In grasses, the most abundant matrix polysaccharide (i. e., hemicellulose) is xylan. In the last five years, several GTs and other enzymes from Arabidopsis and grasses that likely function in xylan biosynthesis have been characterized (Fig. 2). A complete list of the proteins that function in and the mechanism of xylan synthesis in the Golgi and subsequent release into the cell wall is still being unraveled. Here, we will discuss recent progress with an emphasis on results in grasses. Table 2 lists enzymes implicated in synthesis of the xylan backbone, reducing end oligosaccharide, and sidechains. The reader is referred for additional information to other recent reviews on the topic (Buanafina 2009; Faik 2010; Scheller et al. 2010; Carpita

2012) . When it has been assayed, cell wall material from loss-of-function xylan mutants all exhibit improved digestibility (Mortimer et al. 2010; Brown et al. 2011; Chen et al. 2013; Chiniquy et al. 2012), consistent with the role of this polysaccharide in stabilizing cell walls.

At least two Carbohydrate-Active enZyme (CAZy) database GT families are implicated in the backbone synthesis of xylans, namely GT43s and GT47s (Table 2). Studies in Arabidopsis showed that GT43s are responsible for xylan backbone synthesis since irx9 and irx!4 have drastically shorter xylan chains and reduced ability to transfer Xyl from UDP-Xyl onto xylo-oligosaccharide acceptors (Brown et al. 2007; Pena et al. 2007). Similarly, double mutation of two GT47 genes, irxl0 and irxl0-L severely reduced xylan length, but without affecting the reducing end of xylan. The importance of GT47 family enzymes in xylan synthesis has recently been extended to rice. The Osirx10 mutant has greatly reduced xylan amounts in stems without showing a reduction in xylan chain length, suggesting a somewhat different xylan synthesis mechanism in grasses compared with dicots (Chen et al. 2012). A biochemical study in wheat (Triticum aestivum) suggested proteins from the GT43, GT47, and GT75 families are promising candidates for members of the gluronoarabinoxylan synthetase. Coimmunoprecipitation indicated these three GT proteins interact with each other to form a complex exhibiting xylan synthesis activity (Zeng et al. 2010).

Other studies in Arabidopsis have identified proteins that function in synthesis of the xylan reducing-end oligosaccharide (Fig. 2B), which has been found in several dicots and conifers but not detected in grasses (York et al. 2008; Scheller et al. 2010). The sequence of this reducing end oligosaccharide, or "primer", is 4-p-D-Xylp-(1—>4)-p-D-Xylp-(1—>3)-a- L-Rhap-(1—>2)-a-D-GalpA-(1—>4)-D-Xylp. As summarized in Table 2, the mutants irx7/fra8, irx8, and parvus are depleted for the reducing-end oligosaccharide (Pena et al. 2007). IRX8 and PARVUS, both GT8s, are implicated in adding the galacturonic acid and a a-xylose residue to the primer (Lee et al. 2007; Pena et al. 2007). IRX7, and its close homolog F8H, have been implicated as Rha-specific xylosyltransferases, because they act on a diversity of sugars (Rennie et al. 2012). Despite the absence of the reducing-end primer in experiments in grasses, enzymes with sequence similarity to those implicated in its synthesis have been retained in the rice genome (Scheller et al. 2010).

Other recent work has revealed enzymes that likely function to attach the xylan side chains, glucuronic acid and, in grasses, arabinose. Mortimer et al. (2010) identified mutants in two GT8 family genes, gux1 and gux2. The proteins encoded by these genes are Golgi-localized and required for the addition of both glucuronic acid and 4-O-methylglucuronic acid branches to xylan in Arabidopsis stem cell walls (Mortimer et al. 2010). Recently, another double mutant, irx15 and irx15L, was also found to be involved in xylan synthesis (Brown et al. 2011). These two genes, which belong to the domain of unknown function 579 family, might also be glucuronic acid transferases because they exhibited similar mutant features to gux1 and gux2 (Brown et al. 2011). For addition of the side chains of grass xylan, studies have focused on GT61 family members, which are much more highly expressed in grasses than in dicots (Mitchell et al. 2007). Repression of expression of a GT61 encoding gene, TaXat1, in wheat endosperm and its heterologous overexpression in Arabidopsis provided strong evidence that TaXAT1 possesses a-(1,3)-arabinosyltransferase activity (Anders et al.

2012) . Disruption in rice of another GT61 encoding gene, called Xax1, has also been published recently (Chiniquy et al. 2012). This mutant is lacking a previously observed but poorly characterized xylan substitution, an arabinofuranose residue substituted at the O-2 position of a xylosyl residues, in the structure p-Xylp-(1^-2)-a-Araf-(1^-3) (Fig. 2A). Based on H1-NMR and glycosidic linkage analysis, XAX1 possess a xylosyl transferase activity that can attach (P-1,4-Xylp)4 onto the acceptor p-Xylp-(1^2)-a-Araf-(1^3) (Chiniquy et al. 2012).

Other work has provided important insight into synthesis of the arabinofuranose (Araf) nucleotide sugar precursor of grass cell wall glucuronoarabinoxylan. A recent study of reversibly glycosylated proteins (RGPs) in Arabidopsis showed that the conversion of UDP-p-L — arabinopyranose (UDP-Arap) to UDP-p-L-Araf is indispensable for cell wall synthesis (Rautengarten et al. 2011). The rGps are in the GT75 family. The knockout mutants, rgp1 and rgp2 significantly reduced the total L-Ara content relative to the wild type and showed reduced UDP-Ara mutase (UAM) activity. UAM activity has been identified in rice, as well (Konishi et al. 2007; Konishi et al. 2011). Three rice genes with close sequence similarity to RGP-encoding genes were predicted to be UAM candidates (Konishi et al.

2007) . Later, knock-down of one of these three genes suppressed the UAM activity and reduced UDP-Araf amounts in mutant rice plants (Konishi et al. 2011). The mutant also decreased the incorporation of ferulic acid and p-coumaric acid to the cell wall and presented dwarfed and infertile phenotypes (Konishi et al. 2011). One of the GT75 family members in wheat mentioned above has been inferred to have the UAM activity but needs to be further studied (Zeng et al. 2010).

The final set of recent advances in our understanding of xylan synthesis in grasses relates to acylation of grass xylan arabinose residues by the hydroxycinnamates, ferulic acid and p-coumaric acid (Buanafina 2009). Consistent with expectations, in two studies of rice mentioned above, mutants with reduced Araf substituted xylan had reduced cell wall content of ferulic acid and p-coumaric acid (Konishi et al. 2011; Chiniquy et al. 2012). Mitchell and colleagues developed the hypothesis that 12 members of the BAHD family of acyl-CoA acyltransferases that were much more highly expressed in grasses than dicots might act as arabinofuranose feruloyl transferases (Mitchell et al. 2007). Later, silencing of four members of this family in rice (LOC_Os05g08640, LOC_Os01g09010, LOC_Os06g39470 and LOC_Os06g39390) reduced the ferulic acid content in young leaves by about 20%, leading to the refined hypothesis that one or more of the targeted genes acts as the feruloyl transferase (Piston et al. 2010). Recently, the overexpression of one of the same genes examined by Piston et al., LOC_Os06g39390, dubbed OsAT10, has led to the hypothesis that this protein functions as a p-coumaroyl transferase, as the overexpression plants have increased levels of p-coumaroyl esters bound to arabinose in the cell wall (Bartley et al. 2013a). The plants also exhibit a reduction in the level of polysaccharide-linked ferulic acid and show a concomitant improvement in digestibility (Bartley et al. 2013a), consistent with the model that ferulate — mediated crosslinking is important for grass cell wall digestibility.