Phosphatidylserine Synthase Controls Cell Elongation Especially in the Uppermost Internode in Rice by Regulation of Exocytosis

The uppermost internode is one of the fastest elongating organs in rice, and is expected to require an adequate supply of cell-wall materials and enzymes to the cell surface to enhance mechanical strength. Although it has been reported that the phenotype of shortened uppermost internode 1 (sui1) is caused by mutations in PHOSPHATIDYLSERINE SYNTHASE (OsPSS), the underlying mechanism remains unclear. Here we show that the OsPSS-1, as a gene expressed predominantly in elongating cells, regulates post-Golgi vesicle secretion to intercellular spaces. Mutation of OsPSS-1 leads to compromised delivery of CESA4 and secGFP towards the cell surface, resulting in weakened intercellular adhesion and disorganized cell arrangement in parenchyma. The phenotype of sui1-4 is caused largely by the reduction in cellulose contents in the whole plant and detrimental delivery of pectins in the uppermost internode. We found that OsPSS-1 and its potential product PS (phosphatidylserine) localized to organelles associated with exocytosis. These results together suggest that OsPSS-1 plays a potential role in mediating cell expansion by regulating secretion of cell wall components.


Introduction
Cell division and anisotropic cell expansion determine the final size and shape of plant organs. Cell expansion involves "loosening" of existing cell wall architecture with synthesis and deposition of new cell wall components. Plant cell wall mainly compose of cellulose, hemicellulose, pectin, and structural proteins [1]. Cellulose is produced at plasma membrane (PM) by cellulose synthase complexes, while hemicellulose and pectin are synthesized and modified in Golgi and transported via vesicles to cell wall [2]. Mutations in genes associated with delivery of cellulose synthase genes result in altered cellulose contents and inhibition of cell elongation. CELLULOSE SYNTHASE INTERACTIVE1 (CSI1) is a microtubule-associated protein that bridges cellulose synthase (CESA) complexes and cortical microtubules, mutations in CSI1 cause defective cell elongation in hypocotyls and roots and reduce cellulose content [3]. Mutations in a PM-associated endo-(1!4)-β-D-glucanase disrupt normal wall assembly and cell elongation in both Arabidopsis [4] and rice [5]. Similarly, mutations in COBRA genes that encode glycosylphos-phatidyl inositol-anchored proteins are cellulose-deficient and compromised in organ elongation in Arabidopsis [6,7] and rice [8]. Xyloglucans and pectin are also involved in elongation growth. A recent study showed that improper secretion and accumulation of cell wall matrix polysaccharides including xyloglucan and pectin are associated with phenotypes of dwarf, curled rosette leaves, short petioles and short inflorescence stems [9].
In rice (Oryza sativa L.), elongation of internodes is controlled by cell division in the basal intercalary meristem and elongation of individual cells between nodes [10]. During the heading stage, rapid elongation occurs in the uppermost internodes (>5 cm per day; [11]). Several mutants defective in uppermost internode elongation have been identified in rice [12][13][14][15]. One of these mutants, shortened uppermost internode 1 (sui1), is caused by mutations in SUI1 that encodes phosphatidylserine (PS) synthase (PSS) [16,17]. In Arabidopsis, disruption of the gene encoding PSS reduces pollen fertility and plant height [18]. Mutations of PSS in Schizosaccharomyces pombe (pps1Δ) and the fungus Candida albicans (cho1Δ/Δ) lead to slow growth due to cell-wall defects [19,20]. These results imply an unidentified role of PSS in regulation of cell expansion.
The primary role of PSSs is PS biosynthesis. In bacteria and yeast, PSS catalyzes the conversion of CDP-diacylglycerol and L-serine into PS. In the endoplasmic reticulum of cultured Chinese hamster (Cricetulus griseus) ovary (CHO) cells phosphatidylserine synthase with baseexchange-type (BE-PSS) primarily utilizes phosphatidylcholine (PC, Cg-PSS1) or phosphatidylethanolamine (PE, Cg-PSS2) to synthesize PS [21]. As the most abundant negatively charged phospholipid in the cell, PS directs the binding of proteins with PS-recognition modules by electrostatic association of polycationic ligands with cellular membranes, as well as various physiological processes such as the coagulation cascade [22,23], recruitment and activation of signaling molecules [24], clearance of apoptotic cells [25,26], and vesicular trafficking [27][28][29][30]. The cho1 mutant of Saccharomyces cerevisiae displays compromised PS synthesis and defected sporulation [31] and fusion of vacuolar vesicles [32]. In Arabidopsis and rice, PSS possesses enzymatic activity that catalyzes PC or PE into PS in yeast [17,18]. Verexpression of the wheat phosphatidylserine synthase gene induced necrotic lesions on Arabidopsis leaves [33]. However, the mechanism underlying the link between PSS and plant development remains to be clarified.
Here, we showed that mutation in OsPSS-1 in rice leads to defected cell expansion and compromised cell wall biosynthesis. The shortened uppermost internode in sui1-4 was mainly attributed to the disrupted secretion and deposition of cell wall components. Protoplast isolated from sui1-4 mutant displayed a secretion defect. OsPSS-1 was expressed predominantly in elongating tissues, and OsPSS-1 was co-localized dynamically with organelles associated with exocytosis. We also showed that mutation in OsPSS-1 led to reduced PS contents. Our results revealed a potential role of OsPSS-1 in cell wall component trafficking. was attributed to reduced length of both panicles and internodes, especially the uppermost internode ( Fig 1B-1F), accompanied with reduced fertility, decreased grain size and slightly increased tiller number (Fig 1A; S1C-S1H Fig). The mutant resembled the phenotype of other sui1 mutants [16,17] and was designated as sui1-4. Investigations of the shortened second internode revealed slightly reduced cell length and width in longitudinal sections and cell radii Transmission electron microscopy revealed that, compared to the longitudinally elongated and well-arranged parenchyma cells in WT plants (Fig 2A), parenchyma cells in the sui1-4 were irregular in shape and had enlarged cell corners (Fig 2B).
Cell-wall components are altered in sui1-4 Cell wall components were analyzed in sui1-4 to find causes for the changes relative to WT. Cellulose contents in developed organs of sui1-4 plants were lower than those in the WT plants (uppermost internodes, 12% lower; 2nd internode, 11.7%; leaf sheath, 8.3%). In 7-day-old seedlings, significant reductions in cellulose content were detected in root tips (4.6% less), and in leaves (9.2% less). The results suggested inadequate cellulose content in sui1-4 plants ( Table 1). We further examined other cell wall components in sui1-4 plants, especially in the uppermost internode. The amount of glucose in cell-wall lysates, contributed from non-crystalline cellulose, xyloglucan and mixed linkage glucan, was reduced from 65.1 mg/g in WT plants to 35.7 mg/g in sui1-4 plants. In contrast, both xylose and galacturonic acid (GalA), the major components of hemicelluloses and pectin originating in the Golgi body [34], were enriched in the uppermost internode of sui1-4 (Table 1). Thus, cell wall components in sui1-4 plants were quite different from WT.

Defective pectin secretion in the uppermost internodes of sui1-4
We surveyed the pectin distribution pattern of uppermost internode at the ultrastructural level by immunogold (15 nm gold) labeling using JIM7 antibody (15 nm in diameter) [35]. Pectins localize in primary cell walls, middle lamella, and cell corners, and play a major role in intercellular adhesion [36]. They are initially synthesized in a methyl-esterified form in the Golgi apparatus, and then delivered to the PM and cell walls by secretory vesicles [2]. Two weeks prior to heading, gold particles were restricted to the primary walls of WT parenchyma cells, especially to the middle lamella and periphery of intercellular spaces; there was no gold-labeled pectin in the cytoplasm (Fig 2C). However, in sui1-4 cells, large amounts of gold-labeled pectins were observed in large compartments in the cytoplasm (Fig 2D and 2E; S3A and S3B Fig). One week prior to heading, although some gold particles were scattered in the middle lamella of sui1-4 cell wall, most of the periphery of the intercellular spaces lacked gold labeling in comparison with well distributed gold particles in the intercellular spaces (Fig 2F and 2G). Additional, some pectin-rich "clumps" appeared in the outermost layer of the cell wall and protruded from the edge of enlarged intercellular spaces in the cell cap (Fig 2F and 2H; S3D Fig). To confirm these potential defects in pectin distribution, we performed high-pressure freeze substitution (cryofixation). Similar JIM7-labeled pectin clumps were observed in the outermost regions of the cell wall of sui1-4 plants (S3E and S3F Fig), suggesting that the uppermost internodes were defective in pectin secretion and deposition.

sui1-4 plants exhibit compromised secretion
To determine if any soluble proteins or cell-wall-synthesis related enzymes have defective transport to the PM in sui1-4, we investigated trafficking of vesicles carrying OsCESA4, a member of the cellulose synthesis complex, which is believed to be synthesized in the endoplasmic reticulum (ER) and transported to PM via the trans Golgi network (TGN) [37,38]. We then investigated whether most of the synthesized CESA was transported from endo-membranes (indicated by dextran fraction) into PM (indicated by polyethylene glycol fraction) labeled by OsCESA4 polyclonal antibodies [39]. On fractional analysis to young seedling in WT and sui1-4, OsCESA4 mainly appeared in the polyethylene glycol fraction of WT plants, indicating its presence in PM, whereas in sui1-4 plants, strong labeling was observed in the dextran fraction, reflecting its presence in endomembranes (Fig 3A and 3B). This result suggests that a large amount of OsCESA4 remained in the cytoplasm in the mutant, most likely due to failure in delivery to the PM.
To further evaluate the secretion defect, we employed a protoplast-based transient assay using a construct consisting of a secreted form of green fluorescent protein (secGFP) as an indicator [40]. We expected that secretion of secGFP to the apoplast may allow us to detect secGFP in the medium fraction, while perturbation of secretion would result in intracellular secGFP accumulation. Immunoblotting revealed that most of the secGFP was detected in the culture medium when protoplasts from the WT were used, whereas in protoplasts from sui1-4 most secGFP was detected in protoplasts, with a very low amount detected in the medium (Fig 3C  and 3D), suggesting an compromised secretion process in sui1-4.
The sui1-4 phenotype is caused by a mutation in OsPSS-1 that is expressed predominantly in elongating tissues Genetic analysis revealed that the sui1-4 phenotypes were inherited as a single nuclear recessive mutation (S2 Table). Map-based cloning of an F 2 population made by crossing PA64 (an indica variety) with sui1-4 revealed that the sui1-4 phenotype was due to an A-to-T point mutation To verify that the mutation in OsPSS-1 caused the phenotype, a construct containing 2,259 bp of the 5' upstream sequence, the entire coding sequence, and 650 bp of the 3' downstream sequence was transformed into sui1-4 plants. Among the transgenic plants obtained, several lines with WT phenotype were identified, suggesting that the genomic DNA sequence used here was capable of complementing the sui1-4 phenotype (Fig 4B and  4C). Thus, the mutation in OsPSS-1 caused the sui1-4 phenotype. When OsPSS-1 expression was down-regulated while transcript levels of OsPSS-1 homologs were not affected using an RNA interference (RNAi) construct under the control of the ubiqutin promoter, transgenic plants displayed a phenotype that mimicked the sui1-4 phenotype (Fig 4D-4F), suggesting that reduced OsPSS-1 expression is sufficient to generate the dwarf phenotype.
Quantitative reverse transcription polymerase chain reaction (qRT-PCR) analyses of WT plants revealed that OsPSS-1 is expressed in various organs, including roots, culms, and leaves, with the highest expression in panicles ( Fig 5A). We generated a reporter construct using a 2.7-kb fragment of the OsPSS-1 5' sequence fused with the gene encoding β-glucuronidase (GUS). In young transgenic seedlings, GUS expression occurred predominately in the elongation zones of roots, root hairs, coleoptiles, and occasionally at the tips of newly formed leaves (Fig 5B and 5G). No expression was observed in shoot and root meristems. Four weeks prior to heading, strong GUS staining was found in young roots, the basal parts of internodes (divisional and elongating zones) originating from the intercalary meristem, and shoot apices ( Fig  5C). Two weeks prior to heading, GUS staining shifted to the basal regions of the uppermost internode and the glumes (Fig 5D). During the heading stage, GUS expression was observed mainly in the uppermost internode (Fig 5E and 5F). Further examination of freehand sections of these tissues showed that GUS was ubiquitously expressed in all elongating cells, including parenchymal cells (Fig 5H). These results collectively suggested that OsPSS-1 expression is necessary for cell elongation in various tissues.

OsPSS-1 localizes to organelles associated with exocytosis
To elucidate subcellular localization of OsPSS-1, a construct was made with the OsPSS-1 C-terminal fused with the GFP N-terminal, expressed under the control of the CaMV 35S promoter (p35S: OsPSS-1-GFP), and transformed into the sui1-4. This construct rescued the sui1-4 phenotype completely (S5A and S5B Fig). Fluorescence microscopy visualized strong GFP signals We purified the cytosolic and membrane protein fractions in transformed protoplasts isolated from WT seedlings. OsPSS-1 was primarily found in the membrane fraction, as indicated by staining with the membrane-specific antibody anti-H+-ATPase (Fig 6A). Most likely OsPSS-1 was an integral membrane protein because its presence in the supernatant depended on treatment with membrane-dissolving agents (1% sodium dodecyl sulfate (SDS) or detergent (1% Triton X-100)), but not with high salt (1 M NaCl) or high pH (pH 11) treatments ( Fig  6A).
Further examinations in the WT protoplasts carrying p35S:PSS1-GFP suggests that OsPSS-1 is localized to the ER and PM (S5E and S5F Fig). To further elucidate the subcellular localization of OsPSS-1, a construct carrying the OsPSS-1-GFP fusion protein under control of the 35S promoter was co-transformed into Arabidopsis protoplasts transiently with constructs of monomeric red fluorescent protein (RFP) fused to either HDEL (an ER marker; [41]), ManI (a marker of the cis-Golgi apparatus; [42]), VSR2 (a marker of the prevacuolar compartment; [43]), SYP61 (a TGN/EE marker; [44]), VIT1 (a tonoplast marker; [45]), or SCAMP (a PM marker; [44]). When examined at 12 hours after transformation under confocal microscope, the OsPSS-1-GFP fluorescence was co-localized only with the ER marker HDEL (Fig 6B-6G). However, after 36 h of incubation, the ER co-localization was weakened, and additional green fluorescence appeared in some unidentified endomembrane compartments (Fig 7A). These compartments with green fluorescence were co-localized with the cis-Golgi apparatus (Fig 7B), TGN/EE ( Fig 7C) and PM markers (Fig 7D). However, the OsPSS-1-GFP signal was independent from markers of the prevacuolar compartment ( Fig 7E) and tonoplast (Fig 7F). Localization patterns were independent of the position of OsPSS-1 relative to GFP in the transformed construct ( S6 Fig). Therefore, OsPSS-1 appeared to co-localize dynamically with the Golgi bodies, TGN/EEs and PM in accordance with the secretary pathway.

PS has a similar localization pattern to OsPSS-1
Previously, PS could be detected with the biosensor Lact-C2 [28]. To inquire whether there is any difference in subcellular localizations of OsPSS-1 and PS, we observed the co-localization of PS with organelles that participate in secretion. Twelve hours after transformation, although no clear signal was visualized in the ER (Fig 8A) GFP-LacC2 signal was observed in the cis-Golgi (Fig 8B), the TGN/EEs (Fig 8C), the PM (Fig 8D), and absent from the PVC (Fig 8E) and the tonoplast (Fig 8F). The localization pattern of GFP-LacC2 was similar to that of OsPSS-1 36 h after transformation (Fig 7). The same subcellular localization patterns of PS were observed when GFP-LacC2 was transiently expressed in rice protoplasts from WT and sui1-4 plants ( S7 Fig). PS content is decreased in sui1-4 Yin et al. inferred that OsPSS functions as PS synthase from co-localization of PS with the secretary pathway [17].We determined the PS contents in panicles and the second internodes in both the WT and sui1-4 plants. The results showed that PS content in panicles was reduced from 151.2 μg/g in WT to 116.8 μg/g in sui1-4 (Fig 9A), and in the second internode it was reduced from 171.3 μg/g to 142.9 μg/g (Fig 9B). Conversely, the content of the PSS substrate phosphatidylethanolamine (PE) was increased from 1,689.1 μg/g in WT panicles to 1,865.4 μg/ g in sui1-4 panicles (Fig 9A), and from 1,189.5 μg/g to 1,509.1 μg/g in WT and sui1-4 in the second internodes (Fig 9B), suggesting that the mutation in OsPSS-1 was associated with decreased PS content.

Discussion
Cell expansion is a fundamental, dynamic cellular process driving plant growth through regulation of rate and direction. In this study, we showed that OsPSS-1, which is critical for cell expansion, is expressed predominantly in elongating cells. OsPSS-1 is similarly localized with PS in organelles involving exocytosis. Mutation in OsPSS-1 leads the PS reduction, resulting in compromised secretion and defective deposition of cell wall components in sui1-4 plants. Our results revealed a potential role of OsPSS-1 in mediating cell expansion and biosynthesis of cell wall components.

Defective secretion of cell-wall components causes shortened uppermost internode
In stems, the architecture of the cell wall permits longitudinal elongation of cells while restricting radial expansion leading to highly asymmetric and anisotropic growth. The uppermost internode in rice is the fastest elongating organ during the heading stage [11]; consequently an adequate supply of cell-wall materials is essential in order to support such rapid elongation. Pectin, a highly heterogeneousmixture of polysaccharides, is a crucial component of cell-cell adhesion [46]. Mutations affecting pectin biosynthesis [36,47] and vesicular secretion [48] could cause intercellular attachment loss. In our study, cell wall components in the shortened uppermost internode of sui1-4 were dramatically altered, including decreased cellulose levels and increased pectin and hemicellulose levels ( Table 1). Reduced cellulose content was likely caused by compromised trafficking of CESA complexes to the PM as a large amount of CESA4 was accumulated in the endomembrane fraction of cytoplasm from the sui1-4 plants (Fig 3A). Increased pectin and hemicelluloses contents would been explained by compensatory mechanism triggered by reduced cellulose content. The initial accumulation of pectin-containing compartments in the cytoplasm, and its eventual movement to the intercellular spaces (Fig 2), together with generally defective exocytosis indicated by secGFP secretion defect (Fig 3C), suggested that impairment of cell wall component secretion was the main factor causing cell elongation especially in uppermost internode.

OsPSS-1 participates in exocytosis
Exocytosis occurs in most polarized, growing cells, including root hairs and pollen tubes of plants [49]. In rice, longitudinal cell elongation in roots and internodes adopts a similar, polarized growth pattern [50,51], indicating that exocytosis contributes to cell elongation. The results of GUS staining indicated that OsPSS-1 is mainly expressed in rapidly elongating cells including roots and root hairs, especially in the uppermost internodes, rather than in meristems or fully developed tissues (Fig 5B-5F), corresponding to where exocytosis should be active, and suggesting that OsPSS-1 expression is associated with exocytosis in elongating cells.
OsPSS-1 is an integral membrane protein. Its subcellular localization, which was established 12-36 h after transient transformation of protoplasts (Figs 6 and 7), suggested that OsPSS-1 is synthesized in the ER and transferred to the PM through the TGN, a pathway reminiscent of exocytosis. In cell wall-forming plant cells, exocytotic vesicles enclose cell-wall matrix  Abnormal cellulose deposition indicated a possible trafficking defect in cellulose synthase complexes (CSCs). Small CESA compartments (SmaCCs) or microtubule-associated cellulose synthase compartments (MASCs) are critical for CSC recycling [52,53]. Recent work indicated  Role of OsPSS-1 in Plant Development that CELLULOSE SYNTHASE INTERACTIVE1 might be required for fast recycling of CSCs to the plasma membrane in Arabidopsis [54]. Detailed investigation on whether OsPSS-1 is specifically associated with SmaCCs/MASCs would be very informative in revealing the underlying mechanism.
What is the role of PS in exocytosis?
In eukaryotic cells, membrane traffic and protein cargo delivery require accurate recognition and fusion of transport vesicles with acceptor membranes. As an important pathway of tethering post-Golgi secretory vesicles to the PM, exocytosis is mediated by the evolutionarily conserved exocyst complex, composed of Sec3, Sec5, Sec6, Sec8, Sec10, Sec15, Exo70 and Exo84 subunits [55]. Exocyst complex function relies on intricate interactions between exocyst and secretory vesicles, but importantly also between exocyst and specific membrane lipids [56]. In mammalian cells, PI(4,5)P2 (phosphatidylinositol 4,5-bisphosphate) has significant binding specificity with Exo70 over PS in mediating the targeting of exocyst to the PM [57]. PS is considered to be essential in mammals for the exocyst complex to combine with secretory vesicles [58]. In rice, three isoforms of PSS have been identified. OsPSS-1 complements the phenotype of yeast PS-deficient mutant cho1 [16,17] and the amount of PS is reduced in sui1-4 (Fig 2), indicating OsPSS-1 potentially functions as a PSS to regulate phospholipid metabolism.
To investigate whether the exocyst complex was involved in the general secretion defect of sui1-4, we examined expression levels of exocyst subunit-encoding genes. Apart from the Exo70 subunit that consists of 47 paralogs in 9 clades (designated as Exo70A to Exo70H; [59,60]) other exocyst complex subunits in rice are encoded by either single genes or a few family members. The qRT-PCRs were performed for 12 selected culm-expressed Exo70 paralogs and other exocyst subunit genes in sui1-4 and WT. The results indicated that in young uppermost internodes of sui1-4, significantly reduced expression was detected with OsSec5, OsSec6, OsSec8 and OsSec10. For the two-member families OsSec3, OsSec15 and OsSec84, at least one gene had reduced expression level in sui1-4 (S8 Fig). Of 12 Exo70 paralogs tested, reduced expression was significant for OsExo70A1, A3, B1, E1, F1, G1, H2 and H3 in both young uppermost internodes (S8 Fig). Down-regulated expressions of exocyst subunits encoding genes forced us to determine whether OsExo70 subunits have potential of binding to the membrane lipids. Lipid-overlay assay showed that OsExo70 subunits could bound to PS in vitro (data not shown). Although there is no difference in PS localization due to OsPSS-1 between WT and sui1-4 (S6 Fig), the less amount of PS existed in PM speculated by total reduced amount of PS perhaps affect targeting efficiency of the Exo70 subunits to PM in sui1-4. Whether PS content in PM was reduced in sui1-4 compared to WT needs to be confirmed further. Take together, the results above implied that OsPSS-1 might play a critical role in exocytosis through PS. It will be interesting to explore whether other exocyst family members would interact with PS and the underlying mechanism of PS involvement in exocytosis.

Plant materials and growth conditions
sui1-4 is a mutant derived from japonica cv. Kitaake. The mutant was crossed with indica cv. PA64 to construct an F 2 mapping population. All rice materials used in this study were cultivated in the experimental fields of the Institute of Crop Science in Beijing or Sanya during the natural growing season or grown in pots in a greenhouse at 25-28°C with 16 h of light and 8 h of darkness.

Map-based cloning of OsPSS-1
The OsPSS-1 locus was mapped and cloned using 624 F 2 individuals that exhibited the mutant phenotype. Primers for newly developed indel markers based on the comparison of genomic sequences from indica cv. PA64 and japonica cv. Kitaake (S1 Table) were employed to finemap the mutated locus. For complementation analysis, the entire OsPSS-1 coding region and its upstream and downstream sequences were amplified from WT genomic DNA using the primer pair OsPSS-1-HindIII-F/OsPSS-1-PstI-R (S1 Table). The resulting PCR fragment was ligated into the HindIII and PstI sites of pCAMBIA1305 (Cambia) using an infusion cloning kit (Clontech) to generate pCAMBIA1305-OsPSS-1, which was introduced into Agrobacterium tumefaciens strain EHA105 via electroporation and into the rice mutant as described previously [61].

Gene expression analysis
Total RNA was extracted from various plant tissues using the RNAprep Pure Plant Kit (Tiangen). One μg of total RNA was reverse transcribed to cDNA using the SuperScript II Kit (TaKaRa) and primer pairs RTcF1/R1 (targeting OsPSS-1) and ActinF/ActinR (targeting the reference gene Actin). All qRT-PCRs were performed using the SYBR Premix Ex Taq Kit (TaKaRa) on a 7900HT Real-time PCR System (Applied Biosystems) according to the manufacturers' instructions. The rice Ubiquitin gene was used as an internal control. Expression levels were calculated from three biological replicates, each consisting of three technical replicates. Gene-specific primers are listed in S1 Table. For the GUS assay, 2.7 kb of OsPSS-1 promoter sequence were amplified using the primer pair OsPSS-1-Pro-EcoRI-F/OsPSS-1-Pro-NcoI-R (S1 Table). The resulting construct, ProOsPSS-1::GUS, was transformed into cv. Kitaake mediated by A. tumefaciens. Histochemical GUS staining was performed as previously described [63].

Subcellular localization of OsPSS-1
To obtain GFP-expressing transgenic plants, OsPSS-1 was amplified with primer pair 1305OsPSS-1-GFP-SmaI-F/R (S1 Table). The fragment was cloned into the SmaI site of pCAMBIA1305-GFP (produced by inserting the backbone behaving GFP from vector pAN580 into the SacI-SalI site of pCAMBIA1305) to yield vector pCAMBIA1305-D35S-OsPSS-1-GFP-NOS. This construct was introduced into WT and mutant plants to generate transgenic plants. Fresh roots from 2-day-old T 1 seedlings were used for detecting GFP signals with confocal microscopy.

Lipid measurements
Lipids were extracted from rice panicles and second internodes as described previously [18]. The phospholipids were dried by evaporation under nitrogen and stored at -20°C. Prior to analysis, samples were dissolved in chloroform. High-performance liquid chromatography was employed to measure phospholipid components. The analytical column was a YMC-Pack SIL-06 silica column (4.6 mm × 250 mm, 5 μm), and a Phenomenex silica column (3.0 mm × 4 mm) was used as a guard with the diode array detector at 205 nm. The temperatures for the column and detector were set at 30°C and 25°C, respectively. Twenty μL of the following solutions were injected into the column at a flow rate of 1 mL/min: solution A (chloroform:methanol:water:ammonium hydroxide 75:24:0.5:0.5) and solution B (chloroform:methanol:water: ammonium hydroxide 55:39:5.5:0.5). Solutions were used at 100% A~100% B at 0~20 min, 100% B at 20~40 min, 100% B~100% A at 40~41 min, and 100% A at 41-50 min.

Chemical analysis of cell-wall components
Two grams of the uppermost internode from WT and sui1-4 plants were extracted sequentially with ice-cold 80% ethanol, 100% ethanol, chloroform:methanol (1:1), and 100% acetone [64]. Destarched alcohol-insoluble residues were hydrolyzed in 2 M trifluoroacetic acid at 120°C for 90 min, then centrifuged. Supernatants were dried by evaporation under nitrogen and treated with sodium borohydride (10 mg/mL in 1 M ammonium hydroxide). After 24 h, the resulting alditol was extracted in ethyl acetate for analysis of monosaccharide composition with an Agilent 7890 gas chromatograph equipped with a 5975C mass selective detector (Agilent).
Pellets for cellulose analysis were treated with trifluoroacetic acid and hydrolyzed with Updegraff reagent. Cooled pellets were washed with acetone and hydrolyzed with 72% sulfuric acid. The anthrone assay was applied to measure cellulose content [65].

Confocal laser scanning microscopy and transmission electron microscopy
For labeling of the cell wall, semi-thin (0.5 mm) sections were directly stained with calcofluor white (Sigma-Aldrich) for 5 min at room temperature, washed three times with phosphatebuffered saline, and subjected to confocal imaging with a Zeiss LSM700 laser scanning microscope. For transmission electron microscopy, intercalary meristem zones of sui1-4 and WT plants were fixed overnight in 2% glutaraldehyde at 4°C, then exposed to 2% w/v OsO 4 for 1 h. The samples were dehydrated under 30,50,70,90, and 100% ethanol, infiltrated, and embedded in a low-viscosity medium with the Spurr Kit (Sigma). Samples were cut into ultra-thin sections (90 nm) using an Ultracut E Ultramicrotome (Leica) and mounted on formvar-coated copper grids. After post-staining with uranyl acetate and lead citrate, the sections were observed under a Hitachi H7500 transmission electron microscope.
High-pressure freezing and subsequent immunogold labeling were performed as described previously [66]. Immunolabeling of ultra-thin serial sections was performed using standard procedures [67] with JIM7 primary monoclonal antibody (purchased from the University of Georgia, USA) diluted 1:50 and anti-rat IgG-gold antibody (Sigma-Aldrich) diluted 1:20. For negative contrast, anti-PIP1s antibody was purchased from Agrisera.

Transient expression analysis of Arabidopsis and rice protoplasts
WT OsPSS-1 was amplified with primer pair OsPSS-1-GFP-BamHI-F/R (S1 Table). The fragment was cloned into the BamHI and BglII sites of vector PA7-GFP to produce constructs OsPSS-1-GFP and GFP-OsPSS-1, respectively. A LactC2 fragment was amplified from plasmid p416-GFP-LactC2 (Haematologic Technologies) using primers C2 domain-F/R (S1 Table) and inserted into the BglII site of PA7-GFP. OsPSS-1-GFP and GFP-LactC2 were co-transformed into rice or Arabidopsis protoplasts with various organelle markers and markers of the endomembrane system. Transient expression using Arabidopsis protoplasts from cell suspensions and rice protoplasts from rice leaf sheaths was described previously [68,69].
Confocal images were acquired at specific time points after transformation using an Olympus FV1000 system. For each experiment, more than 20 individual cells were imaged. For statistical analysis, the PSC colocalization plug-in [70] for ImageJ [71] was used to calculate the linear Pearson correlation coefficient (r p ) of red and green fluorescent signals. We changed red fluorescent signal to magenta to make it easy for colorblind individuals to interpret the data in figures.

Subcellular distribution of OsPSS-1 and CESA4
Whole protein extracted from WT protoplasts was separated into cell-soluble and cell-membrane fractions. Both fractions were subjected to protein gel blotting and were probed with anti-GFP antibody (1:2000 dilution, Sigma-Aldrich). Successful fractionation, which was determined by probing the fractions with anti-cFBPase (Agrisera) and anti-H+-ATPase (Agrisera), indicated that OsPSS-1 did not appear in cell-soluble fractions. To further investigate whether OsPSS-1 is an integral membrane protein, cell-membrane fractions were treated with high-salt (1 M NaCl) and alkaline (pH 11) buffers to extract peripheral membrane proteins, as well as detergent (1% [v/v] Triton X-100 and 1% SDS) to extract integral membrane proteins.
For CESA4, 2 g of fresh weight, 1-month-old plants were ground into fine flour using liquid nitrogen and homogenized in extraction buffer (25 mM Tris-HCl [pH 7.5], 2 mM ethylene diaminetetraacetic acid, 0.25 M sucrose, 2 mM dithiothreitol, 10% glycerol, 15 mM b-mercaptoethanol, and proteinase inhibitor cocktail). To separate the proteins into polyethylene glycol and dextran fractions, microsomal proteins obtained from supernatant ultracentrifuged at 100,000 × g for 1 h at 4°C were resuspended in fractionation buffer (5 mM K 2 HPO 4 -KH 2 PO 4 [pH 6.8], 0.25 M sucrose, 1 mM dithiothreitol, 6.2% PEG3350, and 6.2% dextran T500) and centrifuged at 8,000 × g for 10 min at 4°C. Proteins in the polyethylene glycol and dextran fractions were separately collected and concentrated at 100,000 × g for 1 h and dissolved in suspension buffer (2 mM Tris [pH 6.5], 1 mM dithiothreitol, and 0.25 M sucrose). Approximately 9 μg of protein were subjected to gel blotting and detected with anti-OsCESA4 polyclonal antibody. Specificity of the anti-OsCESA4 antibody was previously confirmed [39]. For positive contrast, anti-Anti-H + ATPase and anti-Arf1 antibody was purchased from Agrisera. The immunoblots were quantified by measuring the intensity of the protein bands with ImageJ software.

Secretion assay
Secretion of secGFP was assayed as described previously [72], with some modifications. After 14 h of transient expression, protoplasts and culture medium were collected separately via centrifugation at 200 x g at 25°C for 5 min. Proteins in protoplasts were extracted with 250 mM 2-amino-2-(hydroxymethyl)-1,3-propanediol (Tris)-HCl [pH 7.4], 750 mM NaCl, 5 mM ethylene diaminetetraacetic acid, and protease inhibitor cocktail, and sonicated for 5 s. The supernatant was recovered after centrifugation at 25000 × g at 4°C for 10 min. Proteins from the culture media were concentrated using a 3,000 MWCO Amicon Ultra 0.5 mL centrifugal filter (Millipore) at 1500 × g. Intracellular and medium proteins were then separated via SDS polyacrylamide gel electrophoresis and immunoblotting with anti-GFP (1:2000 dilution, Sigma-Aldrich) or anti-cFBPase (Agrisera, a control for protein loading) as indicated. Three independent experiments were performed to obtain the average targeting efficiency. The immunoblots were quantified by measuring the intensity of the protein bands with IMAGEJ software.