1 Running Title 2 Maize Golgi glycome and proteome 3 4 Glycome and proteome components of Golgi membranes are common between two 5 angiosperms with distinct cell wall structures 6 7 Ikenna O. Okekeogbua,b, Sivakumar Pattathilc,1, Susana M. González Fernández-Niñod, Uma K. 8 Aryale, Bryan W. Penningf, Jeemeng Laod,2, Joshua L. Heazlewoodd,3, Michael G. Hahnc,g, 9 Maureen C. McCannb,e,h, Nicholas C. Carpitaa,b,e,h4 10 11 aDepartment of Botany & Plant Pathology, Purdue University, West Lafayette, Indiana 47907, 12 USA 13 bDepartment of Biological Sciences, Purdue University, West Lafayette, Indiana 47907, USA 14 cComplex Carbohydrate Research Center, University of Georgia, Athens, Georgia 30602, USA 15 dJoint BioEnergy Institute, Lawrence Berkeley National Laboratory, Berkeley, California 94720, 16 USA 17 ePurdue Proteomics Facility, Bindley Biosciences Center, Purdue University, West Lafayette, 18 Indiana 47907, USA 19 fUSDA-ARS, Corn, Soybean and Wheat Quality Research, 1680 Madison Avenue, Wooster, OH 20 44691 21 gDepartment of Plant Biology, University of Georgia, Athens, Georgia 30602, USA 22 hPurdue Center for Plant Biology, Purdue University, West Lafayette, Indiana 47907, USA 23 24 Present addresses: 1 25 Mascoma LLC (Lallemand, Inc.), Lebanon, NH 03766 26 2University of Southern California, USC School of Pharmacy, 1985 Zonal Avenue, Los Angeles, 27 CA 90089 28 3School of BioSciences, The University of Melbourne, Victoria 3010, Australia 29 30 4Author for correspondence 31 Nicholas C. Carpita, Department of Botany & Plant Pathology, Purdue University 32 915 West State Street, West Lafayette, IN 47907-2054 33 Tel. (765)494-4653; FAX: (765)496-0363; Email: [email protected] 34 35 ORCID IDs: 0000-0003-2676-153X (I.O.O.); 0000-0003-3870-4137 36 (S.P.); 0000-0002-2135-4428 (S.M.G.F.); 0000-0003-4543-1536 (U.K.A.); 37 Joshua Heazlewood 0000-0002-2080-3826 (J.L.H.); 0000-0003-2136-5191 38 (M.G.H.);0000-0001-6956-4216 (M.C.M.); 0000-0003-0770-314X (N.C.C.). 39

1

40 IN A NUTSHELL 41 42 Background: Grasses and other commelinid species possess a distinct type of primary cell wall 43 from all other angiosperms based on the structures and relative proportions of their pectic and 44 hemicellulosic polysaccharides, and the presence or absence of a phenylpropanoid network. The 45 plant ER-Golgi apparatus is the site of synthesis, packaging and export of all non-cellulosic 46 polysaccharides and proteins of the cell wall. 47 48 Questions: What polysaccharide structures are present in the Golgi compared to those in the cell 49 wall, and how do they differ between two species representative of the two distinct types of cell 50 wall? What is the maize Golgi proteome that supports synthesis of cell wall polysaccharides 51 unique to grasses? 52 53 Findings: We established that a large proportion of carbohydrate consisting of mostly the glycan 54 moieties of arabinogalactan-proteins are common to the Golgi of maize (Zea mays) and 55 Arabidopsis, Pectic and hemicellulosic polysaccharides that accumulated in Arabidopsis and 56 maize Golgi do not reflect the relative proportions that accumulate in their respective cell walls. 57 Flotation centrifugation followed by free-flow electrophoresis provided the richest collection to 58 date of maize Golgi membrane proteins associated with the synthesis and metabolism of the cell 59 wall. However, this full complement of Golgi cell-wall polysaccharides, and the synthases and 60 glycosyl transferases that make them, represent compositions distinct from those of their cell 61 wall types. 62 63 Next steps: The glycome composition of the Golgi represents a snapshot of what is made and 64 accumulated at the time of harvest. A flux analysis is needed to determine relative rates of 65 synthesis and export of Golgi polysaccharides to account for the anomalous accumulation of cell 66 wall polysaccharides not found in abundance in the cell walls of either species. The fate of these 67 polysaccharides, if not exported and integrated into cell wall architectures, remains to be 68 determined. 69 70 71

2

72 The plant endoplasmic reticulum (ER)-Golgi apparatus is the site of synthesis, assembly, 73 and trafficking of all non-cellulosic polysaccharides, proteoglycans, and proteins destined 74 for the cell wall. As grass species make cell walls distinct from dicots and non-commelinid 75 monocots, it has been assumed that the differences in cell wall composition stem from 76 differences in biosynthetic capacities of their respective Golgi. However, immunosorbent- 77 based screens and carbohydrate linkage analysis of polysaccharides in Golgi membranes, 78 enriched by flotation centrifugation from etiolated coleoptiles of maize (Zea mays) and 79 leaves of Arabidopsis (Arabidopsis thaliana), showed that arabinogalactan-proteins and 80 arabinans represent substantial portions of the Golgi-resident polysaccharides not typically 81 found in high abundance in cell walls of either species. Further, hemicelluloses 82 accumulated in Golgi at levels that contrasted those found in their respective cell walls, 83 with xyloglucans enriched in maize Golgi, and xylans enriched in Arabidopsis. Consistent 84 with this finding, maize Golgi membranes isolated by flotation centrifugation and enriched 85 further by free-flow electrophoresis (FFE), yielded over 200 proteins known to function in 86 the biosynthesis and metabolism of cell wall polysaccharides common to all angiosperms, 87 and not just those specific to wall type. We propose that the distinctive compositions of 88 grass primary cell walls compared to other angiosperms result from differential gating or 89 metabolism of secreted polysaccharides post-Golgi by an as yet unknown mechanism, and 90 not necessarily by differential expression of specific synthase complexes. 91 92 Keywords 93 Plant Golgi, endoplasmic reticulum, cell wall polysaccharides, proteome, glycome, Zea mays 94 (maize), Arabidopsis 95 96

3

97 INTRODUCTION 98 The primary walls of plants are assembled from several kinds of polysaccharides, structural 99 proteins, and phenylpropanoids. Cellulose microfibrils can hydrogen bond with four kinds of 100 hemicellulosic polysaccharides, the xyloglucans (XyGs), (glucuronoarabino)xylans (GAX), 101 (gluco)mannans, and (1→3),(1→4)-β-D-glucans (mixed-linkage β-glucans) (Scheller and 102 Ulvskov, 2010). Angiosperm xyloglucans are (1→4)-β-D-glucan backbones branched at the O-6 103 by Xyl residues, that in some species can be branched further by Gal and Fuc residues. 104 (Gluco)mannans are also (1→4)-β-D-linked backbones with interspersed stretches glucan and 105 mannan, with occasional branching of the mannan moieties at the O-6 position with Gal residues. 106 The GAX polysaccharides are (1→4)-β-D-xylan backbones bearing side-groups of GlcA and Ara 107 residues. The microfibrillar scaffold of cellulose and hemicelluloses is embedded in a matrix of 108 two kinds of pectic polysaccharide backbones of (1→4)-α-D-homogalacturonan (HG) and 109 repeating units of the O-2-α-D-Rha-(1→4)-α-D-Gal disaccharide in rhamnogalacturonan-I (RG- 110 I) (Caffall et al., 2009). The HGs can be branched with Xyl residues to form Xyl-HGs, or with 111 four complex oligosaccharides to form RG-II, a polysaccharide that forms boron di-diester 112 crosslinks into dimers. The Rha residues of RG-I can be substituted with branched (1→5)-α-L- 113 arabinans and type I (1→4)-α-D-galactans with appendant Ara residues. Hydroxyproline- and 114 glycine-rich structural proteins interact with the matrix polysaccharides in plant cell walls, but 115 GPI-anchored peptidoglycans, such as type II arabinogalactan-proteins (AGPs), with highly 116 branched (1→3), (1→6)-, (1→3, 1→6)-β-D-galactan chains and its Ara side-groups (Showalter, 117 1993). 118 Primary cell walls of angiosperm species are classified into two main types of distinctive 119 composition (Carpita and Gibeaut, 1993). Xyloglucans (XyGs) are the major hemicellulose in 120 the Type I walls of dicots and non-commelinid monocots, and contain a rich matrix of HG and 121 RG-I pectic polysaccharides (Scheller and Ulvskov, 2010; Caffall et al., 2009). In contrast, Type 122 II walls of grasses and other commelinid monocots have very little pectin, and contain mostly 123 hemicellulosic glucuronoarabinoxylans (GAX) (Carpita and Gibeaut, 1993). A (1→3),(1→4)-β- 124 D-glucan (mixed-linkage β-glucan) unique to the Poales is found at certain stages of primary 125 wall formation; a small amounts of XyG is found but with truncated galactosylation that 126 typically lacks fucose (Carpita et al., 2001). Type I walls incorporate predominantly Hyp- and 127 Gly-rich structural proteins at the end of growth, whereas phenylpropanoids are deposited in the 128 Type II wall (Carpita and Gibeaut, 1993). 129 Because flowering plants have two distinct types of primary cell walls, it was expected that 130 dicots and grasses might possess Golgi with synthase complexes reflecting the specific kinds of 131 wall polymers that accumulate. The pectic and non-cellulosic polysaccharide constituents of 132 plant cell walls are made within endoplasmic reticulum (ER)-Golgi apparatus and exported to the 4

133 cell surface (Nebenführ and Staehelin, 2001). Numerous structural and enzymatic glycoproteins 134 that function at the plasma membrane and in the cell wall also transit through the Golgi. 135 Cellulose and callose synthases are trafficked through the Golgi membrane, as are receptor-like 136 kinases, AGPs, other GPI-anchored proteins, and components of the cytoskeletal-plasma 137 membrane-cell wall signaling matrix (Dunkley et al. 2006; Heazlewood et al., 2007; Nikolovski 138 et al., 2012). Although we have a reasonably complete inventory of the polysaccharides and 139 proteins that are components of cell walls, we have only a modest knowledge of enzymes that 140 synthesize these wall components in the Golgi and how elements of the trafficking machinery 141 chaperone their delivery to the cell surface (Parsons et al., 2012, 2013; Rosquete et al., 2018). To 142 test the hypothesis that cell-wall composition reflects the biosynthetic capacity of the Golgi, we 143 undertook comparative proteome and glycome studies of intact Golgi isolated from etiolated 144 coleoptiles of the commelinid grass model maize (Zea mays) to compare with those from 145 developing leaves of the eudicot model Arabidopsis (Arabidopsis thaliana), as species 146 representative of these two distinct types of primary cell walls. 147 Equilibrium sucrose density-gradient centrifugation has been employed to enrich Golgi 148 membranes for proteomics analyses, with relative protein and marker abundance across fractions 149 being used to infer the Golgi-specific complement within the mixed membrane preparation 150 (Dunkley et al., 2004, 2006; Nikolovski et al., 2012). Because of contamination by other 151 membranes in conventional downward centrifugation, flotation centrifugation in sucrose 152 gradients substantially improves purity of Golgi membranes (Balch et al., 1984; Lending et al., 153 1990). Its rapid enrichment of Golgi membranes preserves protein integrity required for in vitro 154 polysaccharide synthase reactions and studies of synthase topology (Gibeaut and Carpita, 1993; 155 Urbanowicz et al., 2004). We show here that flotation centrifugation provides a convenient and 156 abundant source of ER and Golgi membranes for comparative analysis of their associated 157 proteins and carbohydrates. 158 Contrary to expectations, an enzyme-linked immunosorbent assay-based screen using plant 159 cell wall glycan-directed monoclonal antibodies (mAbs) (Pattathil et al., 2015) and 160 complementary quantitative linkage analysis, showed that Golgi from both Arabidopsis and 161 maize have the capacity to make a broad range of cell-wall polysaccharides and are not 162 specifically enriched in those associated with wall type. Epitopes associated with arabinoxylans 163 were found in Type II maize walls, but the vast majority of the epitopes recognized in Golgi 164 membranes were in non-fucosylated XyG and RG-I/AG classes more common to Type I walls of 165 Arabidopsis, including epitopes scarcely represented in the wall. In contrast, Golgi of 166 Arabidopsis were enriched in xylan epitopes, with few XyG epitopes. Quantitative 167 monosaccharide and linkage analyses confirmed the relative abundance of polysaccharides 168 inferred from the glycome arrays. 5

169 An alternative means of enrichment of Golgi membranes for proteomics analysis is free- 170 flow electrophoresis (FFE), where membranes are resolved from other organelles and 171 microsomal membranes by differences in surface charges in a continuous liquid stream (Parsons 172 et al., 2014). Preliminary enrichment from crude membranes by alternative means improves 173 protein resolution. For example, isolation of plasma membranes by two-phase aqueous 174 partitioning enhances the quality of subsequent separations by FFE for proteome analysis (de 175 Michele et al., 2016). Consistent with this enrichment strategy, FFE of Golgi membranes isolated 176 by flotation centrifugation showed a substantial enhancement of luminal and intrinsic proteins of 177 Golgi membranes and associated ER from more loosely associated extrinsic proteins and protein 178 contaminants. We identified over 200 maize Golgi proteins known to function in the biosynthesis 179 and metabolism of cell wall polysaccharides common to all angiosperms, and not just those 180 specific to the grass cell wall. Because the Golgi membranes could be resolved into several sub- 181 fractions, we showed further that most Golgi-associated proteins were asymmetrically distributed 182 into subdomains dominated by ER-resident proteins, Golgi-resident proteins, and a third domain 183 enriched with CesA proteins destined for trafficking to the PM. These findings prompt a re- 184 evaluation of the relative roles of gene expression, synthesis and, moreover, the trafficking and 185 turnover of cell wall components in transit to and from the plasma membrane, in determining 186 primary wall composition. 187 188 Results 189 190 Preparation of Cell Walls and Golgi membranes 191 In triplicate experiments, three-week-old light-grown Arabidopsis seedlings and 2.5- to 3-d-old 192 etiolated maize coleoptiles were gently mashed in an equal volume of 84% sucrose buffer, and 193 homogenate squeezed through nylon mesh to separate a cell wall mat from cellular membranes 194 and cytoplasmic material (Supplemental Figure 1). A rich source of intact Golgi membranes was 195 then acquired by flotation centrifugation (Figures 1A to 1C). Numerous Golgi stacks in these 196 preparations could be identified using transmission electron microscopy (Figure 1D). The Golgi- 197 enriched membranes were then directed to glycome array, linkage analysis, and FFE for 198 proteomic analysis. Cell walls purified from the screened mats were sequentially extracted with 199 the chelator ammonium oxalate and increasing concentrations of alkali from 0.1 to 4 M to yield 200 pectic and hemicellulosic material for glycome and linkage analysis (Supplemental Figure 1). 201 202 Glycome and linkage analyses confirm distinctive Arabidopsis Type I and maize Type II 203 cell wall compositions 204 6

205 The immunosorbent assay-based glycome analyses utilize libraries of monoclonal antibodies 206 (mAbs) directed against diverse epitopes present in different kinds of pectins, hemicelluloses, 207 and arabinogalactans of flowering plants (Pattathil et al., 2010, 2015; Schmidt et al., 2015; 208 Ruprecht et al., 2017; Dallabernardina et al., 2017). We used here a library of 155 mAbs that 209 recognize a broad range of pectic and hemicellulosic carbohydrate epitopes (Supplemental 210 Figure 2; Pattathil et al., 2015). Cell walls from isolated from 3-week-old light-grown seedlings 211 of Arabidopsis and from 2.5 to 3-day-old etiolated maize coleoptiles (and undeveloped leaves 212 within) were sequentially extracted with ammonium oxalate and increasing concentrations of 213 alkali to resolve their respective pectic and hemicellulosic constituents. The glycome analyses of 214 the Arabidopsis ammonium oxalate and 0.1 M alkali extracts of the cell wall showed epitope 215 abundances consistent with pectic epitopes of HG, RG-I, and associated arabinans and galactans, 216 while stronger alkali solutions extracted mostly epitopes associated with fucosylated and 217 galactosylated XyGs and xylans (Figure 1). Glycome arrays only provide an indication of the 218 relative proportions of each epitope, we used linkage analysis to quantify the mole% of 219 component polysaccharides. For linkage analyses, uronic acid residues were carboxyl-reduced 220 with sodium borodeuteride to their respective neutral sugars in extractions of cell walls by 221 ammonium oxalate and 0.1 M NaOH, so that their relative abundance with neutral sugars could 222 be determined by GC-MS. The proportions of pectic and hemicellulosic polysaccharides are 223 inferred from diagnostic linkages of specific polysaccharide backbones, its branch-point residues 224 and appendant side-groups (Supplemental Table S1). Linkage analysis of equivalent fractions 225 from Arabidopsis cell walls showed that the ammonium oxalate and 0.1 M NaOH extracts were 226 enriched in RG-I and HG, with 4-GalA, and 2- and 2,4-Rha as the most abundant linkages, 227 whereas the strong alkali fractions of Arabidopsis cell walls were enriched in t-Xyl, 2-Xyl, t-Gal, 228 2-Gal, and 4 and 4,6-Glc characteristic of fucosylated XyGs (Figure 2A; Supplemental Table 229 S2). 230 Glycome analyses of chelator and alkali extracts of isolated cell walls from maize 231 coleoptiles showed epitope abundances consistent with their known pectin, GAX, and XyG 232 constituents. Maize ammonium oxalate and dilute alkali fractions contained small amounts of 233 mostly HG and RG-I epitopes, with abundant GAX epitopes present in all fractions, whereas 234 stronger alkali fractions contained epitopes mainly associated with GAX and smaller amounts of 235 non-fucosylated XyGs (Figure 1). Linkage analyses of maize cell wall fractions confirmed the 236 presence of 4-GalA, and 2- and 2,4-Rha, consistent with the presence of pectic HG and RG-I 237 polysaccharides in the ammonium oxalate and 0.1 M NaOH extracts, and also showed higher 238 abundances of t-GlcA, t-Ara, and 4- and 3,4-Xyl linkages typical of highly substituted GAX 239 (Figure 2B; Supplemental Table S3). The strong alkali fractions of maize cell walls were 240 dominated by t-Ara, and 4- and 3,4-Xyl, but also contained t-Xyl, and 4 and 4,6-Glc typical of 7

241 XyGs, as well as 3- and 4-linked Glc associated with the mixed-linkage β-glucans. In summary, 242 the relative proportions of pectins and hemicelluloses determined by both glycome array and 243 linkage analyses were fully consistent with the compositions characteristic of Type I and Type II 244 cell walls. 245 246 Glycome and linkage analyses show contrasts between carbohydrate compositions of 247 Arabidopsis and maize cell walls and their respective Golgi. 248 249 However, the glycome profiles of the Golgi-resident glycans from both species were different in 250 their epitope compositions from their respective total walls (Figure 4A). Absorbance values for 251 each epitope of non-cellulosic polysaccharides extracted by each of the ammonium oxalate and 252 alkali extracts (Figure 1) were scaled by their weight% contribution mass to obtain an estimate of 253 the relative contribution of pectic and hemicellulosic epitopes in the cell wall. Glycome analyses 254 showed that XyG epitopes (both fucosylated and galactosylated forms) and pectic RG-I dominate 255 the Arabidopsis total wall composition, but xylan and HG epitopes were also detected (Figure 256 4A). The abundances of arabinan and galactan epitopes recognized by the RG-I/AG antibody 257 clade inferred by binding intensity were relatively lower compared to those of xylan backbones 258 (as recognized by the Xylan-6 and -7 antibodies), GlcA-substituted xylans (CCRC-M150), 4-O- 259 Me-substituted xylans GlcA- (Xylan-5 antibodies) and Ara-substituted (CCRC-M154) xylans, 260 and galactosylated XyGs. In contrast, Arabidopsis Golgi membranes were dominated by epitopes 261 recognized by the RG-I/AG, AG-2 and AG-4 antibody clades, with relatively less abundant HG, 262 xylan and XyG epitopes (Figure 4A). The glycome profiles of maize Golgi contained 263 approximately equivalent binding of mAbs recognizing galactosylated XyGs, side-chain and 264 backbone epitopes of xylans, and arabinogalactan epitopes recognized by the RG-I/AG, AG-2, 265 and AG-4 antibodies. Apart from epitopes of two AG-2 mAbs (CCRC-M133 and CCRC-M107) 266 detected in the walls of maize, arabinogalactan epitopes recognized by AG-4 and other AG-2 267 antibodies, largely unbranched 6-linked Gal residues (Ruprecht et al., 2017), were strongly 268 represented in Arabidopsis and maize Golgi but were of relatively low abundance in their walls. 269 Arabinogalactan epitopes recognized by the RG-I/AG antibodies were present with 270 approximately equal intensities in both Golgi and cell walls of Arabidopsis but relatively more 271 abundant in Golgi of maize than wall (Figure 4A). Some non-fucosylated XyG (Non-Fuc XyG-5 272 antibodies) and HG backbone epitopes were also detected in the maize Golgi. The intense 273 signals of the arabinogalactan epitopes were noteworthy, as these epitopes, for the most part, 274 were barely detected in maize cell walls. Amounts of GAX were relatively less abundant 275 compared to representation of these pectin epitopes, and non-fucosylated XyGs were over- 276 represented considering their low abundance in the cell wall (Figure 4A). 8

277 The mole% values of linkage groups diagnostic of total non-cellulosic polymers were 278 compiled from the respective amounts in the four extracts, with relative abundance of each pectic 279 and hemicellulosic polysaccharide reflected the composition of Type II walls of maize (Figure 280 4B). However, consistent with the glycome profiles, the monosaccharide and linkage 281 distributions of maize Golgi membranes were in stark contrast to those of the cell wall. Type II 282 maize walls were pectin-poor and rich in GAX; linkages diagnostic of GAX and mixed-linkage 283 β-glucans were also found in Golgi membranes, which also contained particularly high 284 abundance of 3-, 6-, and 3,6-Gal residues, indicating abundant AGPs (Figure 4C; Supplemental 285 Table 1). In contrast to glycome profiles that showed abundant RG-I/AG and some AG classes of 286 epitopes, linkage analysis indicated that RG-Is represent a much smaller proportion of the walls 287 than AGPs. Golgi from both Arabidopsis and maize were enriched in 5-arabinans in proportion 288 to wall, whereas the smaller proportions of type I (arabino)galactan were generally lower in 289 Golgi compared to wall (Figure 4B and 4C). In contrast to their low abundance in the cell walls 290 of maize, 4-Glc and 4,6-Glc residues associated with XyG were high in the maize Golgi 291 membranes (Figure 4B; Supplemental Table 3). Golgi membranes were also enriched in 2-Man, 292 6-Man, and 3,6-Man, linkages not detected in cell wall fractions but characteristic of trimmed N- 293 linked core glycosylation. 294 Arabidopsis Golgi also yielded particularly high abundance of 3-, 6-, and 3,6-Gal residues 295 of AGPs and the 2-Man, 6-Man, and 3,6-Man linkages of N-linked glycoproteins (Figure 4C; 296 Supplemental Table 2). Although linkages associated with fucosylated XyG were detected in the 297 Arabidopsis Golgi (Supplemental Table 2), XyGs were much more abundant in the maize Golgi 298 than in those of Arabidopsis (Figures 4B and 4C; Supplemental Tables 2 and 3). Conversely, 299 linkages associated with xylans were proportionally higher in Arabidopsis Golgi than in their cell 300 wall. 301 302 Enrichment of maize Golgi membranes by free-flow electrophoresis (FFE) for proteomic 303 analysis. 304 305 A diverse set of cell wall-related proteins was identified in Golgi membranes collected after 306 flotation centrifugation (Supplemental Table 4). When maize Golgi membranes isolated by 307 flotation centrifugation were enriched further by FFE, the Golgi membranes formed a 308 heterogeneous peak that migrated about two-thirds of the distance of the fastest-migrating UV- 309 absorbing material, with only small amounts of other membranes detected in neighboring 310 fractions (Figure 5A). Proteins from triplicate samples after flotation alone (pre-FFE) and those 311 after FFE (post-FFE) were trypsin-digested for subsequent LC-MS/MS analysis Peptide data 312 were processed against maize protein sequence database at MaizeGDB 9

313 (https://www.maizegdb.org), using MaxQuant with an integrated Andromeda search engine (Cox 314 and Mann, 2008; Cox et al., 2014) as well as Mascot (Matrix Science) for protein identification 315 and relative quantification. Since our original search, we have augmented public database at 316 MaizeGDB for improved annotation of maize cell wall protein families 317 (https://www.maizegdb.org/gbrowse/maize_v2test?l=CellWallGenes;l=Gene_models;q=Chr1:26 318 50000..2699999).This resource is also posted at (http://cellwall.genomics.purdue.edu). The 319 relative abundances of protein distributions across FFE fractions were determined using spectral 320 counts (or MS/MS counts). Spectral counts are the total number of MS/MS spectra of all the 321 non-redundant peptides mapped to a protein or protein family during the entire LC-MS/MS run, 322 and they are referred as MS/MS counts for MaxQuant results. Total MS/MS counts from both 323 MaxQuant and Mascot were normalized to determine the relative enrichment after FFE (Figure 324 5B). Pairwise comparison of peptide MS/MS counts from MaxQuant among the three 325 independent experiments gave linear regression values of r2 = 0.93 to 0.96 (Supplemental Figure 326 3A). Similar results were observed for MS/MS counts from Mascot search (data not shown). 327 Identification was considered positive if proteins were found in at least two biological replicates. 328 Identification of a single isoform was considered positive only if it is identified by at least one 329 unique peptide. Identification of protein isoforms is complicated in maize, because many genes 330 share the same sequence as a result of ancestral duplication events. In these instances, the level 331 of expression of similar isoforms was considered as reflective of relative abundance. A total of 332 2030 unique proteins were identified at a significance level of p ≤0.05 in at least one of twelve 333 Golgi-rich fractions collected after FFE (Supplemental Table 5) from combined MaxQuant and 334 Mascot analyses, with 1168 proteins detected by MaxQuant and 1781 proteins by Mascot, of 335 which 936 proteins were common to both. Of these FFE-enriched Golgi proteins, about 11.3% 336 (230) were known to function in cell-wall synthesis and metabolism (Figure 5B; Table 1; 337 Supplemental Figure 4). Application of the maize annotation tool at Phytozome 338 (https://phytozome.jgi.doe.gov/pz/portal.html) resulted in numerous mis-annotations compared 339 to our improved manual annotation of maize cell-wall protein database at MaizeGDB 340 (Supplemental Table 5). Quantitation of classes of proteins using MapMan functional categories 341 in the Plant Proteomics Database (http://ppdb.tc.cornell.edu/dbsearch/subproteome.aspx; Sun et 342 al., 2009) showed only 2.5% contamination with plastid and mitochondrial proteins 343 (Supplemental Figure 4). Plasma membrane-associated proteins detected, such as several V-type 344 ATPases and CesAs, traffic through the Golgi, so it is difficult to judge contamination with 345 plasma membrane. SNAREs and clathrin heavy-chain proteins were also detected that could be 346 TGN-related. However, the two maize homologs closest in sequence to an Arabidopsis SYP61 347 TGN marker, were not detected (Supplemental Table 7).

10

348 Numerous proteins associated with substrate generation, polysaccharide backbone 349 synthesis and side-group attachment were identified after FFE, including a more complete 350 collection of the enzymes of nucleotide-sugar interconversion, nucleotide-sugar transport, 351 cellulose and callose synthesis, mixed-linkage β-glucan synthesis, and 25 proteins associated 352 with xylan backbone synthesis and side-group attachment, and 14 proteins associated with AGP 353 synthesis and glycosylation (Table 1, Supplemental Table 8). When total MS/MS counts from 354 MaxQuant analysis were pooled for cell-wall protein classes before and after FFE, relative 355 abundances of glycosyl transferases were generally enriched. In contrast, enzymes of monolignol 356 synthesis, peroxidases, sucrose synthases, and certain enzymes of nucleotide-sugar 357 interconversion, particularly the Reversibly Glycosylated Proteins (RGPs), more aptly named 358 UDP-arabinopyranose mutases (UAMs), migrated with Golgi after flotation centrifugation but 359 were depleted in Golgi enriched by FFE (Figures 5B and 5C, Supplemental Table 8). 360 361 Comparison of gene expression with maize cell wall protein abundance in FFE-enriched 362 fractions. 363 364 Transcripts for 227 of the 230 cell wall-related proteins identified in the Golgi membranes were 365 detected by RNAseq analysis (Figure 6). Certain proteins involved in cellulose synthesis (CesA2, 366 KOR1c, and KOR1d), isoforms of two nucleotide sugar interconversion enzymes (SUD3a, 367 SUD3b, UGD3b, and RGP1b), XyG modification (XTH5), mixed-linkage glucan synthesis 368 (CslF2 and CslF4), and lignin synthesis (PAL2, DCR, and APX5d), had substantially high 369 transcript abundance, but only modest protein abundance. However, two proteins involved in the 370 synthesis of GAX (GT61-8 and GUT1g) and one membrane-associated nucleotide sugar 371 dehydrogenase enzyme, AUD1c, had high protein abundance with modest to low transcript 372 abundance (Figure 6). Although correlation of transcript level and protein abundance is low 373 across the entire protein set, correlation was much stronger among isoforms encoded by the same 374 family (Supplemental Table 7). For this reason, we used transcript abundance to designate the 375 probable isoform abundance for large families of proteins for which protein assignments were 376 ambiguous. 377 378 Asymmetric distribution of proteins across FFE fractions. 379 380 The abundances of proteins confirmed by MaxQuant were strongly correlated with quantitation 381 of spectral counts from Mascot in Golgi membranes separated by FFE over twelve fractions 382 (Supplemental Figure 3B). Once protein identity was established by Mascot and MaxQuant in 383 the Golgi post-FFE, and relative abundance of all cell wall proteins was well correlated between 11

384 these two methods (Supplemental Figure 3B), the Mascot spectral counts pooled from all three 385 experiments were used to determine the distribution of each protein across four fractions of 386 pooled Golgi membranes relative to the fraction with the highest spectral counts (Figure 7). Total 387 protein abundance based on absorbance at 280 nm during FFE was maximal in the major peak of 388 Golgi membranes in fractions centered at F34, with a smaller spike of proteins centered at F31, 389 and a small shoulder at F28 (Figure 7A). Numerous established Golgi-localized proteins were 390 enriched in the fractions of the bulk Golgi centered between F34 and F37. These include markers 391 for cis-Golgi, such as a DUF616-containing EMBRYO DEFECTIVE2756, a DUF1195- 392 containing putative sugar transporter, and SYP32; markers for medial-Golgi, such as 393 mannosidase II, and also YIPF5 that has been observed in both the ER and Golgi (Figure 7B; 394 Supplemental Table 7). Some Golgi-associated proteins had significant representation by 395 proteins in the putative ER-rich F28, such as a DUF2359-containing transmembrane protein (or 396 TMEM214), and transmembrane protein 115 (TMEM115) which co-localizes with the medial- 397 Golgi marker, mannosidase II (Supplemental Table 7). ECA4, another Golgi-localized protein 398 deviated from the bulk Golgi pattern, being enriched in F31 fraction. The general overlap of 399 these marker proteins across the three Golgi fractions indicated a lack of partitioning of cis-, 400 medial-, and trans-cisternae. Fraction F28 was ER-enriched based on the distribution of the 401 markers, signal peptidase complex subunit3 (SPCS3), sterol methyltransferase 2 (SMT2) and S- 402 adenosyl-L-homocysteine hydrolase (Sah1), although some of these markers migrate to the Golgi 403 fractions to varying extents (Figure 7B). This fraction also contained several 8S, 40S, and 60S 404 ribosomal proteins. 405 A majority of the enzymes of nucleotide-sugar interconversion were more or less evenly 406 distributed across all fractions, including RGPs, which are extrinsic Golgi proteins not enriched 407 by FFE (Figure 7C). When the abundance of all proteins associated with synthesis of five non- 408 cellulosic pectic and hemicellulosic polysaccharides were summed and averaged across each of 409 the four fractions (Supplemental Table 7), these values generally aligned with the major Golgi 410 fractions (Figures 7A and 7D). In contrast, a majority of CesA proteins were enriched in F31, a 411 fraction associated with the small spike in absorbance (Figures 7A and 7E). The only other 412 polysaccharide synthase to enrich in this fraction was CslF5 (Figure 7F). Other proteins 413 associated with cellulose synthase were not enriched in F31, as STELLO was highly represented 414 in all fractions, and KORRIGAN and COBRA were enriched in the bulk Golgi (Figure 7F). 415 416 Discussion 417 418 The cell-wall and Golgi glycomes of maize and Arabidopsis 419 12

420 The characteristic differences between Type I and Type II cell walls (Carpita and Gibeaut, 1993) 421 were reflected in the immunosorbent glycome arrays and linkage analyses of chelator- and 422 alkali-soluble fractions from isolated cell walls of young Arabidopsis leaves and etiolated maize 423 coleoptiles (Figures 1 and 2). As the Golgi apparatus is the site of synthesis and trafficking of 424 non-cellulosic components of the primary cell wall, it was expected that the polysaccharide 425 contents of the Golgi membranes might closely reflect each type of cell wall. While a full range 426 of xylan epitopes (both backbone and side chain epitopes) were present in the maize Golgi, 427 linkage analyses showed that GAX hemicelluloses characteristic of Type II walls of grasses were 428 less abundant in the maize Golgi membranes compared to their walls. In contrast, non- 429 fucosylated XyGs of low abundance in walls were highly represented in the Golgi (Figure 4A 430 and 4B). The converse is true in Arabidopsis, where epitopes of both fucosylated and non- 431 fucosylated XyGs were of low abundance in Golgi compared to wall, and linkage analyses 432 showed xylans to be the dominant hemicellulose of the Golgi but much less represented in the 433 wall (Figure 4A and 4C). 434 There are at least three possible explanations for these anomalous results. First, cell wall 435 polysaccharides that accumulate to higher abundances in Golgi might result from lower 436 trafficking rates compared to higher rates for material that accumulates in the wall. Second, 437 lower trafficking rates might reflect diversion of a subset of polysaccharides to lytic 438 compartments instead of the cell wall. The Trans-Golgi Network-Early Endosome (TGN/EE) is 439 often distinct from the trans-membranes of the Golgi stack and plays a central role, post-Golgi 440 synthesis, in the sorting and packaging of polysaccharide and protein cargoes destined for the 441 cell wall (Kang et al., 2011; Rosquete et al., 2018). Designated as the ‘SYP61 compartment’ in 442 its role in exocytosis (Drakakaki et al., 2012), other proteins, such as ECHIDNA (Gendre et al. 443 2011), and ECHIDNA/YIP4a and YIP4b complex (Gendre et al., 2013), RabA4b and a 444 phosphotidyl-4 kinase (PI-4Kβ1) (Kang et al., 2011) are all implicated in trafficking through the 445 TGN. A SCAMP2 protein is a marker for clusters of secretory vesicles that emanate from the 446 TGN and fuse with the PM (Toyooka et al., 2009). Maize homologs of these TGN-associated 447 proteins were not among those detected in the Golgi enriched by FFE, although we identified 448 numerous Ras- and Rab-like proteins, syntaxins, and SNARE proteins involved in vesicle 449 trafficking and membrane fusion (Supplemental Table 6). Whether polysaccharide sorting occurs 450 at the TGN to wall or lytic compartments remains to be explored. 451 A third possibility is that all materials are trafficked to the cell wall, but specific 452 polysaccharides are digested extracellularly and, therefore, accumulate to much lower levels in 453 the wall. A low abundance of AGPs in the cell wall is expected based on its GPI-anchored 454 location at the exterior of the PM (Tan et al., 2012). Only miniscule amounts of Ara-containing 455 polysaccharides in soluble fractions centrifuged from vacuum-infiltrated pea epicotyl sections 13

456 (Terry and Bonner, 1980), and treatment of these sections with auxin enhanced only the yield of 457 XyG oligomers (Terry et al., 1981). In contrast, isolated enzymatically active cell walls yield 458 primarily Ara and Gal by ‘autolysis’ through action of nascent β-galactosides and α- 459 arabinofuranosidases active against arabinogalactans (Labrador and Nicolás, 1984). Therefore, 460 the low abundance of AGPs in the cell wall is more likely a result of their rapid turnover rather 461 than their loss during homogenization of the cell wall material. 462 Regardless of the mechanism by which the distribution of non-cellulosic polysaccharides in 463 the Golgi membranes is established, the larger question is why a species would place a large 464 metabolic investment in polymers that fail to accumulate in the wall. What selective advantage 465 might this seemingly inefficient and futile synthesis provide? One possible explanation is that 466 maintenance of the capacity to synthesize a full range of polysaccharides serves as a hedge 467 against mutations in synthases of cell wall-abundant polysaccharides that could jeopardize 468 viability. Several examples of cell-wall plasticity to maintain near normal growth and 469 development have been recognized. The mur1mutation, which results in complete loss of fucose 470 in the plant (Reiter et al. 1993), the mur4 and mur5 mutants, which greatly reduce the de novo 471 synthesis of Arap and Araf, respectively (Burget et al., 2003; Dugard et al., 2017), and 472 rhm1(rol1) mutant, which demonstrates limited synthesis of rhamnose and, consequently, of RG- 473 I (Reiter and Vanzin, 2001; Diet et al., 2006), have little impact on normal development apart 474 from a more diminutive stature. From linkage analyses, mur1, mur4, and mur5 demonstrate 475 enhanced HG and/or xyloHG levels, and rhm1(rol1) shows enhanced arabinan and galactan 476 branches on the residual pectin (Mertz et al., 2012; Saffer et al., 2017), modifications that might 477 compensate for the lack of the polysaccharides containing these sugars. The low cell-wall 478 rhamnose mur8 mutant has diminished levels of HG and RG-I but enhanced synthesis of 479 arabinans, xylans, and XyG (Mertz et al., 2012). Cell-wall plasticity in response to mutation is 480 also evident within the fine structure of a polysaccharide. For example, failure to adequately add 481 at least one Gal residue to XyG units results in significant loss of hypocotyl tensile strength 482 (Ryden et al., 2003; Peña et al., 2004); loss of galactosylation of the Xyl residue closest to the 483 reducing end of the unit tetramer in mur3 results in enhanced galactosylation at the adjacent Xyl, 484 rescuing a wild-type phenotype in this mutant (Madson et al., 2003). However, null alleles of 485 mur3 combined with mutations in a second xyloglucan galactosyl transferase xtl2 result in severe 486 growth abnormalities traced to a dysfunctional xyloglucan (Kong et al. 2015; Xu et al., 2017). 487 The Arabidopsis xxt1/xxt2 double mutant completely eliminates XyG from the wall, but despite a 488 slight lowering of tensile strength, plant structure is remarkably unchanged from wild type 489 (Cavalier et al., 2008; Zabotina et al., 2012). Enhanced synthesis of xylan, and possibly HG, 490 provide a possible mechanism of recovery of viable plants. Perhaps the starkest example of 491 plasticity is the survival of cells in liquid culture in the near absence of cellulose induced by a 14

492 potent cellulose synthesis inhibitor (Shedletzky et al., 1990). Tomato and tobacco cells with 493 Type I walls respond by enhancement of a cross-linking of a pectin matrix, whereas barley cells 494 with Type II walls enhance xylan deposition and cross-linking by their phenylpropanoid network 495 (Shedletzky et al., 1992). 496 With the lone exception of AG-2 antibodies CCRC-M133 and CCRC-M107, which have 497 strong signals for epitopes in the maize cell wall, epitopes of other AG-2, RG-I/AG, and AG-4 498 types were of low abundance in the maize cell wall and more more abundant in Golgi compared 499 to wall in both species (Figure 4). In a comprehensive survey of synthetic epitopes of this class, 500 Ruprecht et al. (2017) found that the majority of the mAbs in the RG-I/AG class bind to 3- and 501 6-linked galactans. Coupled with the fact that none of the mAbs that bind to RG-I backbones 502 showed binding in the Golgi, these galactan epitopes were more likely to be from AGPs. This 503 inference was confirmed by linkage analysis, showing that maize Golgi had very little RG-I and 504 HG compared to AGPs and 5-arabinans (Figure 4B). This contrast indicates that these epitopes 505 are associated with Golgi-resident proteoglycans or that their synthesis, flux and turnover rates 506 are substantially higher than for exported pectins and hemicelluloses that accumulate in the wall. 507 The presence of AGPs as the major proteoglycans of the Golgi membranes of both species might 508 indicate a common role in facilitating synthesis and trafficking of wall matrix polymers. In 509 Arabidopsis, the AGP family member APAP1 contains glycan structures linked to xylan and 510 RG-I residues and could therefore represent a structural component in the cell wall (Tan et al., 511 2013). However, an alternative possibility is that core structures similar to APAP1, or other 512 fasciclin-like AGPs, synthesized in the Golgi could serve as initiators of pectin and 513 hemicellulose synthesis or ‘glycochaperones’ during trafficking to the cell wall and turnover 514 upon delivery of their cargo. 515 516 The maize Golgi proteome 517 518 The separation by FFE of membranes previously enriched by flotation centrifugation enabled 519 detection of over two thousand proteins associated with at least one of twelve collected fractions. 520 Of these, 230 proteins were known to be associated with synthesis or metabolism of the cell wall, 521 more than doubling the number reported previously using conventional downward centrifugation 522 of Arabidopsis or Italian ryegrass (Lolium multiflorum) membranes (Nikolovski et al., 2014; 523 Ford et al., 2016). 524 Owing to a genome duplication event about 11 million years ago (Gaut and Doebley, 525 1997), maize gene families are larger, with more gene isoforms expressed at any particular stage 526 of development (Penning et al., 2009). For example, maize CesA genes number twenty, double 527 the number in Arabidopsis and rice. Sixteen CesAs are expressed in the maize coleoptile; of 15

528 these, twelve isoforms were detected by either MaxQuant or Mascot (Supplemental Table 7). 529 The expansion of the number of maize CesAs extends well beyond the classical three isoforms 530 described for Arabidopsis primary and secondary wall synthesis (Taylor, 2008), suggesting that 531 multiple combinations of isoforms might contribute to synthase complexes. 532 Consistent with the detection of extraordinarily high amounts of pectin/AGPs in Golgi, 533 proteins associated with their synthesis were numerous. Eleven of twelve expressed isoforms of 534 family GT8D UDP-GalA transferases (GAUTs) associated with homogalacturonan synthesis 535 were identified, as well as three fasciclin-like AGPs, eight prolyl 4-hydroxylases that synthesize 536 Hyp residues for O-glycosylation, and eight family GT31 UDP-Gal transferases associated with 537 AGP galactan synthesis (Supplemental Table 7). Three family GT14 β-D-glucuronosyl 538 transferases known to decorate the AG chains were also observed. As GAX is the major cell-wall 539 polysaccharide of grasses, a rich assembly of proteins involved in backbone synthases and 540 glycosyl transferases for GAX were detected, including eight IRX10-L family GT47E xylosyl 541 transferases (GUT1 isoforms) involved in chain synthesis, seven members of family GT43 542 (IRX9 and IRX14 proteins) involved in backbone elongation, and eight GT61 glycosyl 543 transferases implicated in xylosylation and arabinosylation of the xylan backbone. 544 The appearance of non-fucosylated XyGs in high abundance in the maize Golgi was 545 surprising considering its low abundance in the cell wall (Figures 4 A and 4B). The capacity for 546 XyG synthesis was validated by detection of several isoforms of CslC XyG (1→4)-β-D-glucan 547 backbone synthases and XyG-specific (1→6)-α-D-xylosyl transferases (XXTs). Grass XyGs are 548 generally less frequently branched with Xyl and Gal residues (Scheller and Ulvskov, 2010), but 549 despite the reduced XyG galactosylation in grasses, three GT47A XyG α-galactosyl transferases, 550 including two MUR3 homologs (Madson et al., 2003), were detected (Supplemental Table 7). A 551 FUT1 homolog of MUR2 (Vanzin et al., 2002) was also detected, indicating that maize Golgi 552 have the capacity to make fucosylated XyGs common among most angiosperms. Despite lack of 553 detectable epitopes of fucosylated XyGs in Golgi, several epitopes associated with fucosylated 554 XyGs were weakly detected, and one strongly detected, in the maize cell wall. Although 555 uncommon in grasses, fucosylated XyGs have been detected in certain tissues (Brennan and 556 Harris, 2011; Liu et al., 2015). In a complementary study of Arabidopsis, cell wall-related 557 proteins in Golgi isolated by FFE, synthases and glycosyl transferases were detected for the full 558 range of cell wall polysaccharides (Parsons et al., 2012). The expected CSLCs, XXTs, MUR and 559 MUR3 enzymes for fucosylated XyGs, the numerous GAUT enzymes for HG synthesis, putative 560 rhamnosyl transferases for RG-I synthesis, and the ARAD1 for 5-arabinan synthesis were 561 detected, as were IRX10, IRX14, GUX3, and XYLT for xylan backbone synthesis and xylosyl 562 side-groups (Supplemental Table 8). 563 16

564 Asymmetric protein distribution in maize Golgi membranes after FFE 565 566 While the majority of the Golgi-associated enzymes of nucleotide-sugar interconversion are 567 more or less uniformly distributed across the four Golgi fractions (Figure 7C), the vast majority 568 of proteins detected after FFE were asymmetrically distributed into three major fractions of bulk 569 Golgi, a fraction rich in CesAs, and a leading shoulder of ER. Golgi bodies comprise several 570 development stages, from fusion of COPII vesicles to form the cis-membranes to their transition 571 through medial- and trans-membranes to the trans-Golgi network (TGN) (Donohoe et al., 2013; 572 Uemura et al., 2014). Location of synthase complexes has been inferred by 573 immunocytochemistry for their products, where XyGs localize to trans-Golgi cisternae and the 574 TGN, pectic polysaccharides to the cis-, medial-, and trans-Golgi cisternae (Zhang and 575 Staehelin, 1992), and xyloglucuronan to early trans-membranes vs. XyG and HG/RG-I to late 576 trans-membranes and the TGN (Wang et al., 2017). The observation that synthases and glycosyl 577 transferases of pectin, AGP, GAX, and XyG synthesis follow the bulk Golgi pattern indicates 578 that these Golgi maturation stages are not differentiated. 579 The Golgi membrane has been shown to be the site of assembly of cellulose synthase 580 complexes by freeze-fracture electron microscopy (Haigler and Brown, 1986) and live-cell 581 imaging of fluorescence-labeled CesA proteins (Crowell et al., 2009; Gutierrez et al., 2009). All 582 but one of the CesA proteins identified in Golgi formed a unique pattern that peaked at F31, 583 coincident with a small spike of abundance in the Golgi peak identified during FFE (Figure 7A 584 and 6E). The co-localization of the CesA proteins in this fraction indicated a possible 585 concentration of the SmaCC or MASC small vesicles that are transported to the PM (Crowell et 586 al., 2009; Gutierrez et al., 2009). However, KORRIGAN and COBRA proteins, known to be 587 associated with CesAs (Liu et al., 2013; Vain et al., 2014), followed the bulk Golgi pattern, and a 588 DUF288-containing STELLO protein that interacts with CesAs in the Golgi (Zhang et al., 2016) 589 was abundant in all four fractions (Figure 7F). 590 The CslF proteins form at least part of the synthase complex for the mixed-linkage β- 591 glucan (Doblin et al., 2009). Although substantial data demonstrate the Golgi-associated 592 synthesis of the mixed-linkage β-glucan in vitro and in vivo (Gibeaut and Carpita, 1993; 593 Buckeridge et al., 1999; Urbanowicz et al., 2004; Carpita and McCann, 2010; Kim et al., 2015), 594 others have suggested that at least some CslF synthases of this polysaccharide traffic to the 595 plasma membrane, where synthesis continues or even commences in muro (Wilson et al., 2015). 596 Indeed, a CslF5 showed enrichment in the CesA pattern (Figures 7E and 7F), and not with the 597 general pattern of pectin and hemicellulose synthases (Figure 7D). Thus, an association of CslF 598 proteins with CesA transport to the plasma membrane is possible. 599 17

600 Conclusions 601 602 A combination of glycome and linkage analyses revealed that the Golgi apparatus of two 603 angiosperm species with distinctively different cell walls possess a common carbohydrate matrix 604 of AGPS and other pectins. Further, epitopes and linkages of the full range of polysaccharides 605 made by both species were enriched to unexpected degrees. Isolation of maize Golgi membranes 606 by flotation centrifugation, followed by separation by FFE, provided a rich inventory of resident 607 and transiting proteins of the Golgi apparatus, including synthases and glycosyl transferases for a 608 broader range of polysaccharides than expected based on wall composition. The capacity of grass 609 species to make a full repertory of polysaccharides opens up new strategies for altering and 610 engineering cell-wall composition of grasses and other bioenergy crop species. The identification 611 of a CesA-rich fraction distinct from the bulk Golgi membranes involved in synthesis of non- 612 cellulosic polysaccharides offers opportunities for further purification of the synthase complex. 613 614 Methods 615 616 Plant Materials 617 618 Seeds from Zea mays (L.) maize were soaked in the dark for about 12 h at ambient temperature 619 in deionized water bubbled with air, then sown in moistened medium-grade vermiculite and 620 grown in darkness for 2.5 to 3 d. Seeds of Arabidopsis thaliana (L.) Heynh. Columbia-0 (Col-0) 621 were soaked in deionized water for 3 d at 4°C to prime germination and sown in 4” pots 622 containing Metromix360 (Sungro, Agawam, MA). They were then grown in growth chambers in 623 a 16-h-light (75 µmol photons m–2 s–1)/8-h-dark photoperiod at 23°C. The seedlings were grown 624 for 20-22 d before harvest. 625 626 Isolation of Crude Cell Walls and Golgi Membrane Fractions 627 628 In triplicate independent experiments, fresh Arabidopsis leaves harvested at 21±1 days and 629 maize coleoptiles (and etiolated leaves within) at 2.5 to 3 days were harvested into a chilled 630 mortar, and overlaid with an equal volume of ice-cold homogenization buffer consisting of 100 631 mM HEPES [1,3-bis[tris(hydroxymethyl)-methylamino]propane (BTP)] (pH 7.4), 20 mm KCl, 632 40 mM ascorbic acid, and 84% (w/v) sucrose (Figure 3, A and B). The isolated coleoptiles were 633 stirred gently in homogenization buffer for about 5 min before mashing in a chilled mortar and 634 pestle (Gibeaut and Carpita, 1990, 1994). The homogenates were squeezed through a nylon mesh 635 (45-µm2 pores) and adjusted to about 42% (w/v) sucrose. Cell walls retained by the nylon mesh 18

636 were directed to isolation as described below. To isolate Golgi membranes, 20 mL of the 637 homogenate was pipetted into each 38.5-mL centrifuge tube (Ultraclear, Beckman) and overlaid 638 with 7 mL each of 35%, and 29% (w/v) sucrose, and 4 mL of 18% (w/v) sucrose in a gradient 639 buffer containing 20 mM HEPES [Bis Tris-propane], pH 7.6, and the remaining volume was 640 made up with 9.5% (w/v) sucrose in the same buffer. After flotation centrifugation at 27,000 rpm

641 (131,500 x g at rmax) in an SW28 rotor (Beckman) for 90 min at 4°C, the interface containing 642 Golgi membranes (29%/35%) was collected with a wide-bore plastic Pasteur pipette, and used 643 directly in FFE as described below. Other fractions of the Golgi membranes and their proteins 644 were precipitated using five volumes (v/v) of cold (–20°C) 100% acetone, and held at –20°C 645 overnight. After five washes with excess chilled 80% acetone (–20°C), the sucrose-free pellet 646 was air-dried and suspended in 100 µL of distilled water for glycome array and linkage analyses 647 (described below), and for proteome analysis as ‘pre-FFE’ proteins. A flow chart for materials is 648 provided in Supplemental Figure 1. 649 650 Preparation of Cell Wall Materials 651 652 Cell wall materials from the three independent harvests of Arabidopsis leaves and maize 653 coleoptiles after membrane isolation were washed several times in water and then ground in 1% 654 (w/v) sodium dodecyl sulfate (SDS) in 50 mM Tris[HCl], pH 7.2. Samples were then incubated 655 to 60°C and washed sequentially in 1% (w/v) SDS-containing grinding buffer, methanol at 50°C 656 for 20 min, and then 50% (v/v) ethanol. Supernatants were discarded to remove soluble 657 compounds and the pellet was finally washed with nanopure water. Starch was extracted from 658 the pellet by sonication in 90% (v/v) DMSO in water (Carpita and Kanabus, 1987). Cell walls 659 were then washed twice with distilled water and suspended in distilled water at 4oC overnight. 660 661 Fractionation of Cell Wall Polysaccharides 662 663 Three independent sequential extractions of isolated cell walls materials were carried out with 664 0.5% (w/v) ammonium oxalate (pH 7.0) at 90°C, then 0.1 M, 1 M and 4 M NaOH (each 665 containing 3 mg/mL sodium borohydride) at ambient temperature with continuous stirring. Each 666 extraction was done twice, with 20 mL and 10 mL of the appropriate reagent for 1 h and 30 min, 667 respectively, and then common extracts were pooled. The second extraction with 4 M NaOH was 668 done overnight before pooling. After each extraction period, the samples were centrifuged at

669 4,500 rpm (3900 x gmax), and the next reagent added to the pelleted wall material. Pooled 670 supernatants were filtered through GF/F glass fiber filter discs, chilled and neutralized with 671 acetic acid, dialyzed extensively against water, and freeze-dried. Total sugar in each fraction was 19

672 estimated by phenol-sulfuric assay (Dubois et al., 1956) using standards for the most abundant 673 neutral sugars, glucose, arabinose, xylose, and galactose, and for galacturonic acid. Alditol 674 acetate derivatives of carboxyl-reduced materials (described below) were used to determine the 675 monosaccharide mole% distributions, and therefore total mg of carbohydrate for loading in 676 glycome profiling. Total monosaccharide for each fraction of non-cellulosic material was used to 677 estimate the relative proportions in the wall based on relative recovery in each fraction. 678 679 Glycome Profiles of Golgi and Cell Wall Polysaccharides 680 681 Three independent preparations of acetone-washed Golgi pellets and freeze-dried cell wall 682 fractions were suspended in nanopure water and sonicated for 30 min in a water bath, and then

683 centrifuged at 4500 rpm (3900 x gmax) to sediment a small amount of undissolved material. The 684 supernatant was brought to about 400 µg mL–1 total carbohydrate (as described above). The 685 small amounts of Golgi material prevented estimation of sugar amounts, and intensities were 686 normalized to the total intensities of the combined cell wall fractions. The glycomics assay of the 687 cell wall and Golgi fractions followed the Enzyme-Linked Immunosorbent Assay (ELISA)-based 688 procedure as previously described by Pattathil et al. (2012). Briefly, the cell wall extracts are 689 probed against a comprehensive suite of 155 plant cell wall glycan-directed monoclonal 690 antibodies (Pattathil et al., 2010), as updated (Pattathil et al., 2015), monitoring most major non- 691 cellulosic plant glycan epitopes using ELISAs. ELISA plates (Costar 3700, Corning Inc., 692 Corning, NY, USA) were coated on an equal carbohydrate basis (0.3 µg/well), and assays were 693 done in two technical replicates using a Robotic system (Thermo Fisher Scientific Inc. Waltham, 694 MA, USA). Water was used as the blank to assess background values, which were subtracted

695 from each sample reading. The mean background-subtracted absorbance values (OD450-655nm) 696 were then represented as heat maps using a modified version of R-console software (Pattathil et 697 al., 2012). The collection of mAbs employed were obtained from laboratory stocks (CCRC, JIM 698 and MAC series) at the Complex Carbohydrate Research Center (available through CarboSource 699 Services; http://www.carbosource.net) or from BioSupplies (Australia) (BG1, LAMP). Detailed 700 information about the antibodies used can be found in the WallMabDB database online 701 (www.wallmabdb.net). Absorbance values for each epitope of the non-cellulosic 702 polysaccharides extracted by each of the ammonium oxalate and alkali extracts were scaled by 703 their weight % contribution mass to obtain an estimate of the relative distribution of total pectic 704 and hemicellulosic epitopes in the total cell wall. 705 706 Monosaccharide and Linkage Analysis 707 20

708 Uronic acids in a 3-mL suspension of the Golgi and the ammonium oxalate- and 0.1 M NaOH 709 cell-wall fractions were reduced by activation with 0.25g 1-cyclohexyl-3-(2-morpholinyl-4- 710 ethyl) carbodiimide (methyl-p-toluene sulfonate) [CMC], followed by addition of 300 mg 711 sodium borodeuteride, and then supplemented with 0.5 mL of 2 M imidazole [HCl], pH 7, to 712 prevent alkali degradation (Kim and Carpita, 1992, as modified by Carpita and McCann, 1997). 713 After 2 h incubation, the solution was acidified with glacial acetic acid to destroy excess 714 borohydride and then dialyzed extensively against nanopure water and lyophilized. 715 Monosaccharide composition was determined essentially as described by Gibeaut and 716 Carpita (1991). Approximately 1 to 2 mg of lyophilized Golgi and cell wall samples were 717 hydrolyzed with 1 mL of 2 M trifluoroacetic acid (TFA) (containing 1µmol of myo-inositol as 718 internal standard) for 90 min at 120°C with vortex mixing every 30 min. The TFA was 719 evaporated under stream of dry air at 40-45°C. Sugars were then reduced with 20 mg mL–1

720 sodium borohydride (NaBH4) in DMSO containing 0.2 M NH4OH and incubated at 45°C for 90 721 min with vortexing every 30 min. The solution was neutralized with 100 µL glacial acetic acid, 722 and 100 µL of 1-methyl-imidazole added, followed by addition of 0.75 mL acetic anhydride. The 723 mixture was incubated at 45°C for 30 min, and then 1.5 mL water added to destroy excess acetic 724 anhydride. After cooling, the alditol acetates were partitioned into dichloromethane and then 725 separated by gas-liquid chromatography (GLC) on a 0.25-mm x 30-m column of SP-2330 726 (Supelco, Bellefonte, PA). Injection was at 80°C with a 1-min hold, and then ramped to 170°C at 727 25°C min–1, and then to 240°C at 5°C min–1, with a 10-min hold at the upper temperature. 728 Helium flow was 1 mL min–1 with split-less injection. The electron impact mass spectrometry 729 (EIMS) was performed with an Agilent 5973N MSD at 70 eV and a source temperature of 730 250°C. 731 Linkage analysis was done as described previously by Gibeaut and Carpita (1991). Briefly, 732 silylation-grade DMSO was added by syringe to about 1 mg of lyophilized Golgi and cell wall 733 samples in a tube sealed with a serum stopper, and the tubes were sonicated for 90 min until the 734 water bath warmed to 50°C. n-Butyllithium (2.5 M) in hexane was added dropwise to the DMSO 735 suspension with rapid stirring while the sample was purged with Argon, and the solution was 736 stirred gently for about 1 h for evaporation of hexane and complete formation of the alkoxide 737 ions. Iodomethane was then added drop-wise to the stirred solution until neutralized. The 738 reaction was quenched with water, the partly methylated polymers were partitioned into 739 chloroform, and the chloroform phase was washed five times with water and dried under 740 nitrogen gas. The partly methylated polymers were then hydrolyzed in 2 M TFA for 90 min at

741 120°C, reduced with sodium borodeuteride (NaBD4), and acetylated as described above. GLC- 742 EIMS was performed as described in Gibeaut and Carpita (1991), and linkage structure inferred 743 as described in Carpita and Shea (1989). 21

744 745 Visualization of Maize Golgi Membranes 746 747 Structural integrity and enrichment of the Golgi membranes from flotation centrifugation was 748 confirmed by transmission electron microscopy (TEM). Briefly, 0.25 mL of Golgi membrane 749 fraction was fixed by addition of 25 μL of 25 % (v/v) glutaraldehyde to give a final concentration 750 of 2.5%. After 30 min, samples were then diluted slowly with additional 2.5% (v/v)

751 glutaraldehyde in 0.1 M KH2PO4, pH 7.2, to 1.5 mL. Pellets were collected after centrifugation 752 for 30 min and re-suspended in fresh 2.5% (v/v) glutaraldehyde for another 1 h. Fixed samples 753 were rinsed in buffer, post-fixed in buffered 1% (w/v) osmium tetroxide containing 0.8% (w/v) 754 potassium ferricyanide, rinsed again in distilled water, and embedded in agarose. The 755 membranes in agarose were stained en bloc with 1% (w/v) uranyl acetate, dehydrated with a 756 graded series of ethanol to 100%, and transferred into acetonitrile and embedded in EMbed-812 757 resin (Fisher). Thin sections were cut on a Reichert-Jung Ultracut E ultramicrotome and stained 758 with 4% (w/v) uranyl acetate and lead citrate. Images were acquired on a FEI Tecnai G2 T12 759 electron microscope equipped with a tungsten source and operating at 80 kV. 760 761 Free-Flow Electrophoresis (FFE) 762 763 Free-flow electrophoresis (FFE) of the three independent maize Golgi membrane preparations 764 collected after flotation centrifugation was performed essentially as described in Parsons et al. 765 (2012, 2014). Briefly, a Separation Buffer consisted of 10 mM sodium acetate, pH 7.0, 766 supplemented with 280 mM sucrose, 10 mM triethanolamine, 1 mM EDTA, and a Stabilization 767 Buffer was made by adjusting the Separation Buffer to 100 mM sodium acetate, pH 6.5, with 768 200 mM sucrose, 100 mM triethanolamine, 10 mM EDTA. Anodic and cathodic electrode 769 buffers were 100 mM sodium acetate, pH 6.5, with 100 mM triethanolamine and 10 mM EDTA. 770 Voltages of 680 to 700 V were applied to the membrane samples, resulting in a current of 132 to 771 137 mA. Media flow rate was set to 250 mL h–1, and sample flow rate to 2.5 mL h–1. Fractions 772 were collected in cooled 2-mL 96-well plates, and fractions of interest identified after detection 773 at 280 nm were pooled into four fractions. Membranes were collected from fractions by 774 centrifugation at 50,000 × g for 45 min, re-suspended in ice cold 10mM Tris-HCl, pH 7.5, and 775 stored at –80°C. 776 777 Analysis of the Golgi Membranes by Tandem Mass Spectrometry 778

22

779 Proteins isolated from the fractions enriched for Golgi membranes after flotation and FFE 780 separation were reduced for disulfide bonds in 10 mM DTT, cysteine-alkylated in 20 mM 781 iodoethanol, and then digested with trypsin (1:10 w/w) overnight at 37 °C in 40% (v/v) methanol 782 and 10 mM Tris[HCl], pH 7.8. After overnight digestion, samples were dried in a SpeedVac 783 concentrator. Digested peptides were desalted using Pierce C18 spin column (Thermo Scientific) 784 prior to LC-MS/MS analysis using manufacturer’s protocol. Peptide samples were rehydrated in 785 a solution of 2% (v/v) acetonitrile and 0.1% (v/v) formic acid. The samples were run on a nano 786 Eksigent 425 HPLC system coupled to the Triple TOF 5600 plus (Sciex, Framingham, MA). LC- 787 MS/MS data were collected using 120 min LC gradient at 300 nL/min over the cHiPLC nanoflex 788 system. The trap column was a Nano cHiPLC 200 µm x 0.5 mm ChromXP C18-CL 3 µm 120 Å 789 followed by the analytical column, the Nano cHiPLC 75 µm x 15 cm ChromXP C18-CL 5 µm 790 120 Å. The sample was injected into the Triple TOF 5600 plus through the Nanospray III ion 791 source equipped with an emission tip from New Objective. Peptides from the digestion were

792 eluted from the columns using a mobile phase A of purified H2O/0.1% formic acid (FA) and a 793 mobile phase B of ACN/0.1 % FA. With a flow rate of 0.3 µL min–1, the method started at 95% 794 A for 1 min followed by a gradient of 5% B to 35% B in 90 min and from 35% B to 80% B in 2 795 min. The run was held at 80% B for 5 min before being brought to 5% B and held for 20 min. 796 The data acquisition was performed monitoring 50 precursor ions at 250 ms per scan. Raw data 797 have been deposited in ProteomeXchange via the PRIDE database with the dataset identifier 798 PXD007612. [for reviewers: Username: [email protected]; Password: irDapzyh 799 800 LC-MS/MS Data Analysis 801 802 The raw data (.wiff) were processed using MaxQuant version 1.6.0.16 with its integrated 803 Andromeda search engine (Cox and Mann, 2008; Cox et al., 2014) and Mascot version 2.3.02 804 (Matrix Science) and searched against our custom maize annotation database (125561 sequences; 805 39895172 residues) comprising proteins from maize version 2 WGS (www.maizeGDB.org) 806 using the largest representative transcript of each gene in case of multiple models 807 (https://www.maizegdb.org/gbrowse/maize_v2test?l=CellWallGenes;l=Gene_models;q=Chr1:26 808 50000..2699999), and the common Repository of Adventitious Proteins (cRAP version 1.0, The 809 Global Proteome Machine). This maize annotation resource is also summarized and posted at our 810 Cell Wall Genomics website (https://cellwall.genomics.purdue.edu/families/index.html). Mascot 811 was set to search with the following parameters: peptide tolerance of ± 0.05Da, MS/MS 812 tolerance of ± 0.2 Da, fixed modification with Ethanolyl (C), variable modifications with 813 Acetylation (K) and Oxidation (M), for one missed cleavage for trypsin, and instrument type set 814 to ESI-QUAD-TOF. For both searches, decoy search option was activated to determine the False 23

815 Discovery Rate (FDR) of peptides and proteins. Mascot peptides matches were accepted if the 816 significance scores of their match had P value ≤0.05 and peptide score ≥13. Mascot search 817 results were filtered further to accept peptides with rank 1 and a significance homology threshold 818 of ≤1% FDR. MaxQuant search was also set at 1% FDR at both peptide spectrum match (PSM) 819 and protein levels. The minimum peptide length required for database search was set to six 820 amino acids. Other parameters applied for MaxQuant search include: precursor mass tolerance of 821 ± 0.05 Da, MS/MS fragment ions tolerance of ± 0.2 Da, maximum missed cleavage for trypic 822 digestion was set to one, methionine oxidation and acetylation (K) were set as the variable, while 823 Ethanolyl (C) was set as a fixed modification. Application of Pro hydroxylation as a variable 824 modification in either Mascot or MaxQuant identified no additional Hyp-containing 825 glycoproteins. 826 827 Post-Search Data Filtering 828 829 Proteins identified as false hits and contaminants, identified from MaxQuant decoy and 830 contaminants databases, were filtered out. For each experiment, proteins with negative scores 831 and those with a total MS/MS count of 0 across the four fractions, were filtered. Thus, for a 832 protein to be counted as identified in each experiment, it must be present in at least two fractions. 833 However, proteins in one fraction with MS/MS counts above 1, with supporting intensity data 834 and peptide counts, were considered valid hits. When proteins across experiments were 835 combined, those with an average MS/MS count of ≥1 in at least two experiments were 836 considered as identified. 837 If the same peptides are mapped to multiple proteins, Mascot and MaxQuant combines 838 these proteins into a protein group or protein family. For effective accountability of cell wall- 839 related proteins, we called out every protein from each cell wall-related protein group in 840 MaxQuant results that had a corresponding RNA expression value. The MS/MS counts assigned 841 to a specific group of cell wall proteins was shared among the proteins called out from the group 842 using the RNA expression value as a measure of distribution. When RNA expression was low for 843 a particular protein called out from a group, no MS/MS value was assigned. Protein groups with 844 low MS/MS counts were assigned to the protein with the highest transcript reads, and the 845 remainder were unassigned. 846 For the Mascot data, proteins with peptide expect values ≤ 0.05 were selected in each 847 experiment. Proteins were counted as identified if total spectral counts were above ≥1 and found 848 in at least two out of the four fractions, or ≥5 in a single fraction. For isoforms of some cell wall- 849 related proteins that could not be resolved by Mascot, but were detected by MaxQuant and have 850 expression support, we denoted them as unassigned. 24

851 852 Asymmetric Distribution of Proteins 853 854 For the enrichment analysis, the total pre-FFE and post-FFE abundance from the filtered Mascot 855 data were used because they represented a more comprehensive proteome dataset. For both the 856 pre-FFE and post-FFE Golgi results, spectral counts for each non-redundant protein were 857 summed to get total counts for both, and this value used for normalization against the highest 858 total spectral count between them. For determination of asymmetric distribution of different 859 classes of cell wall proteins, we pooled the protein data from three biological replicates across 860 the four fractions. Abundance of each protein across the four Golgi fractions was relative to the 861 fraction with highest abundance (1.0). Proteins with high abundance and identified with high 862 scoring peptides (Expect values ≤ 0.05 and scores ≥10) were generally used to determine the 863 distribution. 864 865 Maize Coleoptile Gene Expression 866 867 Three batches of etiolated maize coleoptiles (and etiolated leaves within), harvested at 2-, 2.5-, 868 and 3-d post-planting, were excised aseptically from the internodes and immediately plunged in

869 liquid N2. Samples in triplicate were kept frozen and pulverized by mortar and pestle under

870 liquid N2. Approximately 2 mg of ground tissue was incubated with 1 mL of ice-cold TRIzol 871 reagent (Thermo Fisher Scientific) and extracted according to the manufacturer’s directions. 872 Purified RNA was dissolved in 50 µL of diethyl pyrocarbonate-treated nanopure water and 873 quality and concentration were determined spectrophotometrically. 874 Expression analysis was carried out as described previously (Penning et al., 2014). Briefly, 875 the pooled RNA samples from the three biological replicates were delivered to Purdue’s 876 Genomics Core facility for RNA sequencing using an Illumina HiSeq 2000 to process 100-bp x 877 100-bp libraries of approximately 400-bp inserts. High quality trimmed sequences were mapped 878 to the Maize B73 sequence V2 from Plant GDB (http://www.plantgdb.org) using Bowtie2 879 (Langmead et al., 2009) except in instances where the reads mapped to exactly two places in the 880 genome due to gene duplication in maize. For these, a custom Perl script was used to split the 881 reads between the two locations (Penning et al., 2014). An average mapping rate of 80% was 882 achieved over all samples. The counts for each tissue were appended to an Excel readable file 883 with maize gene names and locations. Total reads from the RNAseq analyses of the three 884 sampling times were pooled and normalized to 20M reads. 885 886

25

887 Supplemental Data 888 889 Supplemental Table 1. Diagnostic linkage groups determined by gas-liquid chromatography- 890 electron-impact mass spectrometry of partly methylated alditol acetate derivatives of 891 polysaccharides. 892 Supplemental Table 2. Linkage composition of Arabidopsis cell wall fractions after sequential 893 extractions with ammonium oxalate, and increasing concentrations of alkali, and of Golgi 894 membranes after flotation centrifugation. 895 Supplemental Table 3. Linkage composition of maize cell wall fractions after sequential 896 extractions with ammonium oxalate, and increasing concentrations of alkali, and of Golgi 897 membranes after flotation centrifugation. 898 Supplemental Table 4. Cell-wall associated proteins in Golgi preparations from etiolated maize 899 coleoptiles after flotation centrifugation. 900 Supplemental Table 5. Comparison of annotated Maize GDB descriptions compared to those of 901 Phytozome for maize cell wall-related families of proteins identified in Golgi membranes 902 isolated by FFE. 903 Supplemental Table 6. Total non-cell wall Golgi proteins quantified as MaxQuant MS/MS 904 counts and Mascot spectral counts after FFE. 905 Supplemental Table 7. Total cell-wall proteins in Golgi membrane preparations from etiolated maize 906 coleoptiles after FFE quantified as MaxQuant MS/MS counts and Mascot spectral counts compared to 907 transcript abundance. 908 Supplemental Table 8. Arabidopsis Golgi-associated proteins identified after FFE (from Parsons et al., 909 2012). 910 911 Supplemental Figure 1. Flow chart for the isolation and enrichment of Arabidopsis and maize Golgi 912 membranes and cell wall pectic and hemicellulosic material. 913 Supplemental Figure 2. Panel of 155 mAbs employed in glycome analysis. 914 Supplemental Figure 3. Reproducibility and correlation of MaxQuant and Mascot analyses of maize 915 Golgi proteins enriched by FFE. 916 Supplemental Figure 4. Distribution of maize ER-Golgi proteins identified and quantified by 917 MaxQuant MS/MS counts after enrichment of Golgi by FFE. 918

26

919 Acknowledgments 920 We thank Vicki Hedrick (Proteomics Facility, Purdue) for her technical expertise in proteome 921 analysis, and Christopher Gilpen (LifeScience Microscopy Facility, Purdue) for assisting in the 922 TEM studies of membrane fractions. This work was supported by the U.S. Department of 923 Energy, Office of Science, Basic Energy Sciences (grant no. DE-SC0000997). This work was 924 also supported by the U. S. Department of Energy, Office of Science, Office of Biological and 925 Environmental Research, through contract DE-AC02-05CH11231 between Lawrence Berkeley 926 National Laboratory and the U. S. Department of Energy. The generation of the CCRC series of 927 plant cell wall glycan-directed monoclonal antibodies used in this work was supported by the 928 National Science Foundation Plant Genome Program (DBI-0421683 and IOS-0923992). 929 Immunological screening of cell wall and Golgi samples was also supported in part by the 930 United States Department of Energy-funded Center for Plant and Microbial Complex 931 Carbohydrates (Grant DE-FG02-93ER20097). 932 933 Disclaimer 934 The use of trade, firm, or corporation names in this paper is for the information and convenience 935 of the reader. Such use does not constitute an official endorsement or approval by the USDA- 936 ARS of any product or service to the exclusion of others that may be suitable. 937 Author Contributions: I.O.O, S.P, J.L.H., M.G.H., and N.C.C designed research, I.O.O, S.P., 938 S.M.G.F-N., U.K.A., B.W.P, J.L and N.C.C. performed experiments, I.O.O, S.P., U.K.A., 939 B.W.P, J.L.H., M.G.H., M.C.M. and N.C.C analyzed data, and I.O.O, S.P., U.K.A., B.W.P, 940 J.L.H., M.G.H., M.C.M. and N.C.C wrote the paper. 941 942 References 943 944 Balch, W.E., Dunphy, W.G., Braell, W.A., and Rothman, J.E. (1984) Reconstitution of the 945 transport of protein between successive compartments of the Golgi measured by the 946 coupled incorporation of N-acetylglucosamine. Cell 39: 405-416. 947 Brennan, M., and Harris, P.J. (2011) Distribution of fucosylated xyloglucans among the walls 948 of different cell types in monocotyledons determined by immunofluorescence microscopy. 949 Mol. Plant 4: 144-156. 950 Buckeridge, M.S., Vergara C.E., and Carpita, N.C. (1999) Mechanism of synthesis of a cereal 951 mixed-linkage (1 →3),(1→4)-β-D-glucan: Evidence for multiple sites of glucosyl transfer 952 in the synthase complex. Plant Physiol. 120: 1105-1116.

27

953 Burget, E.G., Verma, R., Mølhøj, M., and Reiter, W-D. (2003) The biosynthesis of L- 954 arabinose in plants: Molecular cloning and characterization of a Golgi-localized UDP-D- 955 xylose 4-epimerase encoded by the MUR4 gene of Arabidopsis. Plant Cell 15: 523–531. 956 Caffall, K.H., and Mohnen, D. (2009) The structure, function, and biosynthesis of plant cell 957 wall pectic polysaccharides. Carbohydr. Res. 344: 1879-1900. 958 Carpita, N.C., and Shea, E.M. (1989) Linkage structure by gas chromatography-mass 959 spectrometry of partially-methylated alditol acetates. In Analysis of Carbohydrates by GLC 960 and MS, C.J. Biermann, G.D. McGinnis, eds, (Boca Raton, FL: CRC Press), pp. 155-216. 961 Carpita, N.C., and Kanabus, J. (1987) Extraction of starch with dimethylsulfoxide and 962 quantitation by enzymic assay. Anal. Biochem. 161: 132-139. 963 Carpita, N.C., and Gibeaut, D.M. (1993) Structural models of primary cell walls in flowering 964 plants: Consistency of molecular structure with the physical properties of the walls during 965 growth. Plant J. 3: 1-30. 966 Carpita, N.C., and McCann, M.C. (1997) Some new methods to study plant polyuronic acids 967 and their esters. In Progress in Glycobiology, R. Townsend, A. Hotchkiss, eds, (New 968 York: Marcell Dekker), pp. 595-611. 969 Carpita, N.C., Defernez, M., Findlay, K., Wells, B., Shoue, D.A., Catchpole, G., Wilson, 970 R.H., and McCann, M.C. (2001) Cell wall architecture of the elongating maize coleoptile. 971 Plant Physiol. 127: 551-565. 972 Carpita, N.C., and McCann, M.C. (2010) The maize mixed-linkage (1→3),(1→4)-β-D-glucan 973 polysaccharide is synthesized at the Golgi membrane. Plant Physiol. 153: 1362-1371. 974 Cavalier, D.M., Lerouxel, O., Neumetzler, L., Yamauchi, K., Reinecke, A., Freshour, G., 975 Zabotina, O.A., Hahn, M.G., Burgert, I., Pauly, M., et al. (2008) Disrupting two 976 Arabidopsis thaliana xylosyltransferase genes results in plants deficient in xyloglucan, a 977 major primary cell wall component. Plant Cell 20: 1519-1537. 978 Crowell, E.F., Bischoff, V., Desprez, T., Rolland, A., Stierhof, Y.D., Schumacher, K., 979 Gonneau, M., Hofte, H., and Vernhettes, S. (2009) Pausing of Golgi bodies on 980 microtubules regulates secretion of cellulose synthase complexes in Arabidopsis. Plant Cell 981 21: 1141-1154. 982 Cox, J., and Mann, M. (2008) MaxQuant enables high peptide identification rates, 983 individualized p.p.b.-range mass accuracies and proteome-wide protein identification. Nat. 984 Biotechnol. 26: 1367-1372. 985 Cox, J., Hein, M.Y., Luber, C.A., Paron, I., Nagaraj, N., and Mann, M. (2014) Accurate 986 proteome-wide label-free quantification by delayed normalization and maximal peptide ratio 987 extraction, termed MaxLFQ. Mol. Cell Proteomics 13: 2513-2526.

28

988 Dallabernardina, P., Ruprecht, C., Smith, P.J., Hahn, M.G., Urbanowicz, B.R., and 989 Pfrengle, F. (2017) Automated glycan assembly of galactosylated xyloglucan 990 oligosaccharides and their recognition by plant cell wall glycan-directed antibodies. Org 991 Biomol Chem. 15: 9996-10000. 992 de Michele, R., McFarlane, H.E., Parsons, H.T., Meents, M.J., Lao, J., González 993 Fernández-Niño, S.M., Petzold, C.J., Frommer, W.B., Samuels, A.L., and 994 Heazlewood, J.L. (2016) Free-flow electrophoresis of plasma membrane vesicles enriched 995 by two-phase partitioning enhances the quality of the proteome from Arabidopsis 996 seedlings. J. Proteome Res. 15: 900-913. 997 Diet, A., Link, B., Seifert, G.J., Schellenberg, B., Wagner, U., Pauly M., Reiter, W-D., and 998 Ringli, C. (2006) The Arabidopsis root hair cell wall formation mutant lrx1 is suppressed 999 by mutations in the RHM1 gene encoding a UDP-L-rhamnose synthase. Plant Cell 18: 1000 1630-1641. 1001 Doblin, M.S., Pettolino, F.A., Wilson, S.M., Campbell, R., Burton, R.A., Fincher, G.B., 1002 Newbigin, E., and Bacic, A. (2009) A barley cellulose synthase-like CSLH gene mediates 1003 (1,3;1,4)-β-D-glucan synthesis in transgenic Arabidopsis. Proc. Natl. Acad. Sci. USA 106: 1004 5996-6001. 1005 Donohoe, B.S., Kang, B-H., Gerl, M.J., Gergeley, Z.R., McMichael, M., Bednarek, S.Y., 1006 and Staehelin, L.A. (2013) cis-Golgi cisternal assembly and biosynthetic activation occur 1007 sequentially in plants and algae. Traffic 14: 551-567. 1008 Drakakaki, G., van de Ven, W., Pan, S., Miao, Y., Wang, J., Keinath, N.F., Weatherly, B., 1009 Jiang, L., Schumacher, K., Hicks, G., et al. (2012) Isolation and proteomic analysis of 1010 the SYP61 compartment reveal its role in exocytic trafficking in Arabidopsis. Cell Res. 22: 1011 413-424. 1012 DuBois, M., Gilles, K.A., Hamilton, J.K., and Rebers, P.T., and Smith, F. (1956) 1013 Colorimetric method for determination of sugars and related substances. Anal. Chem. 8: 1014 350-356. 1015 Dugard, C.K., Mertz, R.A., Rayon, C., Mercadante, D., Hart, C., Benatti, M.R., Olek, A.T., 1016 SanMiguel, P.J., Reiter, W-D., McCann, M.C., and Carpita, N.C. (2016) The cell wall 1017 arabinose-deficient Arabidopsis thaliana mutant murus5 encodes a defective allele of 1018 Reversibly Glycosylated Polypeptide2. Plant Physiol. 171: 1905-1920. 1019 Dunkley, T.P.J., Watson, R., Griffin, J.L., Dupree, P., and Lilley, K.S. (2004) Localization of 1020 organelle proteins by isotope tagging (LOPIT). Mol. Cell. Proteomics 3: 1128-1134. 1021 Dunkley, T.P.J., Hester, S., Shadforth, I.P, Runions, J., Weimar, T., Hanton, S.L., Griffin, 1022 J.L., Bessant, C., Brandizzi, F., Hawes, C., et al. (2006) Mapping the Arabidopsis 1023 organelle proteome. Proc. Natl. Acad. Sci. USA 103: 6518-6523. 29

1024 Ford, K.L., Chin, T., Srivistava, V., Zeng, W., Doblin, M.S., Bulone, V., and Bacic, A. 1025 (2016) Comparative ‘Golgi’ proteome study of Lolium multiflorum and Populus 1026 trichocarpa. Proteomes 4: 23. 1027 Gaut, B.S., and Doebley, J.F. (1997) DNA sequence evidence for the segmental allotetraploid 1028 origin of maize. Proc. Natl. Acad. Sci. USA 94: 6809–6814. 1029 Gendre, D., McFarlane, H.E., Johnson, E., Mouille, G., Sjödin, A., Oh, J., Levesque- 1030 Tremblay, G., Watanabe, Y., Samuels, L., and Bhalerao, R.P. (2013) Trans-Golgi 1031 network localized ECHIDNA/Ypt interacting protein complex is required for the secretion 1032 of cell wall polysaccharides in Arabidopsis. Plant Cell 25: 2633-2646 1033 Gendre, D., Oh, J., Boutté, Y., Best, J.G., Samuels, L., Nilsson, R., Uemura, T., Marchant, 1034 A., Bennett, M.J., Grebe, M., et al. (2011) Conserved Arabidopsis ECHIDNA protein 1035 mediates trans-Golgi-network trafficking and cell elongation. Proc. Natl. Acad. Sci. USA 1036 108: 8048–8053 1037 Gibeaut, D.M., and Carpita, N.C. (1990) Separation of membranes by flotation centrifugation 1038 for in vitro synthesis of plant cell wall polysaccharides. Protoplasma 156: 82-93. 1039 Gibeaut, D.M., and Carpita, N.C. (1991) Tracing cell wall biogenesis in intact cells and plants. 1040 selective turnover and alteration of soluble and cell wall polysaccharides in grasses. Plant 1041 Physiol. 97: 551-561. 1042 Gibeaut, D.M., and Carpita, N.C. (1993) Synthesis of (1 →3),(1→4)-β-D-glucan in the Golgi 1043 apparatus of maize coleoptiles. Proc. Natl. Acad. Sci. USA 90: 3850-3854. 1044 Gibeaut, D.M., and Carpita, N.C. (1994) Improved recovery of (1→3),(1→4)-β-D-glucan 1045 synthase activity from Golgi apparatus of Zea mays (L.) using differential flotation 1046 centrifugation. Protoplasma 180: 92-97. 1047 Gutierrez, R., Lindeboom, J.J., Paradez, A.R., Emons, A.M., and Ehrhardt, D.W. (2009) 1048 Arabidopsis cortical microtubules position cellulose synthase delivery to the plasma 1049 membrane and interact with cellulose synthase trafficking compartments Nat. Cell Biol. 11: 1050 797-806. 1051 Haigler, C.H., and Brown Jr., R.M. (1986) Transport of rosettes from the Golgi apparatus to 1052 the plasma membrane in isolated mesophyll cells of Zinnia elegans during differentiation 1053 to tracheary elements in suspension culture. Protoplasma 134: 111-120. 1054 Heazlewood, J.L., Verboom, R.E., Tonti-Filippini, J., Small, I., and Millar, A.H. (2007) 1055 SUBA: the Arabidopsis Subcellular Database. Nucleic Acids Res. 35: D213-D218. 1056 Kang, B-H., Nielsen E., Preuss, M.L., Mastronarde, D., and Staehelin, L.A. (2011) Electron 1057 tomography of RabA4b and PI-4Kb1-labeled Trans Golgi Network compartments in 1058 Arabidopsis. Traffic 12: 313-329. 1059 Kim, J-B., and Carpita, N.C. (1992) Changes in esterification of the uronic acid groups of cell 30

1060 wall polysaccharides during elongation of maize coleoptiles. Plant Physiol. 98: 646-653. 1061 Kim, S-J., Zemelis, S., Keegstra, K., and Brandizzi, F. (2015) The cytoplasmic localization of 1062 the catalytic site of CSLF6 supports a channeling model for the biosynthesis of mixed- 1063 linkage glucan. Plant J. 81: 537-547. 1064 Kong, Y., Peña, M.J., Renna, L., Avci, U., Pattathil, S., Tuomivaara, S.T., Li, X., Reiter, W- 1065 D., Brandizzi, F., Hahn, M.G., et al. (2015) Galactose-depleted xyloglucan is 1066 dysfunctional and leads to dwarfism in Arabidopsis. Plant Physiol. 167: 1296-1306. 1067 Labrador, E., and Nicolás, G. (1984) Autolysis of cell walls in pea epicotyls during growth. 1068 Enzymatic activities involved. Physiol. Plant. 64:541-546. 1069 Langmead, B., and Salzberg, S.L. (2012) Fast gapped-read alignment with Bowtie 2. Nat. 1070 Methods 9: 357-359. 1071 Lending, C.R., Lippman, E., Bracker, C.E., and Bartnicki-Garcia, S. (1990) An efficient 1072 preparative isopycnic flotation method for the isolation of chitosomes. Protoplasma 159: 1073 16-25. 1074 Liu, L., Shang-Guan, K., Zhang, B., Liu, X., Yan, M., Zhang, L., Shi, Y., Zhang, M., Qian, 1075 Q., Li, J., and Zhou, Y. (2013) Brittle Culm1, a COBRA-like protein, functions in 1076 cellulose assembly through binding cellulose microfibrils. PLoS Genet. 9: e1003704 1077 Liu, L., Paulitz, J., and Pauly, M. (2015) The presence of fucogalactoxyloglucan and its 1078 synthesis in rice indicates conserved functional importance in plants. Plant Physiol. 168: 1079 549-560. 1080 Madson, M., Dunand, C., Li, X., Verma, R., Vanzin, G.F., Caplan, J., Shoue, D.A., Carpita, 1081 N.C., and Reiter, W.D. (2003) The MUR3 gene of Arabidopsis encodes a xyloglucan 1082 galactosyltransferase that is evolutionarily related to animal exostosins. Plant Cell 15: 1083 1662-1670. 1084 Mertz, R.A., Olek, A.T., and Carpita, N.C. (2012) Alterations in cell-wall glycosyl structure of 1085 Arabidopsis murus mutants. Carbohydr. Polym. 89: 331-339. 1086 Nebenführ, A., and Staehelin, L.A. (2001) Mobile factories: Golgi dynamics in plant cells. 1087 Trends Plant Sci. 6:160-167. 1088 Nikolovski, N., Rubtsov, D., Segura, M.P., Miles, G.P., Stevens, T.J., Dunkley, T.P.J., 1089 Munro, S., Lilley, K.S., and Dupree, P. (2012) Putative glycosyltransferases and other 1090 plant Golgi apparatus proteins are revealed by LOPIT proteomics. Plant Physiol. 160: 1091 1037-1051. 1092 Nikolovski, N., Shliaha, P.V., Gatto, L., Dupree, P., and Lilley, K.S. (2014) Label-free 1093 protein quantification for plant Golgi protein localization and abundance. Plant Physiol. 1094 166: 1033-1043.

31

1095 Parsons, H.T., Christiansen, K., Knierim, B., Carroll, A., Ito, J., Batth, T.S., Smith-Moritz, 1096 A.M., Morrison, S., McInerney, P., Hadi, M.Z., et al. (2012) Isolation and proteomic 1097 characterization of the Arabidopsis Golgi defines functional and novel components 1098 involved in plant cell wall biosynthesis. Plant Physiol. 159: 12-26. 1099 Parsons, H.T., Drakakaki, G., and Heazlewood, J.L. (2013) Proteomic dissection of the 1100 Arabidopsis Golgi and trans-Golgi network. Front. Plant Sci. 3: 298. 1101 Parsons, H.T., González Fernández-Niño, S.M., and Heazlewood, J.L. (2014) Separation of 1102 the plant Golgi apparatus and endoplasmic reticulum by free-flow electrophoresis. In Plant 1103 Proteomics: Methods and Protocols, Methods in Molecular Biology, vol. 1072. Springer 1104 Science and Business Media, pp. 527-539. 1105 Pattathil, S., Avci, U., Baldwin, D., Swennes, A.G., McGill, J.A., Popper, Z., Bootten, T., 1106 Albert, A., Davis, R.H., Chennareddy, C., et al. (2010) A comprehensive toolkit of plant 1107 cell wall glycan-directed monoclonal antibodies. Plant Physiol. 153: 514–525. 1108 Pattathil, S., Avci, U., Miller, J. S., and Hahn, M. G. (2012) Immunological approaches to 1109 plant cell wall and biomass characterization: glycome profiling. Methods Mol. Biol. 908: 1110 61–72. 1111 Pattathil, S., Avci, U., Zhang, T., Cardenas, C.L., and Hahn, M.G. (2015) Immunological 1112 approaches to biomass characterization and utilization. Front. Bioeng. Biotechnol. 3: 173. 1113 Peña, M.J., Ryden, P., Madson, M., Smith, A., Reiter, W-D., Carpita, N.C. (2004) 1114 Galactosylation of xyloglucans is essential for maintenance of cell wall tensile strength 1115 during cell growth in plants. Plant Physiol. 134: 443-451. 1116 Penning, B., Hunter, C.T., Tayengwa, R., Eveland, E., Dugard, C.K., Olek, A. Vermerris, 1117 W., Koch, K.E., McCarty, D.R., Davis, M., et al. (2009) Genetic resources for maize cell 1118 wall biology. Plant Physiol. 151: 1703-1728. 1119 Penning, B.W., Sykes, R.W., Babcock, N.C., Dugard, C.K., Held, M.A., Klimek, J.F., 1120 Shreve, J.T., Fowler, M., Ziebell, A., Davis, M.F., et al. (2014) Genetic determinants for 1121 enzymatic digestion of lignocellulosic biomass are independent of those for lignin 1122 abundance in a maize recombinant inbred population. Plant Physiol. 165: 1475-1487. 1123 Reiter, W-D., Chapple, C.C.S., and Somerville, C.R. (1993) Altered growth and cell walls in a 1124 fucose-deficient mutant of Arabidopsis. Science 261: 1032-1035. 1125 Reiter, W.D., and Vanzin, G.F. (2001) Molecular genetics of nucleotide sugar interconversion 1126 pathways in plants. Plant Mol. Biol. 47: 95-113. 1127 Rosquete, M.R., Davis, D.J., and Drakakaki, G. (2018) The plant Trans-Golgi Network: Not 1128 just a matter of distinction. Plant Physiol. 176: 187-198. 1129 Ruprecht, C., Bartetzko, M.P., Senf, D., Dallabernadina, P., Boos, I., Andersen, M.C.F., 1130 Kotake, T., Knox, J.P., Hahn, M.G., Clausen, M.H., and Pfrengle, F. (2017) A 32

1131 synthetic glycan microarray enables epitope mapping of plant cell wall glycan-directed 1132 antibodies. Plant Physiol. 175: 1094-1104. 1133 Ryden, P., Sugimoto-Shirasu, K., Smith, A.C., Findlay, K., Reiter, W-D., McCann, M.C. 1134 (2003) Tensile properties of Arabidopsis cell walls depend on both a xyloglucan cross- 1135 linked network and rhamnogalacturonan II-borate complexes. Plant Physiol 132: 1033- 1136 1040 1137 Saffer, A.M., Carpita, N.C., and Irish, V.V. (2017) Rhamnose-containing cell wall polymers 1138 suppress helical plant growth independently of microtubule orientation. Curr. Biol. 27: 1139 2248-2259. 1140 Scheller, H.V., and Ulvskov, P. (2010) Hemicelluloses. Annu. Rev. Plant Biol. 61: 263-289. 1141 Schmidt, D., Schuhmacher, F., Geissner, A., Seeberger, P.H., and Pfrengle, F. (2015) 1142 Automated synthesis of arabinoxylan-oligosaccharides enables characterization of 1143 antibodies that recognize plant cell wall glycans. Chemistry 21:5709-5713. 1144 Shedletzky, E., Shmuel, M., Delmer, D.P., and Lamport, D.T.A. (1990) Adaptation and 1145 growth of tomato cells on the herbicide 26-dichlorobenzonitrile leads to production of 1146 unique cell-walls virtually lacking a cellulose- xyloglucan network. Plant Physiol. 94: 980- 1147 987. 1148 Shedletzky, E., Shmuel, M., Trainin, T., Kalman, S., and Delmer, D. (1992) Cell wall 1149 structure in cells adapted to growth on the cellulose synthesis inhibitor 2,6- 1150 dichlorobenzonitrile: a comparison between two dicotyledonous plants and a gramineous 1151 monocot. Plant Physiol. 100: 120-130. 1152 Showalter, A.M. (1993) Structure and function of plant-cell wall proteins. Plant Cell 5: 9-23. 1153 Sun, Q., Zybailov, B., Majeran, W., Friso, G., Olinares, P.D., and van Wijk, K.J. (2009) 1154 PPDB, the plant proteomics database at Cornell. Nucleic Acids Res. 37: D969-D974. 1155 Tan, L., Showalter, A.M., Egelund, J., Hernandez-Sanchez, A., Doblin, M.S., and Bacic, A. 1156 (2012) Arabinogalactan-proteins and the research challenges for these enigmatic plant cell 1157 surface proteoglycans. Front. Plant Sci. 3: 140. 1158 Tan, L., Eberhard, S., Pattathil, S., Warder, C., Glushka, J., Yuan, C., Hao, Z., Zhu, X., 1159 Avci, U., Miller, J.S., et al. (2013) An Arabidopsis cell wall proteoglycan consists of pectin 1160 and arabinoxylan covalently linked to an arabinogalactan protein. Plant Cell 25: 270-287. 1161 Taylor, N.G. (2008) Cellulose biosynthesis and deposition in higher plants. New Phytol. 178: 1162 239-252. 1163 Terry, M.E., and Bonner, B.A. (1980) An examination of centrifugation as a method of 1164 extracting an extracellular solution from peas, and its use for the study of indoleacetic acid- 1165 induced growth. Plant Physiol. 66: 321-325.

33

1166 Terry, M.E., Jones, R.L., and Bonner, B.A. (1981) Soluble cell wall polysaccharides released 1167 from pea stems by centrifugation. I. Effect of auxin. Plant Physiol. 68:531-537. 1168 Toyooka, K., Goto, Y., Asatsuma, S., Koizumi, M., Mitsui, T., Matsuoka, K. (2009) A 1169 mobile secretory vesicle cluster involved in mass transport from the Golgi to the plant cell 1170 exterior. Plant Cell 21: 1212-1229. 1171 Uemura, T., Suda, Y., Ueda, T., and Nakano, A. (2014) Dynamic behavior of the trans-Golgi 1172 network in root tissues of Arabidopsis revealed by super-resolution live imaging. Plant Cell 1173 Physiol. 55: 694-703. 1174 Urbanowicz, B.R., Rayon C., and Carpita, N.C. (2004) Topology of the maize mixed-linkage 1175 (1 →3),(1→4)-β-D-glucan synthase at the Golgi membrane. Plant Physiol. 134: 758-768. 1176 Vain, T., Crowell, E.F., Timpano, H., Biot, E., Desprez, T., Mansoori, N., Trindade, L.M., 1177 Pagant, S., Robert, S., Hofte, H., et al. (2014) The cellulase KORRIGAN is part of the 1178 cellulose synthase complex. Plant Physiol. 165: 1521-1532. 1179 Vanzin, G.F., Madson, M., Carpita, N.C., Raikhel, N.V., Keegstra, K., Reiter, W.D. (2002) 1180 The mur2 mutant of Arabidopsis thaliana lacks fucosylated xyloglucan because of a lesion 1181 in fucosyltransferase AtFUT1. Proc. Natl. Acad. Sci. USA 99: 3340-3345. 1182 Wang, P.F, Chen, X.S., Goldbeck, C., Chung, E., and Kang, B.H. (2017) A distinct class of 1183 vesicles derived from the trans-Golgi mediates secretion of xylogalacturonan in the root 1184 border cell. Plant J. 92: 596-610. 1185 Wilson, S.M., Ho, Y.Y., Lampugnani, E.R., Van de Meene, A.M.L., Bain, M.P., Bacic, A., 1186 and Doblin, M.S. (2015) Determining the subcellular location of synthesis and assembly 1187 of the cell wall polysaccharide (1,3;1,4)-β-D-glucan in grasses. Plant Cell 27: 754-771. 1188 Xu, Z., Wang, M., Shi, D., Zhou, G., Niu, T., Hahn, M.G., O'Neill, M.A., and Kong, Y. 1189 (2017) DGE-seq analysis of MUR3-related Arabidopsis mutants provides insight into how 1190 dysfunctional xyloglucan affects cell elongation. Plant Sci. 258: 156-169 1191 Zabotina, O.A., Avci, U., Cavalier, D., Pattathil, S., Chou, Y-H., Eberhard, S., Danhof, L., 1192 Keegstra, K., and Hahn, M.G. (2012) Mutations in multiple XXT genes of Arabidopsis 1193 reveal the complexity of xyloglucan biosynthesis. Plant Physiol. 159: 1367-1384 1194 Zhang, G.F., and Staehelin, L.A. (1992) Functional compartmentation of the Golgi apparatus 1195 of plant cells: immunocytochemical analysis of high-pressure frozen- and freeze- 1196 substituted sycamore maple suspension culture cells. Plant Physiol. 99: 1070-1083. 1197 Zhang, Y., Nikolovski, N., Sorieul, M., Vellosillo, T., McFarlane, H.E., Dupree, R., Kesten, 1198 C., Schneider, R., Driemeier, C., Lathe, R., et al. (2016) Golgi-localized STELLO 1199 proteins regulate the assembly and trafficking of cellulose synthase complexes in 1200 Arabidopsis. Nat. Comm. 7: 11656. 1201 34

1202 Table 1. Inventory of protein families identified after free-flow electrophoresis. Numbers of isoforms identified in 1203 each family is in parentheses. A complete list of all cell-wall related proteins and their MaizeGDB v.3 ID 1204 numbers are in Supplemental Table 7. 1205 Protein Protein Description Maize GDB Protein Protein Description Maize GDB 1206 ID (number of isoforms) ID (number of isoforms) 1207 ______GT31C AGP UDP-Gal transferaseC (3) Nucleotide Sugar Interconversion GT31E AGP UDP-Gal transferase (2) AUD UDP-D-Glucuronate decarboxylase (6) HPGT GT31D Hyp O-galactosyltransferase (3) SUD UDP-D-Glucuronate decarboxylase (2) GT14 GT14 β-Glucuronosyltransferase (5) GAE UDP-D-Glucuronate-4-epimerase (7) GT16 GT16 Glycosyl transferase (1) UGD UDP-D-Glucose dehydrogenase (3) GT64 GT64 Glycosyl transferase (2) UXE UDP-D-Xylose 4-epimerase (4) GT66 GT66 Glycosyl transferase (2) UAM UDP-L-Arabinose mutase (RGP) (6) STT3 N-Oligosaccharyl transferase subunit (3) MNS1 α-Mannosidase (1) Nucleotide Sugar Transporters RRA GT77 Extensin UDP-Ara transferase (3) URGT NSTs/ UDP-Galactose transporter2 (2) P4H Prolyl 4-hydroxylase (8) UXT Nucleotide sugar transporter family1 (5) FUT GT10 GDP-3-α-L-fucosyl transferase (3) UTR CMP-Sialic acid transporter3 (2) EBS GT24 BRI1suppressor (2) NST-KT1 Nucleotide sugar transporter-KT1 (1) GPT Glucose-6-phosphate Transporter (2) Sucrose synthase PPT Phosphoenolpyruvate Transporter (2) SuSy Sucrose synthase (5) TPT Triose-phosphate Transporter (1) GalT Galactosyltransferase Transporter (2) Monolignol synthesis PAL1 Phenylalanine ammonia-lyase (8) NSTL Nucleotide sugar transporter family (6) C3H1 ρ-Coumarate 3-hydroxylase (2) Cellulose synthesis C4H Cinnamate 4-hydroxylase (3) CesA Cellulose Synthases (15) HCT ρ-Hydroxycinnamoyl transferase-like (1) CesAL Cellulose synthase-like (CesA-L) (1) APX5 Peroxidase (4)

Callose synthesis Polysaccharide modification GSL GT48 Glucan synthase-like (2) GH9(KOR) GH9 Glycosyl hydrolase/KORRIGAN (3) EXPB β-Expansin (2) GAX synthesis XTH Xyloglucan IRX14 GT43 Xylosyl transferase (4) endotransglucosylase/hydrolase (1) IRX9 GT43 Xylosyl transferase (3) BGAL17b GH35 β-Galactosidase (1) IRX10L GT47E Xylan 1,4-β-xylosyl transferase (8) MAN2B1 Lysosomal α-Mannosidase (1) GUX1 GT8A Xylan 1,2-α-GlcA transferase (1) GluA Lysosomal β-Glucosidase-like (1) GT61 GT61 Glycosyl transferase (9) EH-II Exohydrolase II (1) Xyloglucan synthesis GH13 Glucan 1,3-α-Glucosidase (2) CslC Xyloglucan β-glucan synthase (4) PGaseA PolygalacturonaseA (1) XXT GT34 Xyloglucan xylosyl transferase (10) PGIP1 Polygalacturonase inhibitor1 (3) GT47A GT47A Galactosyl transferase (3) PME Pectin methylesterase (1) FUTL11 GT37 Xyloglucan FutT11-like (1) PL4 Pectin and pectate lyase (1)

Mixed-linkage glucan synthesis Other cell wall-related proteins CslF Cellulose synthase-likeF (4) COBRA COBRA (2) STL STELLO (GT75-like) (1) Other cellulose synthase-like proteins CGL1 Complex Glycan Less1 (1) CslA Cellulose synthase-likeAs(1) CRK Cysteine-rich protein kinase (1) CslD Cellulose synthase-likeD (4) GATL3 Galacturonosyl transferase-like (1) CslE1 Cellulose synthase-likeE (1) GDPL Guanosine diphosphate-like (1) Pectin synthesis GT47D Glycosyl transferase47 D5 (1) GAUT GT8D UDP-GalA transferase (12) IPUT1 Inositol phosphorylceramide GT1 (1) MPL Cysteine Protease (2) AGP and glycoprotein synthesis SKU3 SKU (GPI-anchored Cupredoxin) (1) FLA Fasciclin-like arabinogalactan (3)

______

35

Figure Legends

Figure 1. Glycome analyses of chelator- and alkali-soluble fractions of isolated cell walls of Arabidopsis and maize. Glycome profiles of fractions obtained by sequential extraction of isolated cell walls, followed by screening of the wall extracts against a collection of plant glycan-directed monoclonal antibodies. Relative intensities of binding of antibodies are represented as heatmaps, with bright yellow depicting strongest intensity and black depicting not detected. Classes of polysaccharides and glycan epitopes are indicated on the right by small colored boxes. The list of antibodies for each polysaccharide class can be found in Supplemental Figure 2. Lanes represent the relative abundance of polysaccharide epitopes extracted sequentially with hot ammonium oxalate (AmmOx), followed by 0.1 M, 1.0 M, and 4.0 M NaOH.

Figure 2. Polysaccharides and glycans represented in chelator- and alkali-soluble fractions of isolated cell walls from Arabidopsis and maize. Linkage groups diagnostic for specific polysaccharides were pooled, and non-reducing terminal residues common to more than one polysaccharide were apportioned according to the abundance of their respective branch point residues. Abundances of diagnostic linkages for established polysaccharide structures are in Supplemental Tables 2 and 3. Protocols for polysaccharide assignment are described in Supplemental Table 1.

Figure 3. Isolation of Arabidopsis and maize Golgi membranes by flotation centrifugation. (A) Arabidopsis leaves were grown as a dense lawn for 21±1d, and (B) 2.5 to 3-day-old etiolated maize coleoptiles were overlaid with an equal volume of homogenization buffer with 84% (w/v sucrose) and gently mashed in a chilled mortar and pestle. Membranes and organelles in the homogenate are squeezed through nylon mesh, the homogenate adjusted to between 40 and 45% sucrose and placed at the bottom of the centrifuge tube. Step gradients with 35%, 29%, and 18% sucrose are overlaid, and the tubes centrifuged in a Beckman SW28 swinging bucket rotor at 27k rpm (131,500 x g at rmax). (C) Enriched maize and Arabidopsis Golgi membranes float upwards to the interface of the 35%/29% interface, and ER to the 18%/29% interface Mitochondria, plasma membrane, plastids and other cellular debris either pellet or remained trapped in the homogenate fraction. (D) Transmission electron micrograph of fixed and embedded preparations of a maize 29%/35% interface contain numerous Golgi bodies (arrows). Scale bar represents 400 nm.

36

Figure 4. Comparative glycome and polysaccharide profiles of cell wall and Golgi from maize etiolated coleoptiles and Arabidopsis seedlings.

(A) Glycome profiles of fractions obtained by sequential extractions of isolated maize and Arabidopsis cell walls compared to those from Golgi membranes screened against a collection of plant glycan-directed monoclonal antibodies. The cell wall intensities are a composite of relative epitope contributions in chelator- and alkali-soluble fractions (Figure 2). The Golgi relative intensities were normalized to adjust for differences in loading. Relative intensities of binding of antibodies are represented as heatmaps as described in Figure 1. (B) Maize and (C) Arabidopsis relative proportions of Golgi and pectin and hemicellulosic polysaccharides sequentially extracted in chelator- and alkali-soluble cell-wall fractions. Estimations of polysaccharide abundance were calculated as described in Figure 1. Golgi membranes uniquely contained 2-, 6-, and 3,6-Man residues associated with N-linked glycoproteins (N-Glyc). Abundances of diagnostic linkages for established polysaccharide structures are in Supplemental Tables 2 and 3, as described in Supplemental Table 1.

Figure 5. Enrichment of maize Golgi proteins by free-flow electrophoresis (FFE). (A) Profile of fractions enriched in Golgi proteins. Fractions 27 through 38 [shaded] were collected and pooled into four fractions for proteomics analysis. (B) Relative enrichment of Golgi proteins was determined by comparison of relative abundance of MS/MS counts from MaxQuant after flotation centrifugation compared to proteins collected in the Golgi fractions after FFE. Glycosyl transferases involved in synthesis of cellulose, GAX, XyG, and pectin were generally enriched in the Golgi fractions. (C) Distribution of classes of cell-wall proteins after flotation centrifugation and after FFE.

Figure 6. Comparison of relative expression of cell wall-related proteins in etiolated maize coleoptiles with relative protein abundance in Golgi membranes isolated by FFE. Abundance of cell wall-related proteins was derived from MaxQuant MS/MS counts across the twelve fractions of Golgi separated by FFE. Transcripts from RNAseq are quantified as numbers of reads per 20M total. Values for MS/MS counts and transcript abundance were pooled from three independent experiments.

37

Figure 7. Relative abundance of proteins across four fractions of Golgi membranes after FFE. (A) Portion of the profile of fractions from FFE enriched in Golgi proteins. (B) Profiles of proteins proposed as markers for ER- and Golgi- residence (in blue and red, respectively). Distribution of abundance of three proteins, ECA4, SYP32, and α-Man (in green), did not strictly follow an expected ER- or Golgi-residence profile. (C) Enzymes of nucleotide-sugar interconversion. (D) Proteins of pectin and hemicellulose synthesis. (E) CesAs (F) Profiles of three proteins associated with cellulose synthesis, STELLO, KORRIGAN (KOR), and COBRA (in blue), with CslF proteins (in red). A complete list of the proteins associated with non-cellulosic polysaccharide synthesis and metabolism is provided in Supplemental Table 7.

38

Figure 1. Glycome analyses of chelator- and alkali-soluble fractions of isolated cell walls of Arabidopsis and maize.

Glycome profiles of fractions obtained by sequential extraction of isolated cell walls, followed by screening of the wall extracts against a collection of plant glycan-directed monoclonal antibodies. Relative intensities of binding of antibodies are represented as heatmaps, with bright yellow depicting strongest intensity and black depicting not detected. Classes of polysaccharides and glycan epitopes are indicated on the right by small colored boxes. The list of antibodies for each polysaccharide class can be found in Supplemental Figure 2. Lanes represent the relative abundance of polysaccharide epitopes extracted sequentially with hot ammonium oxalate (AmmOx), followed by 0.1 M, 1.0 M, and 4.0 M NaOH.

Figure 2. Polysaccharides and glycans represented in chelator- and alkali-soluble fractions of isolated cell walls from Arabidopsis and maize.

Linkage groups diagnostic for specific polysaccharides were pooled, and non-reducing terminal residues common to more than one polysaccharide were apportioned according to the abundance of their respective branch point residues. Abundances of diagnostic linkages for established polysaccharide structures are in Supplemental Tables 2 and 3. Protocols for polysaccharide assignment are described in Supplemental Table 1.

Figure 3. Isolation of Arabidopsis and maize Golgi membranes by flotation centrifugation. (A) Arabidopsis leaves were grown as a dense lawn for 21±1d, and (B) 2.5 to 3-day-old etiolated maize coleoptiles were overlaid with an equal volume of homogenization buffer with 84% (w/v sucrose) and gently mashed in a chilled mortar and pestle. Membranes and organelles in the homogenate are squeezed through nylon mesh, the homogenate adjusted to between 40 and 45% sucrose and placed at the bottom of the centrifuge tube. Step gradients with 35%, 29%, and 18% sucrose are overlaid, and the tubes centrifuged in a Beckman SW28 swinging bucket rotor at 27k rpm (131,500 x g at rmax). (C) Enriched maize and Arabidopsis Golgi membranes float upwards to the interface of the 35%/29% interface, and ER to the 18%/29% interface Mitochondria, plasma membrane, plastids and other cellular debris either pellet or remained trapped in the homogenate fraction. (D) Transmission electron micrograph of fixed and embedded preparations of a maize 29%/35% interface contain numerous Golgi bodies (arrows). Scale represents 400 nm.

Figure 4. Comparative glycome and polysaccharide profiles of cell wall and Golgi from maize etiolated coleoptiles and Arabidopsis seedlings.

(A) Glycome profiles of fractions obtained by sequential extractions of isolated maize and Arabidopsis cell walls compared to those from Golgi membranes screened against a collection of plant glycan-directed monoclonal antibodies. The cell wall intensities are a composite of relative epitope contributions in chelator- and alkali-soluble fractions (Figure 2). The Golgi relative intensities were normalized to adjust for differences in loading. Relative intensities of binding of antibodies are represented as heatmaps as described in Figure 1. (B) Maize and (C) Arabidopsis relative proportions of Golgi and pectin and hemicellulosic polysaccharides sequentially extracted in chelator- and alkali-soluble cell-wall fractions. Estimations of polysaccharide abundance were calculated as described in Figure 1. Golgi membranes uniquely contained 2-, 6-, and 3,6-Man residues associated with N-linked glycoproteins (N-Glyc). Abundances of diagnostic linkages for established polysaccharide structures are in Supplemental Tables 2 and 3, as described in Supplemental Table 1.

Figure 5. Enrichment of maize Golgi proteins by free-flow electrophoresis (FFE). (A) Profile of fractions enriched in Golgi proteins. Fractions 27 through 38 [shaded] were collected and pooled into four fractions for proteomics analysis. (B) Relative enrichment of Golgi proteins was determined by comparison of relative abundance of MS/MS counts from MaxQuant after flotation centrifugation compared to proteins collected in the Golgi fractions after FFE. Glycosyl transferases involved in synthesis of cellulose, GAX, XyG, and pectin were generally enriched in the Golgi fractions. (C) Distribution of classes of cell-wall proteins after flotation centrifugation and after FFE.

Figure 6. Comparison of relative expression of cell wall-related proteins in etiolated maize coleoptiles with relative protein abundance in Golgi membranes isolated by FFE.

Abundance of cell wall-related proteins was derived from MaxQuant MS/MS counts across the twelve fractions of Golgi separated by FFE. Transcripts from RNAseq are quantified as numbers of reads per 20M total. Values for MS/MS counts and transcript abundance were pooled from three independent experiments.

Figure 7. Relative abundance of proteins across four fractions of Golgi membranes after FFE. (A) Portion of the profile of fractions from FFE enriched in Golgi proteins. (B) Profiles of proteins proposed as markers for ER- and Golgi- residence (in blue and red, respectively). Distribution of abundance of three proteins, ECA4, SYP32, and a-Man (in green), did not strictly follow an expected ER- or Golgi-residence profile. (C) Enzymes of nucleotide-sugar interconversion. (D) Proteins of pectin and hemicellulose synthesis. (E) CesAs (F) Profiles of three proteins associated with cellulose synthesis, STELLO, KORRIGAN (KOR), and COBRA (in blue), with CslF proteins (in red). A complete list of the proteins associated with non-cellulosic polysaccharide synthesis and metabolism is provided in Supplemental Table 7.

Minerva Access is the Institutional Repository of The University of Melbourne

Author/s: Okekeogbu, IO; Pattathil, S; Fernandez-Nino, SMG; Aryal, UK; Penning, BW; Lao, J; Heazlewood, JL; Hahn, MG; McCann, MC; Carpita, NC

Title: Glycome and Proteome Components of Golgi Membranes Are Common between Two Angiosperms with Distinct Cell-Wall Structures

Date: 2019-05-01

Citation: Okekeogbu, I. O., Pattathil, S., Fernandez-Nino, S. M. G., Aryal, U. K., Penning, B. W., Lao, J., Heazlewood, J. L., Hahn, M. G., McCann, M. C. & Carpita, N. C. (2019). Glycome and Proteome Components of Golgi Membranes Are Common between Two Angiosperms with Distinct Cell-Wall Structures. PLANT CELL, 31 (5), pp.1094-1112. https://doi.org/10.1105/tpc.18.00755.

Persistent Link: http://hdl.handle.net/11343/235808

File Description: Submitted version