Structure of xyloglucan xylosyltransferase 1 reveals simple steric rules that define biological patterns of xyloglucan polymers

Significance

The recalcitrant nature of the plant cell wall presents a significant challenge in the industrial processing of biomass. Poor understanding of plant polysaccharide biosynthesis impedes efforts to engineer cell walls susceptible to efficient and unnatural pathways of degradation. Despite numerous genetic and in vitro studies of the xyloglucan xylosyltransferases (XXT1, XXT2, and XXT5), the specific roles of each in the xylosylation of the xyloglucan backbone is unclear. On the basis of steric constraints imposed by the active-site cleft of structures presented here, we propose a multienzyme complex capable of producing the xylosylation patterns of native xyloglucans. This model significantly extends our limited understanding of branched polysaccharide biosynthesis.

Abstract

The plant cell wall is primarily a polysaccharide mesh of the most abundant biopolymers on earth. Although one of the richest sources of biorenewable materials, the biosynthesis of the plant polysaccharides is poorly understood. Structures of many essential plant glycosyltransferases are unknown and suitable substrates are often unavailable for in vitro analysis. The dearth of such information impedes the development of plants better suited for industrial applications. Presented here are structures of Arabidopsis xyloglucan xylosyltransferase 1 (XXT1) without ligands and in complexes with UDP and cellohexaose. XXT1 initiates side-chain extensions from a linear glucan polymer by transferring the xylosyl group from UDP-xylose during xyloglucan biosynthesis. XXT1, a homodimer and member of the GT-A fold family of glycosyltransferases, binds UDP analogously to other GT-A fold enzymes. Structures here and the properties of mutant XXT1s are consistent with a SN_i-like catalytic mechanism. Distinct from other systems is the recognition of cellohexaose by way of an extended cleft. The XXT1 dimer alone cannot produce xylosylation patterns observed for native xyloglucans because of steric constraints imposed by the acceptor binding cleft. Homology modeling of XXT2 and XXT5, the other two xylosyltransferases involved in xyloglucan biosynthesis, reveals a structurally altered cleft in XXT5 that could accommodate a partially xylosylated glucan chain produced by XXT1 and/or XXT2. An assembly of the three XXTs can produce the xylosylation patterns of native xyloglucans, suggesting the involvement of an organized multienzyme complex in the xyloglucan biosynthesis.

Plant cell walls consist of cellulose, hemicellulose, pectin, and lignin, all of which confer mechanical properties to plant structures, and are important for shape and development. Plant cell walls represent the largest pool of renewable carbohydrate and the potential to support numerous industrial applications in bioenergy and biomaterials (1, 2). The complex structure of plant lignocellulosic biomass resists enzymatic and microorganism degradation (3). Engineering a biologically viable plant susceptible to enzymatic or nonbiological degradation requires a complete understanding of plant cell wall polysaccharide biosynthesis and structure.

Xyloglucan (XyG) is the most abundant hemicellulose in the primary cell wall of dicotyledonous plants and has many proposed structural and regulatory functions (4–7). XyG consists of a 1,4-β-linked glucan backbone branched with various glycosyl residues depending on species or tissue (8). The nomenclature for XyG structure is as follows: G represents an unbranched glucose unit, whereas X, L, and F are glucosyl units with Xyl, Gal-Xyl, or Fuc-Gal-Xyl side chains, respectively (9). Arabidopsis XyG consists of a glucan backbone branched with 1,6-α-linked d-Xyl residues, resulting in XXXG-type pattern, which can be further decorated (10, 11).

Glycosyltransferases (GTs) catalyze the formation of glycosidic bonds by transferring a sugar moiety from an activated donor, typically a nucleotide sugar, to a variety of acceptor substrates, including, carbohydrates, proteins, and lipids (12). Amino acid sequences in the Carbohydrate Active Enzyme Database fall into 105 families of GTs (13, 14). Available structures indicate most GT families adopt one of two folds, GT-A or GT-B, although a rarer GT-C fold has been proposed (12). The GT-A fold has two Rossmann-like domains that form a central β-sheet, each face of which is covered by α-helices. These are typically metal-dependent enzymes that require an Asp-X-Asp motif for metal coordination (15, 16). GT-B folds also have two Rossmann-like domains, less tightly associated than those of GT-A folds, with the active site located between domains. GTs are also classified by the stereochemistry of the glycosidic bond in the product (inverted or retained) relative to that of the substrate (12). The catalytic mechanism of inverting GTs likely follows the single displacement mechanism of inverting glycosyl hydrolases (12, 17). The catalytic mechanism of retaining GTs, first proposed as a double displacement mechanism similar to retaining glycosyl hydrolases, has fallen into disfavor due to the absence of a suitably placed catalytic base and the failure to trap a glycosyl-enzyme intermediate. Instead, retaining GTs may employ a SN_i-like mechanism that consists of the acceptor substrate approaching from the same face as the leaving group with an oxocarbenium-ion intermediate (18–20). Although structural information is abundant for glycosyltransferases (21), structural information specifically for GTs involved in plant cell wall polysaccharide biosynthesis is available only for xyloglucan fucosyltransferase 1 (FUT1), which adds fucose to the terminal position of XyG side chains (22, 23).

Xylosylation of the 6-hydroxyl group of glucose, catalyzed by xyloglucan xylosyltransferases (XXTs), is the first step in building branches on the XyG backbone (8, 24). XXTs are type II transmembrane enzymes, having a short cytosolic N-terminal region, a transmembrane domain, a stem region, and a large C-terminal catalytic domain localized in the Golgi lumen (25–27). Three XXTs xylosylate the XyG backbone both in vivo (28–30) and in vitro (25, 31–34). These enzymes putatively adopt GT-A folds, have products that retain the stereochemistry, and belong to the GT34 family. Family GT34 contains seven proteins from Arabidopsis (31), including xyloglucan xylosyltransferases, galactomannan 1,6 galactosyltransferases and 1,2 galactosyltransferases (CAZy: www.cazy.org). Enzymatic action of XXTs in vitro employ short glucans, such as cellohexaose or cellopentaose, due to the low solubility of long glucan chains. XXT1 (and XXT2) primarily adds the first xylosyl residue to the fourth glucose residue from the reducing end of cellohexaose. XXT1 (and XXT2) then adds a xylosyl residue to the third or fifth glucose from the reducing end of cellohexaose (32).

Presented here are crystal structures of XXT1 without ligands, a binary complex with UDP, and ternary complex with UDP and cellohexaose. This is the first structure from the GT34 family and only the second crystal structure of a plant cell wall glycosyltransferase (the first being of FUT1). XXT1 is a homodimer in solution and crystal. Structural comparisons with retaining GT-A fold glycosyltransferases and directed mutations support a SN_i-like catalytic mechanism. Moreover, the crystal structure and homology models indicate similar steric requirements for glucan interactions with XXT1 and XXT2, but different requirements for XXT5, suggesting the participation of XXT1 (or XXT2) with XXT5 in achieving the in vivo pattern of xylosyl transfer to the linear glucan.

Results

Overall Structure of XXT1.

Protein expression of the XXT1 stem region and catalytic domain (residues 45–460) was performed in human embryonic kidney cells (35). SDS/PAGE analysis reveals a double band that is routinely observed in our preparations of XXT1 (SI Appendix, Fig. S1D) and also observed for XXT1 and XXT2 when expressed in Drosophila cells (32). The structure of the enzyme (in crystals of space group P2₁2₁2₁) was solved by single-wavelength anomalous diffraction (SAD), using data from a crystal derivatized with K₂HgI₄. The complex of UDP/Mn²⁺ (hereafter, the binary complex) and the complex of cellohexaose/UDP/Mn²⁺ (hereafter the ternary complex) were formed by ligand soaks. XXT1 exhibits highest activity with Mn²⁺ (32), which was included in all ligand soaks. A second crystal form of the apoenzyme (space group C222₁), diffracting to 1.5-Å resolution, was solved by molecular replacement. Given its superior resolution, the apoenzyme in space group C222₁ is reported in SI Appendix, Table S1. Regardless of crystal form, the asymmetric unit has two subunits of XXT1, and for each, residues 45–115 from the N terminus and residues 454–460 from the C terminus are without electron density. The purified protein has an observed mass consistent with that expected for residues 45–460. Finally, a substantial void exists in both crystal forms that could accommodate 70 additional residues.

XXT1 adopts a GT-A fold with a central β-sheet having both faces covered by α-helices (Fig. 1 A and C). The central β-sheet contains strands β2, β1, β3, β6, β5, and β7, all of which are parallel except for strand β6 (Fig. 1 A and C). The loops extending from this central β-sheet and surrounding α-helices define the active site of XXT1, containing the Asp-X-Asp motif (site of Mn²⁺ binding), donor substrate binding site, and acceptor binding site (Fig. 1A). The active site is a cleft roughly 13 Å wide and 30 Å long (Fig. 1A).

Fig. 1.

Structure overview of XXT1. (A) XXT1 monomer (chain A) colored blue to red from N to C terminus, respectively. (B) Secondary and tertiary structure of XXT1. Names of α-helices and β-strands correspond to those in A. (C) XXT1 dimer with UDP and cellohexaose. Monomers are shown in green and yellow. UDP binds to both monomers, whereas cellohexaose binds to only one monomer. (D) View down the symmetry axis of the dimer, 90° rotation of the molecule shown in C.

XXT1 Forms a Dimer.

XXT1 forms a dimer with an interface of roughly 2,900 Ų (PDBe PISA; www.ebi.ac.uk/pdbe/pisa/). XXT1 dimer has a mass of 96.7 kDa based on amino acid sequence relative to 102.7 kDa determined by gel filtration against protein standards of known mass. The high value from gel filtration is consistent with an increased radius of gyration due to the aforementioned loose stem region of 70 residues (SI Appendix, Fig. S1 A and B). The interface between subunits involves the side chains of Leu171, Leu172, Ile175, Ile179, and Tyr191 from helix α2 and strand β2 from each subunit (SI Appendix, Fig. S1E). Surrounding this hydrophobic core are electrostatic interactions and hydrogen bonds obeying the twofold symmetry of the dimer: Gln432-NE2 to the backbone carbonyl of Leu119; Asp168-OD1 to His169-ND1; the backbone amide of Tyr191 to Lys176-NZ; Tyr414-OH to His217-NE2; Glu188-O1 to Lys451-NZ; and Glu219-OE1 to His447-NE2. Interacting residues at the subunit interface of the dimer are identical in XXT1 and XXT2, suggesting the possibility of heterodimer formation (SI Appendix, Fig. S2). Heterodimer formation would likely have no major impact on xylosylation pattern due to the predicted functional redundancy of XXT1 and XXT2 (28, 32). Although twofold molecular symmetry is obeyed across the subunit interface, cellohexaose binds to only one of two active sites of the UDP/cellohexaose complex. The basis for this asymmetry in the binding of cellohexaose is provided below.

Donor Substrate Binding.

Binary (UDP) and ternary (UDP and cellohexaose) structures of XXT1 were obtained by soaking XXT1 crystals with the substrates, along with MnCl₂, for 2 d; the structures were refined against data to a resolution of 2.4 Å and 2.1 Å, respectively (SI Appendix, Table S1). In both the binary and ternary complexes, Mn²⁺ is coordinated through a monodentate interaction with Asp227, bidentate interaction with Asp229, and NE2 of His377. The UDP molecules have different conformations in the presence and absence of cellohexaose. The extended conformer in the ternary structure coordinates Mn²⁺ through one oxygen atom from each of the α- and β-phosphoryl groups (Fig. 2B). The bent conformer in the binary structure coordinates Mn²⁺ through one oxygen atom of the β-phosphoryl group and a nitrate anion (Fig. 2A). Interactions are identical for the bent conformer of UDP in the active site of the ternary complex that lacks cellohexaose and the active sites of the binary complex. The β-phosphoryl group of the extended UDP conformer hydrogen bonds with the 6-hydroxyl group and the 2-hydroxyl group of the fourth and fifth glucose units, respectively, from the reducing end of bound cellohexaose. In addition, atom NZ of Lys382 hydrogen bonds with oxygen atoms of the α- and β-phosphoryl groups of the extended conformer (Fig. 2). The extended conformation of UDP likely approximates the catalytically productive binding of UDP-xylose and is similar to nucleotide conformations in other retaining GT-A fold glycosyltransferases (18, 19). Additionally, the bent conformation may indicate that interactions with XXT1 and UDP are weak, and that most of the dominant interactions are between XXT1 and the xylose of UDP-xylose.

Fig. 2.

Bound UDP in two conformations. (A) Bent conformer, cellohexaose absent. (B) Extended conformer, cellohexaose present. Colors in A and B are as follows: Tan, XXT1 secondary structure; blue, nitrogen atoms; red, oxygen atoms; cyan, carbon atoms of ligands; orange, phosphate atoms; purple, Mn²⁺. (C) Interaction map for the extended conformer of UDP from LigPlot+ (45).

Modeling of UDP-Xylose.

Attempts to cocrystallize UDP-xylose with XXT1 or soak UDP-xylose into preformed crystals did not reveal a bound donor substrate. Instead, UDP-xylose was modeled into the active site of the ternary structure. GT structures typically have similar bound UDP-sugar conformations in which the sugar residue folds over the phosphoryl groups. An analogous conformation for UDP-xylose is present in the structure of XXYLT1 (PDB ID 4WNH and SI Appendix, Fig. S4B) and LgtC (18, 19). That model for UDP-xylose, when superimposed on the extended UDP molecule in XXT1, fits the active site, with the xylosyl residue occupying a pocket of uniform/unstructured electron density. Energy minimization led to the model in Fig. 3, which retains the conformations of UDP, cellohexaose, and active-site side chains (SI Appendix, Fig. S4A). The 3-hydroxyl group of xylose hydrogen bonds with Gln319 and Thr269, and the 4-hydroxyl group hydrogen bonds with Lys207 and Asp318 (Fig. 3). Phe203, which has a high B parameter in crystal structures, makes a favorable contact with atom C5 of the xylosyl residue, a contact that would discriminate against a hexose sugar.

Fig. 3.

Model of bound UDP-xylose. (A) Position of UDP-xylose with respect to cellohexaose. The arrow marks the 6-hydroxyl group of the fourth glucose from the reducing end. (B) Distances (in angstroms) between selected atoms of XXT1 and xylosyl group of UDP-xylose. Colors are defined in Fig. 2.

Acceptor Substrate Binding.

The ternary complex of XXT1 has only one monomer of the dimer with bound cellohexaose (Fig. 1 C and D). Access to the active site in the nonoccupied monomer is limited due to packing contacts. The cellohexaose molecule is covered by strong electron density and has B parameters significantly lower than those of UDP (SI Appendix, Fig. S3A). The interactions of XXT1 with the glucosyl residue of cellohexaose defines six subsites (hereafter subsites 16). Subsite 1 binds the reducing end of cellohexaose. The 6-hydroxyl of the fourth glucosyl residue projects into the active site of XXT1 (Fig. 4), in agreement with the predominant modification of cellohexaose at its fourth glucosyl residue in solution (32). Glucose residues 3–5 of cellohexaose make direct hydrogen bonds at subsites 3–5, respectively. Conversely, the first, second, and sixth glucosyl residues have water-mediated or hydrophobic interactions at subsites 1, 2, and 6, respectively (Fig. 4). Glucosyl residues at the ends of cellohexaose have weak electron density, whereas residues 2–4 have strong electron density with clearly defined 6-hydroxyl groups that unambiguously determine the orientation of cellohexaose and the placement of its reducing and nonreducing ends. The strong electron density of 6-hydroxyl groups of cellohexaose suggests a single allowed orientation of the bound glucan to XXT1 (Fig. 4B and SI Appendix, Fig. S3A). Interestingly, bound glucan chains in the active sites of the XXT1 dimer point in opposite directions. Thus, both subunits of the XXT1 dimer would unlikely xylosylate the same glucan chain, but more likely act on two separate chains. The glucosyl residue at subsite 3 hydrogen bonds with His346 and stacks with Tyr348 (Fig. 4). Glucosyl residue 5 hydrogen bonds with Asp389, Lys382, and an oxygen atom of the β-phosphoryl group of UDP.

Fig. 4.

Acceptor substrate binding site in the ternary complex. (A) Residues of XXT1 in proximity to cellohexaose. (B) Cellohexoase with electron density from an omit map contoured at 2σ. Colors are defined in Fig. 2. (C) Interaction map for cellohexaose from LigPlot+ (45).

Xylosyl extensions from the 6-hydroxyl groups of glucosyl residues at subsites 1, 3, 5, or 6 should not introduce steric conflicts (SI Appendix, Fig. S5). The 6-hydroxyl groups at subsites 3 and 5 are pointed away from the acceptor cleft and are solvent exposed (SI Appendix, Fig. S5A). The 6-hydroxyl groups at subsites 1 and 6 have space to accommodate a xylosyl adduct (SI Appendix, Fig. S5 B and C). Moreover, relatively low levels of electron density associated with glucose units at subsites 1 and 6 suggests weaker binding at these subsites for an extended glucan chain (Fig. 4 and SI Appendix, Fig. S3A). In contrast, a xylosyl extension at subsite 2 would overlap with the side chains of His252, Ile351, and Tyr344, suggesting that xylose transfer to glucose unit N blocks subsequent binding necessary for xylose transfer to glucose unit N + 2.

Mutations of Active Site Residues.

XXT1 wild-type and mutant enzymes were assayed using the UDP-GLO assay (Promega) to measure UDP formation with 20-min reactions (Fig. 5A). These assays revealed comparable activities for the wild-type and Ser228Ala enzymes, but activity levels for other mutant enzymes were statistically indistinguishable from an enzyme-free buffer (blank). Twenty-hour reactions were then performed to better determine the loss of activity of the least active mutant enzymes. The 20-h reaction resulted in roughly fourfold higher UDP production for wild-type XXT1 (SI Appendix, Fig. S6A). XXT1 assays lacking UDP-xylose resulted in no xylosylation of cellohexaose (SI Appendix, Fig. S6 A and D). Similarly, XXT1 assays lacking cellohexaose had no detectable UDP compared with assays lacking XXT1, indicating XXT1 does not hydrolyze UDP-xylose (SI Appendix, Fig. S6A). Mutant enzymes Ser228Ala and Asn268Ala have UDP levels comparable to those of the wild-type enzyme (Fig. 5B). Mutations of residues that coordinate Mn²⁺ in the wild-type enzyme (single mutant enzymes Asp229Ala and His377Ala, and the double mutant enzyme Asp227Asn/Asp229Asn), produce 25–30% of the wild-type levels of UDP. Mutant enzymes producing the lowest levels of UDP (less than 20% of wild-type) are Lys382Ala, Asp317Ala, Asp318Ala, and Gln319Ala. Lys382 hydrogen bonds with oxygen atoms of the α- and β-phosphoryl groups of UDP (extended conformer), whereas the remaining residues interact with the xylosyl group of modeled UDP-xylose. Evidently, mutations proximal to the xylosyl group of the donor substrate have the greatest impact on the rate of product formation. To confirm activity, the Asp317Ala, Asp318Ala, and Gln319Ala mutants were assayed with sixfold higher protein concentration, demonstrating clear UDP production (SI Appendix, Fig. S6B) and xylosylated cellohexaose formation was confirmed by MALDI-TOF analysis (SI Appendix, Fig. S6 F–H).

Fig. 5.

Activity of XXT1 mutants and wild-type (WT) enzymes. (A) Twenty-minute activity assays. (B) Twenty-hour activity assays. Each datum represents the average of three assays measuring the production of UDP from starting concentrations of 2 mM UDP-xylose, 1 mM cellohexaose, and 2 mM MnCl₂, in 50 mM Tris pH 7.4 and 75 mM NaCl. Error bars are drawn at one SD. Blank represents activity assay with enzyme-free buffer.

Catalytic Mechanism of XXT1.

Inverting GTs utilize a single-displacement mechanism similar to that of inverting glycosyl hydrolases (12, 17). The catalytic mechanism for retaining GTs, however, is unclear. Retaining glycosyl hydrolases utilize a double-displacement mechanism in which a catalytic base first inverts the anomeric stereochemistry of the sugar, forming a glycosyl-enzyme intermediate (12, 17). A second nucleophilic attack at the anomeric carbon of the glycosyl-enzyme intermediate by the hydroxyl group of the acceptor restores the original anomer. For GTs however, a suitable catalytic base is not evident in structures (18, 36, 37) and efforts to trap a glycosyl-enzyme intermediate have been unsuccessful (38). Hence, some have suggested a SN_i-like mechanism for retaining GTs as an alternative (12, 18). In the SN_i-like mechanism an oxocarbenium-ion intermediate forms; however, the acceptor approaches the same face of the oxocarbenium-ion intermediate as the leaving group. This mechanism has gained support from studies in quantum/molecular mechanics (39), kinetic isotope effects (20), and crystallography (snapshots along the reaction pathway) (19).

The XXT1 active site has no nucleophile positioned suitably for a double-displacement mechanism. Asp317 and Asp318 are 5.4 and 5.7 Å away, respectively, from the anomeric carbon in the UDP-xylose model. Invariant conformations of Asp317 and Asp318 over all crystal structures and low B parameters are not indicative of a region predisposed to conformational change, yet mutations at positions 317 and 318 have among the largest impacts on activity (Fig. 5). In the structure of XXYLT1 (19), a hydrogen bond between Gln330 and the 2-hydroxyl group of the xylosyl could in principle stabilize the positive charge of an oxocarbenium ion during the SN_i-like reaction mechanism (19). Gln319 of XXT1 assumes a position analogous to Gln330 in XXYLT1 (SI Appendix, Fig. S4B), 4.7 Å from the anomeric carbon atom of modeled UDP-xylose, forming a hydrogen bond with the 3-hydroxyl group of the xylosyl residue. The Gln319Ala mutation has one of the largest impacts on activity (Fig. 5).

Homology Modeling of XXT2 and XXT5.

Efforts to purify and concentrate XXT2 and XXT5 to levels sufficient for crystallization have been unsuccessful; however, the structure of XXT1 leads to homology models of XXT2 and XXT5 that have no major departures from the backbone structure of XXT1 (SI Appendix, Fig. S7A). XXT1, XXT2, and XXT5 have nearly identical positions for all residues in the active site with one notable difference: XXT1 puts an isoleucine residue at subsite 2 proximal to the 6-hydroxyl group of the second glucosyl from the reducing end of cellohexaose, whereas XXT5 has a glycine residue in the corresponding location (SI Appendix, Fig. S7 C and E). The glycine in XXT5 likely allows a xylosyl moiety at subsite 2, suggesting that XXT5 (in contrast to XXT1 and XXT2) can transfer a xylosyl moiety to glucose N when the N + 2 glucosyl toward the reducing end has been xylosylated (SI Appendix, Fig. S7 B–E). Additionally, the XXT1 surface contains a region rich in acidic residues (Glu357, Glu358, and Glu361) and one rich in basic residues (Lys419, Arg425, and Lys427; SI Appendix, Fig. S8 A and B). These residues are conserved in XXT1, XXT2, and XXT5 (SI Appendix, Fig. S2). Matching the acidic region of one XXT1 dimer to the basic region of another XXT2 dimer results in a feasible dimer-to-dimer interface (SI Appendix, Fig. S8C). A glucan chain would thread from one active site to that of its neighboring dimer, suggesting the possibility of coordinated xylosyl transfers to the glucan chain.

Discussion

Of all plant cell wall GTs, only the structure of FUT1 is known (22, 23). The crystal structure of XXT1 is the first instance of a GT34 family member and the second example of a GT involved in plant cell wall biosynthesis. XXT1 and FUT1 represent different folds, GT-A for XXT1 and GT-B for FUT1. In addition, XXT1 employs a large cleft in acceptor binding, whereas FUT1 does not (22, 23). The twofold symmetry of XXT1 active site clefts would require the glucan chain to bend 180° for both active sites to modify a single glucan chain. Given the structural properties of the glucan chain, such a bend is unlikely. Instead, XXT1 probably acts on separate glucan chains, nearly parallel and in opposite orientations as defined by their reducing and nonreducing ends.

XXT1 is catalytically active on cellopentaose and cellohexaose with xylosylation primarily on glucose residue 4 (31, 32). The ternary complex (PDB ID 6BSW) directly supports xylosylation of cellohexaose at the fourth glucose residue from the reducing end (Fig. 2). The addition of a second xylosyl moiety requires monoxylosylated cellohexaose to disassociate from XXT1, rotate about the axis of the glucan by 180°, and rebind, shifted by one glucosyl residue. Shifting (after rotation) by one glucosyl residue occupies subsites 1–5 or subsites 2–6. Binding of ₆GGXGGG₁ at subsites 1–5 should generate the product ₆GXXGGG₁, whereas binding at subsites 2–6 should generate the product ₆GGXXGG₁ (SI Appendix, Fig. S9). Cavalier and Keegstra (32) observed both products, with ₆GGXXGG₁ preferred. Given that subsite 2 cannot accommodate a xylosyl adduct, XXT1 cannot transfer a xylosyl moiety to ₆GGXXGG₁ to produce ₆GXXXGG₁, terminating further modification; however, XXT1 can transfer a xylosyl moiety to the third glucose of ₆GXXGGG₁, forming ₆GXXXGG₁ (SI Appendix, Fig. S9). Hence, as long as ₆GGXXGG₁ is the preferred product in the pool of doubly modified cellohexaoses, the addition of a third xylosyl moiety will never exceed 50%, as experimentally observed (33).

The steric limitation that prevents subsite 2 from accommodating a xylosyl adduct leads to the N + 2 rule: XXT1 cannot transfer a xylosyl residue to the glucosyl residue N + 2 when glucosyl residue N has a xylosyl adduct. This rule can be extended to XXT2 with some confidence. The homology model of XXT2 (and XXT5) and the crystal structures of XXT1 have nearly identical backbone positions (SI Appendix, Fig. S7A), and the partial redundancy in function of XXT1 and XXT2 as demonstrated by in vivo (28, 29) and in vitro (32) are consistent with similar substrate specificities, steric limitations, and modes of action. The residue that would be in steric conflict with a xylosylated glucose at subsite 2 is isoleucine in XXT1 and XXT2, but is glycine in XXT5 (SI Appendix, Fig. S7). Hence, it is likely that the N + 2 rule does not apply to XXT5 (Fig. 6).

Fig. 6.

Production of an XXXG-type xyloglucan. The model incorporates observed features of the xyloglucans elaborated by the wild-type organism (XXXG repeats) and XXT5-knockout organism (GXXG repeats), the emergence of the reducing end of the glucan from the synthase, the steric requirements imposed by the N + 2 rule (for XXT1 and XXT2), and the lack of a N + 2 rule (for XXT5). Angles in black represent the helical rotation of each cellobiose unit of cellulose, whereas angles in red combine a 180° offset angle with a half-cellobiose helical rotation (25.7°). Circles and stars represent glucose and xylose residues, respectively. The model, based on a 7-glucose separation of XXTs, enables a clear 2D representation. Models based on other separations (11 and 15 glucose units for instance) are possible and may be necessary for allowable protein–protein contacts. This model depicts the order of XXT action required for the XXXG pattern and not the organization of the xyloglucan biosynthetic complex. The XyG biosynthetic complex likely contains multiple copies of all XyG enzymes, producing multiple XyG chains within a single complex (Discussion).

A model for xylosylation of the glucan chain in vivo must satisfy the following assumptions and observations. First, cellulose synthase-like C4 (CSLC4) probably functions like cellulose synthase, adding 1,4-β-linked glucosyl residue onto the nonreducing end of the growing glucan chain (40, 41). Hence, the reducing end of the glucan would emerge first from the synthase. Second, xxt5 knockout plants exhibit two- to fourfold increases in GXXG-type xyloglucan and a threefold decrease in XXXG-type xyloglucan (29). Third, the glucan chain binds in a specific orientation to XXT1, and on the basis of homology models, the same orientation for XXT2 and XXT5 (this study). Fourth, XXT1, XXT2, and XXT5 have similar six-subunit clefts (this study). Fifth, XXT1 and XXT2 adhere to the N + 2 rule, whereas XXT5 does not (this study). Sixth, XXT1 interacts with XXT2, and XXT2 interacts with XXT5, but XXT1 does not interact strongly with XXT5 (42−43). Finally, the glucan is a helix of cellobiose repeats, with an angular displacement (rotation) of 51.4° per cellobiose unit (44). The second glucosyl residue of cellobiose has an additional angular offset of 180° relative to the first glucose. Fig. 6 presents a plausible sequence of events during xyloglucan biosynthesis that accommodates the preceding observations/assumptions. Emerging glucosyl residues from the glucan synthase thread into XXT1 first, followed by XXT2, and then XXT5, the order determined by the observed binding preferences of the XXT protein–protein interactions (XXT1–XXT2 and XXT2–XXT5) and the recognition that XXT5 must come last to produce the XXXG repeat. The glucan chain in this model must advance by four glucosyl residues with each cycle. The active sites of the XXTs are “in phase” by virtue of the 51.4° helical repeat of the glucan in combination with the 180° offset of the second glucosyl residue of cellobiose (35). In the absence of XXT5, as in plant knockouts, the product would have a GXXG repeat. The absence of steric limitations at subsite 2 of XXT5 allows xylosyl transfer to glucose N + 2 when glucose N has a xylosyl adduct. Therefore, production of a glucan with a consistent XXXG repeat, as observed for the native xyloglucan in the cell wall (11, 29), requires the action of XXT1 and XXT2 to synthesize the GXXG repeat, followed by XXT5 to produce XXXG, all of which is organized in a specific manner as to prevent the xylosylation of all 6-hydroxyl groups of the glucan chain (Fig. 6).

The linear model shown in Fig. 6 accounts for all that we know or can be inferred in regard to xylosylation of a glucan chain. This model does not depict the spatial organization of glucan synthase and three XXTs on the 2D surface of the membrane. Information regarding protein–protein interactions is insufficient to favor one packing scheme over another for complex formation. The XyG biosynthetic complex likely contains multiple copies of all XyG biosynthetic proteins and likely synthesizes at least two XyG chains within a single complex.

In summary, we demonstrate here that XXT1 and/or XXT2 are unlikely to produce the XXXG-type XyG in Arabidopsis, requiring the contribution of XXT5 to fully xylosylate the xyloglucan backbone. This limitation of XXT1 and XXT2 and the function of XXT5 were not apparent from reverse genetics nor from in vitro studies. Steric constraints revealed by the structures, however, clarify the action of each XXT in xyloglucan biosynthesis. The ordering of the XXTs proposed here, which requires XXT5 to follow XXT1 and XXT2, provides an explanation for all previous observations, such as the low activity of XXT5 on nonxylosylated cellohexaose (33), preferring instead a cellohexaose xylosylated at the N + 2 site (GGGGXG or GGGXXG) for high in vitro activity, the reduction of XXXG-type xyloglucan in xxt5 knockout plants (29, 30), and why the xxt1/xxt2 double knockout has no XyG, whereas the xxt5 single knockout retains significant (although reduced) levels of XyG (29). Experiments to determine the action of XXT5 on specifically xylosylated substrates and to understand the spatial organization of the XyG biosynthetic complex will lead to a more certain understanding of XyG production in plants. The findings in this study will support strategies to create plant biomass with desired properties.

Methods

All proteins were expressed in HEK cells as described previously (35). Heavy atom derivative was produced by soaking K₂HgI₄ in XXT1 crystals. Protein expression, purification, crystallization, structure determination, and enzyme assays are described in SI Appendix, SI Materials and Methods.