraction compared with the Avicel-bound fraction, and with 30 M cellohexaose, 10-fold above the K D , COBRA was detected only in the soluble fraction and undetectable in the Avicel-bound fraction (Fig. 3). Along with the localization results, this finding raises the possibility that COBRA may act at the stage where individual ␤1-4-glucan chains have emerged from the complex and are being assembled into cellulose fibrils.
We characterized the effect of mutation in COBRA on the cell wall composition (Fig. 4) and structure (Figs. 5 and 6). cob-6 has a reduction of 50% in the amount of cellulose (Fig. 4A), and   fewer overall glucan chains in the cell wall fraction (Fig. 5). It has been shown that the reduction of cellulose in the cobra mutant is accompanied by overproduction of starch, preventing osmotic stress by the excess of glucose that cannot be incorporated into the cell wall (36). Analysis of galacturonic acid (GalA) levels revealed a significant increase in the cobra mutant (Fig. 4B), and Arabinose (Ara) and galactose (Gal) levels also were increased (Fig. 4C). Galacturonic acid forms the backbone of the pectic polymers, and arabinose and galactose are abundant in the pectin side chains, implying a significant increase of the pectic fraction in the cobra mutant. To further characterize the cellulose fibrils that form in the cobra mutant, we used MAS ssNMR spectroscopy (Figs. 5 and 6).
Magic angle spinning solid-state NMR spectroscopy has already been proven to be a feasible technique to investigate the structure and intermolecular interactions of polysaccharides in intact plant cell walls (30,37,38). Previous ssNMR studies have yielded important structural information on primary Arabidopsis cell walls (39 -41).
We used ssNMR to further analyze the molecular structure of the polysaccharides in WT and cob-6 dark grown seedlings grown in liquid 1 ⁄ 2MS media with 13 C-labeled glucose (30) to obtain detailed information about crystalline cellulose struc-ture. Spectra were measured quantitatively by 13 C direct polarization ssNMR using a long recycle delay (Fig. 5A). The onedimensional 13 C-spectra show that the amount of cellulose in cob-6 is considerably lower than in WT. Furthermore, the ratio between the intensity of the interior C4 peak (iC4) to surface C4 peak (sC4) in cellulose (40) was measured and indicates that cob-6 has 16% lower crystallinity relative to WT (Table 4). In addition, we observed the peak at 88 ppm, which is indicative of a degree of crystallinity that is intermediate between the surface and interior cellulose (40).
When a short relaxation delay was utilized to increase the intensities of dynamic polysaccharides, the 88 ppm peak intensity was preferentially enhanced in cob-6 relative to WT. We measured two-dimensional 13 C-13 C CP-dipolar-assisted rotational resonance spectra to detect through space contacts between carbons (Fig. 5, B and C). Qualitative inspection of the two-dimensional spectra shows that the relative intensities of hemicellulose and pectin cross-signals is significantly higher in cob-6 compared with WT. We also observed that cob-6 is not only rich in pectin, but the amount of non-esterified galacturonic acid is significantly higher than in WT. A comparison of the two-dimensional spectra revealed additional cross-peaks in the cob-6 spectrum, which are absent in the WT spectrum even FIGURE 5. One-and two-dimensional ssNMR of WT and cob-6 cell wall. A, quantitative 13 C direct polarization-MAS ssNMR spectra of WT and cob-6 cell walls. B and C, 13 C-13 C chemical shift correlation spectra with 50-ms dipolar-assisted rotational resonance mixing of WT (B) and cob-6 (C). Assignments are indicated by color for cellulose (red), hemicelluloses (blue), and pectin (green). Yellow rectangles indicate polysaccharides whose amounts changed in cob-6 compared with the WT and relative to the amount of cellulose.
at lower counter levels (Fig. 6). Those cross-peaks align at 95.7 ppm, which is characteristic of the anomeric carbon chemical shift of a glucose monomer or cellulose reducing end. The specific chemical shifts we observed suggest that these peaks arise from a ␤-glucose reducing end. This indicates that there is a higher concentration of short glucan chains in the cob-6 mutant than in WT, high enough to enable detection.

DISCUSSION
COBRA was one of the first proteins suggested to be involved in cellulose formation (18) but its role has remained obscure. Previous immunoelectron microscopy studies suggested that COBRA was primarily located in the cell wall (19,21), although the protein carries a motif that would be expected to lead to attachment of a GPI anchor. We found, by live-cell imaging, that COBRA was located in particles in the plasma membrane but did not observe any labeling in the cell wall. These particles are potentially the same as those containing CESAs, although because there are a fewer number of COBRA particles than CESA containing particles, we infer that cellulose synthase complexes would not always be associated with COBRA. Also, we observed a difference between the average velocity of movement of the COBRA-containing particles and the CESA-containing particles, suggesting that, if COBRA associates with cellulose synthase, it would be a slow-moving subset of CESA particles or a transient interaction such as initiation of micro-fibril synthesis (see speculative model as Fig. 7). The apparent discrepancy in localization may be due to an effect of the developmental stage of the tissue, as observed by Liu et al. (21). They suggested that COBRA may be released from the membrane in older tissues by developmentally regulated hydrolysis of the GPI anchor.
Live-cell imaging of the cellulose synthase complex in the cob-6 mutant did not reveal any obvious defect in the rate of cellulose polymerization in the mutant once polymerization was initiated, nor any apparent difference from the wild type in the orientation of cellulose deposition. The disorganization of cellulose observed by electron microscopy (19) may not be due to cobra per se, but rather reflect an altered manner by which the defective microfibrils shift during expansion. There is also no apparent difference in the abundance of cellulose synthase complexes. Thus, we infer that the reduced amount of cellulose measured in the cobra mutants, as well as the increased number of cellulose-associated reducing ends evident by NMR, reflect an alteration in a process downstream of glucan polymerization.
The evidence that COBRA is in the plasma membrane ( Fig.  1) and can bind cellulose and individual glucan chains (Figs. 2 and 3) leads us to hypothesize that COBRA functions after glucan polymerization at the step of cellulose formation in which individual chains emerge from the complex and are aligned to form cellulose fibrils (Fig. 7). Thus, we propose that COBRA is a kind of "polysaccharide chaperone" that facilitates cellulose crystallization. We hypothesize that, in the absence of COBRA, crystallization takes place with lower fidelity leading to the formation of partially "amorphous cellulose," which leads to the swollen cells that are characteristic of cobra mutants. The reduction in the amount of cellulose may be due to the defect in the ability to control turgor-induced cell expansion, or to deg-FIGURE 6. 13 C-13 C chemical shift correlation spectra of WT and cob-6 acquired with 50-ms dipolar-assisted rotational resonance mixing time. The spectra from Fig. 4 was plotted at lower contour levels, emphasizing the cross-section at ϳ96 ppm in the cob-6 spectrum. The cross-section signals have been assigned to the ␤-glucose reducing end units, which can be detected when the degree of polymerization of cellulose is sufficiently low. radation of amorphous cellulose by cellulases, perhaps by KORRIGAN. The apparent reduction in dimensions of cellulose microfibrils inferred from NMR of cellulose from the cob-6 mutant is consistent with this hypothesis. The particulate organization of COBRA, evident from the live cell imaging, presumably reflects accumulation of many COBRA proteins in close proximity. This organization is compatible with the idea that many COBRA molecules would be required in close proximity during cellulose formation because many glucan strands must be co-crystallized. In view of our observation that COBRA does not appear to be associated with the CESA complexes, we propose that COBRA acts transiently at the initiation of cellulose microfibrils to align the glucans as they emerge from the cellulose synthase complex, after which the process does not require COBRA participation. Because it is GPI-anchored, COBRA may help keep the growing cellulose microfibril at the plasma membrane during the initial stages of microfibril formation. If COBRA does not dissociate from the glucan chains, it could eventually be displaced into the cell wall by newly synthesized cellulose fibrils. Cellulose-synthesizing bacteria do not have detectable homologs of COBRA. However, these bacteria do not incorporate cellulose into cell wall structures and the fibrils are smaller than those from higher plants and not coated with hemicellulose. Most importantly perhaps, plant cellulose microfibrils are tightly appressed between the site of synthesis in the plasma membrane and the adjacent cell wall, whereas bacterial microfibrils are not. This may create a tight bend in the glucan chains as they emerge from the plant cellulose synthase, which may impose a need for the proposed COBRA activity that the bacterial process lacks. Alternatively, the presence in plant cell walls of hemicellulose polysaccharides, which hydrogen-bond to cellulose, may impose a need for shielding of the newly synthesized glucans during the process of crystallization to form cellulose microfibrils.