Cracking the “Sugar Code”: A Snapshot of N- and O-Glycosylation Pathways and Functions in Plants Cells

Glycosylation is a fundamental co-translational and/or post-translational modification process where an attachment of sugars onto either proteins or lipids can alter their biological function, subcellular location and modulate the development and physiology of an organism. Glycosylation is not a template driven process and as such produces a vastly larger array of glycan structures through combinatorial use of enzymes and of repeated common scaffolds and as a consequence it provides a huge expansion of both the proteome and lipidome. While the essential role of N- and O-glycan modifications on mammalian glycoproteins is already well documented, we are just starting to decode their biological functions in plants. Although significant advances have been made in plant glycobiology in the last decades, there are still key challenges impeding progress in the field and, as such, holistic modern high throughput approaches may help to address these conceptual gaps. In this snapshot, we present an update of the most common O- and N-glycan structures present on plant glycoproteins as well as (1) the plant glycosyltransferases (GTs) and glycosyl hydrolases (GHs) responsible for their biosynthesis; (2) a summary of microorganism-derived GHs characterized to cleave specific glycosidic linkages; (3) a summary of the available tools ranging from monoclonal antibodies (mAbs), lectins to chemical probes for the detection of specific sugar moieties within these complex macromolecules; (4) selected examples of N- and O-glycoproteins as well as in their related GTs to illustrate the complexity on their mode of action in plant cell growth and stress responses processes, and finally (5) we present the carbohydrate microarray approach that could revolutionize the way in which unknown plant GTs and GHs are identified and their specificities characterized.

Glycosylation is a fundamental co-translational and/or post-translational modification process where an attachment of sugars onto either proteins or lipids can alter their biological function, subcellular location and modulate the development and physiology of an organism. Glycosylation is not a template driven process and as such produces a vastly larger array of glycan structures through combinatorial use of enzymes and of repeated common scaffolds and as a consequence it provides a huge expansion of both the proteome and lipidome. While the essential role of N-and O-glycan modifications on mammalian glycoproteins is already well documented, we are just starting to decode their biological functions in plants. Although significant advances have been made in plant glycobiology in the last decades, there are still key challenges impeding progress in the field and, as such, holistic modern high throughput approaches may help to address these conceptual gaps. In this snapshot, we present an update of the most common O-and N-glycan structures present on plant glycoproteins as well as (1) the plant glycosyltransferases (GTs) and glycosyl hydrolases (GHs) responsible for their biosynthesis; (2) a summary of microorganism-derived GHs characterized to cleave specific glycosidic linkages; (3) a summary of the available tools ranging from monoclonal antibodies (mAbs), lectins to chemical probes for the detection of specific sugar moieties within these complex macromolecules; (4) selected examples of Nand O-glycoproteins as well as in their related GTs to illustrate the complexity INTRODUCTION In the model plant Arabidopsis thaliana, approx. 10-15% of the genome is devoted to construction, dynamic architecture, sensing functions, and metabolism of the plant cell wall (Cosgrove, 2005). The major components of plant cell walls include a complex composite of polysaccharide networks, lignin (secondary walls) together with minor amounts (generally less than 10%) of Nand/or O-glycosylated proteins (Somerville et al., 2004;Cosgrove, 2005;Albenne et al., 2009;Ellis et al., 2010;Lamport et al., 2011;Zielinska et al., 2012). Protein glycosylation, through co-and/or post-translational modification, results in addition of glycans (mono-/oligo-/polysaccharides and GPI anchors) that influence a protein's stability, location and functional properties (Lerouge et al., 1998;Nagashima et al., 2018). While N-glycan synthesis in the endoplasmic reticulum (ER) is relatively well conserved in eukaryotes, N-glycan processing and O-glycan biosynthesis in the Golgi apparatus (GA) are kingdom-specific and result in different oligosaccharide structures attached to glycoproteins in plants and mammals (Gomord et al., 2010). The prasinophytes situated at the base of the green plant lineage feature a much simpler set of N-glycan elaborations (Ulvskov et al., 2013) which may represent either the primordial eukaryotic N-glycosylation machinery or be the result of gene loss. Following initial processing steps in the ER, the N-glycans show differences in the maturation steps in the GA. Interestingly, plant N-glycans differ from their animal counterparts by the following: (1) the complete absence of sialic acid, (2) the core Fuc residues (where present) are α(1 → 3) rather than α(1 → 6)-linked to the reducing GlcNAc, and (3) the core β-mannosyl residue is often substituted with Xylβ(1 → 2) (Sturm, 1995). In contrast to N-glycosylation, the primary mechanism for O-glycosylation in plants is unique among eukaryotes and is via attachment to the hydroxyl group of the imino acid hydroxyproline (Hyp/O; in mammalian systems this type of glycosylation is to hydroxylysine) and less commonly to the hydroxyl group of serine [Ser; e.g., in extensins (EXTs) (Kieliszewski, 2001)]. This O-linked glycosylation determines the molecular properties and biological functions of members of the Hyp-rich glycoprotein (HRGP) superfamily and some secreted small peptides (e.g., CLE for CLAVATA3/Endosperm surrounding region). In addition, in plants there is a complete absence of GalNAc-Ser/Thr in secreted glycoproteins that is common in mammalian secreted glycoproteins and whilst there are also some other forms of O-glycosylation they are less common (e.g., Ser-O-GlcNAc on cytoplasmic and nuclear proteins). This overview should be read in conjunction with more focused reviews recently published by Seifert (2020) and Silva et al. (2020) to gain a comprehensive coverage of the structure, function and biosynthesis of arabinogalactanproteins (AGPs).

Daucus carota
Suspension cultured cells Knox et al., 1991;Yates et al., 1996;Ruprecht et al., 2017 Daucus carota Suspension cultured cells   Pennell et al., 1989;Yates et al., 1996 PN16.4B4 Nicotiana glutinosa Suspension cultured cells Unknown Norman et al., 1986;Pattathil et al., 2010 β-Glc Yariv (synthetic dye) Yariv et al., 1967;Kitazawa et al., 2013;Paulsen et al., 2014 an Arabidopsis mutant lacking complex N-glycans due to a deficiency in the UDP-GlcNAc transporter 1 (UGNT1) does not show a salt sensitivity phenotype (Ebert et al., 2018). Apart from Arabidopsis, complex N-glycan deficient rice and Lotus japonicus mutants have been characterized which display severe defects in growth and reproduction (Fanata et al., 2013;Strasser, 2014;Harmoko et al., 2016;Pedersen et al., 2017). Collectively, while the oligomannosidic N-glycans play a role in ER-quality control, the potential suite of biological functions of complextype and paucimannosidic N-glycans on glycoproteins is still largely unknown and the underlying mechanisms remain to be elucidated for the described phenotypes in Arabidopsis and other plant species. The biological function of the β-N-acetylhexosaminidase (HEXOs), especially HEXO3 acting at the plasma membrane/apoplast is unknown. In addition to HEXOs, it is possible that other GHs (Table 1) liberate monosaccharides from complex N-glycans either on a specific group of glycoproteins or in specific cell-types. Golgi-localized enzymes such as the recently characterized exo-β-(1 → 3)-galactosidases (Nibbering et al., 2020) may to some extent hydrolyze the Gal transferred by GALT1 directly in the Golgi (Table 2). Similarly, either Nicotiana benthamiana BGAL1 or another GH with β-(1 → 3/4)-galactosidase activity could modify Lewis A structures in the apoplast (Kriechbaum et al., 2020). Plant α-(1 → 3/4)-fucosidases that can cleave off Fuc residues from Lewis A structures have been identified in several plant species (Zeleny et al., 2006;Rahman et al., 2016;Kato et al., 2018). The only Arabidopsis GH29 α-(1 → 3/4)-fucosidase, AtFUC1, acts in the glycan degradation pathway in the vacuole and hydrolyses primarily the core α-(1 → 3)-linked Fuc. Consistent with the described substrate specificity, the AtFUC1-deficient mutant displayed slightly higher levels of Lewis A containing complex N-glycans. The degradation pathway for oligomannosidic and complex N-glycans in the vacuole involves several GHs whose substrate specificities are already well characterized (Léonard et al., 2009;Ishimizu, 2015;Kato et al., 2018). Apart from exo-glycosidases, plants have endo-glycosidases such as peptide-N-glycanase A (PNGase A) that is active on small glycopeptides and hydrolyzes complex N-glycans with core α-(1 → 3)-linked Fuc (Tretter et al., 1991;Altmann et al., 1998). Certain plant tissues such as the maize endosperm harbor an endo-glycanase (ENGase) that is active on oligomannosidic N-glycans and cleaves within the chitobiose core (Rademacher et al., 2008). Single GlcNAc residues or chitobiose at N-glycosylation sites have been detected on plant proteins (Ishimizu et al., 1999;Kim et al., 2013;Xu et al., 2016). How abundant those truncated glycans are and whether they have specific functions or represent intermediates of degradation pathways remains to be shown.

O-GLYCANS, GTs AND GHs OF PLANT GLYCOPROTEINS
O-linked glycosylation defines the molecular properties and biological function of the HRGP superfamily and some secreted small hormone peptides (e.g., CLE-like peptides). The HRGP superfamily is traditionally divided into three major subgroups: AGPs, EXTs including the Leucine-Rich eXtensins (LRXs), and the repetitive Pro-rich proteins (PRPs) (Seifert and Roberts, 2007;Ellis et al., 2010;Tan et al., 2012;Hijazi et al., 2014;Johnson et al., 2017). However, the HRGP superfamily is better understood as a spectrum of molecules ranging from the highly glycosylated AGPs to the minimally O-glycosylated PRPs. Two major types of O-glycans are attached to Hyp (O) in plant glycoproteins. The first type includes unbranched chains of up to five arabinose (Ara) units added to clusters of Hyp residues in EXTs (Marzol et al., 2018) and small CLE-like peptides (Ohyama et al., 2009;Shinohara and Matsubayashi, 2013). The second type are complex type II arabino-3,6-galactans (AGs) which are attached to non-contiguous Hyp residues (AO/SO/TO/VO) on AGPs and AGP-like proteins (Johnson et al., 2017). Finally, a single Gal is linked to Ser mostly in EXTs and EXT-related proteins. The Hyp contiguity hypothesis proposes that the addition of these two main types of O-glycan is controlled by "glycomotifs" in the HRGP protein sequence (Kieliszewski, 2001). This hypothesis predicts that short arabino-oligosaccharides are added to contiguous Hyp 3−5 residues in EXTs, whereas complex AGs are assembled on clustered but non-contiguous Hyp residues in AGPs (Shpak et al., 1999;Tan et al., 2010). The only exception to this rule is CLE-like peptides (e.g., Tob/Tom-HypSys, PSY1, CLV3, and CLE2), in which non-contiguous Hyp residues are arabinosylated (Ohyama et al., 2009;Shinohara and Matsubayashi, 2013). The extent of glycosylation of PRPs remains unclear with low levels of Ara residues presumably O-linked to Hyp (Bernhardt and Tierney, 2000).

Arabinogalactan-Proteins-O-glycans and GTs
Arabinogalactan-proteins are complex cell surface proteoglycans with type II AG glycan moieties attached at non-contiguous Hyp residues consisting of a β-(1 → 3)-galactan backbone substituted at C(O)6 with side chains of β-(1 → 6)-galactan of variable length decorated further with Ara, and less frequently also with Fuc, Rha, (O-methyl)glucuronic acid (4-O-MeGlcA) and Xyl (Figure 2). AGPs have been implicated in a diverse array of plant growth and development processes including hormone signaling, cell expansion and division, embryogenesis of somatic cells, differentiation of xylem, reproduction and responses to abiotic stress (Seifert and Roberts, 2007;Ellis et al., 2010;Ma et al., 2018). Recently, it was shown that perturbing an AG-peptide (AGP21) in Arabidopsis triggers aberrant root hair development by altering expression of the homeodomain protein GLABRA 2 (GL2) expression in a BIN2 (a Type-II GSK3-like kinase)-dependent manner, similar to the phenotype observed in plants with defective brassinosteroid signaling (Borassi et al., 2020). These results imply an interesting parallel between plant AGPs and animal heparin sulfate proteoglycans (HSPGs), which are important co-receptors in signaling pathways mediated by growth factors, including members of Wnt/Wingless, Hedgehog, transforming growth factor−β, and fibroblast growth factor family members (Lin, 2004). AGP4, AGP6, and AGP11 from Arabidopsis have been shown to be essential for reproduction, with AtAGP4 shown to play a critical role in synergid degeneration and prevention of more than one pollen tube being attracted to the embryo sac (Pereira et al., 2016). AG glycan structures have also been found to be involved in reproductive development in Torenia fournieri with a methyl-glucuronosyl arabinogalactan (AMOR) released from the ovule inducing the competency of the pollen tube to respond to ovular attractant peptides (Mizukami et al., 2016;Jiao et al., 2017). UPEX1/KNS4/GALT14, a galactosyltransferase (GALT) from Arabidopsis that generates the β-(1 → 3)galactan backbone of type II AG, has been shown to be vital for normal pollen exine development as upex1/kns4/galt14 mutants display a collapsed pollen phenotype with reduced viability and fertility (Suzuki et al., 2017). The requirement for specific glycan structures on AGPs for Ca 2+ signaling during development is supported by mutants in GlcAT14 members. AG glycans with reduced glucuronosylation were shown to have lower Ca 2+ binding capacity (Lopez-Hernandez et al., 2020). Double/triple glcat mutants displayed developmental defects that could be suppressed by additional Ca +2 in growth media. Unique glycan structures on AGPs in seagrasses, that include a high content of terminating 4-O-methyl-GlcA residues, are proposed to strengthen Ca 2+ binding and limit the effects of salt as an adaptation to the marine environment (Pfeifer et al., 2020). These few examples demonstrate the indispensable nature of AGPs to plant processes and the important function their O-glycan moieties play, although their mechanistic role continues to remain elusive and ill-defined as recently reviewed (Seifert, 2020).  Table 3. Number in brackets refer to GHs CAZY family. Please also see Silva et al. (2020) for GTs and GHs acting in AGP O-glycan processing.

Variability of Type II AG O-Glycans
Compared to the relatively high degree of conservation of N-glycan structures, O-glycans attached to AGPs display a considerable degree of variation on every level (Figure 3 and references there in). There are variations between different species and tissues and in the same cell type at different stages of development. The common structural feature of type II AG that are O-linked to isolated Hyp residues on AGPs is a backbone of β-(1 → 3) Gal that contains β-(1 → 6) linked Gal side chains of variable length, although there are examples of β-(1 → 6) linked Gal backbones (Raju and Davidson, 1994;Dong and Fang, 2001). In some reports a β-(1 → 6) linked Gal is further β-(1 → 3) galactosylated forming a kink in the backbone (Churms et al., 1983;Bacic et al., 1987). Mostly however, the Gal side chains are modified by α-(1 → 3) linked L-Araf (Tryfona et al., 2010(Tryfona et al., , 2012 and references therein). Additionally, the side branches can contain β-(1 → 6) linked GlcA or 4-O-MeGlcA. The L-Araf side groups are sometimes extended by one or two α-(1 → 3) linked L-Araf residues and terminated by either α-(1 → 3) linked L-Araf or α-(1 → 2) linked L-Fuc. In some cases, the L-Fuc is not the terminal sugar but further modified by β-(1 → 3) linked D-Xyl. While L-Araf incorporated in plant cell wall carbohydrates is predominantly found in its furanose form there have also been reports on L-Arap β-(1 → 3) linked to Araf or Galp as terminal sugars. Likewise, GlcAp and 4-Omethyl D-GlcAp are often found as terminal modifications of the galactan backbone but sometimes GlcA was found decorated by α-(1 → 4) linked L-Rha. In other cases, another β-(1 → 4) linked D-GlcA followed and terminated by β-(1 → 4) linked 4-O-methyl D-GlcAp were linked to this sugar. Another modification of D-GlcAp was α-(1 → 4) linked L-Rha as the first sugar of an extended heteropolymer resembling rhamnogalacturonan I. Besides this staggering multitude of structures attributed to AGP-linked type II AG, there exists variability in the degree of substitution of individual Hyp residues as well as the sizes of the individual glycans. This was demonstrated for artificial AGP-like fluorescent proteins that showed considerable variations in apparent molecular weight between different organs (Estevez et al., 2006). Moreover, the cell-type specific variation between type II AG structures is elegantly revealed by AGP-glycan specific monoclonal antibodies (mAbs) ( Table 3).
An as yet incomplete list of GHs acting on various linkages in type II AGs are mainly known from various microbial sources (Table 3) and are used for their structural characterization. However, plant endogenous AGP-specific GHs have also been described. Two family GH43 exo-β-(1 → 3)-galactanases from Arabidopsis were shown to be required for controlling the apparent abundance of AGPs and their loss of function resulted in a sugar-conditional root expansion phenotype characteristic of many primary cell wall-defective mutants (Nibbering et al., 2020). Arabidopsis also has three close homologs encoding family 79 GHs. One member of this family named AtGUS2 was identified in a gel filtration fraction that showed Oβ-glucuronidase activity in vitro (Eudes et al., 2008). A T-DNA insertion in this locus displayed abnormally short hypocotyls and overexpression of AtGUS2 enhanced both hypocotyl length and root length with purified AGPs displaying lower terminal-GlcA content. Finally, four Arabidopsis loci encode family 27 GHs named β-L-ARAPASE (APSE), and α-GALACTOSIDASE 1-3 FIGURE 3 | Arabinogalactan-protein glycan variation. Five structures of type II glycans found on AGPs demonstrating common motifs and variations. (A) This relatively small glycan was produced on an artificial AGP recombinantly expressed in tobacco cell cultures by Tan et al. (2004). Note the β-(1 → 6) kink in the β-(1 → 3) galactan backbone. (B) This structure approximates the model described for AGP glycans purified from A. thaliana leaves (Tryfona et al., 2012). Note that the actual size of many of the glycans is probably much bigger than the structure displayed here. In a later study by the same group, the terminal modification of L-Fuc by D-Xyl was described (Tryfona et al., 2014). (C) Using the same tools of enzymatic degradation and mass spectrometry, this group also described the glycan-structure of wheat flour AGP (Tryfona et al., 2010). Again, we show an approximation of their model that should accommodate large variations in glycan size. A noteworthy feature of this glycan is the occurrence of terminally linked L-Arap. (D) AGP-glycans of the see grass Zostera marina are particularly rich in 4-Me-GlcAp (Pfeifer et al., 2020). (E) The partial glycan structure of the type II AG linked to an AGP named as APAP1 that is linked to both rhamnogalacturonan 1 (RG1) and arabinoxylan (AX) (Tan et al., 2013). Legend for sugar symbols is as per Figure 2. (AGAL1-3) (Imaizumi et al., 2017). Although the majority of L-Ara found in plant carbohydrates is in its furanose form some examples of L-arabinopyranose (Arap) exist, one example being found in type II AGs (Tryfona et al., 2010). It was suggested that APSE and the AGALs act on these residues (Imaizumi et al., 2017), and apse agal3 mutants showed decreased β-L-arabinopyranosidase activity and increased levels of β-L-Arap, compared to wild type. Apart from a decrease in hypocotyl length, the apse agal3 mutants appeared phenotypically normal. Finally, a promiscuous α-L-arabinofuranosidase/β-D-xylosidase belonging to family GH3 has been purified and cloned from radish (Kotake et al., 2006]. However, since the Ara and Xyl residues exist in various carbohydrates it is presently unknown whether any of the fifteen Arabidopsis GH3 enzymes act as AGPspecific α-L-arabinofuranosidases. In addition, please see Silva et al. (2020) for GHs, both endogenous and heterologous, acting on type-II AG glycans of AGPs.

Decoding EXTs and Their O-Glycans Functions
It is already known that O-glycans increase HRGP solubility, resistance to proteolytic degradation and thermal stability (Shpak et al., 2001;Kieliszewski et al., 2011;Lamport et al., 2011;Seifert, 2020). EXTs are able to form, at least in vitro, a tridimensional covalent network through diTyr-linkages mediated by EXT peroxidases between individual EXT molecules and also via self-recognition and alignment of hydrophilic O-glycosylated Ser-(Hyp) 3−4 repeats and hydrophobic peptide-cross-linking modules (Cannon et al., 2008). Thus, the ordered EXT monomer assembly in plant cell walls would involve a zipper-like endwise association via cross-linking at the ends of the molecules Lamport et al., 2011). Recently, modeling experiments suggested that classical EXTs would be able to form a putative triple helix structure by lateral staggered alignment (Cannon et al., 2008) and diTyr cross-linking, similar to that present in collagen (Velasquez et al., 2015b;Marzol et al., 2018). It is also proposed that EXTs interact with pectins by a simple acid-base reaction forming a supramolecular ionic structure in the nascent cell wall (Valentin et al., 2010), which would serve as a framework for further cell wall deposition (Cannon et al., 2008;Lamport et al., 2011). In addition, covalent EXT-pectin cross-links were also suggested (Nuñez et al., 2009). However, it is unclear how EXT monomers are secreted and assembled into the glyco-network and how EXT and related glycoproteins-pectin interactions are controlled in a coordinated way during new cell wall formation.
Several mutants in O-glycosylation GTs of EXTs and related proteins (e.g., LRXs) have similarities to root hair-defective growth phenotypes (Velasquez et al., 2011; and EXT content and their O-glycosylation levels were correlated with cotton fiber cell elongation (Guo et al., 2019), highlighting that O-glycans in EXTs affect EXT function during plant cell expansion. Furthermore, an in vitro study has revealed that both Ser-O-galactosylation and Hyp-O-arabinosylation determine the rate of EXT crosslinking and hence the efficiency of EXT network formation (Chen et al., 2015). Thus, correct arabinosylation of EXTs is essential for their in vivo functions. In addition, some of these mutants (e.g., rra2 and xeg113) showed enhanced susceptibility for specific root pathogens (Castilleux et al., 2020). The known roles of EXTs in cell wall assembly, cell shape and growth raises the question to the function of each individual EXT molecule (Hall and Cannon, 2002;Cannon et al., 2008;Velasquez et al., 2011). Although the Arabidopsis genome encodes several EXTs, so far only a single EXT mutant rsh (for root shoot hypocotyl-defective)/ext3) have a nearly lethal phenotype (Cannon et al., 2008). This finding suggests either the high redundancy or masked functions of EXTs in plant development, although their role in root hairs, pollen tubes and root growth are clear exceptions to this rule. Several EXT mutants (ext6-7/12-14/18) (Velasquez et al., 2011) and lrx1/2 mutants have aberrant root hair morphologies (Baumberger et al., 2001(Baumberger et al., , 2003aRingli, 2010) and prp3 (Bernhardt and Tierney, 2000) display short root hairs. Characterization of multiple mutants for pollen LRXs (lrx8/9/10/11) indicates they are key components for proper polar growth as sentinels of cell wall integrity in these rapidly expanding cells (Fabrice et al., 2018;Wang et al., 2017;Sede et al., 2018;Herger et al., 2019) while the triple mutant lrx3/4/5 showed defects in cell expansion in root cells (Draeger et al., 2015), possibly mediated by abnormal vacuolar expansion . Recently, a mechanism of action for LRXs was proposed based on LRX8 and LRX9 binding in the apoplast to the Rapid Alkalinization Factor 4-19 (RALF4 and RALF19) peptides as well as to the extracellular domains of some transmembrane receptors such as CrRLK1Ls (e.g., ANX1,2 and BUDS1,2) (Ge et al., 2017;Mecchia et al., 2017). In a similar manner, the extracellular LRX3/4/5-RALF22/23 together with CrRLK1L FERONIA (FER) are able to coordinate growth under salt conditions (Zhao et al., 2018(Zhao et al., , 2020 and LRX1/5-RALF1-FER in shoot and root growth Herger et al., 2020). It has been proposed that LRXs work together with CrRLK1Ls and RALF peptides to monitor the plant cell wall integrity status during cell growth (Ge et al., 2017;Mecchia et al., 2017;Dünser et al., 2019;Herger et al., 2020). Although the structural basis for the interaction between LRXs and RALFs peptides was recently established (Moussu et al., 2020), it is unclear how the O-glycans in the EXT domain of LRXs affects these proteinprotein interactions. Since the EXT domain is variable among LRXs both in terms of length and motif (Baumberger et al., 2003a,b;Borassi et al., 2016), it is proposed that it has adapted to the specific cell wall architecture of the numerous tissues where they are located as putative cell wall integrity sensors (Baumberger et al., 2003a,b;Marzol et al., 2018;Sede et al., 2018;Herger et al., 2019).

CHEMICAL SYNTHESIS, GLYCAN ARRAYS AND TECHNOLOGICAL CHALLENGES
The tremendous heterogeneity of plant cell wall glycans such as the O-glycans in AGPs make the identification of the exact molecular structures that serve either as acceptors for GTs, substrates for GHs or epitopes for mAbs very challenging. There are basically two options to procure suitable oligosaccharide samples for biochemical assays used in GT functional studies. One possibility is purification of oligosaccharides from digests of natural polysaccharides or glycoproteins, which can provide a large number of oligosaccharides in acceptable time, but oftentimes with compromised purity and in limited quantities (Tan et al., 2012). The second possibility is chemical synthesis, which gives access to significant amounts of well-defined and pure oligosaccharides but is very time consuming (Kinnaert et al., 2017;Pfrengle, 2017). Automated glycan assembly (AGA) can significantly accelerate the process of chemical synthesis for a number a glycan classes (Seeberger, 2015). In AGA, protected monosaccharide building blocks are coupled in a stepwise manner to a linker-functionalized Merrifield resin, in a computer-controlled and automated manner. While many different complex oligosaccharides have been synthesized by AGA, only recently has it begun to be explored for synthesizing plant glycans, including AGP O-glycans (Bartetzko and Pfrengle, 2019). Chemically synthesized glycans as well as natural polysaccharides and isolated oligosaccharides can be printed as glycan arrays to obtain high-throughput platforms for analyzing plant cell wall-related enzymes and molecular probes such as mAbs (Møller et al., 2008;Pedersen et al., 2012). A recently developed glycan array equipped with chemically synthesized plant cell wall glycans, including many AGP glycan related substrates, has proven useful for the rapid characterization of a large number of cell wall glycan-directed mAbs (Ruprecht et al., 2017). The same glycan array has also aided in identifying acceptor substrates for GTs involved in AGP glycan biosynthesis such as GalT31A and FUT7 (Figure 4; Ruprecht et al., 2020). By extension, this technology has the potential to reveal the biochemical function of novel GTs that act in the O-glycosylation pathway of plant HRGPs and other glycoproteins.

PERSPECTIVES AND FUTURE CHALLENGES IN PLANT GLYCOBIOLOGY
Major developments in nuclease-based gene editing, quantitative transcriptomics, metabolomics, and proteomics are now enabling high throughput approaches to explore plant protein and lipid glycosylation through analyzing and targeting enzymes involved in glycosylation processes. Although there has been significant progress in plant glycobiology, there are still many remaining fundamental questions to be addressed. Here, we attempt to highlight some selected aspects that are key to accelerating progress in this field: • In vivo N-and O-glycan mapping. The chemical reporter strategy known as bio-orthogonal click chemistry has arisen as a powerful methodology to investigate the dynamics and functions of non-genetically encoded biomolecules such as sialylated (Chang et al., 2009;Laughlin and Bertozzi, 2009;Mbua et al., 2013), fucosylated (Hsu et al., 2007;FIGURE 4 | Glycan array assay for GT characterization. Glass slides equipped with plant cell wall-related oligosaccharides are incubated with azido-functionalized sugar nucleotides and GT candidates expressed in, for example human embryonic kidney (HEK) 293 cells. Any transferred monosaccharide is visualized by azide-alkyne cycloaddition reaction with a fluorescent dye to determine reactive acceptors (reprinted from Ruprecht et al., 2020). Laughlin and Bertozzi, 2009;Besanceney-Webler et al., 2011), and mucin-type O-linked glycans (Laughlin and Bertozzi, 2009;Baskin et al., 2010) in live cells and model organisms (Prescher and Bertozzi, 2005;Grammel and Hang, 2013). This approach relies on the labeling of specific sugars by feeding cells with a synthetic monosaccharide analog carrying a chemical reporter that is then reacted with a probe (e.g., a fluorophore suitable for fluorescent microscopy imaging) in living systems to locate/visualize the incorporated reporter. Despite the fast-growing number of examples of this potent method in animal cells, reports describing its use in plant biology are surprisingly few (Anderson et al., 2012;Dumont et al., 2016;Zhu et al., 2016;Zhu and Chen, 2017). In part, this is due to the capacity of these probes to penetrate the cell wall barrier and, in part, due to the limited diversity of sugar analogs available to replace the endogenous sugars that need to be transported into the plant cell, and incorporated into glycan structures by GTs in a similar manner. Other new technologies are being developed to directly perform imaging of single glycan molecules that are isolated by mass-selective, softlanding electrospray ion beam deposition and imaged by low-temperature scanning tunneling microscopy (Wu et al., 2020). This generates glycan structures at the singlemolecule and single cell levels to directly relate how molecular structure correlates with properties -a step forward toward cracking the "sugar code." • GT activity characterization by glycan arrays. The use of glycan arrays equipped with oligosaccharide acceptors, in combination with expressed GT/GH candidates, may significantly accelerate the identification and characterization of further GTs/GHs responsible for plant glycosylation/modulation in the future. To enable rapid progress in this area, intensive research on the chemical and/or enzymatic synthesis of oligosaccharide acceptors and sugar nucleotide donors as well as on high-yielding production of active GT candidates in different expression systems is required. In this direction, a JBEI (The Joint BioEnergy Institute) GT Collection with almost 500 GTs from Arabidopsis and rice were cloned in-frame into Gateway technology compatible vectors to readily enable downstream applications (Lao et al., 2014). Either more collective resources from our laboratories (a major barrier for individual groups when research funding is scarce) or commercial intervention (which would require the same importance placed on plant biology as medical research where such resources are provided) are necessary to drive functional genomic approaches in plant glycobiology. • Structural diversity in N-and O-glycans present in plant glycoproteins. Although some progress has been made recently, the precise N-glycan composition of individual native plant glycoproteins from different cells or tissues is only partially known (Xu et al., 2016;Zeng et al., 2018). Future efforts will aim to obtain a more comprehensive picture on N-glycan composition within specific glycoproteins to identify distinct N-glycan structures that are causative for a specific phenotype. In the same vein, determining functional roles for individual HRGP O-glycoproteins has been hampered by our failure to directly characterize each of these complex O-glycan structures. Only few studies have been able to purify AGPs and analyze their glycan structural variations in detail (Tan et al., 2004;Tryfona et al., 2010Tryfona et al., , 2012Tryfona et al., , 2014Pfeifer et al., 2020). Biochemical characterization needs to be linked to detailed functional studies (e.g., sitedirected mutagenesis). In general, functional validation is experimentally much more complex as well as timeconsuming compared to the biochemical quantitation of the O-glycosylation levels. Furthermore, small changes in O-glycosylation in AGPs/FLAs and in EXTs can result in either activation or inactivation of their in vivo functions and can have an effect on their subcellular localization targeting (Velasquez et al., 2015b;Xue et al., 2017;Borassi et al., 2020;Seifert, 2020), so the functional relevance of each event cannot directly be inferred from large-scale quantitative analysis. A dual convergent approach between both enzymology/biochemistry and genetics is required to address this important aspect of plant glycoprotein structural diversity at the single cell level.
• Overcome functional genetic redundancy of plant glycoproteins. Addressing genetic redundancy and functional overlap might be achieved by using multiplex CRISPR-CAS9/genome editing/gene knockout technology. Some recent reports have used this approach to overcome functional redundancy in AGPs (Moreira et al., 2020) and in GTs (e.g., GLCATs) acting on AGPs (Zhang et al., 2020). This might be extended to investigate their function in other plant species. • Plant glycoproteome-interactome. Finding new proteins associated with plant glycoproteins, plant GTs and GHs will expand our knowledge on the regulatory aspects of plant glycobiology. New techniques such as proximity labeling (e.g., APEX, TurboID, etc.) together with the existing tools for detecting in vivo protein-protein interactions (e.g., BiFC, TriFC, FRET, etc.) will allow us to improve our plant glycobiology interactome inventory. Deeper integration of the N-and O-glycosylation pathway into the broader context of plant cell biology and systems biology is necessary. We envisage the development of a broad atlas of glycomes across plant tissues and cell types to integrate protein glycosylation features into plant gene and protein databases.

AUTHOR CONTRIBUTIONS
KJ, MD, CR, and FP analyzed the references, wrote the manuscript, and helped on the figures design. RS, GS, AB, and JE analyzed the references, supervised the project, and wrote the manuscript. All authors have read the manuscript and have approved this submission.