Subfunctionalization of cellulose synthases in seed coat epidermal cells mediates secondary radial wall synthesis and mucilage attachment.

Arabidopsis (Arabidopsis thaliana) epidermal seed coat cells follow a complex developmental program where, following fertilization, cells of the ovule outer integument differentiate into a unique cell type. Two hallmarks of these cells are the production of a doughnut-shaped apoplastic pocket filled with pectinaceous mucilage and the columella, a thick secondary cell wall. Cellulose is thought to be a key component of both these secondary cell wall processes. Here, we investigated the role of cellulose synthase (CESA) subunits CESA2, CESA5, and CESA9 in the seed coat epidermis. We characterized the roles of these CESA proteins in the seed coat by analyzing cell wall composition and morphology in cesa mutant lines. Mutations in any one of these three genes resulted in lower cellulose content, a loss of cell shape uniformity, and reduced radial wall integrity. In addition, we found that attachment of the mucilage halo to the parent seed following extrusion is maintained by cellulose-based connections requiring CESA5. Hence, we show that cellulose fulfills an adhesion role between the extracellular mucilage matrix and the parent cell in seed coat epidermal cells. We propose that mucilage remains attached to the seed coat through interactions between components in the seed mucilage and cellulose. Our data suggest that CESA2 and CESA9 serve in radial wall reinforcement, as does CESA5, but CESA5 also functions in mucilage biosynthesis. These data suggest unique roles for different CESA subunits in one cell type and illustrate a complex role for cellulose biosynthesis in plant developmental biology.

Arabidopsis epidermal seed coat cells follow a complex developmental program where, following fertilization, cells of the ovule outer integument differentiate into a unique cell type.
Two hallmarks of these cells are the production of a donut-shaped apoplastic pocket filled with pectinaceous mucilage and the columella, a thick secondary cell wall. Cellulose is thought to be a key component of both these secondary cell wall processes. Here we investigated the role of cellulose synthase (CESA) subunits CESA2, CESA5, and CESA9 in the seed coat epidermis.
We characterized the roles of these CESA proteins in the seed coat by analyzing cell wall composition and morphology in cesa mutant lines. Mutations in any one of these three genes resulted in lower cellulose content, a loss of cell shape uniformity and reduced radial wall integrity. In addition, we found that attachment of the mucilage halo to the parent seed following extrusion is maintained by cellulose-based connections requiring CESA5. Hence, we show that cellulose fulfills an adhesion role between the extracellular mucilage matrix and the parent cell in seed coat epidermal cells. We propose that mucilage remains attached to the seed coat through interactions between components in the seed mucilage and cellulose. Our data suggest that CESA2 and CESA9 serve in radial wall reinforcement, as does CESA5, but CESA5 also functions in mucilage biosynthesis. These data suggest unique roles for different CESA subunits in one cell type, and illustrate a complex role for cellulose biosynthesis in plant developmental biology.

Introduction
In Arabidopsis, the specialized epidermal cells of the seed coat are hexagonally shaped when viewed from the surface, and form an interconnected cell layer that protects the embryo from adverse environmental effects such as UV damage, mechanical stress, desiccation and biotic factors (Windsor et al., 2000;Haughn and Chaudhury, 2005;North et al., 2010;Nowack et al., 2010). Developmentally, epidermal seed coat cells display several unique morphogenic and cell wall characteristics. Between approximately 5-9 days after fertilization, seed coat epidermal cells deposit large amounts of pectinaceous mucilage which accumulates in a mucilage pocket in the apoplast (Haughn and Chaudhury, 2005). A cytoplasmic column is clearly visible in the center of the cell, surrounded by a donut-shaped mucilage pocket (Western et al., 2000;Western et al., 2004;Haughn and Chaudhury, 2005). As maturation proceeds, 9-13 days after fertilization, secondary cell wall material is deposited along the outer-tangential and radial sides of the plasma membrane, eventually displacing the cytoplasm with a volcano-shaped secondary cell wall called the columella (Western et al., 2000). During the latter stages of seed development, all cells of the seed coat die and desiccate (Haughn and Chaudhury, 2005). Upon imbibition, the pectinaceous mucilage swells and erupts from the pocket by breaking the radial walls above the secondary cell wall thickening. The extruded mucilage forms a halo that surrounds the seed (Western et al., 2000;Arsovski et al., 2009).
Many transcription factors involved in regulating mucilage production and seed coat cell development are known; several genes encoding pectin modifying and biosynthesis enzymes have also been identified through mucilage hydration screens (reviewed in Arsovski et al., 2010). Transcription factors regulate mucilage production directly by regulating mucilage biosynthetic genes, or indirectly by effecting mucilage secretory cell (MSCs) development (Arsovski et al., 2010). The transcription factor APETALA2 (AP2) is required for the differentiation of the two outer layers of the seed coat (Jofuku et al., 1994;Western et al., 2001): seeds homozygous for mutations in AP2 show altered cell shape, and lack of columellae and mucilage in the seed epidermal layer (Western et al., 2001). Other transcription factors such as (MUM1) required for the proper extrusion of mucilage has been identified, which positively regulates genes (MUM2, SUBSILIN PROTEASE1.7, and β-XYLOSIDASE1) involved in mucilage extrusion (Huang et al., 2011). Apart from these transcription factors, several pectin biosynthetic/modifying genes involved in mucilage production/extrusion have been identified and characterized (Western et al., 2001;Arsovski et al., 2010). However, the role of cellulose in seed coat development and mucilage biosynthesis remains largely unknown.
Cellulose biosynthesis is an essential process that mediates cell anisotropic growth and morphogenesis in higher plants. Cellulose microfibrils are paracrystalline chains of β-1,4 linked glucose molecules that support the structural framework of the plant cell wall (Brown, 1985).
Several lines of evidence suggest that cellulose synthesizing organisms produce microfibrils from plasma membrane localized cellulose synthesizing complexes. In Arabidopsis there are 10 CELLULOSE SYNTHASE (CESA) genes identified based on sequence homology to bacterial cellulose synthases (Richmond and Somerville, 2000). A combination of biochemical and genetic analyses have shown that at least three CESA subunits directly interact to form a CSC in primary (Desprez et al., 2007;Persson et al., 2007) and secondary cell wall formation (Taylor et al., 2003). It is unclear exactly how all the CESA subunits interact due to redundancy in the CESA gene family; separate combinations of CESAs may therefore be active in different cell types at different times. CESA1 appears to be required for primary wall synthesis because a temperature sensitive allele, rsw1-1 (Arioli et al., 1998), results in temperature dependent disappearance of rosettes from the plasma membrane, and strong alleles are either embryo or male gametophytic lethal (Beeckman et al., 2002;Gillmor et al., 2002). CESA3 is co-expressed with CESA1, and homozygous cesa3 alleles are also male gameteophytic lethal (Persson et al., 2007). CESA6 appears to be required for elongation of hypocotyl cells in etiolated seedlings (Fagard et al., 2000;Desprez et al., 2002), but can be partially complemented by other CESA6related CESAs; CESA2, CESA5, and CESA9 (Desprez et al., 2007;Persson et al., 2007).
These genes are thought to encode components of the primary cell wall cellulose biosynthetic machinery, but when mutated individually do not result in lethality (Desprez et al., 2007). Mutant cesa5 plants have no reported phenotype, but cesa5cesa6 double mutants are seedling lethal (Desprez et al., 2007). CESA5 is therefore redundant to CESA6, with at least one of these subunits being required for seedling establishment. Triple cesa2cesa6cesa9 plants are pollen lethal, indicating they function redundantly in pollen development (Persson et al., 2007). In addition, CESA9 is required for normal secondary wall synthesis in epidermal seed coat cells in Arabidopsis (Stork et al., 2010) and functions redundantly with CESA6 during pollen development (Persson et al., 2007).
Functional specialization of CESAs is thought to be due to both differences in expression and differences in amino acid sequences that could result in chemically distinct microfibrils.
Expression of CESA genes in different tissues has been well documented (Beeckman et al., 2002) and transcriptional regulation may be a primary means for defining the CSC stoichiometry (Persson et al., 2005). For example, CESA5 undergoes cell type specific transcriptional repression by GLABRA2 in Arabidopsis root cells (Tominaga-Wada et al., 2009). CESA5 and CESA6 are co-expressed during seedling establishment, while CESA2, CESA6 and CESA9 are co-expressed during pollen development. During xylem development, all three secondary wall CESAs (CESA4, CESA7, CESA8) are expressed in the same cells at the same time (Taylor et al., 2003). While the coding sequence of CESA proteins is highly conserved (Richmond and Somerville, 2000), a hyper variable region and numerous subtle amino acid level differences exist between subunits that could influence CESA function (Fagard et al., 2000). However, no study has shown functional specialization of individual CESA subunits. Nevertheless, immunoprecipitation analyses of differentiating xylem (secondary cell wall biosynthesis) cells of Populus have shown that two different types of CSC exist, which indicates that different CSCs could produce structurally discrete microfibrils that partition during cell wall synthesis (Song et al., 2010). These data suggest that in differentiating cell types, cellulose biosynthesis may be more complicated than a single stoichiometric model for a hexameric CSC, as is proposed in many plant models (Scheible et al., 2001), or that it involves ancillary proteins that function to specialize the CSC. Functional specificity of multiple CESA proteins in a single cell type, particularly one with a well-understood differentiation program, such as that seen in epidermal seed coat cells, has not yet been explored.
During seed development and desiccation, the secondary cell walls in the seed coat protect the embryo. Cellulose has been proposed to have a major role in reinforcing the secondary cell wall in the seed coat epidermal cells (Stork et al., 2010). By way of histochemistry, cellulose is also proposed to be a component of seed mucilage (Willats et al., 2001;Blake et al., 2006;Young et al., 2008). Fluorescently tagged cellulose binding modules were used to identify cellulose in mucilage, which indicated distinct binding patterns; for crystalline and amorphous cellulose microfibrils (Blake et al., 2006;Young et al., 2008). Further studies showed that glycan strands are localized to the internal domain of the inner layer of mucilage (Macquet et al., 2007a).
Analysis of mucilage composition using the glycan binding dye Calcofluor White indicated the presence of glycans in addition to pectins, but no crystalline cellulose could be detected by analytical techniques in extracted mucilage (Windsor et al., 2000;Willats et al., 2001). By genetically dissecting cellulose biosynthesis in seed coat epidermal cell development we discovered that cellulose production involving CESA5 subunit is indispensible for mucilage attachment to the seed coat while CESA2, CESA5 and CESA9 subunits contribute towards secondary wall synthesis in epidermal seed coat cells.

Cell type specific expression of CESA2, CESA5 and CESA9
Current models suggest that primary cell wall cellulose biosynthesis requires CESA1, CESA3 and a CESA6 related subunit (Desprez et al., 2007;Persson et al., 2007). We examined the expression of the CESA6 related genes: CESA2, CESA5, and CESA9, which were previously shown to be expressed in the Arabidopsis seed (Beeckman et al., 2002) by cross referencing seed coat gene expression data from GENEINVESTIGATOR (Zimmermann et al., 2004). These data showed gene expression was highest for CESA2, followed by CESA5 and CESA9 in seed coat cells (general seed coat, testa and chalza)( Fig 1A). To further explore seed specific expression, transgenic Arabidopsis plants transformed with promoter CESA2 ::GFP, promoter CESA5 ::GFP or promoter CESA9 ::GFP fusions were generated and developing seeds were carefully examined by laser scanning confocal microscopy. Consistent with expression levels shown via GENEINVESTIGATOR, seed coat gene expression was observed for promoter CESA2 ::GFP, promoter CESA9 ::GFP and promoter CESA5 ::GFP ( Fig 1B). Expression patterns of promoter CESA2 ::GFP, promoter CESA9 ::GFP were localized through the seed coat, while promoter CESA5 ::GFP expression patterns were more localized to the outer layer of the seed coat.
Observations were made in seeds isolated from whole siliques 10 days post anthesis (DPA) from transgenic plants.
Based on this distinctive cell shape, we were able to employ a simple and reliable visual screen for altered morphogenesis using scanning electron microscopy (SEM) of plasma (goldpalladium) coated seed surfaces. Close examination of multiple generations of cesa2-1 and cesa5-1 seeds revealed distorted and irregular epidermal cell shape as compared to the more uniformly shaped hexagonal cells of wild type seeds (Fig 2, Fig S1), similar to a previous report for cesa9 seeds (Stork et al., 2010). The most obvious discriminating feature between wild type seed coat cells and those observed in the single mutants was a reduction in the radial wall ridge (Fig 2) suggesting a decrease in secondary cell wall deposition. As observed via SEM, cells could clearly be distinguished by a central columella, but the boundaries between neighboring cells were less visible due to a reduction in the radial wall ridges for both cesa2 and cesa5 (Fig 2), as was also reported for cesa9 (Stork et al., 2010). Radial wall ridge reduction was even more severe in the cesa2-1cesa9-1 double mutant compared to the single mutants ( Fig 2). Inspection of the cesa2-1cesa5-1cesa9-1 triple mutant revealed a dramatic loss of radial wall ridges between cells, leading to nearly indistinguishable cell boundaries (Fig 2). However a mutation in a related subunit, CESA6 (procuste) (Fig S2), did not alter the cell shape or cause radial wall ridge reduction.
In addition to reductions in the radial wall ridge, changes in the surface area of individual cells were also observed. To investigate this observation further, the surface area of individual cells was quantified using the SEM images. To avoid the interference of the three dimensional landscape and to promote planar uniformity, a region of cells from the micrograph that did not extend to the border regions of the image were selected for the area measurements. We quantitatively analyzed cell area using frequency distribution comparisons. Cell surface area measurements were made using a minimum population of 10 different seeds from each genetic background and 30 cells from each seed (n=300; Fig 2). All cesa single mutant lines showed a decrease in average cell surface area: wild type (842.8 µm 2 ; SE of 14.5 µm 2 ), cesa2-1 (782.7 µm 2 ; SE of 13.4 µm 2 ), cesa2-2 (765.4 µm 2 ; SE of 15.0 µm 2 ), cesa5-1 (780.2 µm 2 ; SE of 15.4 µm 2 ), cesa5-2 (788.8 µm 2 ; SE of 17.0 µm 2 ), cesa9-1 (743.6 µm 2 ; SE of 13.1 µm 2 ) (Fig 2 and as previously reported in Stork et al., 2010). The cesa2-1cesa9-1 double mutant showed a slightly stronger reduction compared to single mutants (736.7 µm 2 ; SE of 12.3 µm 2 ), while the cesa2-1cesa5-1cesa9-1 triple mutant showed the most severe reduction (706.0 µm 2 ; SE of 13.4 µm 2 ) compared to wild type. These data demonstrate that a reduction in cellulose synthesis in seed epidermal cells results in smaller overall cell size or smaller seeds.

Histochemical analysis of mutant versus wild type epidermal seed coat cells revealed changes in the thickness and height of radial wall secondary thickening.
Calcofluor White has previously been used to effectively explore spatial constraints of cellulose deposition in different plant tissues (Herth and Schnepf, 1980;Albani and Plancke, 1998).
Based on the reduced radial wall ridge phenotype, we hypothesized that a deficiency in cellulose should be apparent at the cell borders. To visualize this scenario we used laser scanning confocal microscopy to create three dimensional animations (z-plane projections used 500 nm per frame shift for resolution) of seed epidermal cells in mutant and wild type lines stained with Calcofluor White. Consistent with the surface view of epidermal cells observed by SEM we could observe aberrant cell shape and regions of reduced and patchy fluorescence across the radial walls separating neighboring cells and the columella in all mutant lines as compared to wild type (Fig 3A, 3B). To quantify how differences in cellulose biosynthesis correlated with Calcofluor White staining between wild type and mutant lines, we generated cross sectional fluorescence intensity plots. These analyses are relative quantifications, since there is no standard available for glycan staining, and the cell wall is a complex mixture of glycan containing biopolymers. Our results revealed a sharp maximum in fluorescence intensity at the radial walls and the central columella of wild type cells ( Fig 3B). Overall, the fluorescence intensity was quantitatively reduced in the single and double mutants and showed a severe loss of intensity in the triple mutant ( Fig 3B).
The most intriguing comparison was between wild type cells and the cesa2-1cesa5-1cesa9-1 triple mutant (Fig 3C; Movie S1 and Movie S2). In wild type, we found that the radial wall appeared elevated and displayed regions of partial fluorescence overlap with the neighboring cells. In cesa2-1cesa5-1cesa9-1 cells, we could not observe fluorescence overlap between neighboring cells. Rather, a thin band of fluorescence was visible surrounding most cells, and some cells showed regions of no fluorescence at the boundary (Fig 3). When the three www.plantphysiol.org on August 24, 2017 -Published by Downloaded from Copyright © 2011 American Society of Plant Biologists. All rights reserved. dimensional animations were oriented at a tangential perspective (Fig 3A), we observed that wall elevation in the triple mutant was reduced compared to wild type (also see Movie S1 and Movie S2). These data suggest that the secondary cell wall at the boundaries of seed coat cells is primarily cellulosic in nature and is reliant on at least CESA2, CESA5 and CESA9 for its biosynthesis. In wild type cells, the fluorescence derived from the radial wall appeared to be nearly equal in height as the central columella (Figs 3A and 3B). By contrast, cesa2-1cesa5-1cesa9-1 radial walls barely extended beyond the plane of the trough (Figs 3A and 3B). We interpret the thin band of fluorescence between columella in cesa2-1cesa5-1cesa9-1 as representing the primary radial cell wall, and that the loss of fluorescence at this position compared to wild type cells was due to a reduction in secondary wall thickening.
To measure the reduction in secondary wall thickening and to visualize the developmental program of epidermal seed coat cells in CESA mutants, we examined cryofixed sections of wild type and mutant seeds from 4 days post anthesis (DPA), 7DPA and 11 DPA using light microscopy (Fig 4). Cross-sectional analysis was performed on seeds from cesa2-1, cesa2-2, cesa5-1, cesa9-1, cesa9-2, cesa2-1cesa9-1, and cesa2-1cesa5-1cesa9-1. For the developmental analysis, seeds were excised from at least five siliques for 11 DPA, four siliques for 7 DPA and three siliques for 4 DPA, and sampled randomly from five to six plants for each genotype. Visual examination of embedded seeds revealed relatively minor differences between wild type and mutant genotypes at 4 DPA (data not shown) and 7 DPA (Fig 4, column 1). At 11 DPA (Fig 4, column 2), wild type Col-0 cells have completed mucilage production and have begun to deposit secondary wall material, filling most of the cytoplasmic column. By contrast, the cesa2-1cesa5-1cesa9-1 triple mutant appeared to be delayed in secondary wall deposition with cytoplasm still observable through the middle of the columella, extending to the top of the cell. In addition, secondary wall material accumulated in an irregular manner on the sides of the columella. In order to quantify secondary cell wall deposition, mature seeds were fixed with glutaraldehyde, allowing the mucilage pocket to burst, leaving the secondary wall easier to observe (Fig 4, column 3). Although differences in cell wall thickness between wild type and mutants were observed, the irregular shape of the secondary cell wall and variability in the plane of section makes quantification of secondary cell wall thickness difficult. To reduce this variability, we chose the height of the secondary cell wall thickening at the radial wall as a determinant for two reasons. First, because it is easy to identify the position of the radial wall in each cell, variability stemming from irregularities in thickness throughout a cell could be minimized. Second, because the secondary cell wall ends at the radial wall, relatively small changes in secondary cell wall thickness can have a pronounced effect on the height of the radial wall thickening. A total of 20 radial walls per resin block were measured for each genotype from a random set of seeds from three different resin blocks (n=60). Radial walls were measured from the base of the secondary wall depositions to the top of the radial wall thickening (Fig 5). Radial walls for Col-0 were the highest (8.14 ± 0.27 µm), while all cesa mutant lines were reduced to a similar height [cesa2-1 6.51 ± 0.17 µm; cesa2-2 5.71 ± 0.15 µm; cesa5-1 6.27 ± 0.18 µm; cesa9-1 6.64 ± 0.22 µm; cesa9-2 5.78 ± 0.16 µm). The double mutant radial walls were reduced more than any single mutant, and the triple mutant was severely reduced compared to all other mutant lines (cesa2-1cesa9-1 5.49 ± 0.15 µm; cesa2-1cesa5-1cesa9-1 4.87 ± 0.12 µm) (Fig 5). Radial wall thickening was absent in some cells of the triple mutant. Radial wall height measurements for cesa9-1 and cesa9-2 were similar to previous measurements of these genotypes, when differences in the way they were measured are accounted for (measured from the base of the radial wall in this study, and measured from the base of the mucilage pocket in the previous study) (Stork et al., 2010). Based on the wall height measurements all cesa mutants were significantly different from wild type (two tailed t-test, P<0.05) (Fig 5). In addition, cesa2-1cesa5-1cesa9-1 radial walls were significantly smaller than the radial walls of the cesa single mutant lines, and the cesa2-1cesa9-1 double mutant (two tailed t-test, P<0.05). Results of the developmental series indicate reduced radial wall height is consistent with the decreased radial wall ridges observed by SEM as well as the reduced and patchy regions of fluorescence at the radial wall observed when cell were stained with Calcofluor White (Figs 2 and 3).
Changes in seed cellulose content, without observable difference in stem cellulose content further suggests that cellulose biosynthesis is under tissue specific control with respect to the composition of active CESA subunits.

Cellulose biosynthesis via CESA5 is required for mucilage adherence to the parent seed during hydration
Connections between the mucilage halo and the parent seed during the process of imbibition are poorly understood, but could involve cellulose present in the mucilage (Willats et al., 2001;Blake et al., 2006;Macquet et al., 2007a;Young et al., 2008). We tested this model by visualizing seed mucilage extrusion and adherence in mutant cesa lines. Our initial experiment was to visually examine for mucilage defects in cesa mutants using Ruthenium Red stain.
Previous studies have shown co-localization of methylesterified pectins through JIM5 antibody labeling, with structural glycans through Calcofluor White staining at the linear rays that extend from the top of the columella (Macquet et al., 2007a). To examine whether cesa5 mucilage had changes in cellulose staining patterns, we treated imbibed seed with Pontamine S4B and Calcofluor White and examined them using confocal microscopy. Pontamine S4B shows primary binding affinity to cellulose ( Fig 7D) and xyloglucans to a lesser degree, while Calcofluor White is less specific, showing binding affinity to cellulose, arabinans and pectic galactans (Herth and Schnepf, 1980;Anderson et al., 2010). A marked loss of fluorescence was observed within the mucilage itself in cesa5-1 and cesa2-1cesa5-1cesa9-1 mutants ( Fig 7E). However, we were still able to see staining associated with rays extending from the columella (Fig 7D &   7E). To examine this phenotype in greater detail we observed mucilage imbibition after staining with Calcofluor White or Pontamine S4B at high magnification ( Fig 7F). Pontamine S4B staining showed the presence of rays in the cesa5-1 and cesa2-1cesa5-1cesa9-1 mutants, but a distinct lack of fluorescence within the mucilage itself (Fig 7F). Examination of fluorescence after Calcofluor White imbibition (five min) revealed the presence of brightly stained mucilage surrounding the columella in wild type seeds, which was absent in the cesa2-1cesa5-1cesa9-1 mutant (Fig 7F). An additional means to observe this phenotype used consecutive sections from fixed mature seed that were stained with Pontamine S4B and imaged with the confocal microscope. These images were grouped to form a movie that represented a series of cross sections from wild type and cesa2-1cesa5-1cesa9-1 (Movie S1 and S2). These data showed the rays, stained with Pontamine S4B, arising from the columella in both wild type and mutant seeds and a significant loss of fluorescence from the mucilage. Overall our results support a mechanism of matrix attachment between the mucilage halo and the seed via cellulose that requires CESA5. cesa5-1 and cesa2-1cesa5-1cesa9-1 Cellulose creates an interconnected network within the plant cell wall via cross-linking with hemicelluloses and matrix associations with pectins (Carpita and Gibeaut, 1993). To determine whether cellulose deficiency would cause feedback changes in the composition of other cell wall sugars, we determined the monosaccharide content of seed preparations using highperformance liquid chromatography (HPLC). Soluble mucilage analysis indicated an increase in the rhamnogalacturonan I backbone sugars, rhamnose and galacturonic acid, in cesa5-1 and cesa2-1cesa5-1cesa9-1 compared to wild type and cesa2-1cesa9-1 (Table 1). Increases in rhamnose levels were significant as determined by a two-tailed t-test (p<0.05). Arabinose, galactose, glucose and xylose levels remained largely unchanged.

Cell wall monosaccharide content is altered in
To investigate the underlying cause for increases in rhamnogalacturonan I backbone sugars we next tested monosaccharide levels in whole seeds. We ground whole seeds and performed serial washes with ethanol and acetone, in addition to an amylase treatment to remove soluble mucilage, cytosolic sugars and starch, leaving an enriched cell wall fraction. Monosaccharide quantification revealed proportional differences among neutral sugars. Rhamnose levels were significantly reduced in cesa5-1 and cesa2-1cesa5-1cesa9-1 compared with wild type and cesa2-1cesa9-1 (Table 1). Arabinose levels were increased while Xylose levels decreased in cesa5-1 and cesa2-1cesa5-1cesa9-1 compared with wild type (Table 1). The reduction in rhamnose levels are consistent with the increases found in soluble mucilage levels, suggesting changes to the adherence of mucilage to the seed in cesa5 mutant lines.

Discussion
Our results illustrate the specific roles of CESA2, CESA5 and CESA9 in seed coat development and differentiation. Specifically, by reverse genetics we have identified two secondary cell wall processes occurring in the seed coat epidermal cells that require specific cellulose biosynthetic machinery. Firstly, loss of mucilage adherence properties caused by mutations in cesa5 alone link mucilage biosynthesis with CESA5. Secondly, during seed coat radial wall thickening, mutations in single, double and triple mutants for cesa2, cesa5 and cesa9 caused decreased radial wall reinforcement (measured as radial wall height- Fig 5) implicating all three subunits in this process. These data suggest that CESA subunits undergo cell-specific regulation and have functional sub-specificity, extending our knowledge of cellulose biosynthetic processes.
Late in the differentiation of seed coat epidermal cells (8-9 DPA) a cellulosic secondary wall is initiated along the membrane of the radial and outer tangential sides of the cell, resulting in a thickening of the radial wall and creation of the volcano shaped columella. We have shown previously that CESA9 is required for normal synthesis of this secondary cell wall (Stork et al., 2010). In this manuscript we show that this process also involves other CESA subunits, CESA2, and CESA5. The additive reduction in radial wall height in the cesa2-1cesa5-1cesa9-1 triple mutant relative to single and double mutants (Fig 3 and Fig 5) suggests that each subunit plays a similar role. Given the current working models for a hexameric CSC comprising three different CESAs (Taylor et al., 2003;Desprez et al., 2007), a plausible hypothesis is that CESA2, CESA5 or CESA9 combine to form a functional CSC complex, similar to the other such secondary wall complex studied in xylem cells, which require CESA4, CESA7 and CESA8 (Taylor et al., 2003), An alternative explanation is that a degree of redundancy exists as illustrated for CESA6-like proteins (Persson et al., 2007). However, reduced radial wall height in single mutants and additive phenotypes in double (cesa2-1cesa9-1) and triple (cesa2-1cesa5-1cesa9-1) mutant support functional significance for each subunit.
Prior to the synthesis of the cellulosic secondary cell wall, seed coat epidermal cells synthesize seed mucilage and deposit it in the apoplast. Upon exposure to an aqueous environment, mature seeds extrude the mucilage, encapsulating the seed (Western et al., 2000;Windsor et al., 2000). The extruded mucilage includes a loosely bound (soluble) layer that can be easily removed through simple agitation and an adherent layer that is difficult to remove completely even with strong alkali solutions (Western et al., 2000;Western et al., 2001;Macquet et al., 2007a). Mucilage consists primarily of pectin but xyloglucan and cellulose can also be detected in the adherent layer (Western et al., 2000;Willats et al., 2001;Western et al., 2004;Blake et al., 2006;Macquet et al., 2007a;Young et al., 2008;Arsovski et al., 2009). It has previously been suggested that cellulose could be involved in maintaining connections to the parent seed (Macquet et al., 2007a), but there has been little experimental evidence to support this hypothesis. We have shown here that CESA5 is required for normal levels of mucilage cellulose and formation of the adherent layer, providing evidence that at least one role of cellulose in seed mucilage is to form a network with pectins such that the mucilage is held to the seed  Although our data indicate that CESA2, CESA5 and CESA9 are required for secondary cell wall biosynthesis in the seed coat epidermis, these proteins have been associated with primary cell wall synthesis in other tissues. Indeed, current models suggest that primary cell wall cellulose biosynthesis involves CESA1, CESA3 and one of a class of redundant CESA6-related subunits including CESA2, CESA5, CESA6 or CESA9 (Persson et al., 2007). Based on our developmental analysis (Fig 4), there is little evidence that primary wall cellulose biosynthesis in the seed coat epidermis requires CESA2, CESA5 and CESA9, suggesting that CESA1, CESA3 and CESA6 are sufficient for this purpose. CESA5 could function as the sole biosynthetic unit required for sufficient cellulose production to mediate mucilage attachment. However, given everything that is known about the organization of the CSC, and the fact that not all mucilage cellulose is absent in the cesa5 mutant, it is plausible that other CESAs are involved.
Our data are consistent with the speculation that different cell types may require cell-specific cellulose deposition (Carpita and Vergara, 1998). The developing seed coat represents an excellent system to dissect the composition of CSC in two separate processes and an intriguing area for future exploration. Moreover, while it has long been held that cellulose is the structural core of the plant extracellular matrix (Green, 1962;Brown, 1985), its link to mediating adherence between parent cells has been complicated to examine. Our results show that in seed coat epidermal cells, cellulose partially fulfills a matrix attachment role. Finally, we have demonstrated the functional specialization of three CESAs in secondary cell wall deposition.
The process of secondary wall cellulose biosynthesis in seed testa cells can be likened to that of cellulosic xylem vessel reinforcement, whereby the epidermal seed coat cells must withstand the immense pressure of desiccation during seed maturation, and rehydration upon imbibition (Windsor et al., 2000) illustrating an intriguing parallel mechanism. Evidence of aberrant cell shape and morphogenesis in cesa2, cesa5 and cesa9 seed coat cells also documents the role for cellulosic secondary cell wall reinforcement in structurally supporting cell shape.
Seeds were surface sterilized using 30 % bleach for 20 min. After three sequential washes with sterile water the seeds were vernalized for 2 days at 4°C in 0.15% agar. For germination, phenotypic analysis and growth assays, the surface sterilized seeds were exposed to light for 1 h and grown in either continuous light (200 mmol m 2 s -1 ) or complete darkness at 22°C on plates containing 0.5× Murashige and Skoog mineral salts (Sigma) and 1 % agar.

Staging of Flower Age
Pollination was defined as the time when the flower was beginning to open. At this stage the stamen are beginning to grow over the gynoecium and the pollen is released (anthesis). Petals can be seen extending beyond the tops of the sepals. Flowers at anthesis were marked using water-soluble paint, and a different color was used for each day of marking.

Promoter GFP fusion constructs
Promoters of CESA2, CESA5 and CESA9 were amplified from Columbia-0 genomic DNA by polymerase chain reaction (PCR) using the following primers CESA2/Prom/SbfI/F

Ruthenium Red Staining
Seeds incubated in 0.01 % (w/v) ruthenium red (Sigma) in water for 30 min at 25°C were used to test for aberrant mucilage production (Beeckman et al., 2000). Alternatively seeds were shaken on an orbital shaker for 2 hours, then stained with 0.01 % (w/v) ruthenium red in water for 30 min at 25°C and imaged with an Olympus MVX-10 Stereo Fluorescence Microscope.

Cellulose Staining Using Pontamine S4B
Seeds were shaken on an orbital shaker for 2 hours, then stained with 0.01 % (w/v) Pontamine Fast Scarlet S4B (Aldrich Rare Chemical Library, #S479896) and 50-150mM NaCl 2 for one hour. Seeds were then washed 5X with distilled water and imaged using confocal microscopy.

Scanning Electron Microscopy and Image Analysis
Arabidopsis seeds were attached to standard electron microscope stubs, and sputter coated with gold-palladium alloy using Hummer VI sputtering system (Anatech). Specimens were visualized using Hitachi model S-800 scanning electron microscope (Hitachi), images were captured using Evex Nano Analysis digital imaging system. Image analysis was performed using the ImageJ (W. Rasband, National Institute of Health) software. Measurement for each cell's planar area used area measurement output after tracing the polygon via the freehand selection tool (ImageJ). Data were organized in the spreadsheet program Microsoft Excel and transferred to GraphPad Prism (Prism-4, GraphPad) for generating histograms and performing statistical comparisons.

Confocal Microscopy
Whole seed were mounted in water for pro::GFP fusion analysis and Calcofluor White for seed histochemical analysis. Seed mounts were created between a slide and a cover slip separated by perimeter of vacuum grease that acted to keep seed in contact with Calcofluor White and avoid crushing seed structures. Once mounted, specimens were imaged in darkness, after being exposed to Calcofluor White for 5 min or imaged immediately for pro::GFP fusions.
Imaging was performed on an Olympus FV1000 laser scanning confocal microscope using a 63× N.A 1.4 water-immersion objective. The microscope is equipped with lasers for excitation wavelengths ranging from 405-633 nm; Calcofluor White was visualized at 405 nm while pro::GFP fusions were visualized at 488 nm. All image processing was performed by using Olympus Fluoview software (Olympus) and ImageJ (W. Rasband, National Institutes of Health, Bethesda, MD) software.

Resin Embedding for Bright Field Microscopy
Siliques staged at 4, 7, and 11 DPA were removed from the plant and dissected using a razor blade. The seeds were removed and either an insect pin, or a razor blade was used to puncture the seed coat. High pressure freezing, freeze substitution, and resin embedding were performed as described (Rensing et al., 2002). Samples were loaded onto copper hats (Ted Pella) containing 1-hexadecene and frozen under high pressure using a Leica EM HPM 100 High Pressure Freezer (Leica). Copper hats were then transferred to frozen cryovials containing freeze substitution medium consisting of 2 % (w/v) osmium tetroxide in acetone with 8% (v/v) dimethoxypropane. Freeze substitution was performed for 6 days at -80°C by incubation in a Leica EM AFS chamber (Leica), followed by an incubation at -20°C over 20 h to allow for reaction of the fixatives, then the temperature was increased to 4°C. Samples were then removed from the copper hats and rinsed in anhydrous acetone several times and slowly infiltrated and embedded in Spurr's epoxy resin (Canemco) (Spurr, 1969). Mature seeds were pierced with a razor blade and fixed in 3 % glutaraldehyde in 0.1 M PO 4 buffer pH 7.0, then osmicated with 1 % (w/v) osmium tetroxide in 0.05 M PO 4 buffer pH 7.0 and washed 3 times with dH 2 O followed by dehydration in increasing percent ethanol washes. All samples were then transferred to a propylene oxide solution and slowly infiltrated with Spurr's epoxy resin (Canemco). Resin embedded samples were then thick sectioned (0.5mm) using a Reichert Ultracut E microtome (Reichert) and stained with 1 % (w/v) toluidine blue O in 1X (w/v) sodium borate (pH 11). Sections were examined by light microscopy and evaluated for intact seed coat Photographs were taken using an Axioskop 2 microscope (Carl Zeiss) and Q Capture pro imaging software (Q imaging).

Mucilage Precipitation
The mucilage content was determined by using ammonium-oxalate method (Arsovski et al., 2009;Stork et al., 2010). 50 mg (dry weight) of Arabidopsis seed were incubated in 0.2 % (w/v) of aqueous ammonium-oxalate solution at 37°C for 8 hours. To facilitate extraction of mucilage, test tubes were vortexed for 1 min every hour. The seeds were removed by centrifugation after incubation and the resultant supernatant was transferred to a fresh tube. Five volumes of ethanol were added to the supernatant followed by 30 min incubation on ice to precipitate the polysaccharides. Samples were centrifuged in a bench-top centrifuge for 30 min at 21,000g.
Precipitated mucilage was washed with 70 % ethanol, air dried, and weighed. Average weights were calculated based on three mucilage replicates for mutant and wild type seeds.

Cellulose Content and Monosaccharide Analysis
Cellulose content of seed cell wall was measured colorimetrically (Updegraff, 1969) and neutral sugar composition was determined by HPLC. Briefly, cell wall material from mature seed samples were prepared by sequential washing (five times) with 70 % ethanol for 45 min at 70°C followed by five sequential acetone washes at room temperature for 2 min each. To remove the starch, the vacuum dried samples were re-dissolved in 1 mL of α-amylase solution (Sigma Aldrich no. A8220) according to manufacturer's instruction and incubated for 15 min at 37°C. The α-amylase treated samples were washed and vacuum dried. Exactly 4mg of the airdry cell wall material was measured into glass tubes and standards of different neutral sugars (Arabinose, Fucose, Galactose, Glucose, Mannose, Rhamnose and Xylose) were added to individual glass tubes (5.6 µL of 10 mM stock) and dried at 60°C. 35 µL of 72 % w/v H 2 SO 4 was added to each sample and standard incubated on ice for 2 hours, mixing every 30 minutes with the vortex. Samples were diluted to 4 % w/v H 2 SO 4 by adding 965 µL water and autoclave at 121°C for 1 hour.
For soluble mucilage samples, ~20 mg of seeds (exact weight recorded) were hydrated with 1.2 mL of water and 10 µL of 5 mg mL -1 D-erythritol was added as an internal standard and the samples were shaken on an orbital shaker for 2 hours. 1 mL of mucilage was removed and dried at 60°C under a stream of nitrogen. Sulphuric acid (72 % W/V, 17.5 µL) was added to each sample, and they were incubated on ice for 2 h and shaken every 30 mins. Water (480 µL) was then added to give a final concentration of 4 % sulphuric acid, and the samples were autoclaved for 60 min at 121°C before being filtered through 0.45-mm nylon syringe filters.
Neutral sugar standards (fucose, arabinose, rhamnose, galactose, glucose, mannose, and xylose) and acid sugar standards (galacturonic acid), were also processed in the same way at the same time.
Neutral-and acidic-cell wall sugars were identified and quantified by Pulsed Electrochemical Detection (PED) using a Dionex ED50. Sugar separation was achieved using a BioLC GS50 HPLC and CarboPAC-PAl pellicular anion exchange column with column guard (Dionex, Sunnyvale, CA, USA). Column and detector temperature was maintained at 30°C using a Dionex LC25 chromatography oven. Samples (25µL) were introduced to the column from a Dionex AS50 autosampler, and, for neutral sugars, were eluted isocratically in 22 mM sodium hydroxide (NaOH) over a 30-min period after which the column was washed in 1 M sodium acetate (NaOAc), 200 mM NaOH for 5 min. The column was recharged by passing 200 mM NaOH through it for 10 min, then re-equilibrating the column in 22 mM NaOH for 10 min prior to injection of the next sample. This protocol resulted in optimal (near baseline) resolution of galactose, glucose, mannose and xylose (Downie and Bewley, 1996). Acidic sugar separation and quantification used the same HPLC but followed the eluent profile of (Dean et al., 2007).
For soluble mucilage, monosaccharides were quantified with the D-erythritol internal standard after correction of response factors with monosaccharide standards of different concentrations to allow area to be converted to molar amounts. Sugars from mucilage were normalized to the mass of seed used in the extraction. For cell wall monosaccharides the neutral sugars were normalized to the amount of cell wall material used for the sample preparation.     in Calcofluor White solution using a 60 X objective and DAPI filter. A) 3-dimensional Z-plane animation of Calcofluor White labeled wild type (Column 1) and cesa2-1cesa5-1cesa9-1 (Column 2) seed coat cells acquired by merging individual frames taken across a Z-plane shift (at 0.5 µm steps) from the top of the columella to the base of the tangential wall (scale bar = 10 µm). Cells were rotated from surface to side view illustrating the elevation and fluorescence difference observed between wild type and triple mutant seed coats radial walls. A surface view of Calcofluor White stained seed coat epidermal cells for cesa2-1, cesa5-1, cesa2-1cesa9-1 and cesa2-1cesa5-1cesa9-1 (Column 3)(Scale bar = 30 µm). B) Fluorsecence intensity plots generated across a transect that spanned a single cell for wild type, single, double and triple mutants. C) A high magnification view of individual cells from Calcofluor White labeled epidermal cells in wild type and cesa2-1cesa5-1cesa9-1 visually illustrates the reduced radial wall fluorescence in cesa2-1cesa5-1cesa9-1 (scale bars = 15 µm)   cesa2-1, cesa2-2, cesa5-1,   cesa9-1, cesa2-1cesa9-1 and cesa2-1cesa5-1cesa9-1 seed coat cells The asterisk (*) in C indicates significant differences from wild type (p < 0.05 Student's t test).  confocal microscopy) labeled seed coat cells in wild type and cesa2-1cesa5-1cesa9-1. Scale bar = 10 µm.

Supplemental Data
Movie S1. Cell morphology in wild type seed coats. Calcofluor White stained seed coats imaged at was used to create a 3-dimensional reconstruction of a Z-projection (0.5 µm resolution). Scale bar = 10 µm.
Movie S2. Cell morphology in cesa2-1cesa5-1cesa9-1 seed coats. Calcofluor White stained seed coats imaged at was used to create a 3-dimensional reconstruction of a Z-projection (0.5 µm resolution). Scale bar = 10 µm. cesa5-2 micrographs from mature seed were taken using SEM, false colored with the cell area in green and the columella in red for each genotype Scale bars = 30 µm. Planar cell area for individual cells was measured using 300 cells from 10 different seeds in wild type, cesa2-2 and cesa5-2. Both the mutants showed significant difference from wild type based on results of a Wilcoxon rank sum test (P < 0.01).   Two alleles examined for mucilage extrusion phenotype. T-DNA analysis is presented in Table   S1. Scale bar = 100 µm.  Expression of CESA2, CESA5 and CESA9. A) GENEINVESTIGATOR analysis of expression for CESA2, CESA5 and CESA9 in seed. B) Con rmation of A) by localization of promoter GFP fusions for target CESA genes. Laser scanning confocal microscopy was used to image transgenic plants expressing promoterCESA2::GFP, promoterCESA5::GFP and promoterCESA9::GFP were examined at the T2 stage by excising individual seed from siliques 10 days post anthesis and mounting them in aqueous solution between two cover slips separated by a thick perimeter of vacuum grease to protect seed from being crushed. Seed were imaged immediately after mounting (within 5 min). Scale bar = 250 µm.  Radial wall ridge displays decreased thickness in CESA mutants. To carefully examine the structural change of CESA mutant epidermal seed coat cells, Calco uor White was used to label structural β-glycans and cellular features were visualized via laser scanning confocal microscopy. Imaging was performed approximately 5 min after seeds were immersed in Calco uor White solution using a 60 X objective and DAPI lter. a) 3-dimensional Z-plane animation of Calco uor White labeled wild type (Column 1) and cesa2-1cesa5-1cesa9-1 (Column 2) seed coat cells acquired by merging individual frames taken across a Z-plane shift (at 0.5 µm steps) from the top of the columella to the base of the tangential wall (scale bar = 10 µm). Cells were rotated from surface to side view illustrating the elevation and uorescence di erence observed between wild type and triple mutant seed coats radial walls. A surface view of Calco uor White stained seed coat epidermal cells for cesa2-1, cesa5-1, cesa2-1cesa9-1 and cesa2-1cesa5-1cesa9-1 (Column 3) (Scale bar = 30 µm). b) Fluorsecence intensity plots generated across a transect that spanned a single cell for wild type, single, double and triple mutants. c) A high magni cation view of individual cells from Calco uor White labeled epidermal cells in wild type and cesa2-1cesa5-1cesa9-1 visually illustrates the reduced radial wall uorescence in cesa2-1cesa5-1cesa9-1 (scale bars = 15 μm) Wild-type cesa2-1 cesa2-2 cesa5 cesa9-1 cesa9-2 cesa2-1 cesa9-1 cesa2-1 cesa5-1cesa9-1 Radial wall height (µm) Average height of the radial wall (RW) of wild type, cesa2-1, cesa2-2, cesa5-1, cesa9-1, cesa2-1cesa9-1 and cesa2-1cesa5-1cesa9-1 seed coat cells. The asterisk (*) in C indicates signi cant di erences from wild type (p < 0.05 Student's t test). Seed size and cellulose content A. Seed weight was measured as the average of three batches of 100 seed and error bars represent the SE of three replicates. B. Acid-insoluble crystalline cellulose content of various tissues was determined colorimetrically for mature seeds, Error bars represent the SE of three technical replicates from three independent batches of seed as biological replicates. Mucilage attachment to the parent seed is mediated by CESA5 in Arabidopsis seed. A) Ruthenium red-stained seed B) Mucilage weights determined for wild type and CESA mutant seed, error bars are se of three replicates. C) water hydrated ruthenium red stained seed visualized by light microscopy in wild type, cesa5-1, cesa2-1cesa9-1 and cesa2-1cesa5-1cesa9-1. Scale bar = 100 μm (cesa2-1, cesa2-2, cesa9-1, cesa9-1 not shown due to no observed mucilage phenotype). D) Pontamine S4B and E) Calco uor White stained wild type, cesa5-1, cesa2-1cesa9-1 and cesa2-1cesa5-1cesa9-1 mature seed. Scale bar = 100 μm F) High magni cation visualization of Pontamine S4B (whole seed viewed by light microscopy) and Calco uor White (3-dimensional reconstruction of a Z-projection, side view via laser scanning confocal microscopy) labeled seed coat cells in wild type and cesa2-1cesa5-1cesa9-1. Scale bar = 10 µm.
www.plantphysiol.org on August 24, 2017 -Published by Downloaded from Copyright © 2011 American Society of Plant Biologists. All rights reserved.  Table 1. HPLC analysis of mucilage and cell wall monosaccharide composition from wild type and mutant seeds. (average of three replicates ± SE).